ks/0000755000176200001440000000000014673325021010672 5ustar liggesusersks/MD50000644000176200001440000001051314673325021011202 0ustar liggesusers2a5d852616b9ad885f1656da5abe8d34 *CHANGELOG 89ed2edda932b2a7fca9a47ac4648cca *DESCRIPTION fe2c07c00353a611402926a41f8e81fe *NAMESPACE c7be26732ecd99c5cd8e00b7a6dac21d *R/binning.R 52ec8535d7a7ba554292570f2c1a988c *R/deconv-kde.R 13c69114eb039016d7f73a129e3d944c *R/hist.R c6a8cf8fff94aef509b038066fc95e60 *R/integrate-kde.R f95543caac123c1426500159c4c0db75 *R/kcde.R a8dc0b16f5b22eb0c59eb966a0a220cf *R/kcopula.R a2bc07490172f54bf852f9bf919fabd0 *R/kda.R 7f6f4efd52e47642c3f169a7bc03c301 *R/kdde.R d8a7251537f2428ad7a1f7a41b87b233 *R/kde-boundary.R eac35313fe00ee83ef666faf0bee0597 *R/kde-test.R beff4b967d92202b5a52f707f952c177 *R/kde.R 1b6256c05680f6da6bef46836a13b148 *R/kdr.R d22541e9ee20aff41356dd45a09767f6 *R/kfe.R 01702dc25944b2ee2a858306e0f3bfd0 *R/kfs.R 7e7e734c53eeda6391b7ebd4af128638 *R/kms.R bb90646a970cafc78fd71e0549bfaa15 *R/ksupp.R a3a4eeb009657cb64cb3b271960b0180 *R/mise.R 17f443e719ee81441de4fb270ed5ad2c *R/normal.R 44a686dfc4812fef4fb1f1e89022cc85 *R/prelim.R 069ffb307e6ae24b6d113d0391a591cc *R/selector.R 90145f629d37940030203306d26450ef *R/vkde.R bbc3657603a19ee8b50f437afc26618b *build/vignette.rds 3c7be27cb3c3505cb05a32dbdc3429ce *data/air.RData 421d741ffd92ca967c18a8f69a4824dc *data/cardio.RData 0b36987e874b237b5f47264783c15d56 *data/grevillea.RData 4d7c1de5dc3f1f1439c334bdab304544 *data/hsct.RData 2da4ac6e35cfa5c5b9873c2bc2549925 *data/plate.RData 90505be88b0945c0ff1e4fefb71fe405 *data/platesf.RData 538a3d83385e0ed088088650f01931b6 *data/quake.RData 56d0187718be4b9f8f6f076579a843a1 *data/quakesf.RData cfd66d7c906e51e34dcc3138079b5aa2 *data/tempb.RData 000e3e99be5ef662b38c44716894009a *data/unicef.RData 9bbcbfd4701d3cf7b336d8d4a2d15226 *data/worldbank.RData dcb71532f73008bec85c9ab501e62134 *inst/doc/kde.R cbdb42b6418bbc44cdfaa0ce34ceb548 *inst/doc/kde.Rnw 2647d5250b18e380471a1ecfa969fbc3 *inst/doc/kde.pdf 09cc4c7d52b9cb2b6b250f43d290d39d *man/Hbcv.Rd 36b1ae7ef6b788c79f1084fe3d061b90 *man/Hlscv.Rd 94b6c50b5508e697d65df32ec99445ca *man/Hnm.Rd 25eda66bd9de30f99432d3280f72ea74 *man/Hns.Rd 0f57de8a85583ba6bd78a114953f70e1 *man/Hpi.Rd f0b3df039566ccb1f0f03753411a5fd5 *man/Hscv.Rd 82b05328d68c4ed60be7c34ce9d36fad *man/air.Rd 853fee7af33b49718350ef339109f69b *man/binning.Rd c6e0dcfc5c75bdc4cb18136b12e4f732 *man/cardio.Rd b3ff052773e4e3e927b227529d31e512 *man/contour.Rd 313c5646b32d47574e9714df05471118 *man/grevillea.Rd 8adc954be040ce5baec02fa06162fff6 *man/histde.Rd b6c31dd6444d282f38b555cb580e4a0a *man/hsct.Rd d9ad3aa7eb053749870d03a2b3a5028b *man/ise.mixt.Rd cce5f940cef3b91d2285dfe4abd76437 *man/kcde.Rd 2eb237f8aede1bbac14a59c2309a5a2d *man/kcopula.Rd ba64937e9217d2de7d678ae032a6c980 *man/kda.Rd d7700cb9edb9663cf84c57b8c5b5b974 *man/kdcde.Rd 3a8307124b2466038809375e11280b62 *man/kdde.Rd da52d195ecc39a99ac902a647a08e630 *man/kde.Rd 7ba7254f9ed4bef3fb994b729b071324 *man/kde.boundary.Rd 54291c57f977e688a5093f1cfafbbf4b *man/kde.local.test.Rd c1bf17c5681ecfe9241d053d3d3a6feb *man/kde.test.Rd 3e40af3cee40624d05343ad5365ca7ea *man/kde.truncate.Rd 7aad3c1e751939165ab7097037196ed9 *man/kdr.Rd e5072b6e79c8bf9683f690502afa808a *man/kfe.Rd 75b4e70fbc7979b97531f08d2d1f571f *man/kfs.Rd 718adc41fd4a5933472782d07e9d0633 *man/kms.Rd 1515e35869b4b815ee2428b7391fd86b *man/kroc.Rd 5a5d408e97724631f72d960bbbef6800 *man/ks-internal.Rd 2febfbcd2a438705fbaac0ae7e83a986 *man/ks-package.Rd d0c5daf6f5a85ded5b13c54e2f2d6eb2 *man/ksupp.Rd c202c55f433f7b43e905aac384e1989b *man/mixt.Rd 7229879feeec7a251aee675b80a29575 *man/plot.histde.Rd 5e455c586daea5143c697f791aed6929 *man/plot.kcde.Rd 850a5679407630abb5ced9602c730b74 *man/plot.kda.Rd 73be2de7f2c5c2fc34595d244b2587d7 *man/plot.kdde.Rd a505a8a666de0db2c1ff21149d0edc76 *man/plot.kde.Rd 5567c48ecb71e70d6df5df59dabec801 *man/plot.kde.loctest.Rd ca4944aaf4470c6078d49265d45e301e *man/plot.kde.part.Rd a9627de16e87e013210243b219610cfb *man/plot.kfs.Rd 15a1e4ca32c6ee0cd471aaf4d8958f03 *man/plot.kroc.Rd 3912b617887530235767d1f3bd990a50 *man/plotmixt.Rd 12a1015540dc5666926c2bf420b95594 *man/pre.transform.Rd 9a3f68c8db72400289fc1e30cf6a3708 *man/quake.Rd 06a267a70e44e0a8c020601d8238b744 *man/rkde.Rd 0cd3c0323cf61c5095fcd405e0550951 *man/tempb.Rd c0aec91d5d02ba2d6266dfdffbcb3038 *man/unicef.Rd 62a053684bcc75557c31f7cb94af7f6b *man/vector.Rd 5c165aa40eaff1e65228a7109648bfa6 *man/vkde.Rd 5955f3a2ef78e8608d7987c819692f85 *man/worldbank.Rd 9f2af8aa0744c09e3ec177c998361f29 *src/ks.c cbdb42b6418bbc44cdfaa0ce34ceb548 *vignettes/kde.Rnw ks/R/0000755000176200001440000000000014606002552011070 5ustar liggesusersks/R/deconv-kde.R0000644000176200001440000000535114336512626013246 0ustar liggesusers###################################################################### ## Deconvolution KDE ###################################################################### dckde <- function(...) { return (kdcde(...)) } kdcde <- function(x, H, h, Sigma, sigma, reg, bgridsize, gridsize, binned, verbose=FALSE, ...) { ## default values ksd <- ks.defaults(x=x, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n binned <- ksd$binned gridsize <- ksd$gridsize bgridsize <- ksd$bgridsize x <- as.matrix(x) if (d==1 & missing(h)) h <- hpi(x=x, nstage=2, binned=binned, deriv.order=0) if (d>1 & missing(H)) H <- Hpi(x=x, nstage=2, binned=binned, deriv.order=0) if (d==1) stop("dckde not yet implemented for d=1") if (missing(reg)) reg <- reg.ucv(x=x, H=H, h=h, Sigma=Sigma, sigma=sigma, k=5, d=d, binned=binned, verbose=verbose) ## Deconvolution KDE is weighted KDE with non-uniform weights w <- dckde.weights(x=x, H=H, Sigma=Sigma, reg=reg) fhat <- kde(x=x, H=H, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize, verbose=verbose, ...) fhat$reg <- reg return(fhat) } ## weights for deconvolution KDE ## code adapted from DeconWK 0.6-5, author B. Turlach ## R-forge website: https://r-forge.r-project.org/R/?group_id=630 dckde.weights <- function(x, Sigma, H, reg) { n <- nrow(x) d <- ncol(x) Qmat <- matrix(0, ncol=n, nrow=n) bvec <- rep(0, n) for (j in 1:n) { Qmat[j,] <- dmvnorm.mixt(x, mus=x[j,], Sigmas=2*H + 2*Sigma, props=1) bvec[j] <- sum(dmvnorm.mixt(x, mus=x[j,], Sigmas=2*H + Sigma, props=1)) } if(!missing(reg)) diag(Qmat) <- diag(Qmat) + reg/n bvec <- bvec/n val <- kernlab::ipop(c=-bvec, H=Qmat, A=rep(1,n), b=1, r=0, l=rep(0,n), u=rep(1, n)) w <- kernlab::primal(val) w <- w/sum(w)*n return(w) } ## unbiased k-fold cross validation choice of regularisation penalty (gamma) reg.ucv <- function(x, H, h, Sigma, sigma, k=5, d, binned=FALSE, verbose=FALSE) { gamma.ucv.temp <- function(gamma) { return(-reg.ucv.val(x=x, H=H, Sigma=Sigma, k=k, reg=gamma^2, binned=binned)) } gamma.val <- nlm(f=gamma.ucv.temp, p=0.1, print.level=2*as.numeric(verbose))$estimate^2 return(gamma.val) } ## k-fold UCV value for regularisation penalty reg.ucv.val <- function(x, Sigma, H, reg, k=5, binned=FALSE) { n <- nrow(x) n.seq <- block.indices(n, n, npergroup=round(n/k)) cv.val <- 0 for (j in 1:(length(n.seq)-1)) { iind <- n.seq[j]:(n.seq[j+1]-1) w <- dckde.weights(x=x[-iind,], H=H, Sigma=Sigma, reg=reg) fhat <- kde(x=x[-iind,], H=H+Sigma, w=w, binned=binned) cv.val <- cv.val + sum(predict(fhat, x=x[iind,])) } return(cv.val) } ks/R/kde-boundary.R0000644000176200001440000007011614336776116013622 0ustar liggesusers###################################################################### ## Boundary KDE ###################################################################### kde.boundary <- function(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned=FALSE, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, boundary.supp, boundary.kernel="beta", verbose=FALSE) { bk <- match.arg(boundary.kernel, c("beta", "linear")) if (bk=="beta") { if (missing(boundary.supp)) boundary.supp <- 10 fhat <- kde.beta.boundary(x=x, H=H, h=h, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, compute.cont=compute.cont, approx.cont=approx.cont, boundary.supp=boundary.supp, verbose=verbose) } else if (bk=="linear") { if (missing(boundary.supp)) boundary.supp <- 2 fhat <- kde.linear.boundary(x=x, H=H, h=h, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, compute.cont=compute.cont, approx.cont=approx.cont, boundary.supp=boundary.supp, verbose=verbose) } return(fhat) } ###################################################################### ## Linear boundary KDE ###################################################################### kde.linear.boundary <- function(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned=FALSE, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, boundary.supp=2, verbose=FALSE) { ## default values ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w if (missing(binned)) binned <- ksd$binned if (missing(gridsize)) gridsize <- ksd$gridsize if (missing(bgridsize)) bgridsize <- gridsize ## clip data to xmin,xmax grid grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths if (d==1 & missing(h)) h <- hpi(x=x, nstage=2, binned=default.bflag(d=d, n=n)) if (d>1 & missing(H)) H <- Hpi(x=x, nstage=2, binned=default.bflag(d=d, n=n)) ## compute exact (non-binned) estimator ## 1-dimensional if (d==1) { stop("Not yet implemented.") } ## multi-dimensional else { if (is.data.frame(x)) x <- as.matrix(x) if (missing(eval.points)) { if (d==2) fhat <- kde.LB.grid.2d(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, boundary.supp=boundary.supp, binned=binned, verbose=verbose) else stop("Not yet implemented.") } else stop("Not yet implemented.") } fhat$binned <- binned fhat$names <- parse.name(x) ## add variable names fhat$w <- w class(fhat) <- "kde" ## compute prob contour levels if (compute.cont & missing(eval.points)) fhat$cont <- contourLevels(fhat, cont=1:99, approx=approx.cont) return(fhat) } ###################################################################### ## Bivariate linear boundary KDE ###################################################################### kde.LB.grid.2d <- function(x, H, gridsize, bgridsize, supp, gridx=NULL, grid.pts=NULL, xmin, xmax, gridtype, w, boundary.supp=10, binned=FALSE, verbose=FALSE) { n <- nrow(x) d <- ncol(x) if (missing(xmin)) xmin <- apply(x, 2, min) if (missing(xmax)) xmax <- apply(x, 2, max) if (missing(gridtype)) gridtype <- rep("linear", d) h <- sqrt(diag(H)) ## initialise grid if (is.null(gridx)) gridx <- make.grid.ks(x, matrix.sqrt(H), tol=supp, gridsize=gridsize, xmin=xmin, xmax=xmax, gridtype=gridtype) suppx <- make.supp(x, matrix.sqrt(H), tol=supp) if (is.null(grid.pts)) grid.pts <- find.gridpts(gridx, suppx) fhat.grid <- matrix(0, nrow=length(gridx[[1]]), ncol=length(gridx[[2]])) ## indicator for closeness to boundary bound.ind <- boundary.ind(x=x, h=h, boundary.supp=boundary.supp) n1 <- sum(!bound.ind) if (verbose) pb <- txtProgressBar() if (binned) { ## interior points - use normal kernel fhat.grid <- n1*kde(x=x[!bound.ind,], H=H, xmin=xmin, xmax=xmax, binned=binned, bgridsize=bgridsize)$estimate for (i in 1:n) { if (verbose) setTxtProgressBar(pb, i/n) if (bound.ind[i]) { ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.x.len <- length(eval.x) ## use linear boundary kernel for boundary points fhat <- dmvnorm.LB(x=expand.grid(eval.x, eval.y), mu=x[i,], Sigma=H) ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (j in 1:length(eval.y)) fhat.grid[eval.x.ind, eval.y.ind[j]] <- fhat.grid[eval.x.ind, eval.y.ind[j]] + w[i]*fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] } } } else { for (i in 1:n) { if (verbose) setTxtProgressBar(pb, i/n) ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.x.len <- length(eval.x) ## interior points - use normal kernel if (!bound.ind[i]) { eval.pts <- expand.grid(list(eval.x, eval.y)) fhat <- dmvnorm.mixt(x=eval.pts, mus=x[i,], Sigmas=H, props=1) } else { ## use linear boundary kernel for boundary points fhat <- dmvnorm.LB(x=expand.grid(eval.x, eval.y), mu=x[i,], Sigma=H) } ## place vector of density estimate values fhat onto grid fhat.grid for (j in 1:length(eval.y)) fhat.grid[eval.x.ind, eval.y.ind[j]] <- fhat.grid[eval.x.ind, eval.y.ind[j]] + w[i]*fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] } } fhat.grid <- fhat.grid/n gridx1 <- list(gridx[[1]], gridx[[2]]) fhat.grid <- fhat.grid/sum(fhat.grid*prod(sapply(gridx1,diff)[1,])) if (verbose) close(pb) fhat.list <- list(x=x, eval.points=gridx1, estimate=fhat.grid, H=H, gridtype=gridx$gridtype, gridded=TRUE, boundary=bound.ind) return(fhat.list) } ## bivariate linear boundary normal kernel dmvnorm.LB.kernel.2d <- function(x, H, xmin=c(0,0), xmax=c(1,1), ...) { x1 <- seq(xmin[1], xmax[1], length=151) x2 <- seq(xmin[2], xmax[2], length=151) eval.points <- list(x1, x2) fhat <- list() fhat$eval.points <- eval.points fhat$estimate <- array(dmvnorm.LB(x=expand.grid(eval.points), mu=x, Sigma=H), dim=c(length(x1), length(x2))) x <- rmvnorm.mixt(n=1000, mus=x, Sigmas=H, props=1) fhat$x <- x fhat$H <- H fhat$gridtype <- "linear" fhat$gridded <- TRUE fhat$binned <- FALSE fhat$names <- parse.name(x) fhat$w <- rep(1, nrow(x)) class(fhat) <- "kde" return(fhat) } dmvnorm.LB <- function(x, mu, Sigma, a0, a1) { if (!is.matrix(x)) x <- as.matrix(x) if (missing(a0) | missing(a1)) { ev <- -sweep(x, 2, mu, FUN="-") %*% diag(sqrt(1/diag(Sigma))) ev.list <- list(unique(ev[,1]), unique(ev[,2])) delta <- prod(unlist(lapply(lapply(ev.list, diff), getElement, 1))) d <- ncol(Sigma) evalK <- dmvnorm.mixt(x=ev, mus=rep(0,d), Sigmas=diag(d), props=1) m0 <- sum(evalK*delta) m1 <- apply(evalK*delta*ev, 2, sum) m2 <- apply(evalK*delta*rowKpow(ev, ev), 2, sum) M2 <- invvec(m2) M2inv <- chol2inv(chol(M2)) a0 <- 1/drop(m0 - t(m1) %*% M2inv%*%m1) a1 <- -a0*M2inv%*%m1 } evalK.LB <- (a0 - drop(ev %*% a1))*dmvnorm.mixt(x=x, mus=mu, Sigmas=Sigma, props=1) return(evalK.LB) } ###################################################################### ## Boundary kernel estimator using beta bounday kernels (2nd form) ###################################################################### kde.beta.boundary <- function(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned=FALSE, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, boundary.supp=1, verbose=FALSE) { ## default values ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w if (missing(binned)) binned <- ksd$binned if (missing(gridsize)) gridsize <- ksd$gridsize if (missing(bgridsize)) bgridsize <- gridsize ## if (missing(gridsize)) gridsize <- default.gridsize(d) ## clip data to xmin,xmax grid grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths if (d==1 & missing(h)) h <- hpi(x=x, nstage=2, binned=default.bflag(d=d, n=n)) if (d>1 & missing(H)) H <- Hpi(x=x, nstage=2, binned=default.bflag(d=d, n=n)) ## compute exact (non-binned) estimator ## 1-dimensional if (d==1) { if (missing(eval.points)) { fhat <- kde.boundary.grid.1d(x=x, h=h, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, boundary.supp=boundary.supp, binned=binned) } else stop("Not yet implemented.") # fhat <- kde.points.1d(x=x, h=h, eval.points=eval.points, positive=positive, adj.positive=adj.positive, w=w) } ## multi-dimensional else { if (is.data.frame(x)) x <- as.matrix(x) if (missing(eval.points)) { if (d==2) fhat <- kde.boundary.grid.2d(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, boundary.supp=boundary.supp, binned=binned, verbose=verbose) else if (d == 3) fhat <- kde.boundary.grid.3d(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, boundary.supp=boundary.supp, binned=binned, verbose=verbose) else stop("Need to specify eval.points for more than 3 dimensions") } else stop("Not yet implemented.") } fhat$binned <- binned fhat$names <- parse.name(x) ## add variable names fhat$w <- w class(fhat) <- "kde" ## compute prob contour levels if (compute.cont & missing(eval.points)) fhat$cont <- contourLevels(fhat, cont=1:99, approx=approx.cont) return(fhat) } kde.boundary.grid.1d <- function(x, h,gridsize, supp=3.7, xmin, xmax, gridtype, w, boundary.supp=0.5, binned=FALSE) { if (missing(xmin)) xmin <- min(x) if (missing(xmax)) xmax <- max(x) if (missing(gridtype)) gridtype <- "linear" ## transform x into [0,1] x.star <- (x-xmin)/(xmax-xmin) h.star <- h/(xmax-xmin) n <- length(x) gridtype1 <- match.arg(gridtype, c("linear", "sqrt")) if (gridtype1=="linear") eval.x <- seq(0, 1, length=gridsize) else if (gridtype1=="sqrt") { eval.x.temp <- seq(0, 1, length=gridsize) eval.x <- sign(eval.x.temp) * eval.x.temp^2 } gridtype.vec <- gridtype1 ## indicator for closeness to boundary of [0,1] bound.ind <- boundary.ind(x=x.star, h=h.star, boundary.supp=boundary.supp) n1 <- sum(!bound.ind) ## interior points - use normal kernel ## binned estimation only in the interior fhat.grid <- rep(0,length=gridsize) if (n1>0) { if (binned) { fhat.grid <- n1*kde(x=x.star[!bound.ind], h=h.star, xmin=0, xmax=1, binned=TRUE, bgridsize=gridsize)$estimate } else { fhat.grid <- n1*dnorm.mixt(x=eval.x, mus=x.star[!bound.ind], sigmas=rep(h.star, n1), props=w[!bound.ind]/n1) } } ## boundary points - use adjusted beta kernel hb.star <- 2*h.star for (i in 1:(n-n1)) fhat.grid <- fhat.grid + dbeta.kernel2(x=x.star[bound.ind][i], eval.x=eval.x, h=hb.star)*w[bound.ind][i] fhat.grid <- fhat.grid/n ## backtransform eval.points <- (xmax-xmin)*eval.x + xmin fhat.grid <- fhat.grid/(xmax-xmin) fhat <- list(x=x, eval.points=eval.points, estimate=fhat.grid, h=h, H=h^2, gridtype=gridtype.vec, gridded=TRUE) class(fhat) <- "kde" return(fhat) } kde.boundary.grid.2d <- function(x, H, gridsize, supp, gridx=NULL, grid.pts=NULL, xmin, xmax, gridtype, w, boundary.supp=1, binned=FALSE, verbose=FALSE) { n <- nrow(x) d <- ncol(x) if (missing(xmin)) xmin <- apply(x, 2, min) ##- h*supp if (missing(xmax)) xmax <- apply(x, 2, max) ## + h*supp if (missing(gridtype)) gridtype <- rep("linear", d) ## transform x into [0,1]^d x.star <- x for (j in 1:d) x.star[,j] <- (x[,j]-xmin[j])/(xmax[j]-xmin[j]) H.star <- diag(1/(xmax-xmin)) %*% H %*% diag(1/(xmax-xmin)) h.star <- sqrt(diag(H.star)) ## initialise grid if (is.null(gridx)) gridx <- make.grid.ks(x.star, matrix.sqrt(H.star), tol=supp, gridsize=gridsize, xmin=rep(0,d), xmax=rep(1,d), gridtype=gridtype) suppx <- make.supp(x.star, matrix.sqrt(H.star), tol=supp) if (is.null(grid.pts)) grid.pts <- find.gridpts(gridx, suppx) fhat.grid <- matrix(0, nrow=length(gridx[[1]]), ncol=length(gridx[[2]])) ## indicator for closeness to boundary of [0,1]^d bound.ind <- boundary.ind(x=x.star, h=h.star, boundary.supp=boundary.supp) n1 <- sum(!bound.ind) if (verbose) pb <- txtProgressBar() if (binned) { ## interior points - use normal kernel fhat.grid <- n1*kde(x=x.star[!bound.ind,], H=H.star, xmin=rep(0,d), xmax=rep(1,d), binned=TRUE, bgridsize=gridsize)$estimate for (i in 1:n) { if (verbose) setTxtProgressBar(pb, i/n) if (bound.ind[i]) { ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.x.len <- length(eval.x) ## convert bandwidth from normal kernel to beta kernel scale ## for boundary points hb.star <- 2*h.star fhat <- dmvbeta.prod.kernel2(x=x.star[i,], eval.x=list(eval.x, eval.y), hs=hb.star) ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (j in 1:length(eval.y)) fhat.grid[eval.x.ind, eval.y.ind[j]] <- fhat.grid[eval.x.ind, eval.y.ind[j]] + w[i]*fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] } } } else { for (i in 1:n) { if (verbose) setTxtProgressBar(pb, i/n) ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.x.len <- length(eval.x) ## interior points - use normal kernel if (!bound.ind[i]) { eval.pts <- expand.grid(eval.x, eval.y) fhat <- dmvnorm(eval.pts, x.star[i,], H.star) } else { ## convert bandwidth from normal kernel to beta kernel scale ## for boundary points hb.star <- 2*h.star fhat <- dmvbeta.prod.kernel2(x=x.star[i,], eval.x=list(eval.x, eval.y), hs=hb.star) } ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (j in 1:length(eval.y)) fhat.grid[eval.x.ind, eval.y.ind[j]] <- fhat.grid[eval.x.ind, eval.y.ind[j]] + w[i]*fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] } } fhat.grid <- fhat.grid/n if (verbose) close(pb) ## back-transform gridx1 <- list((xmax[1]-xmin[1])*gridx[[1]] + xmin[1], (xmax[2]-xmin[2])*gridx[[2]] + xmin[2]) fhat.grid <- fhat.grid/prod(xmax-xmin) fhat.list <- list(x=x, eval.points=gridx1, estimate=fhat.grid, H=H, gridtype=gridx$gridtype, gridded=TRUE, boundary=bound.ind) return(fhat.list) } kde.boundary.grid.3d <- function(x, H, gridsize, supp, gridx=NULL, grid.pts=NULL, xmin, xmax, gridtype, w, boundary.supp=0.5, verbose=FALSE, binned=FALSE) { n <- nrow(x) d <- ncol(x) if (missing(xmin)) xmin <- apply(x, 2, min) if (missing(xmax)) xmax <- apply(x, 2, max) if (missing(gridtype)) gridtype <- rep("linear", d) ## transform x into [0,1]^d x.star <- x for (j in 1:d) x.star[,j] <- (x[,j]-xmin[j])/(xmax[j]-xmin[j]) H.star <- diag(1/(xmax-xmin)) %*% H %*% diag(1/(xmax-xmin)) h.star <- sqrt(diag(H.star)) ## initialise grid if (is.null(gridx)) gridx <- make.grid.ks(x.star, matrix.sqrt(H.star), tol=supp, gridsize=gridsize, xmin=rep(0,d), xmax=rep(1,d), gridtype=gridtype) suppx <- make.supp(x.star, matrix.sqrt(H.star), tol=supp) if (is.null(grid.pts)) grid.pts <- find.gridpts(gridx, suppx) fhat.grid <- array(0, dim=c(length(gridx[[1]]), length(gridx[[2]]), length(gridx[[3]]))) ## indicator for closeness to boundary of [0,1]^d bound.ind <- boundary.ind(x=x.star, h=h.star, boundary.supp=boundary.supp) n1 <- sum(!bound.ind) if (verbose) pb <- txtProgressBar() if (binned) { ## interior points - use normal kernel fhat.grid <- n1*kde(x=x.star[!bound.ind,], H=H.star, xmin=rep(0,d), xmax=rep(1,d), binned=TRUE, bgridsize=gridsize)$estimate for (i in 1:n) { if (verbose) setTxtProgressBar(pb, i/n) if (bound.ind[i]) { ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.z <- gridx[[3]][grid.pts$xmin[i,3]:grid.pts$xmax[i,3]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.z.ind <- c(grid.pts$xmin[i,3]:grid.pts$xmax[i,3]) eval.x.len <- length(eval.x) eval.pts <- expand.grid(eval.x, eval.y) ## convert bandwidth from normal kernel to beta kernel scale ## for boundary points hb.star <- 2*h.star fhat.xy <- dmvbeta.prod.kernel2(x=x.star[i,], eval.x=list(eval.x, eval.y), hs=hb.star[1:2]) ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (k in 1:length(eval.z)) { fhat <- w[i]*cbind(fhat.xy, dbeta.kernel2(x=x.star[i,3], eval.x=eval.z[k], h=hb.star[3])) for (j in 1:length(eval.y)) fhat.grid[eval.x.ind,eval.y.ind[j], eval.z.ind[k]] <- fhat.grid[eval.x.ind, eval.y.ind[j], eval.z.ind[k]] + fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] } } } } else { for (i in 1:n) { if (verbose) setTxtProgressBar(pb, i/n) ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.z <- gridx[[3]][grid.pts$xmin[i,3]:grid.pts$xmax[i,3]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.z.ind <- c(grid.pts$xmin[i,3]:grid.pts$xmax[i,3]) eval.x.len <- length(eval.x) eval.pts <- expand.grid(eval.x, eval.y) ## interior points - use normal kernel if (!bound.ind[i]) { ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (k in 1:length(eval.z)) { fhat <- w[i]*dmvnorm(cbind(eval.pts, eval.z[k]), x[i,], H) for (j in 1:length(eval.y)) fhat.grid[eval.x.ind,eval.y.ind[j], eval.z.ind[k]] <- fhat.grid[eval.x.ind, eval.y.ind[j], eval.z.ind[k]] + fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] } } else { ## convert bandwidth from normal kernel to beta kernel scale hb.star <- 2*h.star fhat.xy <- dmvbeta.prod.kernel2(x=x.star[i,], eval.x=list(eval.x, eval.y), hs=hb.star[1:2]) for (k in 1:length(eval.z)) { fhat <- w[i]*cbind(fhat.xy, dbeta.kernel2(x=x.star[i,3], eval.x=eval.z[k], h=hb.star[3])) for (j in 1:length(eval.y)) fhat.grid[eval.x.ind,eval.y.ind[j], eval.z.ind[k]] <- fhat.grid[eval.x.ind, eval.y.ind[j], eval.z.ind[k]] + fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] } } } } fhat.grid <- fhat.grid/n if (verbose) close(pb) ## back-transform gridx1 <- list((xmax[1]-xmin[1])*gridx[[1]] + xmin[1], (xmax[2]-xmin[2])*gridx[[2]] + xmin[2], (xmax[3]-xmin[3])*gridx[[3]] + xmin[3]) fhat.grid <- fhat.grid/prod(xmax-xmin) fhat.list <- list(x=x, eval.points=gridx1, estimate=fhat.grid, H=H, gridtype=gridx$gridtype, gridded=TRUE, boundary=bound.ind) return(fhat.list) } ## indicator function for boundary region of [0,1] i.e. [0,h] + [1-h, h] boundary.ind <- function(x, h, xmin, xmax, boundary.supp=1) { if (is.vector(x)) { x <- matrix(x, ncol=1) } n <- nrow(x) d <- ncol(x) ## indicator for closeness to boundary of [0,1]^d bound.ind <- matrix(NA, nrow=n, ncol=d) for (j in 1:d) bound.ind[,j] <- (abs(x[,j]) <= boundary.supp*h[j]) | (abs(1-x[,j]) <= boundary.supp*h[j]) bound.ind <- apply(bound.ind, 1, any) return(bound.ind) } ###################################################################### ## Bivariate beta boundary KDE ###################################################################### ## modified boundary beta kernel - first form (Chen, 1999) dbeta.kernel <- function(x, eval.x, h) { return (dbeta(x=eval.x, shape1=x/h^2+1, shape2=(1-x)/h^2+1)) } ## modified boundary beta kernel - second form (Chen, 1999) dbeta.kernel2 <- function(x, eval.x, h) { rhox <- function(y, hy) { if (y==0) return (1) else return(2*hy^4 + 5/2 - sqrt(4*hy^8 + 6*hy^4 + 9/4 - y^2 -y/hy^2)) } ind <- cut(x, c(0, 2*h^2, 1-2*h^2, 1), labels=FALSE, include.lowest=TRUE) dbk <- rep(0, length(eval.x)) if (ind==1) { shape1 <- rhox(x,hy=h); shape2 <- (1-x)/h^2 } else if (ind==2) { shape1 <- x/h^2; shape2 <- (1-x)/h^2 } else if (ind==3) { shape1 <- x/h^2; shape2 <- rhox(1-x, hy=h) } return(dbeta(eval.x, shape1=shape1, shape2=shape2)) } ## modified multivariate boundary beta product kernel dmvbeta.prod.kernel2 <- function(x, eval.x, hs) { d <- length(hs) db <- vector("list", d) for (i in 1:d) { db[[i]] <- 0 db[[i]] <- dbeta.kernel2(x=x[i], eval.x=eval.x[[i]], h=hs[i]) } db <- expand.grid(db) db <- apply(db, 1, prod) return(db) } ## modified multivariate boundary beta spherically symmetric kernel dmvbeta.symm.kernel2 <- function(x, eval.x, H) { d <- ncol(H) eval.y <- sqrt(apply(eval.x^2, 1, sum))/sqrt(d) y <- sqrt(sum(x^2))/sqrt(d) return(dbeta.kernel2(x=y, eval.x=eval.y, h=sqrt(tr(H)))/d) } rbeta.kernel2 <- function(x, n, h) { rhox <- function(y, hy) { if (y==0) return (1) else return(2*hy^4 + 5/2 - sqrt(4*hy^8 + 6*hy^4 + 9/4 - y^2 -y/hy^2)) } ind <- cut(x, c(0, 2*h^2, 1-2*h^2, 1), labels=FALSE, include.lowest=TRUE) if (ind==1) { shape1 <- rhox(x,hy=h); shape2 <- (1-x)/h^2 } else if (ind==2) { shape1 <- x/h^2; shape2 <- (1-x)/h^2 } else if (ind==3) { shape1 <- x/h^2; shape2 <- rhox(1-x, hy=h) } return(rbeta(n=n, shape1=shape1, shape2=shape2)) } dmvbeta.prod.kernel2.2d <- function(x, hs, xmin=c(0,0), xmax=c(1,1), ...) { x.star <- (x - xmin)/(xmax - xmin) hs.star <- hs/(xmax - xmin) x1 <- seq(0,1, length=151) x1[1] <- 1e-9; x1[151] <- 1-1e-9 eval.points <- list(x1, x1) fhat <- list() x <- cbind(rbeta.kernel2(x=x.star[1], n=1000, h=hs.star[1]), rbeta.kernel2(x=x.star[2], n=1000, h=hs.star[2])) fhat$eval.points <- list(seq(0,1, length=151), seq(0,1, length=151)) fhat$estimate <- matrix(dmvbeta.prod.kernel2(x=x.star, eval.x=eval.points, hs=hs.star, ...), nrow=length(eval.points[[1]])) fhat$x <- sweep(sweep(x, 2, xmin, FUN="+"), 2, xmax-xmin, FUN="*") fhat$eval.points[[1]] <- xmin[1] + fhat$eval.points[[1]]*(xmax[1]-xmin[1]) fhat$eval.points[[2]] <- xmin[2] + fhat$eval.points[[2]]*(xmax[2]-xmin[2]) fhat$H <- diag(2) fhat$gridtype <- "linear" fhat$gridded <- TRUE fhat$binned <- FALSE fhat$names <- parse.name(x) fhat$w <- rep(1, nrow(x)) class(fhat) <- "kde" return(fhat) } ########################################################################## ## Truncate unbounded KDE to polygon boundary ########################################################################## kde.truncate <- function(fhat, boundary) { ## reallocate any probability mass outside of map boundary regions ## to interior regions truncate.ind <- array(mgcv::in.out(boundary, as.matrix(expand.grid(fhat$eval.points[[1]], fhat$eval.points[[2]]))), dim=dim(fhat$estimate)) fhat.trunc <- fhat fhat.trunc$estimate <- fhat.trunc$estimate*truncate.ind fhat.sum <- contourProbs(fhat, abs.cont=0) fhat.trunc.sum <- contourProbs(fhat.trunc, abs.cont=0) fhat.trunc$estimate <- fhat.trunc$estimate*fhat.sum/fhat.trunc.sum fhat.trunc.temp <- fhat.trunc fhat.trunc.temp$x <- fhat.trunc.temp$x[mgcv::in.out(boundary, fhat.trunc$x),] fhat.trunc$cont <- contourLevels(fhat.trunc.temp, cont=1:99, approx=TRUE) return(fhat.trunc) } ########################################################################## ## Truncate unbounded KDDE to polygon boundary ########################################################################## kdde.truncate <- function(fhat, boundary) { ## reallocate any probability mass outside of map boundary regions ## to interior regions fhat.trunc <- fhat fhat0 <- kde(x=fhat$x, H=fhat$H) fhat0.sum <- contourProbs(fhat0, abs.cont=0) fhat0.trunc <- kde.truncate(fhat=fhat0, boundary=boundary) fhat0.trunc.sum <- contourProbs(fhat0.trunc, abs.cont=0) for (i in 1:length(fhat$estimate)) { truncate.ind <- array(mgcv::in.out(boundary, as.matrix(expand.grid(fhat$eval.points[[1]], fhat$eval.points[[2]]))), dim=dim(fhat$estimate[[i]])) fhat.trunc$estimate[[i]] <- fhat.trunc$estimate[[i]]*truncate.ind fhat.trunc$estimate[[i]] <- fhat.trunc$estimate[[i]]*fhat0.sum/fhat0.trunc.sum } fhat.trunc.temp <- fhat.trunc fhat.trunc.temp$x <- fhat.trunc.temp$x[mgcv::in.out(boundary, fhat.trunc$x),] fhat.trunc$cont <- contourLevels(fhat.trunc.temp, cont=1:99, approx=TRUE) return(fhat.trunc) } ks/R/kdr.R0000644000176200001440000003346414547756063012026 0ustar liggesusers###################################################################### ## Kernel density ridge estimation for 2D/3D data ##################################################################### kdr <- function(x, y, H, p=1, max.iter=400, tol.iter, segment=TRUE, k, kmax, min.seg.size, keep.path=FALSE, gridsize, xmin, xmax, binned, bgridsize, w, fhat, density.cutoff, pre=TRUE, verbose=FALSE) { ## default values xnames <- parse.name(x) x <- as.matrix(x) x.orig <- x if (pre) { S12 <- diag(apply(x.orig, 2, sd)) Sinv12 <- matrix.pow(S12,-1) x <- pre.scale(x) if (!missing(xmin)) xmin <- xmin %*% Sinv12 if (!missing(xmax)) xmax <- xmax %*% Sinv12 rescale <- function(x) { as.matrix(x) %*% S12 } } ksd <- ks.defaults(x=x, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned bgridsize <- ksd$bgridsize gridsize <- ksd$gridsize ## default bandwidth if (missing(H)) H <- Hpi(x=x, nstage=2-(d>2), binned=binned, deriv.order=2, verbose=verbose) Hinv <- chol2inv(chol(H)) tol <- 3.7 tol.H <- tol * diag(H) if (missing(xmin)) xmin <- apply(x, 2, min) - tol.H if (missing(xmax)) xmax <- apply(x, 2, max) + tol.H if (missing(tol.iter)) tol.iter <- 1e-3*min(apply(x, 2, IQR)) if (missing(y)) { xx <- seq(xmin[1], xmax[1], length = gridsize[1]) yy <- seq(xmin[2], xmax[2], length = gridsize[2]) if (d==2) y <- expand.grid(xx, yy) else if (d==3) { zz <- seq(xmin[3], xmax[3], length = gridsize[3]) y <- expand.grid(xx, yy, zz) } } else { y <- as.matrix(y); if (pre) y <- y %*% Sinv12 } if (is.vector(y)) y <- matrix(y, nrow=1) if (missing(min.seg.size)) min.seg.size <- round(1e-3*nrow(y), 0) ## exclude low density regions from ridge search if (missing(fhat)) fhat <- kde(x=x, w=w, binned=binned) if (missing(density.cutoff)) density.cutoff <- contourLevels(fhat, cont=99) y.ind <- predict(fhat, x=y)>density.cutoff y <- y[y.ind,] fhat2 <- kdde(x=x, H=H, deriv.order=2, xmin=xmin, xmax=xmax, binned=binned, bgridsize=bgridsize, gridsize=gridsize, w=w, verbose=verbose) ## projected gradient mean shift iterations n.seq <- block.indices(n, nrow(y), d=d, r=0, diff=FALSE)#, block.limit=1e6) if (verbose) pb <- txtProgressBar() pc <- list() i <- 1 if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) pc <- kdr.base(x=x, fhat2=fhat2, y=y[n.seq[i]:(n.seq[i+1]-1),], H=H, tol.iter=tol.iter, Hinv=Hinv, verbose=verbose, max.iter=max.iter, p=p) if (pre) { pc[c("x","y","end.points")] <- lapply(pc[c("x","y","end.points")], rescale) pc[["path"]] <- lapply(pc[["path"]], rescale) } if (length(n.seq)>2) { for (i in 2:(length(n.seq)-1)) { if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) pc.temp <- kdr.base(x=x, fhat2=fhat2, y=y[n.seq[i]:(n.seq[i+1]-1),], H=H, tol.iter=tol.iter, Hinv=Hinv, verbose=verbose, max.iter=max.iter, p=p) if (pre) { pc.temp[c("y","end.points")] <- lapply(pc.temp[c("y","end.points")], rescale) pc.temp[["path"]] <- lapply(pc.temp[["path"]], rescale) } pc$y <- rbind(pc$y, pc.temp$y) pc$end.points <- rbind(pc$end.points, pc.temp$end.points) pc$path <- c(pc$path, pc.temp$path) } } if (verbose) close(pb) ## remove short segments for p=1 if (p==1) { tol.seg <- 1e-2*max(apply(x.orig, 2, IQR)) pc.dendo <- hclust(dist(pc$end.points), method="single") pc.label <- cutree(pc.dendo, h=tol.seg) pc.label.ind <- pc.label %in% which(table(pc.label)>min.seg.size) pc$y <- pc$y[pc.label.ind,] pc$end.points <- pc$end.points[pc.label.ind,] pc$path <- pc$path[pc.label.ind] } pc$H <- H pc$names <- xnames if (pre) pc$H <- S12 %*% pc$H %*% S12 if (segment) pc <- kdr.segment(x=pc, k=k, kmax=kmax, min.seg.size=min.seg.size, verbose=verbose) else pc$end.points <- data.frame(pc$end.points, segment=1L) ## put paths as last element in list path.temp <- pc$path pc$path <- NULL pc$tol.iter <- tol.iter pc$min.seg.size <- min.seg.size pc$binned <- binned pc$names <- xnames pc$w <- w if (keep.path) pc$path <- path.temp return(pc) } kdr.base <-function(x, fhat2, H, y, max.iter, tol.iter, p=1, verbose=FALSE, Hinv, ...) { if (!is.matrix(x)) x <- as.matrix(x) if (!is.matrix(y)) y <- as.matrix(y) if (missing(Hinv)) Hinv <- chol2inv(chol(H)) nx <- nrow(x) ny <- nrow(y) d <- ncol(y) y.path <- split(y, row(y), drop=FALSE) names(y.path) <- NULL xHinv <- x %*% Hinv xHinvx <- rowSums(xHinv*x) y.update <- y i <- 1 eps <- max(sqrt(rowSums(y.update^2))) disp.ind <- head(sample(1:nrow(y)), n=min(1000,nrow(y))) while (eps > tol.iter & i < max.iter) { y.curr <- y.update yHinvy <- t(rowSums(y.curr%*%Hinv *y.curr)) Mah <- apply(yHinvy, 2, "+", xHinvx) - 2*xHinv %*% t(y.curr) w <- exp(-Mah/2) denom <- colSums(w) num <- t(w)%*%x mean.shift.H <- num/denom - y.curr fhat2.y.curr <- predict(fhat2, x=y.curr) for (j in 1:ny) { Hessian <- invvec(fhat2.y.curr[j,]) Hessian.svd <- eigen(Hessian, symmetric=TRUE) Up <- Hessian.svd$vectors[,tail(1:d,n=d-p)] mean.shift.H[j,] <- drop(Up %*% t(Up) %*% mean.shift.H[j,]) } y.update <- y.curr + mean.shift.H y.update.list <- split(y.update, row(y.update), drop=FALSE) y.path <- mapply(rbind, y.path, y.update.list, SIMPLIFY=FALSE) eps <- max(sqrt(rowSums((y.curr-y.update)^2))) if (verbose>1) { if (d==2) plot(y.update[disp.ind,], col=1, xlab="x", ylab="y") else pairs(y.update[disp.ind,], col=1) } i <- i+1 } pc.endpt <- t(sapply(y.path, tail, n=1, SIMPLIFY=FALSE)) pc <- list(x=x, y=y, end.points=pc.endpt, path=y.path, type="kdr") class(pc) <- "kdr" return(pc) } ## create segment of KDR filaments ## x = output from kdr kdr.segment <- function(x, k, kmax, min.seg.size, verbose=FALSE) { ep <- x$end.points if (any(names(ep) %in% "segment")) ep <- x$end.points[,-which(names(ep)=="segment")] if (missing(min.seg.size)) min.seg.size <- x$min.seg.size hc <- hclust(dist(ep), method="single") if (missing(kmax)) kmax <- 30 kmax <- min(kmax, nrow(ep)) if (missing(k)) { if (verbose) pb <- txtProgressBar() clust.ind <- rep(0, kmax) for (i in 1:kmax) { if (verbose) setTxtProgressBar(pb, i/kmax) clust.ind[i] <- clust.crit(hc=hc, x=as.matrix(ep), k=i, min.seg.size=min.seg.size) } if (verbose) close(pb) clust.ind[is.na(clust.ind)] <- 0 kopt <- which.max(clust.ind) } else kopt <- k ep <- data.frame(ep, segment=as.integer(cutree(hc, k=kopt))) label <- ep$segment tlabel <- as.integer(names(table(label))[table(label) > min.seg.size]) ep <- ep[label %in% tlabel,] ep$segment <- factor(ep$segment, labels=1:length(unique(ep$segment))) ep$segment <- as.integer(levels(ep$segment))[ep$segment] ## re-order KDR segments into 'reasonable' linestring order ## experimental j <- 1 for (i in unique(ep$segment)) { ep.temp <- as.matrix(ep[ep$segment==i,-ncol(ep)]) ep.temp <- data.frame(chain.knnx(ep.temp, k1=1, k2=1), segment=i) if (j==1) ep.ord <- ep.temp else ep.ord <- rbind(ep.ord, ep.temp) j <- j+1 } names(ep.ord) <- c(x$names, "segment") rownames(ep.ord) <- NULL x$end.points <- ep.ord x$min.seg.size <- min.seg.size x$k <- kopt if (exists("clust.ind")) x$clust.ind <- clust.ind return(x) } ## rbind nearest neighbour of y from x to y add.knnx <- function(x, y, k=1) { xynn <- FNN::get.knnx(x, y, k=k) y <- rbind(y, x[xynn$nn.index,]) d <- ncol(x) xy <- list(x=matrix(x[-xynn$nn.index,],ncol=d), y=y) return(xy) } ## arrange points in KDR to form a "reasonable" linestring chain.knnx <- function(x, k1=1, k2=5) { ## concatenate the nearest neighbours in a chain ## start with first point in x if (!is.matrix(x)) x <- as.matrix(x) d <- ncol(x) if (nrow(x)>1) { x.ord.list <- add.knnx(x=matrix(x[-1,], ncol=d), y=matrix(x[1,], ncol=d), k=k1) x.ord <- x.ord.list$y while (nrow(x.ord.list$x)>0) { y.temp <- matrix(apply(as.matrix(tail(x.ord.list$y, n=k2)), 2, mean), ncol=d) colnames(y.temp) <- names(x.ord.list$x) x.ord.list.temp <- add.knnx(x=x.ord.list$x, y=y.temp, k=k1) x.ord <- rbind(x.ord, matrix(tail(x.ord.list.temp$y,n=1),ncol=d)) x.ord.list <- x.ord.list.temp } ## decide which permutation is "best" linestring ## break at max discontinuity ind <- which.max(rowSums((head(x.ord, n=-1)-tail(x.ord,n=-1))^2)) ind1 <- c(1:ind, (ind+1):nrow(x.ord)) ind2 <- c(1:ind, rev((ind+1):nrow(x.ord))) ind3 <- c(rev(1:ind), (ind+1):nrow(x.ord)) ind4 <- c(rev(1:ind), rev((ind+1):nrow(x.ord))) x.ord1 <- x.ord[ind1,] x.ord2 <- x.ord[ind2,] x.ord3 <- x.ord[ind3,] x.ord4 <- x.ord[ind4,] x.ord.dist <- rep(0,4) x.ord.dist[1] <- sum(rowSums((head(x.ord1, n=-1)-tail(x.ord1,n=-1))^2)) x.ord.dist[2] <- sum(rowSums((head(x.ord2, n=-1)-tail(x.ord2,n=-1))^2)) x.ord.dist[3] <- sum(rowSums((head(x.ord3, n=-1)-tail(x.ord3,n=-1))^2)) x.ord.dist[4] <- sum(rowSums((head(x.ord4, n=-1)-tail(x.ord4,n=-1))^2)) x.ord <- get(paste0("x.ord", which.min(x.ord.dist))) } else x.ord <- x return(x.ord) } ## Calinski-Harabasz clustering criterion for hierarchical clustering object clust.crit <- function(hc, x, k, min.seg.size=1) { label <- cutree(hc, k=k) tlabel <- as.integer(names(table(label))[table(label) > min.seg.size]) tlabel.ind <- label %in% tlabel cc <- fpc.calinhara(x=x, clustering=label) return(cc) } ## copied from fpc::calinhara 2020-09-18 fpc.calinhara <- function(x, clustering, cn = max(clustering)) { x <- as.matrix(x) p <- ncol(x) n <- nrow(x) cln <- rep(0, cn) W <- matrix(0, p, p) for (i in 1:cn) cln[i] <- sum(clustering == i) for (i in 1:cn) { clx <- x[clustering == i, ] cclx <- cov(as.matrix(clx)) if (cln[i] < 2) cclx <- 0 W <- W + ((cln[i] - 1) * cclx) } S <- (n - 1) * cov(x) B <- S - W out <- (n - cn) * sum(diag(B))/((cn - 1) * sum(diag(W))) return(out) } ############################################################################# ## S3 methods for KDR objects ############################################################################# ## plot method plot.kdr <- function(x, ...) { fhat <- x d <- ncol(fhat$x) if (d==2) { plotret <- plotkdr.2d(fhat, ...) invisible(plotret) } else if (d==3) { plotkdr.3d(fhat, ...) invisible() } else stop ("Plot function only available for 2 or 3-d data") } plotkdr.2d <- function(x, add=FALSE, col, type="p", alpha=1, ...) { xp <- x$end.points if (!any(names(xp) %in% "segment")) { if (missing(col)) col <- 6 col <- transparency.col(col, alpha=alpha) if (!add) plot(xp, col=col, type=type, ...) else points(xp, col=col, ...) } else { xps <- unique(xp$segment) if (missing(col)) col <- hcl.colors(n=length(xps), palette="Set2") if (length(col) < length(xps)) col <- rep(col, length(xps)) col <- transparency.col(col, alpha=alpha) if (!add) plot(xp[,-ncol(xp)], col="transparent", ...) for (i in 1:length(xps)) lines(xp[xp$segment==xps[i],-ncol(xp)], col=col[i], type=type, ...) } } plotkdr.3d <- function(x, display="plot3D", colors, col, col.fun, alphavec, size=3, cex=1, pch=1, theta=-30, phi=40, d=4, ticktype="detailed", add=FALSE, xlab, ylab, zlab, alpha=1, box=TRUE, axes=TRUE, type="p", ...) { fhat <- x xp <- x$end.points if (missing(xlab)) xlab <- fhat$names[1] if (missing(ylab)) ylab <- fhat$names[2] if (missing(zlab)) zlab <- fhat$names[3] if (!any(names(xp) %in% "segment")) { if (missing(col)) col <- 6 } else { xps <- unique(xp$segment) if (missing(col)) col <- hcl.colors(n=length(xps), palette="Set2") if (length(col) < length(xps)) col <- rep(col, length(xps)) } colors <- col disp <- match.arg(display, c("plot3D", "rgl")) if (disp %in% "plot3D") { if (!requireNamespace("plot3D", quietly=TRUE)) stop("Install the plot3D package as it is required.", call.=FALSE) if (!add) plot3D::scatter3D(x=xp[,1], y=xp[,2], z=xp[,3], add=add, theta=theta, phi=phi, d=d, type=type, xlab=xlab, ylab=ylab, zlab=zlab, ticktype=ticktype, type="n", col=NA, ...) for (i in 1:length(xps)) plot3D::scatter3D(x=xp[xp$segment==xps[i],1], y=xp[xp$segment==xps[i],2], z=xp[xp$segment==xps[i],3], cex=cex, col=col[i], add=TRUE, pch=pch, type=type, alpha=alpha, ...) } else if (disp %in% "rgl") { if (!requireNamespace("rgl", quietly=TRUE)) stop("Install the rgl package as it is required.", call.=FALSE) for (i in 1:length(xps)) rgl::plot3d(x=xp[xp$segment==xps[i],1], y=xp[xp$segment==xps[i],2], z=xp[xp$segment==xps[i],3], col=col[i], alpha=alpha, xlab=xlab, ylab=ylab, zlab=zlab, add=add | (i>1), box=box, axes=axes, type=type, size=size, ...) } } ks/R/kde.R0000644000176200001440000012712314606003615011765 0ustar liggesusers############################################################################### ## Multivariate kernel density estimators ############################################################################### ############################################################################### ## Multivariate kernel density estimate using normal kernels ## ## Parameters ## x - points ## H - bandwidth matrix ## gridsize - number of interval points in grid ## supp - effective support of kernel ## eval.points - compute density estimate at these points (if missing ## and dim(x) = 2, 3 compute density estimate over grid) ## eval.levels - compute 3-D in 2-D slices taken at these level curves ## ## Returns ## list with first d components with the points that the density ## estimate is evaluated at, and values of the density estimate ############################################################################## kde <- function(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, compute.cont=TRUE, approx.cont=TRUE, unit.interval=FALSE, density=FALSE, verbose=FALSE) { ## default values ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned gridsize <- ksd$gridsize bgridsize <- ksd$bgridsize if (binned & d>4) stop("Binned estimation for d>4 not implemented. Set binned=FALSE for exact estimation.") ## clip data to xmin, xmax grid limits for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] if (positive & missing(xmin)) { xmin <- rep(0,d) } if (unit.interval) { if (missing(xmin)) xmin <- rep(0,d); if (missing(xmax)) xmax <- rep(1,d) } xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths if (d==1 & missing(h)) { if (positive) x1 <- log(x) else x1 <- x if (unit.interval) x1 <- qnorm(x) h <- hpi(x=x1, nstage=2, binned=default.bflag(d=d, n=n), deriv.order=0) } if (d>1 & missing(H)) { if (positive) x1 <- log(x) else x1 <- x H <- Hpi(x=x1, nstage=2, binned=default.bflag(d=d, n=n), deriv.order=0) } ## compute binned estimator if (binned) { if (positive) { if (d==1) { fhat <- kde.positive.1d(x=x, h=h, bgridsize=bgridsize, xmin=xmin, xmax=xmax, w=w, binned=binned, adj.positive=adj.positive) } else if (d==2) { fhat <- kde.positive.2d(x=x, H=H, bgridsize=bgridsize, xmin=xmin, xmax=xmax, w=w, binned=binned, adj.positive=adj.positive) } } else if (unit.interval) { fhat <- kde.unit.interval.1d(x=x, w=w, binned=binned, h=h) } else { fhat <- kdde.binned(x=x, H=H, h=h, bgridsize=bgridsize, xmin=xmin, xmax=xmax, w=w, deriv.order=0, verbose=verbose) } if (!missing(eval.points)) { fhat$estimate <- predict(fhat, x=eval.points) fhat$eval.points <- eval.points } } else { ## compute exact (non-binned) estimator ## 1-dimensional if (d==1) { if (missing(eval.points)) { if (unit.interval) fhat <- kde.unit.interval.1d(x=x, w=w, h=h, binned=FALSE) else if (positive) fhat <- kde.positive.1d(x=x, h=h, xmin=xmin, xmax=xmax, w=w, binned=FALSE, adj.positive=adj.positive) else fhat <- kde.grid.1d(x=x, h=h, gridsize=gridsize, supp=supp, positive=positive, xmin=xmin, xmax=xmax, adj.positive=adj.positive, gridtype=gridtype, w=w) } else fhat <- kde.points.1d(x=x, h=h, eval.points=eval.points, positive=positive, adj.positive=adj.positive, w=w) } ## multi-dimensional else { if (is.data.frame(x)) x <- as.matrix(x) if (missing(eval.points)) { if (d==2) { if (positive) fhat <- kde.positive.2d(x=x, H=H, gridsize=gridsize, xmin=xmin, xmax=xmax, w=w, binned=binned, adj.positive=adj.positive) else fhat <- kde.grid.2d(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, verbose=verbose) } else if (d==3) fhat <- kde.grid.3d(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, verbose=verbose) else fhat <- kde.grid.nd(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, verbose=verbose) } else fhat <- kde.points(x=x, H=H, eval.points=eval.points, w=w, verbose=verbose) } } if (density) fhat$estimate[fhat$estimate<0] <- 0 fhat$binned <- binned fhat$names <- parse.name(x) ## add variable names fhat$w <- w fhat$type <- "kde" class(fhat) <- "kde" ## compute prob contour levels if (compute.cont & missing(eval.points)) fhat$cont <- contourLevels(fhat, cont=1:99, approx=approx.cont) return(fhat) } ############################################################################### ## Univariate kernel density estimate on a grid ############################################################################### kde.grid.1d <- function(x, h, gridsize, supp=3.7, positive=FALSE, adj.positive, xmin, xmax, gridtype, w) { if (missing(xmin)) xmin <- min(x) - h*supp if (missing(xmax)) xmax <- max(x) + h*supp if (missing(gridtype)) gridtype <- "linear" if (positive) { if (missing(adj.positive)) adj.positive <- abs(min(x)) y <- log(x + adj.positive) ## transform positive data x to real line gridx <- seq(max(0, xmin), xmax, length=gridsize) gridy <- log(gridx + adj.positive) gridtype.vec <- "linear" } else { y <- x gridtype1 <- match.arg(gridtype, c("linear", "sqrt", "quantile", "exp")) if (gridtype1=="linear") { gridy <- seq(xmin, xmax, length=gridsize) } else if (gridtype1=="sqrt") { gridy.temp <- seq(sign(xmin)*sqrt(abs(xmin)), sign(xmax)*sqrt(abs(xmax)), length=gridsize) gridy <- sign(gridy.temp) * gridy.temp^2 } else if (gridtype1=="exp") { gridy.temp <- seq(exp(xmin), exp(xmax), length=gridsize) gridy <- log(gridy.temp) } gridtype.vec <- gridtype1 } n <- length(y) est <- dnorm.mixt(x=gridy, mus=y, sigmas=rep(h, n), props=w/n) fhat <- list(x=y, eval.points=gridy, estimate=est, h=h, H=h^2, gridtype=gridtype.vec, gridded=TRUE) if (positive) { ## compute transformation KDE fhat$estimate <- fhat$estimate/(exp(gridy)) fhat$x <- x fhat$eval.points <- exp(gridy) - adj.positive } class(fhat) <- "kde" return(fhat) } kde.positive.1d <- function(x, h, adj.positive, binned=FALSE, xmin, xmax, compute.cont=TRUE, approx.cont=TRUE, ...) { if (missing(adj.positive)) adj.positive <- abs(min(x)) y <- log(x + adj.positive) if (missing(h)) h <- hpi(y, binned=binned) d <- 1 tol <- 3.7 tol.h <- tol*h if (missing(xmin)) xmin <- min(x) - tol.h if (missing(xmax)) xmax <- max(x) + tol.h xmin[xmin<0] <- 0 ymin1 <- log(xmin + adj.positive) ymax1 <- log(xmax + adj.positive) fhaty <- kde(x=y, h=h, xmin=ymin1, xmax=ymax1, gridtype=c("exp"), binned=binned, compute.cont=compute.cont, approx.cont=approx.cont, ...) fhaty$estimate[is.nan(fhaty$estimate)] <- 0 fhatx <- fhaty fhatx$x <- x fhatx$eval.points <- exp(fhaty$eval.points) - adj.positive jacobian <- abs(exp(fhaty$eval.points)) jacobian[jacobian<=0] <- min(fhatx$estimate[fhatx$estimate>0]) fhatx$estimate <- fhaty$estimate/jacobian if (compute.cont) fhatx$cont <- contourLevels(fhatx, cont=1:99, approx=approx.cont) ## re-sample on regular grid ep <- seq(fhatx$eval.points[1], tail(fhatx$eval.points,n=1), length=length(fhatx$eval.points)) fhatx$estimate <- predict(fhatx, x=ep) fhatx$eval.points <- ep return(fhatx) } kde.unit.interval.1d <- function(x, h, w, binned=FALSE) { d <- 1 y <- qnorm(x) if (missing(h)) h <- hpi(y) xseq <- tail(head(seq(0,1, length=default.gridsize(d)+2),n=-1), n=-1) fhaty <- kde(x=y, h=h, w=w, binned=binned) fhaty$estimate <- predict(fhaty, x=qnorm(xseq)) fhatx <- fhaty fhatx$eval.points <- xseq fhatx$estimate <- fhaty$estimate/dnorm(qnorm(xseq)) ## apply loess smoothing for unsmooth binned estimates if (binned) { fhatx.loess <- loess(fhatx$estimate ~ fhatx$eval.points, span=0.1) fhatx.smoothed <- fhatx fhatx.smoothed$eval.points <- xseq fhatx.smoothed$estimate <- predict(fhatx.loess, x=xseq) fhatx <- fhatx.smoothed } fhatx$x <- x return(fhatx) } ############################################################################### ## Bivariate kernel density estimate using normal kernels, evaluated over grid ## ## Parameters ## x - data points ## H - bandwidth matrix ## gridsize - number of interval points in grid ## supp - effective support of kernel ## ## Returns ## list with fields ## x - data points ## eval.points - points that KDE is evaluated at ## estimate - KDE evaluated at eval.points ## H - bandwidth matrix ############################################################################### kde.grid.2d <- function(x, H, gridsize, supp, gridx=NULL, grid.pts=NULL, xmin, xmax, gridtype, w, verbose=FALSE) { ## initialise grid n <- nrow(x) if (is.null(gridx)) gridx <- make.grid.ks(x, matrix.sqrt(H), tol=supp, gridsize=gridsize, xmin=xmin, xmax=xmax, gridtype=gridtype) suppx <- make.supp(x, matrix.sqrt(H), tol=supp) if (is.null(grid.pts)) grid.pts <- find.gridpts(gridx, suppx) fhat.grid <- matrix(0, nrow=length(gridx[[1]]), ncol=length(gridx[[2]])) if (verbose) pb <- txtProgressBar() for (i in 1:n) { ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.x.len <- length(eval.x) eval.pts <- expand.grid(eval.x, eval.y) fhat <- dmvnorm(eval.pts, x[i,], H) ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (j in 1:length(eval.y)) fhat.grid[eval.x.ind, eval.y.ind[j]] <- fhat.grid[eval.x.ind, eval.y.ind[j]] + w[i]*fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] if (verbose) setTxtProgressBar(pb, i/n) } if (verbose) close(pb) fhat.grid <- fhat.grid/n gridx1 <- list(gridx[[1]], gridx[[2]]) fhat.list <- list(x=x, eval.points=gridx1, estimate=fhat.grid, H=H, gridtype=gridx$gridtype, gridded=TRUE) return(fhat.list) } ###################################################################### ## Bivariate KDE for data in positive quadrant ###################################################################### kde.positive.2d <- function(x, H, adj.positive, binned=FALSE, xmin, xmax, compute.cont=TRUE, approx.cont=TRUE, ...) { if (missing(adj.positive)) adj.positive <- abs(apply(x, 2, min)) y <- log(cbind(x[,1] + adj.positive[1],x[,2] + adj.positive[2])) if (missing(H)) H <- Hpi(y, binned=binned) d <- ncol(x) tol <- 3.7 tol.H <- tol * diag(H) if (missing(xmin)) xmin <- apply(x, 2, min) - tol.H if (missing(xmax)) xmax <- apply(x, 2, max) + tol.H xmin[xmin<0] <- 0 ymin1 <- log(pmax(xmin + adj.positive, apply(x, 2, min))) ymax1 <- log(xmax + adj.positive) fhaty <- kde(x=y, H=H, xmin=ymin1, xmax=ymax1, gridtype=c("exp", "exp"), binned=binned, compute.cont=compute.cont, approx.cont=approx.cont, ...) fhaty$estimate[is.nan(fhaty$estimate)] <- 0 fhatx <- fhaty fhatx$x <- x fhatx$eval.points[[1]] <- exp(fhaty$eval.points[[1]]) - adj.positive[1] fhatx$eval.points[[2]] <- exp(fhaty$eval.points[[2]]) - adj.positive[2] jacobian <- abs(exp(fhaty$eval.points[[1]]) %o% exp(fhaty$eval.points[[2]])) jacobian[jacobian<=0] <- min(fhatx$estimate[fhatx$estimate>0]) fhatx$estimate <- fhaty$estimate/jacobian if (compute.cont) fhatx$cont <- contourLevels(fhatx, cont=1:99, approx=approx.cont) return(fhatx) } ############################################################################### ## Trivariate kernel density estimate using normal kernels, evaluated over grid ## ## Parameters ## x - data points ## H - bandwidth matrix ## gridsize - number of interval points in grid ## supp - effective support of kernel ## ## Returns ## list with fields ## x - data points ## eval.points - points that KDE is evaluated at ## estimate - KDE evaluated at eval.points ## H - bandwidth matrix ############################################################################### kde.grid.3d <- function(x, H, gridsize, supp, gridx=NULL, grid.pts=NULL, xmin, xmax, gridtype, w, verbose=FALSE) { ## initialise grid n <- nrow(x) if (is.null(gridx)) gridx <- make.grid.ks(x, matrix.sqrt(H), tol=supp, gridsize=gridsize, xmin=xmin, xmax=xmax, gridtype=gridtype) suppx <- make.supp(x, matrix.sqrt(H), tol=supp) if (is.null(grid.pts)) grid.pts <- find.gridpts(gridx, suppx) fhat.grid <- array(0, dim=c(length(gridx[[1]]), length(gridx[[2]]), length(gridx[[3]]))) if (verbose) pb <- txtProgressBar() for (i in 1:n) { ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.z <- gridx[[3]][grid.pts$xmin[i,3]:grid.pts$xmax[i,3]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.z.ind <- c(grid.pts$xmin[i,3]:grid.pts$xmax[i,3]) eval.x.len <- length(eval.x) eval.pts <- expand.grid(eval.x, eval.y) ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (k in 1:length(eval.z)) { fhat <- w[i]*dmvnorm(cbind(eval.pts, eval.z[k]), x[i,], H) for (j in 1:length(eval.y)) fhat.grid[eval.x.ind,eval.y.ind[j], eval.z.ind[k]] <- fhat.grid[eval.x.ind, eval.y.ind[j], eval.z.ind[k]] + fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] } if (verbose) setTxtProgressBar(pb, i/n) } if (verbose) close(pb) fhat.grid <- fhat.grid/n gridx1 <- list(gridx[[1]], gridx[[2]], gridx[[3]]) fhat.list <- list(x=x, eval.points=gridx1, estimate=fhat.grid, H=H, gridtype=gridx$gridtype, gridded=TRUE) return(fhat.list) } kde.grid.nd <- function(x, H, gridsize, supp, gridx=NULL, grid.pts=NULL, xmin, xmax, gridtype, w, verbose=FALSE) { ## initialise grid n <- nrow(x) if (is.null(gridx)) gridx <- make.grid.ks(x, matrix.sqrt(H), tol=supp, gridsize=gridsize, xmin=xmin, xmax=xmax, gridtype=gridtype) gridx1 <- gridx gridx1$stepsize <- NULL gridx1$gridtype <- NULL eval.points <- do.call(expand.grid, gridx1) est <- kde.points(x=x, H=H, eval.points=eval.points, w=w, verbose=verbose)$estimate fhat.grid <- array(est, dim=gridsize) fhat.list <- list(x=x, eval.points=gridx1, estimate=fhat.grid, H=H, gridtype=gridx$gridtype, gridded=TRUE) return(fhat.list) } ############################################################################### ## Multivariate kernel density estimate using normal kernels, ## evaluated at each sample point ## ## Parameters ## x - data points ## H - bandwidth matrix ## eval.points - points where to evaluate density estimate ## ## Returns ## list with fields ## x - data points ## eval.points - points that KDE is evaluated at ## estimate - KDE evaluated at eval.points ## H - bandwidth matrix ############################################################################### kde.points <- function(x, H, eval.points, w, verbose) { n <- nrow(x) d <- ncol(x) ne <- nrow(eval.points) Hs <- replicate(n, H, simplify=FALSE) Hs <- do.call(rbind, Hs) fhat <- dmvnorm.mixt(x=eval.points, mus=x, Sigmas=Hs, props=w/n, verbose=verbose) return(list(x=x, eval.points=eval.points, estimate=fhat, H=H, gridded=FALSE)) } kde.points.1d <- function(x, h, eval.points, positive=FALSE, adj.positive, w) { n <- length(x) if (positive) { if (missing(adj.positive)) adj.positive <- abs(min(x)) y <- log(x + adj.positive) ## transform positive data x to real line eval.pointsy <- log(eval.points + adj.positive) } else { y <- x eval.pointsy <- eval.points } est <- dnorm.mixt(x=eval.pointsy, mus=y, sigmas=rep(h,n), props=w/n) if (positive) est <- est/(eval.points + adj.positive) fhat <- list(x=x, eval.points=eval.points, estimate=est, h=h, H=h^2, gridded=FALSE) return(fhat) } ############################################################################# ## S3 methods for KDE objects ############################################################################# ## predict method for KDE objects predict.kde <- function(object, ..., x, zero.flag=TRUE) { fhat <- grid.interp(x=x, gridx=object$eval.points, f=object$estimate) if (!zero.flag) warning("zero.flag=FALSE has been deprecated and no longer has any effect") return(fhat) } ## plot method plot.kde <- function(x, ...) { fhat <- x if (is.vector(fhat$x)) plotkde.1d(fhat, ...) else { d <- ncol(fhat$x) if (d==2) { opr <- options()$preferRaster; if (!is.null(opr)) if (!opr) options("preferRaster"=TRUE) plotret <- plotkde.2d(fhat, ...) if (!is.null(opr)) options("preferRaster"=opr) invisible(plotret) } else if (d==3) { plotkde.3d(fhat, ...) invisible() } else stop ("Plot function only available for 1, 2 or 3-d data") } } plotkde.1d <- function(fhat, xlab, ylab="Density function", add=FALSE, drawpoints=FALSE, col=1, col.pt=4, col.cont=1, cont.lwd=1, jitter=FALSE, cont, abs.cont, approx.cont=TRUE, alpha=1, ...) { if (missing(xlab)) xlab <- fhat$names col <- transparency.col(col, alpha=alpha) if (add) lines(fhat$eval.points, fhat$estimate, xlab=xlab, ylab=ylab, col=col, ...) else plot(fhat$eval.points, fhat$estimate, type="l", xlab=xlab, ylab=ylab, col=col, ...) ## compute contours if (!missing(cont) | !missing(abs.cont)) { if (missing(abs.cont)) { if (!is.null(fhat$cont)) { cont.ind <- rep(FALSE, length(cont)) for (j in 1:length(cont)) { ci <- which(cont[j]==(100-as.numeric(unlist(strsplit(names(fhat$cont),"%"))))) if (length(ci)==0) cont.ind[j] <- NA else cont.ind[j] = ci } if (any(is.na(cont.ind))) hts <- contourLevels(fhat, prob=(100-cont)/100, approx=approx.cont) else hts <- fhat$cont[cont.ind] } else hts <- contourLevels(fhat, prob=(100-cont)/100, approx=approx.cont) } else hts <- abs.cont if (is.null(fhat$deriv.order)) { hts <- sort(hts, decreasing=TRUE) cont.ind <- 1-as.numeric(fhat$estimate>=hts[1]) cont.ind[cont.ind==1] <- NA lines(fhat$eval.points, cont.ind, col=col.cont, lwd=cont.lwd) } else { for (i in 1:length(hts)) { cont.ind <- 1-as.numeric(abs(fhat$estimate)>=abs(hts[i])) cont.ind[cont.ind==1] <- NA lines(fhat$eval.points, cont.ind, col=col.cont, lwd=cont.lwd) } } } if (drawpoints) if (jitter) rug(jitter(fhat$x), col=col.pt) else rug(fhat$x, col=col.pt) } plotkde.2d <- function(fhat, display="slice", cont=c(25,50,75), abs.cont, approx.cont=TRUE, xlab, ylab, zlab="Density function", cex=1, pch=1, labcex=1, add=FALSE, drawpoints=FALSE, drawlabels=TRUE, theta=-30, phi=40, d=4, col.pt=4, col, col.fun, alpha=1, lwd=1, border=1, thin=3, kdde.flag=FALSE, ticktype="detailed", ...) { disp <- match.arg(display, c("slice", "persp", "image", "filled.contour", "filled.contour2")) if (disp=="filled.contour2") disp <- "filled.contour" if (!is.list(fhat$eval.points)) stop("Need a grid of density estimates") if (missing(xlab)) xlab <- fhat$names[1] if (missing(ylab)) ylab <- fhat$names[2] if (missing(col.fun)) { if (any(fhat$type=="kcurv")) col.fun <- function(n) { hcl.colors(n, palette="Oranges",rev=TRUE, alpha=alpha) } else col.fun <- function(n) { hcl.colors(n, palette="heat",rev=TRUE, alpha=alpha) } } ## perspective/wireframe plot if (disp=="persp") { hts <- seq(0, 1.1*max(fhat$estimate,na.rm=TRUE), length=100) if (missing(col)) col <- col.fun(length(hts)+1) if (length(col)0 | !is.null(fhat$deriv.order)) { j <- j+1 if (j==1) contour(fhat$eval.points[[1]], fhat$eval.points[[2]], fhat$estimate*scale, level=hts[i]*scale, label=signif(hts[ni]*scale2), add=add, drawlabels=drawlabels, col=col[i], lwd=lwd, labcex=labcex, xlab=xlab, ylab=ylab, ...) else contour(fhat$eval.points[[1]], fhat$eval.points[[2]], fhat$estimate*scale, level=hts[i]*scale, label=signif(hts[ni]*scale2), add=TRUE, drawlabels=drawlabels, col=col[i], lwd=lwd, labcex=labcex, ...) } } ## add points if (drawpoints) points(fhat$x[,1], fhat$x[,2], col=col.pt, cex=cex, pch=pch) } ## image plot else if (disp=="image") { if (missing(col)) col <- col.fun(100) col <- transparency.col(col, alpha=alpha) image(fhat$eval.points[[1]], fhat$eval.points[[2]], fhat$estimate, xlab=xlab, ylab=ylab, add=add, col=col, ...) ## add points if (drawpoints) points(fhat$x[,1], fhat$x[,2], col=col.pt, cex=cex, pch=pch) box() } else if (disp=="filled.contour") { ## compute contours if (missing(abs.cont)) { if (!is.null(fhat$cont)) { cont.ind <- rep(FALSE, length(cont)) for (j in 1:length(cont)) { ci <- which(cont[j]==(100-as.numeric(unlist(strsplit(names(fhat$cont),"%"))))) if (length(ci)==0) cont.ind[j] <- NA else cont.ind[j] = ci } if (any(is.na(cont.ind))) hts <- contourLevels(fhat, prob=(100-cont)/100, approx=approx.cont) else hts <- fhat$cont[cont.ind] } else hts <- contourLevels(fhat, prob=(100-cont)/100, approx=approx.cont) } else hts <- abs.cont hts <- sort(hts) if (missing(col)) col <- c("transparent", col.fun(length(hts))) col <- transparency.col(col, alpha=alpha) clev <- c(min(c(fhat$estimate, hts)-0.01*max(abs(fhat$estimate))), hts, max(c(fhat$estimate, hts)) + 0.01*max(abs(fhat$estimate))) clev <- unique(clev) if (!add) plot(fhat$eval.points[[1]], fhat$eval.points[[2]], type="n", xlab=xlab, ylab=ylab, ...) .filled.contour(fhat$eval.points[[1]], fhat$eval.points[[2]], z=fhat$estimate, levels=clev, col=col) if (!missing(lwd)) { for (i in 1:length(hts)) { if (missing(abs.cont)) { ni <- length(hts)-i+1; scale <- (100-cont[i])/hts[i]; scale2 <- cont[ni]/hts[ni] } else { ni <- i; scale <- 1; scale2 <-1 } if (lwd >=1) contour(fhat$eval.points[[1]], fhat$eval.points[[2]], fhat$estimate*scale, level=hts[i]*scale, label=signif(hts[ni]*scale2,3), add=TRUE, drawlabels=drawlabels, col=1, lwd=lwd, labcex=labcex, ...) } } ## add points if (drawpoints) points(fhat$x[,1], fhat$x[,2], col=col.pt, cex=cex, pch=pch) box() } if (disp=="persp") invisible(plotret) else invisible() } plotkde.3d <- function(fhat, display="plot3D", cont=c(25,50,75), abs.cont, approx.cont=TRUE, colors, col, col.fun, alphavec, size=3, cex=1, pch=1, theta=-30, phi=40, d=4, ticktype="detailed", bty="f", col.pt=4, add=FALSE, xlab, ylab, zlab, drawpoints=FALSE, alpha, box=TRUE, axes=TRUE, ...) { ## compute contours if (missing(abs.cont)) { if (!is.null(fhat$cont)) { cont.ind <- rep(FALSE, length(fhat$cont)) for (j in 1:length(cont)) cont.ind[which(cont[j] == 100-as.numeric(unlist(strsplit(names(fhat$cont),"%"))))] <- TRUE if (all(!cont.ind)) hts <- contourLevels(fhat, prob=(100-cont)/100, approx=approx.cont) else hts <- fhat$cont[cont.ind] } else hts <- contourLevels(fhat, prob=(100-cont)/100, approx=approx.cont) } else hts <- abs.cont nc <- length(hts) if (missing(col)) { if (missing(col.fun)) { if (any(fhat$type=="kcurv")) col.fun <- function(n) { hcl.colors(n, palette="Oranges",rev=TRUE) } else col.fun <- function(n) { hcl.colors(n, palette="heat",rev=TRUE) } } col <- col.fun(n=length(hts)) } colors <- col if (missing(xlab)) xlab <- fhat$names[1] if (missing(ylab)) ylab <- fhat$names[2] if (missing(zlab)) zlab <- fhat$names[3] if (missing(alphavec)) { if (is.null(fhat$deriv.order)) alphavec <- seq(0.1,0.5,length=nc) else alphavec <- c(rev(seq(0.1,0.4,length=round(nc/2))), seq(0.1,0.4,length=round(nc/2))) } if (missing(alpha)) alpha <- 0.1 #else if (!missing(alpha)) { alphavec <- rep(alpha,nc) } disp <- match.arg(display, c("plot3D", "rgl")) if (disp %in% "plot3D") { if (!requireNamespace("plot3D", quietly=TRUE)) stop("Install the plot3D package as it is required.", call.=FALSE) for (i in 1:nc) if (hts[nc-i+1] < max(fhat$estimate)) plot3D::isosurf3D(x=fhat$eval.points[[1]], y=fhat$eval.points[[2]], z=fhat$eval.points[[3]], colvar=fhat$estimate, level=hts[nc-i+1], add=add | (i>1), col=colors[i], alpha=alphavec[i], phi=phi, theta=theta, xlab=xlab, ylab=ylab, zlab=zlab, d=d, ticktype=ticktype, bty=bty, ...) if (drawpoints) plot3D::points3D(x=fhat$x[,1], y=fhat$x[,2], z=fhat$x[,3], cex=cex, col=col.pt, add=TRUE, pch=pch, d=d) } else if (disp %in% "rgl") { ## suggestions from Viktor Petukhov 08/03/2018 if (!requireNamespace("rgl", quietly=TRUE)) stop("Install the rgl package as it is required.", call.=FALSE) if (!requireNamespace("misc3d", quietly=TRUE)) stop("Install the misc3d package as it is required.", call.=FALSE) fhat.eval.mean <- sapply(fhat$eval.points, mean) if (drawpoints) rgl::plot3d(fhat$x[,1],fhat$x[,2],fhat$x[,3], size=size, col=col.pt, alpha=alpha, xlab=xlab, ylab=ylab, zlab=zlab, add=add, box=FALSE, axes=FALSE, ...) else rgl::plot3d(fhat$x[,1],fhat$x[,2],fhat$x[,3], size=0, col="transparent", alpha=0, xlab=xlab, ylab=ylab, zlab=zlab, add=add, box=FALSE, axes=FALSE, ...) for (i in 1:nc) if (hts[nc-i+1] < max(fhat$estimate)) misc3d::contour3d(fhat$estimate, level=hts[nc-i+1], x=fhat$eval.points[[1]], y=fhat$eval.points[[2]], z=fhat$eval.points[[3]], add=TRUE, color=colors[i], alpha=alphavec[i], box=FALSE, axes=FALSE, ...) if (axes) rgl::axes3d(c("x","y","z")) if (box) rgl::box3d() } } ## contourLevels method ## create S3 generic contourLevels <- function(x, ...) UseMethod("contourLevels") contourLevels.kde <- function(x, prob, cont, nlevels=5, approx=TRUE, ...) { fhat <- x if (is.vector(fhat$x)) { d <- 1; n <- length(fhat$x) } else { d <- ncol(fhat$x); n <-nrow(fhat$x) if (!is.matrix(fhat$x)) fhat$x <- as.matrix(fhat$x) } if (is.null(x$w)) w <- rep(1, n) else w <- x$w if (is.null(fhat$gridded)) { if (d==1) fhat$gridded <- fhat$binned else fhat$gridded <- is.list(fhat$eval.points) } if (missing(prob) & missing(cont)) hts <- pretty(fhat$estimate, n=nlevels) else { if (approx & fhat$gridded) dobs <- predict(fhat, x=fhat$x) else dobs <- kde(x=fhat$x, H=fhat$H, eval.points=fhat$x, w=w)$estimate if (!missing(prob) & missing(cont)) hts <- quantile(dobs, prob=prob) if (missing(prob) & !missing(cont)) hts <- quantile(dobs, prob=(100-cont)/100) } return(hts) } ############################################################################### ## Riemann sums to compute approximate Lebesgue measure of contour set ############################################################################### contourSizes <- function(x, abs.cont, cont=c(25,50,75), approx=TRUE) { if (missing(abs.cont)) abs.cont <- contourLevels(x, cont=cont, approx=approx) num.int <- rep(0, length(abs.cont)) if (!is.null(names(abs.cont))) names(num.int) <- names(abs.cont) if (!is.list(x$eval.points)) delta.int <- head(diff(x$eval.points), n=1) else delta.int <- prod(sapply(lapply(x$eval.points, diff), head, n=1)) for (j in 1:length(abs.cont)) num.int[j] <- sum(x$estimate>abs.cont[j]) return(num.int*delta.int) } ############################################################################### ## Riemann sums to compute approximate probability of contour set ############################################################################### contourProbs <- function(x, abs.cont, cont=c(25,50,75), approx=TRUE) { if (missing(abs.cont)) abs.cont <- contourLevels(x, cont=cont, approx=approx) num.int <- rep(0, length(abs.cont)) if (!is.null(names(abs.cont))) names(num.int) <- names(abs.cont) if (!is.list(x$eval.points)) { delta.int <- head(diff(x$eval.points), n=1) eval.points.midpoint <- (head(x$eval.points,n=-1)+tail(x$eval.points,n=-1))/2 } else { delta.int <- prod(sapply(lapply(x$eval.points, diff), head, n=1)) eval.points.midpoint <- expand.grid(lapply(x$eval.points, function(y) { (head(y,n=-1)+tail(y,n=-1))/2 })) } x.evmp <- predict(x, x=eval.points.midpoint) for (i in 1:length(num.int)) num.int[i] <- sum(x.evmp*(x.evmp>=abs.cont[i])*delta.int) return(num.int) } ############################################################################### ## Generate grid over a set of points ## ## Parameters ## x - data points ## H - bandwidth matrix ## tol - tolerance = extra coverage exceeding the range of x ## gridsize - number of points for each direction ## ## Returns ## gridx - list of intervals, one for each co-ord direction so that ## gridx[[1]] x gridx[[2]] x ... x gridx[[d]] is the grid ## stepsize - vector of step sizes ############################################################################### make.grid.ks <- function(x, H, tol, gridsize, xmin, xmax, gridtype) { d <- ncol(x) tol.H <- tol * diag(H) if (missing(xmin)) xmin <- apply(x, 2, min) - tol.H if (missing(xmax)) xmax <- apply(x, 2, max) + tol.H stepsize <- rep(0, d) gridx <- numeric(0) if (length(gridsize)==1) gridsize <- rep(gridsize, d) if (missing(gridtype)) gridtype <- rep("linear", d) gridtype.vec <- rep("", d) for (i in 1:d) { gridtype1 <- match.arg(gridtype[i], c("linear", "sqrt", "quantile", "exp")) if (gridtype1=="linear") { gridx <- c(gridx, list(seq(xmin[i], xmax[i], length=gridsize[i]))) stepsize[i] <- abs(gridx[[i]][1] - gridx[[i]][2]) } else if (gridtype1=="sqrt") { gridx.temp <- seq(sign(xmin[i])*sqrt(abs(xmin[i])), sign(xmax[i])*sqrt(abs(xmax[i])), length=gridsize[i]) gridx <- c(gridx, list(sign(gridx.temp) * gridx.temp^2)) stepsize[i] <- NA } else if (gridtype1=="quantile") { gridx.temp <- qnorm(seq(1e-2, 1-1e-2, length=gridsize[i])) gridx <- c(gridx, list(xmin[i] + (xmax[i]-xmin[i])*(gridx.temp-min(gridx.temp))/(max(gridx.temp)-min(gridx.temp)))) stepsize[i] <- NA } else if (gridtype1=="exp") { gridx.temp <- seq(exp(xmin[i]), exp(xmax[i]), length=gridsize[i]) gridx <- c(gridx, list(log(gridx.temp))) stepsize[i] <- NA } gridtype.vec[i] <- gridtype1 } gridx <- c(gridx, list(stepsize = stepsize, gridtype=gridtype.vec)) return(gridx) } ############################################################################### ## Generate kernel (rectangular) support at data point ## ## Parameters ## x - data points ## H - bandwidth matrix ## tol - tolerance = extra coverage exceeding the range of x ## ## Returns ## list of min and max points of support (here we parameterise rectangles ## by their min = lower left co-ord and max = upper right coord) ############################################################################### make.supp <- function(x, H, tol) { n <- nrow(x) d <- ncol(x) tol.H <- tol * diag(H) xmin <- matrix(0, nrow=n, ncol=d) xmax <- matrix(0, nrow=n, ncol=d) for (i in 1:n) { xmin[i,] <- x[i,] - tol.H xmax[i,] <- x[i,] + tol.H } return(list(xmin=xmin, xmax=xmax)) } ############################################################################### ## Find the grid points contained in kernel support rectangles ## ## Parameters ## gridx - grid (list of subdivided intervals) ## rectx - rectangles (list of min and max points) ## ## Returns ## list of min and max points of the grid for each rectangle ############################################################################### find.gridpts <- function(gridx, suppx) { xmax <- suppx$xmax xmin <- suppx$xmin d <- ncol(xmax) n <- nrow(xmax) gridpts.min <- matrix(0, ncol=d, nrow=n) gridpts.max <- gridpts.min for (i in 1:n) for (j in 1:d) { ## find index of last element of gridx smaller than min support tsum <- sum(xmin[i,j] >= gridx[[j]]) if (tsum==0) gridpts.min[i,j] <- 1 else gridpts.min[i,j] <- tsum ## find index of first element gridx greater than max support gridpts.max[i,j] <- sum(xmax[i,j] >= gridx[[j]]) } return(list(xmin=gridpts.min, xmax=gridpts.max)) } ############################################################################## ## Interpolate the values of f defined on gridx at new values x ############################################################################## grid.interp <- function(x, gridx, f) { if (!is.list(gridx)) { ## uniform grid if (isTRUE(all.equal(diff(gridx), rep(diff(gridx)[1], length(gridx)-1)))) fx <- grid.interp.1d(x=as.vector(x), gridx=gridx, f=f) ## non-uniform grid else fx <- varying.grid.interp.1d(x=as.vector(x), gridx=gridx, f=f) } else { if (is.vector(x)) x <- as.matrix(t(x)) d <- ncol(x) n <- nrow(x) if (d<2) stop("x should be a vector") gridx.diff <- lapply(lapply(gridx,diff), getElement, 1) for (i in 1:length(gridx.diff)) gridx.diff[[i]] <- rep(gridx.diff[[i]], sapply(gridx, length)[i]-1) uniform.grid.flag <- isTRUE(all.equal(lapply(gridx,diff), gridx.diff)) ## uniform grid if (d==2 & uniform.grid.flag) fx <- grid.interp.2d(x=x, gridx=gridx, f=f) else if (d==3 & uniform.grid.flag) fx <- grid.interp.3d(x=x, gridx=gridx, f=f) else ## d >=4 or non-uniform grid { gridsize <- sapply(gridx,length) gind <- matrix(0, nrow=n, ncol=d) for (i in 1:n) for (j in 1:d) { tsum <- sum(x[i,j] >= gridx[[j]]) if (tsum==0) gind[i,j] <- 1 else gind[i,j] <- tsum } for (j in 1:d) gind[gind[,j]>=gridsize[j],j] <- gridsize[j]-1 bperm <- list() for (j in 1:d) bperm[[j]] <- elem(1,2) binary.perm <- as.matrix(expand.grid(bperm)) colnames(binary.perm) <- NULL gind.list <- list() fx <- rep(0, length=n) for (i in 1:n) { gind.list[[i]] <- matrix(gind[i,], nrow=2^d, ncol=d, byrow=TRUE) + binary.perm w <- matrix(0, nrow=2^d, ncol=d) gridw <- matrix(0, nrow=2^d, ncol=d) for (j in 1:d) { gind.list[[i]][,j][gind.list[[i]][,j]>=gridsize[j]] <- gridsize[j] gridw[,j] <- gridx[[j]][gind.list[[i]][,j]] } w <- abs(matrix(as.numeric(x[i,]), nrow=2^d, ncol=d, byrow=TRUE) - gridw) w <- apply(w, 1, prod) w <- w/sum(w) fx[i] <- sum(w*f[gind.list[[i]][2^d:1,]]) } } } return(fx) } grid.interp.1d <- function(x, gridx, f) { n <- length(x) gpoints1 <- gridx M1 <- length(gpoints1) a1 <- gpoints1[1] b1 <- gpoints1[M1] out <- .C(C_interp1d, x1=as.double(x), n=as.integer(n), a1=as.double(a1), b1=as.double(b1), M1=as.integer(M1), fun=as.double(as.vector(f)), est=double(n)) return(out$est) } grid.interp.2d <- function(x, gridx, f) { n <- nrow(x) gpoints1 <- gridx[[1]] gpoints2 <- gridx[[2]] M1 <- length(gpoints1) M2 <- length(gpoints2) a1 <- gpoints1[1] a2 <- gpoints2[1] b1 <- gpoints1[M1] b2 <- gpoints2[M2] out <- .C(C_interp2d, x1=as.double(x[,1]), x2=as.double(x[,2]), n=as.integer(n), a1=as.double(a1), a2=as.double(a2), b1=as.double(b1), b2=as.double(b2), M1=as.integer(M1), M2=as.integer(M2), fun=as.double(as.vector(f)), est=double(n)) return(out$est) } grid.interp.3d <- function(x, gridx, f) { n <- nrow(x) gpoints1 <- gridx[[1]] gpoints2 <- gridx[[2]] gpoints3 <- gridx[[3]] M1 <- length(gpoints1) M2 <- length(gpoints2) M3 <- length(gpoints3) a1 <- gpoints1[1] a2 <- gpoints2[1] a3 <- gpoints3[1] b1 <- gpoints1[M1] b2 <- gpoints2[M2] b3 <- gpoints3[M3] out <- .C(C_interp3d, x1=as.double(x[,1]), x2=as.double(x[,2]), x3=as.double(x[,3]), n=as.integer(n), a1=as.double(a1), a2=as.double(a2), a3=as.double(a3), b1=as.double(b1), b2=as.double(b2), b3=as.double(b3), M1=as.integer(M1), M2=as.integer(M2), M3=as.integer(M3), fun=as.double(as.vector(f)), est=double(n)) return(out$est) } ## Linear intepolation based on kernel estimation grid ## alias for predict.kde kde.approx <- function(fhat, x) { return(grid.interp(x=x, gridx=fhat$eval.points, f=fhat$estimate)) } ############################################################################## ## Find the nearest grid points surrounding point x for non-uniform grids ############################################################################## varying.grid.interp <- function(x, gridx, f) { if (!is.list(gridx)) return(varying.grid.interp.1d(x=x, gridx=gridx, f=f)) else { if (is.vector(x)) x <- as.matrix(t(x)) d <- ncol(x) n <- nrow(x) gridsize <- sapply(gridx,length) gind <- matrix(0, nrow=n, ncol=d) for (i in 1:n) for (j in 1:d) { tsum <- sum(x[i,j] >= gridx[[j]]) if (tsum==0) gind[i,j] <- 1 else gind[i,j] <- tsum } } bperm <- list() for (j in 1:d) bperm[[j]] <- elem(1,2) binary.perm <- as.matrix(expand.grid(bperm)) colnames(binary.perm) <- NULL gind.list <- list() fx <- rep(0, length=n) for (i in 1:n) { gind.list[[i]] <- matrix(gind[i,], nrow=2^d, ncol=d, byrow=TRUE) + binary.perm w <- matrix(0, nrow=2^d, ncol=d) gridw <- matrix(0, nrow=2^d, ncol=d) for (j in 1:d) { gind.list[[i]][,j][gind.list[[i]][,j]>=gridsize[j]] <- gridsize[j] gridw[,j] <- gridx[[j]][gind.list[[i]][,j]] } w <- 1/apply((matrix(as.numeric(x[i,]), nrow=2^d, ncol=d, byrow=TRUE) - gridw)^2, 1, sum) w[w>1e5] <- 1e5 w <- w/sum(w) fx[i] <- sum(w*f[gind.list[[i]]]) } return(fx) } varying.grid.interp.1d <- function(x, gridx, f) { n <- length(x) gind <- rep(0, length=n) for (i in 1:length(x)) { tsum <- sum(x[i] >= gridx) if (tsum==0) gind[i] <- 1 else gind[i] <- tsum } gind2 <- gind+1 gind2[gind2>length(gridx)] <- length(gridx) gind2[x<=gridx[1]] <- gind[x<=gridx[1]] gind <- cbind(gind, gind2) colnames(gind) <- NULL fx <- rep(0, n) for (i in 1:n) { w <- 1/(x[i] - gridx[gind[i,]])^2 w[w>1e5] <- 1e5 w <- w/sum(w) fx[i] <- sum(w*f[gind[i,]]) } return(fx) } ks/R/prelim.R0000644000176200001440000006230614547755373012536 0ustar liggesusers########################################################################## ## Default values for parameter choices for ks objects ########################################################################## ks.defaults <- function(x, w, binned, bgridsize, gridsize) { ## dimensions of x if (is.vector(x)) { d <- 1; n <- length(x) } else { d <- ncol(x); n <- nrow(x) } ## default uniform weights if (missing(w)) w <- rep(1,n) else if (!missing(w)) { if (!(identical(all.equal(sum(w), n), TRUE))) { warning("Weights don't sum to sample size - they have been scaled accordingly\n") w <- w*n/sum(w) } } ## default binning flag if (missing(binned)) binned <- default.bflag(d=d, n=n) ## default grid sizes if (missing(bgridsize)) { if (missing(gridsize)) bgridsize <- default.bgridsize(d) else bgridsize <- gridsize } if (missing(gridsize)) gridsize <- default.gridsize(d) if (length(gridsize)==1) gridsize <- rep(gridsize, d) if (length(bgridsize)==1) bgridsize <- rep(bgridsize, d) ksd <- list(d=d, n=n, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) return(ksd) } ########################################################################## ## Parse variable name ########################################################################## parse.name <- function(x) { if (is.vector(x)) { d <- 1 x.names <- deparse(substitute(x)) } else { d <- ncol(x) x.names <- colnames(x) if (is.null(x.names)) { x.names <- strsplit(deparse(substitute(x)), "\\[")[[1]][1] x.names <- paste(x.names, "[, ", 1:d,"]",sep="") } } return(x.names) } ########################################################################## ## Basic vectors and matrices and their operations ########################################################################## ## vec operator vec <- function(x, byrow=FALSE) { if (is.vector(x)) return (x) if (byrow) x <- t(x) d <- ncol(x) vecx <- vector() for (j in 1:d) vecx <- c(vecx, x[,j]) return(vecx) } ## vech operator vech <- function(x) { if (is.vector(x)) { if (length(x)==1) return (x) else stop("vech undefined for vectors") } else if (is.matrix(x)) { d <- ncol(x) if (d!=nrow(x)) stop("vech only defined for square matrices") vechx <- vector() for (j in 1:d) vechx <- c(vechx, x[j:d,j]) return(vechx) } } ## inverse vec operator invvec <- function(x, ncol, nrow, byrow=FALSE) { if (length(x)==1) return(x) d <- sqrt(length(x)) if (missing(ncol) | missing(nrow)) { ncol <- d; nrow <- d if (round(d) != d) stop("Need to specify nrow and ncol for non-square matrices") } invvecx <- matrix(0, nrow = nrow, ncol = ncol) if (byrow) for (j in 1:nrow) invvecx[j,] <- x[c(1:ncol) + (j-1)*ncol] else for (j in 1:ncol) invvecx[,j] <- x[c(1:nrow) + (j-1)*nrow] return(invvecx) } ## inverse vech operator invvech <- function(x) { if (length(x)==1) return(x) d <- (-1 + sqrt(8*length(x) + 1))/2 if (round(d) != d) stop("Number of elements in x will not form a square matrix") invvechx <- matrix(0, nrow=d, ncol=d) for (j in 1:d) invvechx[j:d,j] <- x[1:(d-j+1)+ (j-1)*(d - 1/2*(j-2))] invvechx <- invvechx + t(invvechx) - diag(diag(invvechx)) return(invvechx) } ## trace of matrix tr <- function(A) { count <- 0 if (is.vector(A)) return (A[1]) if (nrow(A)!=ncol(A)) stop("Not square matrix") else for (i in 1:nrow(A)) count <- count + A[i,i] return(count) } ## elementary vector elem <- function(i, d) { elem.vec <- rep(0, d) elem.vec[i] <- 1 return(elem.vec) } ## commutation matrix (taken from MCMCglmmm library) comm <- function(m,n) { K <- matrix(0,m*n, m*n) H <- matrix(0,m,n) for (i in 1:m) { for(j in 1:n) { H[i,j] <- 1 K <- K+kronecker(H,t(H)) H[i,j] <- 0 } } return(K) } ########################################################################## ## Duplication matrix ## Taken from Felipe Osorio http://www.ime.usp.br/~osorio/files/dupl.q ########################################################################## dupl <- function(order, ret.q = FALSE) { ## call cl <- match.call() time1 <- proc.time() if (!is.integer(order)) order <- as.integer(order) n <- order - 1 ## initial duplication matrix d1 <- matrix(0, nrow = 1, ncol = 1) d1[1,1] <- 1 if (!is.integer(d1)) storage.mode(d1) <- "integer" ## recursive formula if (n > 0) { for (k in 1:n) { drow <- 2*k + 1 + nrow(d1) dcol <- k + 1 + ncol(d1) d2 <- matrix(0, nrow = drow, ncol=dcol) storage.mode(d2) <- "integer" d2[1,1] <- 1 d2[2:(k+1),2:(k+1)] <- diag(k) d2[(k+2):(2*k+1),2:(k+1)] <- diag(k) d2[(2*k+2):drow,(k+2):dcol] <- d1 ## permutation matrix q <- permute.mat(k) ## new duplication matrix d2 <- q %*% d2 storage.mode(d2) <- "integer" d1 <- d2 } } else { d2 <- q <- d1 } ## results obj <- list(call=cl, order=order, d=d2) if (ret.q) obj$q <- q obj$time <- proc.time() - time1 obj } ########################################################################## ## Pre-scaling ## Parameters ## x - data points ## ## Returns ## Pre-scaled x values ########################################################################## pre.scale <- function(x, mean.centred=FALSE) { S <- diag(diag(var(x))) Sinv12 <- matrix.sqrt(chol2inv(chol(S))) if (mean.centred) x.scaled <- sweep(x, 2, apply(x, 2, mean)) else x.scaled <- x x.scaled <- x.scaled %*% Sinv12 return (x.scaled) } ########################################################################## ## Pre-sphering ## Parameters ## x - data points ## ## Returns ## Pre-sphered x values ########################################################################## pre.sphere <- function(x, mean.centred=FALSE) { S <- var(x) Sinv12 <- matrix.sqrt(chol2inv(chol(S))) if (mean.centred) x.sphered <- sweep(x, 2, apply(x, 2, mean)) else x.sphered <- x x.sphered <- x.sphered %*% Sinv12 return (x.sphered) } ########################################################################## ## Boolean functions ########################################################################## is.even <- function(x) { y <- x[x>0] %%2 return(identical(y, rep(0, length(y)))) } is.diagonal <- function(x) { return(identical(diag(diag(x)),x)) } ########################################################################## ## Finds row index matrix ## Parameters ## x - data points ## ## Returns ## i - if r==mat[i,] ## NA - otherwise ########################################################################## which.mat <- function(r, mat) { ind <- numeric() for (i in 1:nrow(mat)) if (identical(r, mat[i,])) ind <- c(ind,i) return(ind) } ################################################################### ## Permutation functions ################################################################### #################################################################### ## Exactly the same function as combinat:::permn #################################################################### permn.ks <- function (x, fun = NULL, ...) { if (is.numeric(x) && length(x) == 1 && x > 0 && trunc(x) == x) x <- seq(x) n <- length(x) nofun <- is.null(fun) out <- vector("list", gamma(n + 1)) p <- ip <- seqn <- 1:n d <- rep(-1, n) d[1] <- 0 m <- n + 1 p <- c(m, p, m) i <- 1 use <- -c(1, n + 2) while (m != 1) { out[[i]] <- if (nofun) x[p[use]] else fun(x[p[use]], ...) i <- i + 1 m <- n chk <- (p[ip + d + 1] > seqn) m <- max(seqn[!chk]) if (m < n) d[(m + 1):n] <- -d[(m + 1):n] index1 <- ip[m] + 1 index2 <- p[index1] <- p[index1 + d[m]] p[index1 + d[m]] <- m tmp <- ip[index2] ip[index2] <- ip[m] ip[m] <- tmp } out } ########################################################################## ## Permutations with repetitions of the first d naturals (1:d) taking ## k elements at a time. There are d^k of them, each having length k ## => We arrange them into a matrix of order d^k times k ## Each row represents one permutation ## Second version: filling in the matrix comlumn-wise (slightly faster) ########################################################################## perm.rep <- function(d,r) { if (r==0) { PM <- 1 } if (r>0) { PM <- matrix(nrow=d^r,ncol=r) for (pow in 0:(r-1)) { t2 <- d^pow p1 <- 1 while (p1<=d^r) { for (al in 1:d) { for (p2 in 1:t2) { PM[p1,r-pow] <- al p1 <- p1+1 } } } } } return(PM) } ########################################################################## ## Permute a list of values ## ## Same function as EXPAND.GRID (base package), modified to take ## list as an argument and returns a matrix ########################################################################## permute <- function (args) { nargs <- length(args) if (!nargs) return(as.data.frame(list())) if (nargs == 1 && is.list(a1 <- args[[1]])) nargs <- length(args <- a1) if (nargs <= 1) return(as.data.frame(if (nargs == 0 || is.null(args[[1]])) list() else args, optional = TRUE)) cargs <- args rep.fac <- 1 orep <- prod(sapply(args, length)) for (i in 1:nargs) { x <- args[[i]] nx <- length(x) orep <- orep/nx cargs[[i]] <- rep(rep(x, rep(rep.fac, nx)), orep) rep.fac <- rep.fac * nx } do.call("cbind", cargs) } permute.mat <- function(order) { m <- as.integer(order) m <- m + 1 eye <- diag(m) u1 <- eye[1:m,1] u2 <- eye[1:m,2:m] q1 <- kronecker(eye, u1) q2 <- kronecker(eye, u2) q <- matrix(c(q1, q2), nrow = nrow(q2), ncol = ncol(q1) + ncol(q2)) if (!is.integer(q)) storage.mode(q) <- "integer" return(q) } ########################################################################## ## pinv.all generates all the permutations PR_{d,r} as described in ## Appendix B of Chacon and Duong (2014) ########################################################################## pinv.all <- function(d,r) { i <- 1:d^r n <- i-1 dpow <- d^(0:r) n.mat <- matrix(rep(n,r+1),byrow=FALSE,nrow=d^r,ncol=r+1) dpow.mat <- matrix(rep(dpow,d^r),byrow=TRUE,nrow=d^r,ncol=r+1) ndf.mat <- floor(n.mat/dpow.mat) ans <- ndf.mat[,r:1]-d*ndf.mat[,(r+1):2] return(ans+1) } ########################################################################## ## Block indices for double sums ########################################################################## block.indices <- function(nx, ny, d, r=0, diff=FALSE, block.limit=1e6, npergroup) { if (missing(npergroup)) { if (diff) npergroup <- max(c(block.limit %/% (nx*d^r), 1)) else npergroup <- max(c(block.limit %/% nx,1)) } nseq <- seq(1, ny, by=npergroup) if (tail(nseq,n=1) <= ny) nseq <- c(nseq, ny+1) if (length(nseq)==1) nseq <- c(1, ny+1) return(nseq) } block.indices2 <- function(nx, ny, block.limit=1e6, npergroup) { if (missing(npergroup)) npergroup <- max(c(block.limit %/% nx,1)) nseq <- seq(1, ny, by=npergroup) if (tail(nseq,n=1) <= ny) nseq <- c(nseq, ny+1) if (length(nseq)==1) nseq <- c(1, ny+1) return(nseq) } #################################################################### ## Differences for double sums calculations #################################################################### differences <- function(x, y, upper=FALSE, ff=FALSE, Kpow=0) { if (missing(y)) y <- x if (is.vector(x)) x <- t(as.matrix(x)) if (is.vector(y)) y <- t(as.matrix(y)) nx <- nrow(x) ny <- nrow(y) d <- ncol(x) if (ff) difs <- ff(init=0, dim=c(nx*ny,d)) else difs <- matrix(ncol=d,nrow=nx*ny) for (j in 1:d) { difs[,j] <- rep(x[,j], times=ny) - rep(y[1:ny,j], each=nx) ## jth column of difs contains all the differences X_{ij}-Y_{kj} } if (upper) { ind.remove <- numeric() for (j in 1:(nx-1)) ind.remove <- c(ind.remove, (j*nx+1):(j*nx+j)) return(difs[-ind.remove,]) } else return(difs) } ########################################################################## ## Odd factorial ########################################################################## OF <- function(m) { factorial(m)/(2^(m/2)*factorial(m/2)) } ########################################################################## ## Matrix square root - taken from Stephen Lake ## http://www5.biostat.wustl.edu/s-news/s-news-archive/200109/msg00067.html ########################################################################## matrix.sqrt <- function(A) { if (length(A)==1) return(sqrt(A)) sva <- svd(A) if (min(sva$d)>=0) Asqrt <- sva$u %*% diag(sqrt(sva$d)) %*% t(sva$v) else stop("Matrix square root is not defined") return(Asqrt) } ########################################################################## ## Matrix power ########################################################################## matrix.pow <- function(A, n) { if (nrow(A)!=ncol(A)) stop("A must be a square matrix") if (floor(n)!=n) stop("n must be an integer") if (n==0) return(diag(ncol(A))) if (n < 0) return(matrix.pow(A=chol2inv(chol(A)), n=-n)) ## trap non-integer n and return an error if (n == 1) return(A) result <- diag(1, ncol(A)) while (n > 0) { if (n %% 2 != 0) { result <- result %*% A n <- n - 1 } A <- A %*% A n <- n / 2 } return(result) } ########################################################################## ## Kmat computes the commutation matrix of orders m,n ########################################################################## Kmat <- function(m, n) { K <- matrix(0,ncol=m*n,nrow=m*n) i <-1:m; j <- 1:n rows <- rowSums(expand.grid((i-1)*n,j)) cols <- rowSums(expand.grid(i,(j-1)*m)) positions <- cbind(rows,cols) K[positions] <- 1 return(K) } ########################################################################## ## mat.Kprod computes row-wise Kronecker products of matrices ## Returns a matrix with rows U[i,]%x%V[i,] ########################################################################## mat.Kprod <- function(U,V) { n1 <- nrow(U) n2 <- nrow(V) if (n1!=n2) stop("U and V must have the same number of vectors") p <- ncol(U) q <- ncol(V) onep <- rep(1,p) oneq <- rep(1,q) P <- (U%x%t(oneq))*(t(onep)%x%V) return(P) } ########################################################################## ## Kpow computes the Kronecker power of a matrix A ########################################################################## Kpow <- function(A,pow) { if (floor(pow)!=pow) stop("pow must be an integer") Apow <- A if (pow==0) { Apow <- 1 } if (pow>1) { for(i in 2:pow) Apow <- Apow%x%A } return(Apow) } ########################################################################## ## Kronecker sum ########################################################################## Ksum <- function(A,B) { AB <- numeric() for (i in 1:nrow(A)) for (j in 1:nrow(B)) AB <- rbind(AB, A[i,] + B[j,]) return(AB) } ########################################################################## ## Row-wise Kronecker product ########################################################################## rowKpow <- function(A,B,r=1,s=1) { A <- as.matrix(A) if (missing(B)) { res <- t(apply(t(A), 2, Kpow, r)) } else { B <- as.matrix(B) if (nrow(A)!=nrow(B)) stop("A and B must have same number of rows") res <- numeric() Ar <- t(apply(t(A), 2, Kpow, r)) Bs <- t(apply(t(B), 2, Kpow, s)) for (i in 1:ncol(Ar)) res <- cbind(res, apply(Bs, 2, FUN="*", Ar[,i])) } return(drop(res)) } getRow <- function(object, n) { return(object[n,]) } ## Returns a matrix with the pow-th Kronecker power of A[i,] in the i-th row mat.Kpow <- function(A,pow) { Apow <- A if (pow==0) { Apow <- matrix(1,nrow=nrow(A), ncol=1) } if (pow>1) { for(i in 2:pow) Apow <- mat.Kprod(Apow,A) } return(Apow) } ## Vector of all r-th partial derivatives of the normal density at x=0, i.e., D^{\otimes r)\phi(0) DrL0 <- function(d,r) { v <- as.vector(Kpow(A=vec(diag(d)),pow=r/2)) DL0 <- (-1)^(r/2)*(2*pi)^(-d/2)*OF(r)*matrix(Sdrv(d=d, r=r, v=v), ncol=1) return(DL0) } ########################################################################## ## Symmetriser matrix ########################################################################## ## Wrapper functions for Chacon & Duong (2014) Sdr <- function(d, r, type="recursive") { type <- match.arg(type, c("recursive", "direct")) Sdr.mat <- do.call(paste("Sdr", type, sep="."), list(d=d, r=r)) return(Sdr.mat) } Sdrv <- function(d, r, v, type="recursive") { type <- match.arg(type, c("recursive", "direct")) v <- as.vector(v) Sdrvec <- do.call(paste("Sdrv", type, sep="."), list(d=d, r=r, v=v)) return(Sdrvec) } ########################################################################## ## Sdr.direct computes the symmetrizer matrix S_{d,r} based on Equation (4) ## as described in Section 3 of Chacon and Duong (2014) ########################################################################## Sdr.direct <- function(d,r) { S <- matrix(0,ncol=d^r,nrow=d^r) per <- permn.ks(r) per.rep <- pinv.all(d,r) nper <- factorial(r) nper.rep <- d^r per <- matrix(unlist(per), byrow=TRUE, ncol=r, nrow=nper) pow <- 0:(r-1) dpow <- d^pow if (nper.rep<=nper) { dpow.mat <- matrix(rep(dpow,nper),byrow=TRUE,ncol=r,nrow=nper) for(i in 1:nper.rep) { ## Loop over no. perms with reps (d^r) pinvi <- per.rep[i,] sigpinvi <- matrix(pinvi[per],byrow=FALSE,nrow=nrow(per),ncol=ncol(per)) psigpinvi <- drop(1+rowSums((sigpinvi-1)*dpow.mat)) S[i,] <- tabulate(psigpinvi,nbins=d^r) } } if (nper=2) { Id <- diag(d) T <- Id A <- Kmat(d,d) for (j in 2:r) { T <- ((j-1)/j)*(A%*%(T%x%Id)%*%A)+A/j S <- (S%x%Id)%*%T if (j1) { delta <- sapply(f.diff$eval.points, diff) delta <- rbind(head(delta, n=1), delta) if (d==2) riemann.sum <- sum(outer(delta[,1], delta[,2]) * f.diff$estimate) else if (d==3) riemann.sum <- sum(outer(outer(delta[,1], delta[,2]), delta[,3]) * f.diff$estimate) } return(riemann.sum) } ## ISE of difference between two KDEs ise.diff <- function(fhat1, fhat2, xmin, xmax) { if(!isTRUE(all.equal(fhat1$eval.points, fhat2$eval.points))) stop("fhat1 and fhat2 need to de defined on the same grid") fhat.sq <- fhat1 fhat.sq$estimate <- (fhat1$estimate - fhat2$estimate)^2 if (missing(xmin) & missing(xmax)) int <- integral.kde(fhat=fhat.sq, q=max(fhat.sq$eval.points)+0.1*abs(max(fhat.sq$eval.points)), density=FALSE) if (missing(xmin) & !missing(xmax)) int <- integral.kde(fhat=fhat.sq, q=xmax, density=FALSE) if (!missing(xmin) & missing(xmax)) int <- integral.kde(fhat=fhat.sq, q=max(fhat.sq$eval.points)+0.1*abs(max(fhat.sq$eval.points)), density=FALSE) - integral.kde(fhat=fhat.sq, q=xmin, density=FALSE) if (!missing(xmin) & !missing(xmax)) int <- integral.kde(fhat=fhat.sq, q=xmax, density=FALSE) - integral.kde(fhat=fhat.sq, q=xmin, density=FALSE) return(int) } ########################################################################## ## Create transparent colour ## partial alias for plot3D::alpha.col ########################################################################## ## copied from plot3D::alpha.col plot3D.alpha.col <- function (col="grey", alpha=0.5) { RGBini <- col2rgb(col) return(rgb(t(RGBini), maxColorValue=255, alpha=alpha*255)) } transparency.col <- function(col, alpha) { trans.ind <- col!="transparent" col[trans.ind] <- plot3D.alpha.col(col[trans.ind], alpha=alpha) return(col) } ks/R/kda.R0000644000176200001440000007404314547755547012011 0ustar liggesusers############################################################################## ## Kernel discriminant analysis ############################################################################### ############################################################################### ## KDEs of individual densities for KDA - 1- to 3-dim ## ## Parameters ## x - data values ## group - group variable ## Hs - bandwidth matrices ## ## Returns ## List with components (class dade) ## x - list of data values ## eval.points - evaluation points of dnesity estimate ## estimate - list of density estimate ## H - list of bandwidth matrices ############################################################################## kda <- function(x, x.group, Hs, hs, prior.prob=NULL, gridsize, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, kde.flag=TRUE) { if (missing(eval.points)) eval.points <- x if (!is.factor(x.group)) x.group <- factor(x.group) gr <- levels(x.group) m <- length(gr) ## default values ksd <- ks.defaults(x=x, w=w, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned bgridsize <- ksd$bgridsize gridsize <- ksd$gridsize if (!is.null(prior.prob)) if (!(identical(all.equal(sum(prior.prob), 1), TRUE))) stop("Sum of weights not equal to 1") ## clip data to xmin,xmax grid for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] xt <- truncate.grid(x=x, y=1:n, xmin=xmin, xmax=xmax) x <- xt$x; w <- w[xt$y]; x.group <- x.group[xt$y]; n <- length(w) } if (d==1) { if (missing(hs)) hs <- hkda(x=x, x.group=x.group, bw="plugin", nstage=2, binned=default.bflag(d=d,n=n)) ## compute KDA on grid if (kde.flag) fhat.list <- kda.1d(x=x, x.group=x.group, hs=hs, prior.prob=prior.prob, gridsize=gridsize, supp=supp, binned=binned, bgridsize=bgridsize, xmin=xmin, xmax=xmax, compute.cont=compute.cont, approx.cont=approx.cont, w=w) ## compute KDA at eval.points fhat <- kda.1d(x=x, x.group=x.group, hs=hs, prior.prob=prior.prob, gridsize=gridsize, supp=supp, binned=binned, bgridsize=bgridsize, xmin=xmin, xmax=xmax, eval.points=eval.points, compute.cont=compute.cont, approx.cont=approx.cont, w=w) fhat.wt <- matrix(0, ncol=m, nrow=length(eval.points)) } else { if (missing(Hs)) Hs <- Hkda(x=x, x.group=x.group, bw="plugin", binned=default.bflag(d=d, n=n)) ## Compute KDA on grid if (d>3) kde.flag <- FALSE if (kde.flag) fhat.list <- kda.nd(x=x, x.group=x.group, Hs=Hs, prior.prob=prior.prob, gridsize=gridsize, supp=supp, binned=binned, bgridsize=bgridsize, xmin=xmin, xmax=xmax, compute.cont=compute.cont, approx.cont=approx.cont) ## compute KDA at eval.points fhat <- kda.nd(x=x, x.group=x.group, Hs=Hs, prior.prob=prior.prob, gridsize=gridsize, supp=supp, binned=FALSE, bgridsize=bgridsize, xmin=xmin, xmax=xmax, eval.points=eval.points, compute.cont=compute.cont, approx.cont=approx.cont) fhat.wt <- matrix(0, ncol=m, nrow=nrow(eval.points)) } for (j in 1:m) fhat.wt[,j] <- fhat$estimate[[j]]* fhat$prior.prob[j] ## assign y according largest weighted density value disc.gr.temp <- apply(fhat.wt, 1, which.max) disc.gr <- as.factor(gr[disc.gr.temp]) if (is.numeric(gr)) disc.gr <- as.numeric(levels(disc.gr))[disc.gr] if (kde.flag) fhat.list$x.group.estimate <- disc.gr else fhat.list <- disc.gr fhat.list$type <- "kda" return(fhat.list) } kda.1d <- function(x, x.group, hs, prior.prob, gridsize, supp, eval.points, binned, bgridsize, xmin, xmax, w, compute.cont, approx.cont) { gr <- levels(x.group) m <- length(gr) d <- 1 hmax <- max(hs) if (missing(xmin)) xmin <- min(x) - supp*hmax if (missing(xmax)) xmax <- max(x) + supp*hmax fhat.list <- list() for (j in 1:m) { xx <- x[x.group==gr[j]] ww <- w[x.group==gr[j]] h <- hs[j] ## compute individual density estimate if (missing(eval.points)) fhat.temp <- kde(x=xx, h=h, supp=supp, xmin=xmin, xmax=xmax, bgridsize=bgridsize, gridsize=gridsize, w=ww, binned=binned) else fhat.temp <- kde(x=xx, h=h, w=ww, binned=binned, bgridsize=bgridsize, gridsize=gridsize, eval.points=eval.points) fhat.list$estimate <- c(fhat.list$estimate, list(fhat.temp$estimate)) fhat.list$eval.points <- fhat.temp$eval.points fhat.list$x <- c(fhat.list$x, list(xx)) fhat.list$h <- c(fhat.list$h, h) fhat.list$H <- c(fhat.list$H, h^2) fhat.list$w <- c(fhat.list$w, list(ww)) ## compute prob contour levels if (compute.cont & missing(eval.points)) { contlev <- contourLevels(fhat.temp, cont=1:99, approx.cont=approx.cont) fhat.list$cont <- c(fhat.list$cont, list(contlev)) } } fhat.list$names <- parse.name(x) ## add variable names fhat.list$binned <- binned fhat.list$gridded <- fhat.temp$gridded if (is.null(prior.prob)) { pr <- rep(0, length(gr)) for (j in 1:length(gr)) pr[j] <- length(which(x.group==gr[j])) pr <- pr/length(x) fhat.list$prior.prob <- pr } else fhat.list$prior.prob <- prior.prob fhat.list$x.group <- x.group class(fhat.list) <- "kda" return(fhat.list) } kda.nd <- function(x, x.group, Hs, prior.prob, gridsize, supp, eval.points, binned, bgridsize, xmin, xmax, w, compute.cont, approx.cont) { if (is.data.frame(x)) x <- as.matrix(x) gr <- levels(x.group) m <- length(gr) d <- ncol(x) ## find largest bandwidth matrix to initialise grid detH <- vector() for (j in 1:m) detH[j] <- det(Hs[((j-1)*d+1) : (j*d),]) Hmax.ind <- which.max(detH) Hmax <- Hs[((Hmax.ind-1)*d+1) : (Hmax.ind*d),] if (missing(xmin)) xmin <- apply(x, 2, min) - supp*max(sqrt(diag(Hmax))) if (missing(xmax)) xmax <- apply(x, 2, max) + supp*max(sqrt(diag(Hmax))) if (missing(w)) w <- rep(1, nrow(x)) if (binned & d > 4) stop("Binning only available for 1- to 4-d data") if (missing(bgridsize)) bgridsize <- default.gridsize(d) if (missing(gridsize)) gridsize <- default.gridsize(d) fhat.list <- list() for (j in 1:m) { xx <- x[x.group==gr[j],] ww <- w[x.group==gr[j]] H <- Hs[((j-1)*d+1) : (j*d),] ## compute individual density estimate if (binned) fhat.temp <- kdde(x=xx, bgridsize=bgridsize, H=H, xmin=xmin, xmax=xmax, w=ww, deriv.order=0, binned=TRUE) else if (missing(eval.points)) fhat.temp <- kde(x=xx, H=H, supp=supp, xmin=xmin, xmax=xmax, gridsize=gridsize, w=ww) else fhat.temp <- kde(x=xx, H=H, eval.points=eval.points, w=ww) fhat.list$estimate <- c(fhat.list$estimate, list(fhat.temp$estimate)) fhat.list$eval.points <- fhat.temp$eval.points fhat.list$x <- c(fhat.list$x, list(xx)) fhat.list$H <- c(fhat.list$H, list(H)) fhat.list$w <- c(fhat.list$w, list(ww)) ## compute prob contour levels if (compute.cont & missing(eval.points)) { contlev <- contourLevels(fhat.temp, cont=1:99, approx.cont=approx.cont) fhat.list$cont <- c(fhat.list$cont, list(contlev)) } } fhat.list$names <- parse.name(x) ## add variable names fhat.list$binned <- binned fhat.list$gridded <- fhat.temp$gridded if (is.null(prior.prob)) { pr <- rep(0, length(gr)) for (j in 1:length(gr)) pr[j] <- length(which(x.group==gr[j])) pr <- pr/nrow(x) fhat.list$prior.prob <- pr } else fhat.list$prior.prob <- prior.prob fhat.list$x.group <- x.group class(fhat.list) <- "kda" return (fhat.list) } ############################################################################## ## S3 methods for kda objects ############################################################################## ## contourLevel method contourLevels.kda <- function(x, prob, cont, nlevels=5, approx=TRUE,...) { fhat <- x m <- length(fhat$x) hts <- list() for (j in 1:m) { fhatj <- list(x=fhat$x[[j]], eval.points=fhat$eval.points, estimate=fhat$estimate[[j]], H=fhat$H[[j]], binned=fhat$binned, gridded=fhat$gridded) class(fhatj) <- "kde" hts[[j]] <- contourLevels(x=fhatj, prob=prob, cont=cont, nlevels=nlevels, approx=approx, ...) } return(hts) } ## predict method predict.kda <- function(object, ..., x) { fhat <- object m <- length(fhat$prior.prob) if (is.vector(fhat$x[[1]])) n <- length(x) else { if (is.vector(x)) n <- 1 else n <- nrow(x) } fhat.temp <- matrix(0, ncol=m, nrow=n) for (j in 1:m) fhat.temp[,j] <- fhat$prior.prob[j]*grid.interp(x=x, gridx=fhat$eval.points, f=fhat$estimate[[j]]) est.group <- apply(fhat.temp, 1, which.max) est.group <- as.factor(sort(unique(fhat$x.group))[est.group]) return(est.group) } ## plot method ## fhat - output from `kda.kde' ## y - data points (separate from training data inside fhat) ## y.group - data group labels plot.kda <- function(x, y, y.group, ...) { if (is.vector(x$x[[1]])) plotkda.1d(x=x, y=y, y.group=y.group, ...) else { d <- ncol(x$x[[1]]) if (d==2) { opr <- options()$preferRaster; if (!is.null(opr)) if (!opr) options("preferRaster"=TRUE) plotkda.2d(x=x, y=y, y.group=y.group, ...) if (!is.null(opr)) options("preferRaster"=opr) } else if (d==3) plotkda.3d(x=x, y=y, y.group=y.group, ...) } } plotkda.1d <- function(x, y, y.group, prior.prob=NULL, xlim, ylim, xlab, ylab="Weighted density function", drawpoints=FALSE, col, col.fun, col.part, col.pt, lty, jitter=TRUE, rugsize, add=FALSE, alpha=1, ...) { fhat <- x m <- length(fhat$x) if (missing(xlab)) xlab <- fhat$names if (is.null(prior.prob)) prior.prob <- fhat$prior.prob if (m != length(prior.prob)) stop("prior.prob not same length as number of components in fhat") if (!(identical(all.equal(sum(prior.prob), 1), TRUE))) stop("Sum of weights not equal to 1") weighted.fhat <- matrix(0, nrow=length(fhat$eval.points), ncol=m) for (j in 1:m) weighted.fhat[,j] <- fhat$estimate[[j]]*fhat$prior.prob[j] if (missing(xlim)) xlim <- range(fhat$eval.points) if (missing(ylim)) ylim <- range(weighted.fhat) if (missing(lty)) lty <- rep(1, m) if (length(lty) < m) lty <- rep(lty, m) if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="Dark2") } if (missing(col)) col <- col.fun(m) if (length(col) < m) col <- rep(col, m) if (missing(col.part)) col.part <- plot3D.alpha.col(col, alpha=alpha) if (missing(col.pt)) col.pt <- col.fun(m) if (length(col.pt)==1) col.pt <- rep(col.pt, m) ## plot each training group's KDE in separate colour and line type if (!add) plot(fhat$eval.points, weighted.fhat[,1], type="l", xlab=xlab, ylab=ylab, xlim=xlim, ylim=ylim, lty=lty[1], col=col[1], ...) if (m > 1) for (j in 2:m) lines(fhat$eval.points, weighted.fhat[,j], lty=lty[j], col=col[j], ...) ydata <- seq(min(fhat$eval.points), max(fhat$eval.points), length=401) x.gr <- 1:m ydata.gr <- x.gr[apply(weighted.fhat,1, which.max)] ## draw partition class as rug-like plot plot.lim <- par()$usr if (missing(rugsize)) rugsize <- abs(plot.lim[4]-plot.lim[3])/50 image(ydata, c(plot.lim[3], plot.lim[3]+rugsize), cbind(as.numeric(ydata.gr), as.numeric(ydata.gr)), breaks=0.5+(0:m), add=TRUE, col=col.part, ...) for (j in 1:m) { ## draw data points if (drawpoints) { if (missing(y)) yy <- fhat$x[[j]] else { if (missing(y.group)) yy <- fhat$x[[j]] else yy <- y[y.group==levels(y.group)[j]] } if (jitter) yy <- jitter(yy) rug(yy, col=col.pt[j], ticksize=-0.03) } } } plotkda.2d <- function(x, y, y.group, prior.prob=NULL, display.part="filled.contour", cont=c(25,50,75), abs.cont, approx.cont=TRUE, xlim, ylim, xlab, ylab, drawpoints=FALSE, drawlabels=TRUE, cex=1, pch, lty, part=TRUE, col, col.fun, col.part, col.pt, alpha=1, lwd=1, lwd.part=0, labcex=1, add=FALSE, ...) { fhat <- x m <- length(fhat$x) xtemp <- numeric() for (j in 1:m) xtemp <- rbind(xtemp, fhat$x[[j]]) if (missing(xlim)) xlim <- range(xtemp[,1]) if (missing(ylim)) ylim <- range(xtemp[,2]) if (missing(pch)) pch <- 1:m if (missing(lty)) lty <- rep(1, m) if (length(lty) < m) lty <- rep(lty, m) if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="Dark2", alpha=1) } if (missing(col)) col <- col.fun(m) if (length(col) < m) col <- rep(col, m) if (missing(col.part)) col.part <- plot3D.alpha.col(col, alpha=alpha) if (missing(col.pt)) col.pt <- col.fun(m) if (length(col.pt)==1) col.pt <- rep(col.pt, m) x.names <- fhat$names if (!is.null(x.names)) { if (missing(xlab)) xlab <- x.names[1] if (missing(ylab)) ylab <- x.names[2] } else { xlab="x" ylab="y" } if (is.null(prior.prob)) prior.prob <- fhat$prior.prob if (m != length(prior.prob)) stop("prior.prob not same length as number of components in fhat") if (!(identical(all.equal(sum(prior.prob), 1), TRUE))) stop("Sum of weights not equal to 1") ## set up plot if (!add) { if (missing(y)) plot(fhat$x[[1]], type="n", xlab=xlab, ylab=ylab, xlim=xlim, ylim=ylim, ...) else plot(y, type="n", xlab=xlab, ylab=ylab, xlim=xlim, ylim=ylim, ...) } ## set up common grid for all densities class.grid <- array(0, dim=dim(fhat$estimate[[1]])) temp <- matrix(0, ncol=length(fhat$est), nrow=nrow(fhat$est[[1]])) for (j in 1:ncol(fhat$estimate[[1]])) { for (k in 1:length(fhat$estimate)) temp[,k] <- fhat$estimate[[k]][,j]* prior.prob[k] class.grid[,j] <- max.col(temp) } ## draw partition fhat.part <- fhat fhat.part$estimate <- class.grid fhat.part$H <- fhat$H[[1]] fhat.part$x <- fhat$x[[1]] fhat.part$w <- fhat$w[[1]] fhat.part$cont <- fhat$cont[[1]] class(fhat.part) <- "kde.part" if (part) plot(fhat.part, col=col.part, add=TRUE, display=display.part, lwd=lwd.part, alpha=alpha, ...) ## common contour levels removed from >= v1.5.3 if (missing(abs.cont)) { hts <- contourLevels(fhat, prob=(100-cont)/100, approx=approx.cont) nhts <- length(hts[[1]]) } else { hts <- abs.cont nhts <- length(hts) } ## draw contours for (j in 1:m) { for (i in 1:nhts) { if (missing(abs.cont)) { scale <- cont[i]/hts[[j]][i] contour(fhat$eval.points[[1]], fhat$eval.points[[2]], fhat$estimate[[j]]*scale, level=hts[[j]][i]*scale, add=TRUE, drawlabels=drawlabels, lty=lty[j], col=col[j], lwd=lwd, labcex=labcex, ...) } else { contour(fhat$eval.points[[1]], fhat$eval.points[[2]], fhat$estimate[[j]], level=hts[i], add=TRUE, drawlabels=drawlabels, lty=lty[j], col=col[j], lwd=lwd, labcex=labcex, ...) } } } for (j in 1:m) { ## draw data points if (drawpoints) { if (missing(y)) points(fhat$x[[j]], pch=pch[j], col=col.pt[j], cex=cex) else { if (missing(y.group)) { points(y, col=col.pt[1], cex=cex) j <- m+1 } else points(y[y.group==levels(y.group)[j],], pch=pch[j], col=col.pt[j], cex=cex) } } } } plotkda.3d <- function(x, y, y.group, prior.prob=NULL, display="plot3D", cont=c(25,50,75), abs.cont, approx.cont=TRUE, colors, col, col.fun, col.pt, alpha=0.5, alphavec, xlab, ylab, zlab, drawpoints=FALSE, size=3, cex=1, pch, theta=-30, phi=40, d=4, ticktype="detailed", bty="f", add=FALSE, ...) { fhat <- x m <- length(fhat$x) if (is.null(prior.prob)) prior.prob <- fhat$prior.prob if (m != length(prior.prob)) stop("prior.prob not same length as number of components in fhat") if (!(identical(all.equal(sum(prior.prob), 1), TRUE))) stop("Sum of prior weights not equal to 1") if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="Dark2") } if (missing(col)) col <- col.fun(m) if (length(col) < m) col <- rep(col, m) if (missing(col.pt)) col.pt <- col.fun(m) if (length(col.pt)==1) col.pt <- rep(col.pt, m) if (missing(pch)) pch <- 1:m if (length(pch)==1) pch <- rep(pch, m) x.names <- colnames(fhat$x[[1]]) if (missing(xlab)) if (is.null(x.names)) xlab <- "x" else xlab <- x.names[1] if (missing(ylab)) if (is.null(x.names)) ylab <- "y" else ylab <- x.names[2] if (missing(zlab)) if (is.null(x.names)) zlab <- "z" else zlab <- x.names[3] xx <- numeric(0) for (j in 1:m) xx <- rbind(xx, fhat$x[[j]]) ## common contour levels removed from >= v1.5.3 if (missing(abs.cont)) { hts <- contourLevels(fhat, prob=(100-cont)/100, approx=approx.cont) nhts <- length(hts[[1]]) } else { hts <- abs.cont nhts <- length(hts) } if (missing(alphavec)) alphavec <- seq(0.05,0.2,length=nhts) disp <- match.arg(display, c("plot3D", "rgl")) if (disp %in% "plot3D") { if (!requireNamespace("plot3D", quietly=TRUE)) stop("Install the plot3D package as it is required.", call.=FALSE) for (j in 1:m) { for (i in 1:nhts) { cti <- hts[[j]][nhts-i+1] if (cti <= max(fhat$estimate[[j]])) plot3D::isosurf3D(x=fhat$eval.points[[1]], y=fhat$eval.points[[2]], z=fhat$eval.points[[3]], colvar=fhat$estimate[[j]], level=cti, add=add | (j>1 | i>1), alpha=alphavec[i], col=col[j], phi=phi, theta=theta, xlab=xlab, ylab=ylab, zlab=zlab, d=d, ticktype=ticktype, bty=bty, ...) } } ## plot points if (drawpoints) { for (j in 1:m) { if (missing(y)) plot3D::points3D(fhat$x[[j]][,1], fhat$x[[j]][,2], fhat$x[[j]][,3], col=col.pt[j], cex=cex, alpha=1, add=TRUE, pch=pch[j]) else { if (missing(y.group)) plot3D::points3D(y[,1], y[,2], y[,3], col=col.pt, cex=cex, alpha=alpha, add=TRUE, pch=pch[j]) else { y.temp <- y[y.group==levels(y.group)[j],] if (nrow(y.temp)>0) plot3D::points3D(y.temp[,1], y.temp[,2], y.temp[,3], col=col.pt[j], cex=cex, alpha=alpha, add=TRUE, pch=pch[j]) } } } } } else if (disp %in% "rgl") { ## suggestions from Viktor Petukhov 08/03/2018 if (!requireNamespace("rgl", quietly=TRUE)) stop("Install the rgl package as it is required.", call.=FALSE) if (!requireNamespace("misc3d", quietly=TRUE)) stop("Install the misc3d package as it is required.", call.=FALSE) if (missing(colors)) colors <- col xtemp <- numeric(); for (i in 1:length(fhat$x)) xtemp <- rbind(xtemp, fhat$x[[i]]) rgl::plot3d(x=xtemp[,1], y=xtemp[,2], z=xtemp[,3], type="n", xlab=xlab, ylab=ylab, zlab=zlab, ...) for (j in 1:m) { for (i in 1:nhts) { cti <- hts[[j]][nhts-i+1] if (cti <= max(fhat$estimate[[j]])) misc3d::contour3d(x=fhat$eval.points[[1]], y=fhat$eval.points[[2]], z=fhat$eval.points[[3]], f=fhat$estimate[[j]], level=cti, add=TRUE, alpha=alphavec[i], color=colors[j], ...) } if (drawpoints) ## plot points { if (missing(y)) rgl::points3d(fhat$x[[j]][,1], fhat$x[[j]][,2], fhat$x[[j]][,3], color=col.pt[j], size=size, alpha=alpha) else { if (missing(y.group)) rgl::points3d(y[,1], y[,2], y[,3], color=col.pt, size=size, alpha=1) else { y.temp <- y[y.group==levels(y.group)[j],] if (nrow(y.temp)>0) rgl::points3d(y.temp[,1], y.temp[,2], y.temp[,3], color=col.pt[j], size=size, alpha=alpha) } } } } } } ############################################################################### ## Find bandwidths for each class in training set, for 2- to 6-dim ## ## Parameters ## x - data values ## group - group variable ## bw - type of bandwidth selector ## nstage, pilot, pre - parameters for plugin bandwidths ## diag - FALSE - use full b/w matrices ## - TRUE - use diag b/w matrices ## ## Returns ## Matrix of bandwidths for each group in training set ############################################################################### hkda <- function(x, x.group, bw="plugin", ...) { gr <- sort(unique(x.group)) m <- length(gr) bw <- match.arg(bw, c("lscv", "plugin", "scv")) hs <- numeric(0) for (i in 1:m) { y <- x[x.group==gr[i]] if (bw=="plugin") h <- hpi(y, ...) else if ((bw=="lscv") | (bw=="ucv")) h <- hlscv(y, ...) else if (bw=="scv") h <- hscv(y, ...) hs <- c(hs, h) } return(hs) } Hkda <- function(x, x.group, Hstart, bw="plugin", ...) { d <- ncol(x) gr <- sort(unique(x.group)) m <- length(gr) bw <- match.arg(bw, c("lscv", "plugin", "scv", "ucv")) Hs <- numeric(0) for (i in 1:m) { y <- x[x.group==gr[i],] if (!missing(Hstart)) { Hstarty <- Hstart[((i-1)*d+1) : (i*d),] if ((bw=="lscv") | (bw=="ucv")) H <- Hlscv(y, Hstart=Hstarty, ...) else if (bw=="scv") H <- Hscv(y, Hstart=Hstarty, ...) else if (bw=="plugin") H <- Hpi(y, Hstart=Hstarty, ...) } else { if ((bw=="lscv") | (bw=="ucv")) H <- Hlscv(y, ...) else if (bw=="scv") H <- Hscv(y, ...) else if (bw=="plugin") H <- Hpi(y, ...) } Hs <- rbind(Hs, H) } return(Hs) } Hkda.diag <- function(x, x.group, bw="plugin", ...) { d <- ncol(x) gr <- sort(unique(x.group)) m <- length(gr) bw <- match.arg(bw, c("lscv", "plugin", "scv", "ucv")) Hs <- numeric(0) for (i in 1:m) { y <- x[x.group==gr[i],] if ((bw=="lscv") | (bw=="ucv")) H <- Hlscv.diag(y, ...) else if (bw=="plugin") H <- Hpi.diag(y, ...) else if (bw=="scv") H <- Hscv.diag(y, ...) Hs <- rbind(Hs, H) } return(Hs) } ############################################################################### ## Compares true group classification with an estimated one ## ## Parameters ## group - true group variable ## est.group - estimated group variable ## ## Returns ## List with components ## comp - cross-classification table of groupings - true groups are the rows, ## estimated groups are the columns ## error - total mis-classification rate ############################################################################### compare <- function(x.group, est.group, by.group=FALSE) { if (length(x.group)!=length(est.group)) stop("Group label vectors not the same length") if (!is.factor(x.group)) x.group <- factor(x.group) grlab <- levels(x.group) m <- length(grlab) comp <- table(x.group, est.group) if (by.group) { er <- vector() for (i in 1:m) er[i] <- 1-comp[i,i]/rowSums(comp)[i] er <- matrix(er, ncol=1) er <- rbind(er, 1 - sum(diag(comp))/sum(comp)) rownames(er) <- c(as.character(paste(grlab, "(true)")), "Total") colnames(er) <- "error" } else er <- 1 - sum(diag(comp))/sum(comp) comp <- cbind(comp, rowSums(comp)) comp <- rbind(comp, colSums(comp)) colnames(comp) <- c(as.character(paste(grlab, "(est.)")), "Total") rownames(comp) <- c(as.character(paste(grlab, "(true)")), "Total") if (nrow(comp)==2 & nrow(comp)==2) { TN <- comp[1,1]; FP <- comp[1,2]; FN <- comp[2,1]; TP <- comp[2,2] spec <- 1-(FP/(FP+TN)) sens <- 1-(FN/(TP+FN)) comp <- list(cross=comp, error=er, TP=TP, FP=FP, FN=FN, TN=TN, spec=spec, sens=sens) } else comp <- list(cross=comp, error=er) return(comp) } ############################################################################### ## Computes cross-validated misclassification rates (for use when test data is ## not independent of training data) for KDA ## ## Parameters ## x - training data ## x.group - group variable for x ## y - data values to be classified ## Hs - bandwidth matrices ## prior.prob - prior probabilities ## ## Returns ## List with components ## comp - cross-classification table of groupings - true groups are the rows, ## estimated groups are the columns ## error - total mis-classification rate ############################################################################### compare.kda.cv <- function(x, x.group, bw="plugin", prior.prob=NULL, Hstart, by.group=FALSE, verbose=FALSE, recompute=FALSE, ...) { if (verbose) pb <- txtProgressBar() bw <- match.arg(bw, c("lscv", "plugin", "scv", "ucv")) ## 1-d if (is.vector(x)) { n <- length(x) h <- hkda(x, x.group, bw=bw, ...) gr <- sort(unique(x.group)) kda.cv.gr <- x.group for (i in 1:n) { h.mod <- h ## find group that x[i] belongs to ind <- which(x.group[i]==gr) indx <- x.group==gr[ind] indx[i] <- FALSE if ((bw=="lscv") | (bw=="ucv")) h.temp <- hlscv(x[indx], , ...) else if (bw=="plugin") h.temp <- hpi(x[indx], , ...) else if (bw=="scv") h.temp <- hscv(x[indx], , ...) h.mod[ind] <- h.temp ## recompute KDA estimate of groups with x[i] excluded if (verbose) setTxtProgressBar(pb, i/n) kda.cv.gr[i] <- kda(x[-i], x.group[-i], hs=h.mod, eval.points=x, prior.prob=prior.prob, kde.flag=FALSE)[i] } if (verbose) close(pb) return(compare(x.group, kda.cv.gr, by.group=by.group)) } ## multi-dimensional n <- nrow(x) d <- ncol(x) if (!missing(Hstart)) H <- Hkda(x, x.group, bw=bw, Hstart=Hstart, ...) else H <- Hkda(x, x.group, bw=bw, ...) ## classify data x using KDA rules based on x itself ## kda.group <- kda(x, x.group, Hs=H, y=x, prior.prob=prior.prob) ## comp <- compare(x.group, kda.group) gr <- sort(unique(x.group)) kda.cv.gr <- x.group for (i in 1:n) { H.mod <- H ## find group that x[i] belongs to ind <- which(x.group[i]==gr) indx <- x.group==gr[ind] indx[i] <- FALSE if (recompute) { ## compute b/w matrix for that group with x[i] excluded if (!missing(Hstart)) { Hstart.temp <- Hstart[((ind-1)*d+1):(ind*d),] if (bw=="plugin") H.temp <- Hpi(x[indx,], Hstart=Hstart.temp, ...) else if (bw=="scv") H.temp <- Hscv(x[indx,], Hstart=Hstart.temp, ...) else if ((bw=="lscv") | (bw=="ucv")) H.temp <- Hlscv(x[indx,], Hstart=Hstart.temp, ...) } else { if (bw=="plugin") H.temp <- Hpi(x[indx,], ...) else if (bw=="scv") H.temp <- Hscv(x[indx,], ...) else if ((bw=="lscv") | (bw=="ucv")) H.temp <- Hlscv(x[indx,], ...) } H.mod[((ind-1)*d+1):(ind*d),] <- H.temp } ## recompute KDA estimate of groups with x[i] excluded if (verbose) setTxtProgressBar(pb, i/n) kda.cv.gr[i] <- kda(x[-i,], x.group[-i], Hs=H.mod, eval.points=x, prior.prob=prior.prob, kde.flag=FALSE)[i] } if (verbose) close(pb) return(compare(x.group, kda.cv.gr, by.group=by.group)) } ############################################################################### ## Same as compare.kda.cv except uses diagonal b/w matrices ############################################################################### compare.kda.diag.cv <- function(x, x.group, bw="plugin", prior.prob=NULL, by.group=FALSE, verbose=FALSE, recompute=FALSE, ...) { if (is.vector(x)) return(compare.kda.cv(x=x, x.group=x.group, by.group=by.group, verbose=verbose, prior.prob=prior.prob, recompute=recompute, ...)) n <- nrow(x) d <- ncol(x) H <- Hkda.diag(x, x.group, bw=bw, ...) if (verbose) pb <- txtProgressBar() bw <- match.arg(bw, c("lscv", "plugin", "scv", "ucv")) gr <- sort(unique(x.group)) kda.cv.gr <- x.group for (i in 1:n) { H.mod <- H if (recompute) { ind <- which(x.group[i]==gr) indx <- x.group==gr[ind] indx[i] <- FALSE if (bw=="plugin") H.temp <- Hpi.diag(x[indx,], ...) else if ((bw=="lscv") | (bw=="ucv")) H.temp <- Hlscv.diag(x[indx,], ...) else if (bw=="scv") H.temp <- Hscv.diag(x[indx,], ...) H.mod[((ind-1)*d+1):(ind*d),] <- H.temp } if (verbose) setTxtProgressBar(pb, i/n) kda.cv.gr[i] <- kda(x[-i,], x.group[-i], Hs=H.mod, eval.points=x, prior.prob=prior.prob, kde.flag=FALSE)[i] } if (verbose) close(pb) return(compare(x.group, kda.cv.gr, by.group=by.group)) } ks/R/kcde.R0000644000176200001440000006452514547754273012157 0ustar liggesusers##################################################################### ## Kernel estimators of the multivariate cdf (cumulative distribution function) ##################################################################### kcde <- function(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, verbose=FALSE, tail.flag="lower.tail") { ## default values ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned gridsize <- ksd$gridsize bgridsize <- ksd$bgridsize tail.flag1 <- match.arg(tail.flag, c("lower.tail", "upper.tail")) ## clip data to xmin, xmax grid for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] if (positive & missing(xmin)) { xmin <- rep(0,d) } xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths if (d==1 & missing(h)) { if (positive) x1 <- log(x) else x1 <- x h <- hpi.kcde(x=x1, nstage=2, binned=default.bflag(d=d, n=n)) } if (d>1 & missing(H)) { if (positive) x1 <- log(x) else x1 <- x H <- Hpi.kcde(x=x1, binned=default.bflag(d=d, n=n), bgridsize=bgridsize, verbose=FALSE) } ## KCDE is computed as cumulative Riemann sum of KDE on a grid if (d==1) { Fhat <- kde(x=x, h=h, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, binned=binned, bgridsize=bgridsize, positive=positive, adj.positive=adj.positive, w=w) if (positive) { ep <- seq(Fhat$eval.points[1], tail(Fhat$eval.points,n=1), length=length(Fhat$eval.points)) Fhat$estimate <- predict(Fhat, x=ep) Fhat$eval.points <- ep } diffe <- abs(diff(Fhat$eval.points)) if (tail.flag1=="lower.tail") Fhat$estimate <- c(0, diffe) * cumsum(Fhat$estimate) else Fhat$estimate <- c(diffe[1], diffe) * (sum(Fhat$estimate) - cumsum(Fhat$estimate)) } else if (d==2) { Fhat <- kde(x=x, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, binned=binned, bgridsize=bgridsize, w=w, verbose=verbose) diffe1 <- abs(diff(Fhat$eval.points[[1]])) diffe2 <- abs(diff(Fhat$eval.points[[2]])) if (tail.flag1=="lower.tail") { Fhat$estimate <- apply(Fhat$estimate, 1, cumsum)*c(0,diffe1) Fhat$estimate <- apply(t(Fhat$estimate), 2, cumsum)*c(0,diffe2) } else { Fhatsum <- matrix(apply(Fhat$estimate, 1, sum), ncol=ncol(Fhat$estimate), nrow=nrow(Fhat$estimate), byrow=TRUE) Fhat$estimate <- (Fhatsum-apply(Fhat$estimate, 1, cumsum))*c(diffe1[1], diffe1) Fhatsum <- matrix(apply(Fhat$estimate, 1, sum), ncol=ncol(Fhat$estimate), nrow=nrow(Fhat$estimate), byrow=TRUE) Fhat$estimate <- (Fhatsum-apply(t(Fhat$estimate), 2, cumsum))*c(diffe2[1], diffe2) } } else if (d==3) { if (missing(H) & !positive) H <- Hpi.kcde(x=x, binned=default.bflag(d=d, n=n), bgridsize=bgridsize, verbose=FALSE) Fhat <- kde(x=x, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, binned=binned, bgridsize=bgridsize, w=w, verbose=verbose) Fhat.temp <- Fhat$estimate diffe1 <- abs(diff(Fhat$eval.points[[1]])) diffe2 <- abs(diff(Fhat$eval.points[[2]])) diffe3 <- abs(diff(Fhat$eval.points[[3]])) if (tail.flag1=="lower.tail") { for (i in 1:dim(Fhat$estimate)[3]) { Fhat.temp[,,i] <- apply(Fhat.temp[,,i], 1, cumsum)*c(0,diffe1) Fhat.temp[,,i] <- apply(t(Fhat.temp[,,i]), 2, cumsum)*c(0,diffe2) } for (i in 1:dim(Fhat$estimate)[1]) for (j in 1:dim(Fhat$estimate)[2]) Fhat.temp[i,j,] <- cumsum(Fhat.temp[i,j,])*c(0,diffe3) Fhat$estimate <- Fhat.temp } else { for (i in 1:dim(Fhat$estimate)[3]) { Fhatsum <- matrix(apply(Fhat.temp[,,i], 1, sum), ncol=ncol(Fhat.temp), nrow=nrow(Fhat.temp), byrow=TRUE) Fhat.temp[,,i] <- (Fhatsum-apply(Fhat.temp[,,i], 1, cumsum))*c(diffe1[1], diffe1) Fhatsum <- matrix(apply(Fhat.temp[,,i], 1, sum), ncol=ncol(Fhat.temp), nrow=nrow(Fhat.temp), byrow=TRUE) Fhat.temp[,,i] <- (Fhatsum-apply(t(Fhat.temp[,,i]), 2, cumsum))*c(diffe2[1],diffe2) } for (i in 1:dim(Fhat$estimate)[1]) for (j in 1:dim(Fhat$estimate)[2]) { Fhatsum <- sum(Fhat.temp[i,j,]) Fhat.temp[i,j,] <- (Fhatsum-cumsum(Fhat.temp[i,j,]))*c(diffe3[1],diffe3) } Fhat$estimate <- Fhat.temp } } ## normalise max CDF estimate equal to 1 Fhat$estimate <- Fhat$estimate/max(Fhat$estimate) if (!missing(eval.points)) { if (d<=3) { Fhat$estimate <- predict(Fhat, x=eval.points) Fhat$eval.points <- eval.points } else { Fhat <- kcde.points(x=x, H=H, eval.points=eval.points, w=w, verbose=verbose, tail.flag=tail.flag1) } } Fhat$tail <- tail.flag1 Fhat$type <- "kcde" class(Fhat) <- "kcde" return(Fhat) } ## KCDE is computed at specified estimation points kcde.points <- function(x, H, eval.points, w, verbose=FALSE, tail.flag="lower.tail") { n <- nrow(x) if (verbose) pb <- txtProgressBar() Fhat <- rep(0, nrow(eval.points)) pmvnorm.temp <- function(x, ...) { return(pmvnorm(mean=x, ...)) } for (i in 1:nrow(eval.points)) { if (verbose) setTxtProgressBar(pb, i/(nrow(eval.points)-1)) if (tail.flag=="lower.tail") Fhat[i] <- sum(apply(x, 1, pmvnorm.temp, upper=eval.points[i,], sigma=H)) else Fhat[i] <- sum(apply(x, 1, pmvnorm.temp, lower=eval.points[i,], sigma=H)) } Fhat <- Fhat/n if (verbose) close(pb) return(list(x=x, eval.points=eval.points, estimate=Fhat, H=H, gridded=FALSE, binned=FALSE, names=NULL, w=w)) } ##################################################################### ## Plotting functions for 1-d to 3-d KCDE ##################################################################### plot.kcde <- function(x, ...) { Fhat <- x if (is.vector(Fhat$x)) plotkcde.1d(Fhat, ...) else { d <- ncol(Fhat$x) if (d==2) { opr <- options()$preferRaster; if (!is.null(opr)) if (!opr) options("preferRaster"=TRUE) plotret <- plotkcde.2d(Fhat, ...) if (!is.null(opr)) options("preferRaster"=opr) invisible(plotret) } else if (d==3) { plotkcde.3d(Fhat, ...) invisible() } else stop ("kde.plot function only available for 1, 2 or 3-d data") } } plotkcde.1d <- function(Fhat, xlab, ylab="Distribution function", add=FALSE, drawpoints=FALSE, col=1, col.pt=4, jitter=FALSE, alpha=1, ...) { if (missing(xlab)) xlab <- Fhat$names if (Fhat$tail=="upper.tail") zlab <- "Survival function" col <- transparency.col(col, alpha=alpha) if (add) lines(Fhat$eval.points, Fhat$estimate, xlab=xlab, ylab=ylab, col=col, ...) else plot(Fhat$eval.points, Fhat$estimate, type="l", xlab=xlab, ylab=ylab, col=col, ...) if (drawpoints) if (jitter) rug(jitter(Fhat$x), col=col.pt) else rug(Fhat$x, col=col.pt) } plotkcde.2d <- function(Fhat, display="persp", cont=seq(10,90, by=10), abs.cont, xlab, ylab, zlab="Distribution function", cex=1, pch=1, add=FALSE, drawpoints=FALSE, drawlabels=TRUE, theta=-30, phi=40, d=4, col.pt=4, col, col.fun, alpha=1, lwd=1, border=NA, thin=3, labcex=1, ticktype="detailed", ...) { disp <- match.arg(display, c("slice", "persp", "image", "filled.contour", "filled.contour2")) if (disp=="filled.contour2") disp <- "filled.contour" if (!is.list(Fhat$eval.points)) stop("Needs a grid of density estimates") if (missing(xlab)) xlab <- Fhat$names[1] if (missing(ylab)) ylab <- Fhat$names[2] if (Fhat$tail=="upper.tail") zlab <- "Survival function" ## perspective/wireframe plot if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="viridis", alpha=alpha) } if (disp=="persp") { hts <- seq(0, 1.1*max(Fhat$estimate, na.rm=TRUE), length=500) if (missing(col)) col <- col.fun(n=length(hts)+1) if (length(col)<(length(hts)+1)) col <- rep(col, length=length(hts)+1) col <- transparency.col(col, alpha=alpha) ## thinning indices plot.ind <- list(seq(1, length(Fhat$eval.points[[1]]), by=thin), seq(1, length(Fhat$eval.points[[2]]), by=thin)) z <- Fhat$estimate[plot.ind[[1]], plot.ind[[2]]] nrz <- nrow(z) ncz <- ncol(z) zfacet <- z[-1, -1] + z[-1, -ncz] + z[-nrz, -1] + z[-nrz, -ncz] facetcol <- cut(zfacet, length(hts)+1, labels=FALSE) plotret <- persp(Fhat$eval.points[[1]][plot.ind[[1]]], Fhat$eval.points[[2]][plot.ind[[2]]], z, theta=theta, phi=phi, d=d, xlab=xlab, ylab=ylab, zlab=zlab, col=col[facetcol], border=border, ticktype=ticktype, ...) } else if (disp=="slice") { if (!add) plot(Fhat$x[,1], Fhat$x[,2], type="n", xlab=xlab, ylab=ylab, ...) if (missing(abs.cont)) hts <- cont/100 else hts <- abs.cont if (missing(col)) col <- col.fun(n=length(hts)) if (length(col)0) contour(Fhat$eval.points[[1]], Fhat$eval.points[[2]], Fhat$estimate*scale, level=hts[i]*scale, add=TRUE, drawlabels=drawlabels, col=col[i], lwd=lwd, labcex=labcex, ...) } ## add points if (drawpoints) points(Fhat$x[,1], Fhat$x[,2], col=col.pt, cex=cex, pch=pch) } ## image plot else if (disp=="image") { if (missing(col)) col <- col.fun(100) col <- transparency.col(col, alpha=alpha) image(Fhat$eval.points[[1]], Fhat$eval.points[[2]], Fhat$estimate, xlab=xlab, ylab=ylab, add=add, col=col, ...) box() } else if (disp=="filled.contour") { hts <- cont/100 clev <- c(-0.01*max(abs(Fhat$estimate)), hts, max(c(Fhat$estimate, hts)) + 0.01*max(abs(Fhat$estimate))) if (missing(col)) col <- col.fun(length(hts)) col <- transparency.col(col, alpha=alpha) if (!add) plot(Fhat$eval.points[[1]], Fhat$eval.points[[2]], type="n", xlab=xlab, ylab=ylab, ...) if (tail(hts, n=1) < max(Fhat$estimate)) hts2 <- c(hts, max(Fhat$estimate)) .filled.contour(Fhat$eval.points[[1]], Fhat$eval.points[[2]], Fhat$estimate, levels=hts2, col=col) if (!missing(lwd)) { for (i in 1:length(hts)) { if (missing(abs.cont)) scale <- (100-cont[i])/hts[i] else scale <- 1 if (lwd >=1) contour(Fhat$eval.points[[1]], Fhat$eval.points[[2]], Fhat$estimate*scale, level=hts[i]*scale, add=TRUE, drawlabels=drawlabels, col=1, lwd=lwd, labcex=labcex, ...) } } } if (disp=="persp") invisible(plotret) else invisible() } plotkcde.3d <- function(Fhat, display="plot3D", cont=c(25,50,75), colors, col, alphavec, size=3, cex=1, pch=1, theta=-30, phi=40, d=4, ticktype="detailed", bty="f", col.pt=4, add=FALSE, xlab, ylab, zlab, drawpoints=FALSE, alpha, box=TRUE, axes=TRUE, ...) { disp <- match.arg(display, c("plot3D", "rgl")) hts <- sort(cont/100) nc <- length(hts) if (missing(col)) { col.fun <- function(n) { hcl.colors(n, palette="viridis") } col <- col.fun(n=length(hts)) } colors <- col if (missing(xlab)) xlab <- Fhat$names[1] if (missing(ylab)) ylab <- Fhat$names[2] if (missing(zlab)) zlab <- Fhat$names[3] if (missing(alphavec)) alphavec <- seq(0.5,0.1,length=nc) if (missing(alpha)) alpha <- 0.5 if (!missing(alpha)) { alphavec <- rep(alpha,nc) } disp <- match.arg(display, c("plot3D", "rgl")) if (disp %in% "plot3D") { if (!requireNamespace("plot3D", quietly=TRUE)) stop("Install the plot3D package as it is required.", call.=FALSE) for (i in 1:nc) if (hts[nc-i+1] < max(Fhat$estimate)) plot3D::isosurf3D(x=Fhat$eval.points[[1]], y=Fhat$eval.points[[2]], z=Fhat$eval.points[[3]], colvar=Fhat$estimate, level=hts[nc-i+1], add=add | (i>1), col=colors[nc-i+1], alpha=alphavec[i], phi=phi, theta=theta, xlab=xlab, ylab=ylab, zlab=zlab, d=d, ticktype=ticktype, bty=bty, ...) if (drawpoints) plot3D::points3D(x=Fhat$x[,1], y=Fhat$x[,2], z=Fhat$x[,3], cex=cex, col=col.pt, add=TRUE, pch=pch, d=d) } else if (disp %in% "rgl") { ## suggestions from Viktor Petukhov 08/03/2018 if (!requireNamespace("rgl", quietly=TRUE)) stop("Install the rgl package as it is required.", call.=FALSE) if (!requireNamespace("misc3d", quietly=TRUE)) stop("Install the misc3d package as it is required.", call.=FALSE) if (drawpoints) rgl::plot3d(Fhat$x[,1],Fhat$x[,2],Fhat$x[,3], size=size, col=col.pt, alpha=alpha, xlab=xlab, ylab=ylab, zlab=zlab, add=add, box=FALSE, axes=FALSE, ...) else rgl::plot3d(Fhat$x[,1],Fhat$x[,2],Fhat$x[,3], type="n", xlab=xlab, ylab=ylab, zlab=zlab, add=add, box=FALSE, axes=FALSE, ...) rgl::bg3d(col="white") for (i in 1:nc) if (hts[nc-i+1] < max(Fhat$estimate)) misc3d::contour3d(Fhat$estimate, level=hts[nc-i+1], x=Fhat$eval.points[[1]], y=Fhat$eval.points[[2]], z=Fhat$eval.points[[3]], add=TRUE, color=colors[nc-i+1], alpha=alphavec[i], box=FALSE, axes=FALSE, ...) if (box) rgl::box3d() if (axes) rgl::axes3d(c("x","y","z")) } } ##################################################################### ## Bandwidth selectors for KCDE ##################################################################### ## Normal scale bandwidth selectors hns.kcde <- function(x) { d <- 1 n <- length(x) sigma <- sd(x) hns <- 4^(1/3)*sigma*n^(-1/3) return(hns) } Hns.kcde <- function(x) { if (is.vector(x)) { return(hns.kcde(x)^2) } d <- ncol(x) n <- nrow(x) m1 <- (4*pi)^(-1/2) Jd <- matrix(1, ncol=d, nrow=d) Sigma <- var(x) Hns <- (4*det(Sigma)^(1/2)*tr(matrix.sqrt(Sigma))/tr(Sigma))^(2/3)*Sigma*n^(-2/3) return(Hns) } ## Plug-in bandwidth selector hpi.kcde <- function(x, nstage=2, binned, amise=FALSE) { n <- length(x) d <- 1 if (missing(binned)) binned <- default.bflag(d,n) K2 <- dnorm.deriv(x=0, mu=0, sigma=1, deriv.order=2) K4 <- dnorm.deriv(x=0, mu=0, sigma=1, deriv.order=4) m2 <- 1 m1 <- (4*pi)^(-1/2) ## formula for bias annihliating bandwidths from Wand & Jones (1995, p.70) if (nstage==2) { psi6.hat <- psins.1d(r=6, sigma=sd(x)) gamse4 <- (2*K4/(-m2*psi6.hat*n))^(1/(4+3)) psi4.hat <- kfe.1d(x=x, g=gamse4, deriv.order=4, inc=1, binned=binned) gamse2 <- (2*K2/(-m2*psi4.hat*n))^(1/(2+3)) psi2.hat <- kfe.1d(x=x, g=gamse2, deriv.order=2, inc=1, binned=binned) } else { psi4.hat <- psins.1d(r=4, sigma=sd(x)) gamse2 <- (2*K2/(-m2*psi4.hat*n))^(1/(2+3)) psi2.hat <- kfe.1d(x=x, g=gamse2, deriv.order=2, inc=1, binned=binned) } ## formula form Polansky & Baker (2000) h <- (2*m1/(-m2^2*psi2.hat*n))^(1/3) if (amise) PI <- -2*n^(-1)*m1*h - 1/4*psi2.hat*h^4 if (!amise) return(h) else return(list(h=h, PI=PI)) } Hpi.kcde <- function(x, nstage=2, pilot, Hstart, binned, bgridsize, amise=FALSE, verbose=FALSE, optim.fun="optim", pre=TRUE) { n <- nrow(x) d <- ncol(x) m1 <- (4*pi)^(-1/2) Jd <- matrix(1, ncol=d, nrow=d) if (missing(binned)) binned <- default.bflag(d,n) if (!is.matrix(x)) x <- as.matrix(x) if (missing(pilot)) pilot <- "dunconstr" pilot1 <- match.arg(pilot, c("dunconstr", "dscalar")) if (pre) { S12 <- diag(sqrt(diag(var(x)))); x <- pre.scale(x) } D2K0 <- t(dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=2)) if (nstage==2) { ## stage 1 psi4.ns <- psins(r=4, Sigma=var(x), deriv.vec=TRUE) amse2.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) Hinv <- chol2inv(chol(H)) Hinv12 <- matrix.sqrt(Hinv) amse2.val <- 1/(det(H)^(1/2)*n)*((Hinv12 %x% Hinv12) %*% D2K0) + 1/2* t(vec(H) %x% diag(d^2)) %*% psi4.ns return(sum(amse2.val^2)) } Hstart2 <- matrix.sqrt(Gns(r=2, n=n, Sigma=var(x))) optim.fun1 <- match.arg(optim.fun, c("nlm", "optim")) if (optim.fun1=="nlm") { result <- nlm(p=vech(Hstart2), f=amse2.temp, print.level=2*as.numeric(verbose)) H2 <- invvech(result$estimate) %*% invvech(result$estimate) } else { result <- optim(vech(Hstart2), amse2.temp, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) H2 <- invvech(result$par) %*% invvech(result$par) } psi2.hat <- kfe(x=x, G=H2, deriv.order=2, add.index=FALSE, binned=binned, bgridsize=bgridsize, verbose=verbose) } else { psi2.hat <- psins(r=2, Sigma=var(x), deriv.vec=TRUE) H2 <- Gns(r=2, n=n, Sigma=var(x)) } if (missing(Hstart)) Hstart <- Hns.kcde(x=x) ## stage 2 amise.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) H12 <- matrix.sqrt(H) amise.val <- -2*n^(-1)*m1*tr(H12) - 1/4*t(vec(H %*% H)) %*% psi2.hat return(drop(amise.val)) } Hstart <- matrix.sqrt(Hstart) optim.fun1 <- match.arg(optim.fun, c("optim", "nlm")) if (optim.fun1=="nlm") { result <- nlm(p=vech(Hstart), f=amise.temp, print.level=2*as.numeric(verbose)) H <- invvech(result$estimate) %*% invvech(result$estimate) amise.star <- result$minimum } else { result <- optim(vech(Hstart), amise.temp, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) H <- invvech(result$par) %*% invvech(result$par) amise.star <- result$value } if (pre) H <- S12 %*% H %*% S12 if (amise) H <- list(H=H, PI=amise.star) return(H) } Hpi.diag.kcde <- function(x, nstage=2, pilot, Hstart, binned=FALSE, bgridsize, amise=FALSE, verbose=FALSE, optim.fun="optim", pre=TRUE) { n <- nrow(x) d <- ncol(x) m1 <- (4*pi)^(-1/2) Jd <- matrix(1, ncol=d, nrow=d) if (missing(binned)) binned <- default.bflag(d,n) if(!is.matrix(x)) x <- as.matrix(x) if (missing(pilot)) pilot <- "dscalar" pilot <- match.arg(pilot, c("dunconstr", "dscalar")) if (pre) { S12 <- diag(sqrt(diag(var(x)))); x <- pre.scale(x) } D2K0 <- t(dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=2)) if (nstage==2) { ## stage 1 psi4.ns <- psins(r=4, Sigma=var(x), deriv.vec=TRUE) amse2.temp <- function(diagH) { H <- diag(diagH) %*% diag(diagH) Hinv <- chol2inv(chol(H)) Hinv12 <- matrix.sqrt(Hinv) amse2.val <- 1/(det(H)^(1/2)*n)*((Hinv12 %x% Hinv12) %*% D2K0) + 1/2* t(vec(H) %x% diag(d^2)) %*% psi4.ns return(sum(amse2.val^2)) } Hstart2 <- matrix.sqrt(Gns(r=2, n=n, Sigma=var(x))) optim.fun <- match.arg(optim.fun, c("optim", "nlm")) if (optim.fun=="nlm") { result <- nlm(p=diag(Hstart2), f=amse2.temp, print.level=2*as.numeric(verbose)) H2 <- diag(result$estimate) %*% diag(result$estimate) } else { result <- optim(diag(Hstart2), amse2.temp, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) H2 <- diag(result$par) %*% diag(result$par) } psi2.hat <- kfe(x=x, G=H2, deriv.order=2, add.index=FALSE, binned=binned, bgridsize=bgridsize, verbose=verbose) } else psi2.hat <- psins(r=2, Sigma=var(x), deriv.vec=TRUE) if (missing(Hstart)) Hstart <- Hns.kcde(x=x) ## stage 2 amise.temp <- function(diagH) { H <- diag(diagH) %*% diag(diagH) H12 <- matrix.sqrt(H) amise.val <- -2*n^(-1)*m1*tr(H12) - 1/4*t(vec(H %*% H)) %*% psi2.hat return(drop(amise.val)) } Hstart <- matrix.sqrt(Hstart) optim.fun1 <- match.arg(optim.fun, c("optim", "nlm")) if (optim.fun1=="nlm") { result <- nlm(p=diag(Hstart), f=amise.temp, print.level=2*as.numeric(verbose)) H <- diag(result$estimate) %*% diag(result$estimate) amise.star <- result$minimum } else { result <- optim(diag(Hstart), amise.temp, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) H <- diag(result$par) %*% diag(result$par) amise.star <- result$value } if (pre) H <- S12 %*% H %*% S12 if (amise) H <- list(H=H, PI=amise.star) return(H) } ##################################################################### ## Multivariate kernel ROC estimators ##################################################################### kroc <- function(x1, x2, H1, h1, hy, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, verbose=FALSE) { if (is.vector(x1)) { d <- 1; n1 <- length(x1) } else { d <- ncol(x1); n1 <- nrow(x1); x1 <- as.matrix(x1); x2 <- as.matrix(x2) } if (!missing(eval.points)) stop("eval.points in kroc not yet implemented") if (d==1) { if (missing(h1)) h1 <- hpi.kcde(x=x1, binned=default.bflag(d=d, n=n1)) Fhatx1 <- kcde(x=x1, h=h1, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, binned=binned, bgridsize=bgridsize, positive=positive, adj.positive=adj.positive, w=w, tail.flag="upper.tail") } else { if (missing(H1)) H1 <- Hpi.kcde(x=x1, binned=default.bflag(d=d, n=n1), verbose=verbose) Fhatx1 <- kcde(x=x1, H=H1, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, binned=binned, bgridsize=bgridsize, w=w, tail.flag="upper.tail", verbose=verbose) } ## transform from [0,1] to reals y1 <- predict(Fhatx1, x=x1) y2 <- predict(Fhatx1, x=x2) y1 <- qnorm(y1[y1>0]) y2 <- qnorm(y2[y2>0]) if (missing(hy)) hy <- hpi.kcde(y2, binned=default.bflag(d=1, n=n1)) Fhaty2 <- kcde(x=y2, h=hy, binned=TRUE, xmin=min(y1,y2)-3.7*hy, xmax=max(y1,y2)+3.7*hy) Fhaty1 <- kcde(x=y1, h=hy, binned=TRUE, xmin=min(y1,y2)-3.7*hy, xmax=max(y1,y2)+3.7*hy) Fhaty1$eval.points <- pnorm(Fhaty1$eval.points) Fhaty2$eval.points <- pnorm(Fhaty2$eval.points) Rhat <- Fhaty1 Rhat$eval.points <- Fhaty1$estimate Rhat$estimate <- Fhaty2$estimate if (d==1) { Rhat$h1 <- h1; Rhat$H1 <- h1^2; Rhat$hy <- hy } else { Rhat$H1 <- H1; Rhat$hy <- hy } ## Use spline to smooth out transformed ROC curve Rhat.smoothed <- smooth.spline(Rhat$eval.points, Rhat$estimate, spar=0.5) Rhat.smoothed <- predict(Rhat.smoothed, x=seq(0,1,length=length(Rhat$eval.points))) Rhat$eval.points <- Rhat.smoothed$x Rhat$estimate <- Rhat.smoothed$y ## add (0,0) and (1,1) as endpoints if (head(Rhat$eval.points, n=1)!=0) Rhat$eval.points[1] <- 0 if (head(Rhat$estimate, n=1)!=0) Rhat$estimate[1] <- 0 if (tail(Rhat$eval.points, n=1)!=1) Rhat$eval.points[length(Rhat$eval.points)] <- 1 if (tail(Rhat$estimate, n=1)!=1) Rhat$estimate[length(Rhat$estimate)] <- 1 Rhat$estimate[Rhat$estimate>1] <- 1 Rhat$estimate[Rhat$estimate<0] <- 0 Rhat$indices <- indices.kroc(Rhat) Rhat <- Rhat[-c(4,5)] Rhat$type <- "kroc" class(Rhat) <- "kroc" return(Rhat) } ## summary measure of ROC curves indices.kroc <- function(Rhat) { auc <- sum(abs((head(Rhat$estimate, n=-1) - tail(Rhat$estimate, n=-1)))*abs(diff(Rhat$eval.points))/2 + head(Rhat$estimate, n=-1)*abs(diff(Rhat$eval.points))) youden.val <- Rhat$estimate - Rhat$eval.points if (max(youden.val)>0.001) { youden.ind <- which.max(youden.val) youden <- youden.val[youden.ind] LR <- list(minus=(1-Rhat$estimate[youden.ind])/(1-Rhat$eval.points[youden.ind]), plus=Rhat$estimate[youden.ind]/Rhat$eval.points[youden.ind]) } else LR <- list(minus=1, plus=1) return(list(auc=auc, youden=max(youden.val), LR=LR)) } ############################################################################# ## S3 methods ############################################################################# ## plot method plot.kroc <- function(x, add=FALSE, add.roc.ref=FALSE, xlab, ylab, alpha=1, col=1, ...) { Rhat <- x col <- transparency.col(col, alpha=alpha) if (missing(ylab)) ylab <- "True positive rate (sensitivity)" if (missing(xlab)) xlab <- expression("False positive rate"~~group("(", list(bar(specificity)), ")")) if (add) lines(Rhat$eval.points, Rhat$estimate, ...) else plot(Rhat$eval.points, Rhat$estimate, type="l", ylab=ylab, xlab=xlab, col=col, ...) if (is.vector(Rhat$x[[1]])) d <- 1 else d <- ncol(Rhat$x[[1]]) if (add.roc.ref) { z <- seq(0,1, length=401) kind <- 0:(d-1) roc.indep <- 0 for (k in kind) roc.indep <- roc.indep + z*(-log(z))^k/factorial(k) lines(z, roc.indep, lty=2, col="grey") } } ## summary method summary.kroc <- function(object, ...) { cat("Summary measures for ROC curve\nAUC =", signif(object$indices$auc, ...), "\n") cat("Youden index =", signif(object$indices$youden, ...), "\n") cat(paste("(LR-, LR+) = (", signif(object$indices$LR$minus, ...), ", ", signif(object$indices$LR$plus, ...),")\n\n",sep="")) } ## predict methods predict.kcde <- function(object, ..., x) { return(predict.kde(object=object, ..., x=x)) } predict.kroc <- function(object, ..., x) { return(predict.kde(object=object, ..., x=x)) } ## contourLevels method contourLevels.kcde <- function(x, prob, cont, nlevels=5, ...) { fhat <- x if (missing(prob) & missing(cont)) hts <- pretty(fhat$estimate, n=nlevels) if (!missing(prob) & missing(cont)) { hts <- prob/100; names(hts) <- paste0(prob, "%") } if (missing(prob) & !missing(cont)) { prob <- 100-cont; hts <- prob/100; names(hts) <- paste0(prob, "%") } return(hts) } ks/R/kde-test.R0000644000176200001440000003651514673143073012755 0ustar liggesusers############################################################################## ## Test statistic for multivariate 2-sample test ############################################################################## kde.test <- function(x1, x2, H1, H2, h1, h2, psi1, psi2, var.fhat1, var.fhat2, binned=FALSE, bgridsize, verbose=FALSE) { ## default values ksd <- ks.defaults(x=x1, binned=binned, bgridsize=bgridsize) d <- ksd$d binned <- ksd$binned bgridsize <- ksd$bgridsize gridsize <- ksd$gridsize if (is.vector(x1) & is.vector(x2)) return (kde.test.1d(x1=x1, x2=x2, h1=h1, h2=h2, psi1=psi1, psi2=psi2, var.fhat1=var.fhat1, var.fhat2=var.fhat2, binned=binned, bgridsize=bgridsize, verbose=verbose)) if (!is.matrix(x1)) x1 <- as.matrix(x1) if (!is.matrix(x2)) x2 <- as.matrix(x2) n1 <- nrow(x1) n2 <- nrow(x2) d <- ncol(x1) K0 <- drop(dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=0)) S1 <- var(x1) S2 <- var(x2) ## default bandwidths if (missing(H1)) H1 <- Hpi.kfe(x1, deriv.order=0, binned=default.bflag(d=d,n=n1), bgridsize=bgridsize) if (missing(H2)) H2 <- Hpi.kfe(x2, deriv.order=0, binned=default.bflag(d=d,n=n2), bgridsize=bgridsize) ## kernel estimation for components of test statistic if (missing(psi1)) psi1 <- Qr(x=x1, y=x1, Sigma=H1, verbose=verbose) if (missing(psi2)) psi2 <- Qr(x=x2, y=x2, Sigma=H2, verbose=verbose) if (missing(var.fhat1)) { H1.r1 <- Hns(x=x1, deriv.order=1) fhat1.r1 <- predict(kdde(x=x1, H=H1.r1, deriv.order=1), x=apply(x1, 2, mean)) var.fhat1 <- drop(fhat1.r1 %*% S1 %*% fhat1.r1) } psi12 <- Qr(x=x1, y=x2, Sigma=H1, verbose=verbose) if (missing(var.fhat2)) { H2.r1 <- Hns(x=x2, deriv.order=1) fhat2.r1 <- predict(kdde(x=x2, H=H2.r1, deriv.order=1), x=apply(x2, 2, mean)) var.fhat2 <- drop(fhat2.r1 %*% S2 %*% fhat2.r1) } psi21 <- Qr(x=x2, y=x1, Sigma=H2, verbose=verbose) ## test statistic + its parameters T.hat <- drop(psi1 + psi2 - (psi12 + psi21)) muT.hat <- (n1^(-1)*det(H1)^(-1/2) + n2^(-1)*det(H2)^(-1/2))*K0 varT.hat <- 3*(n1*var.fhat1 + n2*var.fhat2)/(n1+n2) *(1/n1+1/n2) zstat <- (T.hat-muT.hat)/sqrt(varT.hat) pval <- 1-pnorm(zstat) if (length(pval==0)>0) pval[pval==0] <- pnorm(-abs(zstat[pval==0])) val <- list(Tstat=T.hat, zstat=zstat, pvalue=pval, mean=muT.hat, var=varT.hat, var.fhat1=var.fhat1, var.fhat2=var.fhat2, n1=n1, n2=n2, H1=H1, H2=H2, psi1=psi1, psi12=psi12, psi21=psi21, psi2=psi2) return(val) } kde.test.1d <- function(x1, x2, h1, h2, psi1, psi2, var.fhat1, var.fhat2, binned=FALSE, bgridsize, verbose=FALSE) { n1 <- length(x1) n2 <- length(x2) d <- 1 K0 <- dnorm.deriv(x=0, mu=0, sigma=1, deriv.order=0) s1 <- sd(x1) s2 <- sd(x2) ## kernel estimation for components of test statistic if (missing(h1)) h1 <- hpi.kfe(x1, nstage=2, deriv.order=0, binned=binned, bgridsize=bgridsize) if (missing(h2)) h2 <- hpi.kfe(x2, nstage=2, deriv.order=0, binned=binned, bgridsize=bgridsize) if (missing(psi1)) psi1 <- Qr.1d(x=x1, y=x1, sigma=h1, verbose=verbose) if (missing(psi2)) psi2 <- Qr.1d(x=x2, y=x2, sigma=h2, verbose=verbose) if (missing(var.fhat1)) { h1.r1 <- hns(x=x1, deriv.order=1) fhat1.r1 <- predict(kdde(x=x1, h=h1.r1, deriv.order=1), x=mean(x1)) var.fhat1 <- fhat1.r1^2*s1^2 } psi12 <- Qr.1d(x=x1, sigma=h1, y=x2, verbose=verbose) if (missing(var.fhat2)) { h2.r1 <- hns(x=x2, deriv.order=1) fhat2.r1 <- predict(kdde(x=x2, h=h2.r1, deriv.order=1), x=mean(x2)) var.fhat2 <- fhat2.r1^2*s2^2 } psi21 <- Qr.1d(x=x2, sigma=h2, y=x1, verbose=verbose) ## test statistic + its parameters T.hat <- drop(psi1 + psi2 - (psi12 + psi21)) muT.hat <- ((n1*h1)^(-1) + (n2*h2)^(-1))*K0 varT.hat <- 3*(n1*var.fhat1 + n2*var.fhat2)/(n1+n2) *(1/n1+1/n2) zstat <- (T.hat-muT.hat)/sqrt(varT.hat) pval <- 1-pnorm(zstat) if (length(pval==0)>0) pval[pval==0] <- pnorm(-abs(zstat[pval==0])) val <- list(Tstat=T.hat, zstat=zstat, pvalue=pval, mean=muT.hat, var=varT.hat, var.fhat1=var.fhat1, var.fhat2=var.fhat2, n1=n1, n2=n2, h1=h1, h2=h2, psi1=psi1, psi12=psi12, psi21=psi21, psi2=psi2) return(val) } ############################################################################### ## Local kde test ############################################################################### ## multivariate local test kde.local.test <- function(x1, x2, H1, H2, h1, h2, fhat1, fhat2, gridsize, binned, bgridsize, verbose=FALSE, supp=3.7, mean.adj=FALSE, signif.level=0.05, min.ESS, xmin, xmax) { ## default values ksd <- ks.defaults(x=x1, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d if (missing(binned)) binned <- ksd$binned if (missing(bgridsize)) bgridsize <- ksd$bgridsize if (missing(gridsize)) gridsize <- ksd$gridsize ## clip data to xmin, xmax grid limits for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] x1 <- truncate.grid(x=x1, xmin=xmin, xmax=xmax) x2 <- truncate.grid(x=x2, xmin=xmin, xmax=xmax) } if (is.vector(x1) & is.vector(x2)) { return(kde.local.test.1d(x1=x1, x2=x2, h1=h1, h2=h2, fhat1=fhat1, fhat2=fhat2, gridsize=gridsize, binned=binned, bgridsize=bgridsize, verbose=verbose, supp=supp, mean.adj=mean.adj, xmin=xmin, xmax=xmax)) } if (missing(H1) & !missing(x1)) H1 <- Hpi(x=x1, deriv.order=0, binned=default.bflag(d=ncol(x1), n=nrow(x1)), bgridsize=bgridsize, verbose=verbose) if (missing(H2) & !missing(x2)) H2 <- Hpi(x=x2, deriv.order=0, binned=default.bflag(d=ncol(x2), n=nrow(x2)), bgridsize=bgridsize, verbose=verbose) if (!missing(x1) & !missing(x2)) { n1 <- nrow(x1) n2 <- nrow(x2) d <- ncol(x1) RK <- (4*pi)^(-d/2) xrange <- apply(rbind(x1,x2), 2, range) tol.H <- sqrt(prod(diag(matrix.sqrt(H1) %*% matrix.sqrt(H2)))^(1/d)) if (missing(xmin)) xmin <- xrange[1,] - supp*tol.H if (missing(xmax)) xmax <- xrange[2,] + supp*tol.H } else { n1 <- nrow(fhat1$x) n2 <- nrow(fhat2$x) d <- ncol(fhat1$x) RK <- (4*pi)^(-d/2) H1 <- fhat1$H H2 <- fhat2$H } ## kernel estimation for components of test statistic if (missing(fhat1)) fhat1 <- kde(x=x1, H=H1, binned=binned, gridsize=gridsize, bgridsize=bgridsize, supp=supp, xmin=xmin, xmax=xmax, verbose=verbose) if (missing(fhat2)) fhat2 <- kde(x=x2, H=H2, binned=binned, gridsize=gridsize, bgridsize=bgridsize, supp=supp, xmin=xmin, xmax=xmax, verbose=verbose) HD2fhat <- 0 if (mean.adj) { D2fhat1 <- kdde(x=x1, H=H1, binned=binned, bgridsize=bgridsize, supp=supp, xmin=xmin, xmax=xmax, deriv.order=2, verbose=verbose) D2fhat2 <- kdde(x=x2, H=H2, binned=binned, bgridsize=bgridsize, supp=supp, xmin=xmin, xmax=xmax, deriv.order=2, verbose=verbose) D2fhat <- list() vH1 <- vec(H1) vH2 <- vec(H2) for (j in 1:length(D2fhat1$estimate)) D2fhat$estimate[[j]] <- vH1[j]*D2fhat1$estimate[[j]] - vH2[j]*D2fhat2$estimate[[j]] for (j in 1:length(D2fhat)) HD2fhat <- HD2fhat + D2fhat$estimate[[j]] } fhat.diff <- fhat1$estimate - fhat2$estimate - 1/2*HD2fhat var.fhat.diff <- (n1^(-1)*det(H1)^(-1/2)*fhat1$estimate + n2^(-1)*det(H2)^(-1/2)*fhat2$estimate)*RK X2 <- fhat.diff^2/var.fhat.diff pvalue <- 1 - pchisq(X2, 1) pvalue[is.na(pvalue)] <- 0 ## apply Hochberg multiple test adjustment gridsize <- sapply(fhat1$eval.points, length) fhat.diff.signif <- hochberg.mult.test(pvalue=pvalue, gridsize=gridsize, signif.level=signif.level) fhat.diff.pos <- fhat1 fhat.diff.neg <- fhat2 fhat.diff.pos$estimate <- fhat.diff.signif*(fhat.diff>0) fhat.diff.neg$estimate <- fhat.diff.signif*(fhat.diff<0) if (!missing(min.ESS)) { ESS1 <- n1*fhat1$estimate/dmvnorm(rep(0,d), rep(0,d), H1) ESS2 <- n2*fhat2$estimate/dmvnorm(rep(0,d), rep(0,d), H2) ESS <- pmin(ESS1, ESS2) >= min.ESS fhat.diff.pos$estimate <- fhat.diff.pos$estimate*ESS fhat.diff.neg$estimate <- fhat.diff.neg$estimate*ESS } result <- list(fhat1=fhat1, fhat2=fhat2, X2=X2, pvalue=pvalue, fhat.diff=fhat.diff, mean.fhat.diff=HD2fhat, var.fhat.diff=var.fhat.diff, fhat.diff.pos=fhat.diff.pos, fhat.diff.neg=fhat.diff.neg, n1=n1, n2=n2, H1=H1, H2=H2, names=parse.name(x1)) class(result) <- "kde.loctest" return(result) } ## 1-d local test kde.local.test.1d <- function(x1, x2, h1, h2, fhat1, fhat2, gridsize=gridsize, binned=FALSE, bgridsize, verbose=FALSE, supp=3.7, mean.adj=FALSE, signif.level=0.05, min.ESS, xmin, xmax) { if (missing(h1) & !missing(x1)) h1 <- hpi(x=x1, nstage=2, binned=binned, bgridsize=bgridsize) if (missing(h2) & !missing(x2)) h2 <- hpi(x=x2, nstage=2, binned=binned, bgridsize=bgridsize) if (!missing(x1) & !missing(x2)) { n1 <- length(x1) n2 <- length(x2) d <- 1 RK <- (4*pi)^(-d/2) xrange <- range(c(x1,x2)) if (missing(xmin)) xmin <- xrange[1] - supp*sqrt(h1*h2) if (missing(xmax)) xmax <- xrange[2] + supp*sqrt(h1*h2) } else { n1 <- length(fhat1$x) n2 <- length(fhat2$x) d <- 1 RK <- (4*pi)^(-d/2) h1 <- fhat1$h h2 <- fhat2$h } ## kernel estimation for components of test statistic if (missing(fhat1)) fhat1 <- kde(x=x1, h=h1, gridsize=gridsize, binned=binned, bgridsize=bgridsize, supp=supp, xmin=xmin, xmax=xmax) if (missing(fhat2)) fhat2 <- kde(x=x2, h=h2, gridsize=gridsize, binned=binned, bgridsize=bgridsize, supp=supp, xmin=xmin, xmax=xmax) h2D2fhat <- 0 if (mean.adj) { D2fhat1 <- kdde(x=x1, h=h1, gridsize=gridsize, binned=binned, bgridsize=bgridsize, supp=supp, xmin=xmin, xmax=xmax, deriv.order=2, verbose=verbose) D2fhat2 <- kdde(x=x2, h=h2, gridsize=gridsize, binned=binned, bgridsize=bgridsize, supp=supp, xmin=xmin, xmax=xmax, deriv.order=2, verbose=verbose) h2D2fhat <- (h1^2*D2fhat1$estimate - h2^2*D2fhat2$estimate) } fhat.diff <- fhat1$estimate - fhat2$estimate - 1/2*h2D2fhat var.fhat.diff <- ((n1*h1)^(-1)*fhat1$estimate + (n2*h2)^(-1)*fhat2$estimate)*RK X2 <- fhat.diff^2/var.fhat.diff pvalue <- 1 - pchisq(X2, 1) gridsize <- length(fhat1$eval.points) fhat.diff.signif <- hochberg.mult.test(pvalue=pvalue, gridsize=gridsize, signif.level=signif.level) fhat.diff.pos <- fhat1 fhat.diff.neg <- fhat2 fhat.diff.pos$estimate <- fhat.diff.signif*(fhat.diff>0) fhat.diff.neg$estimate <- fhat.diff.signif*(fhat.diff<0) if (!missing(min.ESS)) { ESS1 <- n1*fhat1$estimate/dnorm(0, 0, h1) ESS2 <- n2*fhat2$estimate/dnorm(0, 0, h2) ESS <- pmin(ESS1, ESS2) >= min.ESS fhat.diff.pos$estimate <- fhat.diff.pos$estimate*ESS fhat.diff.neg$estimate <- fhat.diff.neg$estimate*ESS } result <- list(fhat1=fhat1, fhat2=fhat2, chisq=X2, pvalue=pvalue, fhat.diff=fhat.diff, mean.fhat.diff=h2D2fhat, var.fhat.diff=var.fhat.diff, n1=n1, n2=n2, h1=h1, h2=h2, H1=h1^2, H2=h2^2, fhat.diff=fhat.diff, fhat.diff.pos=fhat.diff.pos, fhat.diff.neg=fhat.diff.neg, names=parse.name(x1)) class(result) <- "kde.loctest" return(result) } ## Hochberg (1988) adjustment for multiple correlated tests hochberg.mult.test <- function(pvalue, gridsize, signif.level) { pvalue.ord <- pvalue[order(pvalue)] num.test <- sum(!is.na(pvalue.ord)) if (num.test>=1) num.test.seq <- c(1:num.test, rep(NA, prod(gridsize) - num.test)) else num.test.seq <- rep(NA, prod(gridsize)) reject.nonzero <- ((pvalue.ord <= signif.level/(num.test + 1 - num.test.seq)) & (pvalue.ord > 0)) reject.nonzero.ind <- which(reject.nonzero) ## p-value == 0 => reject null hypotheses automatically fhat.diff.signif <- array(FALSE, dim=gridsize) fhat.diff.signif[which(pvalue==0, arr.ind=TRUE)] <- TRUE ## p-value > 0 then reject null hypotheses indicated in reject.nonzero.ind for (i in reject.nonzero.ind) fhat.diff.signif[which(pvalue==pvalue.ord[i], arr.ind=TRUE)] <- TRUE return(fhat.diff.signif) } ############################################################################# ## S3 method for kde.loctest objects ############################################################################# ## plot methods plot.kde.loctest <- function(x, ...) { fhat <- x if (is.vector(fhat$fhat1$x)) plotkde.loctest.1d(fhat, ...) else { d <- ncol(fhat$fhat1$x) if (d==2) plotkde.loctest.2d(x, ...) else if (d==3) plotkde.loctest.3d(x, ...) else stop("Plot function only available for 1, 2 or 3-d data") } } plotkde.loctest.1d <- function(x, lcol, col, add=FALSE, xlab, ylab, rugsize, add.legend=TRUE, pos.legend="topright", alpha=1, ...) { if (missing(xlab)) xlab <- x$fhat.diff.pos$names[1] if (missing(ylab)) ylab <- expression("Density difference "*f[1]-f[2]) if (missing(col)) col <- hcl.colors(palette="Dark2",2) col <- transparency.col(col, alpha=alpha) if (missing(lcol)) lcol <- 1 if (!add) plot(x$fhat1$eval.points, x$fhat.diff, type="l", ylab=ylab, xlab=xlab, col=lcol, ...) else lines(x$fhat1$eval.points, x$fhat.diff, col=lcol, ...) plot.lim <- par()$usr if (missing(rugsize)) rugsize <- abs(plot.lim[4]-plot.lim[3])/50 try(image(x$fhat.diff.pos$eval, c(plot.lim[3], plot.lim[3]+rugsize), cbind(x$fhat.diff.pos$estimate==1, x$fhat.diff.pos$estimate==1), level=0.5, add=TRUE, col=c("transparent", col[1]), ...)) try(image(x$fhat.diff.neg$eval, c(plot.lim[3], plot.lim[3]+rugsize), cbind(x$fhat.diff.neg$estimate==1, x$fhat.diff.neg$estimate==1), level=0.5, add=TRUE, col=c("transparent", col[2]), ...)) if (add.legend) legend(pos.legend, legend=c(expression(f[1]>f[2]), expression(f[1]f[2]), expression(f[1]f[2]), expression(f[1]=4) gridsize <- rep(21, d) return(gridsize) } default.bgridsize <- function(d) { if (d==1) gridsize <- 401 else if (d==2) gridsize <- rep(151,d) else if (d==3) gridsize <- rep(31, d) else if (d==4) gridsize <- rep(15, d) else gridsize <- NA return(gridsize) } default.bflag <- function(d, n) { if (d==1) thr <- 1 else if (d==2) thr <- 500 else if (d>2) thr <- 1000 bf <- n>thr return(bf) } ## truncate x to xmin, xmax grid truncate.grid <- function(x, y, xmin, xmax) { if (is.vector(x)) { d <- 1; n <- length(x); xvec <- TRUE; x <- matrix(x, ncol=1) } else { d <- ncol(x); n <- nrow(x); xvec <- FALSE } xind <- rep(TRUE, n) if (!(missing(xmax))) { if (d==1) msgmax <- xmax else msgmax <- paste0("c(", paste0(xmax, collapse=","), ")") if (any(sweep(x, 2, FUN=">", xmax))) warning(paste0("Points in x greater than xmax=", msgmax, " have been excluded.")) for (i in 1:d) xind <- xind & x[,i]<=xmax[i] } if (!(missing(xmin))) { if (d==1) msgmin <- xmin else msgmin <- paste0("c(", paste0(xmin, collapse=","), ")") if (any(sweep(x, 2, FUN="<", xmin))) warning(paste0("Points in x less than xmin=", msgmin, " have been excluded.")) for (i in 1:d) xind <- xind & x[,i]>=xmin[i] } if (d==1) { x <- x[xind,, drop=xvec] ## special case for x is 1-col matrix to force x to be ## 1-col matrix, as required for eks <= 1.0.1 } else { x <- x[xind,, drop=FALSE] } if (!missing(y)) { y <- y[xind]; x <- list(x=x, y=y) } return(x) } ######################################################################## ## Linear binning ## Courtesy of M Wand 2005 ## Extended by T Duong to 3- and 4-dim 2006 ## Extended by Gramack & Gramacki to include unconstrained b/w 2015 ######################################################################## binning <- function(x, H, h, bgridsize, xmin, xmax, supp=3.7, w, gridtype="linear") { ## default values x <- as.matrix(x) d <- ncol(x) n <- nrow(x) if (missing(bgridsize)) bgridsize <- default.gridsize(d) if (missing(w)) w <- rep(1,n) if (missing(h)) h <- rep(0,d) if (!missing(H)) h <- sqrt(diag(H)) range.x <- list() if (!missing(xmin) & !missing(xmax)) for (i in 1:d) range.x[[i]] <- c(xmin[i], xmax[i]) else if (!missing(xmin) & missing(xmax)) for (i in 1:d) range.x[[i]] <- c(xmin[i], max(x[,i]) + supp*h[i]) else if (missing(xmin) & !missing(xmax)) for (i in 1:d) range.x[[i]] <- c(min(x[,i]) - supp*h[i], xmax[i]) else for (i in 1:d) range.x[[i]] <- c(min(x[,i]) - supp*h[i], max(x[,i]) + supp*h[i]) a <- sapply(range.x,min) b <- sapply(range.x,max) if (missing(gridtype)) gridtype <- rep("linear", d) gridtype.vec <- rep("", d) gpoints <- list() for (id in 1:d) { gridtype1 <- match.arg(gridtype[i], c("linear", "sqrt", "quantile", "log")) if (gridtype1=="linear") gpoints[[id]] <- seq(a[id],b[id],length=bgridsize[id]) else if (gridtype1=="log") gpoints[[id]] <- seq(exp(a[id]),exp(b[id]),length=bgridsize[id]) } if (d==1) counts <- linbin.ks(x,gpoints[[1]], w=w) if (d==2) counts <- linbin2D.ks(x,gpoints[[1]],gpoints[[2]], w=w) if (d==3) counts <- linbin3D.ks(x,gpoints[[1]],gpoints[[2]],gpoints[[3]], w=w) if (d==4) counts <- linbin4D.ks(x,gpoints[[1]],gpoints[[2]],gpoints[[3]],gpoints[[4]], w=w) bin.counts <- list(counts=counts, eval.points=gpoints, w=w) if (d==1) bin.counts <- lapply(bin.counts, unlist) return(bin.counts) } ######################################################################## ## Linear binning ######################################################################## linbin.ks <- function(x, gpoints, w) { n <- length(x) M <- length(gpoints) if (missing(w)) w <- rep(1, n) a <- gpoints[1] b <- gpoints[M] xi <- .C(C_massdist1d, x1=as.double(x[,1]), n=as.integer(n), a1=as.double(a), b1=as.double(b), M1=as.integer(M), weight=as.double(w), est=double(M))$est return(xi) } linbin2D.ks <- function(x, gpoints1, gpoints2, w) { n <- nrow(x) M1 <- length(gpoints1) M2 <- length(gpoints2) a1 <- gpoints1[1] a2 <- gpoints2[1] b1 <- gpoints1[M1] b2 <- gpoints2[M2] if (missing(w)) w <- rep(1, n) ## binning for interior points out <- .C(C_massdist2d, x1=as.double(x[,1]), x2=as.double(x[,2]), n=as.integer(n), a1=as.double(a1), a2=as.double(a2), b1=as.double(b1), b2=as.double(b2), M1=as.integer(M1), M2=as.integer(M2), weight=as.double(w), est=double(M1*M2)) xi <- matrix(out$est, nrow=M1, ncol=M2) return(xi) } linbin3D.ks <- function(x, gpoints1, gpoints2, gpoints3, w) { n <- nrow(x) M1 <- length(gpoints1) M2 <- length(gpoints2) M3 <- length(gpoints3) a1 <- gpoints1[1] a2 <- gpoints2[1] a3 <- gpoints3[1] b1 <- gpoints1[M1] b2 <- gpoints2[M2] b3 <- gpoints3[M3] if (missing(w)) w <- rep(1, n) ## binning for interior points out <- .C(C_massdist3d, x1=as.double(x[,1]), x2=as.double(x[,2]), x3=as.double(x[,3]), n=as.integer(n), a1=as.double(a1), a2=as.double(a2), a3=as.double(a3), b1=as.double(b1), b2=as.double(b2), b3=as.double(b3), M1=as.integer(M1), M2=as.integer(M2), M3=as.integer(M3), weight=as.double(w), est=double(M1*M2*M3)) xi <- array(out$est, dim=c(M1,M2,M3)) return(xi) } linbin4D.ks <- function(x, gpoints1, gpoints2, gpoints3, gpoints4, w) { n <- nrow(x) M1 <- length(gpoints1) M2 <- length(gpoints2) M3 <- length(gpoints3) M4 <- length(gpoints4) a1 <- gpoints1[1] a2 <- gpoints2[1] a3 <- gpoints3[1] a4 <- gpoints4[1] b1 <- gpoints1[M1] b2 <- gpoints2[M2] b3 <- gpoints3[M3] b4 <- gpoints4[M4] if (missing(w)) w <- rep(1, n) ## binning for interior points out <- .C(C_massdist4d, x1=as.double(x[,1]), x2=as.double(x[,2]), x3=as.double(x[,3]), x4=as.double(x[,4]), n=as.integer(n), a1=as.double(a1), a2=as.double(a2), a3=as.double(a3), a4=as.double(a4), b1=as.double(b1), b2=as.double(b2), b3=as.double(b3), b4=as.double(b4), M1=as.integer(M1), M2=as.integer(M2), M3=as.integer(M3), M4=as.integer(M4), weight=as.double(w), est=double(M1*M2*M3*M4)) xi <- array(out$est, dim=c(M1,M2,M3,M4)) return(xi) } ######################################################################## ## Discrete convolution ######################################################################## symconv.1d <- function(keval, gcounts) { M <- length(gcounts) N <- length(keval) L <- round(length(keval)/2)+1 ## Smallest powers of 2 >= M + N P <- 2^(ceiling(log2(M + N))) ## Zero-padded version of keval and gcounts keval.zeropad <- rep(0, P) gcounts.zeropad <- rep(0, P) keval.zeropad[1:N] <- keval gcounts.zeropad[L:(L+M-1)] <- gcounts ## FFTs K <- fft(keval.zeropad) C <- fft(gcounts.zeropad) ## Invert element-wise product of FFTs and truncate and normalise it symconv.val <- Re(fft(K*C, inverse=TRUE)/P)[N:(N+M-1)] return(symconv.val) } symconv.nd <- function(keval, gcounts, d) { M <- dim(gcounts) N <- dim(keval) L <- round(dim(keval)/2)+1 ## Smallest powers of 2 > M + N P <- 2^(ceiling(log2(M + N))) ## Zero-padded version of keval and gcounts keval.zeropad <- array(0, dim=P) gcounts.zeropad <- array(0, dim=P) if (d==2) { keval.zeropad[1:N[1], 1:N[2]] <- keval gcounts.zeropad[L[1]:(L[1]+M[1]-1), L[2]:(L[2]+M[2]-1)] <- gcounts } else if (d==3) { keval.zeropad[1:N[1], 1:N[2], 1:N[3]] <- keval gcounts.zeropad[L[1]:(L[1]+M[1]-1), L[2]:(L[2]+M[2]-1), L[3]:(L[3]+M[3]-1)] <- gcounts } else if (d==4) { keval.zeropad[1:N[1], 1:N[2], 1:N[3], 1:N[4]] <- keval gcounts.zeropad[L[1]:(L[1]+M[1]-1), L[2]:(L[2]+M[2]-1), L[3]:(L[3]+M[3]-1), L[4]:(L[4]+M[4]-1)] <- gcounts } ## FFTs K <- fft(keval.zeropad) C <- fft(gcounts.zeropad) ## Invert element-wise product of FFTs and truncate and normalise it symconv.val <- Re(fft(K*C, inverse=TRUE)/prod(P)) if (d==2) symconv.val <- symconv.val[N[1]:(N[1]+M[1]-1), N[2]:(N[2]+M[2]-1)] else if (d==3) symconv.val <- symconv.val[N[1]:(N[1]+M[1]-1), N[2]:(N[2]+M[2]-1), N[3]:(N[3]+M[3]-1)] else if (d==4) symconv.val <- symconv.val[N[1]:(N[1]+M[1]-1), N[2]:(N[2]+M[2]-1), N[3]:(N[3]+M[3]-1), N[4]:(N[4]+M[4]-1)] return(symconv.val) } ks/R/mise.R0000644000176200001440000004674114465213133012166 0ustar liggesusers############################################################################### ## Exact MISE for normal mixtures ############################################################################### ## nu, gamma_r, gamma_r2 written by Jose Chacon 10/2008 nu <- function(r, A) { ## using the recursive formula provided in Kan (2008) ei <- eigen(A)$values tr.vec <- numeric(r) for(p in 1:r) tr.vec[p] <- sum(ei^p) nu.val <- 1 if (r>=1) { for(p in 1:r) { a <- sum(tr.vec[1:p]*rev(nu.val))/(2*p) nu.val <- c(nu.val,a) } } return(factorial(r)*2^r*nu.val[r+1]) } nu.rs <- function(r, s, A, B) { if (s==0) nu.val <- nu(r=r, A=A) else if (s>=1) { nu.val <- 0 for (i in 0:r) for (j in 0:(s-1)) nu.val <- nu.val + choose(r,i)*choose(s-1,j) *factorial(r+s-i-j-1)*2^(r+s-i-j-1)*tr(matrix.pow(A,r-i)%*%matrix.pow(B,s-j))*nu.rs(r=i,s=j, A=A, B=B) } return(nu.val) } ## gamma functional for normal mixture MISE gamma_r <- function(mu, Sigma, r) { Sigmainv <- chol2inv(chol(Sigma)) d <- ncol(Sigma) v <- 0 for (j in 0:r) v <- v + (-1)^j*choose(2*r, 2*j)*OF(2*j)*nu.rs(r=r-j, s=j, A=Sigmainv%*%mu%*%t(mu)%*%Sigmainv, B=Sigmainv) v <- (-1)^r*v*drop(dmvnorm.deriv(x=rep(0,d), mu=mu, Sigma=Sigma,deriv.order=0)/OF(2*r)) return(v) } ## gamma functional for normal mixture AMISE gamma_r2 <- function(mu, Sigma, d, r, H) { Sigmainv <- chol2inv(chol(Sigma)) if (d==1) w <- vec(Kpow(Sigmainv %*% Sigmainv, r)) %x% vec(Kpow(Sigmainv %*% H %*% Sigmainv, 2)) else w <- matrix(Sdrv(d=d,r=2*r+4, v=vec(Kpow(Sigmainv %*% Sigmainv, r)) %x% vec(Kpow(Sigmainv %*% H %*% Sigmainv, 2))), nrow=1) v <- rep(0,length=d^(2*r+4)) for(j in 0:(r+2)) v <- v+((-1)^j*OF(2*j)*choose(2*r+4, 2*j))*(Kpow(mu,2*r-2*j+4)%x%Kpow(vec(Sigma),j)) gamr <- (-1)^r*mvtnorm::dmvnorm(mu,mean=rep(0,d),sigma=Sigma)*sum(w %*% v) return(gamr) } ############################################################################### ## Omega matrices (for exact MISE for normal mixtures) ## ## Parameters ## mus - means ## Sigmas - variances ## k - number of mixture components ## a - subscript of Omega matrix ## H - bandwidth matrix ## ## Returns ## Omega matrix ############################################################################### omega <- function(mus, Sigmas, k, a, H, r) { ## the (i,j) element of Omega matrix is ## dmvnorm(0, mu_i - mu_j, a*H + Sigma_i + Sigma_j) if (is.matrix(mus)) d <- ncol(mus) else d <- length(mus) if (k == 1) omega.mat <- gamma_r(mu=rep(0,d),Sigma=a*H + 2*Sigmas, r=r) else { omega.mat <- matrix(0, nrow=k, ncol=k) for (i in 1:k) { Sigmai <- Sigmas[((i-1)*d+1):(i*d),] mui <- mus[i,] for (j in 1:k) { Sigmaj <- Sigmas[((j-1)*d+1):(j*d),] muj <- mus[j,] omega.mat[i,j] <- gamma_r(mu=mui-muj, Sigma=a*H + Sigmai + Sigmaj, r=r) } } } return(omega.mat) } omega.1d <- function(mus, sigmas, k, a, h, r) { H <- h^2 Sigmas <- sigmas^2 if (k == 1) omega.mat <- gamma_r(mu=0, Sigma=as.matrix(a*H + 2*Sigmas), r=r) else { omega.mat <- matrix(0, nrow=k, ncol=k) for (i in 1:k) { Sigmai <- Sigmas[i] mui <- mus[i] for (j in 1:k) { Sigmaj <- Sigmas[j] muj <- mus[j] omega.mat[i,j] <- gamma_r(mu=mui-muj, Sigma=as.matrix(a*H + Sigmai + Sigmaj), r=r) } } } return(omega.mat) } ############################################################################## ## Exact MISE for normal mixtures ## ## Parameters ## mus - means ## Sigmas - variances ## Props - vector of proportions of each mixture component ## H - bandwidth matrix ## samp - sample size ## ## Returns ## Exact MISE for normal mixtures ############################################################################### mise.mixt <- function(H, mus, Sigmas, props, samp, h, sigmas, deriv.order=0) { if (!(missing(h))) return(mise.mixt.1d(h=h, mus=mus, sigmas=sigmas, props=props, samp=samp, deriv.order=deriv.order)) if (is.vector(mus)) d <- length(mus) else d <- ncol(mus) k <- length(props) r <- deriv.order Hinv <- chol2inv(chol(H)) ## formula is found in Wand & Jones (1993) and Chacon, Duong & Wand (2008) if (k == 1) { mise <- 2^(-r)*nu(r,Hinv)/(samp * (4 * pi)^(d/2) * sqrt(det(H))) + (1-1/samp)*omega(mus, Sigmas, 1, 2, H, r) - 2*omega(mus, Sigmas, 1, 1, H, r) + omega(mus, Sigmas, 1, 0, H, r) } else { mise <- 2^(-r)*nu(r,Hinv)/(samp * (4 * pi)^(d/2) * sqrt(det(H))) + props %*% ((1-1/samp)*omega(mus, Sigmas, k, 2, H, r) - 2*omega(mus, Sigmas, k, 1, H, r) + omega(mus, Sigmas, k, 0, H, r)) %*% props } return(drop(mise)) } mise.mixt.1d <- function(h, mus, sigmas, props, samp, deriv.order=0) { d <- 1 k <- length(props) r <- deriv.order H <- as.matrix(h^2) Hinv <- chol2inv(chol(H)) ## formula is found in Wand & Jones (1993) and Chacon, Duong & Wand (2008) if (k == 1) { mise <- 2^(-r)*nu(r,Hinv)/(samp * (4 * pi)^(d/2) * sqrt(det(H))) + (1-1/samp)*omega.1d(mus, sigmas, 1, 2, h, r) - 2*omega.1d(mus, sigmas, 1, 1, h, r) + omega.1d(mus, sigmas, 1, 0, h, r) } else { mise <- 2^(-r)*nu(r,Hinv)/(samp * (4 * pi)^(d/2) * sqrt(det(H))) + props %*% ((1-1/samp)*omega.1d(mus, sigmas, k, 2, h, r) - 2*omega.1d(mus, sigmas, k, 1, h, r) + omega.1d(mus, sigmas, k, 0, h, r)) %*% props } return(drop(mise)) } ############################################################################### ## Exact AMISE for bivariate normal mixtures ## ## Parameters ## mus - means ## Sigmas - variances ## props - mixing proportions ## H - bandwidth matrix ## samp - sample size ## ## Returns ## Exact AMISE for normal mixtures ############################################################################### amise.mixt <- function(H, mus, Sigmas, props, samp, h, sigmas, deriv.order=0) { if (!(missing(h))) return(amise.mixt.1d(h=h, mus=mus, sigmas=sigmas, props=props, samp=samp, deriv.order=deriv.order)) r <- deriv.order if (is.vector(mus)) { d <- length(mus); mus <- t(matrix(mus)) } else d <- ncol(mus) k <- length(props) if (k == 1) { Sigmasinv <- chol2inv(chol(Sigmas)) omega.mat <- 2^(-d-r-2)*pi^(-d/2)*det(Sigmas)^(-1/2)*nu.rs(r=r, s=2, Sigmasinv, matrix.sqrt(Sigmasinv) %*% H %*% matrix.sqrt(Sigmasinv)) } else { omega.mat <- matrix(0, nrow=k, ncol=k) for (i in 1:k) { Sigmai <- Sigmas[((i-1)*d+1):(i*d),] mui <- mus[i,] for (j in 1:k) { Sigmaj <- Sigmas[((j-1)*d+1):(j*d),] muj <- mus[j,] omega.mat[i,j] <- gamma_r2(mu=mui-muj, Sigma=Sigmai + Sigmaj, d=d, r=r, H=H) } } } Hinv <- chol2inv(chol(H)) if (k == 1) amise <- 2^(-r)*nu(r,Hinv)/(samp * (4 * pi)^(d/2) * sqrt(det(H))) + omega.mat/4 else amise <- 2^(-r)*nu(r,Hinv)/(samp * (4 * pi)^(d/2) * sqrt(det(H))) + (props %*% omega.mat %*% props)/4 return(drop(amise)) } amise.mixt.1d <- function(h, mus, sigmas, props, samp, deriv.order=0) { d <- 1 r <- deriv.order k <- length(props) H <- as.matrix(h^2) if (k == 1) omega.mat <- gamma_r2(mu=rep(0,d),Sigma=as.matrix(2*sigmas^2), d=d, r=r, H=H) else { omega.mat <- matrix(0, nrow=k, ncol=k) for (i in 1:k) { Sigmai <- as.matrix(sigmas[i]^2) mui <- mus[i] for (j in 1:k) { Sigmaj <- as.matrix(sigmas[j]^2) muj <- mus[j] omega.mat[i,j] <- gamma_r2(mu=mui-muj, Sigma= Sigmai + Sigmaj, d=d, r=r, H=H) } } } Hinv <- h^(-2) if (k == 1) amise <- 2^(-r)*nu(r,Hinv)/(samp * (4 * pi)^(d/2) * sqrt(det(H))) + omega.mat/4 else amise <- 2^(-r)*nu(r,Hinv)/(samp * (4 * pi)^(d/2) * sqrt(det(H))) + (props %*% omega.mat %*% props)/4 return(drop(amise)) } ############################################################################### ## Lambda matrices (for exact AMISE for normal mixtures) ## ## Parameters ## mus - means ## Sigmas - variances ## k - number of mixture components ## r - derivative (r1, r2) ## ## Returns ## Lambda matrix ############################################################################### lambda <- function(mus, Sigmas, k, r) { ## the (i,j) element of Lambda matrix is d^r/ dx^r dmvnorm(0, mu_i - mu_j, ## a*H + Sigma_i + Sigma_j) if (is.vector(mus)) d <- length(mus) else d <- ncol(mus) if (k == 1) lambda.mat <- dmvnorm.deriv(deriv.order=r, x=rep(0, length(mus)), Sigma=2*Sigmas) else { if (is.matrix(mus)) d <- ncol(mus) else d <- length(mus) lambda.mat <- matrix(0, nrow=k, ncol=k) for (i in 1:k) { Sigmai <- Sigmas[((i-1)*d+1) : (i*d),] mui <- mus[i,] for (j in 1:k) { Sigmaj <- Sigmas[((j-1)*d+1) : (j*d),] muj <- mus[j,] lambda.mat[i,j] <- dmvnorm.deriv(deriv.order=r, x=mui-muj,Sigma=Sigmai+Sigmaj) } } } return(lambda.mat) } amise.mixt.2d <- function(H, mus, Sigmas, props, samp) { d <- ncol(Sigmas) k <- length(props) ## formula is found in Wand & Jones (1993) if (k == 1) { amise <- 1/(samp * (4*pi)^(d/2) * sqrt(det(H))) + 1/4 *(lambda(mus, Sigmas, k, r=c(4,0))*H[1,1]^2 + 4*lambda(mus, Sigmas, k, r=c(3,1))*H[1,1]*H[1,2] + 2*lambda(mus, Sigmas, k, r=c(2,2))*(H[1,1]*H[2,2] + 2*H[1,2]^2) + 4*lambda(mus, Sigmas, k, r=c(1,3))*H[2,2]*H[1,2]+ lambda(mus, Sigmas, k, r=c(0,4))*H[2,2]^2) } else { amise <- 1/(samp * (4*pi)^(d/2) * sqrt(det(H))) + 1/4 * props %*% (lambda(mus, Sigmas, k, r=c(4,0))*H[1,1]^2 + 4*lambda(mus, Sigmas, k, r=c(3,1))*H[1,1]*H[1,2] + 2*lambda(mus, Sigmas, k, r=c(2,2))*(H[1,1]*H[2,2] + 2*H[1,2]^2) + 4*lambda(mus, Sigmas, k, r=c(1,3))*H[2,2]*H[1,2]+ lambda(mus, Sigmas, k, r=c(0,4))*H[2,2]^2) %*% props } return(drop(amise)) } ############################################################################### ## Finds the bandwidth matrix that minimises the MISE for normal mixtures ## ## Parameters ## mus - means ## Sigmas - variances ## props - vector of proportions of each mixture component ## Hstart - initial bandwidth matrix ## samp - sample size ## full - 1 minimise over full bandwidth matrices ## - 0 minimise over diagonal bandwidth matrices ## ## Returns ## H_MISE ############################################################################### hmise.mixt <- function(mus, sigmas, props, samp, hstart, deriv.order=0) { r <- deriv.order d <- 1 if (missing(hstart)) { x <- rnorm.mixt(n=1000, mus=mus, sigmas=sigmas) hstart <- sqrt((4/(samp*(d+2*r+2)))^(2/(d+2*r+4)) * var(x)) } mise.mixt.temp <- function(h) { return(mise.mixt.1d(h=h, mus=mus, sigmas=sigmas, props=props, samp=samp, deriv.order=deriv.order)) } result <- optimize(f=mise.mixt.temp, interval=c(0, 10*hstart)) hmise <- result$minimum return(hmise) } Hmise.mixt <- function(mus, Sigmas, props, samp, Hstart, deriv.order=0) { r <- deriv.order if (is.vector(mus)) d <- length(mus) else d <- ncol(mus) ## use normal reference estimate as initial condition if (missing(Hstart)) { x <- rmvnorm.mixt(10000, mus, Sigmas, props) Hstart <- (4/(samp*(d+2*r+2)))^(2/(d+2*r+4)) * var(x) } mise.mixt.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) return(mise.mixt(H=H, mus=mus, Sigmas=Sigmas, props=props, samp=samp, deriv.order=deriv.order)) } Hstart <- vech(matrix.sqrt(Hstart)) result <- nlm(p=Hstart, f=mise.mixt.temp) Hmise <- invvech(result$estimate) %*% invvech(result$estimate) return(Hmise) } Hmise.mixt.diag <- function(mus, Sigmas, props, samp, Hstart, deriv.order=0) { if (is.vector(mus)) d <- length(mus) else d <- ncol(mus) if (missing(Hstart)) { x <- rmvnorm.mixt(10000, mus, Sigmas, props) Hstart <- (4/(samp*(d + 2)))^(2/(d + 4)) * var(x) } mise.mixt.temp <- function(diagH) { H <- diag(diagH) %*% diag(diagH) return(mise.mixt(H=H, mus=mus, Sigmas=Sigmas, props=props, samp=samp, deriv.order=deriv.order)) } Hstart <- diag(matrix.sqrt(Hstart)) result <- nlm(p=Hstart, f=mise.mixt.temp) Hmise <- diag(result$estimate) %*% diag(result$estimate) return(Hmise) } ############################################################################### ## Finds bandwidth matrix that minimises the AMISE for normal mixtures - 2-dim ## ## Parameters ## mus - means ## Sigmas - variances ## props - vector of proportions of each mixture component ## Hstart - initial bandwidth matrix ## samp - sample size ## ## Returns ## Bandwidth matrix that minimises AMISE ############################################################################### hamise.mixt <- function(mus, sigmas, props, samp, hstart, deriv.order=0) { r <- deriv.order d <- 1 if (missing(hstart)) { x <- rnorm.mixt(n=1000, mus=mus, sigmas=sigmas) hstart <- sqrt((4/(samp*(d+2*r+2)))^(2/(d+2*r+4)) * var(x)) } amise.mixt.temp <- function(h) { return(amise.mixt.1d(h=h, mus=mus, sigmas=sigmas, props=props, samp=samp, deriv.order=deriv.order)) } result <- optimize(f=amise.mixt.temp, interval=c(0, 10*hstart)) hamise <- result$minimum return(hamise) } Hamise.mixt <- function(mus, Sigmas, props, samp, Hstart, deriv.order=0) { r <- deriv.order if (is.vector(mus)) d <- length(mus) else d <- ncol(mus) ## use explicit formula for single normal if (length(props)==1) { Hamise <- (4/ (samp*(d+2*r+2)))^(2/(d+2*r+4)) * Sigmas } else { ## use normal reference estimate as initial condition if (missing(Hstart)) { x <- rmvnorm.mixt(10000, mus=mus, Sigmas=Sigmas, props=props) Hstart <- matrix.sqrt((4/ (samp*(d+2*r+2)))^(2/(d+2*r+4)) * var(x)) } amise.mixt.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) return(amise.mixt(H=H, mus=mus, Sigmas=Sigmas, props=props, samp=samp, deriv.order=deriv.order)) } result <- nlm(p=vech(Hstart), f=amise.mixt.temp) Hamise <- invvech(result$estimate) %*% invvech(result$estimate) } return(Hamise) } Hamise.mixt.diag <- function(mus, Sigmas, props, samp, Hstart, deriv.order=0) { r <- deriv.order if (is.vector(mus)) d <- length(mus) else d <- ncol(mus) ## use normal reference estimate as initial condition if (missing(Hstart)) { x <- rmvnorm.mixt(10000, mus, Sigmas, props) Hstart <- matrix.sqrt((4/ (samp*(d+2*r+2)))^(2/(d+2*r+4)) * var(x)) } amise.mixt.temp <- function(diagH) { H <- diag(diagH) %*% diag(diagH) return(amise.mixt(H=H, mus=mus, Sigmas=Sigmas, props=props, samp=samp, deriv.order=deriv.order)) } result <- nlm(p=diag(Hstart), f=amise.mixt.temp) Hamise <- diag(result$estimate) %*% diag(result$estimate) return(Hamise) } ############################################################################### ## ISE for normal mixtures (fixed KDE) ## ## Parameters ## x - data values ## H - bandwidth matrix ## mus - matrix of means (each row is a vector of means from each component ## density) ## Sigmas - matrix of covariance matrices (every d rows is a covariance matrix ## from each component density) ## props - mixing proportions ## lower - vector of lower end points of rectangle ## upper - vector of upper end points of rectangle ## gridsize - vector of number of grid points ## stepsize - vector of step sizes ## ## Returns ## ISE ############################################################################### ise.mixt <- function(x, H, mus, Sigmas, props, h, sigmas, deriv.order=0, binned=FALSE, bgridsize) { if (!(missing(h))) return(ise.mixt.1d(x=x, h=h, mus=mus, sigmas=sigmas, props=props, deriv.order=deriv.order, binned=binned)) if (is.vector(x)) x <- matrix(x, nrow=1) if (is.vector(mus)) mus <- matrix(mus, nrow=length(props)) d <- ncol(x) n <- nrow(x) M <- length(props) r <- deriv.order ## formula is found in thesis vIdr <- vec(diag(d^r)) ise1 <- 0 ise2 <- 0 ise3 <- 0 if (binned) { ise1 <- dmvnorm.deriv.sum(x=x, Sigma=2*H, inc=1, deriv.order=2*r, binned=binned, bgridsize=bgridsize) for (j in 1:M) { Sigmaj <- Sigmas[((j - 1) * d + 1):(j * d), ] ise2 <- ise2 + props[j] * colSums(dmvnorm.deriv(x, mu=mus[j,],Sigma=H + Sigmaj, deriv.order=2*r)) for (i in 1:M) { Sigmai <- Sigmas[((i - 1) * d + 1):(i * d), ] ise3 <- ise3 + props[i] * props[j] * dmvnorm.deriv(x=mus[i,],mu=mus[j,], Sigma = Sigmai + Sigmaj, deriv.order = 2*r) } } ise <- (-1)^r * sum(vIdr*(ise1/n^2 - 2 * ise2/n + ise3)) } else { ise1 <- Qr(x=x, Sigma=2*H, inc=1, deriv.order=2*r) for (j in 1:M) { Sigmaj <- Sigmas[((j-1)*d + 1):(j*d), ] ise2 <- ise2 + props[j] * Qr(x=x, y=mus[j,], Sigma=H + Sigmaj, deriv.order=2*r, inc=1) for (i in 1:M) { Sigmai <- Sigmas[((i-1)*d + 1):(i*d), ] ise3 <- ise3 + props[i] * props[j] * Qr(x=mus[i,], y=mus[j,], Sigma=Sigmai + Sigmaj, deriv.order=2*r, inc=1) } } ise <- (-1)^r*(ise1 - 2*ise2 + ise3) } return(ise) } ise.mixt.1d <- function(x, h, mus, sigmas, props, deriv.order=0, binned=FALSE) { n <- length(x) M <- length(props) r <- deriv.order ise1 <- 0 ise2 <- 0 ise3 <- 0 ise1 <- dnorm.deriv.sum(x=x, sigma=sqrt(2)*h, inc=1, deriv.order=2*r, binned=binned) for (j in 1:M) { sigmaj <- sigmas[j] ise2 <- ise2 + sum(props[j]*dnorm.deriv(x=x, mu=mus[j], sigma=sqrt(h^2 + sigmaj^2), deriv.order=2*r)) for (i in 1:M) { sigmai <- sigmas[i] ise3 <- ise3 + sum(props[i]*props[j]*dnorm.deriv(x=mus[i], mu=mus[j], sigma=sqrt(sigmai^2+sigmaj^2), deriv.order=2*r)) } } return ((-1)^r*(ise1/n^2 - 2*ise2/n + ise3)) } Hise.mixt <- function(x, mus, Sigmas, props, Hstart, deriv.order=0) { r <- deriv.order if (is.vector(mus)) d <- length(mus) else d <- ncol(mus) samp <- nrow(x) ## use normal reference estimate as initial condition if (missing(Hstart)) { xstart <- rmvnorm.mixt(10000, mus, Sigmas, props) Hstart <- (4/(samp*(d+2*r+2)))^(2/(d+2*r+4)) * var(xstart) } ise.mixt.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) return(ise.mixt(x=x, H=H, mus=mus, Sigmas=Sigmas, props=props, deriv.order=deriv.order)) } Hstart <- vech(matrix.sqrt(Hstart)) result <- nlm(p=Hstart, f=ise.mixt.temp) Hise <- invvech(result$estimate) %*% invvech(result$estimate) return(Hise) } ks/R/kfs.R0000644000176200001440000001627514336513742012021 0ustar liggesusers############################################################################### ## Feature significance for ultivariate kernel density stimate ############################################################################### kfs <- function(x, H, h, deriv.order=2, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, verbose=FALSE, signif.level=0.05) { r <- deriv.order ## default values ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned bgridsize <- ksd$bgridsize gridsize <- ksd$gridsize ## clip data to xmin, xmax grid limits for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] if (positive & missing(xmin)) { xmin <- rep(0,d) } xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } if (d==1) { if (missing(h)) h <- hpi(x=x, nstage=2, binned=default.bflag(d=d, n=n), deriv.order=r) ## KDDE for r=2 fhatr <- kdde(x=x, h=h, deriv.order=r, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, deriv.vec=FALSE, verbose=verbose) ## KDE fhat <- kde(x=x, h=h, gridsize=gridsize, gridtype=gridtype, xmin=min(fhatr$eval.points), xmax=max(fhatr$eval.points), binned=binned, bgridsize=bgridsize, positive=positive, adj.positive=adj.positive, w=w) fhat.est <- as.vector(fhat$estimate) fhatr.est <- as.vector(fhatr$estimate) RDrK <- (-1)^r*psins.1d(r=2*r, sigma=1) fhatr.Sigma <- n^(-1)* h^(-2*r-1)*RDrK fhatr.Sigma12 <- sqrt(fhatr.Sigma) fhatr.est <- fhatr.est/fhatr.Sigma12 local.mode <- fhatr.est <= 0 fhatr.wald <- fhatr.est^2/abs(fhat.est) fhatr.wald[is.infinite(fhatr.wald)] <- max(fhatr.wald[!is.infinite(fhatr.wald)]) gs <- length(fhat$estimate) } else if (d>1) { if (missing(H)) H <- Hpi(x=x, nstage=2-(d>2), binned=default.bflag(d=d, n=n), deriv.order=r, verbose=verbose) ## KDDE for r=2 fhatr <- kdde(x=x, H=H, deriv.order=r, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, deriv.vec=FALSE, verbose=verbose) ## KDE fhat <- kde(x=x, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, verbose=verbose) fhat.est <- as.vector(fhat$estimate) fhatr.est <- sapply(fhatr$estimate, as.vector) ## convert from vec to vech because vec'ed derivative ## contains repeated columns so its variance isn't invertible Hinv <- chol2inv(chol(H)) Hinv12 <- matrix.sqrt(Hinv) RDrK <- (-1)^r*invvec(psins(r=2*r, Sigma=diag(d))) dupld <- dupl(d)$d dupld.MPinv <- chol2inv(chol(t(dupld)%*% dupld)) %*% t(dupld) fhatr.Sigma.const <- n^(-1)*det(H)^(-1/2)* Kpow(Hinv12,r) %*% RDrK %*% Kpow(Hinv12,2) fhatr.Sigma.const <- dupld.MPinv %*% fhatr.Sigma.const %*% t(dupld.MPinv) fhatr.Sigma.const12inv <- chol2inv(chol(matrix.sqrt(fhatr.Sigma.const))) fhatr.est <- fhatr.est %*% fhatr.Sigma.const12inv ## all eigenvalues < 0 => local mode fhatr.eigen <- lapply(lapply(seq(1,nrow(fhatr.est)), function(i) { invvech(fhatr.est[i,]) }), eigen, only.values=TRUE) fhatr.eigen <- t(sapply(fhatr.eigen, getElement, "values")) local.mode <- apply(fhatr.eigen <= 0, 1, all) fhatr.wald <- apply(fhatr.est^2, 1, sum)/fhat.est gs <- dim(fhat$estimate) } ## Hochberg adjustment for sequential tests pval.wald <- 1 - pchisq(fhatr.wald, d*(d+1)/2) pval.wald[fhat.est<=contourLevels(fhat, cont=99) | !local.mode] <- NA pval.wald.ord.index <- order(pval.wald) pval.wald.ord <- pval.wald[pval.wald.ord.index] num.test <- sum(!is.na(pval.wald.ord)) if (num.test>=1) num.test.seq <- c(1:num.test, rep(NA, prod(gs) - num.test)) else num.test.seq <- rep(NA, prod(gs)) reject.nonzero <- (pval.wald.ord <= signif.level/(num.test + 1 - num.test.seq)) reject.nonzero.ind <- which(reject.nonzero) ## reject null hypotheses indicated in reject.nonzero.ind if (d==1) { signif.wald <- array(0L, dim=gs) signif.wald[pval.wald.ord.index[reject.nonzero.ind]] <- 1L } else { signif.wald <- array(0L, dim=gs) signif.wald.index <- expand.grid(lapply(gs, seq_len)) signif.wald[as.matrix(signif.wald.index[pval.wald.ord.index[reject.nonzero.ind],])] <- 1L } ## ESS = effective sample size ## ess <- n*fhat$estimate*dmvnorm.mixt(x=rep(0,d), mu=rep(0,d), Sigma=H, props=1) ## signif.ess <- ess >= 5 fhatr$estimate <- signif.wald fhatr$type <- "kfs" class(fhatr) <- "kfs" return(fhatr) } ############################################################################# ## S3 methodfor KFS objects ############################################################################# ## plot method plot.kfs <- function(x, display="filled.contour", col=7, colors, abs.cont, alpha=1, alphavec=0.4, add=FALSE, ...) { fhatr <- x fhatr$deriv.order <- NULL class(fhatr) <- "kde" if (is.vector(fhatr$H)) d <- 1 else d <- ncol(fhatr$H) if (d==1) { fhat <- kde(x=fhatr$x, xmin=min(fhatr$eval.points), xmax=max(fhatr$eval.points), gridsize=length(fhatr$eval.points)) plot(fhat, col="grey", add=add, ...) gridsize <- length(fhatr$estimate) estimate.rle <- rle(as.vector(fhatr$estimate)) estimate.rle.cumsum <- rbind(cumsum(estimate.rle$lengths), estimate.rle$lengths, estimate.rle$values) col <- transparency.col(col, alpha=alpha) for (i in which(estimate.rle$values==1)) { seg.ind <- 1:estimate.rle.cumsum[2,i] if (i>1) seg.ind <- seg.ind + estimate.rle.cumsum[1,i-1] lines(fhat$eval.points[seg.ind], fhat$estimate[seg.ind], col=col, ...) } } else if (d==2) { if (missing(abs.cont)) abs.cont <- 0.5 disp1 <- match.arg(display, c("slice", "persp", "image", "filled.contour", "filled.contour2")) if (disp1=="filled.contour2") disp1 <- "filled.contour" col <- c("transparent",col) plot(fhatr, abs.cont=abs.cont, drawlabels=FALSE, col=col, add=add, display=display, alpha=alpha, ...) } else if (d==3) { if (missing(abs.cont)) abs.cont <- 0.25 e1 <- try(match.arg(display, c("plot3D", "rgl")), silent=TRUE) if (inherits(e1, "try-error")) display <- "plot3D" if (!missing(colors)) col <- colors plot(fhatr, abs.cont=abs.cont, col=col, colors=colors, alphavec=alphavec, add=add, display=display, ...) } invisible() } ## predict method predict.kfs <- function(object, ..., x) { fhat <- predict.kde(object=object, ..., x=x) fhat <- as.integer(fhat>=0.5) return(fhat) } ks/R/kcopula.R0000644000176200001440000002275614336513203012665 0ustar liggesusers############################################################################# ## Kernel copula and copula density estimators ############################################################################# ## empirical pseudo-uniform transformation ## taken from pobs() function in copula package pseudo.unif.empirical <- function (x, y) ##, na.last="keep", ties.method="average") { if (missing(y)) y <- x if (is.vector(y)) y <- matrix(y, nrow=1) d <- ncol(x) u <- matrix(0, ncol=d, nrow=nrow(x)) for (i in 1:d) { ecdf.fun <- ecdf(x=x[,i]) u[,i] <- ecdf.fun(y[,i]) } return(u) } ## kernel pseudo-uniform transformation pseudo.unif.kernel <- function(x, y, hs, binned=TRUE) { if (missing(y)) y <- x if (is.vector(y)) y <- matrix(y, nrow=1) d <- ncol(x) u <- list() for (i in 1:d) { u[[i]] <- kcde(x=x[,i], h=hs[i], eval.points=y[,i], binned=binned) } u2 <- sapply(u, getElement, "estimate") return(u2) } ############################################################################# ## Kernel copula estimator ############################################################################# kcopula <- function(x, H, hs, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, marginal="kernel", verbose=FALSE) { ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned gridsize <- ksd$gridsize bgridsize <- ksd$bgridsize ## clip data to xmin,xmax grid grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths if (missing(H)) H <- Hpi.kcde(x=x, binned=default.bflag(d=d,n=n)) if (missing(hs)) { hs <- rep(0, d) for (i in 1:d) hs[i] <- hpi.kcde(x=x[,i], binned=TRUE) } Fhat <- kcde(x=x, H=H, gridsize=gridsize, binned=binned, bgridsize=bgridsize, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, w=w, tail.flag="lower.tail", verbose=verbose) xlims <- sapply(Fhat$eval.points, range) xlims[1,] <- xlims[1,] - 0.1*abs(apply(xlims, 2, diff)) xlims[2,] <- xlims[2,] + 0.1*abs(apply(xlims, 2, diff)) ## generate pseudo-uniform values marginal <- match.arg(marginal, c("kernel", "empirical")) if (marginal=="kernel") { ## kernel pseudo-uniform u <- list() u.eval.points <- list() for (i in 1:d) { u.eval.points[[i]] <- kcde(x=x[,i], h=hs[i], eval.points=Fhat$eval.points[[i]], xmin=xlims[1,i], xmax=xlims[2,i], binned=TRUE) } y <- pseudo.unif.kernel(x=x, y=x, hs=hs, binned=TRUE) ep <- lapply(u.eval.points, getElement, "estimate") } else if (marginal=="empirical") { ## empirical pseudo-uniform y <- pseudo.unif.empirical(x=x, y=x) ep <- lapply(1:d, function(i) { f <- ecdf(x=x[,i]); f(Fhat$eval.points[[i]]) }) ep <- numeric() for (i in 1:d) { f <- ecdf(x=x[,i]) ep <- c(ep, list(f(Fhat$eval.points[[i]]))) } } Chat <- Fhat Chat$x <- y Chat$x.orig <- x Chat$eval.points <- ep Chat$hs <- hs ## loess smoothing on a uniform grid if (d==2) { ## select smaller grid for d==2 for memory usage reasons subselect <- round(cbind(seq(1,length(Chat$eval.points[[1]]), length=101), seq(1,length(Chat$eval.points[[2]]), length=101)),0) eval.points.df <- data.frame(expand.grid(Chat$eval.points[[1]][subselect[,1]], Chat$eval.points[[2]][subselect[,2]])) names(eval.points.df) <- paste("x", 1:ncol(eval.points.df), sep="") eval.points.df <- data.frame(estimate=as.numeric(Chat$estimate[subselect[,1], subselect[,2]]), eval.points.df) } else if (d==3) { ## select smaller grid for d==3 for memory usage reasons subselect <- round(cbind(seq(1,length(Chat$eval.points[[1]]), length=21), seq(1,length(Chat$eval.points[[2]]), length=21), seq(1,length(Chat$eval.points[[3]]), length=21)),0) eval.points.df <- data.frame(expand.grid(Chat$eval.points[[1]][subselect[,1]], Chat$eval.points[[2]][subselect[,2]], Chat$eval.points[[3]][subselect[,3]])) names(eval.points.df) <- paste("x", 1:ncol(eval.points.df), sep="") eval.points.df <- data.frame(estimate=as.numeric(Chat$estimate[subselect[,1], subselect[,2], subselect[,3]]), eval.points.df) } if (d==2) Chat.loess <- loess(estimate ~ x1+x2, data=eval.points.df, span=0.1) else if (d==3) Chat.loess <- loess(estimate ~ x1+x2+x3, data=eval.points.df, span=0.1) u.eval.points.regular <- list() for (i in 1:d) u.eval.points.regular[[i]] <- seq(0,1,length=length(Chat$eval.points[[i]])) u.eval.points.regular.df <- data.frame(expand.grid(u.eval.points.regular)) names(u.eval.points.regular.df) <- paste("x", 1:ncol(u.eval.points.regular.df), sep="") Chat.smoothed <- Chat Chat.smoothed$eval.points <- u.eval.points.regular Chat.smoothed$estimate <- array(predict(Chat.loess, newdata=u.eval.points.regular.df), dim=dim(Chat$estimate)) ## interpolate NA boundary values from loess smoothing gsdim <- dim(Chat$estimate) if (d==2) { Chat.smoothed$estimate[1,] <- 0 Chat.smoothed$estimate[gsdim[1],] <- Chat.smoothed$estimate[gsdim[1]-1,]*1.001 Chat.smoothed$estimate[,1] <-0 Chat.smoothed$estimate[,gsdim[2]] <- Chat.smoothed$estimate[,gsdim[2]-1]*1.001 Chat.smoothed$estimate[gsdim[1],gsdim[2]] <- 1 Chat.smoothed$estimate[Chat.smoothed$estimate>1] <- 1 } else if (d==3) { Chat.smoothed$estimate[,,1] <- 0 for (k in 2:(gsdim[3]-1)) { Chat.smoothed$estimate[,,k][1,] <- 0 Chat.smoothed$estimate[,,k][gsdim[1],] <- Chat.smoothed$estimate[,,k][gsdim[1]-1,]*1.001 Chat.smoothed$estimate[,,k][,1] <-0 Chat.smoothed$estimate[,,k][,gsdim[2]] <- Chat.smoothed$estimate[,,k][,gsdim[2]-1]*1.001 } Chat.smoothed$estimate[,,gsdim[3]] <- Chat.smoothed$estimate[,,gsdim[3]-1]*1.001 Chat.smoothed$estimate[gsdim[1],gsdim[2],gsdim[3]] <- 1 Chat.smoothed$estimate[Chat.smoothed$estimate>1] <- 1 } Chat <- Chat.smoothed Chat$marginal <- marginal class(Chat) <- "kcopula" return(Chat) } ############################################################################# ## Kernel copula density estimator ############################################################################# kcopula.de <- function(x, H, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, marginal="kernel", boundary.supp, boundary.kernel="beta", verbose=FALSE) { ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w if (missing(binned)) binned <- ksd$binned if (missing(bgridsize)) bgridsize <- ksd$bgridsize if (missing(gridsize)) gridsize <- ksd$gridsize ## clip data to xmin,xmax grid for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths hs <- rep(0, d) for (i in 1:d) hs[i] <- hpi.kcde(x=x[,i], binned=TRUE) ## generate pseudo-uniform values marginal <- match.arg(marginal, c("kernel", "empirical")) if (marginal=="kernel") y <- pseudo.unif.kernel(x=x, y=x, hs=hs, binned=TRUE) else if (marginal=="empirical") y <- pseudo.unif.empirical(x=x, y=x) colnames(y) <- colnames(x) if (missing(H)) H <- Hns(y) ## kernel copula density is boundary kernel estimator if (d==2 | d==3) chat <- kde.boundary(x=y, H=H, gridsize=gridsize, supp=supp, xmin=rep(0,d), xmax=rep(1,d), gridtype=gridtype, w=w, boundary.supp=boundary.supp, binned=FALSE, boundary.kernel=boundary.kernel, verbose=verbose) else stop("kcopula.de requires 2-d or 3-d data.") ## normalise KDE to integrate to 1 chat$estimate <- chat$estimate/sum(chat$estimate*apply(sapply(chat$eval.points, diff), 1, prod)[1]) chat$names <- parse.name(x) chat$x.orig <- x chat$hs <- hs ## compute prob contour levels if (compute.cont & missing(eval.points)) chat$cont <- contourLevels(chat, cont=1:99, approx=approx.cont) chat$marginal <- marginal class(chat) <- "kcopula.de" return(chat) } ############################################################################# ## S3 methods ############################################################################# ## plot methods plot.kcopula <- function(x, ...) { plot.kcde(x, ...) } plot.kcopula.de <- function(x, ...) { plot.kde(x, ...) } ## predict methods predict.kcopula <- function(object, ..., x, u) { if (missing(u)) { if (object$marginal=="kernel") u <- pseudo.unif.kernel(x=object$x.orig, y=x, hs=object$hs) } return(predict.kde(object, ..., x=u)) } predict.kcopula.de <- function(object, ..., x, u) { if (missing(u)) { if (object$marginal=="kernel") u <- pseudo.unif.kernel(x=object$x.orig, y=x, hs=object$hs) } return(predict.kde(object, ..., x=u)) } ## contourLevel method contourLevels.kcopula.de <- function(x, ...) { x1 <- x; class(x1) <- "kde" return(contourLevels(x=x1, ...)) } ks/R/vkde.R0000644000176200001440000002003314336302305012141 0ustar liggesusers###################################################################### ## Balloon variable KDE ###################################################################### kde.balloon <- function(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, verbose=FALSE) { ## default values ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned gridsize <- ksd$gridsize bgridsize <- ksd$bgridsize ## clip data to xmin, xmax grid limits for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths if (d==1 & missing(h)) h <- hns(x=x, deriv.order=2) if (d>1 & missing(H)) H <- Hns(x=x, deriv.order=2) if (d==2) fhat <- kde.balloon.2d(x=x, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, compute.cont=compute.cont, approx.cont=approx.cont, verbose=verbose) else stop("kde.balloon only implemented for d=2") if (compute.cont) fhat$cont <- contourLevels(fhat, cont=1:99, approx=approx.cont) return(fhat) } ###################################################################### ## Bivariate balloon variable KDE ###################################################################### kde.balloon.2d <- function(x, H, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, verbose=FALSE) { d <- ncol(x) n <- nrow(x) fhat <- kde(x=x, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, compute.cont=compute.cont, approx.cont=approx.cont) fhat.ep <- expand.grid(fhat$eval.points) fhat.pilot <- fhat fhat2.pilot <- kdde(x=x, deriv.order=2, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w) h.pi <- (d*predict(fhat.pilot, x=fhat.ep)/drop((4*pi)^(d/2)*predict(fhat2.pilot, x=fhat.ep) %*% vec(diag(d)))^2)^(1/(d+4))*n^(-1/(d+4)) h.pi <- array(h.pi, dim=dim(fhat$estimate)) gs <- dim(fhat$estimate) if (verbose) { pb <- txtProgressBar(max=prod(gs)); k <- 0 } fhat$estimate <- array(0, dim=dim(fhat$estimate)) for (i in 1:gs[2]) for (j in 1:gs[1]) { if (!is.na(h.pi[i,j]) & !is.infinite(h.pi[i,j])) if (h.pi[i,j]>0) fhat$estimate[i,j] <- kde(x=x, w=w, H=h.pi[i,j]^2*diag(d), eval.points=fhat.ep[i+(j-1)*gs[1],])$estimate if (verbose) { k <- k+1; setTxtProgressBar(pb,k) } } if (verbose) close(pb) ## re-scale density estimate to integral 1 delta.int <- prod(sapply(fhat$eval.points, diff)[1,]) riemann.sum <- sum(fhat$estimate*delta.int) fhat$estimate <- fhat$estimate/riemann.sum fhat$names <- parse.name(x) ## add variable names if (compute.cont) fhat$cont <- contourLevels(fhat, cont=1:99, approx=approx.cont) fhat$H <- h.pi^2 return(fhat) } ###################################################################### ## Sample point variable KDE ###################################################################### kde.sp <- function(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, verbose=FALSE) { ## default values ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned gridsize <- ksd$gridsize bgridsize <- ksd$bgridsize ## clip data to xmin, xmax grid limits for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths if (d==1 & missing(h)) h <- hns(x=x, deriv.order=4) if (d>1 & missing(H)) H <- Hns(x=x, deriv.order=4) if (d==2) fhat <- kde.sp.2d(x=x, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, compute.cont=compute.cont, approx.cont=approx.cont, verbose=verbose, pre=TRUE) else stop("kde.sp only implemented for d=2") fhat$names <- parse.name(x) ## add variable names if (compute.cont) fhat$cont <- contourLevels(fhat, cont=1:99, approx=approx.cont) return(fhat) } ###################################################################### ## Bivariate sample point variable KDE ###################################################################### kde.sp.2d <- function(x, H, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, verbose=FALSE, pre=TRUE) { d <- 2; n <- nrow(x) if (pre) { x.orig <- x S12 <- diag(apply(x.orig, 2, sd)) Sinv12 <- matrix.pow(S12,-1) x <- pre.scale(x) if (!missing(xmin)) xmin <- xmin %*% Sinv12 if (!missing(xmax)) xmax <- xmax %*% Sinv12 H <- Hns(x=x, deriv.order=4) } fhat <- kde(x=x, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w, compute.cont=compute.cont, approx.cont=approx.cont) fhat.pilot <- fhat fhat1.pilot <- kdde(x=x, deriv.order=1, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w) fhat2.pilot <- kdde(x=x, deriv.order=2, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w) fhat3.pilot <- kdde(x=x, deriv.order=3, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w) fhat4.pilot <- kdde(x=x, deriv.order=4, H=H, gridsize=gridsize, gridtype=gridtype, xmin=xmin, xmax=xmax, supp=supp, eval.points=eval.points, binned=binned, bgridsize=bgridsize, w=w) fhat.pilot <- predict(fhat.pilot, x=x) fhat1.pilot <- predict(fhat1.pilot, x=x) fhat2.pilot <- predict(fhat2.pilot, x=x) fhat3.pilot <- predict(fhat3.pilot, x=x) fhat4.pilot <- predict(fhat4.pilot, x=x) lambda1 <- 1/fhat.pilot^5*rowKpow(fhat1.pilot, r=4) lambda2 <- 1/fhat.pilot^4*rowKpow(fhat2.pilot, fhat1.pilot, r=1, s=2) lambda3 <- 1/fhat.pilot^3*rowKpow(fhat2.pilot, r=2) lambda4 <- 1/fhat.pilot^3*rowKpow(fhat3.pilot, fhat1.pilot, r=1, s=1) lambda5 <- 1/fhat.pilot^2*fhat4.pilot lambda <- drop((24*lambda1 - 36*lambda2 + 6*lambda3 + 8*lambda4 - lambda5) %*% Sdr(d,r=4) %*% (vec(diag(d)) %x% vec(diag(d)))) RK <- (4*pi)^(-d/2) h.Ab <- (8*d*RK*fhat.pilot^(1+d/2)/lambda^2)^(1/(d+8))*n^(-1/(d+8)) fhat$estimate <- array(0, dim=dim(fhat$estimate)) xmin <- sapply(fhat$eval.points, min) xmax <- sapply(fhat$eval.points, max) if (verbose) { pb <- txtProgressBar(max=n) } for (i in 1:n) { if (verbose) setTxtProgressBar(pb, i) HAb <- h.Ab[i]^2*diag(d) fhat$estimate <- fhat$estimate + kde(x=matrix(x[i,], nrow=1), H=HAb, xmin=xmin, xmax=xmax, binned=binned, gridsize=dim(fhat$estimate), bgridsize=dim(fhat$estimate))$estimate } if (verbose) close(pb) fhat$estimate <- fhat$estimate/n if (pre) { ep <- cbind(fhat$eval.points[[1]], fhat$eval.points[[2]]) %*% S12 fhat$eval.points[[1]] <- ep[,1] fhat$eval.points[[2]] <- ep[,2] fhat$estimate <- fhat$estimate/det(S12) fhat$x <- x.orig } fhat$cont <- contourLevels(fhat, cont=1:99) fhat$H <- h.Ab^2 return(fhat) } ks/R/kdde.R0000644000176200001440000007347214576366115012155 0ustar liggesusers############################################################################### ## Multivariate kernel density derivative estimate ############################################################################### kdde <- function(x, H, h, deriv.order=0, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, deriv.vec=TRUE, verbose=FALSE) { ## default values r <- deriv.order ksd <- ks.defaults(x=x, w=w, binned=binned, bgridsize=bgridsize, gridsize=gridsize) d <- ksd$d; n <- ksd$n; w <- ksd$w binned <- ksd$binned gridsize <- ksd$gridsize bgridsize <- ksd$bgridsize ## clip data to xmin,xmax grid for binned estimation grid.clip <- binned if (grid.clip) { if (!missing(xmax)) xmax <- xmax[1:d] if (!missing(xmin)) xmin <- xmin[1:d] if (positive & missing(xmin)) { xmin <- rep(0,d) } xt <- truncate.grid(x=x, y=w, xmin=xmin, xmax=xmax) x <- xt$x; w <- xt$y; n <- length(w) } ## default bandwidths if (d==1 & missing(h)) { if (positive) x1 <- log(x) else x1 <- x h <- hpi(x=x1, nstage=2, binned=default.bflag(d=d, n=n), deriv.order=r) } if (d>1 & missing(H)) { if (positive) x1 <- log(x) else x1 <- x if ((r>0) & (d>2)) nstage <- 1 else nstage <- 2 H <- Hpi(x=x1, nstage=nstage, binned=default.bflag(d=d, n=n), deriv.order=r, verbose=verbose) } ## compute binned estimator if (binned) { if (positive & is.vector(x)) { y <- log(x) fhat <- kdde.binned(x=y, H=H, h=h, deriv.order=r, bgridsize=bgridsize, xmin=xmin, xmax=xmax, w=w) fhat$estimate <- fhat$estimate/exp(fhat$eval.points) fhat$eval.points <- exp(fhat$eval.points) fhat$x <- x } else fhat <- kdde.binned(x=x, H=H, h=h, deriv.order=r, bgridsize=bgridsize, xmin=xmin, xmax=xmax, w=w, deriv.vec=deriv.vec, verbose=verbose) if (!missing(eval.points)) { fhat$estimate <- predict(fhat, x=eval.points) fhat$eval.points <- eval.points } } else { ## compute exact (non-binned) estimator ## 1-dimensional if (d==1) { if (!missing(H) & !missing(h)) stop("Both H and h are both specified") if (missing(h)) h <- sqrt(H) if (missing(eval.points)) fhat <- kdde.grid.1d(x=x, h=h, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, deriv.order=r) else fhat <- kdde.points.1d(x=x, h=h, eval.points=eval.points, w=w, deriv.order=r) } ## multi-dimensional else { if (is.data.frame(x)) x <- as.matrix(x) if (missing(eval.points)) { if (d==2) fhat <- kdde.grid.2d(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, deriv.order=r, deriv.vec=deriv.vec, verbose=verbose) else if (d==3) fhat <- kdde.grid.3d(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, deriv.order=r, deriv.vec=deriv.vec, verbose=verbose) else stop("Need to specify eval.points for more than 3 dimensions") } else fhat <- kdde.points(x=x, H=H, eval.points=eval.points, w=w, deriv.order=r, deriv.vec=deriv.vec) } } fhat$binned <- binned fhat$names <- parse.name(x) fhat$type <- "kdde" class(fhat) <- "kdde" return(fhat) } ############################################################################### ## Multivariate binned kernel density derivative estimate ############################################################################### kdde.binned <- function(x, H, h, deriv.order, bgridsize, xmin, xmax, bin.par, w, deriv.vec=TRUE, deriv.index, verbose=FALSE) { r <- deriv.order if (length(r)>1) stop("deriv.order should be a non-negative integer") ## linear binning if (missing(bin.par)) { if (is.vector(x)) { d <- 1 } else { d <- ncol(x) } if (d==1) if (missing(H)) { H <- as.matrix(h^2) } else { h <- sqrt(H); H <- as.matrix(H) } if (d==1) Hd <- H else Hd <- diag(diag(H)) bin.par <- binning(x=x, H=Hd, h=h, bgridsize=bgridsize, xmin=xmin, xmax=xmax, supp=3.7+max(r), w=w) } else { if (!is.list(bin.par$eval.points)) { d <- 1; bgridsize <- length(bin.par$eval.points) } else { d <- length(bin.par$eval.points); bgridsize <- sapply(bin.par$eval.points, length) } w <- bin.par$w if (d==1) if (missing(H)) H <- as.matrix(h^2) else { h <- sqrt(H); H <- as.matrix(H) } } if (d==1) { fhat <- kdde.binned.1d(h=h, deriv.order=r, bin.par=bin.par) eval.points <- fhat$eval.points est <- fhat$estimate } else { ind.mat <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=H, deriv.order=r, only.index=TRUE, deriv.vec=deriv.vec) fhat.grid <- kdde.binned.nd(H=H, deriv.order=r, bin.par=bin.par, verbose=verbose, deriv.vec=deriv.vec) } if (missing(x)) x <- NULL if (d==1) { if (r==0) fhat <- list(x=x, eval.points=unlist(eval.points), estimate=est, h=h, H=h^2, gridtype="linear", gridded=TRUE, binned=TRUE, names=NULL, w=w) else fhat <- list(x=x, eval.points=unlist(eval.points), estimate=est, h=h, H=h^2, gridtype="linear", gridded=TRUE, binned=TRUE, names=NULL, w=w, deriv.order=r, deriv.ind=r) } else { if (r==0) fhat <- list(x=x, eval.points=fhat.grid$eval.points, estimate=fhat.grid$estimate[[1]], H=H, gridtype="linear", gridded=TRUE, binned=TRUE, names=NULL, w=w) else fhat <- list(x=x, eval.points=fhat.grid$eval.points, estimate=fhat.grid$estimate, H=H, gridtype="linear", gridded=TRUE, binned=TRUE, names=NULL, w=w, deriv.order=r, deriv.ind=ind.mat) } class(fhat) <- "kdde" return(fhat) } kdde.binned.1d <- function(h, deriv.order, bin.par) { r <- deriv.order n <- sum(bin.par$counts) a <- min(bin.par$eval.points) b <- max(bin.par$eval.points) M <- length(bin.par$eval.points) delta <- (b-a)/(M-1) L <- min(ceiling((4+r)*h/delta), M-1) N <- 2*L-1 grid1 <- seq(-(L-1), L-1) keval <- dnorm.deriv(x=delta*grid1, mu=0, sigma=h, deriv.order=r)/n est <- symconv.1d(keval, bin.par$counts) return(list(eval.points=bin.par$eval.points, estimate=est)) } kdde.binned.nd <- function(H, deriv.order, bin.par, verbose=FALSE, deriv.vec=TRUE) { d <- ncol(H) r <- deriv.order n <- sum(bin.par$counts) a <- sapply(bin.par$eval.points, min) b <- sapply(bin.par$eval.points, max) M <- sapply(bin.par$eval.points, length) delta <- (b-a)/(M-1) lambda <- max(sqrt(abs(svd(H)$d))) L <- pmin(ceiling((4+r)*lambda/delta), M-1) N <- 2*L-1 if (min(L)<=1) warning(paste("Binning grid may be too coarse for current (small) bandwidth: consider increasing grid size for dimensions", toString(which(pmin(L)<=1)))) if(d==2) { grid1 <- seq(-(L[1]-1), L[1]-1) grid2 <- seq(-(L[2]-1), L[2]-1) xgrid <- expand.grid(delta[1]*grid1, delta[2]*grid2) } else if (d==3) { grid1 <- seq(-(L[1]-1), L[1]-1) grid2 <- seq(-(L[2]-1), L[2]-1) grid3 <- seq(-(L[3]-1), L[3]-1) xgrid <- expand.grid(delta[1]*grid1, delta[2]*grid2, delta[3]*grid3) } else if (d==4) { grid1 <- seq(-(L[1]-1), L[1]-1) grid2 <- seq(-(L[2]-1), L[2]-1) grid3 <- seq(-(L[3]-1), L[3]-1) grid4 <- seq(-(L[4]-1), L[4]-1) xgrid <- expand.grid(delta[1]*grid1, delta[2]*grid2, delta[3]*grid3, delta[4]*grid4) } deriv.index <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=H, deriv.order=r, add.index=TRUE, only.index=TRUE, deriv.vec=TRUE) deriv.index.minimal <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=H, deriv.order=r, add.index=TRUE, only.index=TRUE, deriv.vec=FALSE) if (verbose) pb <- txtProgressBar() if (r==0) { n.seq <- block.indices(1, nrow(xgrid), d=d, r=r, diff=FALSE) est.list <- vector(1, mode="list") est.list[[1]] <- array(0, dim=dim(bin.par$counts)) } else if (r>0) { n.deriv <- nrow(deriv.index) n.deriv.minimal <- nrow(deriv.index.minimal) if (deriv.vec) n.est.list <- n.deriv else n.est.list <- n.deriv.minimal est.list <- vector(n.est.list, mode="list") for (j in 1:n.est.list) est.list[[j]] <- array(0, dim=dim(bin.par$counts)) if (d^r >= 3^7) n.seq <- block.indices(1, nrow(xgrid), d=d, r=r, diff=FALSE, block.limit=1e4) else n.seq <- block.indices(1, nrow(xgrid), d=d, r=r, diff=FALSE, block.limit=1e5) } for (i in 1:(length(n.seq)-1)) { if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) keval <- dmvnorm.deriv(x=xgrid[n.seq[i]:(n.seq[i+1]-1),], mu=rep(0,d), Sigma=H, deriv.order=r, add.index=TRUE, deriv.vec=FALSE)$deriv/n if (r==0) keval <- as.matrix(keval, ncol=1) else if (is.vector(keval)) keval <- as.matrix(t(keval), nrow=1) est <- list() ## loop over only unique partial derivative indices nderiv <- nrow(deriv.index.minimal) if (!(is.null(nderiv))) for (s in 1:nderiv) { if (deriv.vec) deriv.rep.index <- which.mat(deriv.index.minimal[s,], deriv.index) else deriv.rep.index <- s kevals <- array(keval[,s], dim=N) est.temp <- symconv.nd(kevals, bin.par$counts, d=d) for (s2 in 1:length(deriv.rep.index)) est[[deriv.rep.index[s2]]] <- est.temp } else { kevals <- lapply(as.data.frame(keval), function(x) { array(x, dim=N) }) est <- lapply(kevals, function(x) { symconv.nd(x, bin.par$counts, d=d) }) } if (r==0) est.list[[1]] <- est.list[[1]] + est[[1]] else if (r>0) for (j in 1:n.est.list) est.list[[j]] <- est.list[[j]] + est[[j]] } if (verbose) close(pb) fhatr <- list(eval.points=bin.par$eval.points, estimate=est.list, deriv.order=r) return(fhatr) } ############################################################################# ## Univariate kernel density derivative estimate on a grid ############################################################################# kdde.grid.1d <- function(x, h, gridsize, supp=3.7, positive=FALSE, adj.positive, xmin, xmax, gridtype, w, deriv.order=0) { r <- deriv.order if (r==0) fhatr <- kde(x=x, h=h, gridsize=gridsize, supp=supp, positive=positive, adj.positive=adj.positive, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w) else { if (missing(xmin)) xmin <- min(x) - h*supp if (missing(xmax)) xmax <- max(x) + h*supp if (missing(gridtype)) gridtype <- "linear" y <- x gridtype1 <- match.arg(gridtype, c("linear", "sqrt")) if (gridtype1=="linear") gridy <- seq(xmin, xmax, length=gridsize) else if (gridtype1=="sqrt") { gridy.temp <- seq(sign(xmin)*sqrt(abs(xmin)), sign(xmax)*sqrt(abs(xmax)), length=gridsize) gridy <- sign(gridy.temp) * gridy.temp^2 } gridtype.vec <- gridtype1 n <- length(y) est <- dnorm.deriv.mixt(x=gridy, mus=y, sigmas=rep(h, n), props=w/n, deriv.order=r) fhatr <- list(x=y, eval.points=gridy, estimate=est, h=h, H=h^2, gridtype=gridtype.vec, gridded=TRUE, binned=FALSE, names=NULL, w=w, deriv.order=r, deriv.ind=deriv.order) class(fhatr) <- "kdde" } return(fhatr) } ############################################################################## ## Bivariate kernel density derivative estimate on a grid ## Computes all mixed partial derivatives for a given deriv.order ############################################################################## kdde.grid.2d <- function(x, H, gridsize, supp, gridx=NULL, grid.pts=NULL, xmin, xmax, gridtype, w, deriv.order=0, deriv.vec=TRUE, verbose=FALSE) { d <- 2 r <- deriv.order if (r==0) fhatr <- kde(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, verbose=verbose) else { ## initialise grid n <- nrow(x) if (is.null(gridx)) gridx <- make.grid.ks(x, matrix.sqrt(H), tol=supp, gridsize=gridsize, xmin=xmin, xmax=xmax, gridtype=gridtype) suppx <- make.supp(x, matrix.sqrt(H), tol=supp) if (is.null(grid.pts)) grid.pts <- find.gridpts(gridx, suppx) nderiv <- d^r fhat.grid <- list() for (k in 1:nderiv) fhat.grid[[k]] <- matrix(0, nrow=length(gridx[[1]]), ncol=length(gridx[[2]])) if (verbose) pb <- txtProgressBar() for (i in 1:n) { ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.x.len <- length(eval.x) eval.pts <- expand.grid(eval.x, eval.y) ## Create list of matrices for different partial derivatives fhat <- dmvnorm.deriv(x=eval.pts, mu=x[i,], Sigma=H, deriv.order=r) ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (k in 1:nderiv) for (j in 1:length(eval.y)) fhat.grid[[k]][eval.x.ind, eval.y.ind[j]] <- fhat.grid[[k]][eval.x.ind, eval.y.ind[j]] + w[i]*fhat[((j-1) * eval.x.len + 1):(j * eval.x.len),k] if (verbose) setTxtProgressBar(pb, i/n) } if (verbose) close(pb) for (k in 1:nderiv) fhat.grid[[k]] <- fhat.grid[[k]]/n gridx1 <- list(gridx[[1]], gridx[[2]]) ind.mat <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=H, deriv.order=r, only.index=TRUE) if (!deriv.vec) { fhat.grid.vech <- list() deriv.ind <- unique(ind.mat) for (i in 1:nrow(deriv.ind)) { which.deriv <- which.mat(deriv.ind[i,], ind.mat)[1] fhat.grid.vech[[i]] <- fhat.grid[[which.deriv]] } ind.mat <- deriv.ind fhat.grid <- fhat.grid.vech } fhatr <- list(x=x, eval.points=gridx1, estimate=fhat.grid, H=H, gridtype=gridx$gridtype, gridded=TRUE, binned=FALSE, names=NULL, w=w, deriv.order=deriv.order, deriv.ind=ind.mat) } return(fhatr) } ############################################################################## ## Trivariate kernel density derivative estimate on a grid ## Computes all mixed partial derivatives for a given deriv.order ############################################################################## kdde.grid.3d <- function(x, H, gridsize, supp, gridx=NULL, grid.pts=NULL, xmin, xmax, gridtype, w, deriv.order=0, deriv.vec=TRUE, verbose=FALSE) { d <- 3 r <- deriv.order if (r==0) fhatr <- kde(x=x, H=H, gridsize=gridsize, supp=supp, xmin=xmin, xmax=xmax, gridtype=gridtype, w=w, verbose=verbose) else { ## initialise grid n <- nrow(x) if (is.null(gridx)) gridx <- make.grid.ks(x, matrix.sqrt(H), tol=supp, gridsize=gridsize, xmin=xmin, xmax=xmax, gridtype=gridtype) suppx <- make.supp(x, matrix.sqrt(H), tol=supp) if (is.null(grid.pts)) grid.pts <- find.gridpts(gridx, suppx) nderiv <- d^r fhat.grid <- list() for (k in 1:nderiv) fhat.grid[[k]] <- array(0, dim=gridsize) if (verbose) pb <- txtProgressBar() for (i in 1:n) { ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.z <- gridx[[3]][grid.pts$xmin[i,3]:grid.pts$xmax[i,3]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.z.ind <- c(grid.pts$xmin[i,3]:grid.pts$xmax[i,3]) eval.x.len <- length(eval.x) eval.pts <- expand.grid(eval.x, eval.y) ## create list of matrices for different partial derivatives ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (ell in 1:nderiv) for (k in 1:length(eval.z)) { fhat <- w[i]*dmvnorm.deriv(cbind(eval.pts, eval.z[k]), x[i,], H, deriv.order=r) for (j in 1:length(eval.y)) fhat.grid[[ell]][eval.x.ind,eval.y.ind[j], eval.z.ind[k]] <- fhat.grid[[ell]][eval.x.ind, eval.y.ind[j], eval.z.ind[k]] + fhat[((j-1) * eval.x.len + 1):(j * eval.x.len), ell] } if (verbose) setTxtProgressBar(pb, i/n) } if (verbose) close(pb) for (k in 1:nderiv) fhat.grid[[k]] <- fhat.grid[[k]]/n gridx1 <- list(gridx[[1]], gridx[[2]], gridx[[3]]) ind.mat <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=H, deriv.order=r, only.index=TRUE) if (!deriv.vec) { fhat.grid.vech <- list() deriv.ind <- unique(ind.mat) for (i in 1:nrow(deriv.ind)) { which.deriv <- which.mat(deriv.ind[i,], ind.mat)[1] fhat.grid.vech[[i]] <- fhat.grid[[which.deriv]] } ind.mat <- deriv.ind fhat.grid <- fhat.grid.vech } fhatr <- list(x=x, eval.points=gridx1, estimate=fhat.grid, H=H, gridtype=gridx$gridtype, gridded=TRUE, binned=FALSE, names=NULL, w=w, deriv.order=deriv.order, deriv.ind=ind.mat) } return(fhatr) } ############################################################################# ## Multivariate kernel density estimate using normal kernels, ## evaluated at each sample point ############################################################################# kdde.points.1d <- function(x, h, eval.points, w, deriv.order=0) { r <- deriv.order n <- length(x) fhat <- dnorm.deriv.mixt(x=eval.points, mus=x, sigmas=rep(h,n), props=w/n, deriv.order=r) return(list(x=x, eval.points=eval.points, estimate=fhat, h=h, H=h^2, gridded=FALSE, binned=FALSE, names=NULL, w=w, deriv.order=r, deriv.ind=r)) } kdde.points <- function(x, H, eval.points, w, deriv.order=0, deriv.vec=TRUE) { n <- nrow(x) Hs <- replicate(n, H, simplify=FALSE) Hs <- do.call(rbind, Hs) r <- deriv.order fhat <- dmvnorm.deriv.mixt(x=eval.points, mus=x, Sigmas=Hs, props=w/n, deriv.order=r, deriv.vec=deriv.vec, add.index=TRUE) return(list(x=x, eval.points=eval.points, estimate=fhat$deriv, H=H, gridded=FALSE, binned=FALSE, names=NULL, w=w, deriv.order=r, deriv.ind=fhat$deriv.ind)) } ############################################################################# ## S3 methods for KDDE objects ############################################################################# ## plot method plot.kdde <- function(x, ...) { fhat <- x if (is.null(fhat$deriv.order)) { class(fhat) <- "kde" plot(fhat, ...) } else { if (is.vector(fhat$x)) { plotkdde.1d(fhat, ...) invisible() } else { d <- ncol(fhat$x) if (d==2) { opr <- options()$preferRaster; if (!is.null(opr)) if (!opr) options("preferRaster"=TRUE) plotret <- plotkdde.2d(fhat, ...) if (!is.null(opr)) options("preferRaster"=opr) invisible(plotret) } else if (d==3) { plotkdde.3d(fhat, ...) invisible() } else stop ("Plot function only available for 1, 2 or 3-d data") } } } plotkdde.1d <- function(fhat, xlab, ylab="Density derivative function", cont=50, abs.cont, ...) { if (missing(xlab)) xlab <- fhat$names[1] if (missing(abs.cont)) { abs.cont <- as.matrix(contourLevels(fhat, approx=TRUE, cont=cont), ncol = length(cont)) abs.cont <- c(abs.cont[1, ], rev(abs.cont[2, ])) } class(fhat) <- "kde" plot(fhat, xlab=xlab, ylab=ylab, abs.cont=abs.cont, ...) } plotkdde.2d <- function(fhat, which.deriv.ind=1, cont=c(25,50,75), abs.cont, display="slice", xlab, ylab, zlab="Density derivative function", col, col.fun, alpha=1, kdde.flag=TRUE, thin=3, transf=1, neg.grad=FALSE, ...) { disp <- match.arg(display, c("slice", "persp", "image", "filled.contour", "filled.contour2", "quiver")) if (disp=="filled.contour2") disp <- "filled.contour" if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="Blue-Red", alpha=alpha) } if (missing(xlab)) xlab <- fhat$names[1] if (missing(ylab)) ylab <- fhat$names[2] if (disp=="slice" | disp=="filled.contour") { if (missing(abs.cont)) { abs.cont <- as.matrix(contourLevels(fhat, approx=TRUE, cont=cont, which.deriv.ind=which.deriv.ind), ncol=length(cont)) abs.cont <- c(abs.cont[1,], rev(abs.cont[2,])) } if (missing(col)) { if (disp=="slice") col <- col.fun(n=length(abs.cont)) else if (disp=="filled.contour") { col <- col.fun(n=length(abs.cont)+1) col[median(1:length(col))] <- "transparent" } } } if (disp=="quiver") { if (fhat$deriv.order==1) plotquiver(fhat=fhat, thin=thin, transf=transf, neg.grad=neg.grad, col=col, xlab=xlab, ylab=ylab, alpha=alpha, ...) else warning("Quiver plot requires gradient estimate.") } else { fhat.temp <- fhat fhat.temp$deriv.ind <- fhat.temp$deriv.ind[which.deriv.ind,] fhat.temp$estimate <- fhat.temp$estimate[[which.deriv.ind]] fhat <- fhat.temp class(fhat) <- "kde" if (disp=="persp") plot(fhat, display=display, abs.cont=abs.cont, xlab=xlab, ylab=ylab, zlab=zlab, col.fun=col.fun, kdde.flag=kdde.flag, col=col, thin=thin, alpha=alpha, ...) else plot(fhat, display=display, abs.cont=abs.cont, xlab=xlab, ylab=ylab, zlab=zlab, col.fun=col.fun, kdde.flag=kdde.flag, col=col, alpha=alpha, ...) } } plotkdde.3d <- function(fhat, which.deriv.ind=1, display="plot3D", cont=c(25,50,75), abs.cont, colors, col, col.fun, xlab, ylab, zlab, ...) { if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="Blue-Red") } if (missing(xlab)) xlab <- fhat$names[1] if (missing(ylab)) ylab <- fhat$names[2] if (missing(zlab)) ylab <- fhat$names[3] if (missing(abs.cont)) { abs.cont <- as.matrix(contourLevels(fhat, approx=TRUE, cont=cont, which.deriv.ind=which.deriv.ind), ncol=length(cont)) abs.cont <- c(abs.cont[1,], rev(abs.cont[2,])) } fhat.temp <- fhat fhat.temp$deriv.ind <- fhat.temp$deriv.ind[which.deriv.ind,] fhat.temp$estimate <- fhat.temp$estimate[[which.deriv.ind]] fhat <- fhat.temp class(fhat) <- "kde" if (missing(col)) { col <- col.fun(n=length(abs.cont)+1) nc <- length(col) col <- col[-median(1:nc)] } colors <- col plot(fhat, display=display, abs.cont=abs.cont, colors=col, col=col, xlab=xlab, ylab=ylab, zlab=zlab, ...) } ###################################################################### ## Quiver plot ###################################################################### plotquiver <- function(fhat, thin=5, transf=1, neg.grad=FALSE, xlab, ylab, col, add=FALSE, scale, length=0.1, alpha=1, ...) { if (!requireNamespace("pracma", quietly=TRUE)) stop("Install the pracma package as it is required.", call.=FALSE) if (missing(col)) col <- 1 col <- transparency.col(col, alpha=alpha) ev <- fhat$eval.points est <- fhat$estimate if (transf!=0) { est[[1]] <- sign(est[[1]])*abs(est[[1]])^(transf) est[[2]] <- sign(est[[2]])*abs(est[[2]])^(transf) } thin1.ind <- seq(1, length(ev[[1]]), by=thin) thin2.ind <- seq(1, length(ev[[2]]), by=thin) evx <- ev[[1]][thin1.ind] evy <- ev[[2]][thin2.ind] fx <- est[[1]][thin1.ind, thin2.ind] fy <- est[[2]][thin1.ind, thin2.ind] if (neg.grad) { fx <- -fx; fy <- -fy } if (missing(xlab)) xlab <- fhat$names[1] if (missing(ylab)) ylab <- fhat$names[2] if (!add) plot(fhat, abs.cont=c(0,0), col="transparent", xlab=xlab, ylab=ylab) grid.xy <- pracma::meshgrid(evy, evx) x0 <- grid.xy$Y y0 <- grid.xy$X ## default scale factor - arrows should exceed a bin arrow.len <- sqrt(fx^2 + fy^2) bin.diag <- min(diff(evx), diff(evy)) if (missing(scale)) scale <- bin.diag/max(arrow.len) ## remove `zero-length' arrows i.e. length < 1e-3 inches ## adapted from https://stackoverflow.com/questions/52689959/how-small-is-a-zero-length-arrow/52690054 units <- par(c('usr', 'pin')) x2in <- with(units, pin[1L]/diff(usr[1:2])) y2in <- with(units, pin[2L]/diff(usr[3:4])) arrow2in <- sqrt((x2in*scale*fx)^2 + (y2in*scale*fy)^2) arrowint <- sort(c(min(arrow2in)-0.1*abs(max(arrow2in)), seq(1e-3, length, length=5), max(arrow2in)+0.1*abs(max(arrow2in)))) nsmall.f <- cut(arrow2in, arrowint, labels=FALSE) nsf.lab <- unique(nsmall.f[nsmall.f>1]) for (i in nsf.lab) { nsf <- which(nsmall.f==i) if (length(nsf)>0) pracma::quiver(x=x0[nsf], y=y0[nsf], u=fx[nsf], v=fy[nsf], col=col, scale=scale, length=arrowint[i], ...) } } ## contourLevels method contourLevels.kdde <- function(x, prob, cont, nlevels=5, approx=TRUE, which.deriv.ind=1, ...) { fhat <- x if (is.vector(fhat$x)) { d <- 1; n <- length(fhat$x) if (!is.null(fhat$deriv.order)) { fhat.temp <- fhat fhat.temp$deriv.ind <-fhat.temp$deriv.ind[which.deriv.ind] fhat <- fhat.temp } } else { d <- ncol(fhat$x); n <-nrow(fhat$x) if (!is.matrix(fhat$x)) fhat$x <- as.matrix(fhat$x) if (!is.null(fhat$deriv.order)) { fhat.temp <- fhat fhat.temp$estimate <- fhat.temp$estimate[[which.deriv.ind]] fhat.temp$deriv.ind <-fhat.temp$deriv.ind[which.deriv.ind,] fhat <- fhat.temp } } if (is.null(x$w)) w <- rep(1, n) else w <- x$w if (is.null(fhat$gridded)) { if (d==1) fhat$gridded <- fhat$binned else fhat$gridded <- is.list(fhat$eval.points) } if (missing(prob) & missing(cont)) hts <- pretty(fhat$estimate, n=nlevels) else { if (approx & fhat$gridded) dobs <- predict.kde(fhat, x=fhat$x) else dobs <- kdde(x=fhat$x, H=fhat$H, eval.points=fhat$x, w=w, deriv.order=fhat$deriv.order)$estimate[,which.deriv.ind] if (is.null(fhat$deriv.order)) { if (!missing(prob) & missing(cont)) hts <- quantile(dobs[dobs>=0], prob=prob) if (missing(prob) & !missing(cont)) hts <- quantile(dobs[dobs>=0], prob=(100-cont)/100) } else { if (!missing(prob) & missing(cont)) hts <- rbind(-quantile(abs(dobs[dobs<0]), prob=prob), quantile(dobs[dobs>=0], prob=prob)) if (missing(prob) & !missing(cont)) hts <- rbind(-quantile(abs(dobs[dobs<0]), prob=(100-cont)/100), quantile(dobs[dobs>=0], prob=(100-cont)/100)) } } return(hts) } ## predict method for KDDE predict.kdde <- function(object, ..., x) { fhat <- object if (is.vector(fhat$H)) d <- 1 else d <- ncol(fhat$H) if (d==1) n <- length(x) else { if (is.vector(x)) x <- matrix(x, nrow=1) else x <- as.matrix(x) n <- nrow(x) } if (!is.null(fhat$deriv.ind)) { if (is.vector(fhat$deriv.ind)) pk.mat <- predict.kde(fhat, x=x, ...) else { nd <- nrow(fhat$deriv.ind) pk.mat <- matrix(0, ncol=nd, nrow=n) for (i in 1:nd) { fhat.temp <- fhat fhat.temp$estimate <- fhat$estimate[[i]] pk.mat[,i] <- predict.kde(fhat.temp, x=x, ...) } } } else pk.mat <- predict.kde(fhat, x=x, ...) return(drop(pk.mat)) } ###################################################################### ## Summary kernel curvature ###################################################################### kcurv <- function(fhat, compute.cont=TRUE) { fhat.curv <- fhat if (is.vector(fhat$H)) d <- 1 else d <- ncol(fhat$H) if (fhat$deriv.order!=2) stop("Requires output from kdde(, deriv.order=2).") if (d==1) { Hessian.det <- fhat$estimate local.mode <- fhat$estimate <0 fhat.curv$estimate <- local.mode*abs(Hessian.det) } else if (d>1) { fhat.est <- sapply(fhat$estimate, as.vector) Hessian.det <- sapply(seq(1,nrow(fhat.est)), function(i) { det(invvec(fhat.est[i,])) }) Hessian.eigen <- lapply(lapply(seq(1,nrow(fhat.est)), function(i) { invvec(fhat.est[i,]) }), eigen, only.values=TRUE) Hessian.eigen <- t(sapply(Hessian.eigen, getElement, "values")) local.mode <- apply(Hessian.eigen <= 0, 1, all) fhat.curv$estimate <- local.mode*array(abs(Hessian.det), dim=dim(fhat$estimate[[1]])) } fhat.curv$deriv.order <- NULL fhat.curv$deriv.ind <- NULL if (compute.cont) { fhat.temp <- fhat.curv fhat.temp$x <- fhat.curv$x[predict(fhat.curv, x=fhat.curv$x)>0,] fhat.temp$estimate <- fhat.temp$estimate fhat.curv$cont <- contourLevels(fhat.temp, cont=1:99) } class(fhat.curv) <- "kde" fhat.curv$type <- "kcurv" return(fhat.curv) } ks/R/integrate-kde.R0000644000176200001440000000721614457752752013766 0ustar liggesusers############################################################################# ## Cumulative integral for KDE ############################################################################# integral.kde <- function(q, fhat, density) { gridsize <- length(fhat$eval.points) ## Use Simpson's rule to compute numerical integration simp.rule <- rep(0, gridsize-1) for (i in 1:(gridsize-1)) { del <- fhat$eval.points[i+1] - fhat$eval.points[i] simp.rule[i] <- min(fhat$estimate[i], fhat$estimate[i+1])*del + 1/2*abs(fhat$estimate[i+1] - fhat$estimate[i])*del } ## add last incomplete trapezoid q.ind <- findInterval(x=q, vec=fhat$eval.points) q.prob <- rep(0, length(q)) i <- 0 for (qi in q.ind) { i <- i+1 if (qi==0) q.prob[i] <- 0 else if (qi < gridsize) { ## linearly interpolate kde fhat.estqi <- (fhat$est[qi+1] - fhat$est[qi])/(fhat$eval[qi+1] - fhat$eval[qi]) * (q[i] - fhat$eval[qi]) + fhat$est[qi] delqi <- q[i] - fhat$eval[qi] simp.ruleqi <- min(fhat.estqi, fhat$est[qi])*delqi + 1/2*abs(fhat.estqi - fhat$est[qi])*delqi q.prob[i] <- sum(simp.rule[1:qi]) + simp.ruleqi } else { if (density) q.prob[i] <- 1 else q.prob[i] <- sum(simp.rule) } } if (density) q.prob[q.prob>=1] <- 1 ## remove possible decreasing values in q.prob dec.ind <- which(diff(q.prob)<0) if (length(dec.ind)>0) { dec.ind <- dec.ind+1 for (i in dec.ind) q.prob[i] <- q.prob[i-1] } return(q.prob) } ## cumulative probability P(fhat <= q) pkde <- function(q, fhat) { return(integral.kde(q=q, fhat=fhat, density=TRUE)) } ## density value of KDE at x ## alias for predict.kde dkde <- function(x, fhat) { return(predict(fhat, x=x)) } ## p-quantile of KDE, i.e. solve for x where P(fhat < x) = p qkde <- function(p, fhat) { if (any(p > 1) | any(p < 0)) stop("p must be <= 1 and >= 0") cumul.prob <- pkde(q=fhat$eval.points, fhat=fhat) ind <- findInterval(x=p, vec=cumul.prob) quant <- rep(0, length(ind)) for (j in 1:length(ind)) { i <- ind[j] if (i==0) quant[j] <- fhat$eval.points[1] else if (i>=length(fhat$eval.points)) quant[j] <- fhat$eval.points[length(fhat$eval.points)] else { quant1 <- fhat$eval.points[i] quant2 <- fhat$eval.points[i+1] prob1 <- cumul.prob[i] prob2 <- cumul.prob[i+1] alpha <- (p[j] - prob2)/(prob1 - prob2) quant[j] <- quant1*alpha + quant2*(1-alpha) } } return(quant) } ## Silverman (1983)'s random sample from KDE rkde <- function(n, fhat, positive=FALSE) { if (positive) x <- log(fhat$x) else x <- fhat$x if (is.vector(fhat$H)) { d <- 1; nsamp <- length(x) } else { d <- ncol(fhat$H); nsamp <- nrow(x) } x.ind <- sample(1:nsamp, size=n, replace=TRUE, prob=fhat$w) if (d==1) { h <- fhat$h rkde.val <- x[x.ind] + h*rnorm(n=n, mean=0, sd=1) if (positive) rkde.val <- exp(rkde.val) } else if (d>1) { H <- fhat$H rkde.val <- x[x.ind,] + rmvnorm(n=n, mean=rep(0,d), sigma=diag(d)) %*% matrix.sqrt(H) } return(rkde.val) } ## plot cumulative probability as shaded region on a KDE plotkde.cumul <- function(fhat, q, add=FALSE, col="blue", ...) { qind <- fhat$eval.points<=q n <- sum(qind) if (!add) plot(fhat) polygon(c(fhat$eval.points[qind],fhat$eval.points[n]), c(fhat$estimate[qind],0), col=col, ...) } ks/R/normal.R0000644000176200001440000017766214336776167012552 0ustar liggesusers############################################################################### ## Univariate mixture normal densities and derivatives ############################################################################### rnorm.mixt <- function(n=100, mus=0, sigmas=1, props=1, mixt.label=FALSE) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") ## single component mixture if (identical(all.equal(props[1], 1), TRUE)) { if (mixt.label) rand <- cbind(rnorm(n=n, mean=mus, sd=sigmas), rep(1, n)) else rand <- rnorm(n=n, mean=mus, sd=sigmas) } ## multiple component mixture else { k <- length(props) n.samp <- sample(1:k, n, replace=TRUE, prob=props) n.prop <- numeric(0) ## compute number taken from each mixture for (i in 1:k) n.prop <- c(n.prop, sum(n.samp == i)) rand <- numeric(0) for (i in 1:k) { ## compute random sample from normal mixture component if (n.prop[i] > 0) if (mixt.label) rand <- rbind(rand, cbind(rnorm(n=n.prop[i], mean=mus[i], sd=sigmas[i]), rep(i, n.prop[i]))) else rand <- c(rand, rnorm(n=n.prop[i], mean=mus[i], sd=sigmas[i])) } } if (mixt.label) return(rand[sample(n),]) else return(rand[sample(n)]) } dnorm.mixt <- function(x, mus=0, sigmas=1, props=1) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") ## single component mixture if (identical(all.equal(props[1], 1), TRUE)) dens <- dnorm(x, mean=mus[1], sd=sigmas[1]) ## multiple component mixture else { k <- length(props) dens <- 0 ## sum of each normal density value from each component at x for (i in 1:k) dens <- dens + props[i]*dnorm(x, mean=mus[i], sd=sigmas[i]) } return(dens) } ############################################################################### ## Derivatives of the univariate normal ## (code: J.E.Chacon 08/06/2018) ## ## Parameters ## x - points to evaluate at ## sigma - std deviation ## r - derivative index # ## Returns ## r-th derivative at x ############################################################################### dnorm.deriv <- function (x, mu=0, sigma=1, deriv.order=0) { r <- deriv.order phi <- dnorm(x, mean=mu, sd=sigma) x <- (x - mu) arg <- x / sigma hmold0 <- 1 hmold1 <- arg hmnew <- 1 if (r == 1) hmnew <- hmold1 if (r >= 2) for (i in (2:r)) { hmnew <- arg * hmold1 - (i - 1) * hmold0 hmold0 <- hmold1 hmold1 <- hmnew } derivt <- (-1)^r * phi * hmnew / sigma^r return(derivt) } ############################################################################### ## Double sum of K(X_i - X_j) used in density derivative estimation # ## Parameters ## x - points to evaluate ## Sigma - variance matrix ## inc - 0 - exclude diagonals ## - 1 - include diagonals # ## Returns ## Double sum at x ############################################################################### dnorm.deriv.sum <- function(x, sigma, deriv.order, inc=1, binned=FALSE, bin.par, kfe=FALSE) { r <- deriv.order n <- length(x) if (binned) { if (missing(bin.par)) bin.par <- binning(x, h=sigma, supp=4+r) est <- kdde.binned(x=x, H=sigma^2, h=sigma, deriv.order=r, bin.par=bin.par)$estimate sumval <- sum(bin.par$counts*est*n) if (inc == 0) sumval <- sumval - n*dnorm.deriv(x=0, mu=0, sigma=sigma, deriv.order=r) } else { sumval <- 0 for (i in 1:n) sumval <- sumval + sum(dnorm.deriv(x=x[i] - x, mu=0, sigma=sigma, deriv.order=r)) if (inc == 0) sumval <- sumval - n*dnorm.deriv(x=0, mu=0, sigma=sigma, deriv.order=r) } if (kfe) if (inc==1) sumval <- sumval/n^2 else sumval <- sumval/(n*(n-1)) return(sumval) } dnorm.deriv.mixt <- function(x, mus=0, sigmas=1, props=1, deriv.order=0) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") ## single component mixture if (identical(all.equal(props[1], 1), TRUE)) dens <- dnorm.deriv(x, mu=mus[1], sigma=sigmas[1], deriv.order=deriv.order) ## multiple component mixture else { k <- length(props) dens <- 0 ## sum of each normal density value from each component at x for (i in 1:k) dens <- dens + props[i]*dnorm.deriv(x=x, mu=mus[i], sigma=sigmas[i], deriv.order=deriv.order) } return(dens) } ############################################################################### ## Multivariate normal densities and derivatives ############################################################################### ############################################################################### ## Multivariate normal mixture - random sample ## ## Parameters ## n - number of samples ## mus - matrix of means (each row is a vector of means from each component ## density) ## Sigmas - matrix of covariance matrices (every d rows is a covariance matrix ## from each component density) ## props - vector of mixing proportions ## ## Returns ## Vector of n observations from the normal mixture ############################################################################### rmvnorm.mixt <- function(n=100, mus=c(0,0), Sigmas=diag(2), props=1, mixt.label=FALSE) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") ## single component mixture if (identical(all.equal(props[1], 1), TRUE)) if (mixt.label) rand <- cbind(mvtnorm::rmvnorm(n=n, mean=mus, sigma=Sigmas), rep(1, n)) else rand <- cbind(mvtnorm::rmvnorm(n=n, mean=mus, sigma=Sigmas)) ## multiple component mixture else { k <- length(props) d <- ncol(Sigmas) n.samp <- sample(1:k, n, replace=TRUE, prob=props) n.prop <- numeric(0) ## compute number taken from each mixture for (i in 1:k) n.prop <- c(n.prop, sum(n.samp == i)) rand <- numeric(0) for (i in 1:k) { ## compute random sample from normal mixture component if (n.prop[i] > 0) { if (mixt.label) rand <- rbind(rand, cbind(rmvnorm(n=n.prop[i], mean=mus[i,], sigma=Sigmas[((i-1)*d+1) : (i*d),]), rep(i, n.prop[i]))) else rand <- rbind(rand, rmvnorm(n=n.prop[i], mean=mus[i,], sigma=Sigmas[((i-1)*d+1) : (i*d),])) } } } return(rand[sample(n),]) } ############################################################################### ## Multivariate normal mixture - density values ## ## Parameters ## x - points to compute density at ## mus - matrix of means ## Sigmas - matrix of covariance matrices ## props - vector of mixing proportions ## ## Returns ## Density values from the normal mixture (at x) ############################################################################### dmvnorm.mixt <- function(x, mus, Sigmas, props=1, verbose=FALSE) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") if (is.vector(x)) { d <- length(x); n <- 1 } else { d <- ncol(x); n <- nrow(x) } if (missing(mus)) mus <- rep(0,d) if (missing(Sigmas)) Sigmas <- diag(d) ## single component mixture if (identical(all.equal(props[1], 1), TRUE)) { if (is.matrix(mus)) mus <- mus[1,] dens <- dmvnorm(x=x, mean=mus, sigma=Sigmas[1:d,]) } ## multiple component mixture else { if (verbose) pb <- txtProgressBar() k <- length(props) dens <- 0 ## sum of each normal density value from each component at x for (i in 1:k) { dens <- dens + props[i]*dmvnorm(x=x, mean=mus[i,], sigma=Sigmas[((i-1)*d+1):(i*d),]) if (verbose) setTxtProgressBar(pb, i/k) } if (verbose) close(pb) } return(dens) } ########################################################################## ## Computation of the r-th derivative vector of the Gaussian density ########################################################################## dmvnorm.deriv <- function(x, mu, Sigma, deriv.order=0, deriv.vec=TRUE, add.index=FALSE, only.index=FALSE, type="unique") { type1 <- match.arg(type, c("recursive", "direct", "unique")) r <- deriv.order if(length(r)>1) stop("deriv.order should be a non-negative integer") sumr <- sum(r) if (missing(x)) d <- ncol(Sigma) else { if (is.vector(x)) x <- t(as.matrix(x)) if (is.data.frame(x)) x <- as.matrix(x) d <- ncol(x) n <- nrow(x) } if (add.index | only.index | !deriv.vec) { ## matrix of derivative indices ind.mat <- 0 sumr.counter <- sumr if (sumr>=1) ind.mat <- diag(d) { while (sumr.counter >1) { ind.mat <- Ksum(diag(d), ind.mat) sumr.counter <- sumr.counter - 1 } } ind.mat.minimal <- unique(ind.mat) ind.mat.minimal.logical <- !duplicated(ind.mat) if (only.index) if (deriv.vec) return (ind.mat) else return(ind.mat.minimal) } if (missing(mu)) mu <- rep(0,d) if (missing(Sigma)) Sigma <- diag(d) x.centred <- sweep(x, 2, mu) dens <- do.call(paste("dmvnorm.deriv", type1, sep="."), list(x=x.centred, Sigma=Sigma, deriv.order=r)) if (is.vector(dens) & r>0) dens <- matrix(dens, nrow=1) if (!deriv.vec & r>0) { ind.select <- numeric() for (i in 1:nrow(ind.mat.minimal)) ind.select <- c(ind.select, head(which.mat(ind.mat.minimal[i,], ind.mat), n=1)) dens <- dens[,ind.select] ind.mat <- ind.mat.minimal } if (add.index) return(list(deriv=dens, deriv.ind=ind.mat)) else return(dens) } ############################################################################ ## dmvnorm.deriv.direct computes the vector derivative of the Gaussian ## density phi_Sigma(x) on the basis of Equation (1) and Algotihm 2, as ## described in Chacon and Duong (2014) ############################################################################ dmvnorm.deriv.direct <- function(x,Sigma,deriv.order=0) { if (is.vector(x)) { x <- matrix(x,nrow=1) } d <- ncol(Sigma) n <- nrow(x) r <- deriv.order Sigmainv <- chol2inv(chol(Sigma)) Hermx <- matrix(0,nrow=n,ncol=d^r) for (i in 1:n) { for (j in 0:floor(r/2)) Hermx[i,] <- Hermx[i,] + (-1)^j/(factorial(j)*factorial(r-2*j)*2^j)* Kpow(Sigmainv%*%x[i,], r-2*j)%x%Kpow(vec(Sigmainv),j) Hermx[i,] <- Sdrv.recursive(d=d,r=r,v=Hermx[i,]) } dens <- drop((-1)^r*factorial(r)*Hermx*dmvnorm(x,mean=rep(0,d), sigma=Sigma)) return(dens) } ############################################################################ ## dmvnorm.deriv.recursive computes the vector derivative of the Gaussian ## density phi_Sigma(x) on the basis of Equation (7) and Algorithm 2 as ## described in Section 5 of Chacon and Duong (2014) ############################################################################ dmvnorm.deriv.recursive <- function(x,Sigma,deriv.order=0) { if (is.vector(x)) { x <- matrix(x,nrow=1) } d <- ncol(Sigma) n <- nrow(x) r <- deriv.order G <- Sigma Ginv <- chol2inv(chol(G)) vGinv <- vec(Ginv) nvGinv <- matrix(rep(vGinv,n),byrow=TRUE,ncol=length(vGinv),nrow=n) arg <- matrix(x%*%Ginv, nrow=n) hmold0 <- matrix(1,nrow=n,ncol=1) hmold1 <- arg hmnew <- hmold0 if(r==1) { hmnew <- hmold1 } if(r>=2) { for(i in 2:r) { hmnew <- mat.Kprod(hmold1,arg)-(i-1)*mat.Kprod(nvGinv,hmold0) hmnew <- matrix(Sdrv.recursive(d=d,r=i,v=hmnew), nrow=n) hmold0 <- hmold1 hmold1 <- hmnew } } dens <- dmvnorm(x,mean=rep(0,d),sigma=Sigma) result <- drop(matrix(rep(dens,d^r),byrow=FALSE,nrow=n,ncol=d^r)*hmnew*(-1)^r) return(result) } ############################################################################### ## dmvnorm.deriv.unique computes the whole vector derivative of the Gaussian ## density phi_Sigma(x) from its unique coordinates, based on Algorithm 3 as ## described in Section 5 of Chacon and Duong (2014) ############################################################################### dmvnorm.deriv.unique <- function(x,Sigma,deriv.order=0) { if (is.vector(x)) { x <- matrix(x,nrow=1) } d <- ncol(x) n <- nrow(x) r <- deriv.order G <- Sigma Ginv <- chol2inv(chol(G)) arg <- x%*%Ginv hmold0 <- matrix(1,nrow=n,ncol=1) hmold1 <- arg hmnew <- hmold0 udind0 <- matrix(rep(0,d),nrow=1,ncol=d) udind1 <- diag(d) if(r==1) { hmnew <- hmold1 } if(r>=2) { for (i in 2:r) { Ndi1 <- ncol(hmold1) Ndi0 <- ncol(hmold0) hmnew <- numeric() for (j in 1:d) { nrecj <- choose(d-j+i-1,i-1) hmnew.aux <- arg[,j]*hmold1[,Ndi1-(nrecj:1)+1] for(k in j:d) { udind0.aux <- matrix(udind1[Ndi1-(nrecj:1)+1,],ncol=d,byrow=FALSE) udind0.aux[,k] <- udind0.aux[,k]-1 valid.udind0 <- as.logical(apply(udind0.aux>=0,1,min)) enlarged.hmold0 <- matrix(0,ncol=nrow(udind0.aux),nrow=n) for (ell in 1:nrow(udind0.aux)) { if(valid.udind0[ell]) { pos <- which(rowSums((udind0-matrix(rep(udind0.aux[ell,],nrow(udind0)), nrow=nrow(udind0),byrow=TRUE))^2)==0) enlarged.hmold0[,ell] <- hmold0[,pos] } } hmnew.aux <- hmnew.aux-Ginv[j,k]*matrix(rep(udind1[Ndi1-(nrecj:1)+1,k],n), nrow=n,byrow=TRUE)*enlarged.hmold0 } hmnew <- cbind(hmnew,hmnew.aux) } hmold0 <- hmold1 hmold1 <- hmnew ## compute the unique i-th derivative multi-indexes nudind1 <- nrow(udind1) udindnew <- numeric() for (j in 1:d) { Ndj1i <- choose(d+i-1-j,i-1) udind.aux <- matrix(udind1[nudind1-(Ndj1i:1)+1,],ncol=d,byrow=FALSE) udind.aux[,j] <- udind.aux[,j]+1 udindnew <- rbind(udindnew,udind.aux) } udind0 <- udind1 udind1 <- udindnew } } if(r==0) result <- dmvnorm(x,mean=rep(0,d),sigma=Sigma) if(r==1) result <- (-1)*matrix(rep(dmvnorm(x,mean=rep(0,d),sigma=Sigma),d), nrow=n,byrow=FALSE)*hmnew if(r>=2) { per <- pinv.all(d=d,r=r) dind <- numeric() udind <- udind1 dind.base <- rep(0,d^r) udind.base <- rep(0,choose(d+r-1,r)) for (i in 1:d) { dind <- cbind(dind,rowSums(per==i)) ## Matrix of derivative indices dind.base <- dind.base+dind[,i]*(r+1)^(d-i) ## Transform each row to base r+1 udind.base <- udind.base+udind[,i]*(r+1)^(d-i) ## Transform each row to base r+1 } dlabs <- match(dind.base,udind.base) deriv.vector <- hmnew[,dlabs]*matrix(rep((-1)^rowSums(dind),n),nrow=n,byrow=TRUE) result <- matrix(rep(dmvnorm(x,mean=rep(0,d),sigma=Sigma),ncol(deriv.vector)), nrow=n,byrow=FALSE)*deriv.vector } return(drop(result)) } dmvnorm.deriv.mixt <- function(x, mus, Sigmas, props, deriv.order, deriv.vec=TRUE, add.index=FALSE, only.index=FALSE, verbose=FALSE) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") if (is.vector(x)) d <- length(x) else d <- ncol(x) if (missing(mus)) mus <- rep(0,d) if (missing(Sigmas)) Sigmas <- diag(d) r <- deriv.order sumr <- sum(r) if (only.index | add.index) ind.mat <- dmvnorm.deriv(x=x, mu=mus[1,], Sigma=Sigmas[1:d,], deriv.order=r, only.index=TRUE) if (only.index) if (deriv.vec) return (ind.mat) else return(unique(ind.mat)) ## derivatives ## single component mixture if (identical(all.equal(props[1], 1), TRUE)) { if (is.matrix(mus)) mus <- mus[1,] dens <- dmvnorm.deriv(x=x, mu=mus, Sigma=Sigmas[1:d,], deriv.order=sumr) } ## multiple component mixture else { k <- length(props) if (verbose) pb <- txtProgressBar() dens <- 0 ## sum of each normal density value from each component at x for (i in 1:k) { dens <- dens + props[i]*dmvnorm.deriv(x=x, mu=mus[i,], Sigma=Sigmas[((i-1)*d+1):(i*d),], deriv.order=sumr) if (verbose) setTxtProgressBar(pb, i/k) } if (verbose) close(pb) } if (!deriv.vec) { dens <- dens[,!duplicated(ind.mat)] ind.mat <- unique(ind.mat) } if (add.index) return(list(deriv=dens, deriv.ind=ind.mat)) else return(deriv=dens) } ############################################################################### ## Double sum of K(X_i - X_j) used in density derivative estimation ## ## Parameters ## x - points to evaluate ## Sigma - variance matrix ## inc - 0 - exclude diagonals ## - 1 - include diagonals ## ## Returns ## Double sum at x ############################################################################## dmvnorm.deriv.sum <- function(x, Sigma, deriv.order=0, inc=1, binned=FALSE, bin.par, bgridsize, kfe=FALSE, deriv.vec=TRUE, add.index=FALSE, verbose=FALSE) { r <- deriv.order d <- ncol(x) n <- nrow(x) if (missing(bgridsize)) bgridsize <- default.bgridsize(d) if (binned) { d <- ncol(Sigma) n <- nrow(x) if (missing(bin.par)) bin.par <- binning(x, H=diag(diag(Sigma)), bgridsize=bgridsize) est <- kdde.binned(x=x, bin.par=bin.par, H=Sigma, deriv.order=r, verbose=verbose)$estimate if (r>0) { sumval <- rep(0, length(est)) for (j in 1:length(est)) sumval[j] <- sum(bin.par$counts * n * est[[j]]) } else sumval <- sum(bin.par$counts * n * est) ## transformation approach from Jose E. Chacon 06/12/2010 if (0) { Sigmainv12 <- matrix.sqrt(chol2inv(chol(Sigma))) y <- x %*% Sigmainv12 if (missing(bin.par)) bin.par <- binning(x=y, H=diag(d), bgridsize=bgridsize) est <- kdde.binned(x=y, bin.par=bin.par, H=diag(d), deriv.order=r, verbose=verbose)$estimate if (r>0) { sumval <- rep(0, length(est)) for (j in 1:length(est)) sumval[j] <- sum(bin.par$counts *n*est[[j]]) } else sumval <- sum(bin.par$counts * n * est) sumval <- det(Sigmainv12) * sumval %*% Kpow(Sigmainv12, pow=r) } } ## exact computation else { if (verbose) pb <- txtProgressBar() if (r==0) { n.seq <- block.indices(n, n, d=d, r=r, diff=TRUE) sumval <- 0 for (i in 1:(length(n.seq)-1)) { difs <- differences(x=x, y=x[n.seq[i]:(n.seq[i+1]-1),]) sumval <- sumval + sum(dmvnorm.deriv(x=difs, mu=rep(0,d), Sigma=Sigma, deriv.order=r, deriv.vec=deriv.vec)) if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) } if (verbose) close(pb) } else { ## only with r>0 ## original recursive code from Jose E. Chacon 03/2012 if (2*floor(r/2)!=r) { sumval <- rep(0,d^r) } else { Sigmainv <- chol2inv(chol(Sigma)) per <- perm.rep(d=d,r=r) dind <- numeric() for (i in 1:d) dind <- cbind(dind,rowSums(per==i)) ## matrix of derivative indices udind <- unique(dind) nudind <- nrow(udind) dlabs <- numeric(nrow(dind)) for (i in 1:nrow(udind)) dlabs <- dlabs+i*(rowSums((dind-matrix(rep(udind[i,],nrow(dind)),nrow=nrow(dind),byrow=TRUE))^2)==0) result <- rep(0,nudind) ndif <- n*(n-1)/2 dif.ind <- numeric() M <- 1e6 max.loop.size <- ceiling(M/nudind) ## inside the loop we need to store a matrix of order max.loop.size x nudind <= M nblocks <- ceiling(ndif/max.loop.size) blength <- c(rep(max.loop.size,nblocks-1),ndif-max.loop.size*(nblocks-1)) ## length of each of the blocks, the last one could be smaller tri.num <- (1:n)*((1:n)-1)/2 for (kk in 1:nblocks) { b <- blength[kk] if (verbose) setTxtProgressBar(pb, kk/nblocks) kkm <- (kk-1)*max.loop.size tri.ind <- findInterval(kkm:(kkm+b-1), tri.num) dif.ind.block <- cbind(kkm:(kkm+b-1) - tri.num[tri.ind]+1, tri.ind+1) difs.block <- x[dif.ind.block[,1],]-x[dif.ind.block[,2],] arg <- difs.block %*% Sigmainv narg <- nrow(arg) hmold0 <- matrix(rep(1,narg),ncol=1,nrow=narg) hmold1 <- arg hmnew <- hmold0 udind0 <- matrix(rep(0,d),nrow=1,ncol=d) udind1 <- diag(d) if (r==1) { hmnew <- hmold1 } if (r >= 2) { for (i in 2:r) { Ndi1 <- ncol(hmold1) hmnew <- numeric() for (j in 1:d) { nrecj <- choose(d-j+i-1,i-1) hmnew.aux <- arg[,j]*hmold1[,Ndi1-(nrecj:1)+1] for (k in j:d) { udind0.aux <- matrix(udind1[Ndi1-(nrecj:1)+1,],ncol=d,byrow=FALSE) udind0.aux[,k] <- udind0.aux[,k]-1 valid.udind0 <- as.logical(apply(udind0.aux>=0,1,min)) enlarged.hmold0 <- matrix(0,ncol=nrow(udind0.aux),nrow=narg) for (ell in 1:nrow(udind0.aux)) { if (valid.udind0[ell]) { pos <- which(rowSums((udind0-matrix(rep(udind0.aux[ell,],nrow(udind0)),nrow=nrow(udind0),byrow=TRUE))^2)==0) enlarged.hmold0[,ell] <- hmold0[,pos] } } ## in enlarged.hmold0 we put the vector hmold0 in those positions not having a -1 in any of the derivative order after subtracting e_k ## the remaining positions are zeroes ## surely this could be done in a more efficient way hmnew.aux <- hmnew.aux-Sigmainv[j,k]*matrix(rep(udind1[Ndi1-(nrecj:1)+1,k],narg),nrow=narg,byrow=TRUE)*enlarged.hmold0 } hmnew <- cbind(hmnew,hmnew.aux) } hmold0 <- hmold1 hmold1 <- hmnew ## compute the unique i-th derivative multi-indexes nudind1 <- nrow(udind1) udindnew <- numeric() for (j in 1:d) { Ndj1i <- choose(d+i-1-j,i-1) udind.aux <- matrix(udind1[nudind1-(Ndj1i:1)+1,],ncol=d,byrow=FALSE) udind.aux[,j] <- udind.aux[,j]+1 udindnew <- rbind(udindnew,udind.aux) } udind0 <- udind1 udind1 <- udindnew } } hmnew <- hmnew*matrix(rep((-1)^rowSums(udind),narg),nrow=narg,byrow=TRUE) phi <- dmvnorm(difs.block, mean=rep(0,d), sigma=Sigma) phi <- matrix(rep(phi,nudind),ncol=nudind,byrow=FALSE) result <- result+drop(colSums(phi*hmnew)) } result <- result[dlabs] hm0 <- dmvnorm.deriv(x=rep(0,d),Sigma=Sigma,deriv.order=deriv.order) sumval <- 2*result+n*hm0 if (verbose) close(pb) } } } if (inc==0) sumval <- sumval - n*dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=Sigma, deriv.order=r) sumval <- drop(sumval) if (kfe) { if (inc==1) sumval <- sumval/n^2 else sumval <- sumval/(n*(n-1)) } if (add.index) { ind.mat <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=r, only.index=TRUE) if (deriv.vec) return(list(sum=sumval, deriv.ind=ind.mat)) else return(list(sum=sumval, deriv.ind=unique(ind.mat))) } else return(sum=sumval) } ## Single partial derivative of the multivariate normal with scalar variance matrix sigma^2 I_d ## Code by Jose Chacon 04/09/2007 dmvnorm.deriv.scalar <- function(x, mu, sigma, deriv.order, binned=FALSE) { r <- deriv.order d <- ncol(x) sderiv <- sum(r) arg <- x/sigma darg <- dmvnorm(arg, mean=mu)/(sigma^(sderiv+d)) for (j in 1:d) { hmold0 <- 1 hmold1 <- arg[,j] hmnew <- 1 if (r[j] == 1) { hmnew <- hmold1 } if (r[j] >= 2) ## Multiply by the corresponding Hermite polynomial, coordinate-wise, using Fact C.1.4 in W&J (1995) and Willink (2005, p.273) for (i in (2:r[j])) { hmnew <- arg[,j] * hmold1 - (i - 1) * hmold0 hmold0 <- hmold1 hmold1 <- hmnew } darg <- hmnew * darg } val <- darg*(-1)^sderiv return(val) } dmvnorm.deriv.scalar.sum <- function(x, sigma, deriv.order=0, inc=1, kfe=FALSE, binned=FALSE, bin.par, verbose=FALSE) { r <- deriv.order d <- ncol(x) n <- nrow(x) if (binned) { if (missing(bin.par)) bin.par <- binning(x, H=diag(d)*sigma^2) n <- sum(bin.par$counts) ind.mat <- dmvnorm.deriv(x=rep(0,d), Sigma=diag(d), deriv.order=sum(r), deriv.vec=TRUE, only.index=TRUE) fhatr <- kdde.binned(bin.par=bin.par, H=sigma^2*diag(d), deriv.order=sum(r), deriv.vec=TRUE, w=rep(1,n), deriv.index=which.mat(r=r, ind.mat)[1]) sumval <- sum(bin.par$counts * n * fhatr$est[[1]]) } else { if (verbose) pb <- txtProgressBar() n.seq <- block.indices(n, n, d=d, r=r, diff=TRUE) sumval <- 0 for (i in 1:(length(n.seq)-1)) { difs <- differences(x=x, y=x[n.seq[i]:(n.seq[i+1]-1),]) sumval <- sumval + sum(dmvnorm.deriv.scalar(x=difs, mu=rep(0,d), sigma=sigma, deriv.order=r)) if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) } } if (verbose) close(pb) if (inc==0) sumval <- sumval - n*dmvnorm.deriv.scalar(x=t(as.matrix(rep(0,d))), mu=rep(0,d), sigma=sigma, deriv.order=r) if (kfe) if (inc==1) sumval <- sumval/n^2 else sumval <- sumval/(n*(n-1)) return(sumval) } ########################################################################## ## Normal scale psi functionals ########################################################################## psins.1d <- function(r, sigma) { if (r %% 2 ==0) psins <- (-1)^(r/2)*factorial(r)/((2*sigma)^(r+1)*factorial(r/2)*pi^(1/2)) else psins <- 0 return(psins) } psins <- function(r, Sigma, deriv.vec=length(r)==1) { d <- ncol(Sigma) if (deriv.vec) { dens <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), deriv.order=r, Sigma=2*Sigma, add.index=FALSE) } else { dens <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), deriv.order=sum(r), Sigma=2*Sigma, add.index=TRUE) if (!is.vector(dens$deriv.ind)) { i <- head(which.mat(r, dens$deriv.ind),n=1) dens <- dens$deriv[1,i] } else dens <- dens$deriv } dens <- drop(dens) return(dens) } ########################################################################## ## Vector moments of the normal distribution ########################################################################## mur <- function(r, A, mu, Sigma, type="unique") { type1 <- match.arg(type, c("direct", "recursive", "unique")) mur.val <- do.call(paste("mur", type1, sep="."), list(r=r, A=A, mu=mu, Sigma=Sigma)) return(mur.val) } ############################################################################# ## mur.direct computes the vector moment E[X^{\otimes r}] for a random ## vector with N(mu,Sigma) distribution, on the basis of Equation (8) in ## Section 6 of Chacon and Duong (2014) ############################################################################# mur.direct <- function(r, mu, Sigma) { d <- ncol(Sigma) result <- as.vector(Kpow(mu,r)) vS <- vec(Sigma) vSj <- 1 if (r>=2) { for(j in 1:floor(r/2)) { vSj <- as.vector(vSj%x%vS) cj <- prod(r:(r-2*j+1))/(prod(1:j)*2^j) mur2j <- as.vector(Kpow(mu,r-2*j)) result <- result+cj*as.vector(mur2j%x%vSj) } } return(drop(Sdrv.recursive(d=d,r=r,v=result))) } ############################################################################# ## mur.recursive computes the vector moment E[X^{\otimes r}] for a random ## vector with N(mu,Sigma) distribution, on the basis of Equation (9) in ## Section 6 of Chacon and Duong (2014), using Equation (7) in Section 5 ## to obtain the Hermite polynomial ############################################################################# mur.recursive <- function(r, mu, Sigma) { d <- ncol(Sigma) G <- -Sigma vG <- vec(G) arg <- mu hmold0 <- 1 hmold1 <- arg hmnew <- hmold0 if (r==1) { hmnew <- hmold1 } if (r>=2) { for (i in 2:r) { hmnew <- as.vector(arg%x%hmold1-(i-1)*(vG%x%hmold0)) hmold0 <- hmold1 hmold1 <- hmnew } } return(drop(Sdrv.recursive(d=d,r=r,v=hmnew))) } ############################################################################### ## mur.unique computes the vector moment E[X^{\otimes r}] for a random vector ## with N(mu,Sigma) distribution, on the basis of Equation (9) in Section 6 ## of Chacon and Duong (2014), using Algorithm 3 in Section 5, based on the ## unique partial derivatives, to obtain the Hermite polynomial ############################################################################### mur.unique <- function(r, mu, Sigma) { d <- ncol(Sigma) G <- -Sigma arg <- mu hmold0 <- 1 hmold1 <- arg hmnew <- hmold0 udind0 <- matrix(rep(0,d),nrow=1,ncol=d) udind1 <- diag(d) if (r==1) { hmnew <- hmold1 } if (r>=2) { for (i in 2:r) { Ndi1 <- length(hmold1) Ndi0 <- length(hmold0) hmnew <- numeric() for (j in 1:d) { nrecj <- choose(d-j+i-1,i-1) hmnew.aux <- arg[j]*hmold1[Ndi1-(nrecj:1)+1] for(k in j:d) { udind0.aux <- matrix(udind1[Ndi1-(nrecj:1)+1,],ncol=d,byrow=FALSE) udind0.aux[,k] <- udind0.aux[,k]-1 valid.udind0 <- as.logical(apply(udind0.aux>=0,1,min)) enlarged.hmold0 <- rep(0,nrow(udind0.aux)) for (ell in 1:nrow(udind0.aux)) { if (valid.udind0[ell]) { pos <- which(rowSums((udind0-matrix(rep(udind0.aux[ell,],nrow(udind0)), nrow=nrow(udind0),byrow=TRUE))^2)==0) enlarged.hmold0[ell] <- hmold0[pos] } } hmnew.aux <- hmnew.aux-G[j,k]*udind1[Ndi1-(nrecj:1)+1,k]*enlarged.hmold0 } hmnew <- c(hmnew,hmnew.aux) } hmold0 <- hmold1 hmold1 <- hmnew ## Compute the unique i-th derivative multi-indexes nudind1 <- nrow(udind1) udindnew <- numeric() for (j in 1:d) { Ndj1i <- choose(d+i-1-j,i-1) udind.aux <- matrix(udind1[nudind1-(Ndj1i:1)+1,],ncol=d,byrow=FALSE) udind.aux[,j] <- udind.aux[,j]+1 udindnew <- rbind(udindnew,udind.aux) } udind0 <- udind1 udind1 <- udindnew } } if (r==0) { result <- 1 } if (r==1) { result <- (-1)*hmnew } if (r>=2) { per <- pinv.all(d=d,r=r) dind <- numeric() udind <- udind1 dind.base <- rep(0,d^r) udind.base <- rep(0,choose(d+r-1,r)) for (i in 1:d) { dind <- cbind(dind,rowSums(per==i)) ## Matrix of derivative indices dind.base <- dind.base+dind[,i]*(r+1)^(d-i) ## Transform each row to base r+1 udind.base <- udind.base+udind[,i]*(r+1)^(d-i) ## Transform each row to base r+1 } dlabs <- match(dind.base,udind.base) result <- hmnew[dlabs]*(-1)^rowSums(dind) } return(drop(result*(-1)^r)) } ########################################################################## ## Moments of quadratic forms in normal variables ########################################################################## nur <- function(r, A, mu, Sigma, type="cumulant") { type <- match.arg(type, c("direct", "recursive", "unique", "cumulant")) nur.val <- do.call(paste("nur", type, sep="."), list(r=r, A=A, mu=mu, Sigma=Sigma)) return(nur.val) } nurs <- function(r, s, A, B, mu, Sigma, type="cumulant") { type <- match.arg(type, c("direct", "recursive", "unique", "cumulant")) nur.val <- do.call(paste("nurs", type, sep="."), list(r=r, s=s, A=A, B=B, mu=mu, Sigma=Sigma)) return(nur.val) } ############################################################################# ## nur.direct computes the moment E[(X^T AX)^r] of the quadratic form ## X^T AX where X is a random vector with N(mu,Sigma) distribution, using ## Equation (10) in Section 6 of Chacon and Duong (2014), and the direct ## implementation mur.direct of the normal moments ############################################################################# nur.direct <- function(r,A,mu,Sigma) { vA <- vec(A) result <- drop(Kpow(t(vA),r)%*%mur.direct(2*r,mu,Sigma)) return(result) } ############################################################################# ## nur.recursive computes the moment E[(X^T AX)^r] of the quadratic form ## X^T AX where X is a random vector with N(mu,Sigma) distribution, using ## Equation (10) in Section 6 of Chacon and Duong (2014), and the recursive ## implementation mur.recursive of the normal moments ############################################################################# nur.recursive <- function(r,A,mu,Sigma) { vA <- vec(A) result <- drop(Kpow(t(vA),r)%*%mur.recursive(2*r,mu,Sigma)) return(result) } ############################################################################# ## nur.unique computes the moment E[(X^T AX)^r] of the quadratic form ## X^T AX where X is a random vector with N(mu,Sigma) distribution, using ## Equation (10) in Section 6 of Chacon and Duong (2014), and the function ## mur.unique to compute the normal moments from its unique coordinates ############################################################################# nur.unique <- function(r,A,mu,Sigma) { vA <- vec(A) result <- sum(Kpow(vA,r)*mur.unique(2*r,mu,Sigma)) return(result) } ############################################################################# ## nur.cumulant computes the moment E[(X^T AX)^r] of the quadratic form ## X^T AX where X is a random vector with N(mu,Sigma) distribution, using ## the recursive formula relating moments and cumulants ############################################################################# nur.cumulant <- function(r,A,mu,Sigma) { if(r==0) { result <- 1 } if(r==1) { result <- sum(diag(A%*%Sigma))+t(mu)%*%A%*%mu } if(r>=2) { ASigma <- A%*%Sigma AS <- ASigma Amu <- A%*%mu kappas <- sum(diag(ASigma)+mu*Amu) nus <- kappas for (k in 2:r) { knew <- k*t(mu)%*%ASigma%*%Amu ASigma <- ASigma%*%AS knew <- (knew+sum(diag(ASigma)))*factorial(k-1)*2^(k-1) nnew <- knew+sum(choose(k-1,1:(k-1))*nus*rev(kappas)) kappas <- c(kappas,knew) nus <- c(nus,nnew) } result <- nnew } return(drop(result)) } ############################################################################# ## nurs.direct computes the joint moment E[(X^T AX)^r (X^T BX)^s] of the ## quadratic forms X^T AX and X^T BX, where X is a random vector with ## N(mu,Sigma) distribution, using Equation (10) in Section 6 of Chacon and ## Duong (2014), and the direct implementation mur.direct of the ## normal moments ############################################################################# nurs.direct <- function(r,s,A,B,mu,Sigma) { vA <- vec(A) vB <- vec(B) result <- (Kpow(t(vA),r)%x%Kpow(t(vB),s))%*%mur.direct(2*r+2*s,mu,Sigma) return(drop(result)) } ############################################################################# ## nurs.recursive computes the joint moment E[(X^T AX)^r (X^T BX)^s] of the ## quadratic forms X^T AX and X^T BX, where X is a random vector with ## N(mu,Sigma) distribution, using Equation (10) in Section 6 of Chacon and ## Duong (2014), and the recursive implementation mur.recursive of the ## normal moments ############################################################################# nurs.recursive <- function(r,s,A,B,mu,Sigma) { vA <- vec(A) vB <- vec(B) result <- (Kpow(t(vA),r)%x%Kpow(t(vB),s))%*%mur.recursive(2*r+2*s,mu,Sigma) return(drop(result)) } ############################################################################# ## nurs.unique computes the joint moment E[(X^T AX)^r (X^T BX)^s] of the ## quadratic forms X^T AX and X^T BX, where X is a random vector with ## N(mu,Sigma) distribution, using Equation (10) in Section 6 of Chacon and ## Duong (2014), and the function mur.unique to compute the normal moments ## from its unique coordinates ############################################################################# nurs.unique <- function(r,s,A,B,mu,Sigma) { vA <- vec(A) vB <- vec(B) result <- drop((Kpow(t(vA),r)%x%Kpow(t(vB),s))%*%mur.unique(2*r+2*s,mu,Sigma)) return(drop(result)) } ############################################################################# ## nurs.cumulant computes the joint moment E[(X^T AX)^r (X^T BX)^s] of the ## quadratic forms X^T AX and X^T BX, where X is a random vector with ## N(mu,Sigma) distribution, using the recursive formula (11) in Section 6 ## of Chacon and Duong (2014), relating moments and cumulants. The cumulants ## are computed using the function kappars, which is based on Theorem 3 ############################################################################# kappars <- function(r,s,A,B,mu,Sigma) { d <- ncol(A) if (r+s>1 & r>0 & s>0) { ind <- multicool::allPerm(multicool::initMC(c(rep(1,r),rep(2,s)))) } if (r+s==1 | r==0 | s==0) { ind <- matrix(c(rep(1,r),rep(2,s)),nrow=1) } if (r+s==0) { return(0) } nper <- nrow(ind) result <- 0 Dmat <- solve(Sigma)%*%mu%*%t(mu) ASigma <- A%*%Sigma BSigma <- B%*%Sigma Id <- diag(d) for (i in 1:nper) { product <- Id for (j in 1:(r+s)) { if (ind[i,j]==1) { product <- product%*%ASigma } else if(ind[i,j]==2) { product <- product%*%BSigma } } result <- result+sum(diag(product%*%(Id/(r+s)+Dmat))) } result <- result*factorial(r)*factorial(s)*2^(r+s-1) return(drop(result)) } nurs.cumulant <- function(r,s,A,B,mu,Sigma) { if(r==0 & s>0) { nurs <- nur.cumulant(s,B,mu,Sigma) } if(r>0 & s==0) { nurs <- nur.cumulant(r,A,mu,Sigma) } if(r==0 & s==0) { nurs <- 1 } if((r>0) & (s>0)) { K <- matrix(0,nrow=r+1,ncol=s) for (i in 0:r) { for (j in 1:s) { K[i+1,j] <- kappars(i,j,A,B,mu,Sigma) } } N <- matrix(0,nrow=r+1,ncol=s) for (i in 0:r) { N[i+1,1] <- nur.cumulant(r=i,A=A,mu=mu,Sigma=Sigma) } if (s>1) { for (j in 1:(s-1)) { for (i in 0:r) { Choose <- outer(choose(i,0:i), choose(j-1,0:(j-1))) N[i+1,j+1] <- sum(Choose*N[1:(i+1),1:j]*K[(i:0)+1,j:1]) } } } Choose <- outer(choose(r,0:r),choose(s-1,0:(s-1))) nurs <- sum(Choose*N[1:(r+1),1:s]*K[(r:0)+1,s:1]) } return(nurs) } ########################################################################## ## V-statistics with multivariate Gaussian derivatives kernel ########################################################################## Qr <- function(x, y, Sigma, deriv.order=0, inc=1, type="cumulant", verbose=FALSE) { if (missing(y)) y <- x type <- match.arg(type, c("direct", "cumulant")) if (type=="direct") Qr.val <- Qr.direct(x=x, y=y, Sigma=Sigma, r=deriv.order, inc=inc, verbose=verbose) else Qr.val <- Qr.cumulant(x=x, y=y, Sigma=Sigma, r=deriv.order, inc=inc, verbose=verbose) return(Qr.val) } ############################################################################ ## Qr.direct computes the V-statistic using the direct approach, as ## described in Section 6 of Chacon and Duong (2014) ############################################################################ Qr.direct <- function(x, y, Sigma, r=0, inc=1, binned=FALSE, bin.par, bgridsize, verbose=FALSE) { if (is.vector(x)) x <- matrix(x, nrow=1) d <- ncol(x) eta <- drop(Kpow(vec(diag(d)), r/2) %*% kfe(x=x, G=Sigma, deriv.order=r, inc=inc, verbose=verbose, binned=binned, bin.par=bin.par, add.index=FALSE)) return(eta) } ############################################################################ ## Qr.cumulant computes the V-statistic using the relationship with nur ## shown in Theorem 4 of Chacon and Duong (2014) ############################################################################ Qr.cumulant <- function(x, y, Sigma, r=0, inc=1, verbose=FALSE) { if (is.vector(x)) x <- matrix(x, nrow=1) d <- ncol(x) r <- r/2 if (missing(y)) y <- x if (is.vector(y)) y <- matrix(y, nrow=1) nx <- as.numeric(nrow(x)) ny <- as.numeric(nrow(y)) G <- Sigma Ginv <- chol2inv(chol(G)) G2inv <- Ginv%*%Ginv G3inv <- G2inv%*%Ginv trGinv <- sum(diag(Ginv)) trG2inv <- sum(diag(G2inv)) detG <- det(G) ## indices for separating into blocks for double sum calculation n.seq <- block.indices(nx, ny, d=d, r=r, diff=FALSE) if (verbose) pb <- txtProgressBar() if (r==0) { xG <- x%*%Ginv a <- rowSums(xG*x) eta <- 0 for (i in 1:(length(n.seq)-1)) { nytemp <- n.seq[i+1] - n.seq[i] ytemp <- matrix(y[n.seq[i]:(n.seq[i+1]-1),], ncol=d) aytemp <- rowSums((ytemp %*% Ginv) *ytemp) M <- a%*%t(rep(1,nytemp)) + rep(1, nx)%*%t(aytemp) - 2*(xG%*%t(ytemp)) em2 <- exp(-M/2) eta <- eta + (2*pi)^(-d/2)*detG^(-1/2)*sum(em2) if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) } } else if (r==1) { xG <- x%*%Ginv xG2 <- x%*%G2inv a <- rowSums(xG*x) a2 <- rowSums(xG2*x) eta <- 0 for (i in 1:(length(n.seq)-1)) { nytemp <- n.seq[i+1] - n.seq[i] ytemp <- matrix(y[n.seq[i]:(n.seq[i+1]-1),], nrow=nytemp) aytemp <- rowSums((ytemp %*% Ginv) *ytemp) aytemp2 <- rowSums((ytemp %*% G2inv) *ytemp) M <- a%*%t(rep(1,nytemp))+rep(1,nx)%*%t(aytemp)-2*(xG%*%t(ytemp)) M2 <- a2%*%t(rep(1,nytemp))+rep(1,nx)%*%t(aytemp2)-2*(xG2%*%t(ytemp)) eta <- eta + (2*pi)^(-d/2)*detG^(-1/2)*sum(exp(-M/2)*(M2-trGinv)) if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) } } else if (r==2) { xG <- x%*%Ginv xG2 <- x%*%G2inv xG3 <- x%*%G3inv a <- rowSums(xG*x) a2 <- rowSums(xG2*x) a3 <- rowSums(xG3*x) eta <- 0 for (i in 1:(length(n.seq)-1)) { nytemp <- n.seq[i+1] - n.seq[i] ytemp <- matrix(y[n.seq[i]:(n.seq[i+1]-1),], ncol=d) aytemp <- rowSums((ytemp %*% Ginv) *ytemp) aytemp2 <- rowSums((ytemp %*% G2inv) *ytemp) aytemp3 <- rowSums((ytemp %*% G3inv) *ytemp) M <- a%*%t(rep(1,nytemp))+rep(1,nx)%*%t(aytemp)-2*(xG%*%t(ytemp)) M2 <- a2%*%t(rep(1,nytemp))+rep(1,nx)%*%t(aytemp2)-2*(xG2%*%t(ytemp)) M3 <- a3%*%t(rep(1,nytemp))+rep(1,nx)%*%t(aytemp3)-2*(xG3%*%t(ytemp)) eta <- eta + (2*pi)^(-d/2)*detG^(-1/2)*sum(exp(-M/2)*(2*trG2inv-4*M3 +(-trGinv+M2)^2)) if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) } } else if (r>2) { xG <- x%*%Ginv a <- rowSums(xG*x) eta <- 0 for (i in 1:(length(n.seq)-1)) { nytemp <- n.seq[i+1] - n.seq[i] ytemp <- matrix(y[n.seq[i]:(n.seq[i+1]-1),], ncol=d) aytemp <- rowSums((ytemp %*% Ginv) *ytemp) M <- a %*% t(rep(1,nytemp)) + rep(1,nx)%*%t(aytemp) - 2*(xG%*%t(ytemp)) edv2 <- exp(-M/2) P0 <- Ginv kappas <- matrix(nrow=as.numeric(nx*nytemp), ncol=r) for (j in 1:r) { Gi1inv <- P0%*%Ginv trGi0inv <- sum(diag(P0)) xGi1inv <- x%*%Gi1inv xGi1invx <- rowSums(xGi1inv*x) aytemp <- rowSums((ytemp %*% Gi1inv) *ytemp) dvi1 <- xGi1invx%*%t(rep(1,nytemp))+rep(1,nx)%*%t(aytemp)-2*(xGi1inv%*%t(ytemp)) kappas[,j] <- (-2)^(j-1)*factorial(j-1)*(-trGi0inv+j*dvi1) P0 <- Gi1inv } nus <- matrix(nrow=as.numeric(nx*nytemp), ncol=r+1) nus[,1] <- 1 for (j in 1:r) { js <- 0:(j-1) if (j==1) nus[,2] <- kappas[,1] else nus[,j+1] <- rowSums(kappas[,j:1]*nus[,1:j]/matrix(rep(factorial(js)* factorial(rev(js)),nx*nytemp),nrow=nx*nytemp,byrow=TRUE))*factorial(j-1) } eta <- eta + (2*pi)^(-d/2)*detG^(-1/2)*sum(edv2*nus[,r+1]) if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) } } if (verbose) close(pb) if (inc==0) eta <- (eta - (-1)^r*nx*nur.cumulant(r=r, A=Ginv, mu=rep(0,d), Sigma=diag(d))*(2*pi)^(-d/2)*detG^(-1/2))/(nx*(ny-1)) if (inc==1) eta <- eta/(nx*ny) return(eta) } Qr.1d <- function(x, y, sigma, deriv.order=0, inc=1, verbose=FALSE) { d <- 1 r <- deriv.order/2 if (missing(y)) y <- x nx <- length(x) ny <- length(y) g <- sigma n.seq <- block.indices(nx, ny, d=1, r=0, diff=FALSE) eta <- 0 if (verbose) pb <- txtProgressBar() if (r==0) { a <- x^2 for (i in 1:(length(n.seq)-1)) { if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) nytemp <- n.seq[i+1] - n.seq[i] ytemp <- y[n.seq[i]:(n.seq[i+1]-1)] aytemp <- ytemp^2 M <- a %*%t(rep(1,nytemp)) + rep(1, nx)%*%t(aytemp) - 2*(x %*% t(ytemp)) em2 <- exp(-M/(2*g^2)) eta <- eta + (2*pi)^(-d/2)*g^(-1)*sum(em2) } } else if (r>0) { a <- x^2 for (i in 1:(length(n.seq)-1)) { if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) nytemp <- n.seq[i+1] - n.seq[i] ytemp <- y[n.seq[i]:(n.seq[i+1]-1)] aytemp <- ytemp^2 M <- a %*% t(rep(1,nytemp)) + rep(1,nx)%*%t(aytemp) - 2*(x %*%t(ytemp)) edv2 <- exp(-M/(2*g^2)) kappas <- matrix(nrow=as.numeric(nx*nytemp), ncol=r) for (i in 1:r) { aytemp <- ytemp^2 dvi1 <- (a %*% t(rep(1,nytemp)) + rep(1,nx) %*% t(aytemp) - 2*(x%*%t(ytemp)))/g^(2*(i+1)) kappas[,i] <- (-2)^(i-1)*factorial(i-1)*(-g^(-2*i)+i*dvi1) } nus <- matrix(nrow=as.numeric(nx*nytemp), ncol=r+1) nus[,1] <- 1 for (j in 1:r) { js <- 0:(j-1) if (j==1) nus[,2] <- kappas[,1] else nus[,j+1] <- rowSums(kappas[,j:1]*nus[,1:j]/matrix(rep(factorial(js)*factorial(rev(js)),nx*nytemp),nrow=nx*nytemp,byrow=TRUE))*factorial(j-1) } eta <- eta + (2*pi)^(-d/2)*g^(-1)*sum(edv2*nus[,r+1]) } } if (verbose) close(pb) if (inc==0) eta <- (eta - nx*dnorm.deriv(x=0, mu=0, sigma=g, deriv.order=deriv.order))/(nx*(ny-1)) if (inc==1) eta <- eta/(nx*ny) return(eta) } ############################################################################### ## Creates plots of mixture density functions ## ## Parameters ## mus - means ## Sigmas - variances ## props - vector of proportions of each mixture component ## dfs - degrees of freedom ## dist - "normal" - normal mixture ## - "t" - t mixture ## ... ############################################################################### plotmixt <- function(mus, sigmas, Sigmas, props, dfs, dist="normal", draw=TRUE, deriv.order=0, which.deriv.ind=1, binned=TRUE, ...) { ## locally set random seed not to interfere with global random number generators if (!exists(".Random.seed")) rnorm(1) old.seed <- .Random.seed on.exit( { .Random.seed <<- old.seed } ) set.seed(8192) if (!missing(sigmas)) plotmixt.1d(mus=mus, sigmas=sigmas, props=props, dfs=dfs, dist=dist, draw=draw, deriv.order=deriv.order, which.deriv.ind=which.deriv.ind, ...) else if (ncol(Sigmas)==2) plotmixt.2d(mus=mus, Sigmas=Sigmas, props=props, dfs=dfs, dist=dist, draw=draw, deriv.order=deriv.order, which.deriv.ind=which.deriv.ind,binned=binned, ...) else if (ncol(Sigmas)==3) plotmixt.3d(mus=mus, Sigmas=Sigmas, props=props, dfs=dfs, dist=dist, draw=draw, deriv.order=deriv.order, which.deriv.ind=which.deriv.ind, binned=binned, ...) } plotmixt.1d <- function(mus, sigmas, props, dfs, dist="normal", xlim, ylim, gridsize, draw=TRUE, deriv.order, which.deriv.ind, ...) { dist <- match.arg(dist, c("normal", "t")) maxsigmas <- 4*max(sigmas) if (missing(xlim)) xlim <- c(min(mus) - maxsigmas, max(mus) + maxsigmas) if (missing(gridsize)) gridsize <- default.gridsize(1) x <- seq(xlim[1]-0.1*abs(diff(xlim)), xlim[2]+0.1*abs(diff(xlim)), length=gridsize) if (dist=="normal") { if (deriv.order<=0) dens <- dnorm.mixt(x=x, mus=mus, sigmas=sigmas, props=props) else dens <- dnorm.deriv.mixt(x=x, mus=mus, sigmas=sigmas, props=props, deriv.order=deriv.order) } else if (dist=="t") stop("1-d t mixture not yet implemented") fhat <- list() fhat$x <- x fhat$eval.points <- x fhat$estimate <- dens fhat$H <- diag(1) fhat$h <- 1 fhat$gridtype <- "linear" fhat$gridded <- TRUE fhat$binned <- FALSE fhat$names <- parse.name(x) fhat$w <- rep(1, length(x)) if (deriv.order>0) { fhat$deriv.order <- deriv.order fhat$deriv.ind <- which.deriv.ind } class(fhat) <- "kdde" if (draw) plot(fhat, xlim=xlim, ...) invisible(fhat) } plotmixt.2d <- function(mus, Sigmas, props, dfs, dist="normal", xlim, ylim, gridsize, nrand=1e4, draw=TRUE, binned, deriv.order, which.deriv.ind, display="slice", ...) { dist <- match.arg(dist, c("normal", "t")) maxSigmas <- 4*max(Sigmas) if (is.vector(mus)) mus <- as.matrix(t(mus)) if (missing(xlim)) xlim <- c(min(mus[,1]) - maxSigmas, max(mus[,1]) + maxSigmas) if (missing(ylim)) ylim <- c(min(mus[,2]) - maxSigmas, max(mus[,2]) + maxSigmas) if (missing(gridsize)) gridsize <- default.gridsize(2) x <- seq(xlim[1], xlim[2], length=gridsize[1]) y <- seq(ylim[1], ylim[2], length=gridsize[2]) xy <- expand.grid(x, y) d <- ncol(Sigmas) if (dist=="normal") { if (deriv.order<=0) dens <- dmvnorm.mixt(xy, mus=mus, Sigmas=Sigmas, props=props) else dens <- dmvnorm.deriv.mixt(xy, mus=mus, Sigmas=Sigmas, props=props, deriv.order=deriv.order) } else if (dist=="t") { if (deriv.order>0) stop("deriv.order>0 for t mixture not yet implemented") dens <- dmvt.mixt(xy, mus=mus, Sigmas=Sigmas, props=props, dfs=dfs) } if (deriv.order<=0) dens.mat <- matrix(dens, ncol=length(x), byrow=FALSE) else { dens.mat <- list() for (i in 1:ncol(dens)) dens.mat[[i]] <- matrix(dens[,i], ncol=length(x), byrow=FALSE) } if (dist=="normal") x.rand <- rmvnorm.mixt(n=nrand, mus=mus, Sigmas=Sigmas, props=props) else if (dist=="t") x.rand <- rmvt.mixt(n=nrand, mus=mus, Sigmas=Sigmas, props=props, dfs=dfs) H <- Hns(x=x.rand, deriv.order=deriv.order) if (binned) H <- diag(diag(H)) fhat.rand <- kdde(x=x.rand, H=H, deriv.order=deriv.order, binned=binned) fhat <- fhat.rand fhat$x <- x.rand fhat$eval.points <- list(x,y) fhat$estimate <- dens.mat fhat$names <- c("x", "y") if (deriv.order>0) { deriv.ind <- dmvnorm.deriv.mixt(xy, mus=mus, Sigmas=Sigmas, props=props, add.index=TRUE, only.index=TRUE, deriv.order=deriv.order, deriv.vec=TRUE) fhat$deriv.order <- deriv.order fhat$deriv.ind <- deriv.ind class(fhat) <- "kdde" if (draw) plot(fhat, which.deriv.ind=which.deriv.ind, xlim=xlim, ylim=ylim, ...) } else { if (draw) plot(fhat, xlim=xlim, ylim=ylim, display=display, ...) } invisible(fhat) } plotmixt.3d <- function(mus, Sigmas, props, dfs, dist="normal", xlim, ylim, zlim, gridsize, nrand=1e4, draw=TRUE, binned, deriv.order, which.deriv.ind, ...) { d <- 3 dist <- match.arg(dist, c("normal", "t")) maxsd <- sqrt(apply(Sigmas, 2, max)) if (is.vector(mus)) mus <- as.matrix(t(mus)) if (missing(xlim)) xlim <- c(min(mus[,1]) - 4*maxsd[1], max(mus[,1]) + 4*maxsd[1]) if (missing(ylim)) ylim <- c(min(mus[,2]) - 4*maxsd[2], max(mus[,2]) + 4*maxsd[2]) if (missing(zlim)) zlim <- c(min(mus[,3]) - 4*maxsd[3], max(mus[,3]) + 4*maxsd[3]) if (missing(gridsize)) gridsize <- default.gridsize(3) x <- seq(xlim[1], xlim[2], length=gridsize[1]) y <- seq(ylim[1], ylim[2], length=gridsize[2]) z <- seq(zlim[1], zlim[2], length=gridsize[3]) xy <- expand.grid(x,y) if (deriv.order>0) { if (dist=="t") stop("deriv.order>0 for t mixture not yet implemented") else if (dist=="normal") deriv.ind <- dmvnorm.deriv.mixt(cbind(xy,z[1]), mus=mus, Sigmas=Sigmas, props=props, deriv.order=deriv.order, add.index=TRUE, only.index=TRUE) } if (deriv.order<=0) dens.array <- array(0, dim=gridsize) else { dens.array <- list(); for (i in 1:nrow(deriv.ind)) dens.array <- c(dens.array, list(array(0, dim=gridsize))) } for (i in 1:length(z)) { if (dist=="normal") { if (deriv.order<=0) dens <- dmvnorm.mixt(cbind(xy, z[i]), mus=mus, Sigmas=Sigmas, props=props) else dens <- dmvnorm.deriv.mixt(cbind(xy, z[i]), mus=mus, Sigmas=Sigmas, props=props, deriv.order=deriv.order) } else if (dist=="t") dens <- dmvt.mixt(cbind(xy, z[i]), mus=mus, Sigmas=Sigmas, dfs=dfs, props=props) if (deriv.order<=0) { dens.mat <- matrix(dens, ncol=length(x), byrow=FALSE) dens.array[,,i] <- dens.mat } else { for (j in 1:ncol(dens)) { dens.mat <- matrix(dens[,j], ncol=length(x), byrow=FALSE) dens.array[[j]][,,i] <- dens.mat } } } if (dist=="normal") x.rand <- rmvnorm.mixt(n=nrand, mus=mus, Sigmas=Sigmas, props=props) else if (dist=="t") x.rand <- rmvt.mixt(n=nrand, mus=mus, Sigmas=Sigmas, props=props, dfs=dfs) H <- Hns(x=x.rand, deriv.order=deriv.order) if (binned) H <- diag(diag(H)) fhat.rand <- kdde(x=x.rand, H=H, deriv.order=deriv.order, binned=binned) fhat <- fhat.rand fhat$x <- head(x.rand, n=100) fhat$eval.points <- list(x,y,z) fhat$estimate <- dens.array fhat$names <- c("x", "y", "z") fhat$H <- H fhat$w <- rep(1,nrow(fhat$x)) if (deriv.order>0) { deriv.ind <- dmvnorm.deriv.mixt(xy, mus=mus, Sigmas=Sigmas, props=props, add.index=TRUE, only.index=TRUE, deriv.order=deriv.order, deriv.vec=TRUE) fhat$deriv.order <- deriv.order fhat$deriv.ind <- deriv.ind class(fhat) <- "kdde" if (draw) plot(fhat, which.deriv.ind=which.deriv.ind, xlim=xlim, ylim=ylim, zlim=zlim, ...) } else { if (draw) plot(fhat, xlim=xlim, ylim=ylim, zlim=zlim, ...) } invisible(fhat) } ############################################################################### ## Multivariate t mixture - density values ## ## Parameters ## x - points to compute density at ## mus - vector of means ## Sigmas - dispersion matrices ## dfs - degrees of freedom ## props - vector of mixing proportions ## ## Returns ## Value of multivariate t mixture density at x ############################################################################### dmvt.mixt <- function(x, mus, Sigmas, dfs, props) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") else if (length(dfs) != length(props)) stop("Length of df and mixing proportions vectors not equal") ## single component mixture if (identical(all.equal(props[1], 1), TRUE)) dens <- dmvt(x, delta=mus, sigma=Sigmas, df=dfs, log=FALSE) ## multiple component mixture else { if (is.vector(mus)) d <- length(mus) else d <- ncol(mus) k <- length(props) dens <- 0 for (i in 1:k) dens <- dens+props[i]*dmvt(x,delta=mus[i,],sigma=Sigmas[((i-1)*d+1):(i*d),], df=dfs[i], log=FALSE) } return(dens) } ############################################################################### ## Multivariate t mixture - random sample ## ## Parameters ## n - number of samples ## mus - means ## Sigmas - matrix of dispersion matrices ## dfs - vector of degrees of freedom ## props - vector of mixing proportions ## ## Returns ## Vector of n observations from the t mixture ############################################################################### rmvt.mixt <- function(n=100, mus=c(0,0), Sigmas=diag(2), dfs=7, props=1) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") else if (length(dfs) != length(props)) stop("Length of df and mixing proportions vectors not equal") ## single component mixture if (identical(all.equal(props[1], 1), TRUE)) { rand <- rmvt(n=n, sigma=Sigmas, df=dfs) for (i in 1:length(mus)) rand[,i] <- rand[,i] + mus[i] } ## multiple component mixture else { k <- length(props) d <- ncol(Sigmas) n.samp <- sample(1:k, n, replace=TRUE, prob=props) n.prop <- numeric(0) ## compute number to be drawn from each component for (i in 1:k) n.prop <- c(n.prop, sum(n.samp == i)) ## generate random samples from each component rand <- numeric(0) for (i in 1:k) { if (n.prop[i] > 0) { rand.temp <- rmvt(n=n.prop[i],sigma=Sigmas[((i-1)*d+1):(i*d),],df=dfs[i]) for (j in 1:length(mus[k,])) rand.temp[,j] <- rand.temp[,j] + mus[i,j] rand <- rbind(rand, rand.temp) } } } return(rand[sample(n),]) } ############################################################################### ## Moments of multivariate normal mixture ############################################################################### mvnorm.mixt.moment <- function (mus, Sigmas, props) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") d <- ncol(Sigmas) k <- length(props) mn <- rep(0, d) va <- matrix(0, nrow=d, ncol=d) for (i in 1:k) { mn <- mn + props[i] * mus[i,] va <- va + props[i] * (Sigmas[((i-1)*d+1):(i*d),] + mus[i,] %*% t(mus[i,])) } va <- va + mn %*% t(mn) return( list(mean=mn, var=va)) } ###################################################################### ## Modes for normal mixture ###################################################################### mvnorm.mixt.mode <- function(mus, Sigmas, props=1, verbose=FALSE) { if (!(identical(all.equal(sum(props), 1), TRUE))) stop("Proportions don't sum to one") if (identical(all.equal(props[1], 1), TRUE)) mm <- mus else { k <- length(props) d <- ncol(Sigmas) mm <- matrix(0, nrow=k, ncol=d) dmvnorm.mixt.temp <- function(x) { return(-1*dmvnorm.mixt(x=x, mus=mus, Sigmas=Sigmas, props=props)) } for (i in 1:k) { result <- nlm(p=mus[i,], f=dmvnorm.mixt.temp, print.level=2*as.numeric(verbose)) mm[i,] <- result$estimate } } return(mm) } ###################################################################### ## Partition for 2-d normal mixture ###################################################################### mvnorm.mixt.part <- function(mus, Sigmas, props=1, xmin, xmax, gridsize, max.iter=100, verbose=FALSE) { maxSigmas <- 4*max(Sigmas) if (is.vector(mus)) mus <- as.matrix(t(mus)) if (missing(xmin)) xmin <- c(min(mus[,1]) - maxSigmas, min(mus[,2]) - maxSigmas) if (missing(xmax)) xmax <- c(max(mus[,1]) + maxSigmas, max(mus[,2]) + maxSigmas) if (missing(gridsize)) gridsize <- c(201,201) x <- seq(xmin[1], xmax[1], length=gridsize[1]) y <- seq(xmin[2], xmax[2], length=gridsize[2]) xy <- expand.grid(x, y) xy.orig <- xy d <- ncol(Sigmas) k <- length(props) a <- min(c(xmax[1]-xmin[1], xmax[2]-xmin[2]))/1e3 ## max.iter mean shift iterations for (i in 1:max.iter) { dens <- dmvnorm.mixt(xy, mus=mus, Sigmas=Sigmas, props=props) grad <- dmvnorm.deriv.mixt(xy, mus=mus, Sigmas=Sigmas, props=props,deriv.order=1) xy[,1] <- xy[,1] + a*grad[,1]/dens xy[,2] <- xy[,2] + a*grad[,2]/dens } xy.dist <- matrix(0, ncol=k, nrow=nrow(xy)) for (i in 1:k) xy.dist[,i] <- apply(sweep(xy, 2, mus[i,])^2, 1, sum) xy.lab <- array(apply(xy.dist, 1, which.min), dim=gridsize) x.rand <- rmvnorm.mixt(n=1e3, mus=mus, Sigmas=Sigmas, props=props) fhat <- kde(x=x.rand, binned=TRUE) fhat$eval.points <- list(x, y) fhat$estimate <- xy.lab fhat$names <- c("x", "y") class(fhat) <- "kde.part" return(fhat) } ks/R/ksupp.R0000644000176200001440000001027114547756302012373 0ustar liggesusers###################################################################### ## Kernel support estimate - contour-based or convex hull ###################################################################### ksupp <- function(fhat, cont=95, abs.cont, convex.hull=TRUE) { if (missing(abs.cont)) abs.cont <- contourLevels(fhat, cont=cont) supp <- expand.grid(fhat$eval.points)[as.vector(fhat$estimate > abs.cont),] d <- length(fhat$eval.points) if (d==2) { if (convex.hull) supp <- supp[chull(supp),] } else { if (convex.hull) { if (!requireNamespace("geometry", quietly=TRUE)) stop("Install the geometry package as it is required.", call.=FALSE) supp <- as.matrix(supp[t(geometry::convhulln(supp)),]) } } class(supp) <- c("ksupp", class(supp)) return(supp) } ## Devroye-Wise support estimate dwsupp <- function(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, binned, bgridsize, verbose=FALSE, w) { if (is.vector(x)) { if (missing(H)) { d <- 1; n <- length(x) } else { if (is.vector(H)) { d <- 1; n <- length(x) } else { x <- matrix(x, nrow=1); d <- ncol(x); n <- nrow(x) } } } else { d <- ncol(x); n <- nrow(x) } if (!missing(w)) if (!(identical(all.equal(sum(w), n), TRUE))) { warning("Weights don't sum to sample size - they have been scaled accordingly\n") w <- w*n/sum(w) } if (missing(binned)) binned <- default.bflag(d=d, n=n) if (missing(w)) w <- rep(1,n) if (d==1) { y <- x if (missing(h)) h <- hpi(x=y, binned=default.bflag(d=d, n=n), bgridsize=bgridsize) } if (missing(H) & d>1) { H <- Hpi(x=x, binned=default.bflag(d=d, n=n), bgridsize=bgridsize, verbose=verbose) } if (missing(bgridsize)) bgridsize <- default.gridsize(d) if (missing(gridsize)) gridsize <- default.gridsize(d) ## initialise grid n <- nrow(x) gridx <- make.grid.ks(x, matrix.sqrt(H), tol=supp, gridsize=gridsize, xmin=xmin, xmax=xmax, gridtype=gridtype) suppx <- make.supp(x, matrix.sqrt(H), tol=supp) grid.pts <- find.gridpts(gridx, suppx) fhat.grid <- matrix(0, nrow=length(gridx[[1]]), ncol=length(gridx[[2]])) if (verbose) pb <- txtProgressBar() for (i in 1:n) { ## compute evaluation points eval.x <- gridx[[1]][grid.pts$xmin[i,1]:grid.pts$xmax[i,1]] eval.y <- gridx[[2]][grid.pts$xmin[i,2]:grid.pts$xmax[i,2]] eval.x.ind <- c(grid.pts$xmin[i,1]:grid.pts$xmax[i,1]) eval.y.ind <- c(grid.pts$xmin[i,2]:grid.pts$xmax[i,2]) eval.x.len <- length(eval.x) eval.pts <- expand.grid(eval.x, eval.y) fhat <- rep(1, nrow(eval.pts)) ## place vector of density estimate values `fhat' onto grid 'fhat.grid' for (j in 1:length(eval.y)) fhat.grid[eval.x.ind, eval.y.ind[j]] <- fhat.grid[eval.x.ind, eval.y.ind[j]] + w[i]*fhat[((j-1) * eval.x.len + 1):(j * eval.x.len)] if (verbose) setTxtProgressBar(pb, i/n) } if (verbose) close(pb) gridx1 <- list(gridx[[1]], gridx[[2]]) fhat <- list(x=x, eval.points=gridx1, estimate=fhat.grid>=1, H=H, gridtype=gridx$gridtype, gridded=TRUE) fhat$binned <- binned fhat$names <- parse.name(x) ## add variable names fhat$w <- w class(fhat) <- "kde" return(fhat) } ###################################################################### ## S3 methods for KSUPP objects ###################################################################### ## plot method plot.ksupp <- function(x, display="plot3D", ...) { d <- ncol(x) if (d==2) polygon(x, ...) else if (d==3) { disp <- match.arg(display, c("plot3D", "rgl")) if (disp=="plot3D") { if (!requireNamespace("plot3D", quietly=TRUE)) stop("Install the plot3D package as it is required.", call.=FALSE) plot3D::triangle3D(x, ...) } else if (disp=="rgl") { if (!requireNamespace("rgl", quietly=TRUE)) stop("Install the rgl package as it is required.", call.=FALSE) rgl::triangles3d(x=x[,1], y=x[,2], z=x[,3], ...) } } } ks/R/hist.R0000644000176200001440000001401414464267002012167 0ustar liggesusers############################################################################# ## Histogram density estimators ############################################################################# histde <- function(x, binw, xmin, xmax, adj=0) { if (is.vector(x)) { d <- 1; n <- length(x) if (missing(binw)) binw <- 2*3^(1/(d+2))*pi^(2/(2*d+4))*sd(x)*length(x)^(-1/(d+2)) nbin <- round(diff(range(x))*1.2/binw, 0) } else { d <- ncol(x); n <- nrow(x) if (missing(binw)) binw <- 2*3^(1/(d+2))*pi^(2/(2*d+4))*apply(x, 2, sd)*nrow(x)^(-1/(d+2)) nbin <- round(apply(apply(x, 2, range), 2, diff)*1.2/binw,0) } if (d==1) fhat <- hist.1d(x, binw=binw, xmin=xmin, xmax=xmax, adj=adj) else if (d==2) fhat <- hist.2d(x=x, binw=binw, xmin=xmin, xmax=xmax, adj=adj) fhat$names <- parse.name(x) class(fhat) <- "histde" return(fhat) } ## 1D histogram hist.1d <- function(x, nbin, binw, x.cut, xmin, xmax, adj=0, ...) { if (missing(xmin)) xmin <- min(x) if (missing(xmax)) xmax <- max(x) if (missing(nbin)) nbin <- round(diff(range(x))*1.2/binw, 0) if (missing(x.cut)) x.cut <- seq(from=xmin-0.1*(xmax-xmin), to=xmax+0.1*(xmax-xmin), length=nbin+1) if (missing(binw)) binw <- diff(x.cut)[1] x.cut <- x.cut + adj*binw hs <- hist(x=x, breaks=x.cut, plot=FALSE) hs <- list(x=x, estimate=hs$density, eval.points=hs$breaks, binw=binw, nbin=nbin) class(hs) <- "histde" return(hs) } ## 2D histogram hist.2d <- function(x, nbin, binw, x.cut, xmin, xmax, adj=0, ...) { if (missing(nbin)) nbin <- round(apply(apply(x, 2, range), 2, diff)*1.2/binw,0) if (length(nbin)==1) nbin <- rep(nbin,2) if (missing(xmin)) xmin <- apply(x, 2, min) if (missing(xmax)) xmax <- apply(x, 2, max) if (missing(x.cut)) x.cut <- list(seq(from=xmin[1]-0.1*(xmax[1]-xmin[1]), to=xmax[1]+0.1*(xmax[1]-xmin[1]), length=nbin[1]+1), seq(from=xmin[2]-0.1*(xmax[2]-xmin[2]), to=xmax[2]+0.1*(xmax[2]-xmin[2]), length=nbin[2]+1)) if (missing(binw)) binw <- c(diff(x.cut[[1]])[1], diff(x.cut[[2]])[1]) x.cut[[1]] <- x.cut[[1]] + adj*binw[1] x.cut[[2]] <- x.cut[[2]] + adj*binw[2] index.x <- cut(x[,1], x.cut[[1]], include.lowest=TRUE) index.y <- cut(x[,2], x.cut[[2]], include.lowest=TRUE) m <- matrix(0, nrow=nbin[1], ncol=nbin[2], dimnames=list(levels(index.x), levels(index.y))) for (i in 1:length(index.x)) m[index.x[i], index.y[i]] <- m[index.x[i], index.y[i]] + 1 hs2d <- list(x=x, estimate=m/(nrow(x)*prod(binw)), eval.points=x.cut, binw=binw, nbin=nbin) class(hs2d) <- "histde" return(hs2d) } ############################################################################# ## plot histograms ############################################################################# plot.histde <- function(x, ...) { if (is.vector(x$x)) plothistde.1d(fhat=x, ...) else plothistde.2d(fhat=x, ...) invisible() } plothistde.1d <- function(fhat, xlab, ylab="Density function", add=FALSE, drawpoints=FALSE, col="transparent", col.pt=4, jitter=FALSE, border=1, alpha=1, ...) { if (missing(xlab)) xlab <- fhat$names col <- transparency.col(col, alpha=alpha) if (!add) plot(fhat$eval.points, c(fhat$estimate,0), type="n", xlab=xlab, ylab=ylab, ...) rect(fhat$eval.points[-length(fhat$eval.points)], 0, fhat$eval.points[-1], fhat$estimate, border=border, col=col, ...) if (drawpoints) if (jitter) rug(jitter(fhat$x), col=col.pt) else rug(fhat$x, col=col.pt) } plothistde.2d <- function(fhat, breaks, nbreaks=11, xlab, ylab, zlab="Density function", cex=1, pch=1, add=FALSE, drawpoints=FALSE, col, col.fun, alpha=1, col.pt=4, lty.rect=2, cex.text=1, border, lwd.rect=1, col.rect="transparent", add.grid=TRUE, ...) { if (missing(xlab)) xlab <- fhat$names[1] if (missing(ylab)) ylab <- fhat$names[2] if (missing(border)) border <- grey(0.5) if (!add) plot(fhat$x, col=col.pt, type="n", xlab=xlab, ylab=ylab, ...) if (missing(breaks)) breaks <- seq(min(fhat$estimate,0), max(fhat$estimate)+0.1*diff(range(fhat$estimate)), length=nbreaks) if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="heat", rev=TRUE, alpha=alpha) } if (missing(col)) col <- col.fun(n=length(breaks)) col <- transparency.col(col, alpha=alpha) for (i in 1:(nrow(fhat$estimate))) for (j in 1:(ncol(fhat$estimate))) { if (fhat$estimate[i,j]>=breaks[2]) { rect(fhat$eval.points[[1]][i], fhat$eval.points[[2]][j], fhat$eval.points[[1]][i+1], fhat$eval.points[[2]][j+1], col=col[findInterval(fhat$estimate[i,j], breaks)], lty=lty.rect, border=border, lwd=lwd.rect) } } if (add.grid) { for (i in 1:length(fhat$eval.points[[1]])) lines(rep(fhat$eval.points[[1]][i],2), range(fhat$eval.points[[2]]), col=border, ...) for (j in 1:length(fhat$eval.points[[2]])) lines(range(fhat$eval.points[[1]]), rep(fhat$eval.points[[2]][j],2), col=border, ...) } if (drawpoints) points(fhat$x[,1], fhat$x[,2], col=col.pt, cex=cex, pch=pch) } ## predict method predict.histde <- function(object, ..., x) { fhat <- object$estimate if (!is.list(object$eval.points)) d <- 1 else d <- length(object$eval.points) if (d==1) { gs <- length(object$eval.points) x.ind <- findInterval(x, object$eval.points, all.inside = FALSE) fhat[x.ind==0] <- object$estimate[1] fhat[x.ind==gs] <- object$estimate[gs] } else { x <- matrix(x, ncol = d) gs <- sapply(object$eval.points, length) x.ind <- matrix(0, nrow=nrow(x), ncol=d) for (i in 1:d) x.ind[, i] <- findInterval(x[, i], object$eval.points[[i]], all.inside = FALSE) x.ind[x.ind==0] <- 1 x.ind.flag <- x.ind==1 for (i in 1:d) x.ind.flag[, i] <- x.ind.flag[, i] | x.ind[, i] ==gs[i] } return(fhat[x.ind]) } ## contourLevels method contourLevels.histde <- function(...) { return(contourLevels.kde(...)) } ks/R/kfe.R0000644000176200001440000002307114336521322011764 0ustar liggesusers############################################################################# ## Kernel functional estimation ############################################################################# kfe <- function(x, G, deriv.order, inc=1, binned, bin.par, bgridsize, deriv.vec=TRUE, add.index=TRUE, verbose=FALSE) { r <- deriv.order d <- ncol(x) n <- nrow(x) if (missing(binned)) binned <- default.bflag(d, n) psir <- dmvnorm.deriv.sum(x=x, Sigma=G, deriv.order=r, inc=inc, binned=binned, bin.par=bin.par, bgridsize=bgridsize, deriv.vec=deriv.vec, verbose=verbose, kfe=TRUE, add.index=FALSE) if (add.index) { ind.mat <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=r, only.index=TRUE) if (deriv.vec) return(list(psir=psir, deriv.ind=ind.mat)) else return(list(psir=psir, deriv.ind=unique(ind.mat))) } else return(psir=psir) } kfe.1d <- function(x, g, deriv.order, inc=1, binned=TRUE, bin.par) { r <- deriv.order n <- length(x) psir <- dnorm.deriv.sum(x=x, sigma=g, deriv.order=r, inc=1, binned=binned, bin.par=bin.par, kfe=TRUE) if (inc==0) psir <- (n^2*psir - n*dnorm.deriv(0, mu=0, sigma=g, deriv.order=r))/(n*(n-1)) return(psir) } kfe.scalar <- function(x, g, deriv.order, inc=1, binned=TRUE, bin.par, verbose=FALSE) { r <- deriv.order d <- ncol(x) psir <- dmvnorm.deriv.scalar.sum(x=x, sigma=g, deriv.order=r, inc=inc, kfe=TRUE, binned=binned, bin.par=bin.par, verbose=verbose) return(psir) } ############################################################################### ## Plug-in unconstrained bandwidth for KFE ## ## Returns ## Plug-in bandwidth ############################################################################### hpi.kfe <- function(x, nstage=2, binned=FALSE, bgridsize, amise=FALSE, deriv.order=0) { n <- length(x) d <- 1 r <- deriv.order k <- 2 ## kernel order Kr0 <- dnorm.deriv(x=0, mu=0, sigma=1, deriv.order=r) mu2K <- 1 if (nstage==2) { psi4.hat <- psins.1d(r=r+k+2, sigma=sd(x)) gamse2 <- (factorial(r+2)*Kr0/(mu2K*psi4.hat*n))^(1/(r+k+3)) psi2.hat <- kfe.1d(x=x, g=gamse2, deriv.order=r+k, inc=1, binned=binned) } else { psi2.hat <- psins.1d(r=r+k, sigma=sd(x)) } ## formula for bias annihliating bandwidths from Wand & Jones (1995, p.70) gamse <- (factorial(r)*Kr0/(-mu2K*psi2.hat*n))^(1/(r+k+1)) return(gamse) } Hpi.kfe <- function(x, nstage=2, pilot, pre="sphere", Hstart, binned=FALSE, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") { if (deriv.order!=0) stop("Currently only deriv.order=0 is implemented") n <- nrow(x) d <- ncol(x) r <- deriv.order if (missing(pilot)) pilot <- "dscalar" if(!is.matrix(x)) x <- as.matrix(x) pilot1 <- match.arg(pilot, c("dunconstr", "dscalar")) pre1 <- match.arg(pre, c("scale", "sphere")) optim.fun1 <- match.arg(optim.fun, c("nlm", "optim")) if (pre1=="scale") { x.star <- pre.scale(x) S12 <- diag(sqrt(diag(var(x)))) Sinv12 <- chol2inv(chol(S12)) } else if (pre1=="sphere") { x.star <- pre.sphere(x) S12 <- matrix.sqrt(var(x)) Sinv12 <- chol2inv(chol(S12)) } D2K0 <- t(dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=2)) K0 <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=0) if (nstage==2) { ## stage 1 if (pilot1=="dscalar") { psi4.ns <- psins(r=r+4, Sigma=var(x.star), deriv.vec=TRUE) ## gdscalar not used because it's an implicit computation without ## symmetriser matrices and is slower than direct computation with ## symmetriser matrices. A1 <- drop(t(D2K0) %*% D2K0) A2 <- drop(t(D2K0) %*% t(vec(diag(d)) %x% diag(d^2)) %*% psi4.ns) A3 <- drop(t(psi4.ns) %*% (vec(diag(d)) %*% t(vec(diag(d))) %x% diag(d^2)) %*% psi4.ns) ## Special case from Chacon & Duong (2009): bias annihilation h2 <- (-A1/(2*A2*n))^(1/(d+4)) H2 <- h2^2*diag(d) psi2.hat <- kfe(x=x.star, G=H2, deriv.order=r+2, add.index=FALSE, binned=binned, bgridsize=bgridsize, verbose=verbose) } else if (pilot1=="dunconstr") { psi4.ns <- psins(r=r+4, Sigma=var(x), deriv.vec=TRUE) amse2.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) Hinv <- chol2inv(chol(H)) Hinv12 <- matrix.sqrt(Hinv) amse2.temp <- 1/(det(H)^(1/2)*n)*((Hinv12 %x% Hinv12) %*% D2K0) + 1/2* t(vec(H) %x% diag(d^2)) %*% psi4.ns return(sum((amse2.temp)^2)) } Hstart2 <- matrix.sqrt(Gns(r=r+2, n=n, Sigma=var(x))) if (optim.fun1=="nlm") { result <- nlm(p=vech(Hstart2), f=amse2.temp, print.level=2*as.numeric(verbose)) H2 <- invvech(result$estimate) %*% invvech(result$estimate) } else if (optim.fun1=="optim") { result <- optim(vech(Hstart2), amse2.temp, method="BFGS", control=list(trace=as.numeric(verbose),REPORT=1)) H2 <- invvech(result$par) %*% invvech(result$par) } psi2.hat <- kfe(x=x, G=H2, deriv.order=r+2, add.index=FALSE, binned=binned, bgridsize=bgridsize, verbose=verbose) } } else { if (pilot1=="dscalar") psi2.hat <- psins(r=r+2, Sigma=var(x.star), deriv.vec=TRUE) else if (pilot1=="dunconstr") psi2.hat <- psins(r=r+2, Sigma=var(x), deriv.vec=TRUE) } if (pilot1=="dscalar") { if (missing(Hstart)) Hstart <- Gns(r=r, n=n, Sigma=var(x.star)) } else if (pilot1=="dunconstr") { if (missing(Hstart)) Hstart <- Gns(r=r, n=n, Sigma=var(x)) } ## stage 2 amse.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) amse.val <- 1/(det(H)^(1/2)*n)*K0 + 1/2* t(vec(H)) %*% psi2.hat return(sum((amse.val^2))) } Hstart <- matrix.sqrt(Hstart) if (optim.fun1=="nlm") { result <- nlm(p=vech(Hstart), f=amse.temp, print.level=2*as.numeric(verbose)) H <- invvech(result$estimate) %*% invvech(result$estimate) amise.star <- result$minimum } else if (optim.fun1=="optim") { result <- optim(vech(Hstart), amse.temp, method="BFGS", control=list(trace=as.numeric(verbose),REPORT=1)) H <- invvech(result$par) %*% invvech(result$par) amise.star <- result$value } ## back-transform if (pilot1=="dscalar") H <- S12 %*% H %*% S12 if (!amise) return(H) else return(list(H=H, PI=amise.star)) } ############################################################################### ## Plug-in diagonal bandwidth for KFE ## ## Returns ## Plug-in bandwidth ############################################################################### Hpi.diag.kfe <- function(x, nstage=2, pilot, pre="scale", Hstart, binned=FALSE, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") { if (deriv.order!=0) stop("Currently only dervi.order=0 is implemented") n <- nrow(x) d <- ncol(x) r <- deriv.order D2K0 <- t(dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=2)) K0 <- dmvnorm.deriv(x=rep(0,d), mu=rep(0,d), Sigma=diag(d), deriv.order=0) if(!is.matrix(x)) x <- as.matrix(x) if (missing(pilot)) pilot <- "dscalar" pilot1 <- match.arg(pilot, c("dunconstr", "dscalar")) pre1 <- match.arg(pre, c("scale", "sphere")) optim.fun1 <- match.arg(optim.fun, c("nlm", "optim")) if (pre1=="sphere") stop("Using pre-sphering won't give diagonal bandwidth matrix") if (pilot1=="dunconstr") stop("Unconstrained pilot selectors are not suitable for Hpi.diag.kfe") if (pre1=="scale") { x.star <- pre.scale(x) S12 <- diag(sqrt(diag(var(x)))) Sinv12 <- chol2inv(chol(S12)) } if (nstage==2) { ## stage 1 psi4.ns <- psins(r=r+4, Sigma=var(x.star), deriv.vec=TRUE) if (pilot1=="dscalar") { ## h2 is on pre-transformed data scale A1 <- drop(t(D2K0) %*% D2K0) A2 <- drop(t(D2K0) %*% t(vec(diag(d)) %x% diag(d^2)) %*% psi4.ns) A3 <- drop(t(psi4.ns) %*% (vec(diag(d)) %*% t(vec(diag(d))) %x% diag(d^2)) %*% psi4.ns) h2 <- ((4*d+8)*A1/(-d*A2 + sqrt(d^2*A2^2 + (8*d+16)*A1*A3)))^(1/(d+4))*n^(-1/(d+4)) H2 <- h2^2*diag(d) } psi2.hat <- kfe(x=x.star, G=H2, deriv.order=r+2, add.index=FALSE, binned=binned, bgridsize=bgridsize, verbose=verbose) } else psi2.hat <- psins(r=r+2, Sigma=var(x.star), deriv.vec=TRUE) ## stage 2 amse.temp <- function(diagH) { H <- diag(diagH) %*% diag(diagH) amse.val <- 1/(det(H)^(1/2)*n)*K0 + 1/2* t(vec(H)) %*% psi2.hat return(sum((amse.val^2))) } if (missing(Hstart)) Hstart <- Hns(x=x.star, deriv.order=r) Hstart <- matrix.sqrt(Hstart) if (optim.fun1=="nlm") { result <- nlm(p=diag(Hstart), f=amse.temp, print.level=2*as.numeric(verbose)) H <- diag(result$estimate) %*% diag(result$estimate) amise.star <- result$minimum } else if (optim.fun1=="optim") { result <- optim(diag(Hstart), amse.temp, method="BFGS", control=list(trace=as.numeric(verbose),REPORT=1)) H <- diag(result$par) %*% diag(result$par) amise.star <- result$value } ## back-transform if (pilot1=="dscalar") H <- S12 %*% H %*% S12 if (!amise) return(H) else return(list(H=H, PI=amise.star)) } ks/R/selector.R0000644000176200001440000017032414336776546013067 0ustar liggesusers############################################################################### ## Estimate g_AMSE pilot bandwidths for even orders - 2-dim ## ## Parameters ## r - (r1, r2) partial derivative ## n - sample size ## psi1 - psi_(r + (2,0)) ## psi2 - psi_(r + (0,2)) ## ## Returns ## g_AMSE pilot bandwidths for even orders ############################################################################### gamse.even.2d <- function(r, n, psi1, psi2) { d <- 2 num <- -2 * dmvnorm.deriv(x=c(0,0), deriv.order=r, Sigma=diag(2), deriv.vec=FALSE) den <- (psi1 + psi2) * n g.amse <- (num/den)^(1/(2 + d + sum(r))) return(g.amse) } ############################################################################### ## Estimate g_AMSE pilot bandwidths for odd orders - 2-dim ## ## Parameters ## ## r - (r1, r2) partial derivative ## n - sample size ## psi1 - psi_(r + (2,0)) ## psi2 - psi_(r + (0,2)) ## psi00 - psi_(0,0) ## RK - R(K^(r)) ## ## Returns ## g_AMSE pilot bandwidths for odd orders ############################################################################### gamse.odd.2d <- function(r, n, psi1, psi2, psi00, RK) { d <- 2 num <- 2 * psi00 * (2 * sum(r) + d) * RK den <- (psi1 + psi2)^2 * n^2 g.amse <- (num/den)^(1/(2*sum(r) + d + 4)) return(g.amse) } ############################################################################### ## Estimate g_SAMSE pilot bandwidth - 2- to 6-dim ## ## Parameters ## Sigma.star - scaled variance matrix ## n - sample size ## ## Returns ## g_SAMSE pilot bandwidth ############################################################################### gsamse <- function(Sigma.star, n, modr, nstage=1, psihat=NULL) { d <- ncol(Sigma.star) K <- numeric(); psi <- numeric() ## 4th order g_SAMSE K <- dmvnorm.deriv(x=rep(0,d), deriv.order=modr, Sigma=diag(d), add.index=TRUE, deriv.vec=FALSE) K <- K$deriv[apply(K$deriv.ind, 1, is.even)] if (modr==4) { derivt4 <- dmvnorm.deriv(x=rep(0,d), deriv.order=4, add.index=TRUE, deriv.vec=FALSE, only.index=TRUE) derivt6 <- dmvnorm.deriv(x=rep(0,d), deriv.order=6, add.index=TRUE, deriv.vec=FALSE, only.index=TRUE) for (i in 1:nrow(derivt4)) { r <- derivt4[i,] if (is.even(r)) { A3psi <- 0 for (j in 1:d) { if (nstage==1) A3psi <- A3psi + psins(r=r+2*elem(j,d), Sigma=Sigma.star) else if (nstage==2) A3psi <- A3psi + psihat[which.mat(r=r+2*elem(j,d), mat=derivt6)] } psi <- c(psi, A3psi) } } } ## 6th order g_SAMSE else if (modr==6) { derivt6 <- dmvnorm.deriv(x=rep(0,d), deriv.order=6, add.index=TRUE, deriv.vec=FALSE, only.index=TRUE) for (i in 1:nrow(derivt6)) { r <- derivt6[i,] if (is.even(r)) { A3psi <- 0 for (j in 1:d) A3psi <- A3psi + psins(r=r+2*elem(j,d), Sigma=Sigma.star) psi <- c(psi, A3psi) } } } ## see thesis for formula A1 <- sum(K^2) A2 <- sum(K * psi) A3 <- sum(psi^2) B1 <- (2*modr + 2*d)*A1 B2 <- (modr + d - 2)*A2 B3 <- A3 gamma1 <- (-B2 + sqrt(B2^2 + 4*B1*B3)) / (2*B1) g.samse <- (gamma1 * n)^(-1/(modr + d + 2)) return (g.samse) } ############################################################################## ## Scalar pilot selector for derivatives r>0 from Chacon & Duong (2011) ## Generalisation of gsamse for r>0 ############################################################################## gdscalar <- function(x, d, r, n, verbose, binned=FALSE, nstage=1, scv=FALSE) { if (scv) cf <- c(2^(-d), 2^(-d/2+1), 4) else cf <- c(1,1,1) S <- var(x) Sinv <- chol2inv(chol(S)) if (nstage==1) { psi2r6.ns <- psins(r=2*r+6, Sigma=S) B3 <- norm(Matrix(psi2r6.ns, nrow=1) %*% (Matrix(vec(diag(d)), ncol=1) %*% Matrix(vec(diag(d)), nrow=1) %x% Diagonal(d^(2*r+4))) %*% Matrix(psi2r6.ns, ncol=1), type="1") if (!scv) { B1 <- 2*(2*pi)^(-d)*OF(2*r+4)*prod(d+2*(0:(r+2))) B2 <- -(d+2*r+2)*2^(-d/2-r)*(2*pi)^(-d)*det(S)^(-1/2)*OF(2*r+4)*nu(r=r+2, A=Sinv) } else { B1 <- 2^(-2*r-3)*(4*pi)^(-d)*OF(2*r+6)*prod(d+2*(0:(r+2))) B2 <- -(d+2*r+2)*2^(-2*r-4)*(4*pi)^(-d)*det(S)^(-1/2)*OF(2*r+4)*nu(r=r+2, A=Sinv) B3 <- 4*B3 } g2r4 <- (2*B1/(-B2 + sqrt(B2^2 + 4*B1*B3)))^(1/(d+2*r+8))*n^(-1/(d+2*r+8)) } else if (nstage==2) { psi2r8.ns <- psins(r=2*r+8, Sigma=S) B3 <- norm(Matrix(psi2r8.ns, nrow=1) %*% (Matrix(vec(diag(d)), ncol=1) %*% Matrix(vec(diag(d)), nrow=1) %x% Diagonal(d^(2*r+6))) %*% Matrix(psi2r8.ns, ncol=1), type="1") if (!scv) { B1 <- 2*(2*pi)^(-d)*OF(2*r+6)*prod(d+2*(0:(r+3))) B2 <- -(d+2*r+4)*2^(-d/2-r+2)*(2*pi)^(-d)*det(S)^(-1/2)*OF(2*r+6)*nu(r=r+4, A=Sinv) } else { B1 <- 2^(-2*r-5)*(4*pi)^(-d)*OF(2*r+6)*prod(d+2*(0:(r+3))) B2 <- -(d+2*r+4)*2^(-2*r-6)*(4*pi)^(-d)*det(S)^(-1/2)*OF(2*r+6)*nu(r=r+4, A=Sinv) B3 <- 4*B3 } g2r6.ns <- (2*B1/(-B2 + sqrt(B2^2 + 4*B1*B3)))^(1/(d+2*r+8))*n^(-1/(d+2*r+8)) L0 <- dmvnorm.mixt(x=rep(0,d), mus=rep(0,d), Sigmas=diag(d), props=1) if (binned) eta2r6 <- drop(kfe(x=x, G=g2r6.ns^2*diag(d), inc=1, binned=binned, deriv.order=2*r+6, add.index=FALSE, verbose=verbose) %*% vec(diag(d^(r+3)))) else eta2r6 <- Qr(x=x, deriv.order=2*r+6, Sigma=g2r6.ns^2*diag(d), inc=1) A1 <- cf[1]*(2*d+4*r+8)*L0^2*OF(2*r+4)*nu(r=r+2, A=diag(d)) A2 <- cf[2]*(d+2*r+2)*L0*OF(2*r+4)*eta2r6 A3 <- cf[3]*eta2r6^2 g2r4 <- (2*A1/((-A2+ sqrt(A2^2 +4*A1*A3))*n))^(1/(d+2*r+6)) } return(g2r4) } ############################################################################## ## Unconstrained pilot selector for derivatives r>0 from Chacon & Duong (2011) ## Generalisation of Gunconstr for r>0 ############################################################################## Gdunconstr <- function(x, d, r, n, nstage=1, verbose, binned=FALSE, scv=FALSE, optim.fun="optim") { if (scv) cf <- c(2^(-(d/2+r+2)), 2) else cf <- c(1,1) S <- var(x) S12 <- matrix.sqrt(S) optim.fun <- match.arg(optim.fun, c("nlm", "optim")) if (nstage==2) { # G2r6.NR <- Gns(r=2*r+6,n=n,Sigma=S, scv=scv) # vecPsi2r6 <- kfe(x=x, G=G2r6.NR, binned=binned, deriv.order=2*r+6, deriv.vec=TRUE, add.index=FALSE, verbose=verbose) x.star <- pre.sphere(x) G2r6.NR <- Gns(r=2*r+6,n=n,Sigma=var(x.star), scv=scv) vecPsi2r6 <- kfe(x=x.star, G=G2r6.NR, binned=binned, deriv.order=2*r+6, deriv.vec=TRUE, add.index=FALSE, verbose=verbose) dls <- (0:(d^2-1))*d^(2*r+4) AB2 <- function(vechG) { G <- invvech(vechG) %*% invvech(vechG) Ginv <- chol2inv(chol(G)) ## direct computation v1 <- n^(-1)*det(G)^(-1/2)*(2*pi)^(-d/2)*(-1)^(r+2)*OF(2*r+4)* Sdrv(d=d,r=2*r+4,v=Kpow(vec(Ginv),r+2)) v2 <- numeric(d^(2*r+4)) for(k in 1:d^(2*r+4)) { v2[k] <- sum(vec(G)*vecPsi2r6[dls+k]) } AB <- cf[1]*v1 + cf[2]*(1/2)*v2 AB2.val <- sum(AB^2) return(AB2.val) } Gstart <- Gns(r=2*r+4, n=n, Sigma=var(x.star), scv=scv) Gstart <- matrix.sqrt(Gstart) if (optim.fun=="nlm") { result <- nlm(p=vech(Gstart), f=AB2, print.level=2*as.logical(verbose)) G2r4 <- result$estimate } else if (optim.fun=="optim") { result <- optim(par=vech(Gstart), fn=AB2, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) G2r4 <- result$par } G2r4 <- invvech(G2r4) %*% invvech(G2r4) G2r4 <- S12 %*% G2r4 %*% S12 } else { G2r4 <- Gns(r=2*r+4, n=n, Sigma=S, scv=scv) } return(G2r4) } ############################################################################## ## Estimate psi functionals using 1-stage plug-in ## ## Parameters ## x.star - pre-transformed data points ## pilot - "amse" = different AMSE pilot bandwidths ## - "samse" = optimal SAMSE pilot bandwidth ## ## Returns ## estimated psi functionals ############################################################################### psifun1 <- function(x.star, pilot="samse", binned, bin.par, deriv.order=0, verbose=FALSE) { d <- ncol(x.star) r <- deriv.order S.star <- var(x.star) n <- nrow(x.star) ## pilots are based on (2r+4)-th order derivatives ## compute 1 pilot for SAMSE if (pilot=="samse") { g.star <- gsamse(S.star, n, 4) psihat.star <- kfe(x=x.star, G=g.star^2*diag(d), deriv.order=4, deriv.vec=TRUE, binned=binned, add.index=TRUE, verbose=verbose) } ## compute 5 different pilots for AMSE else if ((pilot=="amse") & (d==2)) { derivt4 <- dmvnorm.deriv(x=rep(0,d), deriv.order=4, add.index=TRUE, deriv.vec=FALSE, only.index=TRUE) derivt4.vec <- dmvnorm.deriv(x=rep(0,d), deriv.order=4, add.index=TRUE, deriv.vec=TRUE, only.index=TRUE) RK31 <- 15/(64*pi) psi00 <- psins(r=c(0,0), Sigma=S.star) psihat.star <- vector() g.star <- vector() for (k in 1:nrow(derivt4)) { r <- derivt4[k,] psi1 <- psins(r=r + 2*elem(1, 2), Sigma=S.star) psi2 <- psins(r=r + 2*elem(2, 2), Sigma=S.star) ## odd order if (prod(r) == 3) g.star[k] <- gamse.odd.2d(r=4, n, psi1, psi2, psi00, RK31) ## even order else g.star[k] <- gamse.even.2d(r=4, n, psi1, psi2)[k] psihat.star[k] <- kfe.scalar(x=x.star, deriv.order=r, g=g.star[k], binned=binned, bin.par=bin.par) } ## create replicated form of psihat psihat.star.vec <- rep(0, nrow(derivt4.vec)) for (k in 1:nrow(derivt4.vec)) psihat.star.vec[k] <- psihat.star[which.mat(r=derivt4.vec[k,], mat=derivt4)] psihat.star <- list(psir=psihat.star.vec, deriv.ind=derivt4.vec) } return(psihat.star) } ############################################################################### ## Estimate psi functionals using 2-stage plug-in ## ## Parameters ## x - pre-transformed data points ## pilot - "amse" - different AMSE pilot ## - "samse" - SAMSE pilot ## Returns ## estimated psi functionals ############################################################################### psifun2 <- function(x.star, pilot="samse", binned, bin.par, deriv.order=0, verbose=FALSE) { d <- ncol(x.star) r <- deriv.order S.star <- var(x.star) n <- nrow(x.star) ## pilots are based on (2r+4)-th order derivatives ## compute 1 pilot for SAMSE if (pilot=="samse") { g6.star <- gsamse(S.star, n=n, modr=6) psihat6.star <- kfe(x=x.star, G=g6.star^2*diag(d), deriv.order=6, deriv.vec=TRUE, binned=binned, add.index=FALSE, verbose=verbose) g.star <- gsamse(S.star, n=n, modr=4, nstage=2, psihat=psihat6.star) psihat.star <- kfe(x=x.star, G=g.star^2*diag(d), deriv.order=4, deriv.vec=TRUE, binned=binned, add.index=TRUE, verbose=verbose) } ## compute different pilots for AMSE else if ((pilot=="amse") & (d==2)) { derivt4 <- dmvnorm.deriv(x=rep(0,d), deriv.order=4, add.index=TRUE, deriv.vec=FALSE, only.index=TRUE) derivt4.vec <- dmvnorm.deriv(x=rep(0,d), deriv.order=4, add.index=TRUE, deriv.vec=TRUE, only.index=TRUE) derivt6 <- dmvnorm.deriv(x=rep(0,d), deriv.order=6, add.index=TRUE, deriv.vec=FALSE, only.index=TRUE) RK31 <- 15/(64*pi) RK51 <- 945/(256*pi) RK33 <- 225/(256*pi) psi00 <- psins(r=rep(0,d), Sigma=S.star) psihat6.star <- vector() g6.star <- vector() psihat.star <- vector() g.star <- vector() for (k in 1:nrow(derivt6)) { r <- derivt6[k,] psi1 <- psins(r=r + 2*elem(1, 2), Sigma=S.star) psi2 <- psins(r=r + 2*elem(2, 2), Sigma=S.star) if (prod(r) == 5) g6.star[k] <- gamse.odd.2d(r=6, n, psi1, psi2, psi00, RK51) else if (prod(r) == 9) g6.star[k] <- gamse.odd.2d(r=6, n, psi1, psi2, psi00, RK33) else g6.star[k] <- gamse.even.2d(r=6, n, psi1, psi2)[k] psihat6.star[k] <- kfe.scalar(x=x.star, deriv.order=r, g=g6.star[k], binned=binned, bin.par=bin.par) } ## pilots are based on 4th order derivatives using 6th order psi functionals ## computed above 'psihat6.star' for (k in 1:nrow(derivt4)) { r <- derivt4[k,] psi1 <- psihat6.star[7 - (r + 2*elem(1,2))[1]] psi2 <- psihat6.star[7 - (r + 2*elem(2,2))[1]] if (prod(r) == 3) g.star[k] <- gamse.odd.2d(r=4, n, psi1, psi2, psi00, RK31) else g.star[k] <- gamse.even.2d(r=4, n, psi1, psi2)[k] psihat.star[k] <- kfe.scalar(x=x.star, deriv.order=r, g=g.star[k], binned=binned, bin.par=bin.par) } ## create replicated form of psihat psihat.star.vec <- rep(0, nrow(derivt4.vec)) for (k in 1:nrow(derivt4.vec)) psihat.star.vec[k] <- psihat.star[which.mat(r=derivt4.vec[k,], mat=derivt4)] psihat.star <- list(psir=psihat.star.vec, deriv.ind=derivt4.vec) } return(psihat.star) } ############################################################################# ## Estimate psi functionals for 6-variate data using 1-stage plug-in ## with unconstrained pilot ## ## Parameters ## x - data points ## Sd4, Sd6 - symmetrizer matrices of order 4 and 6 ## ## Returns ## estimated psi functionals ############################################################################# psifun1.unconstr <- function(x, binned, bgridsize, deriv.order=0, verbose=FALSE) { n <- nrow(x) r <- deriv.order S <- var(x) ## stage 1 of plug-in G2r4 <- Gns(r=2*r+4,n=n,Sigma=S) vecPsi2r4 <- kfe(x=x, G=G2r4, deriv.order=2*r+4, binned=binned, bgridsize=bgridsize, deriv.vec=TRUE, add.index=FALSE, verbose=verbose) return (vecPsi2r4) } ############################################################################# ## Estimate psi functionals for 6-variate data using 2-stage plug-in ## with unconstrained pilot ## ## Parameters ## x - data points ## Sd4, Sd6 - symmetrizer matrices of order 4 and 6 ## ## Returns ## estimated psi functionals ############################################################################ psifun2.unconstr <- function(x, binned, bgridsize, deriv.order=0, verbose=FALSE, optim.fun="optim") { d <- ncol(x) n <- nrow(x) S <- var(x) S12 <- matrix.sqrt(S) r <- deriv.order optim.fun <- match.arg(optim.fun, c("nlm", "optim")) ## stage 1 of plug-in x.star <- pre.sphere(x) G2r6 <- Gns(r=2*r+6,n=n,Sigma=var(x.star)) vecPsi2r6 <- kfe(x=x.star, G=G2r6, binned=binned, bgridsize=bgridsize, deriv.order=2*r+6, deriv.vec=TRUE, add.index=FALSE, verbose=verbose) ## asymptotic squared bias for r = 4 for MSE-optimal G D2r4phi0 <- DrL0(d=d, r=2*r+4) Id2r4 <- diag(d^(2*r+4)) AB2 <- function(vechG) { rr <- 2*r+4 G <- invvech(vechG)%*%invvech(vechG) G12 <- matrix.sqrt(G) Ginv12 <- chol2inv(chol(G12)) AB <- n^(-1)*det(Ginv12)*(Kpow(A=Ginv12,pow=rr)%*%D2r4phi0)+(1/2)*(t(vec(G))%x%Id2r4) %*% vecPsi2r6 AB2.val <- sum(AB^2) return (AB2.val) } Gstart <- Gns(r=2*r+4,n=n,Sigma=var(x.star)) Gstart <- matrix.sqrt(Gstart) if (optim.fun=="nlm") { res <- nlm(p=vech(Gstart), f=AB2, print.level=2*as.logical(verbose)) G2r4 <- res$estimate } else if (optim.fun=="optim") { res <- optim(vech(Gstart), AB2, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) G2r4 <- res$par } G2r4 <- invvech(G2r4)%*%invvech(G2r4) G2r4 <- S12 %*% G2r4 %*% S12 ## stage 2 of plug-in vecPsi2r4 <- kfe(x=x, G=G2r4, binned=binned, bgridsize=bgridsize, deriv.order=2*r+4, deriv.vec=TRUE, add.index=FALSE, verbose=verbose) return (vecPsi2r4) } ############################################################################# ## Plug-in bandwidth selectors ############################################################################# ############################################################################ ## Computes plug-in full bandwidth matrix - 2 to 6 dim ## ## Parameters ## x - data points ## Hstart - initial value for minimisation ## nstage - number of plug-in stages (1 or 2) ## pilot - "amse" - different AMSE pilot ## - "samse" - SAMSE pilot ## - "unconstr" - unconstrained pilot ## pre - "scale" - pre-scaled data ## - "sphere"- pre-sphered data ## ## Returns ## Plug-in full bandwidth matrix ############################################################################### hpi <- function(x, nstage=2, binned=TRUE, bgridsize, deriv.order=0) { ## 1-d selector is taken from KernSmooth's dpik d <- 1 if (missing(bgridsize)) bgridsize <- default.gridsize(d) if (deriv.order==0) h <- dpik(x=x, level=nstage, gridsize=bgridsize) else { n <- length(x) d <- 1 r <- deriv.order K2r4 <- dnorm.deriv(x=0, mu=0, sigma=1, deriv.order=2*r+4) K2r6 <- dnorm.deriv(x=0, mu=0, sigma=1, deriv.order=2*r+6) m2 <- 1 mr <- psins.1d(r=2*r, sigma=1) ## formula for bias annihilating bandwidths from Wand & Jones (1995, p.70) if (nstage==2) { psi2r8.hat <- psins.1d(r=2*r+8, sigma=sd(x)) gamse2r6 <- (2*K2r6/(-m2*psi2r8.hat*n))^(1/(2*r+9)) psi2r6.hat <- kfe.1d(x=x, g=gamse2r6, deriv.order=2*r+6, inc=1, binned=binned) gamse2r4 <- (2*K2r4/(-m2*psi2r6.hat*n))^(1/(2*r+7)) psi2r4.hat <- kfe.1d(x=x, g=gamse2r4, deriv.order=2*r+4, inc=1, binned=binned) } else { psi2r6.hat <- psins.1d(r=2*r+6, sigma=sd(x)) gamse2r4 <- (2*K2r4/(-m2*psi2r6.hat*n))^(1/(2*r+7)) psi2r4.hat <- kfe.1d(x=x, g=gamse2r4, deriv.order=2*r+4, inc=1, binned=binned) } ## formula form Wand & Jones (1995, p.49) h <- ((2*r+1)*mr/(m2^2*psi2r4.hat*n))^(1/(2*r+5)) } return(h) } Hpi <- function(x, nstage=2, pilot, pre="sphere", Hstart, binned, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") { n <- nrow(x) d <- ncol(x) r <- deriv.order if (missing(binned)) binned <- default.bflag(d=d,n=n) if (d > 4) binned <- FALSE if (missing(bgridsize)) bgridsize <- default.bgridsize(d) if (!is.matrix(x)) x <- as.matrix(x) if (missing(pilot)) { if (d==2 & r==0) pilot <- "samse" else pilot <- "dscalar" } pilot1 <- match.arg(pilot, c("amse", "samse", "unconstr", "dunconstr", "dscalar")) pre1 <- match.arg(pre, c("scale", "sphere")) optim.fun <- match.arg(optim.fun, c("nlm", "optim")) if (pilot1=="amse" & (d>2 | r>0)) stop("amse pilot selectors not defined for d>2 and/or r>0") if ((pilot1=="samse" | pilot1=="unconstr") & r>0) stop("dscalar or dunconstr pilot selectors are better for derivatives r>0") if (pilot1=="unconstr" & d>=6) stop("Unconstrained pilots are not implemented for d>6") if (pre1=="scale") { x.star <- pre.scale(x) S12 <- diag(sqrt(diag(var(x)))) Sinv12 <- chol2inv(chol(S12)) } else if (pre1=="sphere") { x.star <- pre.sphere(x) S12 <- matrix.sqrt(var(x)) Sinv12 <- chol2inv(chol(S12)) } Idr <- diag(d^r) RKr <- nu(r=r, diag(d))*2^(-d-r)*pi^(-d/2) if (binned) { H.max <- (((d+8)^((d+6)/2)*pi^(d/2)*RKr)/(16*(d+2)*n*gamma(d/2+4)))^(2/(d+4))* var(x) bin.par <- binning(x=x, bgridsize=bgridsize, H=H.max) H.max.star <- (((d+8)^((d+6)/2)*pi^(d/2)*RKr)/(16*(d+2)*n*gamma(d/2+4)))^(2/(d+4))* var(x.star) bin.par.star <- binning(x=x.star, bgridsize=bgridsize, H=H.max.star) } if (pilot1=="unconstr") { ## psi4.mat is on data scale if (nstage==1) psi.fun <- (-1)^r*psifun1.unconstr(x=x, binned=binned, bgridsize=bgridsize, deriv.order=r, verbose=verbose) else if (nstage==2) psi.fun <- psifun2.unconstr(x=x, binned=binned, bgridsize=bgridsize, deriv.order=r, verbose=verbose) psi2r4.mat <- (-1)^r*invvec(psi.fun) ## use normal reference bandwidth as initial condition if (missing(Hstart)) Hstart <- Hns(x=x, deriv.order=r) } else if (pilot1=="dunconstr") { ## G2r4 is on data scale G2r4 <- Gdunconstr(x=x, d=d, r=r, n=n, nstage=nstage, verbose=verbose, binned=binned, optim.fun=optim.fun) vecPsi2r4 <- kfe(x=x, G=G2r4, binned=binned, deriv.order=2*r+4, deriv.vec=TRUE, add.index=FALSE, verbose=verbose, bin.par=bin.par) if (missing(Hstart)) Hstart <- Hns(x=x, deriv.order=r) } else if (pilot1=="dscalar") { ## g2r4 is on pre-transformed data scale g2r4 <- gdscalar(x=x.star, r=r, n=n, d=d, verbose=verbose, nstage=nstage, binned=binned) G2r4 <- g2r4^2 * diag(d) vecPsi2r4 <- kfe(x=x.star, G=G2r4, binned=binned, deriv.order=2*r+4, deriv.vec=TRUE, add.index=FALSE, verbose=verbose, bin.par=bin.par.star) if (missing(Hstart)) Hstart <- Hns(x=x.star, deriv.order=r) } else { ## psi4.mat is on pre-transformed data scale if (nstage==1) psi.fun <- psifun1(x.star, pilot=pilot, binned=binned, bin.par=bin.par.star, deriv.order=r, verbose=verbose)$psir else if (nstage==2) psi.fun <- psifun2(x.star, pilot=pilot, binned=binned, bin.par=bin.par.star, deriv.order=r, verbose=verbose)$psir psi2r4.mat <- invvec(psi.fun) ## use normal reference bandwidth as initial condition if (missing(Hstart)) Hstart <- Hns(x=x.star, deriv.order=r) else Hstart <- Sinv12 %*% Hstart %*% Sinv12 } ## PI is estimate of AMISE pi.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) Hinv <- chol2inv(chol(H)) IdrvH <- Idr%x%vec(H) int.var <- 1/(det(H)^(1/2)*n)*nur(r=r, A=Hinv, mu=rep(0,d), Sigma=diag(d))*2^(-d-r)*pi^(-d/2) if (pilot1=="dunconstr" | pilot1=="dscalar") pi.val <- int.var + (-1)^r*1/4*vecPsi2r4 %*% (vec(diag(d^r) %x% vec(H) %x% vec(H))) else pi.val <- int.var + (-1)^r*1/4* sum(diag(t(IdrvH) %*% psi2r4.mat %*% IdrvH)) pi.val <- drop(pi.val) return(pi.val) } Hstart <- matrix.sqrt(Hstart) if (optim.fun=="nlm") { result <- nlm(p=vech(Hstart), f=pi.temp, print.level=2*as.numeric(verbose)) H <- invvech(result$estimate) %*% invvech(result$estimate) amise.star <- result$minimum } else if (optim.fun=="optim") { result <- optim(vech(Hstart), pi.temp, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) H <- invvech(result$par) %*% invvech(result$par) amise.star <- result$value } if (!(pilot1 %in% c("dunconstr","unconstr"))) H <- S12 %*% H %*% S12 ## back-transform if (!amise) return(H) else return(list(H = H, PI.star=amise.star)) } ############################################################################### ## Computes plug-in diagonal bandwidth matrix for 2 to 6-dim ## ## Parameters ## x - data points ## nstage - number of plug-in stages (1 or 2) ## pre - "scale" - pre-scaled data ## - "sphere"- pre-sphered data ## ## Returns ## Plug-in diagonal bandwidth matrix ############################################################################### Hpi.diag <- function(x, nstage=2, pilot, pre="scale", Hstart, binned, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") { n <- nrow(x) d <- ncol(x) r <- deriv.order RK <- (4*pi)^(-d/2) if (missing(binned)) binned <- default.bflag(d=d,n=n) if (d > 4) binned <- FALSE if (missing(bgridsize)) bgridsize <- default.bgridsize(d) if (!is.matrix(x)) x <- as.matrix(x) if (missing(pilot)) { if (d==2 & r==0) pilot <- "samse" else pilot <- "dscalar" } pilot1 <- match.arg(pilot, c("amse", "samse", "unconstr", "dunconstr", "dscalar")) pre1 <- match.arg(pre, c("scale", "sphere")) optim.fun <- match.arg(optim.fun, c("nlm", "optim")) if (pre1=="sphere") stop("Using pre-sphering won't give diagonal bandwidth matrix") if (pilot1=="amse" & (d>2 | r>0)) stop("samse pilot selectors are better for higher dimensions and/or deriv.order>0") if (pilot1=="samse" & r>0) stop("dscalar or dunconstr pilot selectors are better for derivatives r>0") if (pilot1=="unconstr" | pilot1=="dunconstr") stop("Unconstrained pilot selectors are not suitable for Hpi.diag") if (pre1=="scale") { x.star <- pre.scale(x) S12 <- diag(sqrt(diag(var(x)))) Sinv12 <- chol2inv(chol(S12)) } else if (pre1=="sphere") { x.star <- pre.sphere(x) S12 <- matrix.sqrt(var(x)) Sinv12 <- chol2inv(chol(S12)) } if (binned) { H.max <- (((d+8)^((d+6)/2)*pi^(d/2)*RK)/(16*(d+2)*n*gamma(d/2+4)))^(2/(d+4))* var(x.star) bin.par <- binning(x=x.star, bgridsize=bgridsize, H=H.max) } Idr <- diag(d^r) if (pilot1=="amse" | pilot1=="samse") { if (nstage==1) psi.fun <- psifun1(x.star, pilot=pilot, binned=binned, bin.par=bin.par, deriv.order=r, verbose=verbose)$psir else if (nstage==2) psi.fun <- psifun2(x.star, pilot=pilot, binned=binned, bin.par=bin.par, deriv.order=r, verbose=verbose)$psir psi2r4.mat <- invvec(psi.fun) } else if (pilot1=="dscalar") { g2r4 <- gdscalar(x=x.star, r=r, n=n, d=d, verbose=verbose, nstage=nstage, binned=binned) G2r4 <- g2r4^2 * diag(d) vecPsi2r4 <- kfe(x=x.star, G=G2r4, binned=binned, deriv.order=2*r+4, deriv.vec=TRUE, add.index=FALSE, verbose=verbose) } if (d==2 & r==0 & (pilot1=="amse" | pilot1=="samse")) { ## diagonal bandwidth matrix for 2-dim has exact formula psi40 <- psi.fun[1] psi22 <- psi.fun[6] psi04 <- psi.fun[16] s1 <- sd(x[,1]) s2 <- sd(x[,2]) h1 <- (psi04^(3/4)*RK/(psi40^(3/4)*(sqrt(psi40*psi04)+psi22)*n))^(1/6) h2 <- (psi40/psi04)^(1/4) * h1 H <- diag(c(s1^2*h1^2, s2^2*h2^2)) psimat4.D <- invvech(c(psi40, psi22, psi04)) amise.star <- drop(n^(-1)*RK*(h1*h2)^(-1) + 1/4*c(h1,h2)^2 %*% psimat4.D %*% c(h1,h2)^2) } else { ## PI is estimate of AMISE pi.temp <- function(diagH) { H <- diag(diagH) %*% diag(diagH) Hinv <- chol2inv(chol(H)) IdrvH <- Idr%x%vec(H) int.var <- 1/(det(H)^(1/2)*n)*nu(r=r, Hinv)*2^(-d-r)*pi^(-d/2) if (pilot1=="dscalar") pi.val <- int.var + (-1)^r*1/4*vecPsi2r4 %*% (vec(diag(d^r) %x% vec(H) %x% vec(H))) else pi.val <- int.var + (-1)^r*1/4* sum(diag(t(IdrvH) %*% psi2r4.mat %*% IdrvH)) return(drop(pi.val)) } ## use normal reference bandwidth as initial condition if (missing(Hstart)) Hstart <- Hns(x=x.star, deriv.order=r) else Hstart <- Sinv12 %*% Hstart %*% Sinv12 Hstart <- matrix.sqrt(Hstart) if (optim.fun=="nlm") { result <- nlm(p=diag(Hstart), f=pi.temp, print.level=2*as.numeric(verbose)) H <- diag(result$estimate^2) amise.star <- result$minimum } else if (optim.fun=="optim") { result <- optim(diag(Hstart), pi.temp, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) H <- diag(result$par) %*% diag(result$par) amise.star <- result$value } H <- S12 %*% H %*% S12 ## back-transform } if (!amise) return(H) else return(list(H = H, PI.star=amise.star)) } ############################################################################### ## Cross-validation bandwidth selectors ############################################################################### ############################################################################### ## Computes the least squares cross validation LSCV function for 2 to 6 dim ## ## Parameters ## x - data values ## H - bandwidth matrix ## ## Returns ## LSCV(H) ############################################################################### lscv.1d <- function(x, h, binned, bin.par, deriv.order=0) { r <- deriv.order lscv1 <- kfe.1d(x=x, g=sqrt(2)*h, inc=1, binned=binned, bin.par=bin.par, deriv.order=2*r) lscv2 <- kfe.1d(x=x, g=h, inc=0, binned=binned, bin.par=bin.par, deriv.order=2*r) return((-1)^r*(lscv1 - 2*lscv2)) } lscv.mat <- function(x, H, binned=FALSE, bin.par, bgridsize, deriv.order=0) { r <- deriv.order d <- ncol(x) n <- nrow(x) if (!binned) { lscv1 <- Qr(x=x, deriv.order=2*r, Sigma=2*H, inc=1) lscv2 <- Qr(x=x, deriv.order=2*r, Sigma=H, inc=0) lscv <- drop(lscv1 - 2*lscv2) } else { lscv1 <- kfe(x=x, G=2*H, inc=1, binned=binned, bin.par=bin.par, bgridsize=bgridsize, deriv.order=2*r, add.index=FALSE) lscv2 <- kfe(x=x, G=H, inc=0, binned=binned, bin.par=bin.par, bgridsize=bgridsize, deriv.order=2*r, add.index=FALSE) lscv <- (-1)^2*drop((lscv1 - 2*lscv2) %*% vec(diag(d^r))) } return(lscv) } ############################################################################### ## Finds the bandwidth matrix that minimises LSCV for 2 to 6 dim ## ## Parameters ## x - data values ## Hstart - initial bandwidth matrix ## ## Returns ## H_LSCV ############################################################################### hlscv <- function(x, binned=TRUE, bgridsize, amise=FALSE, deriv.order=0, bw.ucv=TRUE) { ## adapted from versions supplied by J.E. Chacon 19/02/2021 if (any(duplicated(x))) warning("Data contain duplicated values: LSCV is not well-behaved in this case") n <- length(x) d <- 1 r <- deriv.order hnorm <- sqrt((4/(n*(d + 2)))^(2/(d + 4)) * var(x)) if (missing(binned)) binned <- default.bflag(d=d,n=n) if (missing(bgridsize)) if (bw.ucv) bgridsize <- 10000 else bgridsize <- 10001 ## use stats::bw.ucv function if (bw.ucv) { difs <- dist(x) lower <- min(difs[difs>0]) opt <- list() opt$minimum <- stats::bw.ucv(x=x, nb=bgridsize, lower=lower) opt$objective <- NA } else { difs <- x%*%t(rep(1,n))-rep(1,n)%*%t(x) difs <- difs[lower.tri(difs)] if (binned) { bin.par <- binning(x, bgridsize=bgridsize, h=hnorm) lscv.1d.temp <- function(h) { return(lscv.1d(x=x, h=h, binned=binned, bin.par=bin.par, deriv.order=r)) } } else { if (r>0) stop("Unbinned hlscv not yet implemented for deriv.order>0") edifs <- exp(-difs^2/2) RK <- 1/(2*sqrt(pi)) lscv.1d.temp <- function(h) { lscv1 <- (1-1/n)*sum(edifs^(1/(2*h^2)))/(h*sqrt(2)*sqrt(2*pi)) lscv2 <- 2*sum(edifs^(1/h^2))/(h*sqrt(2*pi)) return(RK/(n*h)+2*(lscv1-lscv2)/(n^2-n)) } } lower <- min(difs[difs>0]) opt <- optimise(lscv.1d.temp, interval=c(lower, 2*hnorm), tol=.Machine$double.eps) } if (!amise) return(opt$minimum) else return(list(h=opt$minimum, LSCV=opt$objective)) } Hlscv <- function(x, Hstart, binned=FALSE, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim", trunc) { if (any(duplicated(x))) warning("Data contain duplicated values: LSCV is not well-behaved in this case") if (!is.matrix(x)) x <- as.matrix(x) n <- nrow(x) d <- ncol(x) r <- deriv.order if (missing(binned)) binned <- default.bflag(d=d,n=n) if (d>4) binned <- FALSE if (missing(bgridsize)) bgridsize <- default.bgridsize(d) optim.fun <- match.arg(optim.fun, c("nlm", "optim")) ## use normal reference selector as initial condn Hnorm <- Hns(x=x, deriv.order=r) if (missing(Hstart)) Hstart <- Hnorm Hstart <- matrix.sqrt(Hstart) if (binned) bin.par <- binning(x=x, H=Hnorm, bgridsize=bgridsize) if (missing(trunc)) { if (deriv.order==0) trunc <- 1e10 else trunc <- 4 } lscv.init <- lscv.mat(x=x, H=Hnorm, binned=binned, bin.par=bin.par, deriv.order=r) lscv.mat.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) lscv <- lscv.mat(x=x, H=H, binned=binned, bin.par=bin.par, deriv.order=r) if (det(H) < 1/trunc*det(Hnorm) | det(H) > trunc*det(Hnorm) | abs(lscv) > trunc*abs(lscv.init)) lscv <- lscv.init return(lscv) } if (optim.fun=="nlm") { result <- nlm(p=vech(Hstart), f=lscv.mat.temp, print.level=2*as.numeric(verbose)) H <- invvech(result$estimate) %*% invvech(result$estimate) amise.opt <- result$minimum } else if (optim.fun=="optim") { result <- optim(vech(Hstart), lscv.mat.temp, method="Nelder-Mead", control=list(trace=as.numeric(verbose), REPORT=1)) H <- invvech(result$par) %*% invvech(result$par) amise.opt <- result$value } if (!amise) return(H) else return(list(H=H, LSCV=amise.opt)) } ############################################################################### ## Finds the diagonal bandwidth matrix that minimises LSCV for 2 to 6 dim ## ## Parameters ## x - data values ## Hstart - initial bandwidth matrix ## ## Returns ## H_LSCV,diag ############################################################################### Hlscv.diag <- function(x, Hstart, binned=FALSE, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim", trunc) { if (any(duplicated(x))) warning("Data contain duplicated values: LSCV is not well-behaved in this case") if (!is.matrix(x)) x <- as.matrix(x) n <- nrow(x) d <- ncol(x) r <- deriv.order if (missing(binned)) binned <- default.bflag(d=d,n=n) if (d>4) binned <- FALSE if (missing(bgridsize)) bgridsize <- default.bgridsize(d) optim.fun <- match.arg(optim.fun, c("nlm", "optim")) Hnorm <- Hns(x=x, deriv.order=r) if (missing(Hstart)) Hstart <- Hnorm ## don't truncate optimisation for deriv.order==0 if (missing(trunc)) { if (deriv.order==0) trunc <- 1e10 else trunc <- 4 } ## linear binning if (binned) { bin.par <- binning(x=x, bgridsize=bgridsize, H=Hnorm) } lscv.init <- lscv.mat(x=x, H=Hnorm, binned=binned, bin.par=bin.par, deriv.order=r) lscv.mat.temp <- function(diagH) { H <- diag(diagH^2) lscv <- lscv.mat(x=x, H=H, binned=binned, bin.par=bin.par, deriv.order=r) if (det(H) < 1/trunc*det(Hnorm) | det(H) > trunc*det(Hnorm) | abs(lscv) > trunc*abs(lscv.init)) lscv <- lscv.init return(lscv) } Hstart <- matrix.sqrt(Hstart) if (optim.fun=="nlm") { result <- nlm(p=diag(Hstart), f=lscv.mat.temp, print.level=2*as.numeric(verbose)) H <- diag(result$estimate^2) amise.opt <- result$minimum } else if (optim.fun=="optim") { result <- optim(diag(Hstart), lscv.mat.temp, method="Nelder-Mead", control=list(trace=as.numeric(verbose), REPORT=1)) H <- diag(result$par^2) amise.opt <- result$value } if (!amise) return(H) else return(list(H=H, LSCV=amise.opt)) } ## aliases using "UCV" instead of "LSCV" hucv <- function(...) { hlscv(...) } Hucv <- function(...) { Hlscv(...) } Hucv.diag <- function(...) { Hlscv.diag(...) } ############################################################################### ## Computes the biased cross validation BCV function for 2-dim ## ## Parameters ## x - data values ## H1, H2 - bandwidth matrices ## ## Returns ## BCV(H) ############################################################################### bcv.mat <- function(x, H1, H2, binned=FALSE) { n <- nrow(x) d <- 2 psi <- kfe(x, G=H2, deriv.order=4, add.index=TRUE, deriv.vec=TRUE, inc=0, binned=binned) psi40 <- psi$psir[1] psi31 <- psi$psir[2] psi22 <- psi$psir[4] psi13 <- psi$psir[8] psi04 <- psi$psir[16] coeff <- c(1, 2, 1, 2, 4, 2, 1, 2, 1) psi.fun <- c(psi40, psi31, psi22, psi31, psi22, psi13, psi22, psi13,psi04)/(n*(n-1)) psi4.mat <- matrix(coeff * psi.fun, ncol=3, nrow=3) RK <- (4*pi)^(-d/2) bcv <- drop(n^(-1)*det(H1)^(-1/2)*RK + 1/4*t(vech(H1)) %*% psi4.mat %*% vech(H1)) return(list(bcv=bcv, psimat=psi4.mat)) } ############################################################################### ## Find the bandwidth matrix that minimises the BCV for 2-dim ## ## Parameters ## x - data values ## whichbcv - 1 = BCV1 ## - 2 = BCV2 ## Hstart - initial bandwidth matrix ## ## Returns ## H_BCV ############################################################################### Hbcv <- function(x, whichbcv=1, Hstart, binned=FALSE, amise=FALSE, verbose=FALSE) { n <- nrow(x) d <- ncol(x) RK <- (4*pi)^(-d/2) if(!is.matrix(x)) x <- as.matrix(x) ## use normal reference b/w matrix for bounds k <- (((d+8)^((d+6)/2)*pi^(d/2)*RK)/(16*n*gamma((d+8)/2)*(d+2)))^(2/(d+4)) Hmax <- k * abs(var(x)) up.bound <- Hmax if (missing(Hstart)) Hstart <- matrix.sqrt(0.9*Hmax) bcv.mat.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) bcv <- bcv.mat(x=x, H1=H, H2=whichbcv*H, binned=binned)$bcv return(bcv) } result <- optim(vech(Hstart), bcv.mat.temp, method="L-BFGS-B", upper=vech(matrix.sqrt(up.bound)), lower=-vech(matrix.sqrt(up.bound)), control=list(trace=as.numeric(verbose), REPORT=1)) H <- invvech(result$par) %*% invvech(result$par) amise.opt <- result$value if (!amise) return(H) else return(list(H = H, BCV=amise.opt)) } ############################################################################### ## Find the diagonal bandwidth matrix that minimises the BCV for 2-dim ## ## Parameters ## x - data values ## whichbcv - 1 = BCV1 ## - 2 = BCV2 ## Hstart - initial bandwidth matrix ## ## Returns ## H_BCV, diag ############################################################################### Hbcv.diag <- function(x, whichbcv=1, Hstart, binned=FALSE, amise=FALSE, verbose=FALSE) { n <- nrow(x) d <- ncol(x) RK <- (4*pi)^(-d/2) if(!is.matrix(x)) x <- as.matrix(x) ## use maximally smoothed b/w matrix for bounds k <- (((d+8)^((d+6)/2)*pi^(d/2)*RK)/(16*n*gamma((d+8)/2)*(d+2)))^(2/(d+4)) Hmax <- k * abs(var(x)) up.bound <- diag(Hmax) if (missing(Hstart)) Hstart <- 0.9*matrix.sqrt(Hmax) bcv.mat.temp <- function(diagH) { H <- diag(diagH) %*% diag(diagH) return(bcv.mat(x, H, whichbcv*H, binned=binned)$bcv) } result <- optim(diag(Hstart), bcv.mat.temp, method="L-BFGS-B", upper=sqrt(up.bound), control=list(trace=as.numeric(verbose), REPORT=1)) H <- diag(result$par) %*% diag(result$par) amise.opt <- result$value if (!amise) return(H) else return(list(H = H, BCV=amise.opt)) } ############################################################################### ## Estimate scalar g_AMSE pilot bandwidth for SCV for 2 to 6 dim ## ## Parameters ## Sigma.star - scaled/ sphered variance matrix ## Hamise - (estimate) of H_AMISE ## n - sample size ## ## Returns ## g_AMSE pilot bandwidth ############################################################################### Theta6.elem <- function(d) { Theta6.mat <- list() Theta6.mat[[d]] <- list() for (i in 1:d) Theta6.mat[[i]] <- list() for (i in 1:d) for (j in 1:d) { temp <- numeric() for (k in 1:d) for (ell in 1:d) temp <- rbind(temp, elem(i,d)+2*elem(k,d)+2*elem(ell,d)+elem(j,d)) Theta6.mat[[i]][[j]] <- temp } return(Theta6.mat) } gamse.scv <- function(x.star, d, Sigma.star, Hamise, n, binned=FALSE, bin.par, bgridsize, verbose=FALSE, nstage=1, Theta6=FALSE) { if (nstage==0) { psihat6.star <- psins(r=6, Sigma=Sigma.star, deriv.vec=TRUE) } else if (nstage==1) { g6.star <- gsamse(Sigma.star, n, 6) G6.star <- g6.star^2 * diag(d) if (Theta6) psihat6.star <- kfe(x=x.star, bin.par=bin.par, deriv.order=6, G=G6.star, deriv.vec=FALSE, add.index=FALSE, binned=binned, bgridsize=bgridsize, verbose=verbose) else psihat6.star <- kfe(x=x.star, bin.par=bin.par, deriv.order=6, G=G6.star, deriv.vec=TRUE, add.index=FALSE, binned=binned, bgridsize=bgridsize, verbose=verbose) } if (Theta6) { derivt6 <- dmvnorm.deriv(x=rep(0,d), deriv.order=6, add.index=TRUE, deriv.vec=FALSE, only.index=TRUE) Theta6.mat <- matrix(0, ncol=d, nrow=d) Theta6.mat.ind <- Theta6.elem(d) for (i in 1:d) for (j in 1:d) { temp <- Theta6.mat.ind[[i]][[j]] temp.sum <- 0 for (k in 1:nrow(temp)) temp.sum <- temp.sum + psihat6.star[which.mat(temp[k,], derivt6)] Theta6.mat[i,j] <- temp.sum } eye3 <- diag(d) D4 <- dupl(d)$d trHamise <- tr(Hamise) ## required constants - see thesis Cmu1 <- 1/2*t(D4) %*% vec(Theta6.mat %*% Hamise) Cmu2 <- 1/8*(4*pi)^(-d/2) * (2*t(D4)%*% vec(Hamise) + trHamise * t(D4) %*% vec(eye3)) num <- 2 * (d+4) * sum(Cmu2*Cmu2) den <- -(d+2) * sum(Cmu1*Cmu2) + sqrt((d+2)^2 * sum(Cmu1*Cmu2)^2 + 8*(d+4)*sum(Cmu1*Cmu1) * sum(Cmu2*Cmu2)) gamse <- (num/(den*n))^(1/(d+6)) } else { ## updated constants using Chacon & Duong (2010) notation Cmu1Cmu1 <- drop(1/4*psihat6.star %*% (Hamise %x% diag(d^4) %x% Hamise) %*% psihat6.star) Cmu1Cmu2 <- 3/4*(4*pi)^(-d/2)*drop(vec(Hamise %x% diag(d) %x% Hamise) %*% psihat6.star) Cmu2Cmu2 <- 1/64*(4*pi)^(-d)*(4*tr(Hamise%*%Hamise) + (d+8)*tr(Hamise)^2) num <- 2 * (d+4) * Cmu2Cmu2 den <- -(d+2) * Cmu1Cmu2 + sqrt((d+2)^2 * Cmu1Cmu2^2 + 8*(d+4)*Cmu1Cmu1 * Cmu2Cmu2) gamse <- (num/(den*n))^(1/(d+6)) } return(gamse) } ############################################################################### ## Estimate unconstrained G_AMSE pilot bandwidth for SCV for 2 to 6 dim ## Code by J.E. Chacon ## ## Returns ## G_AMSE pilot bandwidth ############################################################################### Gunconstr.scv <- function(x, binned=FALSE, bin.par, bgridsize, verbose=FALSE, nstage=1, optim.fun="optim") { d <- ncol(x) n <- nrow(x) S <- var(x) S12 <- matrix.sqrt(S) optim.fun <- match.arg(optim.fun, c("nlm", "optim")) ## stage 1 of plug-in x.star <- pre.sphere(x) if (nstage==1) { G6 <- Gns(r=6,n=n,Sigma=var(x.star), scv=TRUE) psihat6 <- kfe(x=x.star, deriv.order=6, G=G6, deriv.vec=TRUE, add.index=FALSE, binned=binned, bgridsize=bgridsize, verbose=verbose) } else if (nstage==0) { psihat6 <- psins(r=6, Sigma=var(x.star), deriv.vec=TRUE) } ## constants for normal reference D4phi0 <- DrL0(d=d,r=4) Id4 <- diag(d^4) ## asymptotic squared bias for r = 4 AB2 <- function(vechG) { G <- invvech(vechG)%*%invvech(vechG) G12 <- matrix.sqrt(G) Ginv12 <- chol2inv(chol(G12)) AB <- n^(-1)*det(Ginv12)*(Kpow(A=Ginv12,pow=4)%*%D4phi0)*2^(-(d+4)/2) + (t(vec(G))%x%Id4)%*%psihat6 AB2.val <- sum(AB^2) return (AB2.val) } Gstart <- Gns(r=4,n=n,Sigma=S,scv=TRUE) Gstart <- matrix.sqrt(Gstart) if (optim.fun=="nlm") { result <- nlm(p=vech(Gstart), f=AB2, print.level=2*as.logical(verbose)) G4 <- result$estimate } else if (optim.fun=="optim") { result <- optim(vech(Gstart), AB2, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) G4 <- result$par } G4 <- invvech(G4)%*%invvech(G4) G4 <- S12 %*% G4 %*% S12 return(G4) } ############################################################################### ## Computes the smoothed cross validation function for 2 to 6 dim ## ## Parameters ## x - data values ## H - bandwidth matrix ## G - pilot bandwidth matrix ## ## Returns ## SCV(H) ############################################################################### scv.1d <- function(x, h, g, binned=TRUE, bin.par, inc=1, deriv.order=0) { r <- deriv.order if (!missing(x)) n <- length(x) if (!missing(bin.par)) n <- sum(bin.par$counts) scv1 <- kfe.1d(x=x, deriv.order=2*r, bin.par=bin.par, g=sqrt(2*h^2+2*g^2), binned=binned, inc=inc) scv2 <- kfe.1d(x=x, deriv.order=2*r, bin.par=bin.par, g=sqrt(h^2+2*g^2), binned=binned, inc=inc) scv3 <- kfe.1d(x=x, deriv.order=2*r, bin.par=bin.par, g=sqrt(2*g^2), binned=binned, inc=inc) bias2 <- (-1)^r*(scv1 - 2*scv2 + scv3) if (bias2 < 0) bias2 <- 0 scv <- (n*h)^(-1)*(4*pi)^(-1/2)*2^(-r)*OF(2*r) + bias2 return(scv) } scv.mat <- function(x, H, G, binned=FALSE, bin.par, bgridsize, verbose=FALSE, deriv.order=0) { d <- ncol(x) n <- nrow(x) r <- deriv.order vId <- vec(diag(d)) Hinv <- chol2inv(chol(H)) if (!binned) { scv1 <- Qr(x=x, deriv.order=2*r, Sigma=2*H+2*G, inc=1) scv2 <- Qr(x=x, deriv.order=2*r, Sigma=H+2*G, inc=1) scv3 <- Qr(x=x, deriv.order=2*r, Sigma=2*G, inc=1) bias2 <- (-1)^r*(scv1 - 2*scv2 + scv3) if (bias2 < 0) bias2 <- 0 } else { scv1 <- kfe(x=x, G=2*H + 2*G, deriv.order=2*r, inc=1, binned=binned, bin.par=bin.par, bgridsize=bgridsize, verbose=verbose, add.index=FALSE) scv2 <- kfe(x=x, G=H + 2*G, deriv.order=2*r, inc=1, binned=binned, bin.par=bin.par, bgridsize=bgridsize, verbose=verbose, add.index=FALSE) scv3 <- kfe(x=x, G=2*G, deriv.order=2*r, inc=1, binned=binned, bin.par=bin.par, bgridsize=bgridsize, verbose=verbose, add.index=FALSE) bias2 <- drop((-1)^r*Kpow(vId,r) %*% (scv1 - 2*scv2 + scv3)) if (bias2 < 0) bias2 <- 0 } scvmat <- 1/(det(H)^(1/2)*n)*nur(r=r, A=Hinv, mu=rep(0,d), Sigma=diag(d), type="direct")*2^(-d-r)*pi^(-d/2) + bias2 return (scvmat) } ############################################################################### ## Find the bandwidth that minimises the SCV for 1 to 6 dim ## ## Parameters ## x - data values ## pre - "scale" - pre-scaled data ## - "sphere"- pre-sphered data ## Hstart - initial bandwidth matrix ## ## Returns ## H_SCV ############################################################################### hscv <- function(x, nstage=2, binned=TRUE, bgridsize, plot=FALSE) { sigma <- sd(x) n <- length(x) d <- 1 hnorm <- sqrt((4/(n*(d + 2)))^(2/(d + 4)) * var(x)) if (missing(binned)) binned <- default.bflag(d=d,n=n) if (missing(bgridsize)) bgridsize <- default.bgridsize(d) hmin <- 0.1*hnorm hmax <- 2*hnorm bin.par <- binning(x=x, bgridsize=bgridsize, h=hnorm) if (nstage==1) { psihat6 <- psins.1d(r=6, sigma=sigma) psihat10 <- psins.1d(r=10, sigma=sigma) } else if (nstage==2) { g1 <- (2/(7*n))^(1/9)*2^(1/2)*sigma g2 <- (2/(11*n))^(1/13)*2^(1/2)*sigma psihat6 <- kfe.1d(x=x, bin.par=bin.par, binned=binned, deriv.order=6, g=g1, inc=1) psihat10 <- kfe.1d(x=x, bin.par=bin.par, binned=binned, deriv.order=10, g=g2, inc=1) } g3 <- (-6/((2*pi)^(1/2)*psihat6*n))^(1/7) g4 <- (-210/((2*pi)^(1/2)*psihat10*n))^(1/11) psihat4 <- kfe.1d(x=x, bin.par=bin.par, binned=binned, deriv.order=4, g=g3, inc=1) psihat8 <- kfe.1d(x=x, bin.par=bin.par, binned=binned, deriv.order=8, g=g4, inc=1) C <- (441/(64*pi))^(1/18) * (4*pi)^(-1/5) * psihat4^(-2/5) * psihat8^(-1/9) scv.1d.temp <- function(h) { return(scv.1d(x=x, bin.par=bin.par, h=h, g=C*n^(-23/45)*h^(-2), binned=binned, inc=1)) } if (plot) { hseq <- seq(hmin,hmax, length=400) hscv.seq <- rep(0, length=length(hseq)) for (i in 1:length(hseq)) hscv.seq[i] <- scv.1d.temp(hseq[i]) plot(hseq, hscv.seq, type="l", xlab="h", ylab="SCV(h)") } opt <- optimise(f=scv.1d.temp, interval=c(hmin, hmax))$minimum if (n >= 1e5) warning("hscv is not always stable for large samples") return(opt) } Hscv <- function(x, nstage=2, pre="sphere", pilot, Hstart, binned, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") { n <- nrow(x) d <- ncol(x) r <- deriv.order if (missing(binned)) binned <- default.bflag(d=d,n=n) if (d > 4) binned <- FALSE if (missing(bgridsize)) bgridsize <- default.bgridsize(d) if(!is.matrix(x)) x <- as.matrix(x) if (missing(pilot)) { if (d==2 & r==0) pilot <- "samse" else pilot <- "dscalar" } pilot1 <- match.arg(pilot, c("amse", "samse", "unconstr", "dunconstr", "dscalar")) pre1 <- match.arg(pre, c("scale", "sphere")) optim.fun <- match.arg(optim.fun, c("nlm", "optim")) if (pilot1=="amse" & (d>2 | r>0)) stop("amse pilot selectors not defined for d>2 and/or r>0") if ((pilot1=="samse" | pilot1=="unconstr") & r>0) stop("dscalar or dunconstr pilot selectors are better for deriv.order>0") if (pre1=="scale") { x.star <- pre.scale(x) S12 <- diag(sqrt(diag(var(x)))) Sinv12 <- chol2inv(chol(S12)) } else if (pre1=="sphere") { x.star <- pre.sphere(x) S12 <- matrix.sqrt(var(x)) Sinv12 <- chol2inv(chol(S12)) } RK <- (4*pi)^(-d/2) if (binned) { if (pilot1=="unconstr" | pilot1=="dunconstr") H.max <- (((d+8)^((d+6)/2)*pi^(d/2)*RK)/(16*(d+2)*n*gamma(d/2+4)))^(2/(d+4))* var(x) else H.max <- (((d+8)^((d+6)/2)*pi^(d/2)*RK)/(16*(d+2)*n*gamma(d/2+4)))^(2/(d+4))* var(x.star) if (default.bflag(d=d, n=n)) bin.par <- binning(x=x.star, bgridsize=bgridsize, H=matrix.sqrt(H.max)) } if (pilot1=="unconstr") { ## Gu pilot matrix is on data scale Gu <- Gunconstr.scv(x=x, binned=binned, bgridsize=bgridsize, verbose=verbose, nstage=nstage-1, optim.fun=optim.fun) if (missing(Hstart)) Hstart <- Hns(x=x, deriv.order=r) } else if (pilot1=="dunconstr") { ## Gu pilot matrix is on data scale Gu <- Gdunconstr(x=x, d=d, r=r, n=n, nstage=nstage, verbose=verbose, binned=binned, scv=TRUE, optim.fun=optim.fun) if (missing(Hstart)) Hstart <- Hns(x=x, deriv.order=r) } else if (pilot1=="dscalar") { ## Gs is on pre-transformed data scale g2r4 <- gdscalar(x=x.star, d=d, r=r, n=n, nstage=nstage, verbose=verbose, scv=TRUE, binned=binned) Gs <- g2r4^2*diag(d) if (missing(Hstart)) Hstart <- Hns(x=x.star, deriv.order=r) } else { ## Gs is on transformed data scale Hamise <- Hpi(x=x.star, nstage=1, deriv.order=r, pilot=pilot, pre="sphere", binned=TRUE, bgridsize=bgridsize, verbose=verbose, optim.fun=optim.fun) if (any(is.na(Hamise))) { warning("Pilot bandwidth matrix is NA - replaced with maximally smoothed") Hamise <- (((d+8)^((d+6)/2)*pi^(d/2)*RK)/(16*(d+2)*n*gamma(d/2+4)))^(2/(d+4))* var(x.star) } gs <- gamse.scv(x.star=x.star, d=d, Sigma.star=var(x.star), Hamise=Hamise, n=n, binned=binned, bgridsize=bgridsize, verbose=verbose, nstage=nstage-1) Gs <- gs^2*diag(d) ## use normal reference bandwidth as initial condition if (missing(Hstart)) Hstart <- Hns(x=x.star, deriv.order=r) else Hstart <- Sinv12 %*% Hstart %*% Sinv12 } ## SCV is estimate of AMISE scv.mat.temp <- function(vechH) { H <- invvech(vechH) %*% invvech(vechH) if (pilot1=="samse" | pilot1=="amse" | pilot1=="dscalar") { Gpilot <- Gs; xx <- x.star } else if (pilot1=="unconstr" | pilot1=="dunconstr") { Gpilot <- Gu; xx <- x } if (default.bflag(d=d, n=n)) scvm <- scv.mat(x=xx, H=H, G=Gpilot, binned=binned, bin.par=bin.par, verbose=FALSE, deriv.order=r) else scvm <- scv.mat(x=xx, H=H, G=Gpilot, binned=binned, verbose=FALSE, deriv.order=r) return(scvm) } Hstart <- matrix.sqrt(Hstart) if (optim.fun=="nlm") { result <- nlm(p=vech(Hstart), f=scv.mat.temp, print.level=2*as.numeric(verbose)) H <- invvech(result$estimate) %*% invvech(result$estimate) amise.star <- result$minimum } else if (optim.fun=="optim") { result <- optim(vech(Hstart), scv.mat.temp, method="BFGS", control=list(trace=as.numeric(verbose), REPORT=1)) H <- invvech(result$par) %*% invvech(result$par) amise.star <- result$value } if (!(pilot1 %in% c("dunconstr","unconstr"))) H <- S12 %*% H %*% S12 ## back-transform if (!amise) return(H) else return(list(H = H, SCV.star=amise.star)) } Hscv.diag <- function(x, nstage=2, pre="scale", pilot, Hstart, binned, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") { n <- nrow(x) d <- ncol(x) r <- deriv.order RK <- (4*pi)^(-d/2) if (missing(binned)) binned <- default.bflag(d=d,n=n) if (d > 4) binned <- FALSE if (missing(bgridsize)) bgridsize <- default.bgridsize(d) if(!is.matrix(x)) x <- as.matrix(x) if (missing(pilot)) { if (d==2 & r==0) pilot <- "samse" else pilot <- "dscalar" } pilot1 <- match.arg(pilot, c("amse", "samse", "unconstr", "dunconstr", "dscalar")) pre1 <- match.arg(pre, c("scale", "sphere")) optim.fun <- match.arg(optim.fun, c("nlm", "optim")) if (pilot1=="amse" & (d>2 | r>0)) stop("samse pilot selectors are better for higher dimensions and/or deriv.order>0") if (pilot1=="samse" & r>0) stop("dscalar pilot selectors are better for deriv.order>0") if (pilot1=="unconstr" | pilot1=="dunconstr") stop("Unconstrained pilot selectors are not suitable for Hscv.diag") if (pre1=="sphere") stop("Using pre-sphering doesn't give a diagonal bandwidth matrix") if (pre1=="scale") { x.star <- pre.scale(x) S12 <- diag(sqrt(diag(var(x)))) Sinv12 <- chol2inv(chol(S12)) } else if (pre1=="sphere") { x.star <- pre.sphere(x) S12 <- matrix.sqrt(var(x)) Sinv12 <- chol2inv(chol(S12)) } if (binned) { H.max <- (((d+8)^((d+6)/2)*pi^(d/2)*RK)/(16*(d+2)*n*gamma(d/2+4)))^(2/(d+4))* var(x.star) bin.par.star <- binning(x=x.star, bgridsize=bgridsize, H=H.max) } if (pilot1=="dscalar") { ## Gs is on pre-transformed data scale g2r4 <- gdscalar(x=x.star, r=r, n=n, d=d, verbose=verbose, nstage=nstage, scv=TRUE, binned=binned) Gs <- g2r4^2*diag(d) if (missing(Hstart)) Hstart <- Hns(x=x.star, deriv.order=r) } else { ## Gs is on transformed data scale Hamise <- Hpi(x=x.star, nstage=1, pilot=pilot, pre="sphere", binned=binned, bgridsize=bgridsize, verbose=verbose, optim.fun=optim.fun) if (any(is.na(Hamise))) { warning("Pilot bandwidth matrix is NA - replaced with maximally smoothed") Hamise <- (((d+8)^((d+6)/2)*pi^(d/2)*RK)/(16*(d+2)*n*gamma(d/2+4)))^(2/(d+4))* var(x.star) } gs <- gamse.scv(x.star=x.star, d=d, Sigma.star=var(x.star), Hamise=Hamise, n=n, binned=binned, bgridsize=bgridsize, verbose=verbose, nstage=nstage-1) Gs <- gs^2*diag(d) ## use normal reference bandwidth as initial condition if (missing(Hstart)) Hstart <- Hns(x=x.star, deriv.order=r) else Hstart <- Sinv12 %*% Hstart %*% Sinv12 } scv.mat.temp <- function(diagH) { H <- diag(diagH) %*% diag(diagH) if (default.bflag(d=d, n=n)) scvm <- scv.mat(x.star, H, Gs, binned=binned, verbose=FALSE, bin.par=bin.par.star, deriv.order=r) else scvm <- scv.mat(x.star, H, Gs, binned=binned, verbose=FALSE, deriv.order=r) return(scvm) } Hstart <- matrix.sqrt(Hstart) if (optim.fun=="nlm") { result <- nlm(p=diag(Hstart), f=scv.mat.temp, print.level=2*as.numeric(verbose)) H <- diag(result$estimate) %*% diag(result$estimate) amise.star <- result$minimum } else if (optim.fun=="optim") { result <- optim(diag(Hstart), scv.mat.temp, method="Nelder-Mead", control=list(trace=as.numeric(verbose), REPORT=1)) H <- diag(result$par) %*% diag(result$par) amise.star <- result$value } ## back-transform H <- S12 %*% H %*% S12 if (!amise) return(H) else return(list(H = H, SCV.star=amise.star)) } ############################################################################## ## Normal scale selector H_ns for kernel density derivate estimators ############################################################################## Hns <- function(x, deriv.order=0) { if (is.vector(x)) { n <- 1; d <- length(x) } else { n <- nrow(x); d <- ncol(x) } r <- deriv.order H <- (4/(n*(d+2*r+2)))^(2/(d+2*r+4)) * var(x) return(H) } Hns.diag <- function(x) { if (is.vector(x)) { n <- 1; d <- length(x) } else { n <- nrow(x); d <- ncol(x) } S <- var(x) Delta <- chol2inv(chol(diag(diag(S))))%*%S svD <- svd(Delta) Deltainv <- svD$v %*% diag(1/svD$d) %*% t(svD$u) H <- ((4*d*det(Delta)^(1/2))/(2*tr(Deltainv%*%Deltainv) + (tr(Deltainv))^2))^(2/(d+4))*diag(diag(S))*n^(-2/(d+4)) return(H) } hns <- function(x, deriv.order=0) { n <- length(x) d <- 1 r <- deriv.order h <- (4/(n*(d+2*r+2)))^(1/(d+2*r+4))*sd(x) return(h) } ####################################################################### ## Normal scale G_ns for kernel functional estimators ## scv = flag for SCV bias annihiliation constant ####################################################################### Gns <- function(r,n,Sigma,scv=FALSE) { d <- ncol(Sigma) const <- ifelse(scv,1,2) G <- (2/((n*(d+r))))^(2/(d+r+2))*const*Sigma return(G) } Gns.search <- function(G, f, n=10) { Gstart.vec <- numeric(n) for (i in 1:length(Gstart.vec)) { Gstart <- matrix.sqrt(i/length(Gstart.vec)*G) Gstart.vec[i] <- f(vech(Gstart)) } i <- which.min(Gstart.vec)/length(Gstart.vec) Gstart <- i*G return(Gstart) } ############################################################################## ## Normal mixture selector ############################################################################## Hnm <- function(x, deriv.order=0, G=1:9, subset.ind, mise.flag=FALSE, verbose=FALSE, ...) { if (!requireNamespace("mclust", quietly=TRUE)) stop("Install the mclust package as it is required.", call.=FALSE) modelNames <- "VVV" if (!missing(subset.ind)) nmixt.fit <- mclust::Mclust(x[subset.ind,], G=G, verbose=verbose, modelNames=modelNames, ...) else nmixt.fit <- mclust::Mclust(x, G=G, verbose=verbose, modelNames=modelNames, ...) if (is.vector(x)) { d <- length(x); n <- 1 } else { d <- ncol(x); n <- nrow(x) } mus <- t(nmixt.fit$parameters$mean) Sigmas <- matrix(nmixt.fit$parameters$variance$sigma, byrow=TRUE, ncol=d) props <- nmixt.fit$parameters$pro if (mise.flag) H.nm <- Hmise.mixt(samp=n, mus=mus, Sigmas=Sigmas, props=props, deriv.order=deriv.order) else H.nm <- Hamise.mixt(samp=n, mus=mus, Sigmas=Sigmas, props=props, deriv.order=deriv.order) return(H.nm) } Hnm.diag <- function(x, deriv.order=0, G=1:9, subset.ind, mise.flag=FALSE, verbose=FALSE, ...) { if (!requireNamespace("mclust", quietly=TRUE)) stop("Install the mclust package as it is required.", call.=FALSE) modelNames <- "VVI" if (!missing(subset.ind)) nmixt.fit <- mclust::Mclust(x[subset.ind,], G=G, verbose=verbose, modelNames=modelNames, ...) else nmixt.fit <- mclust::Mclust(x, G=G, verbose=verbose, modelNames=modelNames, ...) if (is.vector(x)) { d <- length(x); n <- 1 } else { d <- ncol(x); n <- nrow(x) } mus <- t(nmixt.fit$parameters$mean) Sigmas <- matrix(nmixt.fit$parameters$variance$sigma, byrow=TRUE, ncol=d) props <- nmixt.fit$parameters$pro if (mise.flag) H.nm <- Hmise.mixt.diag(samp=n, mus=mus, Sigmas=Sigmas, props=props, deriv.order=deriv.order) else H.nm <- Hamise.mixt.diag(samp=n, mus=mus, Sigmas=Sigmas, props=props, deriv.order=deriv.order) return(H.nm) } hnm <- function(x, deriv.order=0, G=1:9, subset.ind, mise.flag=FALSE, verbose=FALSE, ...) { if (!missing(subset.ind)) nmixt.fit <- mclust::Mclust(x[subset.ind], G=G, verbose=verbose, ...) else nmixt.fit <- mclust::Mclust(x, G=G, verbose=verbose, ...) mus <- nmixt.fit$parameters$mean sigmas <- sqrt(nmixt.fit$parameters$variance$sigma) props <- nmixt.fit$parameters$pro n <- length(x) if (mise.flag) h.nm <- hmise.mixt(samp=n, mus=mus, sigmas=sigmas, props=props, deriv.order=deriv.order) else h.nm <- hamise.mixt(samp=n, mus=mus, sigmas=sigmas, props=props, deriv.order=deriv.order) return(h.nm) } ks/R/kms.R0000644000176200001440000002645114547756237012041 0ustar liggesusers############################################################################### ## Kernel mean shift ############################################################################### kms <- function(x, y, H, max.iter=400, tol.iter, tol.clust, min.clust.size, merge=TRUE, keep.path=FALSE, verbose=FALSE) { n <- nrow(x) d <- ncol(x) if (missing(tol.iter)) tol.iter <- 1e-3*min(apply(x, 2, IQR)) ## mean(apply(apply(x, 2, range), 2, diff)) if (missing(tol.clust)) tol.clust <- 1e-2*max(apply(x, 2, IQR)) ## mean(apply(apply(x, 2, range), 2, diff)) if (missing(y)) y <- x if (missing(min.clust.size)) min.clust.size <- round(1e-2*nrow(y),0) if (missing(H)) H <- Hpi(x, deriv.order=1, binned=default.bflag(d=d, n=n), nstage=2-(d>2)) Hinv <- chol2inv(chol(H)) if (is.vector(y)) y <- matrix(y, nrow=1) ## mean shift iterations n.seq <- block.indices(n, nrow(y), d=d, r=0, diff=FALSE, block.limit=3e6) if (verbose) pb <- txtProgressBar() ms <- list() i <- 1 if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) ms <- kms.base(x=x, y=y[n.seq[i]:(n.seq[i+1]-1),], H=H, tol.iter=tol.iter, tol.clust=tol.clust, Hinv=Hinv, verbose=verbose, max.iter=max.iter) if (length(n.seq)>2) { for (i in 2:(length(n.seq)-1)) { if (verbose) setTxtProgressBar(pb, i/(length(n.seq)-1)) ms.temp <- kms.base(x=x, y=y[n.seq[i]:(n.seq[i+1]-1),], H=H, tol.iter=tol.iter, tol.clust=tol.clust, Hinv=Hinv, verbose=verbose, max.iter=max.iter) ms$y <- rbind(ms$y, ms.temp$y) ms$end.points <- rbind(ms$end.points, ms.temp$end.points) ms$label <- c(ms$label, ms.temp$label + max(ms$label)) ms$mode <- rbind(ms$mode, ms.temp$mode) ms$nclust <- ms$nclust + ms.temp$nclust ms$nclust.table <- table(ms$label) ms$path <- c(ms$path, ms.temp$path) ## merge clusters which are closer than tol.clust distance ms <- ms.merge.dist(ms=ms, tol=tol.clust, verbose=FALSE) } } if (verbose) close(pb) path.temp <- ms$path ms$path <- NULL ms$tol.iter <- tol.iter ms$tol.clust <- tol.clust ms$min.clust.size <- min.clust.size ms$names <- parse.name(x) if (keep.path) ms$path <- path.temp ## merge clusters which are smaller than min.clust.size if (merge) ms <- ms.merge.num(ms, min.clust.size=min.clust.size, verbose=verbose) return(ms) } kms.base <- function(x, H, Hinv, y, max.iter, tol.iter, tol.clust, verbose=FALSE) { ## mean shift iterations ## original implementation J.E. Chacon (2013) ## modifications T.D. (2014) if (!is.matrix(x)) x <- as.matrix(x) if (!is.matrix(y)) y <- as.matrix(y) if (missing(Hinv)) Hinv <- chol2inv(chol(H)) nx <- nrow(x) ny <- nrow(y) d <- ncol(y) y.path <- split(y, row(y), drop=FALSE) names(y.path) <- NULL xHinv <- x %*% Hinv xHinvx <- rowSums(xHinv*x) y.update <- y i <- 1 eps <- max(sqrt(rowSums(y.update^2))) disp.ind <- head(sample(1:nrow(y)), n=min(100,nrow(y))) while (eps > tol.iter & i< max.iter) { y.curr <- y.update yHinvy <- t(rowSums(y.curr%*%Hinv *y.curr)) Mah <- apply(yHinvy, 2, "+", xHinvx) - 2*xHinv %*% t(y.curr) w <- exp(-Mah/2) denom <- colSums(w) num <- t(w)%*%x denom[denom<=1e-3*tol.iter] <- 1e-3*tol.iter mean.shift.H <- num/denom y.update <- mean.shift.H y.update.list <- split(y.update, row(y.update), drop=FALSE) y.path <- mapply(rbind, y.path, y.update.list, SIMPLIFY=FALSE) eps <- max(sqrt(rowSums((y.curr-y.update)^2))) if (verbose>1) { y.range <- apply(y, 2, range) if (d==2) plot(y.update[disp.ind,], col=1, xlim=y.range[,1], ylim=y.range[,2], xlab="x", ylab="y") else pairs(y.update[disp.ind,], col=1) } i <- i+1 } ms.endpt <- t(sapply(y.path, tail, n=1, SIMPLIFY=FALSE)) ## extract cluster centres mode.tree <- hclust(dist(ms.endpt)) clust.label <- cutree(mode.tree, h=tol.clust) nclust <- length(unique(clust.label)) mode.val <- by(ms.endpt, INDICES=clust.label, FUN=colMeans) mode.val <- t(sapply(mode.val, FUN=identity)) colnames(mode.val) <- colnames(x) rownames(mode.val) <- NULL nclust.table <- table(clust.label, dnn="") ms <- list(x=x, y=y, end.points=ms.endpt, H=H, label=clust.label, nclust=nclust, nclust.table=nclust.table, mode=mode.val, path=y.path) class(ms) <- "kms" return(ms) } ## merge classes in 'label' into a single class ## label is a list of vector of class labels ## i.e. all classes in label[[j]] merged into new class j ms.merge.label <- function(ms, label, verbose=FALSE) { ms.merge <- ms for (i in 1:length(label)) { labeli <- label[[i]] if (length(labeli)>1) { merge.label <- min(labeli) ms.merge$label[ms$label %in% labeli] <- merge.label mode.label <- ms$mode[labeli[round(length(labeli)/2,0)],] for (j in labeli) ms.merge$mode[j,] <- mode.label } } ms.merge$mode <- unique(ms.merge$mode) ms.merge$nclust <- nrow(ms.merge$mode) ms.merge$label <- as.factor(ms.merge$label) levels(ms.merge$label) <- 1:ms.merge$nclust ms.merge$label <- as.numeric(ms.merge$label) ms.merge$nclust.table <- table(ms.merge$label) if (verbose) cat("Current clusters:", ms.merge$nclust.table, "\n") return(ms.merge) } ## merge mean shift clusters based on distance threshold ms.merge.dist <- function(ms, tol, verbose) { if (missing(tol)) tol <- 1e-1*min(apply(ms$x, 2, IQR)) mode.tree <- hclust(dist(ms$mode)) merge.label <- cutree(mode.tree, h=tol) merge.label <- split(1:ms$nclust, merge.label, drop=FALSE) ## create list where each element is a vector of cluster labels ## to be merged into a single cluster ms.temp <- ms.merge.label(ms=ms, label=merge.label, verbose=verbose) return(ms.temp) } ## merge mean shift clusters based on min cluster size ms.merge.num <- function(ms, min.clust.size, verbose=FALSE) { if (missing(min.clust.size)) min.clust.size <- round(1e-2*nrow(ms$y),0) min.clust.size <- round(min.clust.size, 0) if (any(ms$nclust.table<=min.clust.size)) { if (verbose) cat("Min cluster size merging begins. Min size = ", min.clust.size, "\n") ms.temp <- ms while(any(ms.temp$nclust.table<=min.clust.size) & ms.temp$nclust>1) { nclust.table <- table(ms.temp$label) small.clust.ind <- which.min(nclust.table) if (nclust.table[small.clust.ind] <= min.clust.size) { nearest.clust.ind <- FNN::get.knnx(ms.temp$mode, ms.temp$mode, k=2)$nn.index[small.clust.ind,2] merge.label <- ms$label merge.label[merge.label==small.clust.ind] <- nearest.clust.ind merge.label <- split(ms$label, merge.label, drop=FALSE) merge.label <- lapply(merge.label, unique) ms.temp <- ms.merge.label(ms=ms.temp, label=merge.label, verbose=verbose) } } ms <- ms.temp ms$min.clust.size <- min.clust.size if (verbose) cat("Min cluster size merging ends.\n\n") } if (verbose) { cat("Final clusters:\n"); summary(ms) } return(ms) } ###################################################################### ## Cluster partition for 2D kernel mean shift ##################################################################### kms.part <- function(x, H, xmin, xmax, gridsize, verbose=FALSE, ...) { if (missing(H)) H <- Hpi(x, deriv.order=1, binned=TRUE) tol <- 5 tol.H <- tol * diag(H) if (missing(xmin)) xmin <- apply(x, 2, min) - tol.H if (missing(xmax)) xmax <- apply(x, 2, max) + tol.H if (missing(gridsize)) gridsize <- default.gridsize(2) xx <- seq(xmin[1], xmax[1], length = gridsize[1]) yy <- seq(xmin[2], xmax[2], length = gridsize[2]) xy <- expand.grid(xx, yy) xy.kms <- kms(x=x, y=xy, H=H, verbose=verbose, ...) xy.lab <- array(xy.kms$label, dim=gridsize) fhat <- kde(x=x, binned=TRUE, xmin=xmin, xmax=xmax, bgridsize=gridsize) fhat$estimate <- xy.lab fhat <- c(fhat, xy.kms[c("end.points", "label", "mode", "nclust", "nclust.table", "min.clust.size", "tol.iter", "tol.clust")]) class(fhat) <- "kde.part" return(fhat) } plot.kde.part <- function(x, display="filled.contour", col, col.fun, alpha=1, add=FALSE, ...) { clev <- sort(unique(as.vector(x$estimate))) if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="Dark2", alpha=alpha) } if (missing(col)) col <- col.fun(length(clev)) for (i in 1:length(clev)) { xtemp <- x xtemp$estimate <- x$estimate==clev[i] plot.kde(xtemp, display=display, col=c("transparent", col[i]), add=add | i>1, abs.cont=0.5, drawlabels=FALSE, alpha=alpha, ...) } } ############################################################################# ## S3 methods for KMS objects ############################################################################# ## summary method summary.kms <- function(object, ...) { cat("Number of clusters =", object$nclust, "\n") cat("Cluster label table =", object$nclust.table, "\n") cat("Cluster modes =\n") print(as.data.frame(object$mode), ...) } ## plot method plot.kms <- function(x, display="splom", col, col.fun, alpha=1, xlab, ylab, zlab, theta=-30, phi=40, add=FALSE, ...) { disp <- match.arg(display, c("splom", "plot3D", "rgl")) if (is.vector(x$H)) d <- 1 else d <- ncol(x$H) if (missing(col.fun)) col.fun <- function(n) { hcl.colors(n, palette="Set2", alpha=alpha) } if (missing(col)) col <- col.fun(length(unique(x$label))) col <- transparency.col(col, alpha=alpha) if (d==1) stop("kms plot not yet implemented") else if (d==2) { if (missing(xlab)) xlab <- x$names[1] if (missing(ylab)) ylab <- x$names[2] if (!add) plot(x$x, col=col[x$label], xlab=xlab, ylab=ylab, ...) else points(x$x, col=col[x$label], ...) } else if (d==3 & disp %in% c("plot3D", "rgl")) { if (missing(xlab)) xlab <- x$names[1] if (missing(ylab)) ylab <- x$names[2] if (missing(zlab)) zlab <- x$names[3] if (disp=="plot3D") { if (!requireNamespace("plot3D", quietly=TRUE)) stop("Install the plot3D package as it is required.", call.=FALSE) if (!add) plot3D::points3D(x$x[,1], x$x[,2], x$x[,3], col=1, cex=0, add=add, theta=theta, phi=phi, d=4, colkey=FALSE, xlab=xlab, ylab=ylab, zlab=zlab, ticktype="detailed", bty="f", ...) for (i in 1:length(col)) plot3D::points3D(x$x[x$label==i,1], x$x[x$label==i,2], x$x[x$label==i,3], col=col[i], add=TRUE, ...) } else if (disp=="rgl") { ## suggestions from Viktor Petukhov 08/03/2018 if (!requireNamespace("rgl", quietly=TRUE)) stop("Install the rgl package as it is required.", call.=FALSE) if (!add) rgl::plot3d(x$x, col=col[x$label], alpha=alpha, ...) else rgl::points3d(x$x, col=col[x$label], alpha=alpha, ...) } } else if (d>=3) { pairs(x$x, col=col[x$label], ...) } } ks/vignettes/0000755000176200001440000000000014673274557012723 5ustar liggesusersks/vignettes/kde.Rnw0000644000176200001440000001725414336674020014150 0ustar liggesusers\documentclass[a4paper,11pt]{article} \usepackage{amsmath,amssymb,amsopn,natbib} \usepackage[left=2.5cm,top=2.5cm,right=2.5cm,bottom=2.5cm]{geometry} \renewcommand{\today}{\begingroup \number \day\space \ifcase \month \or January\or February\or March\or April\or May\or June\or July\or August\or September\or October\or November\or December\fi \space \number \year \endgroup} \renewcommand{\vec}[1]{\boldsymbol{#1}} \newcommand{\mat}[1]{\mathbf{#1}} \def\bH{\mat{H}} \def\vecx{\vec{x}} \def\vecX{\vec{X}} \DeclareMathOperator{\E}{\boldsymbol{\mathbb{E}}} \let\code=\texttt \let\proglang=\texttt \let\pkg=\texttt %\VignetteIndexEntry{kde} %\SweaveOpts{eps=FALSE} \title{ks: Kernel density estimation for bivariate data} \author{Tarn Duong} \begin{document} \maketitle \noindent Kernel density estimation is a popular tool for visualising the distribution of data. See \citet*{simonoff1996}, for example, for an overview. When multivariate kernel density estimation is considered it is usually in the constrained context with diagonal bandwidth matrices, e.g. in the \proglang{R} packages \pkg{sm} \citep*{sm} and \pkg{KernSmooth} \citep*{KernSmooth}. We introduce a new \proglang{R} package \pkg{ks} which implements diagonal and unconstrained data-driven bandwidth matrices for kernel density estimation, which can also be used for multivariate kernel discriminant analysis. The \pkg{ks} package implements selectors for 1- to 6-dimensional data. This vignette contains only a brief introduction to using \pkg{ks} for kernel density estimation for 2-dimensional data. See \citet*{duong2007c} for a more detailed account. For a bivariate random sample $\vecX_1, \vecX_2, \ldots, \vecX_n$ drawn from a density $f$, the kernel density estimate is defined by $$ \hat{f} (\vecx; \bH) = n^{-1}\sum_{i=1}^n K_{\bH} ( \vecx - \vec{X}_i) $$ where $\vecx = (x_1, x_2)^T$ and $\vec{X}_i = (X_{i1}, X_{i2})^T, i = 1, 2, \ldots, n$. Here $K(\vecx)$ is the kernel which is a symmetric probability density function, $\bH$ is the bandwidth matrix which is symmetric and positive-definite, and $K_{\bH}(\vecx) = |\bH|^{-1/2} K( \bH^{-1/2} \vecx)$. The choice of $K$ is not crucial: we take $K(\vecx) = (2\pi)^{-1} \exp(-\tfrac{1}{2} \vecx^T \vecx)$ the standard normal throughout. In contrast, the choice of $\bH$ is crucial in determining the performance of $\hat f$. The most common parameterisations of the bandwidth matrix are the diagonal and the general or unconstrained which has no restrictions on $\bH$ provided that $\bH$ remains positive definite and symmetric, that is $$ \bH = \begin{bmatrix}h_1^2 & 0 \\0 & h_2^2 \end{bmatrix} \ \mathrm{or} \ \bH = \begin{bmatrix}h_1^2 & h_{12} \\ h_{12} & h_2^2 \end{bmatrix}. $$ This latter parameterisation allows kernels to have an arbitrary orientation whereas the former only allows kernels which are oriented to the co-ordinate axes. For our target density, we use the `dumbbell' density, given by the normal mixture $$ \frac{4}{11} N \bigg( \begin{bmatrix}-2 \\ 2\end{bmatrix}, \begin{bmatrix}1 & 0 \\ 0 & 1 \end{bmatrix} \bigg)+ \frac{3}{11} N \bigg( \begin{bmatrix}0 \\ 0\end{bmatrix}, \begin{bmatrix}0.8 & -0.72 \\ -0.72 & 0.8\end{bmatrix} \bigg)+ \frac{4}{11} N \bigg( \begin{bmatrix}2 \\ -2\end{bmatrix}, \begin{bmatrix}1 & 0 \\ 0 & 1 \end{bmatrix} \bigg), $$ displayed on the left in Figure \ref{fig:dens-db}. This density is unimodal. On the right is a sample of 200 data points. <>= library(ks) set.seed(8192) samp <- 200 mus <- rbind(c(-2,2), c(0,0), c(2,-2)) Sigmas <- rbind(diag(2), matrix(c(0.8, -0.72, -0.72, 0.8), nrow=2), diag(2)) cwt <- 3/11 props <- c((1-cwt)/2, cwt, (1-cwt)/2) x <- rmvnorm.mixt(n=samp, mus=mus, Sigmas=Sigmas, props=props) @ \setkeys{Gin}{width=0.45\textwidth} \begin{figure}[!ht] \begin{center} <>= plotmixt(mus=mus, Sigmas=Sigmas, props=props, xlim=c(-4,4), ylim=c(-4,4)) @ <>= plot(x, xlim=c(-4,4), ylim=c(-4,4), xlab="x", ylab="y") @ \end{center} \caption{Target `dumbbell' density. (Left) contour plot. (Right) Scatter plot.} \label{fig:dens-db} \end{figure} We use \code{Hpi} for unconstrained plug-in selectors and \code{Hpi.diag} for diagonal plug-in selectors. <>= Hpi1 <- Hpi(x=x) Hpi2 <- Hpi.diag(x=x) @ To compute a kernel density estimate, the command is \code{kde}, which creates a \code{kde} class object <<>>= fhat.pi1 <- kde(x=x, H=Hpi1) fhat.pi2 <- kde(x=x, H=Hpi2) @ We use the \code{plot} method for \code{kde} objects to display these kernel density estimates. The default is a contour plot with the upper 25\%, 50\% and 75\% contours of the (sample) highest density regions. %, as %defined in \citet*{bowman1993} and \citet*{hyndman1996}. These regions are also plotted by the \pkg{sm} library. <>= plot(fhat.pi1) plot(fhat.pi2) @ The respective kernel density estimates are produced in Figure \ref{fig:pi}. The diagonal bandwidth matrix constrains the smoothing to be performed in directions parallel to the co-ordinate axes, so it is not able to apply accurate levels of smoothing to the obliquely oriented central portion. The result is a multimodal density estimate. The unconstrained bandwidth matrix correctly produces a unimodal density estimate. \begin{figure}[!ht] \centering <>= plot(fhat.pi1, main="Plug-in", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) @ <>= plot(fhat.pi2, main="Plug-in diagonal", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) @ \caption{Kernel density estimates with plug-in selectors} \label{fig:pi} \end{figure} The unconstrained SCV (Smoothed Cross Validation) selector is \code{Hscv} and its diagonal version is \code{Hscv.diag}. In Figure \ref{fig:cv}, the most reasonable density estimate is from the unconstrained SCV selector. <>= Hscv1 <- Hscv(x=x) Hscv2 <- Hscv.diag(x=x) @ \begin{figure}[!ht] \centering <>= fhat.cv1 <- kde(x=x, H=Hscv1) fhat.cv2 <- kde(x=x, H=Hscv2) @ <>= plot(fhat.cv1, main="SCV", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) @ <>= plot(fhat.cv2, main="SCV diagonal", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) @ \caption{Kernel density estimates with cross validation selectors} \label{fig:cv} \end{figure} The unconstrained bandwidth selectors will be better than their diagonal counterparts when the data have large mass oriented obliquely to the co-ordinate axes, like for the dumbbell data. The unconstrained plug-in and the SCV selectors can be viewed as generally recommended selectors. \bibliographystyle{apalike} \begin{thebibliography}{} \bibitem[Bowman and Azzalini, 2007]{sm} Bowman, A. W. and Azzalini, A. (2007). \newblock {\em sm: kernel smoothing methods: Bowman and Azzalini (1997)}. \newblock R package version 2.2. \bibitem[Duong, 2007]{duong2007c} Duong, T. (2007). \newblock ks: {K}ernel density estimation and kernel discriminant analysis for multivariate data in {R}. \newblock {\em Journal of Statistical Software}. \textbf{21 (7)}, URL \texttt{http://www.jstatsoft.org/v21/i07}. \bibitem[Simonoff, 1996]{simonoff1996} Simonoff, J. S. (1996). \newblock {\em Smoothing Methods in Statistics}. \newblock Springer-Verlag, New York. \bibitem[Wand, 2006]{KernSmooth} Wand, M. P. (2006). \newblock {\em KernSmooth: Functions for kernel smoothing for Wand \& Jones (1995)}. \newblock R package version 2.22-19. R port by Brian Ripley. \end{thebibliography} \end{document} ks/data/0000755000176200001440000000000014067333146011607 5ustar liggesusersks/data/platesf.RData0000644000176200001440000022352414067333065014172 0ustar liggesusersý7zXZi"Þ6!ÏXÌâT9ð])TW"änRÊŸ’Ù$n³Ñ¡|·OÀʈâ×™…ŠôHP† ‹*1ãvó, :k³Ôµ!âp†õ.c²%ìmá$Lj¹d˜:;Ÿš·;Œ˜µ1)žÞaz3¹«~³ÏC'ÉÕ÷}.Eg²Ü|¬BÎ{³ˆâ;cQ`ÓÛì3"÷3U|O×K8²èHázeÀò`ëJ•RÞ—fï&‘ib7*¯„Õ@-^v–Œ·Ûi°Ýœ/±rÃm"rT˜šÂíÁ˜ó9ù«@}è×Éèùp‡rŒ{›:±G«+Ëé¯FmØ“*ÑæG£‚d28·öÜè33ÑØîEÁù›ÊLïí„88ü&χñ¥RpjÛo9,¬2†¯ãä áÿ>f+(ß$OÍîR»*!=>M\ØÍp¾¤÷Wˆ»ÒJêaoÚ±p[Š­¹í<þãùžp Üß—–|6Ç› ¼c”Ü”ƒ "îÕÐ|ï½[qwS˜~Õ"þ€mµ®î½¶áT%§Ì+o¦î£þªÐn·Ò5Î)‰[_™Û¡åç ´?å «Úri$8-‡¶‹à´‡H^äm? I@ 7âö·K=7‚L‡¸‹²èrÙFëp¶M¢¬$»v#ù t³Ãp±¡Bð½ÜíÙÁÔ.Ù™ë yËJtU^É¥ú Â†‹ý…nƒ-^n ¸A²Í—v UÜ“[¬õ'Ç1êgœÿ JÁÔq¬B&ám~‰nz2$ì LxdKeo°µV – @Nh†7ˆ³“îE¹N? b5j¿*g<§+š0\ˆÛT„Añ‚^è¨Rc…ÂØÉFˆpÂn?8Ptý´†¸\\Ô‚§L|XN~j£mI:> zà‡­v[Þ;ñ*:oÑrc°ì1Õ® -ëû;¼êØ %Í/Rƒ¬hqf¥¼¬*¤hd´þ}â¡ï‹¹enx>:w;þ½ÑØ4M‰ 5€Ì>hÏÛ! 3ÃåÇqTg>>Z·òçÉ$ÓJf±}9ªù/[Ì$~˜W ÆoÏ:¦IÚSD÷³Úˈ>N'Ì«÷1¥v[|¨Še3LÏ‚"ŒüqVüéóâ\õ@Ãiµaó©Bâ,x{ÙÙ:£ß­ µÍšPÕDàB›“ùñîc´3ò]x¾1ã± Ùb<¯>>‚ý ä¼Ao?  ²Ú,)䃜IDÜl“bPô@(óâxÁMÅ®™%þâ\(Lx1,Ÿ99aJuØHú‡çY~ › ÷ ]|fˆ¨…¢ñht‡Á¶mÕØNZPœ¸$O­ÐÒuzÞ˯Hxü œæ/'aó※Øá @«’‹S\é+)R°:uu5dú4‰«ƒaVÞ„mõ†O¶ÿ]Pƒ °ÛÞ—yé‡Ä<rG¼$‡4»!ר>öö+Ræ_Ê“¸Ü¼ü¿í-l8O€o“½9d§¦k»=%ð+Ï‘–»xœy"á_­¡4À#à»Å«Ú”1¤8fyÂ6f€Íå’„µ0EÎ)1Ê ÃH4a1ªh §óÄ_SüX¿‰°jÆ8»U Ó›©Tk`2cg€e»7EU;N¾jôß×#¯Ùï6zÍÛ=â ÊmøPoÙÞ,@í ÐŒªbùl® .ÜÑÚ%ð¤y±ÜÙšÍã=5Xý¸çgŠ'U‡”4¬ITm¤ªTš’ k%x龆4£ÿþø…ÓáËÌ–o¶†&ýñc& ’ÛûCÆìÄÑÌElSéYÙ`®òg›H‡*ü§ðHº ±Û‚vjÝxÒ’=§P]Y‹àÔ Sƒÿño“üÀ«CÞŸý/K×!ð™ F =Æ<øN{~»ð;­ðûê®ÊC,J=u ¡×T#Ôw‘?žÎ*!瀱T6œ;Ò…<Ó““Q+ÆŸF._! ¬Uå'Èc ’Å 0÷×iÞÅC¤ŸâJy³J’VÇ[ÃYÛ« ˆ1FAcVc€úÏa99cf\P(Tþ#èíì¿Ó]{̵À×g¶YÇ£œ¥ ߈@EºïwgùDøDncÓPÖä:ny’Í~iSåbÐ|[mež#ÚIƒýÓxød ÖÏ‘3X§Dp¡`Æ”ÖÐ4ذ8€Ån±l#,;nÖ7ôŒcÀ"eÖýß§åIò7„4‡/«ÿ™Ií ‡«©Æ§§Û`HæÞµ]÷÷Hói$’‘˜7ÏlÉ‘Ch®`X×Ä ¨â@V«)ó¤ Ñ\Sã¦TÔp¥arÉ= œ‡º;¨›ƒÌy_°¹›.ÚcÙäÙeqÑ3Iƒ±>äìb3•4èy®gåCmͪûì=»ˆ+Î=Ù‹³ÀçËCÐC®¦ð@ñ&ä?kUœxÃ×u¸K¡H±GîþÂevŽÛœ¥Éáþ_øí|wO.PN1ü³cK> RªıÇÿLæè–®³FgÞâÖF!O©ÉxÛ§ÝÈ )¨ÄQWO[5ÿ½ …·Fx¤-crXŠaìÿšxI†šía0$£¨Bkh­€5ÄP|7Öû›„4]9|®½ÈäÖNyZßž0`ˆ­%6Wt}âíILÍÙV\DðJcø"+ ¼´KM­8YÄК\}ÊåÁ _иã„)ÖÁ²í=¨aœ¤K½>‚!ì!k7 Ò±C[icˆgÞ¯ç€:Iõ®»§Q?>+uh_‘æW.j0*óÂíãÿÐÏâ1n4dEœ¥U Î@5 M¯Í=  ¡É4âÕl‘Jjx§ \W~ýp¡ö™þUCñ¿w¯¾u€@ÎYCÇP#zl·Í U~½zš‘ØÍÄTYjwð bluIÛÙÕ5Å<Ùñ šãÜÂpkºJªɯ"çÁyÌÛS›|õ®a_oÜ•G¡¿óDp·êâ\ßÓLXM¥“î‰P®~ž­mrÒ9>KtõF?ú¡©ªËxºuÔ/ƒÃh¦©å˜ñöyqe¡ìŒÎÂ}ŽËÇG=<<ê;ÔKŠ‹PïƒÏ6„êd—8‹Â¥VSŽ-È}B wƧ/=½OÝ6=DìR㢠Ã.’*ÍE†\À÷¥ª&Ù ŽŽC&íð  ÉÊ-îùéše‚oå\üÇŸæ†Ïü3“o}ñ³g€攃(iEY“ôÉÂèÖ =ßdç8À¸C‹aå}$Ð ¾…{ lÁt¢¶‚&úMÙ³dFÓhŠH°‘Q¢ZgË0Þª…ÅÚà÷Ö¿µÝÅcÀ~‰nÎQ˜)ö[ŠBTë`#œNÕ+êÊ€Q‰A: ãÏMŠöøŠ´Û\Êc¡)Š­MI„ ä& ¹_`à{¾šêý‰ßòê³d~?oU]‡9K‰§Ýú¸‡w"R¯ï›U‘7´1(Ï”ñàÂL)™]e=S!tºEñ*`Eb¸§¥2„ÕZü^‡°™:ýüôW„"y  éÇý=)÷ãZóéR‹K°)ö‰?j“Çm(]FW¯BH\¿‚¡À$ÍÔ#¡w½£?&ÚpÜØåÃE(š>ÐÛõâ5Óæqø+™–êWÊ‚E…†nŸ¾<Ž c¤1?—Ÿ•eTÒéHŸa\j&(eë–íú‰ú$ƒ¤¤Ì‡ÞN„ŒgLž_ÂO&e˜s?Ù¡ðÚ³‚ÏŒù@ùReƒP™õûÜî#ĦÀ“å•ÏyJzÐ…¾¿D²=Á øjÅÝG|ÒÉuÓväi‹ª1 eÂ,õĪù<$ëÌq³¨Fï&ß>jãÀ®Íf u£í¡™öÈl¿©“–y‰iuT°Öz–åh¼6¿PÊeþ>L«ãäCxž&?뵎+€:âÀ9‘Jj ¹¡÷•™3 ±L߉a/ïÄ+šþ çKÍ <¡‘=`¬|@“ky‰Ÿ¨1ÿ‰O,Ï$ 9‹bůÊgVh®…Òa+N¶Z1ê$´µq£i¼çB“Ÿ¶X¥«®¸ÿkc)ò‹8Ȭlü51| ˜Ä?ØòhÞVJûu½gb †¤æ!áÂ2H§FJCvнÿÈÖp(ö€G¡)[l/˜NøEcA¸¥ið†™“ úŒTðƒÛ…5Ï'ï℞¹¢w<«,Ì ’\«®(=Þ¤P(ÈÐwãÛÑ|L[Õ“ `Å÷dcš=kEG;`'38–÷ ˶ڹtÔÇáë-àæ’blö%(‡PKÂÔƒÙlCr¡k´ÁW§B¢‰Šù.õ I¤ó/ã!;×ëk|°™G8Ú²fp#¥þ˜kŒr ÐdÑ*J {4BæŽR.Zlëˆ/“…^ÄÞ>¸ˆ;œ†)N ¢KêR'àÌÇÀÔ¬>ö¦2 6Ü ênv0½zÛ&ÈËx¿×h»¬Æ`z–Ïg¡QøÄa£ãñÙU¤öyù錯…èW®Â§î\¥fóÚÝuò–jý(pؑߜ)î5F‘P‚Ÿ’4HËîlÛ"0=9]BŸ, …ÎA>Q`¨âƳf£"(wóuÀ»$ɱ“ÃŒIjWX­œTð”vÙN„;ò§àòØ“ßÚÈùx=â³2ût³F´)çÈ×—åbÜÊL@,t›Aí×ù˜Á"z8ÌËhÁš:¯Ž•å´ã]J”QQ}¯[;ÔñBÀL½¸­R…Oªo+ zwÊÂÁøo“)…×0±€yÐ$}>,B¡ð…&}J^Ä ×Zl­3!•%4Þ/©µ†ƒy)–N×Èš«Š^ ¬o_ÏÀ¹ÕsëÑÝ*¨+¨ÊAï„ñV÷uˆ¿ÌR´k6Μg2XmÍ[µ|ÎXv.Ý …ÄaÆ×¼W†Ã3z}*ο Älû©SŠƒôÈù¬ì êØâæÉcöšs]©iŽÃIØÛ‹ù›#Œg·PG¸ÍtdŒ¿¬‘IRäïî° \–N'}soxˆ4¨ø÷¦¢‚_QCâ´³ Qoq(Sd¹P£^ü/ 6±ž0_Ê%ùC>9<ÉÅÂŒ«á¨íu2ölÝ“¸‘Oy< d+To¶ ^#Îü0ìùåxE/õŒ×~o¼@„VÅq³Cê8¡þ Ìòº’1AŠÛÉÂÓ õ a(¢€‡pb_àëLçîŸE“er ­<3´3àû~vòI4op\c¶Ë.¬ôoÍtÑÞ<&T€9:µà›wyCh1O2ÁCÓûçw §Î÷¬o“ŠŠ½ð1ão6¼ÿs ©b¹"ü%çÍšºåm ×/4®ñSz•s!¿¾›e½S0p,—†:iÊÿ£Þï°âÜ~o×)¬Ì¹m5ŸÙ€nECÿ—TÀ©ÞÓCÍ>Fd¹ÖL] áp ×›$¦ª>‡Ú¨·â(\šlöíîÙ“ü]mÈôŠNêIý¾¯¢ÀøÃ/Ä5wáÞXßúhJq2­¥úú|ü~e*µÆ„½³'‹¢›“Á.&î½Ô¿Ö{ÝD³~´ºR*¼ío,ÞËÎÖÚ¢D#$½l”jfÏËw)' Acèü£õVœÓJº¨aã#~l±ÌÞ¯Ènùd!>w–ÎÂžÚÆãï:cƒôü_ÅLvÅ ·:Ê¡‚¡%¢µƒ2ŽêÒœp îЪxv°r)Ò-Ù¹s2ƒmÞ`f‹.Âtncÿ#:#<1DI’ªÙoæ©Z.¸!‹Ü%÷œ·Ø55í3%í0ü÷©fL¾ ô¢wàÕzFØD=šé·’"‹#ÈÐJW¹T2óÿÄ«)¿sJQ—S↮wñÛÀšcln7,]EóðmiÖìÖ°៙ͪ Ž-¥çÆÐçÃNRs¯šue$f;!å!‰g0p hïW¦-œ0Ë[Oo2GŒ4aL5*6§{=¸o?3Õ4n¤›Ëëê@éÉ¥Àu¦~Œ¼\ØkׯÞʉš•ö„pXÝÂkjövŸäqðC[…`ÂñÊý)°¨¤KcWhÓ epq–ú&RäÙ¶ ƒA ‰¯Àn/+Á¬˜5á"×åYZ´°Â¯^üσwu‰Ó~;ìÏ“£Õ’{ò˜¡)ìEÑûˆjDNŽxðc´~òÃ>e +zþBR<ªiAÁ~"G«›½é%šÊµ]Þ߯Méò{ÍÆ©j=¤‚/¦+áBy¸s„C¬äS‹"RiŒ®~Æ1á•caù‘4œ 1"Ë÷šoÑ'ßuK‰äNéÄ#ƽʃ1„RÁ‡·ŽÛ¢v@©w}šƒ¸2ò wr†ç0n†"±õ®$Í5ƒºá¶÷ù6Y Ÿ"ô†ú®gÕ½ a+Äú\#¦½=*ó½âr7˜tA_ÒÃÚžQÆŠÓ€ÿ)‚u͉œÁ.©Ÿ|í‡j¸¾-Y@ö@ÿpÔF]¾W|‰ßNˆlUغ; nŸÞJûon´Æ\ \#oQrN(~’p¡Žúo¬NͤQ³¤Î«îÁÕ¹øLW3 q[”ÒÓEáMÔŠèi´gêï_'z.’×3FY­âÜecÞψ<*î<ÑJº8cžTY÷ëgL’9djéÝüô¼.$âXT¦w€ G¯1j G9Àè]ùQ5‚rû² § •-·¦GÊ%˘5køq8’_éd;[D™¹^Ð0°±+¢Ü~Îïhp–$CY¤Y(ZžëhƒÃÓóá%d- YØ~[9ÍNòŸß«üЃ­ä<…’–6žÃdƒáy½y¨Ã¥ƒ2Îfèó礗pˆsÒ$ꬶ-—XÖgo¼6—Ô|èÔ|ÍaJºÍ:U ¯ÂÜäâúíŒÇ±±XÛ¢Œ~·t^`ÏUoäg¦ÍOÎõdê5Ìkj%"Cs< §a£NîbÇۓ´gõ^øÖ½á ƒ$¶(°y‘ ,Þ¿)R,u¥XdûLgqƒÏ²ÐÜtì|$ñàËö.Ž[0šA`f•›Ý:˜9)\e&Se()øÏ.-¬þ‘–¾r`éÅåÑøyÒI?fèƒϬi`;LHñYÉÅél ­È¹þ"l²ßä=Gs¼»±MòSã n#¼d>-¢ÐõéÍ/òЦåpêÎ ZÒHa@eãñçrPœEnÿ‚Æ´W± ébrËó¶¡ÎêÔwj$¿[ )¾ èQ#I¾ $'y[×47‚5N#‚úí²¹—ú>le4ÍYS­‡ú¥"a™®Tøög3È £pWÇs¯`k2qº Ãú¶ŠË×l…ϱBî€ò<›öDĽ€ü¼'‹Ñêãþ0Ó§W<^53,øÛ¶‹½¦ÁR¨NGL’ÈqÃØ7 #ayej¶N¾Œ¥ö:‰=ªªJÚy*c-”["ENiÑ7eØá'z«mÇRõì`ÛS¤©-?Óã6æ65¿ÈZˆæÉ3ø5tv øqdÆ"`5ZÊVW½ÿgÝn-³æ«uÇ_Íß;!1‰+z€XXø9_Èü(ªO&):w$ü:˜—˜j5ß&–röjo]’rèEûKÛÝa5~á’¦ÿ `uurÖ¶»€m¤•‚¦yÍù3,ÓpŠ®Aõ£°üýèèNóëÚ7ª+àH"g%sL°AÍu2~û`-óך[›Á÷{|}AúX¶9÷6P`±Ò!$ÍÀsßÿ†ÛbìLë5† ûpÄ}Úº~yÛIwÔ ò­ë ²¸Û™›¥)­Ö¨éˆù²è™ÍíI«vZöl» eah~•Þ” 2„ÙÍì9§â°ü§¼ä&¡î";ʶ÷®_œY¶A»<íH%`%U¾Ü¿ã–ì=OزµŸVûâ„@íAÖ²/“ÙÙ:»ƒà :ÏÖ¨éøP`Žå ¿ˆ–Á®âëzDUá&xOî$¿"‘`4ÛÄü…žzœ7åo“Å:ÉIÜ[í½cï×°¦]kr ‡>ßéô梄ñ—&¡ÂíÀŒ „íq!“oâˆæIû„mþ'==æE@Ô½]q]š71Õ|ºÔí®½ÓÕÆÄ8íY‹‰<­Æƒ]_U´ÃPÎ|#k•¹[ÚzؾC¸’þyé@ØÆ’’£dÙ‚çNh®ñø®>1Þ°ßtô0®©c̓ô[vüPÞPsÃÖ´e˜“€I–쾇µÔå¬ØÇ§é¥ÈO0bµÑFw;p-7c°\ £mß·˜ ¼³'[•®¡1rÛ›{<“ZŠÏb8fªKŒà€fù£èÊ\8•Ʀ*}Šn⺙ cgþ›¬ ¦-^GЉs4[цß`8®¾;¹WsICÏÃæXчãPϯhûŠñ˜º/߯¦-€ly+«éU¿è=‚×4»ÏD`Úµ‹æ6“0Â?qC_øD!¢;"È.£àéÓÎÅÐçfâ§)½´ŒkMýù^‚.÷ãÙxr³Sï'R~_ËÔ]å¨AˆŽX¹Ñeâþ¢ýŠÂ•áÖlIç×C¾Z ìÀ{2Pf1´UÚŠwÿ3Ʊ6úîFçô¯¶âY—§- öÞ¯3:®L®#wVjÁ|¸FѹY.z:’ËÎm´9.ËõÎ*÷ðm #Ûyn$3¢Ùrø y€ýŠKo‹ Ìòµô5·•ÍüÝ‘V‹€óqïr'ˆ">áž*„næn4¯R@çÖ‘äþ]ޱA «1l5%!Lzn¢ÂaüÚˆÞ9¨¾×wë2œ‰Ô-„©a%Ú‚N¥Ÿžp  Š>3Öo*6ÈE^I E%[Q*è@·.ØÖÎ7?`áq­ä‘ß!V¯²Â•+›…–²˜ND_nÎåWî ƒðÿÖ ‡fk“Uå ·¿• Vpƒˆ7ug‚¯·ô˜ý·YuP'IÜÑm="hZØËÖÐ e~ºÈä°Ãÿä+zKþ«Wl¸Ÿ¢«e°:É’¾„Ý» ¶ÝâH™åEu'<~›2d“ömƒ~¡ÌV#p-ÚW/ÉQOJAÑ€6ã%yÖ’á¤ò,#%»ÐD[5(¬ÍìÔì².7NÐ?¶y‘‘ÛÖ¤AdàHÍx»fW{á³BX{ÌÆV4ïD< è¿ß'ŽCáwK@Q*#—W†ºM¢vR³D#¿ñˆJ+L”îæø¬³ °Ã²AÆÕå'cf”ýYEz68ºÝ ôZk2 ¿Á"LÌ5\7àüžÏ]·¶ €»µ tþ)­._åcPÔ®´y¨ÛŠ–ûÇiB˜pO!jüöÇÊv³iZø…9³ôŸ…6 OÙÂj4HdÀ Ü ÝØ :"¸ŒÜG,WQ›Y¸!´êÇåRÉæ»°ï²‡_¡’”®¸‚ Õïc8!Èâh‰¶Ù _/2ù ¡Ÿr{—"ÁÞ¥hB„ Íøƒµºö[ÕÂѸêP ˆ4öu4,Åaˆ{Ñ3Q¢ŠÖüûœdà ª±ŒêÇ£Šty¯Þ&ð=¹IE^`i¿¾—“;‹}múó”•¡efjã/ûqPO¹E˜$H€ÍF—‰¡;~”tÈÕ \\ãÉ®2Ì0Sµ¿®ÐXÏýê*GåºíQ°~3"òG Uï¤ø‚÷5 Y‹'ˆu‘c©Ožã2NO€^%½ëtÖã—“mè\Žì0ÔM¶ ¡Î™k3GË¡êAÜ¢$¥LMä?LÊS’4îï˜xdk±LOÊÜ7÷LG×6ÄŒ¤ypéžUõ\”Ox—;‹¥ökÊ@ýoÛñÌ¡G€ü}ÿ*ÆR~€o®Oš®¾Ò«W0ÎMe¥SémIů˜ëµme6t~Ô$*gܳÑD,D®©n%õ‹àCÈŒ][Ið$ÒŸzƒÂq¿è[-RW5;¢9vše± fÌ+ßÏ…”Ïbñ§ë[ö‰QˆFA$EZ¦èãÀ•Π·Åß#1¦1¤¬ù¦34g|k5nW¸ÿÇQåg*ÄâÁö·ƒã)Î&¼@£Ÿ2öŸ %"¯áA=³Ü€y(®ÜCž9ÉSð_mš°²$Ïð-óì¸Ã¬¸Wê>JÛwÐR­y+:¨V†»§Ñ«¹‹¬±„±#ã(ÏF}! „yäEÃŽÇäÒS…0ÒÜég¡ê!¼ë0Ÿ`AÿŽÂÈCOM?û}ªMU§µ°ºd·gÎwͮ˴ ®ì觯ÎäG]Ö–7ñ!ÌðI¥O“n;2í¸cËâRBLˆUOGýóV± @c~ó’<“å"RÃL<(¥°ÿ9R4Z÷ÃÿèšÎ'O™sM†I:1ó Š:xTžféTwg”–ÿ<“©ÇÕZÁŒnB¢@žö¨Ž#¥~‰Øº¤Ûâ ;Ù8)g„*O¡.<|ííñܬ»^\\>'zu•þÑg•¾íc„¿˜nŠgN)¢:9pi„Ûx1”–O ~´}j¡sèÍ 3+R뤵ªßÚ´ü©ì”õôD¶™£ÈŸ­dIZóⲺM~™Mº¢-Æ’ú¼T5jú§ˆÁ/…°^ÙEDªÜåf°.…ºÆþ7•¨ ^ö•×ý Jß]f #˜ÃY1½”În0“¬F†2 —‰»ñc¯*çU\]‚•{a¬!ñùÒÓ‰øÙC‰ÿ>á‰,*.ÁcqÌ%q8Ê) [Þn>ìž®Ü|÷Ö]ð¥¯À k˜Ë| G¡—ˆŠjA°,‰^T¿ÔƲ†7>J¤¬ÛAz¿i^¡q—UBOué%ɽ§Bþ á$/â~B[7Ͻ€Y§Üí´í…Ñr'\-˜§)é¦¹æ¿ , tºÅ50fš—Hà®DêYÐ/”×g nQ³Õ–E¸žUV™®¥؇H°2nݘ*o‰´AäGÐ^ê Ç^dÒ”¡Mé9Ò²Y•Äœk9#åuñì dU¤øL¸¡³¤Aƒ­b&›îÄA{Ÿs¬rÒØäpOðÛ¹Î4yjOHs³Û®[–0sØ1À×HæÚg°´s[Áï|׉^AA¹ƒ®€3ñ€ä0¼•Ibê.ÄÁ˪OU”â?bucuJw’‹)6|÷è…cÄÁ@/jups˜8ê: xˆårÙ¬…½M@Ud¤¯ŒÛ˜)˜ôЉÉÊh{¥áαÃÓÈæ6³gÛ4óÌo‹ ôÉ'‚ÉË)G…Íd™XQpUÞú“‡Î”]‡^TB Fžƒ7$eÁÜ£µþKbök¨AÌëw\3 ÷ÿèfòCêƒ×’ÿ¨Žt|ïÅÆÀVyÕ“tþ'fÃróÄ{½qãÔd‰6©XVKì°B‹¯[ d jãºoãP@ÐñQѬ*š)jl•¼W‘iˆM7køÛY «£ÈÏ“,‰/^Þö*†­xG°ùˆ”èRÞ·Ãj)ËÜ«;k«OUÎÓ% ÿ ‰ô*°Åd ÖOþ‚1GŠ]üÊY­jbËsGyæe¦Ÿ¸…°O&¤eª¸[ߒͼ u£N+¹"à)[há”)VÍô “GW7/¼ŽZ+ƒµlÂÑÙ¼ :úR„µ¶†zÝÙ¥‹É š¥yíÿ‡ÐVãz{p»‡–’æ~Q€â‘z |OðG8½4º:(˜‰Êq ]ÐïîÇ4K$Ÿž‰òÝ·É>Ì4d0(E4ýB $°Wä¥gWñø—M‘JÒ$LUt%»$¥¼RÈ7²_8êp £M:Ð<öpu'HHä»=0°)!Ÿ§Y ü9å±¥X’ý™„Û9‘ÿÿL÷3ÊÍùê¾~–·:OØŒç • 48-0UV+¸ý•U]¢æƒêCúœáœÀh¼.v†ÛŸX°‡9ƒu]F¯æ¦«•Ì € ‰ ï\iKÃ¥<1+~ãȪ—ê4nWœãÚCÜÔÛ)ZdöBöÿ©EŸ°Qô'×C$¿è|Þ”1I™†):°Q+³žŽË·e \ ÿC¡½u¬©Õ›´JeÔï×Y¶±Û…©°¡ïئPòeˆ…©·|yŽÐrÂ*·ÜÛ±AùãèÁ[Ebº° „8ŸDmoEh”[Eæ1CWËNÿøH:ÅëÝõÝY÷a¨.…?W~b¦ FŒÎòqŽ>Ì€:™Èð+RÕ)Ü!Òl9›ÉJ›4ÛOœôÜöÑ"«!9 §Í¾UÇB!`L<è²æÒ&øáóûS£áDÿ„z%á¶-¥ vÝ·Nßv #b£õ²_Öù¢éN ¾^é~j¢BÖQí ÇÚÁ‚"òÆ%’Îð¡´;X`|&Ñ)«4*“F8/Ý⦸¨Æ®úþ[_ÔebßwtÀ×sú³ãw Œåc*¶û4¦¬$²Îêqô3RÇÆ®“u‡—Í÷˜”›„˜äˆ|¢¹Òh×H,ÉzµèÁñn2“ñT‡ˆ—=Sߟa…Ä#EÑa Kë˜cRUÒ@O£^üóÐ!´”,PA5ÃwZ[I7vùê%ÁöÜ”é€,@mÖ|È'y%Q„øÑâ“Üë*8:ÞTn2îgFq1ÇÞӟư¶ßp‘ä áÚ¡;¢ÐáìDeŒ¢ËS`ÉRõNJá‹zªö„B¤gÀ#’Ê2„Q'[4-ó“Ê\‹îxsWÔ1ÈÐ|NcÈá‹n"b^zóFú3‹Û_N/ôòÐÚèáë+º¹Â‹Ž´ ðŸý5"××1èêm²´äÝ%ÆŸ:éå&eÍýÖß6!òè6bh_bÿêy.j·’g4Ã!§EópËg“kê7—ÆßêÎ_ËyìçÁ~øàs9…£G` sÜÚÌS}S §òû¸ÆXÒób3>l²™™Tr×ÛA)éOBv¦0(òÚÂ¥î}q;DÀæ(â­—U®"Ð.?ò@G ­¸{wu<ÿhö‡¸Å¹„õ÷⤉2w…|ð1xÓ±ó…xö‰¿©äàU!|¦ÚÅXˆä%y9DtÃaÀ®Û*$Úuû°Lb÷Dû¡²ý³ïòQ‡jØTÃÊЬ;òº 8%²-–iò”³êË ¶‹B"%æµÉiÂã—%–<ýå Ý~_ÈBø–J ˜½³KJ%+¿žÖª~'?n¿¯6-טÚAÆ…òå“?Ÿ?A©°VCˆ”T¬SCÉüÊýƒçšj;^]”Dy¹Â½S&spÚ‹ÇWÈâa“Mê‹W‰­2Í™ê»Ô¢µý ˆHÛWð•å}³Þ ½ÄNÓ#vÀ@u±o¿·—ר;»ly¶ãçýg/1*Sý} ³;àÄ[|¤ß¾}WðÝXâ2E4Áÿ?mœÌ‘l"Vó÷D½Éè™Ô²C\Õ”´$7Ç„9Ëq‘V¡DkHñ&àíE”/)³=¼²*è‚>íVêG/ÎCî“!õ¸‘þCOIä&6¼žWÂÍßôÿæÏP‚n– ç&„&”ã È%5ôQMëèÉ´ù­o_•á;¤w­ ¯“m¡3G½"jŽc'ÚüÛÌŒ‰¦¹Í½…º¸6>˜åø ¦+¶]ÖaÞaç«£bËs¥, ÔSÁ*…æ¸ò(ì‘æ¶é×Ú ‰ª¼9 qóôõ|°ô´££ËÑѲ®ˆ!®+½Aý¥ø`jôn"íñß& :‡d°ôKì¯ÀH¥¤:°ÿû´¶Æ9-ÿÍЋÈ~Ö‡ŠÙ(çÙ¥2OàÞ¹ çËWï|d§\¿Vj>ã„Ñð¦éB˜õô©©‚ý¶‰°ÕØ»Á; kšãÈ7¢Z !Fä„Aé'Y„¶ªø¹Ÿ¾0/›ª_iF›S9/;“£™-1#‚À›û~'=×ký¼¹UŽ@¬55•×ÜëzOÔÿÖ›ðk±«ŒvÓòÅ—„VOõå©®4Ã-Z>û•<ÛÐg ¼ŸzJ§ä½ Oõ=ù]½lûXÿžé×ዼj‰‡ú¬Çfþ"ˆ3(øu±L‹+½Ø96Ð7/öç-Ì Uÿ÷^!ÌÕ˜á5öÐQ€!÷ìu4Ÿ˜…Ãů`Þ¬:èÀg{&{˜AxE ?ÙÿX'jÇ F~×)a¸?Øòî÷˜@‡Íwš#ó¾…ϵ"2°rÍ’ôºèJßpKÍ?®"zÛvó„Óº/оŽ[oŸoM£U³ctð3³ù‰vɧÓ'‰02Ó¬‰é2cþâ[é !0ïÆ6©ƒS7µ4]øbg“/¤ d%Dúéz@]òÌ¿£GÛÙŠR´7Ì¡¸Æ»ûïA2ÅTøò9ÇP^üDÃu»wÝßš-vu"ÏÔaz(ßÂÇ—ýÊêæ8§ï_:ræ¤;æŒåòDÂTð†Â‘ »¡àôÇ3öÉQRÂyÒ¿R\G´(p3gÑc èJ}¿—B.¬ŠÁ?óðØyEÙ|eP›†‰Ü×c1dÇJB·Œl:–=X N¶^æÍ^ð€ 5›Í¯r3²˜¬ jïßräµ¢ŠÔþí–+JÙø¯:Î9)ßLþª½ëut¶\õ¡­F?¦Uäæ›ƒt¾Î¶Pækz5‡G~ i×mUBâkOåÄÙ6™¢ZÙÖ"íò—¨ý&7Ê—%åOMDÂ.ÝdÒ‰5s™¯6¿ñœ s"Ýú¥ÿlQþRblúÉþöQvëÖÏDÑÞVôEb‘UùEêV(í#½ë^¢ýéÓfÎ[×ÂpÇ)£»î}€fèVAŒ~UKIfi(2¿Ÿ¶µ9®’°7¹{¶šâ¦“t©lmÅhê8Of±¬}gAÕÃ80W£HHÆ|‚ï8ÐËÓC”×6m!)w âbÆo‡ù¥©v2£}l¨§ÿ`&™vS$Þœ¯Þe‹6¤5R´ÓñOáH.‰©lr(f2ÃR‚¿ÔÔèÈÉ=Á§´YåY¬P>¨*¦IŒƒŒ“ÃÜѬÿ÷ƒ¤÷ì#“¿«Œ[±«~˜›ƒÍ8ù”*×h ÅyïÆ}¿{Ð$ÊZeä–ô#ÙCæžÌ¦³§`²g™9 ŽÉ,… ©Eì±5ŸÆ°dœ€ôÈK?}ÿÞ¦Chñr©F loiû½Ã˜A+ÚNÕS÷è+% ”´ Nžî2H_±&ª0P-È…ÊëmÓ[•åðêòÚ™ 1bš]M’•! q­u ò·Ÿz`—†2Ó(ž ^æ'ûVÉÞ¼*i•OÊØ ƒ—æô1!Öä‡ :ù Ö74·¯v<ñÀÑ Qµ¿ö6ˆ”`Ûßx uË;÷Ÿ_ÅæAÒÒNdŸs«s_ÏJôqò)‚²ýç—m};bÖ×}·¢ÆA\2íE*ƒ ·‚yÂmºì– éwnR·AFÔ̽È7J¸}“BÛN䋺®{¨¶Â-£RþŽªÎffñ•…íg29 ÷¸ Ö™¸ð¾Î€q•„ªft=oxwyA¤àÊ7‰ä¯ö Üðª”tÿ«eñB*ü¿Éݸǣ~¨âLQÖ©ÃE‘ç…Ó²ÁWyƒ ƒ›ì Ý-3±\¼mÙÑ<,¿É¢ƒüJ´fúß÷H< ôÂ$¤wŸÆZ® ¹º_œ…÷› ‡¾ûÞn’ŸÛÈ9r̰RôO¬ Ãz2(0mò8K‘ŸV§VÞC'¼2[Œæ„T…$Œ›©¼ÎšÏ6 ꜌å†@M?ˆ§GN¥é8oÊU*½‡1€ A˜ãb›OèiÁY­‹= ÖQ8¼ù7‚Íà 2Ó_+ÈI‹Gõ-k  sØ1ÆÐÃn¢ÇG;6 Rª/âðPùSAƒÐ|×Ýœ†ùbs² õ›þ{NY6’ì–/èÿ"‰ôfÔ¼ž±‚Y³Ä³f"ÝGOírôEo(2ŸÎmskŒŒ½+LÒ“¥sÀóÐóøÖ+Ö?0zŸ4È‘}ø¥8A̵×Ö–["Þ´ý,rë™Îvò‘n†Üõ:y˶ã8„ÊÊÊkååIº kÒ.ÂÑÀý‹|˜¤$Ù$ÔlœzÉ­Úæ"ïÞB^“¾´ŽkÏ4’ªb1‘„î3Àoè&A¿Q“AI“Œ5ǬÃXÉ5ÞYM;̸Qkm,¨.lÇ8ÖiÎiadBòœ¨â—wrVz¦@L±7‹ ýÉRw-³K pÇc¸OSèý`ø|‚}™¦;Í|äaÐu¨"ó~Pm û„¼L ;€@ç>f½ û[eÔ^ŽBÚÜ9$hÁXæ)Ý ŸKÉ;E¾Sõ4"ëDµ~K8¿Óè†çÿTÕìBäÏNâqOƒïˆA'JãVsoÈ ‹î³Ïµ…<ýÆWÜÌ+Ö²â¤~r–dÂî£_²RÑ<Ô¯"m+J3Â%Äß·8D!Å ô¹FàkVÀ1¸ûPЪŸí^jnÅøAÇð¬çCøIõUE±*U÷[¿âpßz¼ %Ì/7pjHHmŽEØiI}–Óïù˜¹3P‘¨ªgïfÕØÛGäñ™²²¢Â§ÁU~ £X4Cpl¯w† 9òË2ʃc¸D´ƒåÖ‚¨yïòMÍ’OßWõiçÐÙSàØqãrÅ5£ý]4BAƒ~kk5¥KˆNÅó¸n½2³àãDÁRõ Ð-…/F÷Oæ´ï\ÑÃTñ?}\ƒÁBïÝ2A‹¢»½¹M™¿4¦fëžíQë'Ñ!—8àyw¿Øüé]`–î]ÃçØj3âÙÊÚÉ9E‘Ú÷èd…l#ù;¥Ê^Ç`Õ5ÌÚàŒ0OF™,÷dh8.€¤¡›õÍz[‡qÞðüTÄè3ŽX¹éM‰žŽÃƒøï`Ïe2ñX꣰ÕG#ÉÓH)3„PGÞµQ j½˜¶Ó%À>í"K)…UV«ù_7›)+æÉ­ü]ÝO¹M})” –N`§Á£k"ƒ¥Jì–S3z׺weI!©Fm»e  d(yP”‰€ø5Óçk«"QË0ì÷{vÒQžï¹äà!Ð%hd^@˜‚¥\͉ŒÍåäÁ<’ÆŸ´_XÀ—ú±©‰¯8¿fO$í[ÑÂp[Ó=5‰ä›kÞ§¦e‹´!ôÇå)«ê ·ù»]Kí&Fö_áž“i+@ÅWÑíLéÜq¿›’Í›·«Pø Ådt“7ï±A`9„¤²‰ažæIæU†üëø ,Sÿ¼hÏ7ˆ¨Ò¼V-!B¢n:5Κ+“ºyÊ#Õ­BÖ øÍHM=úfNèá,DpU¶ìÿ¹ÙÉ'L¯ÁEŽBƒ°òÖºReèàM †ˆ…}&™¤WæC´Àcj›Ÿcˆ2¦Ý¿Ô¤'F;ë+¦O£ ùø˜ŒŸœ¯Äì{~Të •|l£l޳™=¸¹Ëvêä «e +ßæÍ‹0Ö_Ñó|7«×#¹Þ°Hui¤7WÕ¶Zdoo”Aµã]NꙈ—ô`¨š!¬fžKåY¨H¨` ´˜^Ø ˆ¥Ú©ŠàSÑ‹\˜U5M°º«”'ÿ/i3 9i¾›ÏD„A‰xjP î6äR|HüðÚý~.©uó'ìâ´Z;:FÃä'Eê=D"cÒ'ß' ‘¯]îãÞ^"÷ïit1Øö×{`¢7â³FØ[CãˆR-ðçh)×µ~ó`}!lS¹dw°]]Ã3 ‰%àzTç(Êy¹Ñ=¤Ï!+`<¸)CÁ»¿Ý ÔçÈÑ•.ìX,c(wŒy‰ügÃ]^Þz5 çæ8-Ü|‹ÍQ°<÷WÁipͰp¾óò~I0PÖ‰Í<èÐꄦjêrµ·dh8Õ•WªfgT¼XY1 › À>ùl~ #ðán©ìkäú¦Ü–fƒjC7Ê¥Ù7v"Ýò ‰naVi|fƒ£;bZ쑽 Ï|N˜d˜’eØ âN&ºÅª v¬–´4A¾dÑ^ÝâìvWz²žÜä‚¥¯ŽÆÂò%Ä«*šÞªœ¨ÇÜøÕ8Š쀌UM~ÕËåa’ƒ-½'ëÃk12†KrÝ “zt‰œ†IyÁ¯’ÄØî»~–$BLù‹ÄeVïÉô Ùײ•ÌÖËã¹1Ú¬¤ Û ò,aúÞI3à˜pN™ègwüÓJûŠlu.èñzÌmkÏÔº 3yaÿ%|Fk`µt¾^æð–/þ›³Û|¤-1\ð…¹Ô#è&@-ÌtŽãš´K>=I²‹ÈqoÂÂÂt•ÏÍ%®‰“\ÿ2‚à\ ηšUšEQnne%û…4I¶­/T’9P¼¤§¢zC:…ÂX+ fK­.&»R™ËІ†!ØEœnãzÿ-MðÀ!`CûÒqžv€CÂѳûðBP±‹ÿhFâÂ!õbÞ18SÊ•¢(È ÍnãjöKÛZt  ˜fÄhîv=_ˆTëtH~âW7¯¹; RºÖ6{y3ÿ"‡gm1$á4æåˆmŒÃ±eÓ†?IüTFóo˜Ð ßO$¹(¨C4ÍŽB®ŽJÒýN;½gãêUS·®…„E3÷¤ëG™ï……þ ¢¦ KØ$Ü«yñ•ÐÈ£ÂèWÕøÉ¿ )cX[ß×>-7Tm#&¬Ð–(ÝóäÓÏŒI¸”„†ˆCÙ‘“-œ²Šçfô ØGŸn£#vц˜‘ ßáP" \ˆ€vUMRŸ4ãÞqÙ‚«Më˜Öªoµ6äVêlv¸ÇÍD¾3"`·èäEòpÅ;†Û Ô¾ŠXà!i ¶M;´t“q·1/’‚x:Ú|@ÒF7#ý vC³ÇÑEd±P0êý¨žAHtÀbÆN™¦´Žg÷â ' %þ™‡j^ìF ®àÁMKŸIJ²nëë6ŸÉÙŽþÎ9!hþ9ÎŽS"Ïu˦EÌä¸nˆŽÿD@­T çø½ p}¿hÑýTCñäþ/I0ˆ¢¿^çº3tÂïeõüŒmäã&Àdí÷äÝ=`¼‡Æ±ƒ‘Ϥ“5×ç9 NRB`ë¾ E¶`i¢öËfÄ›á’Þó‚º$³ððÁWãßéJ{©U\ŽwǬã]‘dzNg‹žMªDe–nHÊɃT•÷'âè$7‰ pˆg®ÍÇϰaþÙˆEƒ”¡àÛCɦèC±€ò¸Kž°Š—þy¨}â…¤ÇøÃÜY¶…ÞoÉ×kõE¤‚³rч¿W7 [ô‹âƒ ú(Ìêx ¶uIµ¯_æ¼”š@«U¤w|ÇÆéÚ Vþ ;zÒU§…FÖÓÊ]WÈ™„á&î“ß¡NѾ×<Õ6ÅÄ À_ÀÖeÇÑ"Ê5áa¥•Tz:±q‡R›¶ÄšLEq¾?÷b£f@,PûYD8“80øœ*D=|ÕPlço†!\ê`·ùdÚU9”#sÁÍ{ô:û¿C–¢îµ!ÚÝðFR_¶—ÿ/Wú‡pJšKÎè} B†°£ ,mŒêf9dç“"À"v+\ZZo±Ïøhq8²V{DBºey¾HgʼnØRïÉrg§¾ñ† Ì Eò0T,ZÅÃ!P™Lм/?ÀÞ†F®þCÔp˜Ef„_o˜ûEãä å#ªŽêãEdi#N™À³†0ézùÒ˜´ÄSêMkl ƒ ™O^'ô艃¼ûôFŒ–{]^s,JŒ€›"ÌÈÙ„?~«¯&yk×ãV~™…· UvoÉ1dÏ4-‡Ã'ÄçÂüÔ‰)šH1Õ§•lexÚÌC&x%æ’s ë»JÖþ6Ÿ ÛóïHÚ· Ȧ¸Ï…–×füLÙ\&f hÿµÚô]Ñçß̨ÜgI™p˜Kí[ZŒƒ+ŸLYóCõ§„uúÒôH„ʘcPC½Zµ14I@yK€Ó¯ñ^/~ûjÅIç&B xü’ûäyU—=5»g¼éº¹Éð ¯X¾Õ߱̋¢Üè>ƒ&†¿Vb –íêbŸ¦ ”ªä–PÃPâ›oó—Ï71jøTùN“9J‹ÐòabœÍ®£v “ýêîP#£cˆ½ Á þDŠs-%Љ´‚§¾¥ÕÛ4]ÉPX[áP!|$XÍnÓ9®`_C…”O.\iDiÛj̤Ç,”}ÖâœK)#ÿ¤g·¼3bÕD)ºëÖ†¥.ü©¿ýX ÎAOmLJâÃÊ^gž•‰ÞoíʽŠ%œ–lfÓ|Übb·Ÿ†Å"`„‹>Ò‰à VW¬%ý ¦9H|µ²j±k¡h˜€îD¬½ÊIU£ ßÔ  |Ë’¿|ÚPf÷öv[ ^ÙU›)À«jŽÆ&–Ë€Òè°{…¯”q:ºßÙª³æ„ò´À‚ì/¸â9â|·cä÷HO†Sòm1‡^9úòÁ!–04dPwó¢°ÛÀéÌF÷£Ï]Ø+úFÙ vÚ².Å9£2í/ÌtqöÈ!20îìï½[·½ì>ž¿±¬:®LzÇ¢™!ß… ôæpöoÉ¿ªÂÅ{¥E1¥e{ó]•˜‹pr”é\=ÅÒËuøéYp{ãÎK|~³Ø„ÀFf³ š³ûÎô~žGÛîã{°{{8n"- Ûµƒ\U±ÎÓ.Š_ZÚ†±yp¡…·8 @ wga.:œêHxÆAÉ¡„tHƒjeKRªÅ~ñ¥B-Îó|.Ÿ<Ñ=•ïøOÎéù`uôâ‡ÌH>Q;‰jBc,¢o‹l›ÎY|é½÷­ÃÑþ-2«\ùÇ ¸4âGjÙš‹‰Jûu¬¿ô[ ß©NM­CF×j·„pQM Y®a)e¢ösŒsÐ*¸¿!>À®f.B“gûeG!òzrSy-`pÛ }ìAV’<ò9 Σþ,\ÉŸK{sJGþ™Ñê<R°mFÅÃf°œ@y–ÑX€à>¼bíüR,.5Q¶ª.õ»¢Â_šÖ?§À6(ä(C¨5ÉÅ)ÈÇÑ'ÆÐL½¦ÚôÛՙߤËÛrîχãþ…^_¶ÅgÖ¾KPñöÇ ïÔ l ë‘Z¹½ ïrµ;éþQ‰3ó½*D¹'B41X›ŒF"ç‘ÁCíê › °šóù£Þ檭Çúùšc#Hʶ˜Bi7ðµB#¢VÛæålUÒx¼·Qß¿*'b=Šš!;fwT–.ö¦•wš§T­ÏðAã ]̹Jjn‰ýaz¶O ¼ºzTÙBT}=.£$w^›©MÔ’L2fªü†ÎFZ(Ôã‘mô&,´7Ñ´k0‡ºý^xææhïÿ¡m‰÷ó¥¶µÇ']RXÞ•ü“т۪¾†÷àjȨý fÕU|ö§Í˜ ^^íÔÕZã¸O ”ƒÀ×Ü‹Îëï=™’€ü#5†ÈŽÙâÖ+=.ëšêì3b”V Ú¢#“,âN†uHìÒõ\î/Þ‘ó¾Ú[xÕ\uýzÍä”Õqô¢¢ ^rzKZ²HèKßðÀ<îŠWªI?w·:þl/¹žY{C¾ø_Wø-Ñü ïnœö†AùjßfGÕ_vÚç°“°öõ{ô€^#Ï‹I\ôü½Ûî‹ÿÜ'þðf+†¾rÓnóxH¢Öu?|È^ž-î‚&YÔmX½Ô›æÌ´{%ÒÁŽó9œÆV¼v§li&¤‡ %ºüÒ`'^L]9¿lÖ@‰Ý§ðú¸š‰‘zE,ï3Åê×Àcë°• £¢ð•‚¾ †¿†²Š TË¿ú¾ïàÜ£w—ÓµhI(ŸÊ¹sCøy%ŽÄñø€õìCÚéoŸhÿ“u¶­ý.ELõv¿{Ç.EÿV'Û÷Û4:R’ó fcÒ³1©oàùìèg9EàèË&ŒFYçCã½Ð[çëÔŒv³ësÌ”y7[¥HkuyCcâÒTšöÀ-]†É¼ qµÏéÒ³ƒ[Ýrc–ÖßW<Q6iŽ)䛨iñqè$~qÚÆ­hžŽ²o›2øëC䯵,l M±|‚U~…i!*Óœ¾Ï¬ýåÛô+îˇÐwþ‚¥L¡'« m|[:½“}0ùQòßÊJKa¹Z™ŸßAÞ*<¯Or<>¦0‡¸8Ø U¸ûßpr[„æä¹\a`Ë)èœìIcTØ#ÐÇF^wy7/L©Õ¯Yxáüê©f€e½=G3 –ÎÉ$‰®Àb˜0”ùeá!›ŸxØÑü’Ⱥî_fÅù`À—%“£Ê"suT 0?úrßF®ž£WN;í)Á&ЭáÜ…êÊß•C[ OE™3&ó=·ã„èþõqxZ›ðâEÉŽN)‹êj$—·'¾_â媕ïì3)é^.ö2 ñä³ý˜r+†v®^fØ<¡UØ«eÝmš1«FϽ¢ Ù°bЄ¶r>aN ÇÛêÈ=›>eÒŠvÞeéþÞ#Qöÿ7ä7t`%•&I›ÉOå·ÔåP'³Ì]ÕJ²‚%\’6üLòŸ›3æ\Â{sé)ý;§·_‘²íc“â¥\©ì·õre¿Ñ@¦Ë€,vn"o~ÖêÈØ(¦¾eÕ!üÞQ)Â5âµY÷S¸ª‡Ý©QÔ^ÇÝû8þá™í 룫¢¸£¾ê1xξŽ9ÅCÊ@ÿ¶8K*q!4¡‰`½ÂfàS¥t?8î?â·¤"J ´²:í¨Ëa J–q« r™ƒÐ>-ã~ŒŒ•€zùeWÿVQ—× Ã÷·0OmÊ]÷ã씪6«RÉä•Mòƒâ;@´ù}N¯d­…½§ ϘÇÁ<.Ñ rn¾´6 ãgµ}q ßôíÛ‰î7ƒCkõ0ynäªPKж’á«Qw=±Ñk5Í0|Š`fCrà×÷EE2JAWF!iŒ©`¦ä ¿÷ÀyW4ÌmIögô 1Ö°ÉuXTâwÈŸ«4¢´6û®¯ãžù+½Ý‚ AËsX²^kÏPS6ƒRØVØ‹. ¡7PØDöo¦5Ñœs¯³¨ÿ]óŸÇéË`ÿæ{¶§¼>§6—Iì£àÉO”îÎ!fs³N,èZ#0·‹è NÞsŒ‘t56š4L®dÆ¢WoÌ0J¼—áZ¬˜ùccìI ¼^©Î¹Bâ”`êʥΠ_#,yûVžóÞT´·L”“¡Ùf{†ÉN3¶xܾ¿™IÓ0 ê`qËéú)°’¢Vw/–^ðYßíÚC0Ÿmòê²pHX}ÿÈMÒ·úÝX[zEœuÐÀØ„>ù>Xvj‰ mL“HD %`^èl¦ŸH#õäðê«jŠÐMææÀŽ„‚KHoÈ›Dºä³kj¾Ð>SÔù„ó*\=×Çž4êX*š´¡•K³ó,ödÓ–d¹ƒ öO— ü±‹ˆî´ 6£—¤2Sjà}ºDš¥zÐÓ蟦ýWïd5àÀò¨ãˆ±¯H¾Z:ôVtÆä€PGðð.nmÅ4$Òê <–kÔGçÌv¾¶•-ˆˆˆ}5HÔX 9o–ä±2“½ORC{‚n\óÒÄÄJEZgÜGÃ& ½èe«†ll$G3*²ä])Îm÷è»Atžo²¤ùånx\ f¯†lmH¿E£óVÛš&2!ßSŸôÎ8•ûü¬rÛg,é;Ó?²”®QÞÓkŸjN <Ó»0¼^üÄyºáŒÒ7Y…mñiâÞíxMH  Âûèßhó ‚ŠŽ¬§?“·¨ýõF éËV3làj=“çzI- ¨'™ šoÖm³#â‡ûç!WjÝ`‹¯-~}HG<´u’OOEôƒ…ïíŽû üyA/m·¨.³ì…kϪ9…#G’ÓR,@Ê8cê ‘SqDòàU_> ]Jž>‰—à“òDº˜¡å/nIdq¥74ºëÌÖïýKZÙ .af8t¯ˆ¯WªC… ïÜ¢«Œ¼ÀäL²Ë õ&?ΡãFQ‡¼zêÎ"ˆªXÝâ{®ÙIùAÕã$âàeÞ=ì¤h¤5” ”vé8z=-Ž`{ZéÞ)1~.—fѽ8Ÿ:_ÜBlÄþ»ly»†_Ë‚nQÌ0îüu„z;d4±ÑßbÖg¡(|1Qý.…7Yçúà¿B©0'C_Éi ùtˆ>‚׿18èà$͘3B˜®SÁñ)y”í(õšókØŽSDŽ U§Y¸‘¢ÕExRð Ì9û2ÎéZvš Ðí•g U@¡/FšÄ&¯ª­¬ß@h÷Šê‹sMxDLt“’=ª‘.†ÉäNyƹT ?x{£$-¥ŸBÛ<|X!+±Ÿp3­¬ž4µ÷îÆj¨•|´ùIéݱ'Ƕä×\,éû% Û{ÙÛìcy°Z$6-ë[¤Šl'àÜF2ì}œõÏÕªå©éß?+êc÷Ò{ŒÏú;>°7ðÚêbt·u=³\Û/÷ø„ωOž£lCCP…°pˆeAä÷âkü 9=ÁMËÇóOÎÐĦ˗Q…› ¤æäÍ{»­~ú+Ö+ªÿK–Ãò=ú/ŸKè2$EŒ;*ºŸ„N^¯W/Re*ú?ÐÔÄê1K îÄ!fREÕ],>œ©¤ “è Ôs¥•R²Pq[CHÂfÿàðÏŠ¶ØvSUâ-a¸‘`t3iºþTÛ6ÙzççZZ¸Ÿoö»éwtx3¹KÐQ¿¹­î¡¿:)5ËÑ¡Ö?ûXÞõÅ ÿ4Õòd>­Öñ-ët“i—ž}u*0ˆóÿJp4åj(ý>ê1ú³ŒŸÀ,‘n´›ö›Ï2霖ífQ (÷3ïsÞJÕô/ c’=Òþœs_¢h«KH€ @ <Ñô¿ |…Ï" œºz¡²UHÍm‘:.ÏXòW­ ;¥‘Ÿ’°0m9Àn y;ô®äÏ÷Êü´9‚˜XÉ}ÄoŽVzAè„L­ð(„hµÐøµëtæFðWaÞÚ˜ÏÚ›ºÓ³”¸­ˆz™Ân¯{ï³H.½­ïVö6ó9c‰ZY÷6¯|ÐMæêKó4Љ0£¯­–bˆ#ñhèÙ4À-‰{¾¸[?'!ê±ò®‡2‡¡öû^.çÓjµ£A…º!-ŠvñVÙ“y/%"*:³i!œ²y}F3¹ÚØ~{´™êEôVr3û 'ùú÷²p©â2”Ž'šõ—‰i¸gÊÖ®ùúôb` Õ›bmMæï™'BÙ«XˆE«Œ§–Mìã §ŽçÚÃG*ùÆžBªŒ"ã¿#ÆÁÇØŸ— km{ó™V´à, ¹îP¼nÌd¡ |²ð«aâ«éñçÝ¡~²®i%>hÇýMˆ· NQ‡LÁØÕ»îÑ? .ÊR±!í|F¶Á¯Ëe#ÁmÛú(ü¥ zaä#E7ÚJì°«%xbG©…ûŒ‹ßÂ`’kN‹«Ï™¡‘ö§°#S’6(¯*⬱f[‚ïò‘¬õŠ;´(ªYYT§¹ôƒ‹é¸Jõ©{u+÷åSŸß%í.ï­,œfÞ2ç{¶*nZtFê3­¬WZ·2á·@‰$¦&Ýç¡iþ=ë¹à¡ó|:|ù^_«[^¡4äZÄ€âŠÃ< wÐWf» ¦àËÇÔYRÃÃéOnÔ’q/_[}.³êÒ¢õH' ñ±56˜å;sª(e:æ­Õ^$~ºFÈñ‰˜e…+ ˜!ÌåØÛØ_i9PtT…2$J¨1r¸M"µ4j43|5ì2üö*â>Œrô¼YAå ¡Iù5Ó+eÕ(Œ—š¸L¬Â<¤°Ç±wàjáúâÙºÛå®Ï!2ÿ è¹ì˜´)²¥-C=«¸G®eÿiŽ$-ÐÚª‰–IÇõhP˜”|Eðœ‚00™¼Jjàz¬Mµýw½Éb”ë«dMÂÀÄœYpïßze½jæ] ”Åë¼°b Á]ò¡Ös¨)Üx¦|&{0½•÷›‰)„~R¾”&Qz"r‡T–-ÄsTDåv&¥r;,N 6À­èɥߋÍ¥/pÔ­C`í)3&“…•ë”h¨4úMíûÙöûÍÈtp—P¿2‹ö]!—«‘ ;ëêÓ»1ßl·ˆ8{¿Ø HlŒûÐW†q®«¹X½“Y;2ƒ.C|ªþT¯!Íß5•~2…‚ÊqÎôØǤ"ˆRú³§@[ä0Î1Ä‚Éqþßm7”:d‡"h6*åIJ 7|3Dqˆ7öÉ~GÃà‹óþ»BÃ{Là]¨FX¶’Sµ=%6Àé(UÝlãÒwĪ¢ñ›ÞÊR‹røvTqÓëBÄ(ö¬š3Nc]÷ÖbÔ'N9p¬Ð*]Â×~-©ÏCjƒ99abLe/{m{êeZ¡?ý7¸” ËM…`çL 8ÔÀ2Bf²¢”àAzoŸÏ+H°±ž¥yÓ®=§i/tÁ«™Œ³+èRÁàG»òð³÷ œ±¡ŸZ›uÁ[þ›Ó¶úÃþBú;£Â,ÑgíüÜúT”{wûo`Ÿ=œ&€þð#iiâ½Õ›~ùa‚‘-" SÛLtl¶|.‹ûúÁ§ðü$¿&wfѨA£$HŒeù–5 ‚ÅÑ‚ÔâtG¢6˜úÙN“&$U³¥œ¾°Tn­§+ÆÐIÕøÚLnø…ÐÓÍB[á;j¶Wn˜üõÖG,—S:PXra£ø Ÿ~¼u•0¬¬k ŒôkTBÇŽà•€ßN­Û~“©yk|™Ñùÿú69ÚǤɲ÷뚈,Nát^ëÎÉ7σSL‰7 zadp3gc¡Áû)練/XHH?¥í¿ó Á/÷ýÒ×°bYÌQèÆEî8ÙxÓw—üh!>"žšò{•™Œ£Å»[ÇÔóSÉ–×nЖÖeI© ”æ3Ÿ¡oS%pØ”+åíœVÈ-|ÒrÇ.2•ÚØž+ÚZJªRíG;°\ž–(…n;7o(ºò`ÅlÏØòÀ}uæ'$}¼¾zùo1qÍ‹˜–ÃûQEàЫ¯ÄÊû öM\WjPu ÐÖJÓ ‰j“ýu5À¸Q¹Ò Y97hÎãšEÁËØ|Hl!Ž1¥ÑA°1'œÁ„é¡?'Y/þ½äËI†õvÜ['­4¡VWª: ¡ô» òðY #Vp8-ÿ’0€ÃrZ¦´ÌÆ1‘˜&"Éÿ0íVzÕÆÜ õÂNQÞƒ…)™%ÉrXÒäUçþST“ÕqúÖ Ÿæ]å]ð6c?ç,Ö8Û-ÁÄÞPø¿¶%=Wí\ìÇ/Ì]”Å‚ÇÑpfö‹?ÔÙ}™ÕÛÔ¬=ûJ¢ÈH“X­¬Éþïx¢(Ñ.çià›K¡6âÿÎå~zy÷àV‡‚ZIë »ÿ-øúMØ åP¸oЀ\SxÚŒÂcnß W„ëoxÀ^Ô¸¡úWÒ)Þh¡[agjerZ\!¶?y%矔ý‡ÄÐ8ñQæA‡&.øùƒ|.Ô$TÃ?» n ü¤x žSR~ÔI|BŸÑ>(–È„¿¥™³ä¥>Í(4$È©™H‡X–‰ŒÿF CÀ&¢ÖHž¤¾•™ïÅEST­SxŽßO¶?Ð,™T€ßJ˜™‚Øplˆ $µñz¼;$ƒ€hó(ØõÒd¥ þß »434§éx/º³ÄdÞèÔ ›DP|ÐÏ3$ÛÈ0L[/º¤•ă ¸Ó6Íqö@n ©1=A·uŠT<!JxqJTÚ°+%èPåò¥!j ë¯Ú‘fâN?æ&UÛù .þÝò¿óì ÄWH‚ËÛtNNYÁRX¦ÎH™®+L[Șs,.9ösÄ6V òã g€"r³L|Wä›6޶üéÓæZ@É”FG¶j7T$ÄÖÛãÂß> bÉÊîD—߯Ñ,#z­2GóKª„Þ!Lëé@3,ÀK6 ž°baWE>ÜsÀPhâž]·¸åUy6ç:¶ÃŽãÂðÇØ!xALÚgßp×¹$øÕ‚UNY~m‰ª‘O :Ðib¡ oL´WäY­“¼n{bloƒÝ"z@x6òVÜ3ˆÒsVì‡u-9FØÛ†,í;“k“Ò„ºIÒ–.E&öJ ,ËRX¿¨É…&¦%Aü+â½wBº{ÆåÔV´ÿN‰`ÖÈEC±j2WpÔÁnEPOÅÈ+íraPJ_^\Šgá^%zIKÁr¸HÈc˜ÃM[àMšnd9jxüdÃÖlÈмÏ0_¦™¥$M¬_ C¿è7-+Á¬–‰G}`ÊPÂØ{¾ËÍ•¥Tƒ[ìA©Ñz‡ð€Ñ¦ØÝlö[ïktYÓN¶t HW¿b·mùÈݧ(Âw8 3O©j›¨w½—q[¥ô4½s÷fé<´Ç𬠨d]ÐA‹2pËØ÷Q2Ö×8?ÈËrBïÌ Ø¦GgñôÈarñ}¦°ö:ËÕ£²YSQLÇæò‰0™£i+¦ '{BFbÚ– ±A¦T×­TàÏáUå 6Øî§ÖÓó§§·áçâ̺śš3 *À¶â¼ÝXüáΘ•`¤QRW ?9ê¨ýA¨1™Wò¯+k‹Üš0%"%OаG¶èW»ÈÐt!ô+s¸ä〠KN.Øö¶l„Ý={V•€ê­ Q—¥ýË=>— b®Ñ€¬ß©:!_h†;µYÇÇéƒRŒctj0¶$ÈïQ0—]*8X#¦W®V¿´ßÉ’ûi-É^×2ôM>Å4"Ù”[›EndP…àIÓÿÅÁųíàié1!=I+.ÝF˜Î²­¼‡œ ûÈù@€Œ¨(qFIM›…iY‚ocÓ»ÿK:Ù‹tº&d²¨pÛþ ^®/$ÖÜî­›OܾI¶ì‘ޝ7¯P!½ynegu_…–1ªÞENõ˜W™vžœóªNòß)łʟùÍš±ga¦=5&#Íô<:c ‘îÿ°Oôãt\=uLPim8¾YEÕŠ*û§C~à 'kpˆDJßϱƴM=©Cü…#¹þ•Ê šJ_»£X‹‘ðeß#)Ñá]û] ½ˆÜuª¾µl‹»]/’l}D£Ô¹ë DáÃK¸ÎÀ”lþã–”¨µˆ¨Ÿ2 'rŠ|2nýg¹g(E„é}ÁåzVT‡ð‹¨,Ÿ>Š›ÁØ%. r)t³ðA¬}[Uñ‡­“ÆÚÝâ6Uœð§α#¦]ˆHÏÞæx‡ â?AY)–?žfùìPrxðÐ<±ÎÀõ›@5áÈÎ3/Ü)“6]Mxo{«ÙHP´­c¬÷°¦–tC"2+މ÷2K?¿zaÒ›uÒü˜×R8¥öo”Ç»‚nÒMø A/K=á¡*c¯–Vx¯wjQ;ê°Ð5ÔF½S ­•°ëh4ž°Ü’ûFTñÈM K³¾õ\¤Ö8 Ns‡ô K –°DôßýÍlþhì “á¸×°3‹™OÇ >ÁcêØÏySE·²Î\º[$?öÛsÕéH¯ŸçÌÆR¬ìœõØ×4r.Ž#Ñù1ŽR„»5â°g%u ÌõÛ³¥µtž >\¬µ¯è7ãí³ç§-ûÕ´¬ºYÕ⦈ó¬ OâÓŽ;‚[³v²Â£èlÞQ@IC%9çî´Sn©³OÕ¨>²Ÿ¨2˜ª!劾þ©»•å35L¡†€òã‚}e›q›þÌd¯Xã ÷µó[_ç]Ù1¿EˆŒSólcôQ%>µ ~y¾€ì­K 6Û½cÐoø5t†ÿÊ£ï ;Aó{ ²hÅw®³x04וOÊHQóe#´’O¦¸®ÀӪ敯`èò탧ˆ¤åÄÌGbïB$”™zÉ9dšòm¹”8»·…x$ãÏÿµh»*NZ3® fÖíÙÖâu+ÒÈSëm)Ù©G5îðìÐ÷î½#z=~&Á ù‡V†›•‘ã3É6³ K%íañ¼Òpá8)fªÂЉ¤Èœ þs·•(äO\8¬ºl¤ÖN Ñ+Põ‰’ùÃL“ýg”‘Þèzl*Ó %seõ¶ÃÔüÞ)k #£¾H„˜ñ¸-ÞºˆŒôgÓûc†Â§Ù¹·iÿLïXi…*«©ò gÛØiž€¶‹¨€ò1!gÞ¬þ౑*‚±¹19§oQØËVˆ¼  >oO_=ýλ?ÒZýÆ·ø·˜ £{^3Žþï¤všôÁMG”5‰ä†ö¢hª—ZW¯´ÖF¼°·7•¸¦`ãµsNT}Üyî½×²ôÒÑg„]‡¹3 &سûî¤NÂ…?œÂ™¦vƒjËváæs¥PQgŠÚE”#[¨ì\åR9p†Qv/È Ÿô/nFƒî½#»Âª"Èi…˜Ìð³ÐRö¦øf¸·H ¥g¶ž¯”ØîП\lL•ÑIùÑü=`½¢`Zr²8Fs>c4·A®–ؽvk`Y12,ý˜å ÂŒ¡)6Æêîe™?6Àkwöè—åûXX.¨­¤Øºåª^cM¾½©^G{•[7ÙzÅYO~m$ßo(ÒßЛ‰†E÷Våäš—„ÈA$Ý‚Þ$±\÷‹„:ZâÇöŽ ÎÝÍ?xFºÐ±¹æÑtF·-we'ðÙ¦à(ö²o÷W×R˜è¾“í3ª”2ɳ|cÑ;»JçySJŒ¤Ö¸ÂaÛ+&vEB«§Ž@n/©µœbƒÙƒXt‰Ó-)\¬)0LŒ{G—š?ÿWWF#‰¾ç Äá0]¨‘Dÿ{M\up#±xl·jÊÑô”¿¹Ê–Ì7ÝŒaŒïu³åI\ܵ§âwU À°wU÷É(^-y=™ûÍ`ö¸g»¯7 ÷ÙIÖ<ÇÏ—ÎÈÙŽ>%9Ý0¯Óð°)lKïŽЄÖXm«˜S̤±¢'9 x}ì@î9g#ˬԨ2þF~?|2 “éÜŠxŽ/ÚÆ››Tè³Lk”õ^ÈÛžµ—<§|E ÈHJ…!Í “<½zìšÙìH]3Šm*Ù%½ƒ7ÁƒH3~7¿È6¼tÀΛw]³i,zpIÄy›R¶ð.+¿:y÷ :{BÈ…àÏ8³ÖÖS2»ÃÃF·×_ò䧉=ÍýN—yöÿ!Î79VÄL¦Ô¼+ÇU*ÑÙŠâV_ˆÉö3pPȨÏ4•_T|€(ðËî>Ásï¼¥:Þ_³ÄíDó3IpO¼QõäO·‹iªŽô98LzT0µ—í.©Š/žghnÄXá,Ò6ÄÍø›Ø&÷ƒ úié@à"¦¶Ç°û¢3Qþ¼~8ÆÞ©FE rWÜ<ê´~Ð?J‹_ÎÚéº^S–ÒyæìÙFÔà©ÃÕò¼ÛøOÆüûã"G„ýÔ„ܽ»ËB¢®/˜•–=÷•/Òtš7ñ[aGûܱ+6¶u¾¡•j&0`…!Çw/DO]nf`TæÖpaµ%7EnèÑé\±›tøÇ•Ô~¦&«ü¡%Ý‹He9Ü¡píǾcOŽ¿m !…B<7¦¾ º®8!–¡»Žíd]T_žǪ¬â²äD’s¢0õTi#Ù› Ë…öIG¢Š[Ÿ–ÌðÙc'¹Æ–a½iqzAЕÒe:( IÊH€¯T=œèfÊËL¦®ì~€ðöJ`]l¾Ä>CÓ¹R \žu–˜Q¯Ê'2aøTŠ¢vÊҩ퀓û¿{mˆëÚC­ú­~³V÷ ‹ˆ21#fG&þr‘Ÿ¯÷# WXßç†Ä);¦#ºÏ^f®QsKt”Á=бuu’E^t)«)Ëq° ýÑÏwÈçáU!JŒí}¿_Ó†? qÓ¼$zOÇ£Á@èçÛzˆ奺ia4Ç_äÛLq³´XíïA÷Á[ÿcõ{j]m¦GãÖ§AaçéÖÓNå9G$üìßjO°Á¾prØ:½PUd  æÄ¡%ÆYÌj¿,ä]R3<ù1°f¡ÇÁÿŒ|{<°¤u2pV€É÷'-гîfp"q¨]¤k°ÇÍÛf,ÑŽà bެf êB±Q•®í²¿ÅŸ¼hz¼›ô<ôç5žÕsÙ\k NW½BbæÇÏÉãÿ䤋ùÛ!ª¡¼³«˜ÈË‹Y{u0V²®³Mù#Ç ”' i¸RùÐèlÿFÑ€•bär]P)´a›ééà/{!li°@¸Wäy3˜èx2> Q|—ŒÙ_K­‚õÞ3ÚX³y•`UÇ`¶þ@ŠA§³wOz[øåç6س¡å£æ”¨¬Õ¥bJOœæc­ í¶5ƒ:•¿íòñQ¬^ËÆPl"¹jçlHáDbæÕŒ‡Ÿ'EùäLë£]¨/H²q #:ÕODЬ5}§C‹@ §_µ`oz¼sù«|[bѾY׋4*'æªçK ÷’…Ÿb·DŸ”fÚfÊtžìàŒ5¼âÃ:+„‡ã$j¾—Ô±§¸z'¯p1Y$`Œ²#¾…O•‹•æv@”¸ÄYçÖÝâl[yñGröþ]O—ÃŽ¹Ä}|‡¹ÏÑ‘¦m¹µgª¯W~Ä­}ïði¼=Ié[ý¯æ¹„űdJèÅPAæÄÏ›¶äÆgÖ nßJ–GÞV ¯óáò§öw'»€‹FŒ3;•¦ýûX DRTEÛŘTÅ4æÍS-M«uV–£…&¯Èoí¦^:wµáSîùBŸ&F2¨Áþ-þV-²çp.îY=:ÆÕð‹®I‚õËö%añý ¿cÌÕÑv)—Úk^Èãì÷ΰ RîYç)Æ æ—Ž5SêÅ[dÒD»OÎJ¾œ ¼–á~©â;¥÷çÿ>Ê Ëe‰Ü+в Åb~JpýI`=< =²PF-¬.ÔH`ñ#ÆûwÔ d¶1^D·ÑÛ4$vf¦T²:éÕkk¡ŸU’6XnB^MDSNaŽ:€›éÝE¯ƒäMŸRë}¿š‰'=r„WÜr£V·=ÒŠúSbó]•ßT?þ¸\£'`Ûg=2ϬÓë•Õ³réyé C‚c‡çÉ'6pRÆŸ·ˆFîþ±%w¢Ô'ˆ8ìá}gZrá’\­û}e^•Ó½8ƒ¡ËU½“?”ϼBöãþE¿;+Y)j›[«|q®Œ·Ýë!Ká$€QÇ*g+™BX–'ÕV”åz!ô'¸"xë7€õÌëðŠß6ÜÙÎÎLZlÉØ†ØVeÔ'°öî*¹…uB¾¤ôкPZÂLmi8¹¶ÍOðNšI6¹%4W^^wöz áÍ´s$Éór \q;ªàÖÖ|"§WÚdƒÌ<-¶qéTú[ÐÝf£YKÛ–½.ã NnËa@1ì[zŒEX#Ñ' jVŠˆP•ž°5VA™‘ÑG+ÏNyÑ¡öÜCãBoY.%•¥êvÁ•yO¼ß„hRnf-ø¨X#özè˾PL®c„žb¾¼tÂ5ÖØ7™auÊã{kž®{úI*´5¨ÏJý|E.†È!l9Ìuµ"äÝc™P„ñ$–¬5Y,|xaUêée ¤Í‘jœIÂ!ú…¤’ÂÀEÈÏ´¡¯8ª¢^DU È£DcM»1Êó>Æœ…9v$W'êëdm2,Ý a®Çà] Ðo1£õ˜­¯­nÖ§2ô2c>w}ueÃðšò‚ýÅÓ(« `뫲k@žú9Œc½9û­ÿÜ—úöô-+hc®¤q‚r¤:3Ð2ÜŽÖ—!647÷/A^U¨Ê¥“¶Ìîùð"‘€Ø ;¢‡"*ÙÝQço]Q»ÅY᮪ºÞw˜ÿµO`SŒ²¦g [=ÔH.÷ÞFoP·Î¼qF+zþ½èTÚœƒ(ZØ(Maçœ6 W“Ym=ü¬àgûÿjäB•ñZËâÌEN½$\PÐB·­f”öË+«éÃO>—J”¡ŽQãc\ ¥6^¹%¡ÎÄîí)6UŽíXæßÁåñ6U/ 4T°8qÄéQ­˜½âG&î÷ÜÓ@ü&>Ü™OjÒ”w]Å 2Ãe3ºŠ9¹‘íjFagg•íq¶£ÍFíȸ@»­`&P]¯ŒËàÌý˜öœ¶Ž¡aüìÑøúÕ³\ vN®Äͨ‘ØhUGgK™x¾¨éKNˆDÒLÌØÏõŒˆ¸5Ú¾ªmôKŽGKO%6 'à\å\0r$ÞXpžw(³šF.‘ÑŸf$ôZ¡_äŒ1,ï XøçúìD„Åš/×oÓÂ2-º¯G4Þà ×6q¥K:øo@Õ×)ÒÝÙv‹ïÄF_9Øã ÉÍ£^` Ó},÷_+KÏWÆ‚Á,¢à ¸€R6¼¯?ÅÒèÂ÷›Éµ»tâÃ>à:ãÓaú¥<®PÏ葈7p&kUˆ[w:ªa™SNyÅ5•æÄ.O±L4¨å貉H»Ð(œâ£*ù[.Uß͇S»¥A¯c¬@íH„ fû¾…½OcÎýå¿Å8 ] Å$Ú±[¨@𙥣´ð pþáôÈ3uüš£$FRù»¦&äé_»á*2ÝôÜYˆÞÖÿ‡Ëçæ¥o3m¸Æß^Sà)•SÌF6…O°†#‘‹.mÑéòžúpWáÓ‹ºÍq,Þ^š‰ŸZ¹µ"QéÖ–jTœéù„ÎBãA…Â)+™_^èáŒw÷K:Ø¢f÷Ü+%Gd½m~&„óäNÖŒ¯T‡$ÝðˆÏžÛ%7ûÕ ûøb ›¼c­h“â#»8Âè²KÖÆ%õÓU†ŽdŸ4NGŠüv ÄVÑÝî*z¯˜kŒ“¯ÏÕP_ ‰é‡e5Š6ìñ|l!ýŸ³lj¾85I“ëå¼ ·}otªU›c¤$§ÿ œA ”šAÝÜ{!|`ðÿƨ¶ð:÷ÐâÇösY ©dÝcLi@0—ûôîÆÉÚUÕÎZ_:2¯u(’£¯eëBM‹h¶õZ»Læ×¡Ö‰?™,JPh‚w(ð-´ž#ùáï*¢¥|edëšç¼ËÃï+Î5´XlÈ1âM'tõ˜£ÎÔ§Ô„‰Bü09iQÿÈ^Ôä´ì®ûBÐ1™-“-ƒÎ•iÉÚ.oæ¿+MÐ<§”Ï)ù6…K—;Xt—?Ø}¥‚@éúÀ*©Ó!õ,¹„ ¾œnÉÿ2y}AÝû°iíeÐ8 X8('<¥žÛ×;ç+ÇG¸¯;.ûÈ⦳’s"£#ŧ™ƒÞ5ƒ1XøÇãØ2ɪ‰C[`éÛË'Éë3É Êý¬¿fÖš¿çÿújª ú_Tn? Tß7>ñÛ¬ùÞñ`Ãç„ÿŽ£y|ã»l£|(xÊ(‘w&DíËS¤b‚´Ä¶¶òpµ]aÈ6~‡>>þ¯gž¨©&Bd]}JÀÊ*>Þ±ü Ôcà²xÌ»=p’­W:Á¬­rnã£8p øNÕ¼:X>¿æò«ï-ÊZ¿ŸÌù*J¢nÊœ´!Ygô}P\z$ê«`É0WŽçÏþ$×QòzØyºÊ®ÈÀ-²|n1ÿ¢u8¸u{ªÇÌɪHíòšÔf)Ã~)u2Ø]*ËÅ,—RòJE4ö‡W23¬<«ÀžƒB3ŸPt9Ê*a( N?ª?U†éƒZÙáÓløš y=ì°ÇãkõÚà`21=M8r Îð7ªÏ2‡ŸºgæQáôÔìŠ)ütkϪ£Dw°,F=æUHæ7[)…ê~J’ÇÓåãÙ›>#>oñZ}.›7='ñüz=«k± §“¾¬>™w=!Ó lÞy7Ëœ8ÌvùU£X‘ÈYOI¨#sx9¡a³Ï«p SdG"òok¬NÐ+ÙMq¡jyC»½Ÿ¨°·§)h¨`¿°û¼Qï°cõÁEµ&h(FKIÃ~§. ß>œJØëxíæE’ÿ”w'»ƒÒ"p¤Ø¢booÝ!5«z51yÕ"|ê©n&¾¹%ðÝ»¸%O2óâØ dü"¢F„©‡-Dšvl|.”Ñ™/ü5"âi¦í*&(ù¢ú’¦?ÿ;Š*tl·jµk/è)uéá8þ÷F ôo®kН¡êHúgkxÎŬI¿éÄ´ÿ!™"{@ÒSp R]~ÐÇ—°EÉÝ/~n¼m!ã²³ªVX¶l¡š¥Þ €Á ü±04p¢†à‚êÑ8íF›ÂŸË&Â`=  nÊì8ßÇú½þٲȫœläÊø¡ BÅrñIbìÊòˆû¢M©Á{{!q+¾Ã¡ÈWcrérÊØ,ßgüåÁ×zÉ#ø…ãÖCXC ňuööë$ù”É~æ3¢§ìUü }ê)_b~ªÐ½b³¨p•þÓb÷O˜'(ÂÝÊ›²hR:?ü/Î[–Ï`,SþÜNmr2$rÞ+V_ÂyÙƒ›£û 8‰£¹!Ùøü¦Ž·˜Xsø8ïagû·i;6üXqxãG€‡;ô}Ïê—° ÌYzê,Ð)CL²'û¹rrd½þÁ©•‚¿Sæ(ÈÙ»Š‡¬M»ž(ýÏMè£N÷»’Íé3Pð×Tû9©[Kû3Îö•Šï½Åk(>ۣ샡ùGqá/¼½› ®Ý¦R~9‘ß*+¸Ô¯M*l¤ m£ð&¬çªÖ1ë;¯@}ÀŸìUKŸ–.›$ÞÀ ¼ß.Û9¡Óóš´Û僕|jrxR‡¥$¢Â€({Ò~ï„Ø`ãåî19S7<Ü¿ú»$¾](•†z eÜ»ŠF €’A^Y­«Ò™lD[¶œ/¿MKy]¼µ¸“`äå-¤÷´8´ ·m{H¡çMY‹*ºèpCÁ¹©ÿÔžzëµyÕdb¡Ý†‡X0›Eð TÛ¢ iú|ÛÄdëöåxUê†BS¤y©mƒÛ–<¯˜÷È´Þ*¥Ÿ›¾O&Ödí™M Ër£w2R›ÊÒ^‰Ì=^RÛ%à«‚ÅŒO¾thfFh†¯}CÖ¤,Ú±Ôáñ–ÐéÖÜÁ úI¾Óø4×$†¾Z buÖÍ & 9¾tkhF±p‡éï_uw*;OÔ,:=jO_aç4’Qœó5À¹V© iŽž¦9EFŠÝ—.ûˆ¯ÂâÁ tcŽ˜?ŠBláÕ§]ÉbûáÍÇ}'$!0ìÓLTÔü•ssm÷‚>4“µ†ašS‚ÿ6"€Xpô÷ÅâçRq|©;~zsä ;`Šnæ­4ØzûRË=†E7.JQ•ŸA4zxˆÌ2¯ñ)púÒçÐÚmüÇM—Ó:PE@d§Gt"c2 hÄªŠŸèµ(÷,#äo,FU…÷´ç Ç ×†¾âèÆtÈÀ@ÇvyÜÆÂXã4V÷æí3\žEy/)¿èi‘êÅœ]"ø¶«+ùlÓô…ý°¿œ¥Ø‡Öèf¹nÖ™C´DÍÛÛƒV^JºL‰îLÕ¥–çáwàu:Þ}YŠv¨|ž?ö ÈãZpü´5ÜÁ R’eLBÐào:;ãVV1ùÉÊeدõ\¶ ö1lú¶Û»dŽrY\†²Þ¯ð!ÛJ&™þÌOæ$O&XW7EåÒo±)w4ÃF.}ÏŒx¹ŒãÆ7úLb¦e*B˜fÌúžQ¤ ¶º‚(±gŸP!ÜÑs°¿Âeè!aÁ9ÏÉ\Ö¸B°]Õó?–íðÄ×S!I[ç倘噉XÎ [n:Ù¤MX£DðÙw‡^ùtÂo+àDËãJ´5x*†IÐ|œDcºŒP£H—FŸ•ËŒéZ=!k½†½rßP±-KX ã·x†¢ç»‘“5¾&;L6tÃ9:KU€p¾ëܺMÍ! øU>Óè‹]&бŒ¹<Êw‘&ãäݪÂ}ùó\2RÛ£lzڨȢK´Än;íÀ¾ÞÒå;Èã¿LÃl |?ò×] <ú½·Oå¹bõìSn&@VŸDùøYUkŠÃl÷ 6Q'ÑàUÞ0v<‹)–¹^ò¸ÏL^s;‰ßw—d|[Ý6ï2BÒ."U :±5०ҮÖ%DmÚ†N,š’YònÍJ‚Ø×àû%|è +U[t–¥[î_ΜÊ[µP:‹™vʽŒÓà¼ð‚ÌE|yd8Šš ƒÐ3®‘áv ½³àð…ìéåò,ì‘Â~@DÃpg†oCJÙCó;q-´,¬³ž@¼HÑtní;°VVÔ&;ÁpðwÆKϽ§•ÉÃ8žÝ11À OŠT]wðÿL—(qý¦µ¨©Õýç"¬p}vp„—$|kXÎ8³M=Á08î‰Y…\öþ“~r÷©(ô„nã5g©çÒ˜ ;Žgü2MçQÇàb¥yj¢êmbž- Öµjv$bD<lêNÒÅÏcožšÕBßõ#õi¬¡á[cH}m–ðrµÊ„nô{̨;RâYš O1åÍa¦C)å¯ÒŸáu%.´öVEE?©1™W޾!‚KKÝŒfùF|§°hJ—¹NÝ­DÓæ\"|W|ñVŠ9!ˆVÚMpªîÌÙAw¿vvð¹ »cºô ÒÞ½…±â¯Êû”bEûH±-É…«>2ýy ŸÝÌ4R±‘RrÑ]$ÂBfŠrUÁ™™~g9<å=©™dÔ~UÓðÉÐe X_7$›±$‡D²P¦ž4[j£‹&/­½wVW'ªiŽo,½ÒŽž&qÇébz3ÉEdú¦Ú¯_„ 4}—æ^=<èl³£W4»uß|¶«Ëœýß\œ¨úC WŽ;AOâÝä.Âb`,лfeôÜ@m¹Â ím†øºHøâ`×/¼°&Z3¥eökpépÁ°áÛ–w‚€øŽÒÈÄN*_ôƒø"²&o1zù8¶mœÞ*-0T±¡fžë´XªÅªì ³Õöû;¢ä7‰“ë fÌÄ„çªá%…y`UºŒ8)%”Ü`©¾_ι~kkñôUZö‰+ ÞíÐìÒ&L'lE¥¿|Û6¤&l ßq:uyþùú>äe­÷dþ?•’¾,vH¤)æSÜ‚¸‰Å^ ]Õ–{L©im°}†·¢P˜OæX­è…s¿V¥Õˆ½øïW~ÂÇA«}Y,b:B¼¹¡•&¡å2|t·v· m2rjTèCwØ ù—š­B«ÎÒåå]®e8ž8"Cõ_’Qò¨)b߫˜MÊ4ôny³ÍQY_Ï.â¨ÊÁVuXK¤æëh{¤½œxïi* _ž)Õ‘ÔDÓ"¬um×c t¥ÝNeù&’z/©Ž\§Ü0H{„µðè»;êγe?ëËÎ:\+:9þ kKxÕÓŽÆÑL ­±‰WÊ ¸ –óQ²æ0†W=Я‹¼S‚ó>¹Š‹’Þ‰ár­…‘ÍFÖi¶_ ³×oqeûÔ0n\懴¦ò¼ „@ýF¯A}è¸ÈËUŒ]Ö@2óæ…ÈYz•ÔÜ4s)ἑUµù iRœa©GoëYq¶¿ÝðJA©káÀº ,V´%?ÇBªÓ6ê9”¬Íò;Éùb¦ëìUì!¥øÅòåDscŽC…662#ŒÇ±ò~9©©“pb) Jty*Rbͼ­QàÏõIÁ4³-;§sg§v›DùÖÏqxP Ÿ™–yÆ%DL¶»±$Ï öoçót×ýæ4‹À¡U*œþ…Ø KŠr YvÙ¬Oº°û¥FatÇÞQ.Ä$Ÿï·v ¸Bn•´ÌZ•¥Y¨ÓDXnýÀÚómØo*<&_)*ªÞ¶ËèÎ}'½óK zKFæ/r]}VâcT$ëå‘Â&‚±µÆ¡ hÂÎ(¬õiþºÙªM3Àf=‡†_W, Ÿ~*“#/]JË]`«L‹åx$ƒYá 7E ÅaÇè¡>Ví–5P÷É(HŠ<“.׌öÁ7½Ñ[9#£d›wqï@ìk”6Zq‚”¯»_˜¥NpãwpÏIö‰_¸Ï>jùwÍDïÙA… ë%ú¸äl<ë0 ¾ lÁ7䫆ßñ 1±@O1Mä¨ù¢ùqZˆm V´ÓzX„¢ýüÃ@¶!÷¾Þ}]5SIoÖÇ»®S˜«N'7áÃNhH›ÍÇà{ê?½»*Ū¨àù9W-V‡4=®cÊ×·ovN•^j¯ SC/H54§V®‹<ó¹Ó¶FRŸú7ë?³&hl(þœ–K…„ë'„A÷õáOk Ÿ6ÜÏ)Ü÷à Ø¦œã2ÜüOðAîM Åß³ ÃЉ:¼ãš#9;L2X‡•üqaÉ/ü ís²• í˜äWÑkÚl¥ˆD×ëœàY"mE1kúU¸ëÉUaMFá@Áy½Ó!áA7Ù0ò#ÛVkîªâù=À?œáè\ƒgP´[QÿܦKÕnA[ŒÖSè}E&6ƒ|çžù/ÖØß²õŒ´•:òˆ÷î°­³ÔLË×fõ<›ÊkkÍëJr½“.ÛHÛÐÖ¼/;ò•d¾­¬*[’˜«aëq\Рš —548¾diÌ’ù;¨I^uMÀ0—}I -lÞ1àkênèeh"U4;dEÞâ.Bã¶fƒP´»ªÄFO¢N¯™Ëÿo‹åån7£é¨LôùÎk`’tÀ%k@Æ:ž <åerbñ÷ý„go¶Û Nç¿MÇrç´rÅ °ØÜWZØ–ÙìË Þ¿¬ú–Mà \H&’ÚÞ7™¹þ$j­º-ÒU[Î(y«zTð¶6£ø5Œñð±h!j¾üèØa®:Äœ+ vÓ®´¨›^ 'HZŠ-õm½{„ky!ß\­Õ*Ù.ú§H®N8š&¯*´Þþíý%_ÖóZS66 ÎÅ;{XÔ¾.‘äÿYH:Î…¢Ê. Ç¿l.ð¿=ã)V­êžv¸õ¬ZÒoƒˆòËxZ+tâÞ€,àH”‡QÕRÇt®Æ2>=øB/ExÜ,ÐŒ¡·ÜªgGc70ÙÊŒõDæ„Çú‘¡G<°TV¼‚ «ÆbI Ú3¹ÂQJèSQéYC¯\¿îpã:‡ðÇS,OÐ-ˆÏ)-"E­Ê{-½Çÿ¥Žº÷˜°1˜Jc]ú'Ž…·ëZQ8ü}ûùà8jÞÌë0µ\$s A–Õ>(µZ;ûàƒå€ûh¤ríb »ØØaˆÀâC˜>* UõÆ¿—%¸–™â§z%ödRôá…œªËó{C©¿¨°O”"rÝú¬Päbö5ióÙWR[ZNÝ&ÓÎÚÞt:ÂlLkÜ õ0©—šCt§w¼Òc°¦ŽOO~å¿Aõsø[ÐmlÂaWn¢ÆŠH½˜a²ê`V…B“7áØ•§Èp-GÃG¹!×’jl•+Ð@ðøÒø÷çFxI'ß’vD¤;üÄuëgW -n¸s¯˜BÙÿ©NR>"œëËîBM`—ËrÒÈSGÏuO•~ ¢ ˜ÁãsË[[úû‰!ž þ¨ƒÙ<¼›+ @Ëm§¥ÐIAÖŸW¶€<ž'ŠÞz8“ˆ¹Pñ*8áÎr˜eDB¦­UjÁÑøÇø€Û÷“«×Ö(Í|¬š/ ª7#ëzKyœº•÷!/f™x~‹ jþ] ¨ë†¦,чba'$)nÈxk¼Ø6ÚEÓã!Ç1W¢‘¸ j†&O/_z]ihyÇ^ÖÖ.ÅÀž3ÊZP‚Ž(ÂÒ„V@­F4e^QSÛ((Œ´ïhîšþ&µXf^`% ‰@s’ ßUe -« ró”èëXûá`úÒ¤9FXgÑê>Ö¯ùmn’J‰/N×¾g|êèï 8ï߯®–.âÞ OÜ]ˆþw‡RÿH#dq‘Øe¨0íYBå(´v•¹P± ‡™ôá¸m[8;"ú7ãP3ƒË¾Î:GÚ»q&¡À—~X¶±†ð±4<~msPs (’C¡ú'¥‡f§!ìp$†g2Þ 3'‚UœŸÜˆÇ^–8 ×NΆÀóýpÍU%péobNz‘Toè høÅ§ÍÁgÕoÃ.oÌØ_QMVqŸÍqq%îþtÁU½pÁdghæŒqê¬bÃà‡Ðëq±}¢oDÒª ¾Ýª“:Ï}ðè–½y÷ãÕ³¬Ñ.áIÀÝS`-ö¯ èåì_îË1¬_m¿´Å>/\wàä‚\c 7)h÷­³U¸íŽ2ÖÿBE¬ÿ$+?Òd³g~R õ×+~N\aLÒ °®"åÝvh<ÝxÝòBº¦tUIâ}o}ÔZ©&ˆú­S(£ÈDLoA³ ¯ÖŽwÄÝO¨aòXñF1”³TßUm"È;«•Ãc~>ÆRÚ8®¹¦ÞHÔTÒs7ý/¹XUæóñme§¥x˜”œjî3Éþð«(ƒ„ãò dóh÷•hMÞY'9£íÉ]dÒ²ì¹Òäû,ºz- OñB¢_F°Õ–2°=2éC’™Ž%ñN&¥¬Ý«Pã®Óœ¸€¿&E.j¸iW¿¥©Âʯ–mJÊå¼ä´ûâü ánaYÔÓ'?½š={^Í*(d2‹íqð:C±šzÝ> 'îDy»`¾_•|úWvòת†¶<±½»Ú¡Ô 1öIþþ l!¯žª×i¦íÊ»b2ËÁú|Hy1Ý1 `1½ùgÞf—Ù¤çŽä(~ùÒOû¬ð¢´^ˆ}oÝU‰OÊ>µ¶ÆÒtu1kÓªý¬ã3P¥²µ4ìàjÁM¸ÚÜ‹s+¦Á2à_«ësGfuùÞŽ¯öì¤'À_å¬|A¦³'¨a®#XèòŽt®úÜ=ÚŽøp+Å£û†®’Ú„vñû_ƒ[vtl «6çZ ²û­­THy4`›”´Ì£Ðò· ÆF`¨*{Ù18TEG:Ýžhä<ú¤€ñ!1AZµ6¨OßúêJ<â£=󌯑¤"–äæñ€¿ÿ|Ó´ÖDV¬ Æ(äû­«b‚rz¥­‚´AH(ÀZÆÉÜ ’ÌEI÷©ŸuÈÐDÝÂI¿.òÝôãè#–ñKyî­þ}ªž@×Ê®})¯Ú]ߣMÁFÝx¶>ÞùQ-âÿÕf¸u6»Ep\t—“—+öõµ™27Gð7sºÏ¯3ð°óÐC6“e*ñ° t`SÇ”7õ•|ÖÒúâ—˯ZÊ‹³Â0­\—1TÚ‘Ú—˜LCxô̼>ïå½þC¢H4}fÚq ¡%«éIù¼•Nžj§TQMÜ1`@.#„D@u•…,¤Šy´g¬Î¡9>}[T— Û…kßðÚF›©Á>߃ÅY„Ï#Ö†oÁoåuÆfîºLj JnNz8ɾ«<ú3¾ÒÅ/—x³i˜“v½íÉ@´,õ¥r¼¨x•l!ÐI´."2Ã7éó°Ç]°f¥)ŽfEÞÆsË«çl"hS/C™V€êeŸ×r‚bF+¹œÅûÊæŒÛãÂ8ý¾À)A’nöQ‡„ô§ýìÙkü\º|yæžÚ,Jµ‡bŠý´£PZ…Ü4E¢× ¢sZÈ­è­E‰dèyH’ b­±Î^Å/=!¿×‹«òë&¸5†\íß"5 ñA.C4–r¬d@'»hDpýt°ºJÞêV…šë’oERºûGªy£–@j}Ÿ™E¼ÓG-»òô¶"y‡€e^øh7(Z&lƒ5 ‘Cf¶r|žõœ3@ŒKqggö:±S´_6€smà%Wüp´‰ s^±ØD’Yz_ìÇ×HsØ‚?ÃTžyŠs«aʺü*ÅPöHmê²¹§Çò§×J3ɳ^K ÏôcÐYíÍ…ÝmèÝ¥lãù÷¨‹U‘$c4v"ÈîäS tàŠ8ho@ðý£´¢"¾‹üiBâÞø1éçp)l8åGRŠJ˜ÌýÚÒ†QÎH|_«“»b¨ÖþTË~ú»ÐÙʹ–О@Nœ~ŽñÆÆ(Φîú í‹áEÆ )i' [.Å ô5Q¢ãtà·›¢ÑŸHÈÉJ=èΗâÄ€Õß"²€.Y&9AôQ˜¾(“ ž¶Ê®>Ï1v…œ­_ç ­Š¿ë6.–ÍyH"^!X0(sóǪàpyÄa¶Ô) ¨Æ ze? 'è´a ”(ë> p¡óóÝ z!šgU!ÀÅûÙ3ÛUãݹÝâÈKÝ镺ê:_~$pR}ZÚLöÔúª)ÇÖÂErL·ÈûIKX;NžÊ8ÿ·l¡·Êîu¥ÛUâWfSI;àÂ2öÎöPaæÂ/aZU¬Ø¥À$c…èÕ󊜕Þö´zÒ_`$¥:GÈm•õ§ X-Ärd_› õ˜¼ÁUiYs}­3/ŸÊ¸çUÑGpÍìzÆÒ£Ê\iµØüNýVÓˆxBP¿Jp×ÙnÏè–ÖØ/8Ôx…"*r·¬H®%™ ŽªdæRÏy—ÿHMuér'7·|h§Ù1iÃr5ññá#I5|61_Ž wØòñgÕFº)óOÀÄA±°aàgs½pǵކð:ÛfŽ#‹­9I†ïUß‘òÕŠÿÖšš÷çñ¸î¤ìeù=T4x#B€wÆXý¦ Jš½X«òí±=f|ã“P„åqùe¬ƒ–R°*qVòº‡+’,vùÀœ›‘)ßfðdJ=ŽÕîóÍィñfc/¤×«‰;|×ߺ’c þ½ŠåÃâj iäƒ&«‹yH7í$¼~¡ooƒðêÏ]AÆ?: Я<>ƒš3„MzÌqó”-ðå'eé Õ©ójÙÃÐŒlz±IsžÌÒJг<è¬]Wƒ}Á8kß+æ[u¥Ð¨±µìTÙ’„bdŒ@?ÍŽ¿ï ªð‚7¨@¿×kÂ:›:ÚH4-];ìWð¡Ý1’’ç£7½¿ŠÙ ž( )é L‰ÓrëÖß2TœÍêR^ÕÔxœ®a/„¦‰2V_2ŠÐÖ<¨î¥›ñ—žY웋ÆÿáqÛzvø Ï¸$S»g) `Üð4}¯Äu‡ZÔ-F61MPAuEéV•¦ 7¶¾U_K‹˜ ëøŽœoáè½4”•ÀnKq¸¿„<2;š(Wãk´ôuÅ;'‹©8Œ' ôxÒ“xùßòÈ.¥üK4^yÝŽMµ%Ǖ⋋N mJQoÎ…{)ÇäëÇùë,•³ºµ¹)h ö2µð›(÷ÌÍ« Íiø=b IÅÙtYZ®m@‚Ä1÷Á *œµ':΀j?2.âM§™íóx ¹ÆÕx†ZÙç¾A–?©?(q¼hÔE‹Öjçàa§P»GËjoMúnÜ£øúÞ¬à8­F-[ê©A7ÔÜDÒµŸ¯7Ž‘_ª,)ëþ=88vÈþz*k[Þs Ï CQ÷–zʰö+®7i/f°¹¦è­B¶Åü¶>t$$5̽&õ$‰Îæjw¬!£»Â5Æ»X¸K‰ËÕ_õ“ƒ‹J™mte…¤µñÊ>cÁ”ÅNâ÷WC¿¬çs9NÀÏ„÷ùÅ—«Šƒ‰%) c4yëȘÅêiÉ >áf€Á ³í¡_½ß4ŽxxXJ«ðÿKÍþœÆ¹;hêc2nÌ`åêi.Ä$&ÓLàôëGü äo,#ÖR¡ÔÇòt­PQ“0 ŸðDHîºÃê=–ÌR$~Žëu†×ÉÔSŽÜPG 8P Sêúˆ[Þ *úÉÊ÷Ô’j=a{sDÛàá•I\-u̓LYPXGÁ3Œ°¢[ö[(Á'Œ¯‘õìHu×À/=%9ˆoÿ`ÎQÝh{’‡UZ¬V´Ó0©`ˆˆ+„Ëû¿¥ýèÍ)˜—¸ÅZt˜¿ÕŠüRVÌ{£9«Ö«0“øŸ i[L@Øp,_¨XN½\—˜©@kH€¶Uyúµ]fˆw®)ŸýåLW¬PÕÇØüM ÓågÅRçÌj½'_JÙ×Û6õR_ùôÌm [àâ¡óŠ“Ýøª-Ôoæñ¿ie3 ÃÎ!xs¬³KéÝ¿õæ:I©¾œi꓆NÉSîÏ)„jøà1–|•ޝ3›„¹jFÓiÅ´ÔÓW+A;YzÖ¡Ûòýêž/fbä7öXji=w"ÕÔŠ³÷y­«yQsñ-jãá^ì_ÛU&åLˆëŠmý? …xº3ë2Î<Ny£8b2u½i¶ž иRÔ¦âô;(÷D·ÁÏØ ¨'æ_ÍîtoÔ× °izû«]dÃLµOç濾X£G&ÐóL;>õùPi?+J…¼²öo+ërÚbÐHhNpÕbwÅÞ/&M =ÒèsUIpX*êòBÑoÕmŸ¦¬ÿp.¦×X ]`™/j.L…ÛÓ¼øåñ]ÔÅQ«éê4‰ÉÌ»g®av$sÅš_&oj§­¯í*#†m›ZIT!yx`T$iõx¾Z)ûèì?êÙD< Ôd.™ü@tàðŽË'&¦ÐC›$>@,Ü»eÌùe« ^Øo¯L&¹ùF{Ói¯!ò§ <´½›“¹Æ/.>¾+*V‘9ðÅZHù¿œhEÜ ³r)¼Ø@nMS+ìGŸÙtV5Y5XàÇ,#œ߈9XÊ/ø‡ÃšÑï¢!hçëÉ(–¶à;à$ê H¹¾ $+ãi¿WÔýoUYuìéIôÎ?k‘ã¨Õ‡­.¸ØŠ "bC½óFL=Š˜³”>—©R ÐÁ15Ï;RŸ'R1Ö¼¸>u,yõíìäsɤ{cT°÷Ø`^åÒÉgo‡Å¬òë¶àŸ…S‚$áfIQÄh“™Ø­¿»Ádƒ\¬A!££hÄYÊ/›S¬Næµ½„Xùó½\,{cJ£4·¬5ÊÙdý9B϶çЉ•¬ùÄs»†D&cwpùw'#ùÑñÕlº2¢'zkf|ÓÏ„å¢Ág‹õ3Ž, ¸©wè¶#ñ[pÅ.h”kŸ/tbþ4KJ5p¯ æM:èYˆtß…²vÅ4í÷ªû–rßZÄŒ™Wnþ¨ÓÁµHu( 7w ™¸YX¥4 ¢Ì6”¡½…×[Lv­ûJpW:öÝâfµf·A!Kz·QÌðDþ¼ÁÛFÆ2ÄÙ ÿÙ¶æzPq ÁWÄÑÚˆÒy§¿þã Ò¦OØÕ)ýèÝ$%Q=Ä·Ë:›É[ðër¸.¾"FzxXiE.íéÜï ù½â…ê`7*=èApyR‚(¼Ö]êc˜éÈt€¸ð3y›÷0¥8ge}Å&0’ssËŒgUØ—ø¢P e²N’uTZV¾ž!‡Õ>@‹eõäAÑ —i‹V—µÔ žsÍœÚÔIÕ\±Pù…µ(‰<£e6‡HðŸQ`'©áÇ$ñÎ{Yz.èÏbm(GY¶! vD†ÆÄßc%8|%ePÀBýÕàCnÑ:ìê€ýgNy¸3bqäÈÉìs=ô6‡€TA‚¸ó”èü6 ,ÆÅŒ˜¤y’–òÿXJmýPFzÂsgÚz–—Íú!ÊÜÏky×cE~‰5Y{;$Ÿ¼öO*ㆻó(~SüS¹ô•¥ÙŽ ìTNrV±«3xdƒ ‹Æ\\ú6Ôòé…XuŠÊ\ca>ˆÕ-þ¡¬ÈjžÄ6ªh";”â@øqå©fSK2],ˆÄ… ñÿð–Uñ#…2}ç0õOŠ—ü éQß/u ’”ÜÓÕÝ\- Wôø«Ÿ$RÀÖ.ëñ¨î¬(/ÕÐ5fˆu/\x“Û™¥UtYˆP›kÀG‡º–·ºÂ(Pï2-ü˜Ld… ÝJ…°‹a/’¢-<†Þ¿3^‡’¥‚—RM©ÜOVAdVETÀ‹wË`xdkdZ&×4ù™÷Þ @“ýýFÑ”C˜z§ÎÖu©3kíç–q búu[‚ËÐË{‰½Tãëâ}R‚(§ÜD_ ž$¿Ù…Vw¤uø«¸ŠúŠI²Æ·KFFŠdÛ‘À 6UÎ%oÝ(GScV¿å='…åEÌk\EÉTKpÚjˆõ$íïÜ.Šb2sC›Ft׌ÿŽ)·²7h}D;p¡É›¶ó„ÿ²F²Û,+Žhë>ð‹àÝÞo—Mæpä-—zD¨¿…·i6B<Ê" ÛÏÐÑ8_ÁÜfRé _íRùC;ÉTŸ3‘Y.y/õƒ·P P02~w{Þ"ú]ß:¡Oµö¶} ҔϱÁYä8’æƒZ èˆeof¯ÅÖŸŠ˜»u$VXl¼ÒZÒÙµ ¥Ñ§œÂÑ÷ªÀý¸«Z¼ð÷¯4µOsåþ:e ÍÍe«©þ&ÌoÚÂH):ç¹y— ÑWrþUŒÙOåU¬ˆ”¶\ÊŠiÞÀèöCò´uzq©‹áæn‡DãÛ~ù=÷'û±©nU¾Z â.Ærñ`1MÄ­k~ó6G()Ö*lœ˜§#…(hÇ`·¾O‚ñ${p57B)§½A¿¬7ÅÁ˜ü›L §°ÚGAõpº%Í/»,¦ÍI¶\ØgB‚ÑÐ?ÍÕŒÁÊþ¼]9äR„š®ÍNP T$® `ü¿Œ‘bÝ6î¨ûCÔ8Å7ܹE…¯«| @OÆg’ÊÚZZ$^Âh9`@MC{$ =’«Ã_ùôrŠç<ð_oÂô6>ÙÏ!M(äË\‹7ç1ïU½îOý,÷Ùˆ_6éþF*rÅ1¬Œ»²-¾Î»îH"­8šh„Þ„8û‹$9_Þ«qRY 3ƒÿIÓß÷Ej ÁuÈ´8“ž?výç3õÙ¶·Ý¦_ûôÀÙGA"%TÞÙµ–(‚²^b˜]&(¡õ ®šïqÃé‰ð~[§¸¶džlr æíà›{»îMŠlªÑRãC}j`ŒÇ5Æ™ŠºVˈ“¥Aû*¾}4÷¶Ï ½®›¬†²zƒhõU„å¸)]—[›8nï°öÓòºOèÝýr“‡UIQ“†áaž4{ÔŠWôi\ß"»$Nå4HMü$.7J”ÖÂÔ¿frÝÍØƒ£¡`„õð#_®Z#ªúפ5Bÿ>vµ4~_)ijˆƒowûÃL1gnmŸôO»©($±k,X¡k!âb¿M4§\û…i¸ ôgL§Ÿœ/9æ1ï‡ÿ½ì„{Àvâ`W’h=xÀ_“Ö¼36Mf8,ÓFª ýÙÞ[Á‹¯¬Qð°d3K\9v¥x|cÅî°QÆÇ_õ/p7#Y{H7Ðbi‹ù¨Ý5Ì1óÎ⛈‘³¥gõ染@ª*õ¬´<ªr4÷ïÉj“æ “ýìÞ’0–u¾Q9aœ òÒðñüƒVµURÿp÷G]«ÍÛx 1§¾¦‘JìPiN?Õ «äqCmè²|Ñ5ZÁàðižTƒ/˜cދ³Ðu0Pžá _•`3cÀEë }(KZÄxJe¢0ìÚ[œzCÔ–º±íë‚Um*ãIu"ž( €Sïiå] £–Lw³®M†3Ú­«‡K!‘Êg!_«U0bg)žbC׋2S¬„*Ãüбe8ÛxÝN.||^bJ­¥ Åhñ'@o;“ÝžiúïÍ |ñ®/KþŽtÚ’ÌŸÖÞƒu¬©÷‰®wC®‘¹e†·âÿ‡˜W ŠH#û 5«7N‰5°yEá'ƒ®#(ñlË}Ð BsßGkid¶:¤ Îj !ªJ+ipÑT¾ìº{ÁKy«5bÊN|Y­fú(åožgM'Ái]ð™4\«]Ì¢9«éò˜nOt@ƒöC¯ËI.ôjj@‹HvƒC»w£€W&­úðPÞÝ,–å×Pç^ϼÖÒZžx™²Â*Ϙ]'EFÄÌÔRvà§ÎM€•õ}š[Ãõ¬¯€ >ˆë`µñ¾Í±e¥8?!’mIš6Žó9~b,ÇÄ`) *Jª|õ³ýý¤Pí^ áN ñ£œ‚a« ÞòÿeãmáŒïuÐÇ=ETÜ«bñ§eœÌu¿‹Þÿý#KkÃׄHòMq?±IÓ Ú·¯€F‚2âJ%üð“êyAò$ìr,/©›HŽ5¼×ÃÉY+† §ýHÛë1p™%ÒúÔ^÷ƒ°n;Våøõü%<:ð»t–[,é.¢6B®Kéɧñr(Š þoÆÈ2¿|Ç‚!;Ž7±ü;whí¥,2ÝÃqeý¥T1«š.0ÅYœWå).ÿ 9W|½ žË¨—&'F–µlxÓóàÓó·%²ÒMN_·eRC¹"6ãª{%¥§vÿx€ªN»ãPülõ(õÏӂ†;Õ‚[†¨Ô¿ƒ<ÄÆ@°ây36û½ñ«]³ç9‰W›ôRtÙíC•Ò‡„OŸPq[h£‰{C‘¢[Ûºä;b<).‘bUÝ c DB¡i§ @cSfŒ``P"´A²òŸêG=J~ÍJÛESDœ©ž  ËÝÙ@ICs¨!òu phȳХÙüÆè›­/ÉØ'U#oDO—cŸŠå 2}cÄ_¢´é|Ž}zè°ß‰kÔ9ºj^à4‹®HðJF«;‡“š´2š¦A¼QHZP¨¡-0êõâ;rvHçåßa’D}èlX¶Úi¹=#Nœ¹e§º€ÕÓ©¸!å:†´¯,{ W>ÖXäF\â)œ_iN9KåȹaUóüªqx73”å_>:$q.Ÿ˜àÇÌ’ Xi»¸~zy3­Þyt”ÎMñIJîÔ÷VÜl Ø/6¥ê¦jçѯbJøÁ¶ykø$ÆK©Õ£ü£ê)¢‡ÊOÎÚ¸×ô•¸ë³ª›òëCZ!ØÅœêÀøÇƘbáOk̬\°}è¦ TÐ@~»>¨î•ƒc&úádl^ßAìöOñb?ÔoóÏl'ßÆÇˆÂç`˜ êJiü§Á…pÕ²¨Ï¿"ÈI† 5œî{u+ÓH:°¹pË4`¡I¡s;‰gCí§ðbÖô–€Qc\b«×W}s_Iº±™yR¤%ÜýÞâýš ¸l±ô)ü´¢ÊÐvÊ_øÉê#ß¶?Â5Æ+VÓ׬ –ÑùEh÷úU¿¸-AB”çša&RŸÖcT÷’dj²"<»>e%h´ÜÉãU&XééѾˆÀ‹è“æ¸@±Î€ôIË"âÍK8Ø7½Ã52ªÉAÎÍ·uóM˜§#ÊÓ­´BÄ{ÁgªÒXë//™˜ƒOhŠéü±?1’0ËuTk$„€®éÀK㧇£i¾Þ¹0´m’ò0¦ò/Uèñc÷QWðlI ¥NÅAýdçJ‹¨G&Õµ²¥úWÏšó®Ÿk_øÕkîÀ¡q’AVRjÀOTãi c±î¶D«} Ž$ÄtÚ‚—Ï·À¸D2P¼x*}¢Ï½ l qr´¦ÌKÌEåD÷Ú½ãEi“¶­ên+dÓ;ljJ>ªL’k™JåŒ?“aOärÌvç¬øiuu~Q‚°Þùr ù¸×b,¢ÝçðßÚèY0Ò·ÁÔëÃ˽‚|¤:—q+Ge1ñE]5 æ2ÓK1 bš=…ø^Þ0l8‚ ØÇõbKw‡„·‚ªN†ý$îá„vÇCšÌ\D_ ²…º÷_P)P6Õ—œ¿Ì&”)pyUUiÎ’_w;(ª±Ú­Èûi|¨Tµ@¸õ,E)ÞÐlQØ›mï ÝàÖ Ëý‡þáû$tá%èÉ’ÀÝ̵9òÎXÏÑW-,ð cñÉ TxvÜnÆæ:ÔAŸ@Ÿ¿d=‡,ZU ÀG$ôµ]˜b=é©ÍoèÞþþfŒà üTñR\œKË“¦j˜Ý†;‹•º ÝNÁ^žR-kg$ˆ—GæX±=›=¨«à»Šè­ä¿ à‡lí}œmT`K"ß#×6@:J{÷Ï—±-šy«¾Ä ,]TAÁØc×sŠlÅËÌLÑOÈîsLT0c<ƒú½ÈÖ¾§²2-ì—ä˜m¾Èpþ±[hô'÷{¼sî]§°âMê,AV9-›ùßã¯óçžÕÂWV§Ö­Eô“±°O[QÄì¯K…*góĬW5…V½ùÞ’âCVTr–b9\@£†·zÙâ[ìXO+Mœ3{`¥?š’Qýú¼ê%þÚíl;&¿61˜]å@¢â, ÊkYäæ`°€K£ãEö¶Ç%Q8ä¼`ñ¹wÔgβʈÆ"móFrMNšZ$i{Ó-G3õûÎs£³¯!o·ÊÿÕ@È˺,¼ë/h[æ€bš0äpÿä´qd–ó¶±Q·Ì¿àÅîêÖð)ìm‹WXÛ¦xáº)îFú™<¤?T¨Q¦Ðº\ìæ)AI¹ëÞ¹b¦Ù'ˆ,öLípK:Rh‡»Ý)%0áMˆgØ»ØqÜ´Ôìs£~&æmÇe(Ìo‡Á×Ê~3I]å0:)B!]LräŸ\§\ªu… ñÈÉ l¸"rtJcèÀM‡g‰ì*èh(e6Ù¸Ž!y™[/gZò³¡Ò¾d#q'¥D`ù¯…ެq`b 'p–üÌ»HívƒD”Œ0*&m.ËÖõ?¾Ñ#a6—Yêd åÊ?ΠïÁ2®)Öp¿ 2A’B»OE²wü^eˆÉè) œ9‡®ÐÞ’Ïp¶ÌâZgõÁ¦è·J‡ü¼íò©ôMýNÆ4ÿ¡—Ä;èÍDºìO‘ZÔæN4ŒtE†}©Ž[£ê ~Ôäǵµ?Ûü A§Á¾Pþë¾üÀñÀ í!#ˆÅˆ4ûaÀIŒ¶:Å‘ %/% HdIª ðÐä+ÚYJsƒf÷b˜rÈkâ(Ûà6ÖSœëûÎ;úM94 ARç1A]gXøJÇ&töT¢ãí«Ù{wd[;M0ÆîÞçÊÄb¸"¬³¶Ý¹3­.‰/ xGFYUž[*1Så€K bŽ ·¡÷õl†…¡¡GAZ~ÛlZF$¬êo$Ú¬eøÕ‚ ¯Qï]xBèF'¸fv6$æAþYÙVÿ4Ûðžnw‚Ïš‹( «ïm´aêpR8¸ÕþÚôóÖ)L”ƒwBýÈ1oÛ4ÅiÍVàPnTª9o¥ìš9r±¤,mYˆjªe Þì>]¾˜N(Q_jþ5,‡þR<ù¤»íáÜZ ˜†s¤]Ô É‚üú.ŽšOÖ‰>ñž²+–¸]ã%ÄÜ‹òD‚DØÂÝ-\y…¾&?;,®p=­r0yû&«ôùt"1™óÛµ÷m3ÛȨҧÂñŽš7ç´ÂÐÆKJ8A.{ˆN™±¦Ç¾…À4„*è¶|¾Ü|/þ_gI3"F^“]Žcâ„ÙÙ‚‰Úe\T vA4¼›rse—!Öй©–ó’ÛúXÿr2ì_oÜ ¸ ecíC r1m§õ/Eð=×çh¢‹àÔOeÀ¤XD@<þ³U²›v«ìfæ£âºã (˜Ç¬ ;5þ[Qè¢UÑ!ˆaë^Óñ'¸ãN£p_yú~ÀÉη”  †+«XcŽ÷“v›U5zÔiÀ²¡v½ÎÊâY@ÓÃØãg¶ÎC^ ªpÛá^âJΕ,¹¹NÎ:þã¹<'|@àÔýÝNQÒÑó(øÚ²˜…zž|!°óªÅ½(‚´M 3¤;ʳ@»rVø¿„•¦þ¾Á;õMÏ6¸íâv¿¸7¢ãÁŒ0+\ÃuÓ-®eŠJNo[:V&6fh—•éñíòE_B€Ï'9•ÉÊH˜»QòMÔu ªN¯¯k£ÅÞ„¦ªºNü“‘cà–‚Ée¨q?-U­¶vP ‰_’íz\,C´TÍ÷ÇŒ†ñD=Àø ô@|ÙÒÌzh_U‰}aÓ¤+Ý¿£=ŸM]oýzZŽÐÐf­q# âÖ£“d%>4¹6OµÝjÎñ¡ôX•üg·ÀÊF¿áWëëaüŸO+ÕíÖòyªŠ *Ö„¨-¿œø!3½+k{Ò|E”)î¾.JíÂg‘FD™Ÿ¿ªäï‘1xàŽ„HcýòÍØÿ§¥ÿ~L6ÂHÏ»Ç9I¦F²5j[£_Ä.ÊíZG]Z¾ ¸>G,)HOj‡ŠG>zmNtË× ¢¬éfúŸQyЫw™³¾DÍMᎢ„åª2ˆ9ÇRãm%ñÍZ®~5λƒkõ(£mgËÔ¿ÑœPã©AK'ÍÊšU1I˾U½¾·3á»Î4½kQõW¢blSoè (:FRÄykœ[ ƻնî»!É1=<\ ¶Ö­ŸîårwÌ(J™ÖÔ^öܳÆ64y&„žÄÙWVȤGIýˆqS\ÆdvµÃ)ÇU !¿‡ûàÄü«WAü,éwä(=~ÛwðÚê}ÎÁ·<øJðÉì!>ƒ¿UÇ€ô?ÿ“ê1J p¿ÞÇßi#>¸´!?;ˆº”3B—ñ:‹\èÚ£À<öãëò‡¤ÓÅ^4Û]¬È­u\ÐOïÍ|—Ž'_¯<¬ 7Éuˆ+#oaªÿ‰}R©Ý°„'î+ê{åôJ0ŒÙ‰êwD<ƒ|]L=¦$îÊŠ ?Únyš°×XÚKCg\ŒL2“.OÂRù´§2ïL6ï>„ÌÞøíþdðè¹=<&|ûØSÄéÆxê7Ýùºš2Q½‡ð¼o\…î x¢rúb+Åù þ9ØSƒ“¼gªe¥Ð|#) QpŽØÛDA&4’®©!%™bVòùˆÄ4XA[ùˆ_?Àð€ù²§µÅÇ‚±ÜH£Öý¢b N*n6zÃòæ‚–uå ð•"/ënBi=t7+=°b´ßP·ºfð_-Œ–çÖÏÅ¢¼£*7órÜ;9lùZi’AÅÕ<¥2}3°)@Š ÂÇ£Ž§¶§èÜ\P±C›ꈀsætYWAÎ>Ý,°+!g- ™¹Œ>¦F¯“‚žÂàX›&‰Ý„ÿ*:H¾•÷Kˆ\Å›WŸ_þZý¡Cr ìË/´‘Ï]Léö 5»:ú ô zýdž`®7ÖrÆ“mBüß ïn‡˜¸ºÉ\\»0QiF4É}Ò˜Ðlê"}K¾÷§ìD™£íyà>`lüë73—ã}¯Ø†ê:»¿?ç7?_E¥w„NÕ^f™`¼{=!£L £Àç7p¸çKw¤é†íó×xÌUà]ƒr/¯HóÄê$ó¤ê`€“u7ØŽh& 9wUUù­}ÇÍÓöÓÒ2Ë:¶ðé"€çhÅžä!4‡”¾í¢ñ‘«oÏ.åtÿ¾Ã@woÄ™jû”*1$‚†Â¨ÍH wp¡"9/Ð.€T(RïM޸ы€?µh}[$çæV Ò£u=Â:aMmyºw;§=D1ã­íFô#È®\hrC|ƒ¥ª´•2y¼ À_ü¹—×ʨÈatl„³·àéšñ©÷ÆoÍCˆôk>,~Š· Ÿbî‹|Âi¼u¾&1Ûº[ÃN”œNe0óQ–›‚°ÇŒ¼{¾\ÃÏ¡>¹‡Í*ٵט×(ÒµTœIuè8ˆÍ0+®±ÓùX³º* ö_‘Œ5_ÍÀ’]}[0«d&-J?ä–ò>CÕé±ÅtjÈ<ÄjØš’üÂV5‚±ÇutoF`P€·$|œvœ¢*d'%èz8IQÅ\¹Â2ð ¹áåúªD4ÀÉ^»Fëµ¥¹¥éÑœ`ʺ§þv0$üdîä”ÄA#˜ï›—çÐ/Q‰ˆÅ {` iû¥7Å„ß …ïaŽÒ‡Kea¤Q[T§±±¼PÕ27A‹Ÿaý4å·¦…;űÏÛÔäÛžF¶·LFŸÓ+¯O(ñù8³xrÌ!ó{Å+²Ñ°ø×lÏÎäIwÏïù€|d‘sR À}~ZXwÉþs¾TIÈå8#W‰îIÙýû ëê?гþéH.pX®öí {"F0¾çÍj:(¤Ÿ XävTêÎõÐ&§„¶ôÃ|Ü‚gÌ|{ìäÆu2¦¸C ’N»4GþP¼ƒ5ÅAA(,ú‚š²„Õ#G¹…¸Áñ§×Ðθ%p`×;òÍkûÁ“W—Ùg8_m.ÏèÂË,¹.ÛëŒìøíȲ‚µœÚ“Þ8T½Ê|÷P!(¥U¦}Á&´áÆpŠ|8¥1~E‹æc½`¼6¢4"2äVà-ŸyÍebçdøèÕÓÁsÊÉû0qË\÷ +Þˆã@jO"‚ÖÂH—üŠŽ÷ƒCÌôNËt œ„¢Î €A5Á‰Fwöû¼½v`@3šõþ–Fâ}–õ>¾~†Î[<~—þ:Ewc#—›ÂØ] 'ñæ•~bü„<‘دUÃCb51,令®ÆpÕˆÄôeÚDŠ6“Ìxwˆÿh}0(Y¶ï/H1\J°S*êqºžrÛµBiêqàfêeäˆô•Ò¯,ÖàßÊ8‡”!íÙ£(Î ñæÓ)YQÚ£÷h7)¹ …×£ÚPZMaR­›){“VüÓçd1i?{RÕ)¹™z>'- ˜þ`òW¤Ü‹m 2²OÓNÌ+æ!%… ¢µ&Ýl"ÓÚx5Ù'ZCÌùdâjâJáfÄX©DoŠ“ï8Ó7EjП“£ÁMŠ¢a ´”ÙóÛîQùîû±­ÿ™å-äSgýÒÕêä릜¯Ç£õçõƒ1°ÔÓèlZòT¥‡Šû*îÍp!b¯)ó8m„-àšoῨ^ô -ÖI6ÆÑ“‹šÀˆ=Î× ƒãý¯ÀE«‡ÔYãa€_öÊMƒÏm躿 ¬üŽV»Å}ÕWZv!$€äÊVΖMüî_.à‡ÖRÖß/]Š·ù3ɪÿ€ç$æ@£·¯ÑuZ[¸¢(–ˆ\䨈DøU?yOsˆûpcy¯pS›Ë€¹Õ÷º/Ñft׿ñ*ò:ù•‰šAgÞEU*rêЦÊ‚H½„ÅõÜÆÐ㽃‰áW'çqküÔ&¿9°>µm»`SPòó8pN%2K`öÑB÷„  CÔhÌĤ‘‰ ÓzÖLnZ1oZ‚'QéìTuP»‹w>{ļôdš:¹êÊ1ØÔÀ(òùÑÃÈ…E+Pd¥lª2q†lKMũ¹¡>0‰C©©—}±›ÀÄê(AealSC7[ Ø-?"ýŸº` ÊÍçcVŒ]EE„åʼn'|z\eÞoY_vSSY„¶zFÐcܪ«ÈÈ(eÒ~RôÖ»¦•‡?6 hC¦5µ¢”ïk5ªHõ¯ÅO51"úá› îI±É“+áM°mc—¹¸+˜2e-—D ¹|©?õ<à]Vo2åF~jPƒì^êmøöºTHC¯©ѰÔh$IÞ/1ŒؾìD]8§Fvözõ«rµš é‰ÉðMBB²¹î ßá2CEH—Ã/ê)–?‹R¾àvöÉ·8þ‰ÈQ5Ç É9'ú† öåä8«I72 ±%›Å©­ñ,W­Ü^£æâòa¾ ÍØÑGŠ qš"] ]€çÊt´ˆÑ<é÷ýéh$¶Çò7“9:%¡ãÖjý7ÌŸåµâ½~Ýü3îz`™Žt62Ï«»Ëb?KáØq;!Ë_ãÔÀqpÙÑO¾”þ˜[y2uñ)þ ye€öûG…m[9ð!@±Ÿž¾b*Cî74mCd¯ª†ŠÄ¦7’äÛÐppî8Ú0þË­§÷ÿ9ñò-Km¢¤µU/ÐkfÝ÷’o,i/š$ÝÌ<jâsÊ×ÊÆý euûø¹kjhîzæYÁ&½¹ÑŸNØË–'ˆL.tsð¾OAéö‰‚%1ÉÀQ$ÅîFb.ó¤S߀ó‘mbkôîùÍ»˜.¨bL³kŽ‹åE2½egedñÏcu˜dλi”ÉüûMhZõ¦t\ýþZ2Q÷oþÒPü:ÜI,>:éàN8Déü¡=QC]£‘(G'ƒâ?SÑé µ¼Òr2ÃàÙ‘‡Zô>cat+Á¬¾gy¸K75òݘÁ¼yæ$ÚL®´=»³ä52DM\½úoøšûû/Åiýº¡xû·Ü¶$Ô¬y×3!ïDK¯üÚÇ…Çù I4'U(²/&™¤œ89aù8ò˜á8èáÕQ3$ýÿ¼¯ 2ò¬Æul+@nR›1EaMüO¡ê?f£-ÂðÓnb¬CçÀˆjNl€ €ë ;Ê_@HÎw“mÛ)ѸQEµgec1hmÇxÔéE¬NËéˆ b-õŠ|L¢,Z·a¥Ñ°ö¼™‰~ƬŠDãÕ”½,±ð4&•i;ºÜ.ŸjCÝC­4éì•Ò§“Pü(tV<-(„Šb $T—<²Û€aKåg}ºÐ=ͦC}ök/RØÕŸé}ï½2r8 v®%1¹eõÆkw«lú4MJÅ`uÒeªœ@’åWo ·H”àt$£ÀpÛûEmû3t‚˜üO o6\xWF°”µªø¤QSçRK•ä©Ý§ØÞÃbŠÕ$øS¾ÂÀªÄ&½ÈYЦhL< âUAu–ÃÂÀl*ó¡¥oÄÁ)@%ÁÇÔ@ǼyÌYàeª8èÌêÉðÙÄré3Jœ˜ú’뎔÷â‚͉„Exܶ'=˜ ÄHr„ï°œ°ƒùs"íï%o:¨s’…`ŸXkäõ†¯F8ÚWcÃ0H7íì=8®‘øÊYÜá_î|vqè¿e­ž?]UôžÁŸä÷±¨ÚyáÜX÷‘VÔË–w>ø[—ÛuÑèJÛ»D“i™BÑì‘b×e àd‹Žü=ÝÇc,; ƒ††ëüñÁs6LCrËÓAyóðÁtKŒ,ÏÌß˨!‡9aÈÅ5 {ëÆÞƒâ.‘|[¤6Ã0M8QŽsõÌèM^p{\Æò>jú¥}6×îX{ÏÈ?I¹Xä%gÑG­rôî~XAO5ü§œw`úÚÑ8œL§@¢¿cÑëµÁ#Jçëú9W“â— ¼¯älA(^ìkÐT¢¨seYnÝV ×¥l¨Û ™]Úú#cÄy†ÀœJ\¨å„ÁâRIQ¦á‹o•>©B&ŠœGœ²0uKnè• 5ñçŠõj^4¬¡T»¥<ˆ[¼‹²ßms`i›€•WÁå³ÎñÝff›Ú_ù[·$ ª+”å©>`\3Á.“GÈìq‰î`ºAË1CP;Ö¹×{™/ ÃMÊœÝÐ4&T, JµsêƒÂr-™ü&‘ÅëZðõϲáf¢œ2" É}ï+ìW¯'št¶laϤ£Ä´ºÝožt¾Wž~¿J‘(ågÜÙ‘ÂÊhøÿ–KÈ&ü„LÒÑûÍÂ(>?¨øØPíW\–¾'±¨vÑió[`Õ5/)×AÏÓjn»¤Æ"ʼûÚ#¥+ÿ¼Ê:wþÔ’ç‚¶)ë zVð¦Õ.bû›#úõ…`¢^ &¯–Œîmß›|2⟡"補ª‹‡dd£ž¡x楬°`ìæçb#À$Ø2úÌŒ.ýf\(§©Ž…qr¯¶BÜ}½B'`zž}VM£Y.ƒVŠÎãȡܙ.OkW^‡XÍ4Úû6ƒKŸŸiõªª.¸ d"‡ ÔIT© Ü`ç½Îôu»VÛ6 ç6“³¼È÷tàÅÒ¶F=ŸJ,S1QÌdÆ¥Þœ#1<> rÝi~ô Èc‡B œê¸"Ù»ªY± `f€°. Ú‡=+Å Ëk çÏǶÊýcÏÖAáT]ݯ¢g]«k ˆ´œô£ÂË5É­Ëà£:?­LéüSHIÅÀ§_Ý+·Á¥& ¬¿ïBßà+r° ›ùçÀ¦vw"«/}Ó7±¨Ì(Ð{áðc„¬¥¼ßu”þÏõàÒ²w’'xqì_Je×'/ŽºHúßÃl¹8ãüHߊÃç—ËRéûàG屦12>ôé¢j 0.Õb¹ó¨èRXà\¨ÑÏ3Áî}§"ßyx”eÕ=Újƒéám¼xž­~šúA’C4;Ø·Ð9ïÒʖس¬ü›[–¥í\ÑЃñ-“3†ë4¨ËɶW–Ùs¾}r¸½xÍVÈhC% q"ü¦jU÷øMî]ÀX†±ÆX(Þl1CÅ!j¾ŠÍI=*M½Z‰-d)ú5U)Ø82qQ›#í›[xgÁqÿío¶1_!àŒß2ËÖ+6÷É Áç92äÁAô–bÀà>žR9sITÄ"†Æ]ÅqSó¯ýwâ•îGH“³ ¯ýõa‚‡ü06—E¦ -þJHœÐq>º7¼T¡Ëbb&¨ò³ ËžÛyÄ{¼Ÿçu»ãó”³ ü#%z¦tÆë”ˆù u(=cß]’»ø1\xð ··ìp§åRõ‘ØúYk¿MÎám{#z8c?_•oò8(ÄK¥R3L¿.ÞÞC¤iÎòª,ÎBýØU ×#ù´ ­ ˜å#oÇôS——à{ÙJÙ1{[lùàãèRq8ESúPîixLI’ljLÞ ·nêÕ³ˆ/¿LæôÞMË€Ásîþv”Àd¢öLmÄEa(ûiot»U°S"ª«ù­Õ6Я¿Ô…‚sq£ žÓt=Û­¿7‡ PŸCLå6Ññ\ÀÁhŒr:ÐA÷˜0ÐB-QWéBâùÛ6#ñNW  ²åPéæ?ƒ »[ŸÞ÷/Á¸xG–À©û ÀÓªÎâaÝ: 7Œj·àéùm€ù·42€n|)ga#› ª !q+W`bðƾΠp2ö{A± ¨PóOrÁâk]6ºRGшq¦!Ç; œþŠsŸ0T:‹ØO˜Æ Ü!àÚî¤ÊùkòNËFiܦ³²»ä˜zîMf "`Q:´±›…E9É—Ç©oY “œ§øˆ”“ÀÛQ‹d:P•°âŸJ¸¥¶˜}Ëq´«:éÎTgÖI‰üáv£Å€ ª 4Šp™bSˆ=`QüÐÈ/òiÒvòͽÞ?¶àLî}mÌØYWŠ„ÙóÒ|\$œõõ|PI4úûöž Íÿ¤ÆÍéˆJC›­Ÿ 52j[©“$¸Ór|YèÉè Kh›®XÖÌdˆ2~¯J²¬µ¤ «døWÈÏÍÓþ¡Ý2}(镃+gí¦ T±Œ_r–š¦3t ÿ€÷=Ýë}Ê êóvœ†Ó"~:Nb ßÉÐUÎIÂïVœõ êc« “Ç&Ä(˜*_IK\8{‚Üò$ñ?i!¾t%aêª]¹‘JF`Ì ˜WØwµ …"–òÛôv+^-,Ëw¡CSP¤{º8<¦Ì®µÀ±Ì–O#ÿ8ÜTÎ Ê\}Å×™ú~øÎ^ÿÇŠ3µÃÈ™~ïA¬ )úL£ˆ~Ïl™ˆiÝëŸþ!(«¡ß…@½ q._Nää·Ž™JgÜKËè5³…ƒŽÛLyO²>&í/Zf7ó^Ô[“>Ý ãÀ€9¨’M7.úÇ ïg^»¹T5ÜTÙC­ÿ¦þ\äÿ _eë²RìÑB ÚõSêfÿît„ê}Tó¨¥Ûz,ª ÚÌ %>­CÂby&¯Õqˆ_^,&kôÄiŠBiçµèð¦îrNïÏEѸ®]‰Cè^I“|þæ·4’‡±LV“Þ³QÆSœL»§ ®õ”™ã­ÌB—ÙcÜ–,”ÈUðNÁ4œ± ª¯)™H#Z(»L7%#=ÑUv Z€ ×ñã¿öîþ÷ú?×GÜ|Œ.sù/j<Õeñ28Lh¾ùN,‰`ÐIüö°Œõ·i≊øÉÔrøî©YÎ\ñ*^JU9ö”àqì½Ædz9ØÎ$‡KPz¤Žé¹!]©oïJpžfë¤[l߉‚?æH3XŸÕë]ƒšÏÈT¬q¨ïîB{$Œîþ¶ÖÈ"lMñý+ 4é*S%GðP°ù·yŠÈás­öD1嫆¯øè‘^øE?Åü­°ÅŠq¨í”Í-âª`†Ð%D+$°d¨?ÖSÐ(Rf÷è‹)âJ^Ohqh,Eî=È—VÛýSüU½LÂzã{Ùœ0¤HŸë{ ¦ÈS{ˆß4Êþ²ç(þ¸ u7ÿú¾‰ÎŽààpBQ®…Vö¤ðq#IÄBt ×ßÍ›QM®ŠLÃ;:­ à?ì^ˆïÞ&JÂ÷ÛÿGá¹ËfËåþžÍï˜uë´(ÿ¹›X«§ ‰ DÑè4á@DËbŒæ ö‹I¡¢ãySj•¾Õ>Q˜äšï0<æñ ²9Máëv•ЀÜư“2ƒ5k¤Hèt# ÿ²öÙz–½è);zÅEUÌY¥¶)Þ;°ÂnzÇÒæ.ƒ³Yáíc(Uèû¼±ç0Ôª*`Ú/Ó©ˆeƒF\šüQ\ä–Vãžã2rƒà™NÓ>Õ ÀÅ»Áè®6zgñóõQïmáþm:n¼”©dÛ=üÐ?!&H6¦ûÐÌó…;ð8ôZ„‘Ь§–Dgð-¤ù>=6^¯H¹þâ.)ƒô~õqš}¿ ?üvÊňìáË<$þGm.‚¶Ë¿ÙŠJ±Ú‡÷”Ôˆ/»d ä…уÞ3žo€Õ-ÞHoÚ´èmñ/•©‘Û® š°Œl“ãkQ\K@ÓXUka²àhvÂxD»rV•ÉZâF-Á¹Ä ÁjÚüR—c|'†`Ï =Ç)¯ê^™Ý&YQ»àD¬?0}DNzTñz âõÕ¾B€Lë\qêcp‚ˆ;¡ýî¢]„«60·òå ¢1TÇ6Äzë§ó‰¯áì{iž˜ÍÒ_©/ç4E2”ÀD@Ò|èCL™'ªWz0úEûÌ[DeN07Úµ Mzèƒ%|„ÈŽH¿›Eõ¿.NO{ËCxÛUÛm4â÷¶GÇÌz—&¤&"Õ·ÀÌܹ*¢C|=•¡ t,âÅ0ÛKw³4ãj=ŒÔ÷–Û€}Š“Ð$ߤÅûÙäÞæï„è,K˜~=`¸¹Ë×ì¶"of™X0Å3…o-‰90d‰õEŒ¹Ivªn•R[Ù6^(c‹Ù@ÆF±µýÿDì°ØÆ½&+Ê Û^UØœ@5üѳMuäæ¼¼!?3ªî_…Hn± Z½þ2®?ÍWÎôxµ ôžúoHqC1ÈúJúe|&µAŠ‚N¥]ÊÓ)ž †N8µŸ“®xðÑ4¶üa³~Úˆˆo³Ã"rY)¤4œ ŽÖ1ÝwÌÖ:¢áQ)å,X!:B†7ùß›õyÞ˜Šgo† šMŽn|1 Íò<‰`̻ƸX‚L­ª¦Ô“^q™wR‡¯™‰#»Ö»ûë÷tÏrҋЄÎFó³ÝÍüIN@Õ;gºå5ƒØÔ pœ/ß }ì‘òì´>ц«v:u±ÑÜiçï‘éuÆ_j¹—%eèyêD‹æ;%ùð(ÑÒÈÙ€?t9PÆQÀÆÍF8Cz€I»àÖÁߪ¡Ç?Ò½3,ÏÌ«¡§Yî/ 7âK²&þ® w—§·ò±quNÓ^bÓ¥}”=ØLjÐâ˜>.Ñu L]޵>ãJb#æ{qHÚhò¶Eqï¯ñ‰ÝrPã\—. ²Û^– òèû²åÃhÙ÷#P_"á4]Œmg€Ý{ü†r’Ráý ý^¬täš]OæÎ8ºFÈXxDúéX·ÿ}{¬jYžú÷àp䨻?7%qL†‘„ÎåýÖo´7ˆ§j‰ª`¨ÍþÙžžáì‹Lýú4fNÃ+ò|Œ[3÷d…L}÷áÓA÷És<è‡$ ã-Ç¿þÕid%y…Ñ—¦ÿQµ(Žh^UªB”$ƒ ¸O‚O„"¾ìC´BqÐóî‹’ª<€Ï‡Xë[‚ºï Ž,jMŠlõŽÖ¢”š' Ò÷½6ˆµ’9HYxÙ›…&Bš¦Ãq*dĤ`D…ÙF®?žyëV ±Ä´‹wÌ6¥9Õœšm&‚„¢JÖ¬Ê ½‚">Ÿì¡Š<ЇÏÑýv%icÕž^õhàÅÎn@Ók²F0âѸìo©¤»†Ûž£1 5éý‰øOÑ6ˆ˜d"Qç~ìoۜɫó„í’)Á¢.3„ W†~³pÕFê€nHÞëÀwpÚæG‚O³±J7„Øò€k-•ª}“ç<–n¦ƒIûc¬Éë&Ê6ѽÊk«‰ÿäPÜÇ´›„ÑD/ Õ¶WBŸw.9vxƒñíÁÜ])ïGíðš†è{³ö2¥£é¯„øŽÝ¼”9$_õ¨×•QØ%¬ÈY)LkN8â´.J¾ñÕÄ´B·ÌftC²Žg;;YËÐiq°€ªížöm€?Ø{*HöýÊáK‰°‘3ƒ±Fv;ÞEùšØýdZºÆVÕö+ Jà%»žQ€¿Žò®¤¢sý£€gM…„o1ñ˜Ó·|,e;4†ˆ4”ht°ÜdFíç&¯EfÜb)™6zKÁJG‰ÖDö_ùyuÁüƒ{–ÙEŠ}¢åxÒcžP©¦AsC&GvÔÈ>oňãc“seäÖIrñœ7¡f xùüZÈãE iÈÍ*= ó¤Úêâ vϪ·Zâf[|Ä{ª‘—«>»-hQ>¡P¾PwbcÁð¥4gQ—¬á>nzŒ):ûåŠJunÍ ×Ç3€Ô3ÔÜ  °4P±¹yç˜Ã?úÄO k€löñéqã“R4~N5š-T8~¸Ë¦õðÌê5 õ6AüîÜ8X°JÂç¯D€6öÚUüù¢°ˆ%&~ÛöÚ¨BCÙ´Æž}£ñÔIyÉ£‰ýGCù®•;^rY@•nxWÝwM}_Ñ— È™G1•,äEî5Ûø‚ôS¥ ?Ù©‡AŸ‰(¬‘SËc•O 5óx‘$ýMLRð "cš6jÅšê¸Ñîìš*`Þù3: ¨gOX•[ýìØÐoLˆÆÑ"UV,3ý Êc¼Ï­ TR->&P ÁÿÇFÅ–¼`±pШa€¬Ð7õ5>W“ WdÒd¡‘©X¾’sõ&­ÞIÑàZ€Œ¥*¿ã¶Ì‘¿‡ÈÖV@W¨FnªÃýí&͸æàaêeø]&Y’êË€3 ÚÈW~§ã) 2~ic§N—œ^_úb0õç׺ÿb°ÙWOö ^:ª0c|ú(ÖÞ+oÅJP,5ÐvÔè·Kéä/#ÆJ`¢Í›HõŠ~T­ö.˜ }ºk®T]M±K©_.pgi½Ù¬VÂIÛÝNõÓÍ@ø†Æ@"}×÷þ/™ƒ>¹«RÜÐÃ*@B¹=ßd‡Ͼg¸D5ç‰ZûÉèv ™.P\ÀÒ,÷L_¤!ÒŠÎsŒý"ϽªT2±¼,õ}Ãñ­äpˆIçvÀÁôùJ›F`GK†÷Wex<߶̓–~4\)-ùу|9w¶Ò= åw;ŸKÏ{ƒ±3¼«ͨHÎàÇÔ(p%+3Eîd¡ëú•`øYjGòµÿvâ5ͤÉCábQi œ§¢Iãÿ;WQïAÙcp⟲ex²åLÕY4º¬6'›<ئÖ)ZMt×n 9¿ èÐ+oƒ¥ù:וSž3 ÈÖU*Uk4æì= ½Óµü>‹Î_*áÞ÷±\Ôety‹'‹Çs®=r4Mµ'-C£VªÂW¬¡àÛÄ–°©þC©¡eüÜ)"F‡‚jÅ’ö|T’ë\. Œëqx¸_Nòñfþ`fÙ·uíôl¥x€v‚kdéÒ ‚Ä›VѪgR%˜€qx6H«:\àj½2y×$Ž&'ËŸTÐu+ÿ)ÛlÁàÍÖ¥8¤uÒBÌ1Ò2¡|Ù‡ŠjrÚ¶îJd¯‘ÇãAݎ죕:=ž¯ý×ÞX¾nzóõÖåÝgZdzX†àAÑ©µ¾êš;‘€‹×ÜFñ;é3פê¡$&=j:Ð+ðÝ­4§J~? |‘›Ší¡6]ú>'rØ›½€tAÖUU|Í—0×4Pœ€7ô ÞX7÷Ž`;Š,vCÊäÀ™ƒÄOÇßÍÁz$ö'dj`»ûQ7–|„i<\žmº=—^zA@ñˆµ@àñ~ÛBĆHGcÖ´ðÙʆ».dÅŠ?""<-ì˜eŠW·¡¡ˆ}0…´—§W°IáÉÔñ<>Þ†‰¸¾ Z¨¼Oì¾9[™ÞIÃ_Þû¯ò`¹í3@;#sž‹iDéI=§ºùù)'ΈÅ~†Þ÷ÆJ[ªàÐ5^ÂáÙ†Ü+ô«:©—ÑHAîʱèÚ»¼oÑÅ&ÇʪåË$hQÙYªÓ½õŽ p‘üxPíÁÍÒ¿²›¶§~ë·‡þÉ ¢ìª¸Aü"‹é¶f¨I‘cic—Æ•LÉULì¶ÊbÙy^ËuÜ­@úä@ж¦ß¸¸¦I;;D‘ØÆ”p³ÿÛQ1•÷¨†žà¨·œÊ²‚Ôßdªtçˆ}úrø_^4h…ð·{ÄÚ\Af ÐÕmó ÏòŽ€îÁ¹bs.â®~Fhb‹’Ë’e¡Þ[ZËîÃèn!ÄðÜ´åö ÉŽ#)€z*»õçN}P½`íaÑîõ$Ç2(” ËÃV­ÕK ?îŽùüÊ»§¿3 JÅfÄõ„S*ÉXµ*8±^ &X·€­ãµ‹÷ÙÊZâyY‚—Ãñï§Œ™¢(]Z¥´Ä%=µG‰é5(†ŸØ'iÒG%º›õ7´9¯ž,ŒÍ(€ b~]ßUl› M&Œ¤=צ㎌ÂúVKYiîY>Îsê‹Ct.–𣀚邯ƵB ÆB ´­^óc(R)d`çá\h¬1\â`¼‚€o¬îÂàëöãe”Ê@úOð¾§Zó“ßVòåe…Dƒ­mÝ­>o]ÂÿSWd%¨ÎñR–ŸlÍÌKšDwÌܬ)bV0 o295k¢KÁf؇Ácj{»÷” FjùáPrÇ–lI¥ŒyI=纃y¥—à˰Ä?¬ƒB—å’â¹Ô+;»†˜EÅÒNÃS‡ŸCõæ©°S±kµåS.›ö+22xÆŠTØ`Ýã¯Î‘&€›©È¹ÛI%rñBb~Š9DB;!‡ŸÔ&­3:¦u¬ƒŽtUZ8//2 »¾›>̦T”Då„]Døêî®ÚË™_ ÏœD±F5|—q7g{†.ÆCò4[$çý I”ò1ü±pã•©›ï/#Ž4A¨ü»ZÚ¶/S˜Äc!h «àŒÈ6ËEÚB°¸šÝ鳋6>'&ªÒ¶2gÛ󒬿Â$/Nu®³åßAjp‘ND‡Û^=c^Žì§†¬U›Ï¿ò$lêpÞ$éÂoL6¥Ø–ù`8·pÚÙ-õc#JÉHì¯Wz”+²üëÅ®§ -‡àñÃ9hªL£îýŠ}P©ZþCà&]q@×2Î(1É/“¦Òmµ6 8ØÓ)»ÿE$CŽãHê•h7Pÿcf'Kâa_Þ´/šN4¬Dææ{¼Á¸ÿ¼=áFïé&Ùû‰NJÎû{Ý, */I> È‰Ä NhExFßEWSíÈ=›ê¢xʑȎcß4ŸèŽ}ûN:»ßùU:æÄ4cðÁRr÷êsáœ1“ò‡ÀP0²bÃd'g˜ «ª4íêÜ÷šçŒ¨âñ!òÀOí·UOÚÿ#.ŽM¯³wèQ;ŽvêªÖúmÝh@¾oÉ.Ф¶ «è³‰ÿ têcQNò6v…Õ’ÕRë·+L†ßú7³:0¡ŠÔ7ºˆžÿÕˆOšv×CŒƒr0ørr÷d ¸òm(†m%9|ð ]Ì|Œ¿ ¼+ô:=@žq¸8Lk Á8ž¡ýéÃöqwEÔžO#äy2\à1èçiÅäñkÒ„¾}ÕÏØ^gØ£«‰oû$j?|M ¼¶þD¤}T’ Ôß1òXWÚ›À‰£zЪ¥ÿšê©…úû Üã! åÛXñ%O¶á×ùwÛr˜eÅlü)¤@M Î7-~Ž“e¸,è®î°e0ä‘+Bâ8:ųfÿIv>¶ŽÖ$ñ¦‡Ì©ÞZÐMpÀeÌ“¬Uó7u¥›X ˜fâ©Ðø»óP»ÍâüËg‹Z£0ãpÕ¢:h‘'†H“l•>€l¹‹íõ£R“ú%&ÆÊ4ŸÈ1Gp‰ºó’ö´Ÿ{Žù¾K21M r‹V±t饩óJìîEZ€6.æ!¯DÄÊ€!ÊÉýø!j2|µëmëáCÚ€0(Ìè!A_m€”x!•*u¦•ç¶Ì‡ J.¡Ö‚Í•"XKÒÉÑ ×Íø K¿=‘ÜÈkúœ§C`† Ý|±3á³Þî :mSŸNþ@fNŸ¥‹FÜP}§d^$;Oô;ÙŠîe ¾kõ/û¾dÜ5ÀRCΆ9yQ6œ­´`Ýñ$8]ÿ†ïïý9CG×/ë˜É@W•ÏÌð1]ÖÿÄk›Óº‘-³ >ÂJñJËPXOòø8îsÞEÖZ=a/Þ\ϨdKaMp§SKù@ŒÌq\åÒioI^Ô‚s+zµÌ°´etŸÈC:.ïšÝ“€Ï×6m>…³Ø&¡•ÊÄ&°•\Qé´1­×¾tq«ýŸ´BŲִR÷à~…1Ú(½ÉòÓ®°_“à LXM_X¬¢K±ÌÖóWG;Çéü³ó š9¯Xߟ¼;5ç[c_Ó€wF Ú+û²ò<¯‘¬¬‘ØcFxŽ–hg!Þ"¸åg¤Ï©¦6Þ ì³›ªáòOØ‘áõ7Â\/ëB³aTËÈPú°òÙx(¥'„c¸–{n«1ôú¤`Ory»_«LïÃ{Ÿ+)P»}gQ²¢Ø¾ n/[ˆòQ þ>) U8#Qi¢ü†­»‡&ì×Û™ýìýZDþ.»¯à<ø&i¿ÈÁÚ%§»1÷°ÉiC0òùÏÚ[Š4&Å÷híC´Pƒï@ø2£ƒ¬ç ßšiJû H ¸zQßâ,CC§â1Ûz_ÐÇÆ…¦õ2‡k%ë_Кê¯Ô?Ãç—±"&¼©X)‘þß™”cß¡‰´Ðh÷>ã>…1}‚B*Òp¤ªßjwêèGEtd˜à­o팮„¡¨Rë¦_ÙÑkýHgâ †M`qa:âhG´3 eÜN0Ó¯tZçüÖ€—Ó‹»ikÄú!`˜þɱhúÛÒdihœuj°ƒ°PT% ãoª‰©õ“Óod$v\ôsHPè÷sÅzÎ…¥X&‘‚¡Û–U>->Ö‚€½üBÄÓWâŸ\ë 9«ÑÖ²¢%»È…v2ZÎÆ½ÒùëËFymxäLªsöË][8šsⵑ2á‚ÞŒ$J2v2"â%YûÒ\“³~X]RPm|9Fökije¦^ûTÂÉŽOÈëÎ-ª%NÕ'ÓÁFlF¤%ð¬p„ú½HIHmáWÆ@$G•´7@Þ¶‘ù%D ÎïnqÂŽÅjÑoù¥Ú$sÏÊ„„ž;>|{ߨ·¶®ü m»ý»ô®:Jõ‹%¨¬áˆð,F¦¿Hkýùñà¾^†Š³>3þºá˜Nê8Ù—#¿úÜ‘ Í1\\MŒs©ÚzÛ–3]Ed¼ÜÌx7ªRý€)€fM‚œ°Æú@ŸÁ§Tfñ4_âr@ã$€n ˆB‘ò5äXõ±* æ¡NŽ„Ì¥ÊÁ7Yª=ãÐŒÌsKµcu|Ü2xK!."¼ÖÌ篨ÁÄZä˯@ŽyzN:q@§ñ'¥-Tôà]…Ez;?ž´‹hû”Âtô…ªMÀ‚—FÔ¶/òaÐDä3IîÔ0×ìpK¥þXe^ÂŒ2ÙŒ±9ë$ õùšaŠÄ²L)põ+þE‘`!Fa§Ao%ÔSiâ/N¾® w.ˆCT,ã‘#ÆcíªmlI"Fy CCW8f 3 ¤ÜˆÌòi/ha•ãzs?NõÇy\s{¤¿þ„’ÀKHmNì÷¿Øk¹èzëŠü ”×Ò-Œ‹›³­¨iÜ9Âa¬ž@Ù׈VǦȱ_Þ+¿ôÆÁ­Ô[ð%_–Ih~m«®ê~ôK²«þG}–3²ºµªvŒË}ÛɦÃÙàÆ-ÎÕO±Ü®¬Û§»ÍýNè!Îö(ZÌëòUzÖvµãú?ê½ý-ÿe¬ [¶š*ë’T»‘|?©6þ¼¿ç¤Ÿ‘öÉñ«iŒK‹Qž}«ZUîNª`nÿJ_Ntb­#kCÝÿω~ßV º^_óiÁm ys… SȦ·¾ü0þщ‚šRJzx°5V¾þ*SuÂò¶öï+Å©‹–Ž£lÑ…·^¶‰¼}$}  ©œ¯¼DûT³Înr´«P6¬î#®ûÇNQG-léö™fÞ³þ{Œ j‘*_üR!0Nõ­eˆ×ÃYá‰néÂlÐ°ïØÆóϨr ¢j¬¦ÒY‡G3äéÓU Ì?õø!Õ$Ûg†ë Þrš¸óÔ®9˜“1N‹“PÆÓ[½ ¹„y†‰Ï_bÀ©t‹G‹¼J«Î°‰¬TâÖ5»”Ë'ÿŸì£“`ò?(öŽŽG×ç?å&ÞÝ;˜†8ÜTò@“ëãÁ‹w[ e>Û¹Îp½ä›Õ—ò2/¦gi{fœž¦µoÌz/ÏçƒF&¹\àÚWÓ’º‹­ãÞ2Nü]XTòㇶx%´ÎñbÅèh-i8VÐ{bc5©ªš’ð9BŒ‰Šè8â·žc“Ãý»Þ ò¾ÒržÉψ9½'U…‹â ë ">NBˆÈ|vð-Í“Äcÿb=?Ûq2Þd2»Ç»ˆ·$k×÷èc  Y­ýÁ®u>¢Yë´ÜÅ@i’š#‘?ï²ÚÂTA$Lþîƒ= ÷¼{9 ß·þ¶<ØhEúmظ’ys¸ïâ‰Û2< -7èZsÝpÌyGÀCOƒüè6¹Q!ïX˜1Øa½q7]ƒîèäx<²"wɪ.Þ}ZA?{^Çê FïŠI"Ÿ(JܤÅüAœö=ë[ÿ0bLÁ–Ó?¹ñ‹jòY·Shs)ð IìäC6†!’DΆOaš‘âf‚ýfOmJVávüÿÖv NßñË$”o–…D=žU“ØbI騲WOŸ¿XzQyœ^tÅ’R"=%›SW‡?“y6Qâ »,H8ZîàuVÛYDil8 ³“ÃüQ]Œ@¾¡• ž®ë‰z4´¬:Sc2e1j3ð©#Ü0—á` o ‘ šiÕçËaYÞ«¯®#ŒYØöxÁ!•ø9™íØò5TÙM9ö1)åçv¢z¡1  ]!kl³Ú$Õ^D¤sP›Ûc@“´pÝÉX|P¯„“ÍÏ7(ôö/ ý‚\~Ó2ñÌpR÷¯¯Ç1:7µ н^"ò\skÍL˜èYE›ÉŠW¸#–“«>XX\¼ZËBz1ƒAíUÈï½Í}ắÔ&žM×pP[R鮣ŒªA£å«óç ¤’àÍ;±’U×W"¨ù³ðÌ,jªß êèº<΄ ¦m3râK*ÔK¹¿+ÁGë]í ‡³'Õ?8íœ5Ï‘‚-å‹é]Sæ¢ICtš"J&ÔäДv‡«=’Ôõò”¸ˆæð’“3”%6‹à¶ÄÒ᎚Ãy‹4NUO”¼†hLή~å{ê•Ó*ˆürC<àÔG$ÖÆÛ±ÆwàÍ@ Åéí×[†[õóE¡1‘]8óË'_¦°²Xš®…’È€ô¢‚ûÍ¢ÒÝ>9z1.¨‡y!þïÉöÍ0‡øÁ­ì¨«Y®ÁÕM „¸ùÀüœ4N€¸¾•O6°]ý¬ëV9{@ ÎèiqüPnÉ|‘†«ºô2Z„°Çý$ÒòˆñT«ø’ÑÇ¢†ÂDW7VÖ0uþ ÕMöŸßaˆ»;|‚ iÍYá\ÊKûÁáœ3²"b\)É/¡ðÅÔ¸Œ,+uzrÜ•äŠñ I)ÈwÀúE¨.8~e2Â’êPŒ|û¯Þ˜…]Ü!w?˰cÝ jŠ­­}0ðÿ3ÃQ›T%2É’èE-[®Ýp-˪Váº×˜w\÷Ä‹yù8ˆŠ¾"NͦSQ3AÙ½Kž`ù¯ÊjΉKžµ™÷ÜJÔêÅÇ~Û ª¸Y?ìffgjÁ©à[ö"ÇëàÇÂ=޹° …ÚŸÆc µ¹6ìyþ§ÏÒŸñ½nYŠulõ(™ûÈ{¤÷#{SZèº2ž“¡o쾛ᵯÝc¬WEñŒ6Þo4³¸Ýqj`°€‰RÔ¿amâÞµ\Mjé?t¨«o‹DL{áR'쪚l)‰@» œ2·H„d _kïp™ŽY´^V‚ÈBøŠhÔº ±¶c¿Biç» Òx $"¥ùŠÝ¸?e¼Ú¬¶¬9ï•Ù5~#—FÉŒýÈïĪ]Œ T±# ¼ÛX û‘G ùŸ“’¼zyM¿ôKƯd—È”K°ê §È’%ž0ksmmý rëh>ÉÓSÛÂHåQ°ÞI žÛ /Ûªi©Ø›Ñ½Ríù[ õüûbƤêTBs¼Õ¸›&ͬH3YáÑ'’? ¸**Ï4è½ÌÅ!D¶¼<ËÞ\Ä|ßBFuسp`qìÅôbŸ 8ÔÆ’“Âä>¡*‰åuñùÆ3cNô¶µ"æ¶Cr'ðÊîÇ8ª-¨’YM©ˆJ"%–¸Ì6Óv];¨/§ê^Vé-É›óÃûÊ¡îOÜ ‹[J–ûVý­0w㯲¦ǘUÖì:!d]»Û1e¶©„<å¯ CÐñí5tœ¨Ì°uº®CßÁÜTÂ7Àe|¾©{È1ùYe÷Êg²FÆRPM­ˆÓJ/,Þt¸HéT 3ñ@¼¦ë”Fgaå(ñ‰&i„)ÍƒÐæøµc;-Zt¼#ÎjÌìiïÔŸ6F—01лXPËMDù&· –éDÆØbÞÆ°—Xºš$d}Ì%ý8DrÁÈßÄ34Búù^CY Ï Zr——›¢ô7yÐh7£Q»$W™™¡å¶I…ª†rÜññúCe¯U„ê ±~È”¦P#}ð¡I^9^ž:·i-HŠÊÑ\Ú §@Q”/®3 Ï’±ùåLÔJ¶¥wc Ðv÷G†Ñ^w '2&=ú‚@ST: Uµ-Cƒækh ïWB›À§ø§GGDÒ è^r+‹DÈÙH¸¢Ù£b)uË™‡K%µ#ˆ¥ÏÀ "d½ ÷U\ü«>§;Äã–æŽV°9>*3ðGð`è½²´¸ÂÓÜ Yt`>^ë‚DÔ‹´'DÈB ¡Zlñ9ƒêG¥é[ÓHžgéE`¸Â¬C¥äî-«q(¤¤SúÈJϵ9ǪI¾ Ú3ŒÇÏ2‹«¢Øð‚à°ñœm\è=ùÔözgŒ®ô÷â\V>$1¼º`a´ºÑ÷åÜÊæçbDgßìuêÏ솲³r·ye{P.J;A Kx|[ Ûªû@£~ Ÿõ¸3í)Ø{@ñË_ö„&ßµ–¤uŽ¿¬€±~·§™{ØÿVóQFÚg´ge¢ £U¨•D±amÆ´rG±;)æÞ+­áz$‰µR5V,+«³8dFqq@£–dÉš˜ú†÷;EÊ*È`ÑÊÍ j<ûmÓ]%:½™ìMÊ F†“Ò+ië#ʱA¥òä’Ù‘'9oÓªþÛiA``höd–eg˜°ãØ#f¿mâ\bØk¥Ò;u¥÷ûᩈ܈ˆ9Æ·4œëâ¬7àÎkaÌ¢Xš#Àˆ’óáö¬y[^w7ëü¶£zr\6‰l-ytܽ±ÌŠãÁ ±ÝŒžÓ¥l‰ŸŠ±0–ÉÀæâX¸,[Z_øñÐ62ŠÙù°'½ÊÎí5.OO¢¦ÚÞÓ( ÿD}ê*ḚvØDJ2‚ŽØ;Úˆ})¤;0PˆÇŽeñµû4P-£{È”ñz{ ³JÜtPþðn¬í¹¨==0sK¬e|UÀØöŸ›óPÁâx6/ŠrLox†f",#Ll¤Jr°Ö½\2z‡ `²îpkügsÚù/¹ ±Vµ†rýì åoœ–‰¶!sÊ>6sLaÂëiË®Ût‚Ñ'þ†c£‰—˜ôZ€:Ó)²lÐm~ i†'׿#vS:X}¥äa"M:t%€HºvdSËýCÝo(ˆ™þVC–REŽ;RÑÎëº)ñ‰ž.I}õé%¤U]mèôçp%ýéŠ@a”¶«³´ÞY´€'ÓPEt¾pŒP £6úÆ÷2Ô‹„ˆ¤C–M…8:F^F‰y¿T5A–…†J!¸üðà&ôß´¦hï Qâ¦^4eü ø²Øò“¥nS¾)•SÁÉÿ»œ´ÄÈÒö›Þ\ ÓdzmCµpérA›/qö¬k Ë<1tÈX¬¸Pþ‰Ø6¼‡ e;U[d1 sëÓòaZÊ|¾s$Ê…[páSŠy6×ð˜îoí­}òúd}¤¶ y©Rز ѫїV =µ`JbÖ$Ô°7¯0vR¤†-Rµ¯ûäÿ¾Ò·Î8ºòØ9³'ݰ;{ÉX¬«:GoÕ vZ­+>Pü$œè‹W,‘øíLóµw]ÙßiP¤ý0”yVHs®2~€t£¾'²àŒs3åûªÐìüQdŽž(? üûЋX1«‹º¾¥¡ð]ʾng#tÖdŽ=—qº©zÞ']ÃÌËÔÌ;.™'1bÍS¯¢Ð¶’oéýä+ÚØûkéeÀí-Ö…|p ö®lÍ&áÁa×4™u!¯f77£ÖI`]ÿaÛf©m[©¼”ÖDél!mŠTq+Òu¯ÚÇ»lÄ;['ÜUÏKn%ÔíæwÇÜ#PA<’F­š"_¤Ã „6‚Cê±z…à<*æ¾U óã^ÊŸ–ÖÙQ®¹È#ïŸ;8Œ¡ç@«–¾ðL(n¶Äî^Á¯³!\¾”+{«üÎ'¡í«Þ ²ß¨bxBìY"Æ$½HrÉ*VŸb“I¸×¥öœÀ¢$nƒÛ°ñÛLÔ†8±#t2þb¡»“  ) Ø#•Rï·m²¶ìx¶Dåò=^¼Œí™ŽíÍ)À5F# è0õ®‡É2Ó²ÆyPPß»N)7äždˆFÑ-ì;BBªd éäBPAÖ…òS¼žÄuš²áGó[Ç^ËV)a4‰Š åáha«Ð ž}uê=ë?#Ñ€eºQx3Ôë@UCZŽF¼æ&•w‹Ñ7qU`''б Æ5ÏÓs+ iìZÂwCä¿ò\ØEEqL% ¡[¾„#jûeÈ@µlJjýÍ È/•Ξò$:Œ!Îþ溣ÓÖSFtÀéßç²lÊìË“UQ‘Zô¿„9ýú‰ÞÄCƒ>CfZí _pÚpXsŒ«fÚ¦ ‰"Á>Mºâ…° [·eÓ§ž‡‹ÞÕk„x'²æR´‡±à}knڽü]¬»Â\Ö }fÒ9àgÙ’…V¯ÿ‘¿ø£ïldz¶ÎoÒòTéw .K±îWK}µ•gÙ/ΈgúZ„."íñóeùf¢H#¨ÁØÒÆ>JÅv”o¤¡yÎEOVĉ ~€òÄqð1ŽU ¶¬ñÌ^T ¨8tkݱDzIS½Ú9žƒŠóVØ(&NÎú‘oh.Ù8ÉÙlÝkáIvOç†òŸf„ 4ùÀ–j=LÙ#„i?\ÿNã×¹ȬEuC(ú'E|ÿþ !HÂÆÂy¶7ßk’m]tŒ„FC!¨<öÜYÖë9™†S3È Žús̸Æß2jO[(7&%8L¤+¿´««%¾/Ùà y²:ÌLÜ|ø•sÅtJẖÝUwy³&w$ÌE¥×Gƒwãa&Á á ±YW§Ur‡ï¨ëþYjÇá‹b"õA*f¿â(o/UÛBóPAXé8‘°]ü!|%õ$Ùÿw"ÜûŽ †ÓÿðçaHÆ(<¯íVݽ±‰t08P d«®Ù€½›:Ê25•K˜‘¬ { m›«Û5@ï~#`½åD½xA¸Hþ\q @ÝãÊT}„â]$¢—€Ö´}4þˆXp«mÚd‘$ÎT1¬ï`½OP¬”¦Ã]½v”Ä=+¾:Nœ1M‚IÔ–èÄÇ1ÓÒ.ÿ†¨€4ÜHb‡'v 6)oÓCøÄ&bÍ~ããÅl}KÈäÝ,ÇÓSãsûÀ2 iz |n>Kÿ gÂ<¥…‘‡%u©ßè•ãH½zÐ(mÖV—vžèZõ×ôó«1×S~¯i£¬Ñ-œ&²¾ˆ\fykÚå´ì‹†„5)—E„ôò7œX™Q"æ”% ÀptQ!àÖIzÖb3ÆÙ~ÄÞÎÊ[aMU⟼v½‹é¿ÉÃe¯ˆ¼‡da†8 ¢O@:Žj$Ú >nt¿ìûª7Ðã&P«'ýªôóý\Öφ–%Ÿ»j[®zŸ"ÃùC¾´!¡•þ‚:÷v˜ZöþaôµoãgUBŒ¾—^‰M¿×´Ý‹î.û"-ÀwQêí‹JÈ%n¨ÝN&oh¦*Ÿ'5šHl7û¬yÆJñ¨Lmç^LGBmSF§(ð"gÀxôÆŸ2Îö•z¡ðô´è6Jµp"‡cæøp¤¨‘oà>ËíÜHHÔÛ¹ž#„l¥hVåó(lîw\·iÚLu^kÒ*ûrçømCš +ü=1 Ž„ ®;9IÙå·= 1CÁ#R\3üÃ`XrÔu¯²Ìýþ@ Íù€¥¾A3vPXC×÷R¦ma†ÎYôÎwɼE+娼”;±â÷„þm^Õà*ô ºc|±ø†LN6Í÷l<ÇŒúÚ¬ >È Ã~Z1ôÏ úÝ„+f2ñç,¢ð)iÃ!ùªÞ¦¼'bû!¼Oï{M+Á|D™¤3­uÉlÜ9$ý,è4;È!ýgÖìÆ¡UªÅö¸9¹*ˉÕ•ž—ÈÛÆÊ« Kš€È•üƒ¼{ØaÄò×p°ê°0y ]UéN¨ISHá;lQ_wˆd‡Q;¤v…ŸD'µ³°K‹=©ÛYKå#ŠÜÕ;/ìA”Ð,ƒ¦âÙá(Ó›iEÛd'ă0Ç%Ÿ”¯äÞHoÖ·V6Òp¢œ¼äÐ]Þ£ 0÷<~‘Î6xì ⃣N@¹ÜÕ€M.eª X²ø‚2l7?ð>˜…ǹµd°«‘ëfs±<Ùä9`¡q4ØõGHäC*yEÂGà<¥Ý%ç‚XŒ!Øæ§×Ìá¸1cß­À‹ÒÞ°fੇÁy‘8És”AÖ›B1D€ôe «w³Þ²5Þ¼R°Ü¹ËBœîu­Û¹ iTæUú|‹/¨µ!~Oã §ÛvQXàÕv î&øÕî]ˆVzv꣦lt1šéàŒƒÊð«¥KiÁë»sÜ}©›âÞ5 mîÅeÆþi¡åxƒ :kdØÜ‡o¥±ÎJáaNÆpßð5}è+« EaGüHc·1óýŠÖ“f>³~2Öº¸Òü{æ›Æ™Ó“B{ë£PìºÂOG§¤S¶±{ãS†¤²(*ß@à D ‰MLEòˆkže£ ×F«ë³sÆ™R/Ë–µç§ lË ^$ˆl¤ö©Ë{')^_ˆñ ~ÇNtŽ3ÚØ{¢Ž*'.câ=.ò#¬jS²ôò8~íɽP3•ôfް~Cر£õaÜoã0‡îë-”¼zõ §BÛo…Ünµ Åõ ’°yj “ ÇìLçêN¬D™Çh|ÀÞp:€Ýà‰÷17 lN„˜nj(. +íáØV9Q‚¼?Äë‚8;òI4™æÃœâW¤§Æ¨3„D‡7¢çž–2ƒ’שälÙUÓ|­‚ôÍ«ÞË‘·+DùŒS¸¾™Ô-‚}ë©$`•= ¬ Zøò>Èas ÙJ¸ç´W CùÖŒˆ°x8Ù–ÃhEÓ3Ç å -­õôÀ2!ÕïtmËÈLÂ&My¤úb!on|ç#åg$ÇÔøñ±é<%1˜ <¦¡(L( ƒ Á=ˆ ýhì½ø;ÞnA:éD•CÑ㹘 J¹'ø~þàÅ|²cî'8]äÃ…Doš4°xc$7š.ÈÇíÀëãñ“5óÐÍEO¥èÓ‹¦OYåglþÎãɉ©îDî×þu6`ÃçÔÇb͵§…³®*?ÕtQ@ßœÝeòdPyã\h)UŸM¨ê…1l 3¸|“Q×ãº0ºLÝtðPRãÕ³ËqbG3¸£Œ÷˜8¢ZcDƒÝõä »¿‘ÜÇ[À¥ew…>>?û ª†««?ôGýž]pæÖø£¥§èÿX1Ðtþv—Y€\OÕ×ðxE,è­Ÿ Âó‘ÌíeQpê7·WGûááüeFæí±ÃBÅ_Ö¸ÐÊ×Ç–#>®0rvñ³xWDƬĩ'Œj¡Ëˆ ™vñ:·•­€8šÑ¶xae2wZ뺣Ô*3€rÒ$TUûuDÜèpµ8þ{µg/醸Øtö€ÒŠ1Œäx8¬l2³"ð´vID~»uår®bE˜ZY±—Ì'’€c#¿Ü›u}L‹&ˆW15ÎðmÁg*>B±óíá —óÙõÉx†œ~¢‘.FˆžòÀDÚÂà°Î‹‚ ÆD»í>0 ‹YZks/data/worldbank.RData0000644000176200001440000002570413216563676014527 0ustar liggesusers‹u› 8TßûÀǾoc_bì²UÝ׾ˮ´“1ر¥¢R*Z(RJ©´J)Q i·´RIÒ·(í´ˆ(ÿ#3wÔïùÏããÎÙßóž÷=ç½÷¹à<ßR|¾8@à' øÐWA~ô HCW±d“N¦Çò(C]¥ÐõéŸJã QÆx+„Bx¢At¼1b¼s „$b¼4B!‹Cãý* Je„ B¡†PGh ¦ 4ZB¡ƒÐEè!ôC„Â11 a‚0E˜!¦#ÌK„Â111 aƒ°EÌFÌAØ!ìsG„Âá‚pE¸!ÜO„ÂáƒðEÌCø!üˆ@D"‚EÌG,@„!"!#– –"–!ȈpD"AAD!¢1*b9­†ˆCÐ D<"ÁD$"X6" ‘ŒHA¤"Ò+鈕ˆUˆÕˆ D&b b-b" ±±‘؈؄،ÈAä"¶ ¶"¶!¶#òùˆˆˆD!b¢±±QŒØ‹Ø‡(AìG@”""!#ÊGGÇÇ''åˆSˆ ÄiÄD%â,⢠qQ¨A\@\D\BÔ".#êõˆÄD#â*ââ:ââ&ââ6¢ ÑŒhA´"î î"î!î# "ÚíˆGˆÇˆ'ˆÂ?ûË!…i”$ m|Ådÿ”NäJ8DEÇéÔD™ÎÉq !·¥’yÉh OJ;Ä¡T™N $Ç1x•è‘ &“›v G3hÜ”¼Ef“Idz$É‘Ì gGr‹Ä˜ÑTJçuÄŒ£ðr`²Ãy•Ù‰,&™6I´ñ <)îFa†“©Ëñ©H:’cÈqäDRP …Ûå1ÉTnqG2=šFޤ$ÆprDÇE$G2ñ™ÉžœŒ¦²ã¸SEIj·s!G$þ× ¤Q>%rÜùLH® ¸ÒŠû s&'F¹®,ìƒl4·³ñT*ÏÆÓ‘Ô$ îœ(Í­+„¾³È<1˜‰1dÚ¿»0¶„IS›È¡â»œ°%…ÁÕ‘¢5‚9a“HpJ¤)•È].-’‘„û”ƒNÆ[ŠúŒïϼÃ^¥‘>)ÑL¯9“7@ÒÆw‡6¾5ø ŒÃu#⋼€wNùRâÉ\›–ð¥°b(ÌÉs•ò¥$“œÈ´ µóê%“Â(äIþ"æ‹ÌŸIFÞ1ELø ÔÉá“”/ƒÉŠ!q‚(îñø'“¤#—CUÑý·Ö…Qy2™ëqÒ¾ )‚FNL¤FQ)\)çÅá~%êGŽÑ ù!yñmÉ ‡/´9Efã³úkGC]ü™wPA? “Ûƒ„:Š©ññ¨:.!ÚµyõCóaGó”ëǦ0YŒñƒ•»XBþdoaq“LŠ€Â¤†q{¡0ÑŽÀ‹5P„oÐB“#Pñ@ò„úè¸é"G bÄQþU~¨$‚Ï5 É@2;’úçpäy{à¸yáR ¢h’g~”Ôˆ †ÏY2JA!/É›‚ì›k T´YÇ3ðXAe°\d&²]’‘3›…ޏx”2ænK4ä±ÿnK¢ãÙ“a™Àñ ‡ño`%ˆ4ÇóÉ@9Äp¢,\ð‰<Þ. 1‘ÈæíÍBñ¼X,I%y£;w<€d™‘¼¨,VâUúR’¨‰x]TäÍŽÀ;WÏ@ë€bz’çHŸ<_Íñò´("üÓ2|ÒÄÁ;Éž„& 'ˆvd³<%'G¿“‰Àdd)xŒ˜Le¥Mø2÷| L÷+<š¬iñ òrê_î"D¦§M²GÑ òäá$‚¨(À4E‡‹+’`#7í ´q᪠B¢S#É‘æÄ'ãõD‚Øè¾ E8箿IާÆï„&É¥2ž7±NèÈe$þ»M±“È4Üσ£'ùŠHpìøMW^…`:•E‰œPˆKy®{iN™2åH7’âä²&U f²'mâÁiᔿT)B¦£È‰+“Tò°46Š6бíˆ×¡RXt<ÔR ¡¢p7t’Q°Y ×zˆ¡Héè’ûG n(¢àj=tüyw…Pîþ ÂÈ“îDèqáäðdÊ?7ÉBöUÂÄ=2~çEŽ@Ñ1úö›0ñê)Ö9kUŽ]‰:6|>Îdkw&ˆo˜ÛÝ~½ju?a|F¥ ÜýY1iI)öý¸@ËþôxPöÊ_ìÙDlãáiG3±ŸõêA~EoaÖ]+¢;Ì×½ë?¸ ´’²[ã½AÊ´]Óá¹(´|«8Ó,¾¬³~wÐ{|W[ÉzL”Ôž·>µ‹=|©¯-@»²'xÇ/Qìg­°AÐ=ì…±iv&¨Ì“µXñr1Ö·îÊáògÓ°áÂ_ÁE  r{©ÙÒØåÀ_0£O‹& |†¸GO›È°_¤» Ö¹A©òÓ!¬®ZâÚî;·°Ï®¸wžÁÞ¯¤G½ÒìauJb…“/vßhyZeÐ30qÙÔìw ôе>¼ ïÁNë vnÈÄÊâûÓ´´q½“íÙfAíñ‚èsô1lh(dÖm§.¬Å¼ØeŸJv슫úEÑAì?5e»ïÒØˆ[~ˆôÛ7X›°¹Õ·¾æž%3Ä–½®%Þ±ÂûU^$›ûôžUï\“¸4~{¼ÔºÔý7›³”ʇô[êla_Ò.ªQ‡>ð·‹Èy.ž B ÕžénÛ±c?c®¬Íí{™ƒñ©ÎØ¥¶顱gÁDu“Õ6­a¬j¥WKÍL0Lp<¾:Ölþnz¿:Ž- …ËéBC v ú#¿ª=6xúÎöB•ìÎ3?õ¡Ù@X&Þý9u´Û?M¿÷!ëøoaëv¹ã éjÜõÄ3 Œþc-K0ÛN][žM[Æö¦_Î(<'k]\oRs¡ kl:ã2f‚ÿ³½‹ûö¼Áš_›™^H¸‡ íÏ;Ð~”n~K}pß äÏÐŽå‚°ÂqìçmÛC·*€Ÿ®w!` hÍ d>Ùg¯·5§€VVïõ'6ޏ>µß”Ö˾òµ»±N ÛÄ@xIJrä4èìŒõõ´ñPe_ZC'Lßw]íû±ùXÛ`O…WÂMì•LñiFˆôÖ×ZÍü †§’+4óáýÚ©¹‹Çkb_-Öÿ–߀Ý2³1sÿy¤½–äZš¬¹'_Î%øØb~ß›5vbÍß´–.Í•ûí²ƒ]Øà W#ëé@4`_žµÜ¬{Þî>Ù5s¾=¢ü‰õÞ²XíG•³m¿±sîk/fhí¢X-ïšÌª9Á"ع!ƬÁy ¢«?ÓÊþ06æLo ߆õçèÜÌ­1Aûs13‚@†ß¹ŠoËaìë“å+¦&œÇÉyß_+>õc&ÊÛ‘wŠÓF®Çc#w2²F¨XCÓç Ê®ØíÒ/:|9‰Ø×‡E_zûÞ`רG+®:‰€ž¶`Kaä0›ÑhSœ0h¬1¯ßÙ‚½ûíUyR=«Zå¾âláIì?+3E7¾“`~‰6såY|\ð+…3,ð´µµ%ÝËs7öI“Øç\{ô«~ލõÝ!¶¶Î9£Ø›×ÛÚ3‹…±wýCsº¼Å±Ñl­´ÙÄjìýõˆ]ß/éÄ;Óëu@|ôòþë­jx¿®ü4ÓîÊù ûx/±îáR0ö¢ï1ºu;­’õ˵ø.öMôP¾ê4Ü^_Þ3$–'èƒ%1G;ôÍ6ìåÃRݘå9 néeuøîP½S˜œ‰]M2ûàO+9·'Qów—ã㩦,þ6zß4¹ùüö8Ï«luÇvJºMÐ¥§tÖ 6,tš æÎw<–ËÛoëTï¬Äú$g’Wê]ëéߪ-~ã‡ó¢úðz|ñ+Ž&µÄ-å+®>¬RVB7® 'bïËG†’;A‰Ø¥PÞ9ˆv.ŸïMáL/ªb­Ø£¡±´Šgë°«m}úbÉÆ@Ôõo•–À®}‰Ý•qh%Ö–põè—aìsjÈÇ{wÁqûGo­‡ `3ûØ*ó9 ÇêsËê¸FKÄ6÷LGû¯xdVÖ{M\oµ²" +ìAUòðo­¸Z°fÈŒ$ÞVÐ&ƒ×mú`!4óá'7à_²/é#- düí>ÈÆ^¾/³Ü4ßT.(?"múõñ$¿ o=üf;­ò7/>;‚}Œ_ú1g„5[äÜ.Û\„½Ö‰ØŸ-†qÏ5X›(–m»¼ Îßíž&i UnÑË…qýõ^9uñæ¨Ìm÷ /‡Û¢ïkNîʆ&#öbUl!¯ ýùjÀ“Ff´vÙAøD–[£¸Y úcv2¡Vñô7JÍkhç¿°6{Ë4h¯.ßV¸ël²&÷v>·‡;ÁÌý@w¨9pCº.í1¼ñL}¸dõ*8shΑ—[„ {Èý¡mÛøö+ »ûLŒÍÊH~¥ùQ ­Ï–@åè-½ÛWipÙ¯£(ç·“·æÙÞ„žèeÁÚNPç*>#ÿU¬»Ã.+®ytùc³4t œf;õZÃ&Øœ:£>‰¹òOT>.œ oéiñ4 c|þ«²ªO2W]-¯6.ªÙ‡ç·¬P|¸knTݘ޻ Ëjç\Vüq²ÏmfŸùVìž ¢œP8þ3Á¹oß\É|,1°Cò±”c+Ãô %ZéÒöA*T·ì5:¡Ü‹÷Û­lXF\fív`)j?ó}HÖ€B‚ôæúÃp‰–æ’µ Î×R{VùèeGëËPÆœº7ï ”?-_XærÅÌqÞö2ÎʨZXÃ}‰xû’Œ)dK[³˜f}­¯.¿º §ÇÞ/XaòúZ˜ !ï‹¡7âóì"†.Ï,âãtÕDÈ0ì¾¢à¿ÊÕHK6üÚ¯éüE§œ`WËU–¾ã%èð=±y¢zWz»”eµ_éVàý½8t¤¼’Ë…£æ—¬…ªÜ›Û;·9˜^ñ¹Û-2»›çZÀÈá”áËWû¡?²'®Àp6œ}<èygG;¤g9?Þr5 2âf^Uê¶„k‹_ÜÀ̤á9çí½ÎÙ°ñS³íµ5ÒÐþàcȽ–8vNz`aÖ+Øù®Ü\2M ®äø;¼9Wùts+îñÁ‘8÷;òªLØ×䡸_)ÛA ÿ² 5­ ÊSŸ}½¹³j ê %3áÈw¾½=ט­]©@¡áz¨:¸w…Úa ²êe›œãÛàžÑ±mxùË«! /8vÏýmdVȆäƒ3â¥nÂÁCÞ*šÇŽCMóì¥|¼þPîÀ‹y¥¼s®í„âJÉ«; ŽãXµ{ÐøeÃòê÷$hœódƒDm._lþy9Ï*4Çæ ;cïw^Ì-†Ý;o_Ïn%»óÏ­áØTŸüþB9hj;æûývÔªŸH›zÄo_cXî<ó5?ì¸Ù+°·2`ó¾”j¨øyíbõ3ø¨:ãmט4|O7nhé–ÆÛmòhôŠÒ° ºÄÙEF’h=WméKª„­—/Ú°5³!ó…õÁm78½æLè£Y¸ßÙÁ_rTN_}á¼Ø9®íœU)láçm\½1ý9Þo…A;h(Ö¬Xãc-ß'^­-RÉ„W+ŽÈ]^yzDrd•Y~Ðç{aßO¢ 4Ÿp#P‚]ºóy› Ç $Ý–îë…ë|^¯’ˆæ°KíÔúVCž]ºKüö¹Ê®ù|=ø(ï1ìÛM~î E[ÌãsC†![0q}MôWü\SaQ_æuH‚`TT´¶¿ˆX.rûkŸÑŸðÖÙdx,#!Ôžy›-Ñ$¤ëÏ~(õQØÒÒR%{_@,”¥6E „“nÐøp;ýÂoípâ/ìUqòU@h5 d­y ª˜å5Ø¿EÔŽLP2i¬\YGÇFë;”øY?p?´ÿΉ7 rö<«:ûjƒxµû“Yt¸Íë’÷—w pï×)¶Ž4öQv‰Ôðþ_ b‘ú$óä/q·úv¡ß$âO0$öcáßÒwyƒš’çÅCž {Í6ëªâ~i»¨;2 nöóŶ’@vëy%+@L{VܺÅnø|Ä£ßÙL#aU3¿­Ñ´àÅU ›.×üŒ!9ËÈ•@máÑ>­á liyRZØ êlrÞr3H,¦W˜O5"ÑD»2™¿A.'gÝû}Nõ3Ô»2†A˜–²­yš2—tü·Ýå-mKž{ç\ÃGì ÔÀoàrÏkIuÇß7øóáíÓ^ŽÅöVÊð§î3ì—ì(Ñ]² øj|Î}…;‹ö¾Ú‹¤k/÷¥f°f£ÓÀ 0XèqèmW$¾¾¢Q?ŒwîÁí¢=×óm€Pêø¦åã&|ž‹¿´u5ÈùmÉ ¯M—¾j])·SÆO†ÀD9÷¨åÝÚ§G®Ò²ø<úR™¿ãoeâýÈî7:ýkȸè] †Î éÔçÉëA±~ï°yQ (©øV]œ* Ò«v§¼ãì°pýÏØ¨æEyûJlìnè5ů‚ –{rýs>!P)ï|ë3x¤Æ_¨xp4§è§è…ƒà9ʦõ|¨½?¿< šïOÕc/«¶úÿZ‡ä¹ÜInp¬ï"Jý8< ‚éÚóo€²ã–Öƒ ÂôlÙ(û„Þo4\ŠÇ¡ ÅÒÎÏLÍbórmÁ:Ðuh\z³x(?´-Tø$ù&÷ú»@åI~Y©èu ˆEnþ˜$²UåûN)*ÊÈœ}«ažXõ¥ƒ:¼ýMÅt´Wøt,®_¯4Öš 3NN_yd‘ðר25 ı.æ5—ûVüò\š‘£%(š´¿Üð´cÍRþ;<4Ç^ñÑ`ŒB?;ìjη¢¥,ï ÞW÷z2š+„Ц׀¬i…ã%)E DomPñÞªG½¶†edàrXxƇ'nÕ±‡ÏI_”n‚ÂÈ.âšoÐH‰ëNö.Æë),ÞÐxòË:Xú&]åtösßÌ£K^lÆË…²kµŽÊGC’kµ<𯬮©ùœ ùتõØÈo½…a¡¸|&Ìõ]Z™ÎM×°u°Z×ÐèÅÔ6ÝTP˜&Ìš&œ†{/vî{* ú!•î³W­¹%%ƒù×õAõ®Üy+b/²GÓ³7îÏÑjßE¯këÍnÒ^ò¨þ72ufrLmPÖJϽ‹õ¿Þ)D\sä¤×ùÛŃV ¹TuË~Pž2»zQŸ‡¼_Ê‹¶C² ûnؾ±ë#6jºIÓ[©¤E¿0¹ Z9ùB­gŸƒª¼üaktßÒŸ0_o/\ëé]Èë¹¹hÐzt-w’§\ÂËÅΪ¾éN­UåJßóO7·V˜7|S¬ï'Qζ é®í#U÷؈–Eï2ï/Øk¶[‹¹?®§NQÙÒ‚_õc+î“ÏUE¹Wúi~ì»ÊJ÷•?AȯjÊžµØhwâGÅéø¸SŠŠâ͆4@k߬9G»Ê±wåëïÙFö,Ý*ø™ïÈg/gw÷ÁpjÏÚy‡°ôⵡ¢Ó7UqÐÞ|8ZJà˜FEEü|øï—Ô§Fì*ÖÝù§4›1P@ÞD“º m·íž¼ªÃƶ 4ŸÙ3‚ —lùì|dÊ„-ßFæ©hüS‚}Í êùÜ ²çŽvħl•A¹«æìÄû× ½‘ùÐ䈉]ù#8Ú`fêÐùûü\PJZ˜\cú ïn/É=÷?×ô¢±±5|m ±{ÞžCOÇ€¨ä‘Ú$çÏíïwêgÒû·i q„¶Ô#- t>ÑÑç%hî~öëÐtºº@|öa§»ÿÚm¿ÄuN?ŸÍŸ ZQÏÖ9¹±Bµój}šõ5ŽôOý˜gúÿY.RÕ Cà¹×e4ýë't8ç´iCƒÔtÊ•tz6½×è±VŸ÷\©ò¼:ºûU^ ñóâÒ»Õ•ï€ÚæäcúÊÁ@ •Zóݲÿ™-mö} „޾Ýi ›ĨÜr½•*ËÇ@ñ±ÉçU ˆ¬Ë4ø4"ÞR+ìšq¿ûvW‡µ]M¬Aª”rÛÅM‰wÞÕ îˆ'€â¡ªGÅ@ÂóWän~7l44smÖi0+•^zÞÀû>ÓçÿCd/Ö´¬+š _¬Ø+ùøÍ6—öUßú=‹û5=qÆ~$šë¬¨a¹[|ne‡@ÁuÑ—á¡3  úÜ%hÁ, ½ßƒ‰\9Žæ1À s¡€žúî1õÅá °N°0»XZ}óù³/óüô­|Çö´@€â³ŠwoóîǧèǾKwû Á%Ãk+:Ð_ü!ûJ0oþóf?Êÿ’ òÓ‹5\~¶âȽd }s´0m–îw‘°÷ ~0oØ~/¾uÎØËåíßÏÖäªx±aÆ‚š©=B] 0}}M Éeµ¡7‚¤Éo|¶¬uÉ|AkßLP”JuZâ×2‹§É¼Ð™‹“FK:@»zÖêòüxP¤+,óÖëýO—\i}ÊÖ§«ÓeËAðÀúvûg¼øCâ$½â~Ü;ú”l”ì­ ššÑ•z^êè ‰²¥ã×@ªL>&vWL$D ö®˜ò?ë";OðôY"듳¥ÌW}ñQ=ך¾×Mï{ÇÜî[6±$Þ9 J]»m5:÷†u+g7zÂt±$ª÷u^¿Ùè³A$Üœd–>{†æ‘òÛâH»Ï7_¯ª2›ß»æÊ볈3´@ìÏ>°Œvÿ¶öOá':NÝâµ t#iViÉT¼_¥íOå­z FR×ûø±$ýæTòâ Í'Aã9¼ç?:åSÚ×€DQÏö=ØÇµÃM¶‚ wöæá‘¢;  8\&º¶„.ÖÙŠ‚ÎfMÍîcÃÀq-ÁoµÞÜ¥àô£'h@šDT­‘M|61O€þ©ôÅŸ–w¨ëûSúWóìëÞözr On—5ŽÎý†íáwNÉøòâÈŒa%Âu w•úëPé% ­E7Aü‡qÕ#á@,x”Wð(äDÇõ©R¶‹ÛÒðçØïcÙ OZAÙ bŠdB:m_ms¥S7Y¦¶e ÐÔûó;ï9¢É!÷g×B–± $lï6ò- ×ù#y#c ˜J^(j«æjÁÖùõ Èÿ‹ÿ—ï¾S&¬r¯‘à3ltH§ìTÒ ˜²äSѪ`^œ$?-´[ÚtS§Ë˜·¿·#¦é,Alû›ö‡ŒhPëÚ¼âòP±y£Ö/¸cÁ‹‡ÄÒýwì‚6ÌÔ8;ÐU”çg™7  ó ö‚ü£êï ršÑ² Ck-n8>|ʼnö‘£åR`ôìë˜à÷O`L¾Ñ |D¤ÓÖj…Ÿ,ƒ©cïe¼žÉŠž›B@U¢¼€×+P 4ØW,|’a&2çmÃ@õ•5íä÷ƒÿú.—å™psUÐz¬+¨šÃÓ‡Üçgë’ Ÿƒ˜ÃLƒ1Å ™ ø ‡ãëó~êôÐS0[YÅpm$X–3tG×HüOÿ\½Ym6xæbΪÕg[Ld@Ü&äéîNü\S˜ðWpϘ¸šüã¿A0‘oš;¯{Ndxrê{pê'íÏ©Ø=3>åV,’ÈÞ œr?n96–˜©) Dn~ÃÄÕ題@ÀË<Š794@‡¹gÿÜ8‚láKB|Ì]QÌÖëš38íær®n/·.2ú~Hœ´ëÿ÷Yš5àÍ-ÿOݰD'ÒØ@ãD}®Ü¢-u\Ayü?ÿÌœ |ûö;êÏ׋§ÖWíÓïšÁÉàO>8›¼»s®÷"(qÆùj5!¿ÒD;pÚcl/òSˆs?lÞ½š¢r‰`.7-m„÷|Á›.ûµ#v ø— 1·BHâcǨ + ÁÕûQÒµ=‡_'-~c({ö4Ørö;Ë×ãräxÍ?Ä™?§^Yç{Ÿ+A`Çé'Ä4~Kfº²Òº¤åÒq{àêUÿráà´å}€!å" C¿ÛIÐáÊõÐJüØ ðûg}¼;&Æ3âèñGßBË6Îòº §·Ÿéœòɉ«?ŸÛ —y› ô»~ÿÙ½ §ûXÜ>þmЧwýj¾û›Œ•áúôŸõqï×Úiàeñ>é…X=8Öè¥L¿CàÕ¤]y) à8±Þãqõ¡GøëAý;i•–ùx¿~ 6,‡B`‘ñ·üþŒ1‡˜:ÜOd_[ íßf‚œ4Ç¿ xÂaÞýw77I×@Hæ[«Þê ðU#^z¿A¼üew(n˜˜·/g<N?ûDzrnúÝíΕ÷/‡Ô‰rÇ‚mêE7`&G¿\û ’Ÿ˜¿ÑØïö²eeàpÇuäöË;à&uå×ÒŒl›°Wìóu/£Ö+äé3ÞÜ9ƒûyï¸yh>§w½íè iÿ¹ÿÆ Ú:yz ]ºð…ÅÐz<íþwÿx¾Ë„>€»6WÙ7Ž€#ŸÛ)Ô3Øsò}‡‹¶}ÍX^%c½±"¼?޾ùÜòå/-Äí›ëWœ«;G^kcçh›)«À­a"_—£gUÎuÇoU’ä6ì§¾Ãßò½íëÑ¿öÆÔ‚’= 2Ñ?ö3ŒNèS§éÆ7™þXž]qü׌Sog\×õ‘91o ŸVÑ&A|?³ÔžÈµpK,ÜYÁÞayÏ‹Ä`&g¾Se&ü†Í:‘1÷®Õ5|’œ~l9W/N=§œ‰r—lª•Ä;u®¿áòpý}ªðgƒºÎ½`Wåwþˆ‚$rÆ žïßËh9 þôŠt ÉYàq¶ùÓðº˜ËÙu8vî¬1¡‡ìñ‹18pärÞd\°‡„w%Û²ÁôîÄ<8ûLŸhOàžk6™oöf5‚õw·’{Ï@{Žhñ +~Žvß0nçïÇåç7 ¶™²¦ ¹ï}*øfÞ«——†:ìåRæ»{ pÉ*=×p6|ciS“å6PÌë]ooËþÊ Ùʵö !FÀsë-_äi«X{cßõϳæÅ²A§ûüá*“#Øë±oøSäóTíAÚ’2óxìâwzŒ ÚG¦ä=óÊc‘Ž«º3€´JaÔÎÍ„.Í ïÉ…„¯ —"Žc=jÎ;¿Ž”¢eÊñ´ì],zò¢ó'ihY@h-¥æîÈ×ÇÐlŸÚ†ý0+Ï>{ÇœØHì3Éž˜L¬ýî—üÙ«À‹QÒ¼KÏÜo‘63p=HßyIö÷ЕSò¬³juÊì; µœe \K“ñUÊz-‘ ![2@­.Õ¢ôÍõË0ûtgfå“@<ðÆ­1äÇ„…ÖÔ¼ã=ÇzæXI¾3¸¯´³Ä†6 T,k­ÅËMj.ÙW¦*uƒ²‚nƒî«Ûr»HÅ`*Z­Å'=Œ…_ ¦½:÷?qŽŒ¤ãŒkÍšà|.uƒî|<_¿njÿ:è¡ûGƒ¿ea¿ó|’ß󞿽žúL;²lö˜q´+îX”?„½ó+T¬}` ³…nìúæ  â]›ÇøŒfóîSj-ñû>0øã[éA+ß °²Ö³Õœè×fy$_vKܼ%;I€aGÜttÁ oÊkzT8È®œ~Ïa7Ú×›³ EŸÊáý™[\aÛ׃¥ãÇ[¶ÞɆmñgúò@ü¾çšùÑ»ÀnfLuÛ–‹`šuCáp‚o}vøJI]ò]/]&Y LV²NU­ÀnëòŒz﹈ö]a'‚ú*ÏYˆµc?¯}ª³ÛÁ{ÿ$f´©fÞ°þ¸àôعp")oEû„§è»Mg±Á%>ËB–àû!6$«±®9 .Ž}S&=?ú.f û_Ôœãò»T.A‚Hê†Eèç½71Óu¯ý7{Ð’öž« Æ v„äÁo VW²£nÑu ¢+¦æao—T«–þÈÂj׉ºiÖ0%øraN˜ÈTïmR0±, ¯JÿŸu™µ©3Ìt^ öûLnFõ2!˜îC’œKˆý¥ÊO¦û†ßu²ókµ àõY¦¯ì@v0P‰]¾ó ï9L›Â~³ÛÀï±>—úÑóúÔñZ5Vz6"qz!<ÿoïojÛ&ºÖ qϱQE·µs è%ÎG{ú÷œ¥«ß=ª° lãÁ,ïÕÐ=þsÓ¬ŽP×$½Mªû©´|æ!ùjì ŒÖ%¬z‰·›3ÃíF|É`½;eÀ ­î·zª€ñ’ÿ´*i˜Ûx{g˜íZ°:ä(¦óë¦ÇF\ˆj~óßÅ‚ŽÀ«ØinÕÿ3Ÿ9…Äs”˜-L±Ñû:#£ËÞªÖáëÚû£öÖëkØ÷ç=ì,Лsc†½³!XÍj»_7b ¶¥rûwD™ƒâÅü]Ÿ6ƒÕöÐrÓrÞómõ\ß%Á?—‚Ìâè€?﹃¾ö»&1 ˜J-¸½ÃhÛ¿¯OŽ¿¥Ë}}R„û~gƒMgá?`pšgÉ}_ÔÍÙÏ,‚OÅwð''šÉHfq'F¥GÑ&¿.J¥³(L:…û»Qr´Ùøk¯œW9Æ¿½)If‘Í¢˜ãïÿóe4ŽWâñŸóg cccãA×ø«ž¿ÿßü>ç<ks/data/cardio.RData0000644000176200001440000006066413265504354014001 0ustar liggesusersý7zXZi"Þ6!ÏXÌä>ax])TW"änRÊŸ’Øâ…§ "©Ù"ÝȵdׂéV¶Ãy\í“ «ÎØiíÃD6çqCÞ*ËhÀ›³i<$ÈwÓFj›e˰\Ÿ¥+qÐÀy'1“³z(ƃ:—<º:ƒÇE”7TtÉpƒ‰ÄÇá²¢ÝÇÑ´¡¦àW¤º† ’£u‚‚hȊѬ›…¢nSŒÄ¬0¼­PnH§²æ‹J!‚¡¸ŽD…´àáAÈõŸ6 ÝQ}šµ'jk $hŽqë¹ë°Î;¿2e·dÅàÿÿ;ÙmN˜‡@«38¦—Éu Xúa‘ÜX¯|6 %öhYÖhÚ*žŽ ‡›Ìˆ)FÄéž«zÅp¾»0È(Òݹ ié$¾ùŸ\L¸N¾Ç›Úãš—qLîyÛ]Úˆ”ùŠÛýéʘÿ²ãá<Œ‰awKÒ¤¶e¸ŠV údáõ¶~‰°œô1°6³ôd aTëÂOÚ÷¶ž}ñnÍHJ¾a>^GV'q›Ão‰OöUŽ ¸ËìõÓÚƒ]sÄ& ¤fX ´£^Ò/ ÖtI!þ³lï·»z¦ˆ"BŒŒF6W¸$|‚5ìK)mÍ‹m®~DoT¯‹pEÛ¬Âb?¤ít‘7ªh€h#Ív™ôè5m×ÄO6J-JXÖÞ›½7ÁñÝüÌ Y >Ÿñ¤‹¥ùJNãåÂÆCÜÀ @I¼.iùiž Ï =yp¯/ãD8²£ªc2¤ZM­í_b=¦1˜Ì,/X×ò¬D‘îk‰®Eøo_D~|Ö…óÄeÔùü 옴JÓ7Í2<@ÙÆÈˆŠ´Ã5^¯êƒKHò‰tf“áZ+:'FžÖº )Äû¼&Àóª ='SZB8ãïññ ÌcÈ.òI·´¬,¤| tŒg4ø¤Ý²\)ò'oïË覎UòÞ°N'HŒÍã‹ð˜.fw,…Ü|º¦§0í4WÓh†biì‚ù«lzsoS_«‰Ñ‰­ð¯`ÍBà&úPšñ|ò_³šþ‹ÊF…Çßá¼T=(ß𷵿…cî{;¡­¥^&†ÁGŠû|êjqm}ÂÓ‹ò¯„0êAJ»¶O B&Ii¤ƒ@\–ÛÍZ¹W Èåü¸9»‰ûæýs~G SêÄ/Œß‡:L'ßG'¦¼5AQ“µ¾QéB˜llÉ®ERûl×}D²`”il^;¬Mpæ+YL”£¯¦‹´œ£÷8õiŽÐð´‚Z@„ò?ñ½=Ë]Æ8Íw¿ŸKx³?€Nˆ^1Tiüh'i”H]SJ7ð“/„À wû£ tÙ}KœgìÏí»NU/Z`S±*“±\ValruS_³qöõ*VÝ¥#*ª¨™žf}½”˜^}WÄ£dÀ¥ƒyA¯ }n±LbÌÜ!@zlyY¢Žò±3Ð߫ʟòX±›S†D>Ô.fë2òj¦XŸêµòXKœ³ø.ØWkȬ_ÖiÅ8òžF9£³’ŸVaÈL‚R·VÁÑî8ÇgÚ­BÇy²Ù/zxã‹}6g'PdsEи²Ú6h¸É7bfÁÎÇî¡wWþB†¸£¶å(‡½K3ÅŽ6˜t-IãÛovÒð¿„i ê“ùÁ«…” ÑUOØ íqwÌéq¶¶Â>±°ãK>¹ÔxºíC=Ë3û—„“_à YnlØ,ÕTã,ît>lußIĤGØ`ÕKÁD5_¨ ØÐw н¥ÁÑ*Á±ôu!g« Ø»µwÎpñ •Klê„,›0ÏÎc.Ð ò€õ·!Ö~“Ïš¥X&M#d7 ùƒöûÁÝ~ Õ_ÓG`Kx×gÞ~&­G^³·Š…øFéYÙN h`"ÖÿƒÔræ=ý¾ù‚·j¾=1t/éÅù — ôiRB¤yØ»­ÿ†LHD1h¦¼ÑÈd*d*(þS~܇,Ìc‰’A¡uàš9ØÐb0 iNhhžÂ Ú7n¾¿:þÔª—˜Nq oš¤*ßZÅ·,Š µ»3_PgÓÝ]¿Žÿ?X²ºAƒ°ú~‹H4n~;wý·ÿÛƒQ£ÐÅõð;'[9K©†ü“ªzü¼ †"nrKá“M>ºÙîÕ/rþgzJÈÿdý±*ó HQŽBbFΜû”¨šFÞÂW6渃îz²ù¹±¸—ÓpŽaOGúdÀ3ÌË­ês xˆø`ÄdŸEŠÄ2F‘ù†Ôå»ݽܵ‡:0‹ ”w-¬}žžŒÉÀ”tÃPƒ€u~_ö ¸¡£ñÁ¿—¯´éœx+®G޽¦À¹;a¯:„7Êcõþ•ˆ¸-ÑÇã_JcLÅ?¨ÍRa S‘I¯A8°_.auüßAn¾®nן.É‘P´ñ”{*ZÖ¹z¯Ó‰Ó—…§TG>ïx¸%&he©mÞ¥E¿1#@'ã\âú0É,QY–Œ¥Ù?è«V$ƒŠËu:lÆ$jíè*ÀÂ*þ‰oèí\::kè¥~yÚ®|áô¿Ùì—(8î+ׯѮô, ¡ÒE +³MKܤ)ϳüX{}V ¯ @ÔÂA¼˜$ûŸBa±s‘³I ŸLýû«ßÖÚÝË[«èËLÿövúôÙÞÛmŸGœÀa;ŽNØC;L6DýIQÓ´;f”h<ßZ›JvíGWT$«P³U:ÿHl!Äßs€÷ÆÎëHƪù§ùP{ ”zb.%°T}—ÖþjåMÙê+²ßèõŸ²b¥ª…Ø´P÷˜¼qŠÇš36ʯŧ3A_ùþÝŒ±p0Ú¦0QZ{¹_ß6¨J›ðe“äЖ5ët§'iqyÍ0ı>Ù5¢Ù®Þ””‰4)0Q¦8e [² ¿„ÐcÛ›Qøµ¤*Ù |&B\ÐiT™’HíÞ¤„íx}—W‘cÿµKSP¶ ïžAqòùá^¤ø‰ÅñPÜÁnû´…†NÉÞW™ÛgLÚA”6•(òâqÜ.öœÌ2¯P Ò‚žú+f@Öq+ùMqS“Î`•AyuÔQ©l'ÒçÆƒàÌqNHPJS¯¸¡Ì›Mo Á>ЧbÒ¼pãîqóðŠÅž(õqêÃKÔeJ|*îX&ÀøZÞ'’þæ’/2-–ìhš=š`úgDÀ½©÷öéd>YúÛ£Ð>[6 ʃ„ÿ  =ËIûcQé-‹±—!|³l≙¦rz½Ÿ§T­¥ƒºšæþ‰H*Á›Á»IܤYì͹#¾¢1qQ"¸+!y…0o-û]Å:Ä‚6  Ù,\à Gîå Ð$þqù§[YÊ ™««>gº üÜxIÊ ŸíÀ§Õ‘9øŸ0ß²´_ãmàáÀƒeZS1³šýö,¸x_½»¼e=%Þ•*àm×CtPÐûûÅ]Þ#<öÞ_•ØUhŒ³FL…X9~VÂüv1È3÷Älò/]ÍWo`}´ê{ rC6ÎßänÊ´A±àÂhÝ|„¬‡e¨Cðïws7ûÇ>aÓ"|VS1ñÌ߀Äx)ƒ/%LêÀ^á;Ÿ|¥Ýò iÝûsè÷3K$Ñ1 Á£˜,œHyã1M¹€ÂFË•µh¨žzœWX‡6šA ñ F”dî¹¢üá½i/®o¯šö}kÉŒë’õ9„®Ûåøˆã<´"l½d‡¦S–»€Ã@ÒŠ(zúœÕ‹àoO+oÝ5çhd0ð£f®9~vÒ{JQ—õ¹$|7ÐòYƒ z’ÖÌ)oŸîšâFB„½…¶§ÂíÒ¾ò Ìü"çm5¾MÚUͲîHÎc›ë)ÃÖ¦ÑfçXÜ?¢Ò8èè˜y‰“Yj"~ß–50ÇR<ãé{šeÕ>¿4‚7"ÚVÒ8ß÷ï@ š×µèÑ[Îoà­—þÏTײ2:‘»èT3BsgÔ!ŸtÁÁ‚ÿJ~®Ò˜òù€Å÷´¤÷Fm3)‚i("ñÞÍ$ÿÖ¤óT1¥¿N°ê™µüÆU¬mPòõ¢øÄF`ø ž٣¹WÒ ·7ª9?xuXés9Aû¡x¢Èœ­:kGÎ@Ç¢ eâ¡¶˜›D‰ZÞñ­#ÚxèåÉ"¸=#Fuì’ N 㬻ÄN†-(QÃÇòÐüˆ(ýÈ…ƒŸ‰S?¼NÙ’…YC ß½˜½Æ%¥­§¿g„¦d™aN›®Ó°k ÛÌdö­|)_j‡‘¸‹{Yˆ7ç¬Ë˾˜ÙŒ{:.‚±µ‡†ýÏS½ä'A°àÏžbžÆµKh…ñs 6¿’¼Rb ÐJ>Øm-¹Çj$| ê`M|›ºÕÄæÒ°PßÉB!³z$$FHÉd›Ú5ОI¤ˆ©¸H䑅鮽Y‹ý¬gùð¥#÷8NœB†å¹µä ¾ ‰„Ú©ÉOùWÍÜ Y€1蜲è—bsYÏnþá¨âhα=’÷®µo¶½ïnŽ.”š_º‚žD,ÿÌ”1ZpAéÚ¹†Úì}¸B·‡ìÖPˆà` }A6‡Põ&]u¿½³V™cÔ"àÜXŽu¾m±Q=5 ŠFV>uwhá“^ìNP7dg šÌ_ Ù÷P(ä×ÅÓÐÉ ²IIøÎ£¨ãtGûsž9Ý¥žçžÐFÈ%®4ý FC`'_ýU<ðiËSF¦wšúš§íYƒþ•¹7¨QŽÕ}Ïœ¸b*¾¼7øë 8•F—| ÖbË3ŽŽaG-FÂôîÜí÷Às$ʪWßÀ“¬«AªpÎJ Èy“].ògþŽý<» á£xåP% !“_Šû(_’-mlÉójmôÈ7éº@:x7òKŽó5¥?J[Ë³ÌØ&‚v'y~ÈÅSÐ*çyDH¶ª £øµ¶ýÍÍàç±T„ïߣþéyôµ³×%èºìþQè~ã:Û7æÝ'ÛÙe{ #éþ´1ƒ—A|‘4®fÏÍåö‡%GËÅ- §´0FdIÅa»ç=/L§1&ùM«·Úö„G‰ÙÑ OX{:Òr bUmn Äz6g¬r³Œ @aœÖ;u &‚õ•¯O‚L­†–K\/o|¢BÎSxüN|ï›;=³¦±ù} ùÿ…øÿ‰¢ä¶¼ú®Í €]PW“ðþQ¾»s¨ó¦!}¥·,=áÒÕ‹È]Žæ½ùxc]Ë6^ư}ÉýÀåõñ\ÅeX>Â4B_ÓgÔÃ=Έ<æVŸædLu²õ³ð†›Ü:©¡x‰·dâQ” 9{LqÅŸédÞ|ËT¡“Ƀ™èÇÜç=e¬÷¸,IîÈ4 iÄrÌî4•¹f%×Á½V±÷·&N+ÓNßÇÊe¢Jcñézz¬{ü£ÏéBGpV›.²X,!õÓì¶ 2¤q1Í×ÂåØ$5PË+t<_4CýAbÐû<örø'ÜnÍ1áû“³!Ó§>üy•ˆ¢)S‚(A ó¼$äåRUæ” iH£sÊ®:i»ì‡'!{  3ò*ž7P;ÒØ5 NP/¿r Ûg¢¨ùðD\ìÖc*¹¡ÇzÞâ'0q1òëÛú@F©Õ®iö §­r{¥R×:‡:f‘ë›QÍÜ[ ¬OðÐX¬öAà¦Òœýj”ä]6"–­„Fö+2ÞÓKT)ˆÜƒƒq/çv}Êh7¾Ûzç‘ɯ©}ÿ^i°Û¹4SêCq/؇²'¹môå„ÀÄÑ•#o]Läzƒ5AE6c¸_f^wä€ç1 ïš‹„2³˜çžl½0îÚ¡úÀyž–ß I§Úª¦v©©¢Æ2‚ï¯Ò¤Ç®½lÃIdÃú‘K¡HÖ"IÓF|&ˆ§m €+£!ê¢5Þá£(Ý cÎ^Có^æ¢Å¹ô©Õ+{™ºúÊúU¥¹^<9˜ÂšlÜ.cRè«ßéZV1øJÆ÷:i~ª•Kõ€ydYRó ×W9 (‰ui`ÞÕ6òó´ö lß ;"Nó-ò0ˆÝœL៙!ïÖfçHÁo§®ø3›7çXižœkuUjÞ¸ Ûï’‰f\ˆ¶t¢Jqnêè¡•mbÁçsuUÅwuîhaq'ùBÁr^oxWœ†ï­ÑÏDâG/°ßÆ=©¡a'‰¢RÀ¢«:*í˜îåE:yÓÖ‡â£þEÏrÙª^™”N,©RÕd¥ðnIrʈoîZ€r‰cß÷8>î¹”XZ¼˜Õ“·î¤ÄbħP°âá±øì¸ð4§%Ÿò?¾$ÍæWC:)Øbª¾cÊȹ³{Ý@¸ •_Å“ð]/ðE9©™ÖM×qøýH‹–v&¹xâªø÷c `{¶Ç?ÇÖääI¹D³„„"ÅŹÃÙ,ƒ:'Ï4œûeÛÙ„8ró°’pdhˆOŠ©‰|£÷# ‰'ËSºYFû'™RR 9?©%û{W#×øÿåÓx€²ïýà"m­Óè×êà%‹±LÜ CçãŽtÆ¢„YÌÀ¡UP|ÔZY¬‰—Ë4ì˜1žR9ÃïJ—éþ«`|sýÕ…­ìôàçdúy @ýÈWpxIMOÍÃÁǸõ<8½:–Ì!ekˆ™ùJú¬ù066®V-8ñ@”˜UpTe[ØŽ°(ÓØ|“]%:Ï*Ó©>ƒªÃ‡ëÈKà2 OƒËcÁ.>¾üÈÅ…eC ص ÂQ!š¥©ÂÙEú_*,Ù‡fÆàâBK?NêWä7þ¡„àp6nñx%Ѹ¿Z·¡,{òV |ƒ«5}ÿ_j±S—ÄC}Ôš3XÒ áÒ~âráÎùéÑ7òžíœ};gEëºíx4Ãòö@–¿ïÚbÖT±)ûpVèùß¿°ä9P§uäÄ¢»þ}ÿà×O{³jÿ8?«I*Ì$]⺛N— Q›¹.=ŒÚôº+ÎÇÓ62«ó*‚f—âÓ×<õsžG‚“Î݃8€×C_Ym! Çöÿÿå• è_ý «+H–nÏ–N<ýYüèË~Äs÷˜oTfÈı(@‰+¢€HËkƒ¿åºæ#žù(m þ/mdÒ4V0’üP cee하E£Ñ‹exwVK~£]UÙl⸢†ŸÆ]ÇeHAhÛ‘¼—~ðhw‹#Ì_w6(Qx=çÈÌä´Ø‚@9,^$sð•boV¦Hapˆ¿H, ’ ¼öz`Ó=®ÿý“M>e½ÄãÈcQ@ÞÎ$= Ñ!ë„FnäÙ°ojT°ùx é ¸ѳ¶ •Õ£ºÇz34ÜMWêÓ¼ð«'YÃùÉëHy×ô¤Ö·¢ŠÖåø÷Ãx—®êéY룶$ð“Ë^&y×¾¬x+ä(6µe|MúuÈ/#ê„5¶„¾h#¡Ö§†D)šåZ˜/@Z˦ǒ%#ÊñJæýj¢­Ÿ.=ùHÔñq`I™Ò£ãÄEkMÞ9.ï(˜‰(]E1iÂfsŽ·ÀVÈõÌ7y?k«¤ÿ§ñm—ˆ¢¼ÈyJÙ6G¼ãËÐ9=·Ê”èL@bîÌ.Úþó× 6¥lYøde™ÃîŽr´êË (ˆh"öDQI]|ºûá×aðлÏ´áúÔZ óz8dÏcþRtÖFšòO8Üx#uLŽÆ @KDl9¦‹Na1Úù¥ïÆ‘6¶4aÀhf©KwìÊ"ä`—æùM¢¸¹å¹æPL_Æ€¶½Fªv}¸WŠé¹­EŠ˜Â;#žÇ] Y•º¬ë¸`:ð]‹Ë_Ð}CÓúP@Û 7Ï ž£;¤¤[ìHõþCF8€nƒîL›0ãŠÅ™•I¾ãô‘®-ÅCW7Z/t‚†Œ¢ÍSbæjtMÛu׌~«Ëù÷±W»˜tõàM”Öq )l¼p^=6}%®JÝäÁ¤M¥¹¤|‡¬ùʲañïÊ49ÿ4Ú¦¹ &œyf¨}'v ð´©¬cHÄŒnYÍ‚ …ÁŽ‹2ùn¾¢ïÝSPŸÚ„ŸüϹâ:þx0“Ä p–‡cm„˜”ì@šŸÇD|GÏÿE,âZqZ ¨ ‹“O¤´ö®¦¢wo’1}Ð3#ÈH\t*m¿d4e¢f>CWÒLKk°&†-xó.j0 |™‡ÈÅäËcVf\wiÌyw_Õ/V©‹Q€J½½ðD½ø‹ûh;)6ÌaaAöP\³‚¶"êáuåbÁ˜;!A§ÞF‹–:–#™€T,)ו^+!\ ÕOA“ÿ×P=/8`@²À2,­zföׂ­KÁ%ó«Ðf‹öˆ­ÔBäÖ!õT™å˜\¸‰g…£GÞÅ&дZשLÁ»T›5@wqÛû@„ð{<Ö ÍâϦ/NÛY B/‚Ît ®¸Ký•ü1‡eJˆ˜Î¡'áœØÞ£+um UŒÕœW‰VÑí±u1Ñj­Í%¥«luοä¶0˹ÿ}¼Ðܱä¤Sà¬e¡º”÷tIœ dê«üpì} ?·_‰ÈÏý¾óUÆ}}­Dƒ=‚p–—×Ny“ŠÓ³.®´SkýÜ|M)A€!ô}hKµ;—ó®?ô€6%µ8yW?[šÀŠZ€Ñ4šjàº{KŽ£HÈqŠ0n>‡ àÆ«,¥rüð+,•žÙz85X?:°a~ø ý×=˜_¯Ô»Š®X ‘SwcíHeE9Üb¸.¹]³Ýc¾èô“­î1ÖT$/Ôž/ÛrmçQÍ6ÉcÁ.TM0ØpY¥-Wv€7qsy¬žì 9CÇü°UÈ7¢ÿLªyæX£½é××ÕØÍ±i kG½4® S6.Or‹LÀ¥àÒz± 7Ü/Om ýá[b¬Ïê¿Ê„Äöw¥Ô,ã,îŠ{ZÇ·zæÐøoh"¿OF‹Å&ß®Üy¨ü @ÅL:ýÐ]G¶Ö£"ˆM­“gæVgö)¼ÜLs("·÷:,ƒu ªð1Ì9plõÀM_9é´̸¢  %(¦2Ó9f„ÐMʲžàˆæLXÉ>AIjOc¶°otx_ZÈÁ…S í]ì>Mù“3‘Ùï°@˼vÑé:vK‘1™ª*f8•+g„g\Œ;ÿãzšøÖ¶|.í˜4ùJþ¬5b]^¯¡a…º¼*Bû[’ dŽp/Mén»ƒ•Šš0Çêãºøsó<Ú®Þ§Õ~?½í9嚆½a.e˜¤Š¤…÷âÍyfäÌP!üQ®¯™« L ¬0lSG‹³‚Ò*³aR²¼ÝGX—¶˜(°B¤Ë~Ù…=ËËÞwöÈÁŒ÷kE¤½5;„WŽ" òÿÆL õ"¸m(ÓNO»«Í¼¾sÄ5úK&õQÝ?æ[g’¤Õq™v0ìÆ·`ɛҞ:³$³Þñ{°à,4êA¥ šÌñŒá™«Ç`*Rƒ´pÿÌ>G½UÒÕÙ9''?9V—òÐÚ§€œ2>”˜ÔÞ4íP»Þ²jõ_ø".Ó×?ÐxÅÕ'îu¾8 Ä\%”'|Ÿ¶ìªª…1jŒÞ®Û©É¬ýh® âî%ÂôëA¤¡3ÂçÀÁüqÉØ ú0¤añ ”DpœË­Ñ™aãóz(zí \3R\ Ê "*–w·@ñvxt,UÂ¥8“¾>ÆÝ—6Á’ï[ö CE&é—æV4¶Ç¹2wĤØÝ«'fé«&¹€ŽL; æÑƒ+Ï›$Ã1XG5M—ÂþC¨>›…å§õW€j@®¼p†ÑðEîí.BE«u,wXD4Vh"€È²ÀKìðãgýÊ]Qê*“_ÀÏ7’vù«+\7D³›°;ŸÊJ1º Üæä®c'’t÷£ßÓaEf†£:Ë7÷W.ç)r,ðáïCa|ñbD"p–®22©;NidĵgYàB}JSej~Øø cAÇÜ1Ò”á ÓÍH¥G& „ªöŒsœH¼: u÷Œ9|°m¢Êx‹µ„¯Ä=‘(¼ÂÓó D!ñy•’@ƵüCxÆÅÁî[ p_;6ûk[î4…-Œœbbqòu}¾mµòðm¿ÖÌ×鼇ù9Çôy¹¥>Å-TÆþ8øcHèÓ”Kˆº°i:‹Ï2Ïý¥äâhyÇ·:Z¼Fô‰Â_BNꉬâp’m¥a…!¿†ùõ¶(³’)?²Ü[IËk JëéCì8`Ë‘qn°`ûßÓÚ+eATèyTü ;Áá«É¤ ]b™{I§mÝ<ÙïØË1'†?FÐ0•ìþͪÖ3 Iyo ­8Ÿ©üè²ÿØ”YéOͲOT ˃a(ÚO®æ¤ËcFi“fDÇÏì’ÑK]† g¶¢û®bͤ“€‰òªgbzüÌ&[¥ åqÃ}ÖëAä΂óO‰ÕZÇØé¶@ø]JöÉ)¼Áé0×iÖÈ€¥ð!~”åÈ&ÙÕRÆÙ¬¼¼=]]‚;KÚeV< > :Gä,©E#3Îw“E š5«D3KçÄò Y[¿¨µwÿ$ÿ¥þ QkiÛŸúBûvLüvÙ$•ê3äG2fæZ{ Èl.k^[’’ÒM(¼Õø/χŒÛMèßÅ„mŽŽÈ¸žÝK§'~<ðŸ1ê)GžÇ¿xŽÑ¥Yª¤ˆvý´®½ûés© 3öÝz§dŽÄ@r?®ù—…@,yOýøpçzCš¢Ñ€Ø£¤[kA) ,5¥zaÜɳ{«-Ûƒí˜W/ãÉœ=þµC^¿›žÓ¬—NúûС77H¨q>êjDWï$Dæj–$›“¢y›fuÙîuß[\o"t íó\?ϰn㸼ë Ûx©/ïAÁ„%²4C“  Ÿÿåø®b$;Ž·wÁ\¯Ô,F! ¿ÇËã±_pÖc άe‚Ù/DX6Dá³¥:ŽbëŠUü}:Þ4º3ÅVSÐÚ0hUò¤U™ÜÀF¥÷ApíÅS6J ÷Ëý÷ˆœqåeY@¡u8Ëv4'®5üÀVGÝ‹Îz‰ã3b«Nø`v²,Ø/ ºÝéxií W¸içÈ~iþ ¦š5¾©¥µgDòóÚê 7Á²…¶-€Ÿ•šY…›+fbú:µ#èñ-í±n-“ñ¯Ž¼ÞÂuâïÒÙ æTöy‰yº:ìáœ7<‚j¥LʤÀ h¾ùP| ˆ*}* ¥ëÛ2"#£Ê™Pâú+5à^bT3ñCæß5ÅéÆüxì·O‰/¼=é9&ì¾=T ˆBï4rЈ~¶ vRì:ï"…Çß¾g ýgTaõ‘ài|mÈ|­7×:KøÝÀŠØc:Õ§Ã}úøÎÔ¿¸krìóXpó>W—*ÙMþå鶸W°Aëÿ-¯¢3¥ËŽw°¤ºpaÿy=’w¢™Ú="nmã°Ïëª9úùLkÔ³Úæ‡€¯ÙYF/Š2Þ6Áuú÷Ü´Ÿ1¼0Ѱ¹Wº$Î1Ö…Uœ­Îó³s¯4 c*+«P”ÄÜsúSfÚŠW@ -Ë@›•Wï(/FÏ,ã~¶gò”º=ª_VÎ)"ߨV¬à èà{aó±<ÿ°~÷–ªâ7µrⳘÀ¡<¥º ON.ô¦Qæúû@ɶKÒŧÎßœÜ ‚ËÝd5ÏgV×ÉnY‘;Æ\2á®wßaRßjÛˆMÍ;íyö]öÒNֻȅî¤H£ 2s†±½üÃßa(O7‡Å6ÚZµù!£×éP>’õXr$k5C³÷ê‰&ôSHä°ÒÜDÂŒ7&ÕúÏföhºý VAž¹#€­©w2·ÿË5‹06È`jT0å¬Tšàh6¹/~Ö-d1Ld“ ÔýT—ò›9"ÿsΣ¸PUPIÄLQÖ‚¡T3L[²GS ⲉªQj~UÜósk‹<]˜Cr<àZà*Ø„ê¿K:UuÁj1B9‘‡ž‚ÓÆ¯ôh$B–IÎ\-ïæÁnÀußч۫V‡¶r QhùIN“V»š ÚhåVçDPûì b’TTÍ›6Qdš™¿LáÜÙí¦U霺ˆ‡ËÁù§úìßLôY¢xÖ…¤CœÜ¡€Ë±,+mD阻¬VoKª7å‰_æuÁHràë ¾íâÎÔÈî¯>*Ø.2æm$_4C~¶Tͺz-³ÃÕ ¤Vu™p4¡÷årûê£ñ1¦›|pSQñ⫾ãbs•´Xœ!%Ñ›‡)¡E‚ãç¦)ü”n›U u"Õþ¿ñ{$¡+G‡•v’ ÙuÛ''ú"l]nFùÚ­¬ÂŸ¬þï>Á’~tÕŽ^Såá~ïédâËlל íwC°×­ÌWóaèp3:2Ní® ‚˜W‡üÒÌ¡k˜)4“¦º¡;s6úþYFU,!g·†É..­Ö;WãÀ#³(Nâ6ê²Z×.ošz*lß8¤î»(\~ÜÄž M6ÑøøI¼˜ÕÃpåá·4þù#é4P Œ§srŽûþ¨E2صW»=yvr/ÓÞÕX>Ìjס­Ã-ƒˆ°ñÎA ìàX‰Ê•f/&FJ…ðì<}'iHµÃ7ƒ'L°Rb¼ì©h¸90ÌÖÞ‹¥’û’Kþ´£¢½ýiD®Â^h± aøíYýÙÙcˆH¬p‰ÆB‘ž÷hñƒJNW«vÌ\p»ˆÞêöP»©õApá. Žß‰ÚÏ Ü s²a3åëZûH¨z=Æ{Ïž*-÷uG7Ëû°É´ \L<¿h‚nÍÙWØ,5L¦² HæÄx] I@åX3»Iº­á">Ÿ£Â²º ÏNÙ“X"rúMEÄìÍµÈØ£±–¯˜Ð| ßñºòfЀþ£Þ…UY|ÌÛ{=jÛòKdÄÃùh¯ÜÜ…£¿­6½‚j- Íu²Hý<ή‘OA!ÊD›êá&‡V”Ò†˜¤«{L,LbwÀGØÎórÕõ™”èŒùלŒ½¶$Q^';#éNm¨IÚt.`AÎ-e88,^cj6ùè…N¶ÛÏ17”X —Vº¹ð$KçIĈ'=²œ0¥ôé“)ùzNæh9¡póQk=¾FŽè%ŸÅ_Ågý„ï„+*NÔ»wÇ×&ïÀÊ^>j£®©D»›ÚŒ)mËÈÜk–ëÜ«Aƒ0 "xÔîí¨† 6*"]"×”,‡UC-X¾"¯EÉ@­¬zÏ·ùml…቟|è<Ò •êÁ%ÌÌx|j³¾Ý® Ï§ %<ÅŽ‘M ZSK7FžÑé»û0*·#n„mTD«Ã’¨44@]ÆLÝBzÔEä)XÁMÍ](J¨Äž¥ìK%·¦xiö Ä"ð#XÜ<ß„ø×‹]v–3ï8“ç~š3\Íe‚×§ý”Ó­åí~SÀU‹Dïÿ2R³DcáH(±Ù;;vA¸(Awv^˜*ËÔzî0lÕÀ¸nt@gáØ˜Í Ìè:L¦5r9ÓïÜø¾È´@q‡˜‘‡Rœ%6ƒUy½/d°ŠWvF¯[ütH+qoœLŽCù}<Ï·¢#¥5$×r_¥ËÇ ¬) ¶Œ1k´(KmÿÄ ~ìw¯¾ ¿±êþÚä„yḭ̀ÎÕ`jV˜—óeé> 6óß–Ov2l£ác¥lOS}:âé3• ]­j]ðD`D.ýV,k _ÈW YÈ™ÓpºC6ÄwõNP×?/ ²0Æ¥ö Õç2)ÄI!^ëd”rK{á< °Et±o¹È{M o)35±Ež8ªáÓÆºà»½Ñå*Ür1¨¾Üá¨]o—§Ú§­ý Á5Ma^*Ó2HÜÊχå´®?4ÈŒÅ)®Aå‹a>ŒÎw š×&¨½œ,ï`þçKvëÌ-"5=•#ʲ`P ¾JhZãkȽè¢s|ó}Q%›”Ð1UZ$Ô¦,ïW(–l¥\؆,MÖyJñÖiϦu3¢[QÂüÛë4•y¥N¿Bׯs:½5J+jc°NÙL½Jˆ²S‘À(K<ÝA¢`ÊužÏç×›¬K7L$¼g˜{'€Lp€ né8ÆÛ!pe¢`?«TÎ'G“hq5pйw¦¿A…y\4[í¦†'$Š8&«µ×ñOåt½?³½=n»ž½Ê?\q~äë3D éÜB«X¾¸Q‚J„çæÝ:ÀŽØÿ‚/QV™4²­]?ð\@ó57A‘HtŽÄñ}ZK¬0JâêJ¨ ÑÖi”vä>ké( «ï3~§ƒ*0©ÞCýkÃѵ¢ | K °ŠÊ,pÒò£Ù(e2²üV†c®Sa.¦`Ú¢µkÏØ¨#õ-Îi>Ç#m÷¶^ã„Q98/ Ô^{×{°i²êÀß/#%ã½v­l/ûT«+‚œ¼AÒé>õæ0ÿ䟔¹¡êÿfŽ3rŒ}|”â¾'gxˆaI[µcBÅ.‡ã]pÒAŒ^Î]@º?:ù ¦Îdvƒ¶=g‹ ›Rí]êåø¸ägée Ð'2DèÚ0†;—„ÜŽ ]À^Åw”€ñàãµÓw˜ø¶Lýœ´´k0u)XÉh™žZÄŠèl‚z›7å!Ydq8ÉÁÃjú÷.UâcÔ¤fÚØý⊋‰×ƒœñww â„)ñ4§;g7› Ú·Póá#½K0‡ôÀعýÚÆâYKô£Ñ€„³Î?Úu –Òú]¢©éËDK¨nOö>}AìV"¶ Þw{¶ê½ÄÍ/ÆQµipè2‡—\"£=û;{ Rç mv*+­º>Fº9½Úk}f`éLŒ¿˜²9ZÀ{tm™óáUÐtè}}qï ™›¦íwA¶nÙE™4DáÇZZC[Ò·g6Ö‘*&™¾ )!=Ãã½QÈwgl!] _óTEF(}£ôîßš/4™1(>äЗ(ý m fD”cªàŽ6ëêÀ Îwwh‡Ú¥BH´øµÇT"š ˆíî{@y¼¤2Fw_Ö?&—q½Âj`n z3ËÄtˆ2=P´æs¯–€‰oÑÈøùKã ŵ´vN˜EºÁè7R{ >Ù¢b:âô•­È]kídSðÄ<ÊÊŠ`ë–¨d¥ª æÖ½¯1Dx®”s#«™pÄÕ³\Àwg.n•iPª ôDxõ@Pgëð(±]yÛs§Ÿ5¬eâÖ6#vØC4a[RñV^êj~÷¸Ö’Ã^P8ÉŽ6CS|¶þ‘g‹&g§“ŸÑŠ*€=ºT5L|€O¡®^LhÿzmB"Ÿ1}¥óA(×.òÁ’C6Ï©üt8}¬ƒ,'c€ÁZ#µ¾–u.œ‰‹–=í¹>Ÿ­¦ sÙ–(¾ðµÿ>Î &ây±œ8¨…f¢½–ÐÒãJùÀ¹Câ·ÏЉhkļDÒ6 ZTê6"ç §ù[áaDI».Må4NZ)`ÂI,Š×¢?&^äeпËÌ*ºLœ½ž.ºn… ;…§˜¨¸ÅOfêSÞ <Ô䆺 ÛŽ™iŸ&-6r¿À1ŸÛ2AF–ý½YªlþâÊI÷l<êËŸ2>5²Ôl0žSÉ!S”èÇø¦DlNô¾H>üy¿¬v Ôwj—ì)ߨ<ß wø`fV¨Ìn€ŸŽ’¢P„„ÅÞ4œ¡ &¼ø]¼ÒE®"òöUaöÜß`§<琢Ée˜¥Áè¾Ì—x´Ý+|&ü9\nX’¥M.ˆœ7Ky•ð0Ùe4½»Àž"–¬Ú1|UG’Bø!®I÷Ó>¤ö6´”“Qƒ/ ,l÷§Áž$u©£ptcS¶aÜS¸Ï¢ ì»´–&¼^´„F0¬Æãí Òâ±¼g›$v,BõžxLƒèúïï]ùäÝiwš©cžvÑefùUO‡ªgð™ —ÄGj8 ˜Euÿ†)›•Ÿû“C›`´*MV5K׿´:aNÔ1†«D‹¨”;úEÔ³…)}Pð@È…Voâùßv„>6ØÝ娥G¬¯J—»¤L!ðhì µ±÷kV6ÞS˜MÉÌæç‡)ŽÃã>dŒdREÎåœb—Ç¥üý¸‹E+î•GûNì[°çQ’ÊÉåÔö=)Šáù…îaSVü~$ žÑ-óµÅ¥gÆýœOšbú'@.™R& î÷ª¥ Äâ $>“8YÒ;ÈÍ6{+ר(=V=„\®42v÷ÝqY¦{h°PÁî»'} G »òá?<Ó(ƒÐr{N1WŽ™óŒ3h®:d|BaêE·Ý¸PÀ†qŽ’£”!Yü_²¦C G > æÜ²Ffó}.PÈ×_Ht#µ/ÇÈè ~†ŠÌ–h“„;¦5éqöT°A2žPðÆLRñM,šhÎ[p¼:ËYž^l6M¹Ì•Ì›×+ ÷¯Ü™¥ËÌ9C’q9èÝ Ä»§ØßfûS"&Y@•'M3 Äl^¯DŒk£%4TgÛ³“x€=µ-Ê|«—2Ðà#ç»{'ózÂ9ÁÈÖïÏOîbï 3KRd¯?qâ1ÔËΔËÅô@é´)gñy‘Ik³0ÎwÁË–y9ØO8fÉS- 3Ñ¿M· ucÜq.«õ €h¬•0ûܤü\»^K½†£2±‘Š…·²ÊªÆªª¢Î`îŒj¨´ºT‰Y)LI/=!t?A3ŽÊeQÍk¦ ønÓÏç³ðu7ŸŸù’Jq¦È|>ª{…$ºB¢Dz͵AX,u”9d†¬ÙòûóEàC ÈWóÇá,Μ‰°ßò´N×c|<üØ÷lÿm ¤eñ»®L F ÜÛím ³å}2ƒ ÞûHWëv‹âöS7yG!Æÿ«hx±*£#°ˆ¿ÿê× ºÇ´ñ{¦ÿkRå”é^£ ’De ÖF#ÁŒ²\ßì 8p(N™ìÄ-9åÂup.¾aäÿœcGã2ûy<)ÀGg'–(á@‹‹|Žø]]ËŸdAH*ª´îÊóW8¿5ùÇð±4q¥Z™åãN•Ô=s)¾™XnÎ ­þzR닸N¾<ÀÔ`”•[bîÒ9ú¦n‚ÛÁ¶­¤e4*(Œ6Æôì–ÍÉt"ê0¡ÔÍÔ…¥ù œÃxZø@U‘ÇsáÇÎ¥%2Î#@øQQUi™FïõŽw×pšhýöi¥ÆŠÅI¸è“’HÄ":¸¸,Ýÿ"×–é1L‹ŽëÓó½|¥v|A<ë^­.Ê$ùÛqaMü€@*¨§Äª\U…iR¸”ûß9b™Ä_æ IçJ¬FÕ:cÈz–Oú<¬¬LI°‰ZßékSL_Æþæ«°Í0;42Ä)qRkžem’€ìþäæœ<¼÷×>2àsP¸H×»Ý- FÐ@ûåwÞ›ªq1õÁ!6ÍkÂ8-C ³Ðm\¥òu¡è %; /mÃ…kÿâÔwTzC 1¶ÆUƒáUÈ1_•«^æÈ9¹,¶¿2]>]6eùpTƫʪ¨'ºÞ÷SNuâA[úȯûa>Ìz5«­þüîÊ/ñëf7B H;ëò¿Ï‚XŒ·ß’³ã¶ß¶Iˆý½úfä}0É5ºÜïóá^ê÷‰~~›¸RÚ»jmzAaÈF”Aé|â‘øÞµgºú'ãâùêL?\4¬w>’,_p'aíô¾i/¥œ”%‹dÜj~MìOcà€ª<˜~ÍB«y½C-:NåDq)è®´Íîb!Q`RÀ'(äuR,;î{Ÿ #c}Ù‚åjº“† •öµë¡ÁÒpìEüq#ùÀZl!Ûê0J1±¶ÂLÄ÷~-€ÈºäpH„O~šp*=ø‹,_g«92_ÓáÌ)ZRÎÔ\Ç‹”ê´W£ù1ï Iqâºõ~XõW¾ËëôW¯[žïCÙwêÒdɉMŸ8ÅYY´Qs¢Öšj)eÂ͉AF¯²Û?/+ݳ \˜?{ø6õ¬àÝËeN{*—Tó™Ô]ï[Ækí|kÏ0üqpÙ$§æ¹Žz9Ž‹õDåëk³ö>¿fNß}¬Vãá"‡¾6M]œ;z‡”u&äö« n„ÈŽGîÂÞU¡Xní­­¡¼­Õiüʶ!¸+ÞÕB…÷fžóFýþ+çé0U ²Ò> õjå; YETå§Ú` H¦A<4\àÈÝ饸 ÅŸÁÂaÕ49uK‚g"ljÑ/ŽÜ³ '6ð—´Ý¤gN)††üÏêcè!FÈ-–ÍW%…ùé ¯IöÐ5È6zB!fTRO•ï¨ÖX½uÙ²9þô}nW¹c[ìÐÌîvøæBéÕ™3Ë;…–c³Ç+xš#Šð+L¹ŒÂì s^BÃ#ÄJ~öØI z_ÿnzÐ%À=&¢8=Z@ËF€ž‘Èeè=~Õ›;øtëüå ²F衳-z+›¥BŒÉ@dÿ“cQš>To^]`7>y»j£Ø øˆX›ePŠM©X_¿Îg#îÇŸNù¡ä˜O?gø)Û…}…Ÿæ7Ѭ¿cŒÖ×þû¼âQRHO§‰Wm®Oˆß_e3ø-Dê ÝÜø>+ú` †•ÑRøãû„‰%0@:u˜Mó‘™_foH›¹éœ¸‚àzòþÇ;ÑqïIz–;´>äœÁñª"\ÊiZ“’'l}ô¡ÉЄä{IxÜ!D9Jw‚ BáÂeÿÃýæb€’bІÎTƒ©‰MbvÈÔbõ¹êÚ†ù×´‚N¥ ªäìœÌ¯êÄãÃ.SyMyU×qiÌ»í§ó) w™Ðø=¿Kà/Öu-íJÇe*f“fBIlŽ+MÈšp ìÚ>ëçO}_þ)`=J·-¿¡n/- §G&íI77Ƹ”9©4Ò&à¶O¾] ŠNë)ͼ€%ˆ„Nò oE¹Òe±»¬ê5 1¢Ê¢y ;9ð{¸ØÃO÷ã©_ØNƒ‘J @Ö†ç`=I78r1*ö½ÃÏÊ'²˜Îº¤úÒ!l-ÎwˆI!‰î ³¾1É"xøåzÌj ‹ènç5îå¤$þÌÍÃE‰/`g0¯i#èGO̦ùb•À‚T%‹a¨¸;rÇP\Th©`¯ê‡»=¹g$Šé‚l„^î<èÈüßÄAÞK!0í‹1Ë)åö ®/B[å@ M¬wÛ^º8øcvb ¢{ùÇ€¹4ÍSÙ‘oÙd.}e—¢@jIô4÷‚ª«I4!ÿ,z'Zïáã]«†— xÉ©­€@Gf}€ A2¤\ô‡yc›Þû+°ÞKÅÇÓý$hÜr" Câ6 þ9 †²åô4I¼"Ç}Pzñ~ˆ|õ+„Êâ| ØXv˜€FÀmª|º:?kE°l±:ªq{#¹ÇŠm™gg pTT±uP‰\ÿÛcBýÊDš¾=rëjtˆšõX,±ï—Ù_C3ÂíðÝ[âÁ¨Î‹·LaüýT‹W=t®ta®šÓkZ=¶â:/ÞIÑ[Èð‰îe9¿ÿyx]Æ-ë¥ FÐE<^pu~ŠY ã/Ïf§SÂA'f— ´iúŠEg§ÿš3`·!—q¦æö‹wu×_=õO ÔK!cÆî?ŠOY›Zí )Iâ.åè3Ó´n “¼xH´¿ùŽ÷rÐUáLi¦XÝÄßd“ÑÁù3ñâÆs÷b~ù7ˆ¦ìyõæ™÷ì‹öBEÕ#mgzIæ¬ï}hj´L:åA[Ï}=€/œRéP©ä½Ý°ìæ<”Q$é©AŒÔ,`ÀõF kQçz_¥ÿØ~—:gûçú{üÍý©:Cõ¹Ú¥V€Q!¬Vv‡áÙEØFù§cýÕñ´èë0~^XÙHËóNCö۰Éáù"íÖfGùÀ?ó’+¨œùÐÍifιÅÐ<Üo!ÐRÚçÒ­|%é¾ Ï¢Wïœüd?«íBw-ìcyêÖîfîh#²àæÌ´Ë‡ÃkgÂo/8«ÂF-¡‹NÈ ñ—ÀÒE&;\ýÉI9fInü¹4-²qªî¥ˆÅšL/Œ·ô[™&O!XŽ#ÙÍl:¹ò¯ÀÄU¶ûµxÉX„2JÞrÕëßËjU¬Q2ƒ¼ÔÒ7㡘STpë%µ~‡ý{‹ùEÛò꼬lQ̈% ‡N¡¹Uê! £Os ð#ÉÚšéš*8ë>{áÝçæ•ˆÂÝ‹éá"9¿nPnóm“½ÝÈ`® ÁÆ®1*î°¶,ÓÐl‚]”Aî™_«cƒåѪlhù­Rÿ¹Oiª„í&ˆbòú°éŸïDŒ† ît¸ÄmSlBqüÑù%¸íÑ%_¸¢ÚÂ7v±Bþzþ† ¼Ùô}xµå`âIŒ>ÊÀ ¥–ÿ­r·ñ`/ç‹àºT›ëT ôpÉX”Œ¨Ý—ï$Uy¡Gš–3£“­ûÃi~<ñ‚ëôYḭ~ñióiôzSW'J83qÚÒwT˜Œ×)gÕ¸ÇÓ¯‰„5±¥BP'ú% HߢöïÐ"¥¬5Sn0[³vÄÿN’ë¼%Ü/ÿð›EP=/@- Ro±£œ·ûßö.{†o|]|@õÒ¢±´Ù€aÂúÏ ÑiA\PÃ]-Yñ}פƒoÌ7s®t’ì&’£4œB)þUzQA e4Ь¼ ´Û̉fŒSg£ºxvK\þ¤L¦ÔâÐcWxc“ûÒÈèØ[S4Ñ&Ê $YNÈ«5ÙXÃèCY¡²ôK¦âû×ïn ꛓ‚´T¨ מ¶µ™Þü­-Gë9ÕÝpMÛЭ­/ y&|¦¢ö ¼²Ý:EÅBá.ÍBëœçßÁbV°)—+Þ#·¦©>|ƒ˜xè€ÎÝlIg&¯­ â¯2NëXˆ’É›©kEÒbs‰,ƒÜWÙÉøýVÉyÚäð-Þg$ñfPY'“‘@ þK2“¬sL|A Ùg)²†O®ÍšÄ!6å-ÑàüíJt‰ë@ð´¯(×½–¹ Ž@ZÑP>¶)%ëØß«üú¨ò.vù캋*ISdT&ôLß¡&ð„°ubäaÜÞ‡½±›¢Z³͞ïšKÇúkB¢ÉHíB:3hÜ?7<رz{&ðè¿”ù5?½™~W •Ù‰Ðȱ_*×ä~º^o¤@ î½AB),šgðHºŸÁ-È M°Óh¿Íúo@å 9ñNË’ ÙPûft2׬³—˜¡Ì Teo®ƒóqÌx“¶z¼{5ó^JŸ½\y‚Ùñ¯}ÒÞ²"Žˆ: ¡h0ÎÜÇqæÒ¼O#=“¶D0Ü{¦^޳D”­~ƒ ’Âø|Ètzûyv±;7f,f¸CíPÉØ÷°C#*î,Ã…€äN…‰Va¤6³* åpR“lDòý"^êÜ\ž¸ý•¹f=×)MI$(ÙÉ+×)AÕÎ>!л‰å§<û‹ÜAÍ…/¶_gúÜy!3lR7ÑwÿÄa½ÌEýuVpHC¡àsOó°HÏ™Žôß1lô,?°}%ÉÌvĸXš^Ê©Ê+.ôGÙÐg¹P‘µÀ0\Sþ]£À†°k.IûºsyUŽãݾ± daàŠ™ä+4½åµ—bÂù.i ÿ™ŠðHóÅ-W¢ÁüæecÑ·)ÝÅlÎKàÀpFÖåÜ—)AÚÙ ýßC]zX×Ò€ìl6:Í>"«BÇ  ¥@»g‹¯&@˜»Û;K¥#­ÅMÅÒÒ ¡ΈAzÀ4o/ §é7X•B¿Ñ—à9S ɧJ!!ßnX‹hƒºÏÑxðõ—.Œ¼žC¶,x­mr[#2WØ3öiRð<ª7opí…0}º•f,œW?O/Wm¡•È/'îýU«N­ûЙðCÁ§˜~š¾!£KßP[×Ã)NS!¶,nÖœ)²Îèf£Aôf`ËÍ€‰Ó‰‘4íænò+LJ¸1è7rØûÒ J`×VæÏÐì[÷O‹MŒÉ,GØ›‰s‹xxyòɉ‚åÇ«@$¨*˜•6¦B|ÌO ´fcC˜ÙF‹kÇômÛû‰½»ar+,”.ÈÚ¦ô¥ä&Afs ò‚Ç ™«!eF€æ¯ G™‚Ç\»yÀÇÈå…¸túù*"ºòtQQô2qÔàº0ц¬*¯ÏIÐÔ:E°«—;Â7ì#~?¸` ñR@¼6œž•UA½ØŸÀWàŸiôáU5­bç#5÷\к¶o±{Q¬ÚKQZ-yàÃ.Ð ËÊkb¿C…ú&Ňÿ98(¾/ÙµnS}]è!lÏ-gd‚ä«’EÜ6Ue†(*¨ï±ž‘@Ë+Ë4ÑÚdzĪã3ÉÄð0¯.Ö—)k+¬7â †‡É²”X—çc<)ŸUØY%û…_÷¸'thzœOÃ Žž¶©Â(:aÂÂ!lhþ9N4NgTš¬å¶;¹`™²å§wÒ $R‰ZºãˆX¶±%ûª×."mÐv2‚‰W&Ú{÷›Q[Üðõ.í­cŠXWÄG´mä7. ›v.ssw0ɰ2ªÓœ@ú¤nèPœ³å9­ ž‹`ÿÄ1¨•6;þôÔ5òòëž¼Ø=˜fYð3™{E:D<‹Mb<2ûu>BHJOlÜ­èÄĪ磘õY9«¥¨ÑÆ®I§ç Íü*??ëãKã (.—­iME!Çþúu—ß4WdÀg{Å“™S¢/,¨[ÎÍe¯SF‰gªïº¨úSpÿçÓÎ؇šâÜÛxã} VX©°šûÆ9ß'BÇÑ—•ÝB¢éjòãµéÔvÈ1  ¤çÄf}a ÛïI„"Â÷ R"&áˆS!Çò)¹ß ÚpC¤¬!ȺSTCòÂ))¢;÷$`ùÀª¨âÑx¼îû^èøå¡ïëp<†‹Þ<Á“T¾2m'£I©…XõíÌÅ“ëÉåÙ:Ô’,ö©Ñó›×,]h„Aò !Y‘ ÿÜí'b]Ê$´5 ‚sÚÒ=T¶w´’Ê¥;TM꓊9EVÄž¬¸6 ¢Ë2MA%—ü—9ÿ¦Ç½„Ú“*¦þ '9Ì{Cΰ¡†Ö)°½wXSvç1ÿs¶²çR²G»I²†]äjºf0¬ÓËö¬]Љ¹7à?meЦgÖ‘{}ë>y¯I}擽dõÂú+, Ñ'õ*À#ËhÏEhÅì+F¿÷c¢øM( N`}ý9½ü¾OÌËв eÚ`ƒ]‹ïò®£"Ñ-ä·…—‰I¤¬9FhÑ/]â[®KÃVsÓPÏ(L|vSEÅ¿)eÆx¨X`±×yu‰»^´„“z5eÙXUŸ‹Ä”ÛnÄ3òVˆ;쪌ÕS5y~Lj‰§ æ6Þé·ŠÔüu€;vãw©{‡aúB—%A«ÍŸ©ÏAé"¤ Êá“‚\ocƒ=W›“÷–שo“ÖQE¥ðªn#Òêü &ÛVßk&/:®`ö^Ê@J§;+ÆZúýƒÊ¢YÙþpkZ–gê/z?èÉ&òedÐo…¦­Ùk<À8Vñ=ãî…–<ÓÚŽ7”âŠÜsŠÁ}€vU1›Ú˹j3†É½Ó_îôÞ†|SoÇ$2»d†—;÷ØÜE 7©yw8 ²©PniGŒÑ:½¸ˆOî[Ës Žé×Eÿy)?ðŠVÙõ†-3)p©‰þ#²]}ÒÁWÒ×!ø.kÎüW#T»„»JlGÆc¥{]Ç\B¶9tçÐÓ¨B®aÊh¹Êí –¢„j}øzãããc/eNKÆž¡¤SM|ýmê×e‰*…?>›¹ß3¬[AäÒ•qD—Ô0nPSñô$ŸÕ¼‰=ÐdJ×W£†È U–pÚXïîKÙy–ëO:p+'Ê¦Ž Òú®d𠄉—­„÷¼5•=Ѷ;t5æ‡î)o°G %~ɥŻ; [LϨ“°T$K Y¾WÞS*­Õ ;”Œ|_ Ý&ÌË£JäÃeZbj5…¯®¾m;ã^zGû"­Gl`bÅ£Ø6Ã-¡š‹‘B(…ñ’Z˜uäÉyyP)s GêÆ¸jNO†#5#ÊŽŒÕÇzàéU$ÐYË/>ÝÁ«áÏS¹çÕnåƒ  ÀÛ«´®¬CÌ~+-)o;¨€Íú…Ï&=Â…v“ù =DVAk_å~­špwH7!VŠgð†Iºx-þm( wîÇ #Ë·0˜ÜßyæJÕY|;W}Í©¨ëþñïHY£o¹Öm´‡úœ>#À‘]ÍÀûEãfñþRÈ›C{Û¦íJîXß~í­¹ÜŽbœçþKv© ÃY0 Ò'ô7SØÃ¥¸±#Þh•–“u;Òe.êÝUùº6h r©j‡bOÿ'ï' ÒtIq縰h%þÁÖ{¸{êÿ¸ O¤Z%ñKêÿ?[¡¨/Œ†o½l`ÛíÆb‰ãö& eÿ{_˜À¡~²ê'@slš ±gþrhÊ«ù¾—p|øõcç'äâþÙå1_9r‘7ûvò Â}7*îp*p褔séH|’¢™3&KØ…n ’ë¡Ö>ˆ!× ÍDàiÇ“¥?ªnâ´Üø¤•=¶KŠ¿’>â*—OÜó¢ÐQÞtήÛ];ýX´TÜR]AÃÜ)¡p VgE¨÷tôÁ}H¤àSNÀ¾÷P^šEÞÍçßò<ÃV½•Jdwœ­¥säZ^x¡—KóªGö‘¼4YB»bÄ\ƹ‡å‰1–ÖÐEŽ’P0ã8Κ†lE¼z˜dd' Fë°Ènžé?ÛXƒ˜‹™îóΕîºhF˜ã%*˜-,ÝGëƒl?ï‹ ÚZ¶6¬Ïÿä½Ðè}Í»Ÿ–g ­3ü%W5š‘x)#J“¶ë[‹½3â^Û¨ªKºÿIÎÅ­é(ã@&SÀð£âÂQŒYD¨,U›5eþBëHKfŽ×„ÈÂÇ… „ ™Éć{yHùzhfϯ@¼¢ññÚ{ÒAC_=ÛüžkÔû©¶“: ~7·Ô½Ð2ª¯èÿ”Ï-u[ÿMꥊ¯(‘ȧ„2ÊáOêèÒ2A+çªp)"MhÀb¢<“pSntÞ2ÙM?8Ý:Ôÿ·oÓ,´ëðBÿ7<ðŽLpS‡a1JÊa_ÅîŒ&|ëí¯qÈåã]üËÕðwŸ¬Ç}§Zâÿl%øá}ZV2«¿}…Ë7êwVì¥BÁ¡Ý.v/Ã)µ–¸}° )ç를ǽ+’åÊüüQý¾Õ¢td±ñÝ,’;^Ï´Aúæ 9o[X‡¿æ+áÉÏPŠ8¯b׆h`$ÊF¢³Î"õÄd©5x%ˆ}ƒÖ+Tíˆlï ì7 ¥æK7ݰ}´ /ß=éyá˱ ~ctxu»dçWmõn` ÈÊ÷@2Ùøª^ॲæH¿åÙaR¬³ö¾x÷ö§^—¸€–á hÉ'½IAµãjF"'”nÁ” \¼ÜÆ/´f±ÉŠÎP…ù˜0¼~7:<ílZßLXÉç‡6L’äúNgÉóð¢ɰÐIÔñU¥/À¨NľzÞTÈÂñLß­GµTÑ·5é²yøT%e­ÓIÓ„dg7{‰ðjùk@ó¦ì‰$Þ­ºkx°çq?t±Û•Œ×µ±vsÀ%ä.Z«êìgm&µ`ëx„]®d§ñ1“æµÂ¿’¿/ÞB/ÅS +#øœ’GÓÜÃX¸[ØŽ¹ÆØ}úxXEïg›¦F1‰™â>uñÊú&sì¹2mØ;î&fúÚU×\Ê\žosg“Ûò#¡ÌÞ’;Kc¾ØÓó.•*]¯jѵäÈöñ‡!ù;ö¿lt¾ü²© Mçý-ï%1‘­©,µ*8+NÜ‚oiYá^ù¡¸n¬{ð¼íéâY¹!~hÈ {z f¬æò¦k#ZÓ€†ˆ¼Ë3œAÓ¨'[µ:TÀ;µä8ÑýqÍn8ÍAŸŠóëÈm¤ä炎B!‹ éŽ_P¼ÍR’Å= BYž„&EÌó¡CaoT[g¨›ßéE#à‘+°Q’Mhè&—EÜzý ”çmÜm d%¯\NiEÓcf"Zñm™€}[!‚ßÌU®ÏYù´OBªÅº çˆÔO ñÔÛ)9;úõÔÍ©¼‰Ù®” ²’i»9ÉÂN4ï£Z!d¶’¦ôØM§}FKTÌšI)G|zìPaÒ0SQ¤¨Ý¥‹éœ…+Êmt¨í#Ä<îMßx7’ÁÅå¼¢Ñ/nÛ™ EBÜËï”iÓ¡÷G+jÖ²G Uâ”)wúmÂOU+Ró“âê«ÌtŸ Ö•ãz¼s-5‡ÝQðWÈRÁÜ=¥%®\Š 'ø–úáó5Ó™TòþN¬Ë)¢v‘ù)Q¥ 㲈r¶ô=¶eâÞ´^ÞΩ: ª`UáSÉ6ÿ£Îj>0 ‹YZks/data/unicef.RData0000644000176200001440000000164611541700632013774 0ustar liggesusers‹]”ÏoUÇ7©×ApàR$$ ¥¨M”b;Ʊb›*NP[Ö8û¼~ñú½ôínŒ âÄ!!Ή$$.Hܸ!!$Nüœ8ô@|7™}»éá³»óvfÞwfçíÁîý›•ûÇqVÕ®Wð˜[ÅeÅqÖ˸"%ÅØq®/üþ{àSðøÔÀ¼î°ïà1øü þåû貦ø¼ ýe~¯}ÌÏлò6îÏ@\÷<Ï~_ƒcðxì²®)ûÿÉZ¹Wñ^çz;œ#οÞwÁkàð&¸ n€MÐfß[` ¼®gìç½Íïkœs4À» Éyë¬u‡c·™—9ßûÞåïëy ´XÏë¹Í1;|oñÞ{Óæ}nerÖx­ÅÜaÝ×/úsiêòŠf"Àús1tñbñH¹ÂÔ¶Ù¬ÖÏĆ/ÇbC|rúD|Ùèùf6G‡ƒ®ÖÇÞ„” BR¼T¨+Oû”lR73¡dbVêKaF$O¬{¥AÊóÉÁ„Wò ØlI”æ.6´/Ïl¶j#2S©høÚºD&R®d³Ô¤ÙH»6$¶…Ñ:ÉXm ù(¼–kNÈMÄ4'Ø ÉÝÔ3mt`ßi”jÇÆp÷€íµ¦ÅнÖ9ÓÒ$™K»'r¤£0Q÷tëQD¡6’üa;’JØZF†Æš¥V8‘úÔVQh£*kÛBÏšù6¾Hê™M»vaÕ2(J]i@~¤ÕWÚÓÊ ÙŠ;*me†V"H7Þj‘WûÚ„“á¾N«¨ì/Œ·Xff%×%ÛÍbW:œØïØ•#aÒÁé‘KÇdÙ=òiž(ÍÁ’©o„öQ:vÅžö]}fÛÙ‹¿•ŸI®—qCÓ(öÐ’™Ý+ß§ä'u÷%TÙºûÒ‰kñÜHçíM³§cýF4ì‹ùå^8˜“r­ÞP³;VRCî@¿yí©ö1jØ |DÚ.ôŒÒÊÊ#‡]RS«u¹VL~°ÈtøNä%­¥CRËLs‡ÚŽ{õG/>ÙÙ“äek8š’Voåh9—Ò?"ÄO%ó@ _âý ò4 KÏzáaväKålD£¹xòGwìSü¤V’d.…´968ù°þ‹ù­(â©ks/data/quake.RData0000644000176200001440000007404413265737216013650 0ustar liggesusersý7zXZi"Þ6!ÏXÌâÞðwå])TW"änRÊŸ’Øâ…§ "©Ù"Ýȵdׂú»”9•‰ƒ¯ÔÄï.¹Íõ“–4×-ö>ú¶ûŸƒîœêÎXÏ]÷8–çeöWù#†@ê½hZQàf€&-Öñ1&ŸGÿÁ‹úÂÚek4y…#‚ãí•U ze?oÃýj˜TUpÑëò÷Áa!ÓQ›…Võp÷¼&ÑŽû½?záŒoÓFq.û~êÎÈÇ@4'îJ4q•Œé9*¬šø â[.¥¸%Ãu´PÊ-•ÿ4ظɷ»øcOt“uÆŽorhØUh_ð͹y³R¥ãŽþvéûÈzÑœî˜QÙÐË?E`=ˆëK º"Ø ž<²áò²• <‚“ÝàÁÈUy/‘Z…F$.wxHf ®úÚG=íßH'†ÓI­"e7ì¡üRš©^ÊŸ|e'ý4 ÷@lä£W­©ß1Ãðl¤Ç¬MΠ·“†Ø~¬¯^Á“ØÅÆhâîiF ñJ"pnõ5Nôpœ8¹MÍ{gw–Dð¶§áÆ/‹þà |ökd\gN8'“¹€™ï¿s<¤lïf׃š$\j,–@ÛSžŠÔ·ª†¼ßê„­[P„tG«KuOl2ïbê7ùoþ¤Âì‹uÊ…ø‡%¾ é;Å[€ã bB#(Eœ["mþhƒ:»­è®È–ì4%¯ybºÙ*ËÎáù+QPUêí.K¾h…À£Ø[żírñKñ£{&ê*ìM«æ ñÛC0 ñG Èœò;ùs¹ëÛ”Cª@+"¯ïk#RJÅÒ <3¬ Ï«pP®>C®EŠ©¬4Ë=ÞÔûöókïçB’fHÑ*>)Â`Có‡÷ˆòõÊv èÖ3;"„t[È#pžUu:ŸlmQ8ªË”ã™çú&j¬l‰Ÿ˜Â hŸ ¾œÙ©8.T÷õ÷®Î¼˜ˆe߇·-5´ý6ÉÐ((ôaìæ‰\­8ûæwj‘™ß'Ú0FØæ¬±6ôÍ»ùE)>P!“tgµ®#¯YÌ­k¯PØn¡e¾ñSæuŒYù¨•dîïù)kºH²Ö5,E§Ð·íd!ö£Ìïf"& «"w˧ËáÖ$ت¬ˆgtR5ו‘€GƳ^&N:ã<éîo¯ÄònŽ>ð¶0¦uP5ä¬BæÈúM{+ƒ¯áç'-ä¶j}?¯Öiÿ¯¦\ܲ?~•—¢„×-éÞ*Zö+†è/Á¶Òò|k‹ÉflC©#õÀGÈEÜäsAÉB½_Z8­Ÿä³>Ë´9jÊk^WéÄÑ8¼lT`3Mø"ÊV©õ_©äå4ò,a,|™ÓÚ™brv?!>–CŠP±`9š¥}@®{É ø³H·å¯ ˜‹ _¯»2y›=X1¹ïç-»¥àâ’UB1o‰ÝNÊø/m<òTä"!”'‡sPùJÀR¶Pôë}¦Ñõ½-ŸôŠ`":6M`´ D&£²P¯+œ¨¯|2ÆgÔaÉ>qÈ(jÛQµ;VñáX/#¸–,žÄOPÅ;û‘Ïïå¢×pÙÀ´°Uœ ‹öïRIA´‚%ûµ²1ðôgMr2û}ÓM\» nÚ;]K.HAÖþ±§u¿Ý9ßy>Ÿñ6® š;S+oVHdT!çohO¬kOH%/þþ> {qàhãˆ[-éÂ\©Ã¿GŒè#ΰ9*%íÏÇ*lCn¿UŠV§éã7IÕ1otà$ÑîÉbrÔøxwæu,%YÀå8…—… žÕ¥Ì¹bª(ßq˜EØj:Þ)h¸Ê毺ù+Û¯©t°Q§`3Q‰ ¢„2Ÿ<06~ì‘d¼ƒýÚ&×f<¥›ÅßÊÉ%2Êþ †À xŒÛá„vO¢ÃÙM³ÈK4Óu…s‚GÓ8µñ÷ûÜmQ¼ÿCxyTt‚¥Ë‡oùúÅÉM–4 =îBòÞ6mãØŒWT_c±ÁRDŠA™à*'ò‰O±ùˆ¸øÿkúÔþ´7\Óô¯Uvd0[ ãHo¤Îßzñ8Q$ìƽ­Gî.¦ó;P «”ÉW‚ï^àÂFrÍûé¨^Ô‡òúgU%ß÷Ê?àz®«#ÁËdv»I|?µ¼p¾÷Õ­½ÿ,eÑò][DUˆ • „ÙF@WŠ™VþÒæª!«]Aµ|~æ„jïj%»lÜù4*˜PÉ*œ7”ÅÝ]uÜÔR:ˆÔüÊÜ+‡*¢Güè± ™œFIoKÀ—|o_õPi¨˲?¸t½Yå¶ÝÝÃiÂôÑóq‚`o,”ƒà*ÆO.XkkJÿüç¶bvß÷õKÃRlE€ez£ÉrÞ%(ý¢FT½°½Àm®ëìYÉÏŸ I¬éE¶òԢͭ-±Ò åL1¥Du²_J;À’ö& ÉÜ’/¯I µ‹x¯wØÙÂF¸¡TåÂ)Ï®ñXc¸öo8oûVv~…s¡qmꊯbª¹ª!³‚s¸¸½'6kâjt =qS•P£ƒ—ÑM³ïPlRɤüMÏ-°µXä“·1ýG6æDHÀÀOKJ.ºœ;”Í%Úš?æhχ¡îK³rvÄ|Uß9 u>2nJ”×(¼r·Ðmt±}ïà :ÎDù–Fªü¦œ¾wį[l=MWþº@–)ðM"Y:+Ø'ðkð°Áb²>ë‹Zž1ÍVF…;Ž@Ÿ|fЛÂ22ÆL²î_Á“È¢ÎÀÚ+pÅu¤ Mé‡fÍlFs£;ê³DÇ"y«nàcxà¡^ž«OVï^ 0vp^(ÕÁ¿€3êG†s3›]ðšðAíf$SN_MÏ &}èÚT_÷âËÜ?™A¨¤TÞõta¢üjmÚ©®§hÉ„ ÎEðù³³IË@ ׌Ií:Kä+íc_Z*O̧85úÉ¡i>aã³wÒEYQ¡¿ãÈÆZgQ4{ÅüÅ»‹ ÷ÊΙÌÂ\úÇaÇOœæT•µ g Õi·(§ÌŽHåJ§­Î÷“ ¡T.ÑŠ›§ø\oß«v­‚ÖQ%X¨R»¿EaËS¸(Ù0çõ›yÏ!Þ¬±1dŽŒ±pÊ@„–btÀµŒm³×¿¢†Iôâ•ÝWòÇÉjˆlx¡äxÚ2,nËÆMl™ürY–T·g{I˜‘–l•[{õ –J‹Ú“5[ÛT£ 9ig¼àXëgv÷AlÌFnnf¦Láœ0ƒ!áÅ´/4N4Ñã;ì:RQ¢\üQNqæÑZ×µAT ½»êq8âb‹—³ ÐíhIå0¥]ÆE'’e¶XÀh5€`,ö"?òáK0¼Ë€Ë†4òb€Ïô¦FS'Z/â'–N~ÔšÃß².ÐJÉkFŸ•ˆ'Ø«ÜIÝùA©[PôÒ²Ûù=ª8Ãe—Û¨2wÚgUNñšõ‰EÉÒÒÑxŠ~8+»‹¥ms'€rý¤#T­ ›Èžpˆà§pÝ•¸úz±ìøƒÝ#€ôÿ3ÙxÌ„ˆÄüýfì ÐHHîëâüAwÄRKô!½¬EÀ#+Ðá§š²@aQ¿mРùEü­n %i¥¨M ÌK¥ ŒQ´ç|î&jL¡’rƒŽ‰Íu-›|HMÎz¶™s½žóšj;à ÊþÝSe?Ž?é™·Àá> ùë¸Ê–jMÞ *FWjâC>²È€,)'W\K> ɧÞù¦ü .uÛ5ßÛ:ªàN#Wé5LcysÆ _ùðtò(ªUAè ”< Õ#.CdƒÂ>g!teØç{g,1ÛñèUŸ?2¸(2Psvò¬³åõÙá|æ+w±¥WÕM¥æÇ7¤yj±xÙò5Ý¡Š{sÚš}ḟìÀ³d“¾ö'ÜÒ ‹mí0vV©ñíÝ5'Ç} §“l»T·cV1W;W^~Q‚‹~0¹y‚“Û’tïp³q}·ÂÿzÈÞùæPº“7±<ÁÍaqý’ PKÑé¢1]ûvÝ€çªÓªxqHDÞâøÆœî£¨<:VOçu±ìÞúR¯ûêÐÒU¶ÖTXÄ{BDϪîË;e í4 9&¾¤ž)OíxÀ ù9Ð1Ø4K`ä9ÀÿÓÞÛ3éJìÏà|þ‘,jÀ2À€°lÉ4uÝ2¬ NˆÏƒ2ÚüÌy3ë‡Ù¸šÃ+)4%ßÖKÌ VL²ð ¼Ÿ±ž€ßxö½òq,b@—æ·ü¾-ÝÀ-Ðvð¿á¹7y§Dafw6°n_m±XmÄ$:ÉhæÌI:L+Û•,×ë¾¹ Ú¨ÙvËMš¾2КJõm©·W¦x¤ñþî5BÍ ß€Â+$Èu `8/œt¯Á¶/Q/ùÝG I48âãù‚ϲ_ÒÚ‘Ákڪ؂ú§`õ GA£cÆ—°ð#­zþhÛá‰ÝªÙKÒD”³:^³7Î¿!2"ÊS§^‡u ¹ßAвîa¯ÍI‹ ZÉýŽMkf­òžÖšÒAáä!x$+ÖŒ“3iŠ'5V±fê3K™Ê¸g!ßTªP ð,g)%VJty~lªL ›ßo/H&ŠÉº€ìÊ…=ØÐ|! .mÓ $‹nêX‰Òq‘¾xôw„ÁÕõØü™ÞòÆé åj$J“ŒÖÍ J"ißôd=0>É DzNI÷s”î#è¬Sµù›†…"Âà,ç<…MÛG}£otG!s‰ÜÁ•ÒÍî&)]ƒO„ÆÙtÇ‚!^@Hžá·eO5̪ÛRTY}NŽè¤µóŠ;µ6up—µ*,`5gyn·’¬šÇaÁ‹©´ÜÏC·ÿŽj¼°žœ¹~ô¯óÒ'gN„¢à÷duS_qÖLâkÕZÒù@ˆ¸zºŸ©ì+ûRŠV/:s…Ñ‘I¼ršµß†ˆñâù·G— {ÔKù…)f*ýó.úd IJ)ø¥¼þh¨—,s„QÙäg|ŠÂik,¾¥+›Å ußA¸ác6§çûT˜ÌˆNb†ÊtôŸØLÁ!üê–FþžýÚ$üên’x´Ë½àöC·ŠÜ”}ëXõFКL]Ê‚’%}ë¡ô«$ ûMÓW¿‹J‹Ë èmÅ ¸8IóPعØ0^wfñ¿<¥ÿ º—T®+¸ l’”íGD/<Ê>mv¿Ú$!¨mò/G€˜¬ä£Žðòù†öÈ}ëTä¹é ýˆÐ€£N¢h¸ÛÄ!ݯÄß*=ÿÖÄõý”ÐÆêÅ»ÏûEedšàNˆ¨i©hë~†C­´[ ÕÓT`«fÓ!lnº`²U°èó`È–K‚4¬›+(w”ç%6ì$‘=ż#Qø“$Õqé£"^êÀDg6½ &½Ý3Á0œÞ[±äh=\-¼+ÙhOš^nõ¶" ÁH¸â—ûî ªfö‚æy¸Ù¯½ËýcÚlŽ˜üz³æüaŸ ¨N„½ÿÖ³s µM¿‰øT j×ì®0 ‘y"]|ù”=4cxMˆIÅDVš%ÞèaK(s(*@l)…Û¥2Gà@ù±ç¦L l¾r|Ó'Ê@…Xω—Þª (ÆùÇþÂz0«RµAŒÂÙæLþ28C÷ŒF(«È;0¿gȇµ>€´qÏ]BçŠ|É”‹Ú‡J†ˆÃNÞN ƒ×/ìÉÊ F ó¨µÏëZßõ¹‘òˆ,dõûØ !:»Œ2ÍU™DP4Yco[,%:˜¦éÏí,ÿQÝ7ä9¢¾@2Ët´ÒC•F—QÔB'¯’s!ùö¡µÖ fBí}à9× ™väÒDZFŸ·Ñy…¯oæÍ ÜáÈ¥ÈoÅñy‹Ùg¨3H½Úó¸ØjÊšwæ* 9KS39"3“,l+Ùå[ý||_ôªQ‚}Ü˵È^>êI¸bt¤éZãàÞi=è…fwÃ× )Ðߟèö ö—>òCÕ‡]÷š250¿³Z~ Ph».D2ôM{QM¢¡N Å -\® `ŠîÜÀ”_個q;¨ ÎU%ýeûŠ“ t8ž1ü³Ä:FìÍá÷&™ÒŸM "9Üå ù×.<Ù1Ÿ˜3æ—ëO#èAÕy—¤Púä² jØh|uœÎóªæ*"`þⱸ¦Bª=®F+‰’Ë-qØâëy1šBÍG,f­eŠkñÄŽHÇé)¼9“ž6]9ö"óG®AšÃ(澦y¨l]Öœ’3lb7ÝÝ?•,–€ÔŸ×sJ Äçm6ÝÆÐVUËLãòn^Û§pÚÔµèñ¼'¿ïxÐ:PüÉV¨ ª;æ 5Á´´Þ¥±“ Dˆ~‰‚TžkG¦Ó„7Xb·u¤íS…ºÛCõ »=¨h‰œ±…MŠåì6Ô D¥£‰ß"b‘m†vl-…fô@GÖPí°Ü‚ ïºaH"ÄäÌ©ÄÁC¼üʼnÀJ:cœõ‰ç9Ôü>õضCmF°hÈ5‹iæôÇRr8¢«ÂjxËø\'¿·…¢A7»ä+ÓÓ«)ö÷s9/†…Ëk`ËÛšÙ9=ëfyën%)Ö—f®eå=Ó!pì¦áA2a8 ó6ä§Û—ùKwô³íC&9© ßÖ’âǤÇs"aO-ÂNbTÂÞ ·&…p#¶õ)RL¹¨šq92¿®à“ÚIô¶¨LAÁ»ÑÍôc€·Ô4{¢½P.Ð)8â£MQ4Ó0$:`6VL‹žiŒPYàáÙm€à‰—ÛKºKzÆ®}v˜ì›2w„Ô5fXÐ AÖ™¹Šç¦|÷*ÖùchÙ2ßC®JX¦ö€ATÈï¸ÛÁdIfŒo>JçÐØðmýpy*œ–ppCd£#ÓNÑÈeÑœ“ë[¦ÙSÎ\…«Ø§Vr 7t•¶ ڀDŽx`^tWŸuA””Á Å覿ö±:Ó§ Ãær :Mkg£çê:HøÍPüCÓŒ»"ÊÊìEß{¡ àyfï¹p¯'—³€8únÁÔJ+ÑÙJP¬[¸;ý^t'ÞáÀzo¹ Ç—¿‰š¦ÖÀÕ\Fã+C+#6cÿ]G;¦Ž‰µ6÷újlðù¸Ë^I»¨˜È*Ežú•plú©õTlVê±S¼ˆ¢Ý’yÏrfeYOÏÔªhÙ´ 0À ãd”"z{†‹TæÇe@î€tô®ôÃÉ&£Ùn££™ÞÝÏ`’£¹Z@J{8.öGÝß–¨Êÿ¢ì³œê‚xSm]í×z’-të{0 ã‘ݳ3óuûCUÀ—¡aǸè3%ewÅÕ ;Ú¼ógûÄÍhA*VŠi¨ƒ«,/¾·2¸e4"2é~§P—+ô:»úê´1T áDjÀü:цàãy ðã—î[ŽøÐàâmæ%ËöÌ X¨B7Á™”…‹D>¦r“`X}!¨6\V1ÀlåƒxIf̸˜3”7¢Âä4µÖF8x#ü*ƒOŸCñ©|]2/raÚ/ÿðE}'‹ŒzãÑ ¾¹ï+V #Æî˜†õ-Ó¿$±öh»#=Kv–€#É¡¥.W’ç_ËZA’Dµ“ضçÌ} !qV‡EÀÈÈ'ü”—W8{Y’;në$­.ÝBÔ™³êZtžf.·­äèÿÖ9|q³–o7÷ŠñßÖQ2Ëözú£Î”v¸×aßÛ@ëœÝ(ÏI(ÏPŸã|µóÞöÒ¶Ï¿xèqŽÍ”cÞâ•8ƒ<î #6UZ±™‰úz­OðwI/e5Ò0(´ÅK#]:­-KH½F­¿”ÑS4ÇRrÓÞ½}^Îyº´¹z Ëq¶Ð;R•NZKL°÷šµ±½ÉJŒ<Íð)ùb\-ÅGDG+e`¦·}ãœúéÀˆwÝfZ •ð çøâ%}é>%ñ]ituA‚ŠD0ãV eÁt‡KîÄN@ÖI*˜•”Kì;Â|åN߯=?²~_ƒ6$ ,ñlwÀc.³-׫®>@è[ø Ð3«Mm|÷H€±ù—`2œœ¤ë¥,x®WÈŽíèu,œøQ¿é\œ­š-õOÊüyŽ^ † ­˜?w¾{ÈJ¬ÄÖ"Œ¨x­ø2õö"OŸÇræÝF7ÆWXo Šø9XRö‰cç„<˜>Ö€s;­û7A¯¿îŒâÀÝQ_öàüÙ/¼Ü㚟£åê ³½j–KkJ1IIÉ+ÿدA×n‚Òz È( ¸Ù)¸7WZó1?õød4ŠÐ(¯ÑŠ6 p6%ÄÏŽåQtÏò.*£MþІ/µü* Í­.¢ Sko–›!аXÒ½x—::ü%¦üT§9xðï$Ct>+=e8Þ0°p±q¸Ÿüôb¸X¡æîˆUn §“O ÌÐkzkV›Imç]›$j^§ÈUž­výû•Û¡vpÚ¥­È's?o$ÇRÈ¿+™xU1Ù!»Û›Ø8¶¦™%×ê¾üô§¨FŸ²‰­ê¸¢4±,ܱŠò¾½j lô¨­æŽ|Æ•jÂçÊI,-˜"Eû‚D ydâr2Y›3^Ùx ¼ù'B2F¯°}@ÿáŽh¡9I }kDuŸmJÆÔô¡t ÇRSa¡È×÷Š"™pg<ªK¼®CSµÛS•îƧP“)§¤©Ûf®Vì'àf²¨v‚i/¾÷}ËÁßAKs€ÛôÅÂPý9ÄÂtTIÛ̆6l>U‘^°üÝ;Å6ÉOd2†8%“^ðÓ{YÙñ·Ú«JýhÖ%@ ,¶á6N‘Å à‰:!c=oŸ¬¬[Wu2q7ì¾¢ÅSßGÞ9õÔÈäí™R–ÄbPtíóã÷¾Û½Æ­ˆ§] Nd];œfYFt¼t俎Ûl05øÌ=‘PòʹMœVÒ®"«}Z¯ ËbbþWc0r†‚jQ{8 ØVZ,³ûj™½·œl2 > ;³;ïswž..Àʘ§RCý‰0ÅÇvµ.„Ô!Ňc’íÕåíIÄŽ¡ÃÑ¢`S8ËÂàÀ¹|,:ä½U¶½[¢Êå=8»e¿FÄWp$«¢žçÖÅGhý, ÉýRr-@… ©€ør¡9螺SFAvÛ†Ùs[)s ’ÂÄNrÖ€Cåý©¢.Âì4º¹ÖÞ0ü…\-ùÏfí•L°~hÊ|ÑÈB †'‹uvéÁ`u¹ZOЖHòU}PÜÃY“’é"겺gó§ˆÃ¥‰õèO[ÁL­‹e{¹î:úU8;wnRxS¶ Af kÞE«”AEq3UK§àž%àr("öDŠv‹¿b·>µÉѾe|íƒI”¢µeyŠ8Á¦âo¡£8ªö@5šˆOÛkm9gùIÙY¹øËÚ¤ ®M¦fGÑ=Þ ƒARÝWúÄ7Z)¦È9_íRbÊÒ8‹ÍÝbBòŒúå—­Y™° µËŠç[úÆzaFwÍî\m®ÁÜ= +PœFÄD•¹ÃÈÖm›Ù¹†lM&1Šé2 ¯¹€x¯©mac3\:“/¨Ä­ˆ·,î-nÇbùdßÊ45ÐËÑ( çµöÓͰL,8*ƒHŒ¥AüŠö ã]büûÅ3ÉyȺE}íƒ8™÷¹M_ÄùaÆø˜€ññÃÆe/)»WD·¦^û3¡œƒyóq%BÄæo¸„§›ëð¡çcLá”c"KGç ´¾¯¤L`ëêíöÑæ˜±¹ïV—ÌÖv×¥zCµ?"«¬H˜ˆŽ\´ðFë¡h—Ež™§7Ôâ9mN§µ‰GHíÃz¢ÂÄu¦´-îÛ„þóü¶X…Bâu|ªîOÒÑG«Ôõ%¢„+$Ë_¬1g:¬Ýï:˜Î_b¦æ•‡Òüzœ W½"Vmç¨ÅœìQ‚‚˃Ä¿cé;5NjÜYk²‚Š @à— H8í¼Õre 1pàFë·œòð÷kÃŒÆê¡k/=uëp/4u+Ë8»ƒv×Ù Ó{²˜‚„¸K 6$fÒ«5—œí”©’þ²ˆ>–>F#ÍÔ lÕù乜ÿM«§£*´å¾aÅ'éÊé¦Bž[ï×Ë»g2oƒïR&öp³£”"F"áklUëh„›<êSãôwµHEëïìÊÄ„^~h(˜9ö4¥RåRd:ÓLÛ€Ú+ërƆiÒƒ²OE¢¹ä¯iw?÷ßcL|ìL ÀŨêv¾L]ü˯:Œ|9¶¡˜9´Ùð]W]h)Ó²w“¡´õfž} w¼yÏö“ª—Œp©.ƒÀ”ç{õT¯ZbêTL‚Œ]#m o†¶RÚ=ªQxòüñRÿ=ª?‹õrµ7Ÿ4$±:„¼ÄX®Ÿ5Êê<ì¶l‘èUjó«C«>(BpUúöªòm'†ºgëÁ°î“è› ¾ÞëïV¦¼¹LÕ1ƒ½·C“÷ÚAÇúÝPÐ7§ó€¡zy™IáD·$ƒ !ßRÔd@#Ê>ãóia4B…Ò¬ZQòO¶à^þ_’Q‚”½D§ Ó – /»Â"NXξ…Æû»VÄUæ³éª#{¶/Újðù¤Œð€ÿd|õ]h%¤ÎdeÇÌM“/âzk§ÜCm%VŽŒï3Wd¦cØÞ:F•^ PRÙª@+³à™G=?@Ý€~ì4S3M&®"0–(³û^íiœk'<ÍÜ-LbYÔ­ÌKZ÷§ŸqwL¬¯ëQwh}&ê'±éŽ·óèÒ4•l÷!•ÌB’nJØ+É6ø£a®œ1À WÊ…å §o ë">аí-váÏqÜl¾÷î'¯Óû‹E$5J|Kœ‘$FY32þ€¿W†Ö;²œí­›lZôÈâOEâ$hj,F}óŸ7_Gwù´+º]¾ÑÊFzïˆÃF|?íˆOâgÔÙŸRªü9uBÊþuf½lI»–°ÑAü°O µ½ ¬=Ð6‘æ7òû³ò¹tŸÝX !¹©–\ ìº]_|*]}תû‚€•¦€Ã‹Aføy°'™´zw¢±:ßß“Èy°·Ò² Qí`ÁZh &¦ì6íÝ+€ö‰n%nϦj—3u@N¯Ž/®GyÆ«4“/Üštb•¯÷b@í,áÌ\þßÙ×7ƒºqÙ&Xôpuìr Œví9ux–<¸o—ÙéDÚ˱֭C-sŒ›ÉYœ*gÞ‰ÄtRöWR…5)ÀÕó¿zð­÷Mª×dšœ¡“$É}°£F¤wNêC¾%BðÕ£~Ù"‰µF£HàgßÏÑ.Œàêð1 ¡Nκ°XoþA9§ïÊ9$HÛäõ”5…åºË}­é½£@Òk&XÀ½Îƒ˜à„Ò»šžúæ690`Jrs£$äO­˜Ri¨v»¹N³J0q¤a÷=Ey4jÈá1öL_Öü‹Ar`XõøI!mž¡CX” 5h °'µÌeÙ³t°ÆÛKû¤áÖzò‰È+àúv<Ĩ Œ„ÅËÇÅ!$6fJOZ:.»àÆ8sÜÉÕ\Ö"†Éɹ«BSM¡ßllc‡áæ'Vq£ÝÒÐÎEŽäuVonB~¼ùf-BþÙ•£Þ8$«åâS ‹qǛ˨jóapa&ªrž b1ýÓâv°prAÁP«R»ð¦‰‰¢“h]É×zæý% êKi¶ÆvŽø ‘et—ÄZpJÏ9 0aÍûðjjÆë6¢ÚãúM¥¡³'«éYA¿¯Ü…ÁÈ/ML©SY£Ö=Š/8´zKzÅUW®Š¬e¬îí  ¯ ¨Ò¸³´×CÁ¤¾v ŒÛÞÆ™Æ2K%Ëd_mPç8ÜŒ‰²óº¨„¶†Ÿ‰–„Õn>¤*âàa¹(UhŸW£’4} uˆ "R"/péÀ`ÄâjZÙFT›XØþ^%1'ÖÏ2Xh§ûïꤌu‘™ÀJÇ-ðBñ+íÁêžÞg߯INw<*1Çcí)âÔ¢Z%‹Ø‚˜¬„ÞÛ'—[¡ðÙˆ<ÓÄarJåø=úâJÎ_&ã,ưìÚÂ#š¬èú:â²`”䯬ACÿËÞ4”ÝM8›Àv¯y[Æ|·ÎgU³ðú8*74ÖîMì ŒñæWæòLîôÖïœvC 9ľ'?ÉéÿÁB¹ö ®R-ö.ûˆ£ ^¦^<¨Ì\Z® ˜/‡us<ë£MÖ¤‹±f‹øF7ˆºVd$ë&^v»;³8ñSò¡–2B:æ¬H¯ªx¼Ä{ö¢”.oH\ÛÒÍt»žJŽ%Ú=;©òà-í½ÔU›ÊÞ¶9%gäÞOfA3Ÿ?/‹«&#Róë$¶-Ã(ªCa}Ýù×bZ¡ùHñ*ñ–Ó:Jõ®‹qsŒ¾z‡ b¼ëˆ4G¿¯@[¡‚Ï 'ûBƒÔamš[å[À¥Þõ°÷~Àr¨w)]¶•¾gÆcƒ¤”„J¸Ü‹çB«8,‚ÚèǶȑ#ÞI+-©ÏïªTj7ÝÃ9ŽŽ÷YtΛ-‘>÷Þ™&¼â¬)¤¶V{kQ0¥)Ÿk4XS½ÿDèv!§bWÑ®“1n=ßô† *CöôŽèi gÜŒÓišj Ë‘Í Àzò„îüfõ¨å´A±°½;¯ýJ±¬‘ øû¢ÆbqÙÊ–È(×ÇV"3{[ a4bH“ÛÝyI³½¦àP\Všàçö8×ÊM†ªžÌh].Â(­¡gc¼;IÈåÃÛ%ìî°Seͬ€è'f´Þe%¸>Öù½t‹hOZ°Y¾ÅswRR,;ùýtíFî¶9ÏשC\2rŸ _~µ ¡édHPÊÒf‹‹µµÆg«jžÞÿ¹³åè±sf£l Â=¤j³™ÿ¾ªœÂvœG^PÞâXÛQ£ý{µbômY[® ^'êÃédûÖÍËOçêƒ  †=cCMý·Ô$³´å _Ó#qq£EÂ"æ¨?~±"~û¢¤þªK57csÒ9Á ‡‚sp¡Îbt³+{ 8ø²1öáwIfî`ÍÐz  &xÏâ§·t2™ðr6æ±µ·°ì±[©s·êFaeŸ2+g±¦éë3'T¸‰©ÿÏpÐdé"âíR%™¯Pï!×VyP‰W¶® þÁ…“ƒ“lÌ%4 `Hwª-\hÅL·«Òy¿b8ßrêÎÉ4$°IY#WG9\+©Aâkd9{Y> /?‚ìą΄´†9MaëxNxµ pS¨žÕ†é뫯uX…z_O«¶iæ`–Ôcš¾Æ/{›ÔÏËd-8ÜNt`’¯ê°w¼Zâkò_6b£Ù>q·Bƒˆ£&½$8±%¼ j¼:ã >[iÞ[â¤Ò T[Á4oÃo~؃D¼zPHg‘09õslU.}Žð17…œýg~JwÒv”&ОÐ|Zãî½eZ4X«—gŠ» vpjÛ˜¬ç2¡”ƒƒj^2cNhŸ…˜ÃòØL®]B%©UËß9ˆ)ýÞ¥[~¦<¦LCX8¡¸Çtl¥ÈÃŽŽ+Ð’‘†kaó¡~Κ­E:BŒw¤|*œ%Úû¿éÜ£ŠäSøÏÿHC•±Æ{ ) tZOh0Ϧ˜Ã,³ÙëÜöÿ»öt˜; õåã펋²äælÊè‘víd”ÄUbz]áKϧH£;BgS”£Xn³ÊÒ£×ø½xCf“ƒ;ôùÓrDÒ›Þ'6æÚ¬ ¿•¥¨¶ÆÚF¯§“]Úsž…ÝŸ=ŠxF°é]hÕµ’”£3ƒE¸ª ø$H‹=ü2Zgå|º–”®æ!XèÖw¸¤n¸S‘µdj„*lGj$ªžÍÑóÞ%!{MšLç²²Ðl8)¾¼åb;Ó2(¼Ýd^¸¢yV"ÜÔŽ]ïŽl·µû-ÑJW³Óõ»{#,y–—cë>FûÍ7 ö‹síWy0ÈG’ý†½îäó†K[;i5 W¦m_+Wî0ß»ÅRkK1OˆðÊêݳõlvÜîúà6ÕZiva›£ÙñÃõˆDB |É ¡8@0pXMG†¹è–Ÿiß´Ú&ôpA=‡MÅÕŠ·õZ˜~‚`µTB¾– ÿ¼êñ‹ÕyÖæ`4k“‹·-%JW!ÃÏz/DŸ†HÆ*Os˜bÖÈ|¾Ä¢Öw-½õ±S©ÙMDË”DñzÄ#ÄEó§Ù#ÔÏt%hidF”:[i¾ì6¬þü=fIø„6O¯zÉ2 3¼Ã.i€˜’õ1f hs½ QT ûás†Ö¹x÷„ý„>R8T>,TYÁexa±‹¤‰w2縸3ÅÊqrýf~NG,"\j»„øjšÉ¿ŸòCÅHiQlrygáDÙY;IkÙ£^ù[À¤¼XÏ,uã Ȭ6]´®MþÏœ3iÑ·€ìगšDæÔw¢|hasDã<ÌÎ?•5*^á!Õ„s6ËNµ ‡á-ÊUŸh¯8K0ç_/„ò}±ñª+C;Þa kR•±æÁª¾÷ù·º§{U“ú]Ì^̇Nðsg"¿øXÂv¥2¡s\’1ÑÓ )‘£¿íҊʽòoºrzÃ}ެqɃiÁO&º‰9𗚎QA…0³É¸)øŠ/iDÔ¯4Ï' ކ­ø)˜"³Mj%e°šåòP‘üExZyÅ7r•t¡sÉçl0’ÄpsÓ‹ÄÂ}ýÇ,G~bZú€þL{Mß[yTu(Ô¸5q:чæþÓàÚ­´²tƒ ¤dý‚åÝûÛ…/0ç¨ÛЦv¼Ef3DTü›ÛéKøÂ6¦˜rҴ϶s?ÿùã5º/?ë¢vU–ÜFý¿K·œ®¦/Iepkópôªæ²Óì˜î¼q QHÛo~¡'ÈÜ9B°-o7FߪpžÏB7ÖÈký®’ü¤SÈÑ•–Æ0³À¨@‘úI¢Áï½ÙD™¯ê<>ê¸~éYhŒ?#MoÂBx¢‹’ íHüÏÞ‡ŠÅm¤ ã¬ß·ÎüÚ°­· ôÿ™|Á´%þR çÅó>pM m»9Iž&uÍŒéžêú>I¾®^ïüI1ÛøÉ—ÑækÒBÜ9ö®­E®Ÿõ¬-`˜y#ÍWõJxˆ—ª­þõGÆöp`­ýÊáõ‰‚¯ìJSá =OœŠ?Ècúö×ê«^ÀÕì04ûØhv‘ØãUèäż#ƒA’*ØÂê4‡Ou/·uïV{<Ñê/Ò ia(pÑž‰ûêhbˆ_ù¢Â Öm^Ø€JhºVL)`7½YfåÊ›ÑûÃ,8«ÒQÚûÂÖ: v„ˆW †…¢{†Ñ¿›Š¾0!ÓôcTÈmåb•¶høUü@Oì_JÐßKáKiýíc£¬÷ AÂ`äJû¹•jЇÍ!Îæ‹â=·feƒÃ1’QU^Rïÿ6%}rÝîZ¾‘"$‡J†ž-ñIŠÌ·Ç"üz'ôdy-BúÿǺé÷ö¸_ÇbZÃÙ@5 ìEÒÿ„ÑÔ*Â|ݪÅ;¯C)ª¤?ÿgÑÏÌ£’®‡û¬2Œ³bJ%§q»’2ƒä¿›‘;¿±Ýàç•|óд$3`­v1£©ªç }Ê¿è•I2ÊœÍ8O«a"1¡8þ-$«(ìbjæÛÈ zƒŸx¼âåÃs;èÍoï¡dááN«Ë»À°w€ï&á¨YG‰ˆ…ÑJ„‰mü„S¨{º¾Ë§Ÿ°h\N4÷òœ ;è7úuÑLêã":™eÎ'ÈÓ“^¡>d‚DÜIâj,-]©§¿»y¿M:²eÔF°‚MýJ… &Y[pA@h@µQþ­»õ ‡,õ DÝß]HûÃ|=Ö§ool*tWà…`7“ïÆ?[£È[Û¿ˆfŸKÓ¯U ŒÃ%:ʰ–ZÚ‚«“Š;S-yGGÍɼ#m­Œlr“ëŒÈï2L®$w1îȺ %Á…N · O!æAX˜ˆà²eÈç†Qä5¼á k&0ĸ ^…›Æïþ6—<ŠõßfšãV ü˜¢np²Y ã¥;!à†Džþ«êÁïý6;‡2Ñ]ü£Âà}ÉZçW–ÛèÜxë§ãà1À-@À¸nP‹ÉÅÂØ‚Ñ0´|ƒ3—ôG Jw ` º¸9²Wäù¥þ"ÂtëÈ¢én­¤ý¡>ËóÄd„å_žØHŒª'§Öq±·,Ĩ–ÎÇŸï—™ÇC)‘÷¶¿  ¿:’Aþía§tVÖXÁmðAÓ{j ÝkÄ`¶T©JQÎHÙÈj†ÊS>sSWw§quÿù™îT@ØV»ÑÞ*¨WÛ}µ¨p•3l} lN„]Oq4§È»?"¾ëúŠ$ Ñ nW"TלSsÞ|kÚ­@Óë© l“ô×6Á0 MÆ=EšôäFϬk+÷3ËÄmž§h0¬Uñí<’Žæ.QÒÖ™Õ-¹ ©Æ«„£§­Ä»ÿÒÒÇÍ6΂#èkÓh|*/‡Š¨£ÇYû8ý‘ÿz —7_¥Í5ÆBî¨Ä“°7-$ &X¸ÔÃÛ­Y«W]÷[mqÓ’ dĪý±b48?eŠBÛkQouêkU“€ ¨²¾žÑ̰ۻ2LŽLÎpðzWfD{züÕèõTwáK‹®Rð‰ÕV¨AÔËv­> Á%™yŒL ’œ£a¾Ä 5„÷EÕ’Óiu–§S•Ô]œ<¯j$e ÿëìùŸ”›Ð¡q d[ƒãÆ4·røE·Ñ–ÍU5aG‡ÝÕV1'J Ù|†Ç– :æÅÊéV‡c׃Ä5ðÑéýç²ã=W>Ç6"þ¥Oô_ï­î‚ýNß‹£›¿4m!03RKÐ?û„K•YôöJR |Åú¬7Èô¸¯¡¢wMÐ}B'>EGõ_zÍxÎç6¢ô ¢»\>ÊϲäÔ;ô@±òy!&DžiSL#Ôˆæ,5®ÕÖÔ®¼<5‰øýâÖ‘êý¥5?·ÛBå›¶ð)ªOå¬bׯ’LÔ#±4ÒqƲBþâŠ*îXé—¦°¸¥ã'5I#ø[ph.ôDû幄¯qÂr}¬ÌÏôÕ»^éÌ–pr5üÇM%œ·F‘.Êž ¿wdO“Vdv¤€DqËæmµs3=Øx¡©æÈìé6tb+ý v–ÄVPmã ÄéÔ ì+Ro®oàÿ!ÃM²œ_®í(z|*)lg‰ÌiXpuå¸(8 %ë½HÎé–2jZ‹]¤²@Ë÷¥O5¤W¯4–œ3*êhóHGÑÝö/i¿¸öÈ{gíP¤OP‘¾ôéwXHfY3^¾¿cÐÒ†÷ZúÚÛçmó¯Rý(K¦ÛÚdHšÅ¯Áàýl~"ɱ'–a¿Í–“êq9ø¢5íGµdö #órîw5Ôã×ÕÇ“/FaЖ[{[ï2*p>böõ«–T»–jÁŒ{Éi ¯bã“pK½„ýÕׄÑÝÍÈ…¿ê;DE'Ã8ÿ¶®i¼VN¹(lh'Z×úáv}!´ZÔÓæÏ?+h€’:jF6ö5YζL0øÝÚù ~3¬Û£©ˆÕ5§¢Ë·ÃéKÐèÜûãˆØô¥{·EC‘ñýÉ…ô\ÐÃÙîíÄÇeÄ<­á'd“¹ðW- e[·øs¾@Άzæ xa“6õú[¾ ¿Ÿ\ºY{ÁÀ´/TGBpï %D^ö,C<™Á3 BiÔÖLc¶ºvT è)-»îöéAµÙßhF¸”GÕbʦž”?-DM~á}—®FruÙö"X—‰æŒki¤ä'­·{ÉŒô-¥‹ÐåÝ*l\&†ê%ðê¡æ|é´»`ë½Á0þ$…b*>§þ³E=ëó.×ÕDwØ’Y̶à65¤V9¤·/FékQ ñVóx,RÙx¤ëjå3TÆŒ}îcÊõn4ÄfˆÙ8y,(—K¯x•~U(@å4"T(CŽ8a]9It²îgá4Y5cû‰\|LÍÌ~ïOâí­LF=NÃ1zç-öнjш!v:M\¡i?¥ßÅ2²~õ#=°#¤¨U³QI8ªO¡œ€·föœ_7.iÀÖècß[— ß²(Ű·!£ÞCsï%É9ôÞ€‰·ÅàåíD`׋‚¡SàEÖ¢á÷â"üÁÇáT!æÝØ1Ìé™ÔÒ#Ф±t±xðñ]lÞHïÝ1ËÚÈHàŽQà¿ýÕD¿N¢(‰!îö¼Ió¿HU­OuvÐXVLFm=ŽRêïî˜ óô}'Êk°L¦µÉ ×Q¤G|ª®…¡<ˆ—ß 1Ú¥0B¸<‘½%Þ©fPÔ§]©ÆQì~ê5d×Q. ø³p!WŸAç‘h[ttɱ¿xš”Ž+ 3TÚ¬7D ¡×£‚ÍBëur½°ÿ'å—A›Šun|’.0µK³ß [>¤ãÀCÈ6—?/=çb+g¬d™&ZÇ޹Š‹õ¶%Û)£þJ|FÃÜö‚¤£Î–üyØÄtGS$ƒÀƒ8j¢ÌÑ¿Æ5o\ÂuÞ5+ E8;ûÕ]X æD þÌ8GÆqßö ‚ÑW®$ëû¸ë\/‘ ©£ý’[³DnÿËŠö;ÇW_¡1;øGØl¬!ìLPzkÀí¤D1hóàÅ?¿“0V† {ëGŽWC$¿Ëÿè‚EÛ8là7쮫þ\gðJcŸ 4Í`º{7¨æ$zu *e´:í9½#nhà領ü#t¾ŸŒËC]Ü´+§aõbÄÃÛÚþ&ÚѤ¹Ö "ýý'u¦ígÞ· 4n5ß 3Tü©$4y"3â¥ä}Ñã0ÎÙ4CtÇšïÊžSÝî,ª²hRs „ë¼J¨öÖ˜s÷žroŠbcs|žÜŽÕö9Ú¹ª?… >fNCËd³\SÛ¶\+\éKž'§ $))–Xóò‚DRz ºltDȨˆ»bqR³\ÊjÔ(’ÊÏîös¨Ö •%•]?ÄÐu|\Ýê%®ˆe¼ì!LÛTTm²uRoÙ¬ÖHëò#/¯%g˜ï ®ÿŸ;Ú›”e(ÍûТc§0=NñF&†(P–îkJ¹ãw{| ¡KEè²±wƒïÄ¥ÒûÓáž@ãÂXû¾CFìäs$ ¼&ž¿5¶Ý"š+Rà % û@/^ñ³3&ÑDÐ(æT ¢ç[åÿ½û°YÉÏyñk|âsx…•béO‹·;¥‹ƒ*‹"˜FS‘ÍE¼–~DAÚÇ,ظ¡,Šg†„£äæ¡+ð2ÏŽ-;³à¦Ü¦áIv m† GÍ$ouª‚¦§Ù‹@ÆÊš ;A~\} “3<qB$˜t­µ%×fm6P—½ü2¡dæóˆ‘ÈY•¾ƒ§?AEJJà¹ý§Ù¨ƒõ …¾€sGí$lf _TæD—‘×FKX‡(Ö!zNæWÕ¨«:› rQfÐKãjcoìRbsß1© ð>àI|DK^·qyzU*…”‘l…´,aÄ.ÍÎï ¶éž³Ö‚¸ñµHF`›*HI‰DyŸ¬ØÞ‹ÐÖ%ÉÊ¥ÝWäXÃt¸î¹ô¯ø[#Ü{çäØ}ÜñWõÙÜž µ?œ¼Íí/i,‹á¿><ëQWY®aè¯cA s…0\U&ÒB«Ø8üî¡È ¡³D×Äùõä{XcýMå#$ûn¹Bçå'tcú»®Ítè{©è“ëQ$ Zç»<0°²‡ç‹hŠ?Ô_P…·ˆETMݦÂ>ŽpÝ÷˜ù[¾d¤² WÀ<}}*¯ßLsø£‚Ö­A_e)˜žåÈD‘4 Z0|Ïö­˜üè@ž˜úhÕÀnÒv÷.r‚¬¾Qµí²š¬irÍ~'(ö ÉÉ9ðшzÛbÎåӔ˩GT¯||…±åF¹iç '奄žHÝELè åâïU±ç"ÀÓ‰œlî æÖ¢ë,ùqÈmÚÜ7:ø$Bš¦®q‹­8Ò¼ùFOª¹ÜTÒ~–»9%+aN* þk…./ÿ±¥V_L .ÙuòO-bÕ¾þ™±!S–‹kNåø}]ȬA<`{f»AÇðáìvDÔCëö ·Ðßf¶“Š!G_¬:µ"Z^½é àÉFM=•/ö½Ç5?ÇåœìS*׃ö¥1âÿUÉÍ•ô9”¹ Ž¡ ³ò'?U“o.IZê$y‡œÛ=j! 9ö>w‡Cí¬D¸a2Àg÷uϹŸ¸ôH+n 1^•‘Hï·;g–{¤5—ß•çÊ»%QôãÆêèæÌ‚ùÒˆXúëi•ãjØ ˆ²ý¹ø1‡•$)tXedé—As¥;LùO§ Y%œ¡ÓôWY÷¦Œæ+4öTR ÈÕãHÒóß¾üd¼únFÙ싼üs¡;åO%B#Ñ©ðÕ›ÆV­÷Œt·>iR«÷{ªÓ¡|§˜ó'D!o‚.±sñ1ÓO“˜š@ÐènÎû.3{'6.¥c0¾ÆâuØžì€D{ ŒÝðJjŠj;5јlÜ9{ìΆ¨{ïZBQPµBg‹„~Ïd‚cлJ¶¶Ãò«Øó/KPìæº‚2!‘  àð K¤ØêŒ]¹o æÑ„k~/Í¡8H‰v ΧH_Š’Þ‹ÑÙDOojU’16]²‚b„sú«§ÛªC“$ãØ³¦¿q®:¨ß?‚ ”ÇB,yWö,ŠID+µcxLŽ3¢k9¢RÚDOŽ”¼‚áG¡ŽÈ·Z;tçQZOI܇XD„PhaE‡!½#µerk’º9É]kaÙÅ‘ÆýiU¬Ñ# ©.øÆ:Ïïî7æÐLN©!Äò4ªÿ~ŠËúOç²vP Q/ E~3N#>¹Mú|Óý}„©$è‰{¡à‹Ñ ˜&rI扞…˜¡Jp†9C ~-]_wp… ÷8´Þ(‚z c %ûVIÍßtcØõ]œùšú¥¡Ü—aëQº²Žk©ƒfWÞn¨[ýfk}ä mÖ0á' Hm.užœ)Áºæmt(X}õÏZ§PF)Y¸éu†:e†móüáÀ‘G‚Gÿy1²b$ð‰‡`1a3éúr¹@úÉpÌ€âÜÞÅ¥ø 5H S2W•ܲFõþ*±íÚ…1:ÈUÓ\Ïñ§DÊ"X_ðÊÖ… cV\©/¦'—ܯ«2v¼x;-¶þiÏ›¶2eÔK\hœ{(°ÙtåLEæŠ*Þ]Š%·“Z©0ï(zÖ DŽí;@€óœnïòó ê öG~wÓ½‘ûÒǾ¯çø:YÃ@9ùdÖœ¢!CÆN#%Љƒ>Ï Í'ýÃXÃhYslÄÊj¦6Aø¾ñì|®~®2"×/r`ËΕ!¾WüÁÒkª¡+ˆ‡V÷Û‹€¹ %í¦<Ø ûß~â¥R³7Ó(6îÑFµ|æ@9Wš·©(ñc*‰ÿ•Ëeã·Û5Û -ôÎyú½@2Á¾Ñ¡„aWš>Ìè7]F#ñÚ=BÓ§_æÀ/Pø;ͺÉ/@ ´œò#q¢Fõc¤¯©Xeo›t‚„“ ¡¿³»wopƒñƒTýÂ…™ÄKå]Áùh_æû“È©C‡I9à\y@b#è5|5U"û}ÜØ¤ÎÔ€®Ô}[TWÆÊR¿”_P+ß™al忨xŸ_èqâÊÅ›¸úö>«èøM¢Kq㤽pV1ÁFÊ­½|€ éÖ_‰!¾´[T·!G¬ Õ££øÈÐiÀÞ¥Fl XN+Ÿù½¥*eÉË9žÑ«6 ºX5ºÅÎ{/ô½‡§›yÏðJ{§«©FûTyàs¼è$UŒÛÅ‚&X/·ó šxRáv¬ a;j”ì kKv’ä·€6­vŠ»‹ÑÄ6pzaC¸–¤g Ѫr ¼Çº‘1‚ вì´ ,Æ,¹™ ¾x¼Nºy€ Ãã<ž¼îœs3ŽJ•—Œvx®> Ôαªø„:\Y•©•`Ì¿'Äå sí®ï3u4Œ-–n?Á÷ƒñŸÑ-ÑÑVÐÕ'j²È@YuA´TÝk‘¯çû©ûJ}BÉ€£½M‚dR圅ÒêóÈ9-*$•ìýêôåç†Ì¹¬TKÚBeT¿0ÊÖ<#9á‰>yí‰A] <©L6,8ìÕ.1¥ƒÓAÖ[÷5kâæàÄZo—vû j\QSB‰ZAØRËÙÇ÷h_tZ‘HÍ€º•Kœ©Gšœ·T.·$áZàÕÜN ó<ÆãØÜUÊ•JÍ’ÏE8C¦à$P³~†åψâHåuèã•£;+>Z5ÐBà )ÝË GÏÄp&+}`í&š?aå|‘¢§°p–PøEhüJÙ#¶¢1¨ ôrv»¡¿SbÚÊ{X;“@Ùǰ„EEÚ%‚O¤œ&Þyaƒ1Ìzq¥„cñ½Å%øCGàiÇéf¤Ú†ßá äFyŸw/;)âÉõjM€2ä1I‘UvñPg\0ü#˜§-M)Sß³«E©qp4ç5õã8š#†ÓÆñõt!þß¼VÁÔ»[¤â5¯„ëQ±›£ãù“ƒAd<Ý&˜ü…¹õuV¨^L£ô dW€Þçˉ¿Û‡_ ã@J7o¤ nhá8´½Œ'ØÉŽ¡†Ä‰äQhÚµOŸ|¦†‹r8TÙ8F?µ0cÃˤՄt}4®:#§ÍŸJ »± Y|@nH-NtŽ™î¿OÑòÉðΪ>>û¤nŽuBªÿ>Ʊñ,˜:Ú ýóŽÇ7ô<ŠÐ¸Qëùó“Oçß¶½0b¿Š ]´¹ùq .ñgop—~cT²%6)Âë gµ,]ç2®—`¬MiGÿµ¡Ÿ2W5€ö¯¼¤§‚IT@•‘ܽoç2H-5¶Ÿ™“£~ÑÖ‚5û †@´>©D§ë^[s‚")šY3€Z'ˆ[¸)ŸdC|7ñ&‹^¸FY]°‰± 4;A8±ž¢ùMy檩%¡œ(¦2Fã…_Ãh£òÑ%3<‡á{øÊu á’·.)ôz$jøÖ4öûÒ¦³‹†/¤å8¸dÿkbš…Å-äv“åÐPìza,ü>˜¢?¼N†¹ó–xµé}“v !.œ¢¤á쌇úO/»©ƒ{WXjM™ÅëÀAµ'UªÕ\T~R+ôDÌÛ>½1Îen±’QšT…w—À…–(iÚ„¢éÐüCÁùŒn–túêlü”’¨ŸÏGЀבX’ɼF—ºÖŒ u¤° ÅRËx‚äpÑ r:‘1×±o±'j®Ü÷vÿÅ(9¤²”Æ!—ÊÒ¿;¿¦ºÓKõñ'X•EÙJ´Óšù~%…àr®S`¶òE,› Q”®¢»l…01}UÚ†3“_…ÂòOwpü_]¹Í'3£N¯Órp‹É‹ØàÈ™Üqao—ä“ 7IIùœ~pø¿Å¢ÙþÊ#æ•‘ÿ‚—ã"0—]\#¤¯ã>¡¼Òƒ^¯÷p‘óˆ~„†êÏÛJ¬tvsx.Œ§‚ >j˜­ï «ÏÍRžz{'mÁ ì$V˯NNñNþmhû ;4±ËôDÆ%Ц€ïK~Î%ßG¶Ã«× ˆ9£O£n†x=,)x8!v«ú£R”@³. ýZÙåiMrÅÿQ!=O¥ö XJà ÏtŽ7„,EÿÃÈ߉Ìôkè/žQ†¨e*ãu%ÀJ(^ÍèiÕƒ®ŠÛ‹ºóž?Ýü„àpsiûÁaj¼˜µÄF¡ '5èÚÞ^¥¬Ú /ƒZÿ¦ºj`Ÿ²(C§°q¦„Ti°Ò?©eµÇ¿þsÞ`—6åÑŠˆÍ±åÍk>âCËÉ9ÒQ/7ñجÆÀÆÕß¹}Ypß—ýUDlò„¥,¨inþ‹ 8ÿš»VrlÀÅÕXÝÑ»ÎèJJ·`å°ÝÒÄçÖêxó"An®IB%mAIÆÑ©8)¡ÐØ‹7q>Ù“¿ÚÉœkã8"hýzcÓl€jV·ÌÂaµ°^1»cüÛç¥Ô¿·g roQ­ÔšUÙšrD>LbŸƒëÖ² Lš#!ú3\rcÝ l›–±ÛØü¬!šÏúÊD:!(ù72é=c"þTé^Û…Z=8YÕÿ¡Œ7þòt­ÚR¦Y*Q;CþÑ« ¢¢ÌÈú´žèwÈE»k¨éõÖÁ"}^ ?²бâ½H`è͕ҲÖFøCüx9‘~æóu;5˜9[ Œ­/‰ “§»÷è•Ï,Ç5‚²«AgquUáž-3ÏzSÝ C2@=†IþÓJ|Îêþ–lìÌ¿ô'Ÿñ\‚P ÷¤faG?8}Ýua.î J þ«l%ÔÒ(œõ·ñòÃ’ ê=,<Ã4)º§¼6‡±¨ƒT²„áõÛoú¿¦ù@æ†pú£ñ›±ØÖæ°èV€üœÌœ&'|Õâ-aˆR”õ ¨6DZT}"¨¿÷¡Žv„tîµ÷œ“%=+>X7Þœ*ÜfñFúžYlä$¯/ V»G”¶üjýìÐé—­(’ø¯ÐèÎ\ØÈŽÎ^òx x¡RÖX~°Ô3»&±€ûÚhÀW)Ú“Ëiˆ·¸¾Úèá$öp_¡ã|TUS½óUjôAÊÈ'8CæÆÅï_i¤º 2³M^ß=qBóêÿ%û¡Èòæ·ö‰Mé¯;Æ[óþkM±Ò)y§;žt»'`Öxúlk©$|RóFš¶¨b2œ,liDdÈs£{¾ŠìÄ‚¤Í¦VÞ’-ˆ¿ê—èæ2È嫯¦Ž ˆDH\£J׺û¦ÈÌ«•_³ôU8àðÜ_Åj ›²õ¥¤Ç—,ZÄ{ )a½®`7.oº¤DE‚V§ŽBGõ_³Ò@û4 ¼‰ƒ'äŸæ#LäH˜´9­Æ¹Ë;P8Œ zr…Qc©ˆAó…7Ã/¦7˸nÀ îAÅí45èæ# IÈc“ó]OWEú4sÇÀ­D\Ú? ;6u¥„¡°UUêDJ:T ÄŽr²óü­‘U´Û7µö¼ ÑðsøxÈBRxÜ*94Q—$?¬_avùˆ> †›:jœNÉ)ׇ{“ŸŒó©Œk…|°yÌ—õa-ç‡P‘=&à÷ ­¼¤¨9+—Š1vïufØ:ƃܲkÃÕSX¶(Áô¿÷ñ•!wÛ)Æu"ŠKNdÙ0w-Ø ªZ¥ Ež(#}ßÔ>¿_¼@ªªø3ÛÄ2öÔç½>7eV~þ¥¥ØÂ!Œäw¦–t¬¨ÆŸ[ߪ-îø "ðÿâ¤OÔoCL“ŸF‡oÌHT-wòCƒVÉ­»G[”³Ñ½{v7YŸLA2ë#ž±6-ÅŒ‘éü„mã[½¹Ñ#UÜHoÆ9Sµ\‚¡‡=ÐbR¢Ñ~%’ÎýäêíPfgÛt„ƒ]â„v‘@·á…`‡ÉÉüT¼13˜A¶~Tlýë;1µž®&=#ªü^s—¢~ã?\áP=¼<Ùó§ôÔaQMý¿gÏ“l˜ØoaŒq2†˜½ŸÑ£qq×—ÀaätǪÈRØG2uYßИܴ´tCûûä3ÛW:Ž+}1иdO&|dž’«wx%Œ¢ôîÎGï]ƒ‚Å—!uhÏmáÝRÓA±†i=¥fÍõ—µ0´9÷ n•‘«;ÏþÄŸŽgm;·¸!¨ë2þÀXx®Ò¤ï<#I6ôíoc,[ÝÊ –Að¬d^$›Öbš'°3 >¡µ#€iúƒ}Ñ´†0ÎrèlYÛnòÕ>³Úä¬N‹É­­/‰¥ãu_ÁÙà’§Òor–5$I+Hª£!YâÁ“^f•ßœLþçÀ/Íwúy¬ï@üÿÀá5ñVB‰ ‡´†…E »­_Y ´j€žÌzxq%t¼O“»¶ñÑeyO·]¶,ˆH´?MF®½ PÒåÛâã.jÔÃËþ8yT Á­xhæ_é®uN‡)²Ú|0àŒIÐnÆÞýef2úþÿ§§Ç˜A2`NÃ$NÃöæßƒÛNâ–v sœ¶^ÿZn´?‡…¿:.†&‹£_(¡qA·y:Ž Õ\?ý·•ýGd%Aˆº¸;~lW©Vå·À‚z _íÅÈæ-ä8šyŠõê `tÈî©Ey.x§¤'sçM@¹ÿ†õ£LÃ#ðŽê3‚8ÿýoúS±á¹ Œ†wtü6ÿdØè…¸•ßšHç4£X¹À—”‘àUü© @³rEŒ¸Ÿðkº»”ksèä1ÿÊXD“Ç:Oý5ÊæMËÛá‘]‹NâäD D,›³qÍ/]8ÿ–Sºb»e«Åøiõù÷:!ð¯íúNÀªÖð‘ð¨rꣶézOad+”¼ßDµv6ôBPŸ×Ú¤q’F¯Ëï°[Ü émâ“ RPõ:¤÷Ð+-«<}Ëר­†' ARÑrr›É1U«yÀÏ”k÷@CÎÀ…3ï—ý [œÑªßº´!šÂû_¦é·þ’‡ÿýgZ$´Q“N5ºTÐcïèHsüH¾,’˜Õk¼ƒt7ùðÂ8Å®B«ZÕ$Ð. ˜·L•0Ï„<'”ÑO@SP{~œËÄ{]&­â÷-]{ O“Yõ n18RŒìÁ”%KìJ¨¿>úŠblÖï@ýqHãÐQgŠ2j`ÖŠî÷Ñæ)d¹FW/|N$TÏ0Y§+1©'Û¨¼·ÖʹuýW¥ÇöYc]o„!{ÍÛ䜞0‰xG»Ì¹g…Öd\´áÜîÆ¸õ]³L™@aÏôe…‘Pí¾OÚâáÀ!ö J„£©ÍGÀ·ÙŒš 1ü‚ÿ>€Ž¿Ý·Âk{Wì¸H[÷×®Ð:_ï÷î(¨9ë‚$lëþ%÷Ï}æq6³Y*OGSpéyJ|‰,ÿÓY¥ú –í4stIëOM²n¯À‚Íï5ý `_ž=‚‘þâ–¡> 3RÒå{ã1éªÀÛžêH·Ç¿‚<~-}ÅCý).¿Ú"J±šÀ^NT­ ÕÆ_V‘0Ô^©l¼e„ÀÙ·Í£µ´ÃópbrcÃঢ0csTR¥§p¹Z2Þ¬ž"‡Fôæc+7ø¸˜i…¨ „«¾®ÃÎñOÒX9jÝÚƒ=ã¶›kÖßV©~«}ßiT>¯ê½hg˥Π$²S«¾ÞWBÈÝúÁø',°ÂÂT‚¨r`W°<Êz†ïî;áñBQ€*2õ„[]Ø‘ÍYð‚Úñ} ;K‡;"ºÇÿ×Ò&& À/Æð´„®Ý`àÎ9ïˆüb&¬,xIcÒ=°‰žžl»?}ù ÁÅv`£²]¹Þ•IÓšÆv4ýÀG–üRîïñˆ±ç·CÍÉ Jf—zº.‡ˆçv’`˜eŸð¢ _Â]T±s®V¥§Tø6«¢+ñ F(›¶!ñeÚœÊÂËzi~«œÜ¨0e-ô“çÜçÜ~ZÞëèNczÖ)óhbR€g®“ÅÈ)ár´%¼Ž|Rp{§éú€‰‘£G.óR—;$ò‡ØÚˆCf ÑÞFøbsþàÄsÀ7fkÒ5›íçÔáí¬4zôyh߉Úㆫ£Ju’,Pyù<‹óÐ7#½Oà!L²©AÂäàü;{†R¿7UX$p&Á+f^"^3 a76DïÍ!Kúà(ê·S®ŒÄóÚðÏ“žm·ñç©,'Ÿ!º¾b¾-ËqåÐmåƒ#O‚¹ ¼ÓOR¯ªƒ]6vÍœËGØ‹ÌùY¦Œ“!w[ޏõ²8ð–A’x9±EP!GŒû•ø¨²Ó¹£YÑá‰ñêÔxpW¿Äß×5Ø?ým£®Âü³X…Ÿ´[^$ ¢Âœ¹Î4Å#B*u}Ä|ô¤%¹&ÂytÐ'î>ÿx|vcÐKÖoØ]°*¡;áÚÒææËù¼ò%ƒÉ&iìŠÝ ¾9b•.enqx¸íÁq _”{HI®ƒ<",æ6]§Ø¹dÆA©$¯¦þj݇æ3“¡¯¢¦×°ß»«DðéJÊXVIš²0`zÐà->Gy‰›^¡Œ}¼J²êÞóÞt"ÜÔ—×îaB¬Âj›"ÞTê>zÙþÆÿ.†Ï~ÒP›A+ííe]‘ÞÑWþH&¥vHWÄW®Ü'‡ýmˆ-D =WxûR7©,tX»’ãRÇôžÀGs¢0‡í â‡xc4>ºŽçè¢^Ô|-8!Ïè‹ µ[d åÏä9„äÕÿ™'#C¸äÚÔ¡’cõÛQÖEuÊ%µæ_„ÈÄm¶–<µ³/y~\8LS|&»F—Ÿ@½š\\b×d Ä;6ü»·@1嫦[ú°›¬L»\½&Švðb¤„ÚQs–/}á:ø¨c»ïÒ¿cÔzmÆz©SeÍÁ°Áá÷ä Q”E:û“Æ$vƒzòY˜ý9G*6¡ C4c<GÊ…VÊé¦ñ˜¢]vS!Íè nóáQФ{Üâà]õ38ø·-DÅå¯nå!•µ Ï7ÄÆêU?ÑZ…¼íA¶5_›.OÉÕz“#Ô.Ø¢*ÍÞB¥zö²–ÛrðÉ‘5жOæf‡oýlÏÐK ä’ûæ:™NSþM2¥§¹xc3›oƒ´rø‚ÎTTHŒµe´pbqò¸ò¶­î„ý+mÀ˜¾yGE„3¼Â™ hchÈkqO`ĉö’MÉOZ°íŽ*ïK%Ÿkœd¾¯ùÁaHæTXaÂóƒ_ÆK“ =5gñ&íò¼¶ŒË)‹³8üÌmúÄÖ_Ôú¼ £uчÈI²Q”Ò†æê×{˜f/ØkĽèsüW"Î9R–¶-J!>€Z‘Œ¶$Zp°ŠwIS,ó&C'^(‰ÅEm™.už¼XµÕä´êÁyùûì»özÊÔ^R½°à|Llÿ)Olnä(Í6Î`]«£Ä<ªuë»ÄÓâ11<(uò¼Eó×úõÀªöEÿ‰íÎsî$Ø~V$·è/î«©{ô¤`_P3$bc"èÎyV÷þqm¼¼LSþ)“×Àìhªs Ÿ!¢C™'¸·5*k\¼…æä›\1çùÙ» EiIh>.1tq ­µ›‚»]*½¦¾e=KµP8µübYàÖŠƒ¡mHŒöM$éϾ¨P¨¹¹2ã‚üI ßW¿Ê«èX”0Ë➈ÄÚ°bÏB²»>Κ^Û|néÀ ¯ìˆêMý¢VOS»êCÙÊç¡S,2’Jèeú6ý\Öe«÷ ƒ.råîÉ”¨ÛÃÏmŽšß8vÏb'WBÈ#@vßkí0²zÀz]¿iaøNA£¸sy δ4ËúY¼¥|£Y¿Ëb ZcUþ ñ^­¥G´g³Žò~ÏÆÐö"Ú“¸(²Û4ÇR@‡<žÚœSTÆk_Ò"¢(uúò!0¿¡³^ƒC…Mé´8àî[sÁ2ûu H‚½›ÿáË6H+Ã3ŸÂ² šxc‹^¹Y ˆDwŠ|ÒKÕpÊÐ\·î¥I½[¸;JE;YsþæêQ|«MNnw-Š€f9 èÔ§Sq±µD˜çŠ%þXÑ‚ºOíð’VÕNŒ–(´nM… Æê¥îü Åë¥i÷‰]-|çiUvˆ*.œwàžUÝ%¿nƒIƒU{iÂ.¦èïÎ>ãS¤î1qÝE…<Ö ¸¶3ŒnÞ,8Iïï](ä-¡6w$*hhÄŽ\VÐ*V‚Í[ú›\jŽ?ä>*ÞòN!}ŒHΡSVŸ'òbØkâú?>¸2é ;ŸäܽænÀBVåó¤ÍÙ-o†xâ;2NÉ_åû ×døØßãž´ËÄ ¾!8±HJ\\fÀ]5Þ^Æ%g£¦?Š€‘E³’TÎn¬Âç,j§?˜‘ŒÐ ómjóÜ…­Ò}·{D_®a»W(;¶ß @â¼@d¬lb5c™§¶¶ [§s΄¨ø+šûËhÉE†¬“„#ŠÐÿ‰¨î‡îÚR^ŒÀ"ßbFßgïø¬NÛT¤‡-¼‘ ¿ñ]aI¢VåܸÖeÒ%më÷nêÄ ýjŠ F¸®5,vý=î…ŒMNä6‰J[þ»‡¼Å´“«6n^(L™Z:½Ó£³t£tÞ˜Ù|‚"»'‘²¢ÙOe‹ªRBEÓ9Ã,ÔóN ¸:ÌGðVÈ }xß ÙÅ<±“PÉœKሚ!OÊNipPŸd˜è«#{=Žbú÷RL‡^kä¸B|¸Ö$/¯$Ο¥og}¿Ð &÷ýVEäCvë“^[ ¼·vº”üe¨E»B¬ÜÄËûåC±NøâK=¨³øì+Ù@¨#» âå?wû•üóTªP¡³#Øk“¥§a—·ŽáMMWQ3“]39—„ë‘x«’qÏM§rÃu~•Î;­xeMË¿½ëZÿ7…ç^×Ü­¢kz1,ÄÎŽsĦï›z6ÉQcnël«€'¯}2ºl±Êœäãlû0G¿æXšö¯í¹UÖ1Ëûc^]4Ú#–­ZÞXf~à£ò>å^Æõ`óÔžËBê'_¼g“UtS½“À—6å“FF‡^~ GÓ>L›˜¼g!\{©±ü¾•ä—¢ÂèP3wÀ!ˆ^lCKjœøä U—Ê‹8<³ÞÅ(‹Zo.wñûq÷Bõv‰üŸ7š Ö¶1Œ®¹R‹bidaÏnãºé˜O­-*ÉŒåG-%UP²fþ|ÝÏB=ßq\°›kœ……µmxÓ¡]Úæ|»PµÝ¤;T-öÑ|þ[G-Ž/ÃV) —>ñ®Ot ž½1~jA¹ª>wN.~hŸSš£X§¯êxŸÄ§í+&ëöaæ áCÇíŽ GƒoclÖ9\³à)¹ C,¹Ðž±´jÆŽÖÞuAŸü‚Ýë–ÖõBIüYk8‘ºÄ¼,‘¸ESì©a~¾¦‚†Öž‚×Eà(¹÷oh4U]øË$¦3—ƒ_j²Šœùã Üpµ7Ó÷’ˆ=žÙÝŒ\x¬ô(¨GΛÂL,Å×€[hTVéê„ÈÛñü«RãgDÊÈRçaÁ§E1*:DUöØ[+š:4;¤éù“@OäIÐìvc-ž½y²)úÛ{ mÄÌå\Ï™1аxæq´Ç…Ô«‹§ÙÈï†L¡ÙÜYM[së<¬vHºnb®Œ%µº‘V<ßÓµ,‡Æ=s¿bp-Q§Cëê=h „R¿µRŸþ©»A ZQRbْϽÑçUÉ¢£Þºò4É™pߨXZÚ}O~ôIácm€\õzÂÞÒUF3H‡µë6:ÓŒ Iž—Êœ{´VÍôU•eº±Ã‡iæp3RÖÄï…]nZmÓ!¾iÄ£Yr%ºVø+aCá½}?pªÓb³pJ”ƒw FÕåâ«§ÑY‚V>:ÏצÜÃ!wˆ ‡4aéH?Å­‹‘eܧ˜!sÔj¸;í`l“ø1¸ÆwlððŽP›Œ_Bí3™•ögð&Ò3 ¾á¬(.üÍQ¦2…wD~çäA£$ ™Ü,æUb‚“§Ðó*µmø‹¤ïNÒCÔ¤xŒ±¿(×bsººd’lÝÏ 9MŸÅ?·Üµ‘Žå# Ða¢—Ï"´Ùð²Á¨ÖÿßÅä ¯œ97Î’|ÌuÏnôAèZðœκS‰ŸXÒp–1Æ^·ŸMå<4 sT3!9 h¨Œ)ÈO=ìÞŒ? ¿Öø;àk3WçµÎģѻ $µ¤ƒ†¹øÔì¥×ç|Ð =6| 4Q<>6Ým:žOŒ8ûxNâs/¹+4v6¾Ä‰ö½º9þëT_it-`n­ÝrÈÓføŠ *J°+|ñI·„7NÎ0‚¦£»ÿkíanQûK÷êÜ=ä¥Å¤¤ ªNùɳ[¨µŽTÎÇ:r{æÊéHöŸi¡H¬J©æ"¶¾jÀ˨ʷÌè¤Ù7UøtLŽÛ­[Â7¥HÉÿñ4јên25^¦:r¤ø”/éx)F¢‚¹šaöƘNaÀ‰rwqƒzâ–ˆƒ.>ÏɈj×&(Tþè—8Dº)¦¦à“h·w8 ž[ÎZá¶‹0^:‡¨îKÍ¥â³Kd,!Gк‹ÖKVén§š;@¨]X»üÚÀ5eªa²Wt+²s¼ÁNn]ÐY®z¸çÇý 1k‡?¿5Ýf€b-fö4зC®C¨¹]ÅqJËóy-Ì¢a2W܈Oäú‡¼$9êQðËÞ,7/W5Æ›/ð0çdmHù°aè:tê\¦ÎŽ·Þ¿¶J:û¨œ©÷­ç-&ûgÏ‘Ü6#Kº¥ç½¹ô/벞Ëáñ¾…në×¶è°’ÏJ¶®ŽnY]žŸbèŒèuƒq? ¢Wê4nç‘D& Æ»©bTò0_à#û_,?rÄKÜgSHùjô«Å’k¬$”`™¿óÀp«I—ü ¢IÏ„ÛÞmÔÂðÐ56wÈ¢çù9©aÓÔØ±ýÊé¹—¿:Ó¢—®nÛ˜Q-oZV7®w÷GñÊ8—ðÅpÙ)‹ï–]$‹PÔ±6ŸJ/)›Ê&NÑÖþíâ—Á/Y7î2`ýÈCj&p§œ–÷4Ùaº©˜æø–õ~Z¡dœÇ•¨C?i·Ô#D?B›Wê½ULáÔdüMË*Én­±ì¡£`¼.)ãkDGŠÎ{vâ“þ²,¯*Ùä“Éþ] ä‹JÚàÇ•›ü:L} Ñ¥M—$ig¦íó™šA¸Ã$g+Tx¤¤÷÷„„Õò*Ÿ2T #Ää¤Rìõ°›(¤:j¡/‹6s«î”BΙÚm€ü]¨¶#Ðüš©¸H‘ꑉP4øYëž £pýƒƒ’R˜•&šºjR¤s.‹zrž¹0 ‹YZks/data/quakesf.RData0000644000176200001440000011332414067333122014161 0ustar liggesusersBZh91AY&SYN¦+ 6¤ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿà øtúx. s瀓€ô Ð(*‚}NMO{)J€@’ª J !A‡P ¾h@ÛAf¯mÉ_lí‹Ù­5V[y¦Ø­Ù÷¶©§uÏvëNØ+Bžá¡ôúï°×Rmž»€ä÷: dQ^øn…>»¸¯§q­˜õ¥æ×·Üë½o¬ö÷°ú}{²ö€x|6T•c.©º¦f‘AÆ”W]˜3aÑ­màx¼÷º¯wum]6ÚŒ U ÐbŒm¸ôW|€{ÁpA÷=U-µ¾îu®vuW&U(tݲêÚUh4&Á 24 5<2di4ÀjdÀF†A“&2@1£LÐÐÐhh xÑé£Bi‚20¨4BLF€À˜M2dhh4d ¦M ¦“@ÐŒ©íIø˜©±OÐSÓ4§µ6R`ô zž“OP ÊfSQêdh=A§”Ðd!A4Á2 †#Àa4Ðd ƒ!ÐÔÚ™' LF#Fšbz'¤14MžPÉb¤CG¤L©$"jOÔÒa &ÔxSf¤ò=©6Œ£zI¦‡¤zLÓ@d  4Ñ£C@Ð@4Р¤‰“Bj=©6¦£Ô “24h4h4mM™4ɦ©¡£I0MÔÞ¨ÚžM £&i6F¦ÂOjdz§‘šž¦yýªŸJ”BšdÓA hô&¦Ò6 4ÔÈÓ)†MM=êLô¦ÔÚ˜ÊyMMOI mê4õ1è‰ê3DddÉèLOQíBi¡£ÔÄõF††ÔÓqñ€úÞš ø×ü50 ~ÃÐkú…@”¦¨–Nh‘4ITwU¬ÌÈqH)—Ѫ¢wMIîX<ØPˆ6€‹8-Z8)D‡qôäPL¢ rD¸–)¨\˜ B‚^ ˆ—«$¸¥RiBíC L:IRÑX™¤éfÑ$’I´D#jª†!ÌÀ,$Щ1 EtèÒe%DƒIŠ8“ A@4ÿ¥\=\²57Ež€–±¥beÜ™OPDÄ ™Ì–®)Æ+šÃ ÔW €EbqEö£‡—ü—¼ÆD¼‘Z¤dtõpAá Äo"{“ º·Ö‘?­hÚL–Š—D 8vDÞYÝA3—"¥ÞHw^® á{²ˆ¤ 0AÐuÀâ7T‹âK”ñ·ŽÉ×ëçmÇ„<¸Ê0] Éï0ªñ(_T¬˜‘\§{²®´RØàD‘z¼=É4%«88k QíäöŠ'p/•y£ª`îc¢ˆÃ" 5—|qLsy­ê˜+¥eAí7&‘`# 03Zbº¸Ð6»ä݆2 ÃäÇuEªœãpÕ±qâ‘i:ú EK“÷ØF Þò±˜¶*TËá¤g®LbÔ½ÖY©3JSäP%a›GÚeÕòb£Ý0 ÁtÔ!\ñZQ©VTYJb  $‚¤2\.­ˆ¸*_sÉ­×™†3ZúSj…ï×å¬&¤â„²•DڸšE&ME  8V3ÂDº9ÕA˜Ã˜Àdu| ­‰ñFT[~×"4ºÖ"ˆeÑÒ€ 8hЉ<·Ï1ü ö^5 /Ì=¤<ÿ6_ó^´âŠPRGP®§¢—fšútôuëõºœÎÌö:½W³_3à÷ÝAy~©Wû¤Òéc+­Wõü|Í ~øÛîrëè>n؈2bªÄtyZ’_>¢‡Â†_ª² $n!~eŽç‚#U~ªµrgÔmhTñ ÁÍ>ᄜªwè2=ã]"ÒÜð58&"¶žþNŠŸL-Ó³ÊvÃȵ¨¦|WÖcƹ¬“ÂU¼H^1ŠA^J*Žã ™ÐtälÅž©ñ <9†-Ÿ‹”k»µQϰÚÅ›àʰ¥fwÞñb_•™©MT“ë~4&L¹^kâ&KsEž&a–¿ZÑYYÐVŸw³¾¹ð_¤pÖÇ„áˆÀŸqqAf\O1€æ8C¤QR”7-n ʦ}_*'šu I-CD­x5-S”kœ¹O9<]3CVvKžË¥é铲J‚öêUâÒÛœ'’Ý®#ÃÞðæá_3\d˜­7]eé] îÔû'®]ÕÉF#KIáÂæ¤‰ïNqÍ/\æàŽ˜k‘rÉ!Û‡ N“ß!ÅwIfdR¡îˆänÕÄéÅÄ„‹,—QÁâ­Ô£Q\uÀ©HQNŽqâÎQqLRuÍ=rât*ŽEBäU]Êej*î˜.JqŽSN[L¬“]¹Ç…e'Bq˜jÕiX¢‰H”IRñ:ž¿×ëýAjuîN³ìæpGWÀ÷t¹ùûºQ©3‡£µ?cU|ý~)\™k…gĴƹ Æ Îæº•–“µÇÚ]‡ª+Š.[õ±£¬îóÊ¡•ÅÉã· ‰Df‘)iŸã·ÁkÚõ1'"îK]Iø7¢—G¢›†ô¨íΠ×W#5¦]°´Üwà¤üÓ‹›„ÛîE5ªYˆÅ9›Ñ,ºÍb¼ê wÌ"³9•Ô´Ú›ójÐn|8]zºêÁšŽ¥¸È–ÓtËËU¡‘À™øM¥ƒ®Pªµ¾+2bŒî@á|cGBÙïãdÞâ‘–Í9ËX°ÚØ¢Þeúµa¯_2«¶&ËÓ™_ú+*]ëðæ3ÿÖšnAI3ªÄŒTæáÔͦ0É»rí:t¨Ê¹VD4ñÖš–§\íÏF¥ÅNvî(²rœ±Ä–äÀÃmd7æmùWnµS«CÊqb ³eõ´ò*ßÍ´»62f<½šÖfÔ¹‘zCxö«ÚVAÝ{iVo`©{6äŒÑr=(Ôuœ½·™g>5|¹ÂlýkS©žPat6åÈXB¨¹&Ål¬†2u]ª».]¦‘ëÃÞ^µäG¤äPU›‰Zf\ÈÆã½ÂÔrãZ#Ô¹•§nÝ}]<¶îÿ!ÓÍ£y͉W_$gc\1³ñìcWTj Õ¿=P@ÞTØ"Ƚ%¸,"Æòä´ÑчŒ¦ãÔ¶ñÄäõÃÑ£¹³›.œË.´õw­Í[/*þƒm¿©Œë2!næfk§!¤ÑgU]š»v›º¦¶+‰TÔ0ó’λiõ£YÍo]²¿ÅR…©#{cJ̬Q¦ÝÚѦʣ=K6K²'Á$ðS·¡"lj:º5¢»9¬Ì­‰LG¦í½&åá½™ ¶s-ªV3©ÃWÑ·Q†6 ™Ù,€@ˆˆ ×¾D²[ši݉)àj¤ ¬•›£F ‘ÞZ™/&4$¤™² Ñæd&¾æc¿]ºçüLdžd ×çæMthÐ}1¬Ê³ÈËΣ–/—e±—†ÔÀ+Â|ÕÚ_*òC&ï:ö;øªï".’%­<#AŠ\Ù C%n -ŽúKF3û pþÆBë ¿‚å© gÐǨì¡Õ#ÂñC+ȵ³-¹¨&øNµ4 &"4àFi¨ämÄÅ?u©ÒVhSúuxìÏäÔ +Žhøñ•›-ž+f7?ø‰>©Ql[K>uwÝ46ÏÑ¥ùß[—ýV2j²œV³þYzåH¯9Îó9kÆOÊó錖dd³ö'¶¶¸rJ)õ3ØÁKú²¶¼”èô?mëíiü¡—Ipz±¶Å› ¼‚RÝö”›ˆa:qÕÉ:f±œHCèº{ÖI¿–´?™ðg¹&•ø§§-úG-™N[)޲Ø}º«8ÊúlþRaÐDÄ#ÿIè^YÛOõ&«æðOº®@€ #ãt» Mœ#qÛj¤‰©—Æ ó”Λùƒï´õj ‡uÑ2VçÙÔeŸž¯©Ò¹öò²»f7:D—íéFùëWÉsC¸ÆñìÈBdwÁè?©¥ÙŸMuÍß§iuèDŸÎû¾Äö/Tx7ʆꇧEQÔ´J­;-[7²ÿ on^–ïÜvΪŸ?ä†MÜÌ_ŸŒÃ‡pìüh mìzr\<¶$ÞÊA:ï· i:“q/&Ó‘gjYзù›•Ç…ÀñŒ¦àZD|j™Ÿ›Ö²ì=ìïaæÙÜfö5ý<¯ƒÈÖÆþ7Þ{û;ˆ"€² "DDfÔ›;,ôéÂ÷›¯ ­\4Ë’=tc=0;¯È“ôW̃ Œ2Î*,À’áŸFvtïÆÛÔ]Mq9êùËÊJÌ ÜÕêÇS¤õV¯ÎrVzÉ:èžpIUî­HÖdâü*ké¼ÊZÍ5nü½ ™:mî3dæØ…µ ÎEÝIÎOŸ9º=k÷*ÙBšÔ²ô|´\Å+ý¥¶Vç,ÇiÐ]«úL?{؆Žäè0[6–=it9UtUmŠ– –v1³Ò±ˆøP“}=ªÍŸ.ÛÞZ&}…Ù˜ª’qÝ\­¯«­ó˜åH§iÇžœšŒZ©‹¯QiœÇ:Fá”k%øO[?iæ&Qj®½f4‡ýó0CýsÉs:\ŠŽÍBëGjFr6'{·k°ÜB:*k$Tn3Jxë›F£>&é^Z¤ëu›ÆËÂ@£Ô¡i¼;Ütç,úìLC»ÑÿŸ|›q¿åž‡Áœ[mXF:Û©×s.z8sQc«a…½³$ èéÈØÞÉ©aùš¬:³ºOì ¼ùkÄšˆ *Û£w™ƒÌMÜðÛÖ<0Ù¢_Áä¿'ˆª¥ÉrÆÃÚ¶ª?¿)}¶ûÉù¡5“/ú8o.6–ý–B|Xì-OhjãmáFUfÛžÞz]uq²q–y:fB§-JcHá0š¨lÙ7¹«¦¤Yt\Kš‰}4IbN3Órú ^úW–ýî^´õÅ+Õ’´ÎTÍÕ%ƒCho2X£9ˆó”¢eT=µØN?E¦,vÒúqüVÒ±ÿóÕaµÁÙDUT‘5 ¦ßë9ƒ®ÊçZtº«"72žß«‡úì“ð}T¯…gœhßFYß¼ËÎHæk23j=ŸO+U.3â1Qå&[HýÒSXÒâgß_Pé¼¹”‡û3ôÕ>•Û$@R•òá° iœƒ¸¥ÈªE;dä˜5È^Ñ»T,>_×DZôã³Ú"%ªZJ…ÄRûtÎL„Ëb4ºl°¾Á³ˆÂXyóÙO]ŠJ@«ûç“—¢¾«š› …Ü÷õãK`= û÷`؆$´‘–‹.)þ–rËDzœu`qT 9.Äž~”©,È¡ŽvV•u,Ó›ò×~ܺQÛe˲gÖD>*ßïœãSš  çS3£˜DD9ªË:·}s”opÞú6°Õ/ñz=Å.ÖÃþTÛÌ%ƒ¿˜½p¸Úï3ãÓ¤»‚ª-«™¥1ÄÐåSQ'ƒMö µ1ŽLÞì‚ÚP¡‡¾/¥3:@¯ÅÆæÎ#ãÿJ{=´eµ–³ÊDÃem¬š™«Ç-9ôÿ¿™ÚwIÅcþ”ÌP‡ ¹ñÞõÜŽÖdˆ¤èqn¢ë ‘‡/õEG]'&.ïœ'YAÖc` F2ßå9B´>çtÔ£ÖO)ŒÕÚ: N.Ù¨ànâò'~&TÖa#c †ê‘1hÊg—~(dÍÎn÷›[íþû—ˆÁƒg ‰N܇¶ÍŒfè§ÁEC”«‰vU¹|?#¼ŽéKL%òV©m µ©ÉkB;2ÕÓš‹9øßC“~e]lv 'ë/žËæ¨&pÒ–éq¹ôeÁ+•sRÎbã£Jk °Î†CwØÆcÖ箽¶OìFjëAA¿±"oÏÜÝü¸ùÿs÷þ€ !ZJ `È€‡|èNÊ š•eˆq@\¥gÙ˜•x˜j›z‰yúë’£~8"½;ÿ¶I(ÉÀjOˆÇ-üeNñm­eÒ$ZS=ТĢØÄ ,䲈¿Ý•ßf**)ÂNƒÍnò"î.Vq²f}éÁå¹Ì‚ aeü “ãê†P´»©;ôlJÌh¬ö½!ªŒñº¬.¬ùõ[?·îü¿GÀY™Hï—æï¸_­lOQÄóøÜ6UÒÈ ÂÅwÿ@‰"œÔTùñûPQäy:„‡3ÃÐ òœïSQì}qI‚ˆSâÁ‡“ø5±ZyÍëjümõª¢¢ž¿Æ:õW z¸ucçÅÆîi؉úµªý(=$xÿSßÔ4Û<ýA[È €¦e|bÔ‹.ϳ- ãúòú Éÿsç9Rê*ŠrÝ6ž¤º×ÔNfèÑç;­ž€óvÛ}Í`Ô]_;üøzÕØ€Ñ—K·ðäÁ¢ÿ#ìÖ%kwðÀ,t)0Y·¨AƒÄ¥ÜæšSXS&0Üi Óv¬Öo–Ïû^<5h>½dçtÓM¶šD#£ýyú?±Ëº²³ºEßò©!R^_ð 9øÞbjBKó`£%õ޼HPœìü£œËÕWí÷5jG ÐÃ:eà»»rÝßäOðJ2§éýGü]¥RGYøy68M*¾Kì¿zMz®?rcãi(ÒËá;ÖOí~’‹°Ó½¤‰w×x“&T‚Éõ$C<' àè”IÑR!èîë¼ÔLÈ?Háiã}ñ°ÿáÿ2ú¿²³ÁKe¨üˆô¿óJœ<ª>Žu‘A%-•b U-¶e¢ c³¿õ.g™®á_öjE~÷ˆò s¼ÎemâŒn‡F^–ÝËàv9oêBRpø* 7—_bª+‹P ÅÂFé3ìÎV¾mL˜JDRwµÈEÿòÓMšµä_¹}&,»­¨tñÓ 6öãy MÜÖÁ&wÔ£¿±ÊÁmu³Š •]~Z.¢§ñÜüyóß ŸWy`U Sõ>åãò>OÛ@ÀlõÝ•1üÓ_Ë#=›€’˜ˆØ«¼}“´Ñ-^†4ß[8 ¹ èñÍ¢ûM‡ënü¿í^§]Àéflò ×÷Ë׿C­—yÛšØÿ‡ `nêQÁüiNpé^w9-¥_øñuŽÒñpfTøê¼J *fÅ’Ê 0m_j%e2Z°À¬RHÈR« à@£±×Œ÷MÈxÑiú~ÃÙÚƒE•žÊç2w¬/ÅÙZOºò¼†]ÙBu%:OÝKêe;*9 C¨r†ÐÃî%ö°ÎÓ å ž ×Ü–r?ËÂçÓw•K¬wÔÏže7iO­K® õöê·äc}‘äl)&ù¢—oðíÉî9wŽß)O¦·ßF(9œÝ÷´^\ñý¤a¤£8pä¸ö.W¡^o™ãRºî²øûY¯Çæi^ó““§}†~” ;‡sÅP`qÝÝK—rÏN¾Û‰Ö‘¹Ÿ'ÝÓ]~fG6BÏkóYTa‰Ï,”çp0Õ*iäžbÛý-ç9ÌVç²²¤t•¹­ mõ9[Œ1õU¨·= ýCÈ}#×λ}®áͼµØÉ“O ÔíHäLÜ<çFZ8f<½$þ&ê=è}=&.àw÷@}¾8ú•§/+ËObàô#œxú™òc¤¥uÊãe²ÏÑb{mk_Kj[»Ö;°ÉéRÙw‡r½'eHg9e³uð=]”ßÎ]#6×,]½ÎVíq¼¦ÕÈÉÈìåè¸+xYH’†vSÐXaãv[·êñL^ )=|µþJiäÞëLƒßòëÎÝ«¨·r&1Ÿhïv¥o½BÂ<‡ÈÈŠWsA^¿^VEwÆgK%õ·¬Ý¼Ïž§¯Ýk¿ÍïìyÛï[Ç×裻Ÿ³,«<ª^Ï{îVÍ-ó¹*Ý’œŠ§òcÅ‹Õå̧øÜ®Ÿ™¬u}Æ›­Ö_s?ÁW‡3œôjI¼o~êüZÞ|ìKti{ÙðiÚñ{Xâíùq{¸ëõ¸¼ºÏÝRÉj®W^-¦›Æç[‘®ÄV¹ ¹%u²h³Í‰Ï5'\åjo¨x5O¥­¼ß&ú ódø»>ü=ù™§Ñ§{‘[»–Âç΋rQl24[ßgÓÃ3ÍøºO›N Oã5”Ö6îç§°â£<)4Þ¼»#Ö»iÏ^£Œ¬‹ÍÙî5èãXħ3º(óŸƒËãÞ¯evÌäŒÎ§ ËÛÝÏy°×Rë\×Éÿ‹©ÏÚáê{bT èæJ´ÿ5eË• Åo˜ÊõB’l5Oo'%WwúÍt¶YäÖí¾·²o2;‡?³<{çnNŒ´uÞ•¨÷UˆÜ+ýù./¢¥+ï¸ÜÔP{‹Š2ê&^<òaPmz\;i}SÈån°›6̸Õ/kÍ]S¥µKŒ] 4cÑ*ð{º¬àðNmã«n{¡]þf÷ž¥¨ô]õùßmüuío_{“‡}1Ô;„iMld,§g%Ûè[”–'v‘»â¢š}=6ô‘á·+ž²<‘v•ÆY¬ƒH§pO†B'ÒPò:á<ägusZ÷ÚÍð/Óh´×Ž jçž|íOf|ïoÖ¹í.öµ¯±¦Òj<*߈jïQxÜíLÉÒÈœá[iç\Ëz÷PßèÅ"_EìM ÚèëmÒÙ.Ïc}vA¥+¶—½Çw9c‰ƒöôÝ-TŠØ*¿k›Ó¤‰õ¹½ç¿FÐ׊VH“»èê·Y¢•·K[¼ŒÎ2ÇÚ»Íø0óù³mzžéwùœhØd Éß󏨰$ŠUô÷8yä?v7§2¹Vâ*õ"ÇÞÜ£-Ž×Z¼»½®¾~Ñ…¼âÇ–ÆBÑÒo2Wá]rVO8¼gmËF±äù%ósÜwñ¥¢× ×‹Ãj^O±eç£B o:ØÎàŽ×Ã˪Ë'tm'ftc4¬¼3Êï5/—éÝk‘ßs4S&Þ¹Þôü *)ƒû§§Ú[YÕUNÇQ^ÊÅo‡¸ù§"º\Çõ¶j%R§:ŸE¿ØÓ¤DÛ9î¥|RzVnöÌvàe·Þ˜–$2– »[ªñôùl»›Í•·Cƒ;™ÌÚi«æHãvóä‘Õíp¥ÕÞ{1;5cšTnïºÐÔó È0h·èTjnù\ákÁ¹ÖíyõüÑH½“F0ø_#ÆŽYµR·o_s¹ê4_¼íDök‰øÄyþì|]:AÉpñv0ã+ªC÷;œH¼¬ÖåV»%ŽÎ·½ ó×]š h.*šÆJÖÙ×k[¾½Ýv‚Bú.(Æÿ‡‡oU¡¾ ¥]5 ̉§koOWÝô¯»eÚÊǃGõ{{…ý}+ºžL{Yýþ·~V{³ÈĺýN” ¥is¸4Tï ÂK¦Ó}D|IÁñÀëgUgÉçÜâ°ö†Þž¤‡½0ywç:©ç;¦¯n›ŸªþÒ˜>‹”SzTIq%­^-i/f>ËW2ؤüwø³Òáäg¼°~Ÿ¦Ï‹´ã¹+óé.ñ÷2<Òœ^B‘—ççvi7¾ÄS:w±äŒêݽåE,Ý=çms˜aŽC¹»ÜÿsÜcŽ*f;г!Tðü*wZ_.Úòzú ÷Dq‘×»ò$j{ˆ ±zG[’¤”€tM½Îêîx=4šßÝ^ÿE1®h]aC'mÎyuÁª?V‡¿'ƒ9fpïŸÐ>–,x(¶¯±Ýìmë^VSçqyZÝ>š‘Z]…ïDGakŸá}Os•a}H¡.HZ3 .bF§Rá̶wÚÕÎãIÊÎÛ“A&>;#¾²}3-dú4íú>!à±¶âÆç¹ )AIžžÒªyü±´âfñ‚õÈWù[îþ¾Ý*þ»Ú#\¢ª}ZAîøy¼yjW›âEÞN ƒ#›:ç£Óõ÷ZÌåhiTº¼|ôi¿Ñ𴣨j—79f«.:y …Œèpß½R9g—=óŒEèî—„ñ žû'_ÖØÞ½«‡½ µÖÇáï©BÐIHäNu±§ŠÎfæÂAËéó^:zzùñ8ð=¨+ÓMΧ|Kòw=¹Ã®!`<]Ú¤6ßq¬Ï‘LkváÃîßÕ{ü¼5öãù•á^§!AÜ|++¤¬ÇpŠ?Éuö<ƒu9\­–X¥(ãŽimN}Ũ•é'Ä¿Ãæu=v«Wæëp±ßòu·v¤_åC¼ô9xÆr<~ g˜n¿'y®äpß.yOXŽomM6¯^'¥—' Ã÷Ò¶ûÅ'ÇñtÞúCèõßdoøt˜¢2Ÿcð¼õT¿H¦«—ÝzúªmmS4ØB:{œu?›àÕgÏÙÙ{«%Yfaöå…¥áøi|wØo“¸Üo ú¹ÝêŽdìVî[]l¯+ÉaßPFf‹5úùËÜY ¾[Ý¥–0j×ÙÔ˜¸|¾Šm{ùH½~\M ØNÜKfêéÌ&RÞ… ßåÊÔ\Ûô)X Zʵô–Ûä!þÏAL(Ðb¿£ƒ£ábdGoŒÕI¯Cë&zád5ò︣î‹l(…×òSøiîEÝo}<üú0ÕK\L’L$íx6h®žfëÍúëxTÌ׳›ïûý¦WÂËxëra'Ûèrz}×½¤Æ“ÃÖ¡ÈB‡…–¸ÇàÆ9xðã7“ÕcOâèqZ^dYÎÕ… á1Ùä­W–‘¥^5ßC]Sj…~ê¶–øx$Xm§º÷³+ÔÇÆãÐ2æ7,åcop)iŸ©qÞêX+x[Pú*W-øUǘÌîæ}ø=îG™™2'ÞÓÐ[/¦¬#·ÑÕÖ¨¿nLm¤>ÓÚèùÄ @—ýokzÖãïãkðjD^¦Lm§†Ú/쿃WøÏÄp导nvA]®×we®/,Lv´^Ò.3i‡ÉÝ’x<ãây¿ÏÇáçüù_ïÎæGC¸û³"5¢7ÙôFKÓ±›ÃËïò#UñçµAîdµœ¨dD@d=„ß_2ÅŒPÄø¥ŠÒÃÆüK€,,†+)‰ÕYð†¿Àú:Ñê—Z]o£Œò>¯[•?+L2zpˆˆTRVsY¬æo¬Í[ׇCâê?-¥÷+„tƒk;¹Þ÷´±Ìj²/ÞÕoÏÏÿrù|¿õØðp½Ñåox{!ön 3™îû³wfóy½I RÎÀj3Þ…ëÛþBü;³`ïŸþM‹ùݦ¿ñ’ü:GÊ?]ùû¿ðŽõñî®·}ùt{ó§ˆ=ÙÓÑ7];¶:ƒ¨ê:Žðº;ʸGR—EÐ]xÚSÈý¯žoSª6;³Ö÷}×õï~õá¹½ÿFÀ¤vyÝ>w:iîîÂì»»»K¶ï:™×:éÆìÓ½€vØ0SÇ;#^‘_Pê´ujªjû`Ù9àÏqS<åÏxÏ`¶Ó>çÓ>ᇊíݱçÁÜÁ¯À¼oð¼KË™x¶«¬X:Ãû¥ù­ti£ ·ÊJ©² ®¼ÒºWK¥Mxšbø½½q!òð7/‚ø/‚nM8é¾þÃô5¡<3ÃO <åÔ>v¤ÔŽÍ7O—ª¹ÿ¾ +¬ /žkK&¸5θuƸݧÒ¥ô¾‘õGÔ=@=E=@õÔCÔõõáÓý0ú§ê ézÞ¡z—¡z7£zÞ·­ëd|‡Ø; ‚b¶O$MÒ¦ù/—t§ÐÙ3f<² ÕKCËj§˜ùæˆí ¢í µ ¨mháÔrÓj¦a0×ÀpÀLÀ\T<å²»`Û8p6èmÀ¥Ü!¸p;Ü)¸JPÜJ`®˜+‚¦*R»¡Ý4‹TBª•Q*ªUF¨µ@* UJ¢ÕR•k-e+!XÈ5‘+)X² á „„á.%"aˆa‚a‰†8ba€a†¸`ቆ˜èyàâ WV¸%q à5Ñ®!\B¸…tO£+¢W®+i+ªâªR("”‚R-*¢RJ¥(4¨R.T”ƒJ´ªR#H… 4¢R HJ%""´€ÒH…(4‚R¡J*€nÔ1Ð, Ø–, XP°©aFÂ… –,ذ#`RÂK À-€K6 8 %•[*¶P "6Dl¢YE² YA²Y7h ¦"")ˆ ˆ+ˆ€b ‚¸Š2ƒŠ"b¨âˆâ€ª¸ª¢(®*ŠˆnÐq\`Æ `WÆP dPÝ‚¸â®:‹Žãª˜â8!Ž"NIP@WWªÖUòUEE<«ç8ª)ÍPt=}m9ª:¨¨¦Cð—%´è³td£ke)Ѝ‚€b H2 BD  [@'… dAN½lPCA©äA@I?|¸’/ô"/Ó d[Ž"ØR ;^Ç•QEÀŠª#‘W( ˆåÈ¢‹œ ™QEAŠ*£…r9Ê(¢äAQrŽW 9È¢‚ˆ.QUp¨®rሂ&W*ŠgeTr‹•Uȹr"ŠŽ\ˆår¹Ê(är¨£‡&Dr¹UTEr ˆª* áTÂ¢ŠŠ9EË—aUr"Š#…ErˆåX@ŒŒ„ƒ!}å%ªs 0^p÷öô8õPóΞA2âA²f‚¬‰‰ÕmI&¦ 3-dB©šgBåÂØ­’"É$„⦴á)va ‚I„ª$Ö”…ÀçMZEà 18 ز’%hE“"6Ñ0ÈÌ («…bq@æ&¡HuJ°éPYÓŠš0£¨EqV‘’ÉDµQ"¸„dIj¢ ‰*§ .Cf°ˆ»(:%jL¨çDN‘EÁ"1šÓ„†Õ æK5•V¢‰˜ª±9²KK.J…ª©V’T¨ÂBM*å4âZ -«,2àQY$Ijœ”HÑ’œ¢$-#)Í e¨r"¸AF‘€I"£20Öu)™ •&rÎbi³BHª*N—eÂ,P‚:gU•q¡R•V(f %Zd‘š µ³†‚‡[SNe¢´N\¸fÐÔ.Г‰“³K% #ó·sSƒ “Sïâ[*fh ©¹¼¯zo"»È“BØPr #Œ¨â!B4é3‰Òj€œ­šÌevd´àštÍh ´9‘F§JçQ &†jÒl8ApC—b-:Lå@Z©¶\¨å*ÛNØôØÆ‰~95H…ùEQK×È„ ¨" ”påTLá”Qsœ äp ®r¹EŠ¿Ãq,æ!©ÊŠÖ„TDFes•QY'dUÊ4–UPEAGQgB*ª¨’*¢är¨årŠˆ‚*ˆ£œŠ +…DQ]”A\¹Drª8B¡p¢—?Ä;ÓØÆ:ß6~›ëyøŸÝßÔ}ü§ów_äɳÐý{Ø¿Ÿ÷x¦IgOøðyY¹ F+¥OŸ¿ §*"»0MX"¨™—è­b%`¸«¶PÀ»8àÁÇdq°qۆ㠆k.ÚM;\2HMR˜Ë¶\ˆNÁÄ!ÀNÛiÂåÆœ. N]…P'qrË ‰4­A3¶*Ó·Li\qP4™r‹²9ÉŒ¸5™ÄϧrlºpTæƒqÜ8æ4‚É™ hhµZZ\HRÓe†m"Uœ3&TA+qà›/ĈÅ›$Â+T¹@œ¡I­I(Bäâq2ñ1:¢´Hª¸Bc"e“a‹fUdI“Qg;.Gh*Ѐæ­@šxåLñR“¦SàS(H%aBLå8Ùq±”@D¢ H¥H¢‚‡m@2$]÷Ì -#X|þ?àÖÈ%° G?ò¿¢p8Tž —bBq;"`<²0 pp ¬‰""¢[ƒÌ­E7"©ÉáM‚¨  ÓõÿÍøÿ“aÆŠdüÜâ²~Fïò/ujˆ¥ÿÅ¢qsrk»?‹¥ÕcP2 ²èÆIL(îâHÓñDÒÑKAJ °.Ž|!„HÁ$dY:nkNYV’t„Β¬):tlLŠY,ÖµDà’g,¨w»‚qQ ³$íPB¸”rê`œ’N\.$œ¤­,  .5RJèeBv¨M*j´éÐòŠ.XEÄ9UTDRE%̘V‰3XRZ$Ó± "ΜN«Zt„®'OqÜA*èuYa\I¦råÂ9E2Â‚Š Nk¸E@©$C.VZAËq8­b¡Y]JŠe3[N9¬ Zi•I©¡IªÊPÙL Š…D‚é¤\ ¦T;L&б§Seq#DâlâD™Ò¦Q¡dP]š’dªtÎP’v¦ZËE!Ur®3@)” Bb—H•´Î„ Ó„ˆŠŠÂ*˜™Ä“TŠ —I–im3•E4¢"2Š„ Ð-¤S —(°ä@TNÄ®ÚL$XijiV…¨Ð(L’.4âAIE$$2$äI$U'"E”Ø\iQ2†œ¹v•%d \.5iZƪ]–@˜A¬ ’uFÓ¤pI†!r„€»NÄŠÉ”ÊB v\[H´‹¶UM ,åWaQhPÃB"8Y3YvGÈ ÇH); 8(¶«eÙvZ…ÅY¢L. ° ¸‘r€\‹–$'ºu™ÂÎ0Òȉ;’¸Ä¥*ŠÍ¦F•WHI§T)d$š`aˆE6Ê$ÄÕ€¦$h$‹…œ¦‰KéÀº©)(”ÍiË8Tv$Õã(ŠÉL¸]ª2™t‹*ABM2¡$4Ë$å9v„Ñ0iDUÙr"˜….G š´ÎÃS¢%Ì.Ê£¤bM— Ñ !E¤E£Nˤv–Ñ8ÒŽÓ‰4!!"ã ‚Î’t䬈¢˜¨ ‡ ¨‚#0.Qh´T!Tá˜Ë€P•'*Ud…Ä™­…’g.bY••…4 Š6Rr&\»1U+;-0»U”ˉÙ+±dáWp¦Ô@¼£­8H\Ô±Ôàij 4“…„Ë‚TÐ. !’L*‚ˆà ((*®•˶S) #Š… ‚BIÙ¬Tba]9MLˆC¤]”ÚAEQ§j…’EAt©EYPZÑ”RvP«E©2m2¡:E5A2(¬ˆ«‡”“Rƒ ¹Iœ»,0; ”J')”U’E„®sjŠ(©B¬Ô©;4 DŠTI9¬¸P\¦YMf˜Z°¤,Õ8ÍI,–™gBHJ€­(¤%A+–’4¨Ž]¢°¤Õ8G€OŒbQ¤š&r()8ÓˆLѧ#¤¨kHN‡ ¦)dv‚”gjŒ(»  Pœ &P™œ«‚pŽY Ò ²à‡h@\¹¬L‹©‰fh2:LªÍ”…Dr#2¨.Ä‹PH“à¨*R´f"d iˉ2“´!$åË”ÐÕ2,™+d¢Bvœ,“¤$2åÂÙ7\ò„Ò.3@Ö ÊÍg)"‚FˆIEÊ(mÅ9>jèm¨©ÑCaœÈT³3ƒì÷Îúã5Èo×ro±7ìÅifb.@In’A‘6omÔLwæD·ò(pDºw^ꕪx‘ b˜"ºÑ¦÷eª'6p²~PÁë@PL8:#̶ˆbHɦ=½æÑ+1H€Ö À£Ä§ß/)Ê …˜K"þö÷ÌÝg~×áKíSZ„Ù«ŸÔwè¶š•7¾ih¨ä<"S4‚ÅR óëú½ßµwœm—ÙÏ'«WOô:uã' {ÁÊPJÂA¬K­ ÔäÙb]LÈ©œAäÄò3(€vQý¸‹‹â@("aEWôûlK\lBб±DCr â,@² ›1C£ÒŠŠù·³üÌ][„D1"‚†Á@=z *–AE:D·ÀÝ4´"È0´QôPp&]ÊÐ*cdö:Òú뎱òèÖêô6Ù;nd·ÂÏÚª÷¦œpó÷G;?fÌ¢ÙÄ¥šÚUAÞ0AFhënŒLaþ6#TAÉÐÐÈá7–Reü›¹ û]Žæå8x‹s$MöDAd‘’I! ÓÓÓ\ú“A€€*,ÏÔ|UZ0!LHJÄI¬£˜ïii˜¦þîšG5ð-У"ým¤HŒùc÷*«Z¥~Ý÷*"Šh  ]ìåTúÍææñ ²H§?ÀÃCpðeà^<¾GŽqó+#r=#³¾ú–æ½*YÖ¥HÚ “E‘`E  xD&pª u$¾»åÑcv«£>ïÆÙ5fóDñ­ƒ>f%_­ÝÞ3عÑÙÓöv¶Ý<êJSéÎ$6åžeÖP‹X†ó{oJ„¢+¸L¼ÿ½v#c€%Ò6& ´-F” !Ä´–„™¯ÓGMÇ ( NVÙÖ×,å Û¡ŸJgeáeáÌíËwF”5ž@Ý¥èfT¼Ž,ýç´<¿ó[Êe‘ŸÿlïÈ¥¥µšy¼,Z®9ö£ï–é·sX"FF˜““0fIÉç .i!ƒÁ‹çÙQ8 W®XVúÎrFõõûÑ–½1N-Õ“‘@¬¯ÜN&mka!å_€PìˆqÌмþ§ý¹˜œ|,4A%ý(ZÂH’ªslö$ë³Ñý×RûAf³ú5Í4ÕîúÎ[»'ºóÄ~ÝGœ¡¡äîtqX'Ú(`@Ýío"o ø>ÒäåÚrü+ÛŸ+î&A‚ãmTµýMÌ1ìïóÐŽû``,†™€º&LµþðÛè§>|sžŽÌàhéYè`¤‡Bž6V½LÛ«4¥’D$0Â"I!ºI"HÈFŽëý€Ã¬—)_ ð(C‘|’5¬ ºéïußBñ‡œ5³ý_ø¶,îó¯›Xùî ;%aä~¦U·c{ÿ´§Á¿þŸða¨.–ÆöBDŒ!§bQâê¢q¤ƒ""k 52¼˜hôóˆžCÄWn×\ÜX26{—D+Bd³’FBBJ9DÊÉN—‹u¸1;\l3)èô°XɃôˆµ­vù”fAhÚŽ¯@âœ=9³l÷o‘shóìl‹êŽ9ža£ ‚"1–ž¿×qÓ ‘<žQ°³Î€¤·™²K•Êû­tA™ƒ2327€rÄ’WØ”ì¯ÜÀãæ•¿AD±˜‰„ÂÐbLù F¾UX®NÓíNÛh§ @ñýÅï£aÊ …!¾ÏÑtú÷>ÿØÛêøA"$ƒ$„! # !í#‘A5¹Mö ëàpL {?g7èÝ8§¯ËKÕ­ÌÚHz«6æ¹ØšN(CæTm‚ÊúêÄ8D„’B2wðˆ6OÊ„Qí¿wÜÓRPƒD›“¥Ù°*GجÆ_3]mƒÈ¯n§½¸vÖÜÇ?}=Ÿ}UtÜÿ}5T5mß¡gRñ±*ý¾e·Ñ6¯÷–Íל™[B2Î63 ‚Á×&nn–6îd>Aäl´Ëæ»NA ³Kõ¬¢b˜ã^D¹jW‘±È1ñÜsF&gF½ cß)€Ã7Ù=pP«FìýH@êëHÈHiÎÍ-¶„ŒÖè»åô’ȶģXLrÃø)ÀóGÛóÊž¥õNìî /¢½jzîþÜÊ(û4¸Ã·èã8êÒ°Ö×mÈR÷{÷3(ɽ€Ð@`Á¿P“!LÜä D‘adU‘Q. .ʹW(}:bp" ’HÂ@a#$“˾{¼]÷ÜFàëèw¶ô]”|†ò.QDEÊç+•!$!BAà`çl]zH{þ6*jú­/cö?c„—øZUÄÉ#2”2´tzy¦_¿Èôz¸Þ‘Àä¾9¯†j RÞ’ŒÿÓMFÆêQõI%Ð ˜DÓå$^z¤$ÌÌBf@øä-˜åLØ÷kµìHÃCAªÅ9Ô÷×dÓfòSˆûP¿’†T&p7 `”ßA¤#‡±Z‘ <®²ð[£e&·ÍÆÿ_ùmͼ´€e ¢Ä‚@æŒmŸ`'0ýÖúç ÅÙ§Zµ¢+E›äM(¥gˆ¤§‰wU„EDª”R\ªŠÇ:Ô“g‘ aB .B³¥95‰&ó#¿]åD\ ª—Ãåè‡ê’Y3œXò¸¹0ÆÁm¥÷ëÏÊmð½6xŒÓ&‘…Lœ2âéuöÃD³-•$…W0"ê|=Éüœ?³†½wöó‡¾w„†£úTo>­Ž=‹µ}­ÙXé",ˆˆ¤FÊøf@¥ŒÂˆ‡ˆx†ÐÚX˹mÂ²Ø 1kõÔ¶#F×'ãPa`d0¿Ë—¬Å%Ì„79Û8¶ŸWOÉʰ $I1À5 H554Û›¾·C¸ð,´’$4ЄŠ{ÇÉäEE@\»á»|iUDk™‘ÈT ¤ˆ\ñ«’óõª ÒQ©%ÆHSÌKÌòn³?û‹ÔªÀ{¼ïàË¥ºÇ»ÆËö¹Â‹¤;¢kšðdeˆ…ó•ÁBëÄ}ßïÒ³×Ò‚oð âÖȵ­«ÛvÜÍ»înõ0SªBBjfæææ‹uχÆÈ¨Yãq²Ï¬åã*—^„—éD‘<Ÿ{G×E,š™üO‘m©#!"’HÂH¾üP3ùú€Ø"}Ïìþß2ÙJ}ü-ÙøŸ§ø/Íéõ}^¯Áü.ãôk8Þjîù4;è6£¥mI:° A–ÐØè{Í¿õ¸˜ èBs!IŘÌ=·–ÄÇ”’22„ˆNœ8ð­øÜ™]¾Z!áü>–6 ׂ´§« H‚­¦†^K C?£Ežv_äé4^&êáÚ”ôNÖǬtz;=.©˜‡˜dV{½®÷z¥Õb4l3ø½m{ËüS£aiÖÚ¢Õsm‡^ìYcò] ¹ eñw;a;¤ì@ço,¦ÎÕËÞPõ½ÍÒ—Ó•CƒæPÙ"’HHI&pÕp^ \} »Á ñeéÇ·éúnßo{m¾oë:çKóÐ>Ê>9|´hQTõ`Ea €XlÂHÞÿÔëÏ…Ïé%Ù»­ªó@‰ôlX@Î¥ŠT+7ÖjÝ`ÚG&#ÁØÝYgO‡ˆµ,ƒíéB]í¤Ò)‰^s¾ç«’À}÷{Å .Ät¼œßŠmÑáüVÛñ|­J² z”„‘†¦]C*ÊB,ï;Á«µm É ‹·äΧ‡‚â𢔋¯‡òŠ…I‘ ÅO¦¥¢–[Ý(6¡3•ÄÖâýµcˆµ. ñ¸¹9‹~ë²:¤çàk8ª} åž{"GŒþ6ÏØwM×Åøž_­½P²Ú4 ø¡@“>¿à=ÇÕuÙW*ù5çð$sô¾…Ç{:ÈáEv”ìD’œþÁÐ"ÞÊãà¬@ÇáÐñ"9œ÷çÙ$t©ÏÊ@0Ú© HK5)P'Êǫ͛~B7T„!#4¨F;ž«­ÌÞæâÕÅ‹Üvû²ÄÀI$Y#$˜ØÉÐÔu»8TP;ãÒ¸ö¼÷$|Vï ±°[õoÐZ]%#"}[sÔÐó$)(ÌYš¥"¡!™•kH,°œHà­% }¶¾±!G¤^sœÅêyÀåaTÿ¯-ÂIoG)eð¸ÃDç¥EµK$ã×r8Õ‘‡Z²ÉÜÜžþó±Û³Àx1tÒ K»È¡‘&)zJ¤”i˜¥%… HNÆgƒXPADR ÖB/BðÏ.…@žŠCOŠaÏSGuš—±[õv) öÞ¼¨†ŒúF>j½ô,ºý˜wŸì[/˜¸ÉÓˆh”q§ÅI¤× \Æ<Ñ×\‹)‰¿¶Ö4éM¿¨×iö².SndYkFäÆÝ†N‘°×ÿ—AømÑ™*;BKVî,ÆLÉ-ÏMî™úxä^z²øô£¢;ñögÏ‚¥z´ªßK®Ã=#¤©è‘nc1£ëýBLLA4AWjÜR ùÒ¶¤ßf®+÷.äJÜ‹\X ‹ãétÆÄÏ(ª~aÐ- Zq7£§¯wIÂá+YZê©kœðÑÖ £ƒÆJ_8ٖΉ¬¬•$ÑÈk¦)H’½êêoO”J &0Œª øWâ< ÍÐh ]=-y)ÖÉÊêVH«ør6‰øî¸ôúr2è%n´x!ÏŸªeùúžmÂ!d£"NÖ¢ q <˜…#$f/6•H’!%Ú>ß>ÆÙ"Â)I"Å’2$“Ø!Ñòüƒ¹[nCœèW·Ä Û w/)àtôíVèÂ,‚›—Zƒ­ÛÑ­ê tåfîFF5|œ”3 Å-ƼâÔ§´ÖuÕ(œšé-þšPJWøÝÒ·T›‹ë«£-·ö¾§«‘Óâ"9XjèS³ò7U[ ½Ô£Y6èÒ8x½ÿ#ƒ…r÷Ü&‘ddÀw°8ÀÛ7)tHĤ^,¬w\¾ÛA¿>Ù±n‹J1zSÆÏïj'¯‰~EÐÓ¢a„ìB~È: [toJ@15s7ÈÊ.ŽS¥ãûþ×¹­i™É|ßí¢­ÀjòèÐËÌï‹Eá±ÕÕ)tèlÕK è‘p¢œXV ĉéA+;Fñv­Ù&~±Â¨øÖJ¯ëæß±Ëëƒ,Éóî“å> îÜèêÁ¼×@‡ø€Êáåþ•ðð¹Ô¯Ûhµ\ö4Ñ c! …F›3ùå¼*ÿ‰IÐöx@ÎK¥;ŒQ½N1¤ì5;Œu¥+Á²ˆyhÈê¶)€rÀ’I%ƒìu uh‡­ÈãwÏ|8áf$ʼnŒÐÏ'·ºÅèRŒ„Œ- ^W)± ¹tßËÊÀ¬CRƒ!… DŒEÝäóì±]ñ¤»¶§Àð¾/›…¯£xp£'ÝÁ)Iúø”=Vpº7”çåÐK’.‰~ÌÅÊ7VªŒÐÆ%aœ.Õo6v<-Ë‘G{ª¦ ¾òQl€!ƒ †óVµ>|ÙÈ€™{Sfý‚øR“Z É$ì5Êç+œ»(#‘ª‚™Q¯ÄϺì_*µ€Hƒíj‡>ÌÈRžt SÑ×—Ë“œª"¨™ÜÜTKDÎ.Ëé;F8¿1ñ¾ŸÊpϽîÝÎ=«8$~ºR2I1Ó@Ü4 AKÆlS_@ dA+ PýOšSӯ–YZ”­³øó{NÂÍT¥!#çjõëi6Ûgï ð»»ú ŽKÓw¼Íþó£;Ð$¤Èæl“;ÎÝcØØÞsÛ&~1ª€àœ%. ÑSÌóè c „ õü^gm¬!,¢‹&#Õð€ä8ØÊ—^É P‚ò܇A‘P´Kˆ¥:¦‚ÁR†sÐ콿™WI”ân:»ÿKÎåU£:&L(ÖmŤÁ7Õ(x·›ßþ´´oJ£q(~ ´4àe-€y»BÄj™?äbÚý˜¶^@Æ2"TÏ]sqˆJ©íGn¹…ÊñøþHúÒ™Þñ\qãÁLˆ¯RuôºèºN<W*ç9Nº…¹^p¸-Ô•ŸTЙ(ÉÎÉEŒ§³¥ÑyÍ©c8÷ve¦†ŸÑM/ܰ`ÂNeS@ú:5*ÏA­òç¹ê[Í…ÒA§‡î×öuú8—êiøHÅæ[àÜžÛŸ"û7òïcÇmïŸÌ¼vÀî,¨ŠbYÏ FõÛÊvM(óæÀ{d‡çê|òðÖI²&[HNÌ1‘ñ"ïZ 0‡Ü· ;ã÷~Þ:¿þ §Èö°pÃS“ ×Y*…H^ÿ6µ€w’ÔÈ€câR‰««A„;ÙFÚîüzžæ-Ä›T…Èy3p5…½"ͤ1±ë;<†#ÈtÃ_¼@p­#±ÎždFË$„ƒàý/kÙ»nìhUìÑ¢à‰Hœ-Mo[„³C“!'…Û†¹ Èõ4Ò£XÈ^BßÐòsç¶ÓBo^ä§v)ï6MßAŸHÄö7-Ç«Åꊰ½¯½i ,=µà:&ùöL޶ÕH’@òqr´ß.™w/gGæo)ŽoËáß2›šœêC¼ÒWÉÜ Í@HaúþØÿ»Œ³-ß,d’öÑêߩҨq©!Z÷ ÎtµWr^Gg”ö¸¢¸ÇÝ\z{Uô7´apmÙ©bíÝp*yG¨s8ºšâ"ZHVâõ´îÓZ3»7”X"V}ÿ=qú¨XU7©|©–àøÜÇ0€“èwãÆº ðÞ@. úO?Ÿ‹wõ(<àþwNÊ›ÕÓaE\È 5ÌNOÇi2&V –M~±ÄŽIÉïprÚ/ÍA¬Ðst<‰µ‚®¯—¸ó< ¢ÃÓœ 3Éõ¹$uvhí†å÷÷Èv”˜i¹‹á€Ÿ[2c~/ºFðB@ÐÅñl°9tqêaG¼àZ¹°"K ûÖG=tªå ånt8Vh[¬™X£µìâ³M…øOTúWÑ›C—u@4ø0h;hä,±©ƒ¡Ô©?> ¼nbçñ{WwqL²f]$Üʰ°4ü) ‡zjG«þ$D{uëgz(÷Ú8ÜrNà¸"ìFmQÏÃúÿ>€A];Ÿ[£ÀŠ ÈɪôwÔ‘%–Gâ‘ÖD‘\÷}·Ã­Ð¼ôÁb ÌÏ’‡[¦}\„¦3y3ÔH©ï¤vöÑØf;«ÈötT£]rŽ©þP1Ók‹Óæq.¥¬”í¸ĵ 2'µŸµ¬7‘R²Â9§ Œá÷9˜câÁÊTÓQw a‡…NÞH`KòSüS N8 Apνf†-í,gÔUGìäú,0M™6ˆŒ$šA mPéÚ[]ŽJH8mõHPR@¸Ôkõç½Ëó´èt̬sŸâa‚6’JI=Ôzñ!J"™Dä:Õhµž×Mèyœ|Wá¬8Évâ¯7¯W ó;¼ÞÅ”)^I]~‘¬ëí«uÊâÀµÞÂÿÒ,(h¦UN Û;³êÓ×ÓÛü_‘Ä­Y®LäŽ?ÁîüÅZ¶edPÖzKÇä f ‡rMJ.6„jü´7”.²êF§ÀÙiýΖ ÂòIˆf@â+š‚ ˜9û½LªgéRé‹q=Þ'‹4@™ñ³¸<'«Ú‰žm­AOåà כּµck•Üò÷Þ'.{¬4p¢ÂÒ…a!“JIK"s§(v3ú¬"Ô,ˆ€Š,PBÄÀ6ë¥åßåolH¸+ƒ±,à‡ “§$ {mGEudA¹"KÂôú=­UAD ûÙãrzO7gá÷ëâo±¶@ l„.§`’å¸LÉñ✽DT¢PÒõnÌØi<ëd¹^BlÞ¿¯ ›!C½çÏg—ÖÝmraxÀ¥+¹×M>ú¾*à/^ȤDʼnð6Ü߃SÔ·_ÛØR>&—9’2 \¢ŠH7ÄÏd¶“½6¨"×8.åÉn¦Ë_#CpµÃŒ«DϺÖr<-¿;Û³ Ÿ¥¿ÒoóÔ¨¤ˆ ¤,Çw&*éqúß5kü‰&œq”zä;–¼~ 1¢ÌãÓ¥-¨2…š8ø¹íž•~p»œð·xÊÅeöÐü…Ýô^w;Åz« ³}Ñ™´÷:ÿ+½ƒ AÑÜ;¦ ¦(”A)ÓI8,é^WÂÑn2]}Éb ¢™"ËÄÕyþ.*²2ü·}Éu mH|Î 3HJRƒ3Ók™é»àϹ¨àh ¼~¿“?f ¹¾m„œHH0”„eÈ?$¬´÷9_öúÍ›Ž¤%ÐnçÔÚu2˜f`B ôR¤‰šœ/¶îÖ_v>ÓÍF“‹yªÞé¡–ŒÀA´„'ž’Г‹H*Ìàwÿ“éh"œ.î{8 FßÑÚÆ»£ÙzoÈBÂçN€(¢HN“„ÎÙ±µ1J3„þÏs9Äñ©f¨dPÚëy^,z{es]pNI×Cê¸xótžYÚ§#Èì³$è>Ëì¾Óò¾—ß÷}ý<MX]ãäs—Kóúðó ¤X”œsNÏ4u}Ø °! ‘V.éœfo'›«µ§ë;é¿Öåã‹ ŠP÷–ŒúïSŸh4FàYìMiõtÒ(Zµ@"Ä–.™ý^‡œ:ñ5©dî䜳ÖËãýU½óuÄ+I@–÷nv_ËÉç@p»ÈÅ"–G‹è¾|Ì«$’EAoÄwø¸ùi ÝúÒ÷%É–¼ó÷Þ³y!PO¦|göLš A+bôìõ¦ ¡³ú·qá>Ù~óGÀ@¶ÿê»ePJ‚¸D5 G_Ø–rAcÙÓ õLHÞüÛwPàBd×k3ÛîÈ``D„žR==‰ ƒ"»î!ë³/w,9ך)*ÑC!Äm¸ZÔ5N<íbx@ÉMÇ׿8ŠH¢M\™äÃ%ÖX&¼ÈŸ™–>yøÀTOáä®™¦ÛA3móþoÐ{Þ¼½Åß6⻎QOmÔî¸ËÙÙÛñ{´¥j€/8Zɨ,7ÛmD ß"á^òüŽãe ž¿`ý¢(H$P q¡·!ËAhp†‚¾È©öp¼3e× óÈN†rþR@¡`N0X9³7Žˆ!›èñzñžðôö†$^Y²ºÚɆ•‘Î#ÔŒ(îÑÿz›Ð =ªIˆ¿¹¯âÖG9&±)f\š ǯôsyã¥Tü?›ì8ƒñˆ{2Þ“Å`ù‰ÈˆUžâ]Þ”÷'—S[;ɰL7±<,KNœ3%Ä{y.áÂÓÜS¹Ô²î­ªkàò*wrñ¶‡jUô;5|Y·î.i— Üj¿_GÝL«(‹ßdS§-…‘(C•1ó¼?ÛÞcÓéâEÛëÿ/Š=C¾ëß2Ù=J*$FÐìµT¾‚¥‚Þ8¸‡Ý¦î凾C+—ÅÃnd=o 8ú„ù×uìvæ÷<€uð¶8ìúTï`ÎtÄ0@sÄ.\½‹—ÃT$ñvŸXàhæu·D®¥‘""Z ­ø $€ËˆÈ¬‚¨È*„"ª’  æót")R ƒBPò²†¤„ì»)… ²"ˆˆ¨Ø&Ówå ßTÉtí||8ÙM»Ï¦ÇºúoĹ­‚Û‘SßFˆŠH’Ôòô!RGï”LÜ ƒú -¬W)h€B>(eò<ÆñZ·ÂØÛŸÙZøX2šfE®Ý‘nÿ1 »£yðo¢"@†ÈÝAã@²vXŠÉyccLÔüè߈¦((jÄMˆf0V‘pAßàâo{ÜvPÊeA¸ÈùJïåÞ/ÁD k•@,€2„Š)" Œˆ’H‹Œq<ÜŒììä È%ü,½»Ȫ}Ü)N7Êh)X Š š0þœr0¶ù;»oÂDBAÄ$EÌ€,a^û_×_–O(4YÉm!®žýhD RÚÎ|_Q`ˆXæAM( gdBDd :'SO`±dBEI’MH¨R&èïx8 ›ýúoÒð}ö1wž‡öýáú×&DVB@n ÐäQ‘D1Äʦá4¿ …Ñ,€<¼šÜ lO™e’$€H&6 `Hâ„‘ OÀ!$ƒ"¤¤ÓÈtôôÝ1nQûÝlšŠó"¸ss`?3í¹Ÿ…ês‹‹X¥½Ý>ߊ’! \ï4%H2(oðß gær½ÖL.RÒ×EÖo~ßÞ9:8gþpþæêbÿWÌäèb*†db†|=T„dÕ”„ Æ€":w¤ø›[[J^ Dç@L({8*_’È,ˆù‘)wG?¶6w{÷ðNÑõº¸Üû1FÎ<Èƃ  BJ2¢H&ÄPÌŠYG.ø[ø-l1à'æx´0¢@üN>¾­Žè¥â¡«¤Swó&qÀ±"þ¥Q©:cBQšñŠT$’?;6tdX"$ˆ 3 T ÈDâ– ø|cž›kpób9Ùôà.ƒöÝ·£hq`nöqtëbÛX!‚•Ç)B…¶ Xg9¹®j–ó×Aû~5üªcPTSG¯Î-åÈ€]‘1CHV@ˆ‡h9‡š”Ïxw9}ë¯îF#3òÿWäf™¨»¸/]êÏàv¶¤/‰µ c[àK"±¼«‹`u¦î‚Ù€;Ž-‹gb€RÁŠMÌ‘ÉÊʾâÁʘX(‡söm[`7'Õ÷? nCóíOÄG-ž[©Úιոˆ 0&̈[ ÎŒ×á^áÃxƒ®m`\{è7Àû]W£Ýun=œA 7¬[ÂæƯpk¸ß;O°­x¸Pš Ú¼ªwk ›û÷V㬩O©Æ>$¶» »¾’ñ¤:ßvžq>ƒz ÉÍGç@‘!·’Üì8댳+Q*œ†ƒÐ1È-K´¤d´ˆa”¸WQv—œéÉ£ããaðc–JÝTlƒáOôímVb¯ fBÛu½9ÖÖ¦lԘšJL\Y ÒÇHlÍeˆYÇ+dÕ„*Ìo3û¨ï¥®=eËý^keÕòhhñQÄˉ7;ýÇqÕLyÆ@£õ$ È´ÂèhÐÍnêw˜èê_É´Çæè÷4 “bOx–>°ôµ./Ûú^ßî¿FÕTÐЂøGk}h¦øëoÃ~§ nÖ%pNkÀà»@ ¢œn‡Nó„V(cÊÒʸ­˜ÝîÞEÃç\aJ蚎~xg€]s—) ©•ì€E;«3ºæ–GN× Ž‡B•’sjW}ó•÷?,¥M§÷du¦Æj޳9O~C}†Ñ×·¾†ý—H®[sí€ ˆd'Æ&ÿ.=‡/WªYk_¥9 íàb€r@r.ÈdG”$d$nDò1|›nÀª*)…È3.¸òxþê¶HH^/G§{.nŸjÀ¿ñ]¸t‹”Òk縎ržóÌyÁÉö„²y<ÝéÃuÒ ˆïu|upU‚"®WeWí^[˜DG €¦p)‡áKiM—Q=¿“á8 áU¹Ê" ‘püJ aFFD Ò(ãÕþ&ÆZ}HQ4T䪱ôƵ°´i^jb‘î¿Ö4¸cŸ½x*àDðdf2Öü~k°†çjhê ®œ¦êH®ýñ šœ/^ PPG›Ý²Š8WÀ-8;Àícw=…ùxœ#”~ Ç´EûËåDQq‘N}†>C¬J‚>Â!ò% ’$RF<µaÇËÊL¤ƒ°®÷nPüÊ죟Gº=o»ù?q×ØÝ·£¶=ÂÛÝÛ;hmUça»°Ë0°$’KHl‰w‰`øq ÊÑ»rR÷¥çH ›Ñ¦=—á;¸ýØì™CÑ!8_JÊ fg‚@!¨ •KþRœº{T{hå6ÃÞ†õǹúö9t¶¾óD2^¿óË8g ø ¢—´WÏ©XovhVC—JL%)ÆÚ£nÀRæ`åýO£Í¹Ê—½|¡!# “(uáÔ²‘„ž/ŸØ¨Øhálr•!o3ÓváÖvøŸV³‘JtO³ê;÷~%ð0G 4$„¹ôï~e?v–³æ%Œç7q¿þ‹ÙsöºwÛ>–Yý~×MøíúÇHˆ|˜!n¡–%®ÆÈžñ6\¾¬ílwwAíS]Ѩ°–Ð` ð•p+‚er¼d†â”” tã_¼ÐÔº²`Ë‚àÍäbاßÐ=ƒò}΄‰ê‰BB8‘qº;5`C—©óQ=~éÝ Ÿzù"ÈK¿cöé?WóꬔüßÄZŸ¥v­('q·©Á½zõ ½ZëÓæ%-{:éèZ9HP?7ÈŠŒXêB@c­¾¦ú5Ò¡±J2†ÜGÍâRÈÈÌ‚|=äAÙw¯ SƇ¥öå~0õ›Ö($Ć"@ö8 /R@‘‚@†#ÞÍ÷.B´îUðù™çJ¢#— 9UG(¦’Fg‰@шt#c ®Cúêò30.½’@D Ô޾ߓ·æô=WsRÞ·7RÒ¯¤ôü ®“ððù^ÿ/ÞÖÐS5XALéD"Òað¨!X€»(¿IB;dð¥l¨ì‚yÍ…Êá97ÞÜo ;tíÒ`GÕ€q5õÓ\[¼IIP‰ Û|oÍônKˆ„œ¥:øP¬|í:Õ]8.òžtW=wâøìüÃæ‘ó—°/$frЋÁÇxÆý¾Õ@±®Ä`n ý~-páBí,Èy,#°Jn(MæC<-H0`Çá”RÀ½¬U}Ù8G ݈mm*ÈZŠ}Iû¿‡‘SÉü:ømè·õ][L—y’¸Íhº,ݶ“SÅe‘MäÒõQ‘ ÷3Eé;yyÒ|[ƒÀè^ÃÇzmð. Ć.°k RÎÎÅCŠA²IX}Ö×·ö‡ÛsmâÁ>$áEÿ¤æCòzE1¢:d÷Tû[ßBËSüDDË <Ú •‚gEi˜ÓÝPÈH !½ôÅÝ;¤±{ÞÜÑ/KÀ¤}žÖ§ãeÔã'xÛT¸¥ì÷nµçí]xõä­’Þ¾"dC £#«‰'uÛìÎï¾°¶ ‹¼cXI{>ú¨îã¥óS{är6´Çɪàÿë•QßK$…½llÌlûš b§¥öŸ™Ƥ±#ÍC+Æ^T󿦄hY4h8-“ÎÕ,"’œË´½Ç¬ò®ö!®H‰Ä ·r åñrDËoêň‡¼¾”ccšJgh¶4 Ë1ÿ¸/po\OèÿÝÑî 6"ˆd!•ÂV3D™1âɇŸÖ«²l²€H•Çœ% õ ä¤ žUµ@ññ<šiØ„–ÉìáHcv9}ÙÅ@ñ®RôS0>‹ÆY$1"òX¶”Ffr+·Óì+R3M‚t§x0`)ün4G§>ÍksÚùeÈ"{¸ÖwªdEÈ¿Ì$Î…!_æõúVÛ$eê’3Xásö¿#‹Ü›Õn²ÿPÛl¦Û›æn(ªŠúŸ§ç€eA<ï–õüÝ)Ú!ÛÆËè®/˜À`ÿ!ûãü4!+48p¹‰ÿ*qæjh¥P¸¹Î/Ò30Ò6tÐÏïèâ]sŠ/"‡Jø0‡"$ˆŠ)ȵúêÉÉ"†YE¦œ›Y.:äVPVrEráÅ‘lqãÉtÈòáÆJ¥&…Iia­(:®×2"P‹ZI—¤Šõ$ä“Kʨ3°‹Qœà‘„H Í(¿à‹’ P&«ÖÂ,(ïB”„ý7¥Ñ[åë4$¡{²ÔaZ/šû1Žr`$ecWFVR|© ¹"D@œ!æ‘|)”E¡¥%æË/‘…“@´üã9ŸT÷V“ˆ&‰œT–@¶j±‹¡/"ÈDd*&2\ó?PŒ²‹M3Y²‹kSŒ”¿%‘ £‚W&J°”؃™'¿ ú~Q?axº¿Õß7^ò;ÙÊT"o%Œ>¿ÝƯãíiË[PÈMäö‘ÏØÛÈóyIó‘MK@ãúì®*²òý7PÜv~r‘±ª’ÇÚœ‰Ü椸#¼çÑ®çÑ.si‡õ fÜ Z|¬bŒÏ‡mßÍ—÷ä¨H{­f%GB^›N†ÁؘÜÛÇ9¬·_ŠÕ¤-'MüÅ ¬ÏŽÊ%[“L-K+~½§b4©wìþìa!ȰšíLƆËFóÙv-Òz½f/ø";¼’íö}Ävó‘é¦9h½šŸÂäÄí6VFÅÁ¶‰ÙG—­§ó­©\^ Ì_>Öc+¢ªéÙaj—“ùIÊ]Ø|e¥\-·€EŸ¯ûËhÍP±¢Å6Ö„©¸›ýÞUÎ7&mJÙ‹‹£¼ž;Íë*ŠŠxyu(Ã×ów»¯e¾ÈÇÍñû ¢¢žg”8&&ÿÙëôâü,kª\ùU¿çfaf®ýˆ„(DP“`>ìUT…Ùéä8§@°tý¶g;g§ö©Oa(BŒ~íŸÉëüŒ~oö©<–À éR’<:r:µÇ™Úà«€ ——}KýoFPwOÕy¬Gÿ·‹7­•ÖÜÈý:BýÂ!ె1”€ŠsH!xò2Xqf¾LŽà{^:xžÊL{=§¾{]ãÕü낟ÐdL¹•ÔÍñYóYÅzà€öt.£¤Ž-/?_àä}NIÈðzcŒóÒ˜^;ÐÕ>9GU¿""ðl¬øéÎÕÂàäT§%ê‘èÄ@½0ÇÏÃv´ ,g,gQ%5Q%<.{5-3õ"ºþ Åã£éí­¼ù·˜.#;¤-ñÇMO¼ ó7K<ù€À˜ {)@²é`ÅF’^öãC¿í.žÎ£ICOX"4®CŸ§äRÆ­°·j×\®ß3¢¶ì¶Õ›}{·¸#çðÝO}¢´î";˜AŠ•TU†7_¾Á6?Ÿ©)æ,çu·;<´åó€”´.J„ DHìœ}[éýù‹Œþ%®~ùøð{<]†'öÙUôZ¦ñ^¼?§ÔàºõÆ ›3°k‚®DM¥Ò±òcU·Æš‹Œ„Õùè5!Ÿ€ÐÃݘž‚#¡Í–‘; Ð&!â^7 2tàååJBôšü$@ü=޵ZΘHVe(µ³¾6ø-°ö:×›3xúÍŸ§yòßjóóÍî%Ç}[&H(wÏZˆ<.ÂßqºÜ޽ôZðó*=k3¸}×k•‘ˆ$D"<œÿÝÜW—DŠ(%H.½†ôõãóæ(tD…¡ôüÌÛ[{-½‹VX‡{ùŸÝùÙR\ q`}“e¿áï©«|I»âÙðvË‹oËOú`@!ˆ lµ,Ô'⃠;YD­Ë›wn!¤©êþ‡s‚—Ý9ôŸCõJl•ûsG¦m«ænå8ëýÛѹùÜ}5è£KÞ?" Ó¨ɱr"/±ïæ0 ¶íûÌ x ùuyŸØI˜¹¶‹™ÎÙ 2HA¡Ö,WÎψgíö—ËL×ÇŽ§DòÜvßž>Ö¥ù[m¡î~G·§ôýÿ©È ¯¾"K†‹LñäId/ªÐ»|]Wný&cÄc"ÁVa“fY“Ó¤0-0 ïÔA  E)F-Œ¤NœÊÁ’µ8ã'Oœn›ó ;‹9RË-£2FÅ }½èeHÔQòvyob=ó¤/"ÀX³d5ÓävùuM£65"ÆAN¸¸Œˆ,¨ "K\'«Æy¶V0f6^¥ü®¬Ö£<©Ôë$š0>/ãv­W¾@Í!¶Dc9E€ÞiÐÇÒÛø1ì8÷¤IÿuO²˜i âYEð‚<,6+ôÜS>‡Ð•º"ê'-¡Ÿ2Ê|ÃfBed2eû&Ñ@u¼›*]³@F2žd?ó½-AƒàË@Nš42PŒlÅ&uÒb$·?žpŠ”ÕÁ{„Øú ëi%v'¼Éçr‡Y·z†ÃHÒƒ­šT ¢Ê7 q,úªÇ”`ßRÂ’\gä#B3³“7‡šÇs÷ŒªÛ£g»"»;Tæüåk‚ÿvJm×$¾¦•90¾¼eÜŽÄiŽH5ܸ֒$*:ì,ë NThóa –ÚŒKv'¥¾ü‰ Kmö#@ÛñvÙX¦¸Ô…µ£"L™Ë›qäöRñ¼Û“[£1ö§2úNBÈæ<ëQ£.÷¹. ïÏŸ9–æ?BZf2˜„&c¿ð(L5ñœš¡!¸–Y¸Ò”ú&Ç`àö")Jœ$9'A¹&¼wWb”䊇!ùaahz UÉÀüªOxæ´½ ¹Ö›Ž´tcÏäÑî¥è†dú!ÒNTÜŸÈ¢ê@èË¥°9ðíçy¢üI~=´C‚vÿá§qM,Mê½Í{Ì{ánœ§ÞrèdÅÝFØdtèþLqb=©“5/жcÔËŒÍrÕª -! ¨µåŶX Ïèм°ls3BQ Nò‡{ua“õ)È¥f Ñ (]2fÌL˜²ôY‡$mƒ±*KfFÊyAy‹RlX E„ãMbUÉ$‹‘%†Œ…11ªa—[ôkPÒ~.´½#.ŽDW&:ZåQ[­£ƒ ›01£Øìþq¡— ØŽDpE‰Ÿ—\ˆ‰¯D ‘1àˆaE1&|oÄØWuG.!™l†÷4n $Ww±n€VYÑ–EÉ‚H!õð1 %Ñuâã@ºÐÇ—~‰òb&_ƒd X=Œy0 6Ánˆ™P,аëEÁ1¢:1Ñ‚€bîðê°ÐKc‡0F4`Ð$Ûh‡Ƀtp 4ˆ˜ñ: 1à­°0¦”؈8")‰/Cÿ{ª$Lˆ¥ äÄ F ¿€˜ÒDÝCzdò(7â=ü ¸à†DPÈ€º0/ÄI>$BØ¥ ‘Zã@˜´*Ð/Dpã…ĽE2 %±RÈ-ИŒ‰“M.,/Í0À ò½Ç¸ï=ΗK¥âvý‡¼ “ò~DI…^Án&o‰ë0$Ô0þg·@Ývÿç“þ¡A÷>þ‹¡Ÿï}Öµ£-ÖØ64Ý5KúÞUÚþ_²+½„'oŽÏAšÅÆþ „ØŠE@Š’ª—®PLfu^o¿Ís ¾»Õ)Ïðž6sô›[÷Ãô¡î¢Rij‰)òroÁŸðodµbÀ+€Õd^\À¢é9Õ@¨â‘~ü?AQÑ•Ì`ðpvuÈû…øa~Ùö‹š%ú?¶˜\ôëRëþE¯¼ý‹Ôýù=ý¯ú,}£"&Eo´“ÆZ§³ÛW©X…b><šô>·'Ø#Ï¥ùÑò<ý§,h ƒìÏ\%~Ãn2ê`wçÁMì€¢(€DâA.ú;yŽ9ûÿXqÕ 9ï›ñSÿcÖäÂó¾8Â"®2(— n;%ˆ—h­«r†T¥@[¼°Ó/…ò^ÛüO ó×­Äà#ÐDW¸D! Òˆˆ<9î?Ý5gýôùªþûø—ñ¶®KÖyϸãGèºýé@FÞ„’FJ%ê2><ð@ ÞVGÐî:§Iúãß9‰û€"&ðX½ÙŸùÉR¡9…káìÔ¯]”€÷\ïü¶ú›[U>ñÔÝ¡nÀ>ßãñn§uáW^®4ØJ!E2BQ+25©«{n¢£ôÈqŸ Ž<‚®æµƒ3BÕVÿöCs=,~Ç‹ü÷@",¶®[¶¹ÎÁ§_.k"x¹›Ý‹ßç·“ÀÈ2ý´E@ãìyoécäæò)€D\ðõçïˆ|âhbW)H¨ŠJÉZlŠÅ4 ´Â:DZRQ,ɨÿW×tsù͇µ9UIîyÂ>‘.*Ä¥}O •E®îɬ۴\;¸qÜHW»~’!ÏLY!)DXª(™*†åÌ£…ã!=Gˆó2Œ4P¸5!© !ªš%ÈU†ý T (g OC'òà2f÷Ofìò ]„iôF?Á?§j±UÆS‹þž bήQ\/ ™ÄâŰ¼EÆÒùvrN+ÓÛ¸kj©Åv€  ˆ&eÀ%ZúK² A¥_ x(ctä\Xq‘IYe©zøË°ltsÁb5¯êå\_-0Zi(d!Ô´Lx–`65 ÕJÀëªo·ï OJ[dZ^ÊWâ~ýß™Šš¨ÐcÉúÐ?#œ‘vóýSY6QS/ÁQÜȪ3×ïëÍ Ie·PÃlBëpÆá’ç9ŒïÂÛNáߧﺨ9|yk|¯N}œ4Ù9ùúYjb¢‚[RØŽ¨U"¤Ö+¹Þ³ût-ÇŠÔjSdËTG7í"Lg“e2›Dz·×‰­‰í}×b8Æ´…#î¥<™í}HmJ¯nþ=TYnTеIeòüf¿**.Ï¿NR“„‡Óaœ‡¤Î¾G²› m™–nÓŠkr"aj4ÞLYÜ–¼âò|ã$—?Ÿ©îó2O¬_Ïé±®ŽÓ œaÀ6}ìE¢ò_²4ôÍ4 ¸ÙõåêiÍðÿ(òYŒfuÙ|¹ø“‹%ùÛt©2“¸é鿱M´¬©ÞšÂp ò–qj,‰CcÅ¢;úG³“‹Øz}ù£× *ˆƒœ)•UEÊ9r¢¢(Š"¢äʨ¹Q\¢‰•QvQp¹EQAP@\äW* ¢«…EȨ¨¹DEUÊ ¸\¨ärª9UETGeATÙG"9AQDQAQÙDEUpì¢ ¨®r åD\©•s‘EÊäTTUUù£æƒà ˆ.0Š›+”L*®Ë…vʸU0.PÊ ”Ùv]•]•3”Q3‘W((¨®Ù¹2¡”Trå˰¸s—Šår#²¦QG""ì¸\ " ™×x¼9ð?#ö~I~é$@‘„ Nã·å_åü›½–FBvÏÞ÷n¦­Tæ  åb‘ÚQ®-œuyJ³pq=–w|ôu“–ˆbÛ2Ø"X«»—öˆÖ•™û,†_‚ÿ?:d¡(ˆ†gv?éz•ü…Ž*9߃Kf¡Åds7é#×kšÕ»Æz2g ÈRDAÈIb ^9Øc»‘ÏÉöÜmš†ûäù³Ù¼ž”8ü€ˆºþ«ÄBµHµ[ÏsõyÂBÿjÌ“0š]?¿=V–µmu#>hÒ"!s§ˆ1Ä:Þ™ùøFúÜIÙ¼3&ÄýµË‡ŒË#j âý$áâ=µê‹GËO,ÄXcÎ/mÅšç;IW¼9g-C"%ȼ~¶ùŸ ËŽïǹ¤ì˜ðˆa CßÓòãô'TžïbšZ"yÿü0§ªÒ«€âZoKýw<èoÞ»á¦N»¶˜uì-÷é@[L%ŠXˆ„èÿO ë2Íõ ‰TÅî·²ßÀ»%­ƒÃf"ˆZn5üöóUíCÞÊNÿù""!ìÄ»"šû™üÿDp]]R'[clJVvm/š „ì¶"j|ã\d¥Îu¾Ód˜{gtˆŸ¸ )±6S`KöK)Ùß7±«`-סè SØ}‘Uð¨5„h×Áýq%ÜÏ´ üNöŽ5L¶?:]P2H†Áô"ò,/²Ì›R¬‚äV¿.*!QX÷ô˜V¬R´Ts." &DYÅCŠÊ„Š$¡&!w¾³sõü¯aˆf?:32†sï¶Œ´Ã%E/#yù¦1|ä¦bÂþè噤‰k-*.6À«SÈä\žF»lƒMùÈñzï^Ø£ŠR8rÑ·™=W#gÒz*>MŸËÿY_üŸ“þ6ˆvÀ°…›–P¹`F ›±.Y™]×ø½_r(ˆ€XR“Ô‘Yϳ…Iì‚ؾEGì…I~«ývÿù!¯ØÚV[­“å7°ùß[ù6нs¢P[ £»×õÝkåñ=×Wðxük“o?›@5s¹àe{ÍB“ˆód¿  ÿ]ÿÏÚ9fk^Å ^¿þôXàk¹i½¦}ÞÚÆÑ¬žÑò@"™°–N–Q¡«š•–Á7A”® è]lÀ ÓßQ¾žcžZ7.Ç~:ü;±µÏô§8Æ1 €'¹ètð²µ~ãúZÿ?ødž,Åñÿb…—V–•L;hg¶Ekn¶§¦V{†: D“›Ç¤:êé\ÉÎ#Ÿþfv%:Ù lÙ³d„r YWo33ÞÌŠ¾tžŒ¶×Ñq w }ÜGwÀ QÅ!G•ï¿èËðlÞi ؤ%Bß±Øbö÷튶¢þ·!5)x“õUYQŽ·8 l*)™¯˜ $DYkýTÍ éÌ•}‰’’ÉZôȈFÁB0ˆ‰D¨þZDe“ÁNÃîÑ>õspæs6Öož!_¶X¯šÓF ”Ýʆ*–ˆPÇ2®%ÎÝnÜ@fZо¹WTÎG.‰âzÞÕ¼mƪµK[òžºm6,WÚ¿â×2ËQ†JMù3”çb¿êß6²Z“©" Ì**J!hª²äÅÑ¡ïI—ÊZT„³*jϱ,8ÃH~B‚`š•£”m²ÂÐ’ØHë&½¡ö¸BR5GÐNùˆü¤ÑûÖ&¤ý|yÂÆ4öj3 Ñ „º­§£úÉw„nÚrÒ§'÷Q­ä$í’×y¬ Ä¢-QDT@ Öß‘sïP| s÷R›>^¶é$èÚlÑcèì›Ú+Ÿ@€•]â.ªsÍ/—á$R@KÉl¾PÛß»*›Þ£ªœœ¦¹éb2À‡*ÂÇ©³±ý¿Ó:ÿ6¦-{xOöÛóŽéhÛþZäî¦è4}ëè°D@NH3»èÔl×±Ä{s—ä£,¼ð¼™K­Vbƒ^éœ÷ÒDE™¡„[zm DCÉ›<8çpÛæSnåwnJµÏž" ŬЇYí<½9î>Ýï[|Ú÷ñ4ðoív?wò;IØÌ »•á;Š WjcÜ¥~ÐŒB±V;ÙùÛùþmGÖNoK1|®o ¥ϪᲞ¾·„q‘©€)¥4Þ¹>Ã@м=°á÷Ì4£ìÝÕñ3ËûXŽ*ä½Õ & ÈÖà­Ò“)Þæ—…GjÀ€>gÏü»rþ]•Zž×ƒ?öÈ€Œ@§.£?žÞS$Y¦@ðô˜T§¸NkMsÚTÓZ¸õ¼ÓÝÉï7;ÆM~mË8û¹¹?µŽ;X[BC8ÜŒ÷¦ÊÜa,]Çò‚ðùWgdVFÄ@C“áR¥69êÓþø›Wƒ×Âm_ê8¬ˆ¢K‹Ü>ØÖ<~ÒlI’ D$ºÝ¯Ý›¾¹îîPP<îý!ƒPeؘÒ‚ÅÉ{ ia´áL@1¿[4Èá)†N f Û>ù ëö÷Ÿ•ó”ÖÒ¨ïÎÕOX€ÃA£Oñ©QÞ*F²:»¯-¾ùd&¬Žû‚Bgbü° ƒ+‡ðŠ7„Höl6´|>ñS)@ÜEä.8M’ÜÆ c-æ–ðËr£ŠòGãgdºÜæí¾O›Û ·¹7ª²³~cÓšÔv#ÂØ:|ú(ðu ôŽ©8ÙJU€rk)aÏ1-XølXd>»/4率Ó\ˆÑ>†3ý³E5.ðkubk.Œš>eÌ5*bÖN(¬¤Ã÷µCî–)Š«!2ºbUÒ¯œ) h"èB ]\ñ$Fºõˆ qú•Ån!ÊùÖ—J¸ˆñ6`Iår€fû=¤ý¥3»Õ ÌO³ÿ  ?°}Ÿìùž•96⧺á…À²I¿ãñ^¸ÖÁE«RH²è/Ÿ€ÛŒaÃ$gQqU߯­ës¸+H 0è'¨Acå®Qßö¶õ]G¨Uƒ†øV bý]wÞù!†Ð‘¿d"µß#;ùD!Ü:²³lYf7§øCQ_í;o61¢a‘Ü©*eªSs>gÿnä>åÔ óÈl¶[[ìkl~'Ãßÿ¶}éTï;h9Ð~ïÁ<°äád™öYÌÚÒYË1Ôg²Ë©XJ­™êÛ½<'HÖò¹T2Ù+Ë*šüÆJes[£Ê(áXû¦wâ­ª¾¨¤eõN‰ya™[³É+h°‘Ôðîø Œ÷ aÚ´ûQkãV?ÓG¯5Öof «©çÇXh(ts:Úݳ}ñDþaÄ­>-ž¯±®·üúÓèè±ðnÛ‰:”’ FE£Ù]1"_ßBz†±ÛÌçm,¸3é·´ø~§¸Ô¬¤8f¢Â;*Obôõ;A”™‰Î´“1 ¤¤nHï'ÝÕº—L 79V½8Y\ÈL?wŒ›¡œ§› áz›»°¾ü‹²bt óq8û¿½™Á˾•'°“Üš‚$Ã$Ús}|SÇ€©ê]LÁý8ù-2˜oÌ鉵 €¦Áz©(5äzŸD0;ÓBr1» '¦ 3è´œ˜QµÑĵ ^§ÜŽÐ1˜ó>ÃPÜiô‹fôŒ0z1vR.,§õ9}Þ=8̆çxº5j¬¿;ÍÐWò0t<) 2>9£”UT)I›áhÎAT‹ˆ‹ÀîU wƒÆqÞ&Å8‹MîÒÿ—BœÓli Ûé¼<{Ÿ™£5ù5‰" W­Hñë½í2<ÖˆWL‹ÉóV°E{{-QâøM‘{Ȧ7€ó,ha‘C ËúO¼b6Ñá cÎÐÈz Fúy÷ ¼ŠoˆWâ€Å¬!i–Z5XóÞ+ÒöZxñúýŒ}u¿ÕÈ® µ0XŠ_{ò©ð,êepMíašžEñIY_Û#Ф·´@B½½¼]fçbÊ,TqNŸ“‹¨ CÛ^¥´‡k“ÊéåªÙ(¶’ÆÜRGš•£üŒÈ=ÔÀŒË8ïÞøwvèÜ ¾¦­ EmÒ ~Ϭöý­;`Ïö/õyØ8–QŠ8D@Å”t º„D 03¡NK¬®cw$LèÄÒí¤»çäÍ¿ï£l ÷ì4r¾§]=©µ3O`ÛçÞ'ÖÜ-øÍîßò¹nkèÖ@®718˜Ä0¹ÌR½ªËöp*sô~Ä»ÇÄ)®‘+!ˆd@§h²vû¥âòt™ç†¼7 ˆäû6ý>ý4)?BºîŽuªßã|½ÒM3žl#l'œ‘X"¡!nÜyr—Í´_0~#ÌËN¨'ÉðÝ=oö4dßàÍÎÅТó’D@h²,ޝËÛ4“F•) ôù:¸³õ=Ëcn›éBÓÚ@ ››¨ÉP,±`X`ˆŠƒK‚¾¼²Žj!rö28¿zÃq]~l‘$IPÜ*¢OÖ\ò§]Ñe¸}þ¦ýFó’Q’Ïã’‰%C Zf¤Aª˜Øò‘ Ê#‹1HŠJ2µb!BoÁñOÖ']Z­BGóë“@«Z„E •ª94’JŠ+ÔMÓÌr´«%#Û›Š§¢¤8"§)*®îNÿ‡nqïŽåx÷<˜Ò,¾Œ¸¥EÐáé?Y!!(Lb]®UP üj0™˜¨‡Wæ˜K¹NèR*~‹É"+ ½Æçôj»¹ÌÄ+ER%äú²ÿþ{ëZÔ‡ùk%q Ú¥ªâÐùrmså\¢PrÁ˱±,k.Ĭn²Ê¡€ã Õ剡‚aµ ¹@-Š1=(‰¤ÞhìeÞ%é ³œé$L UÃ"ïÌBmq{Ý a.ì½)¥8gFJ3|ç{¢¸=ÀßRÞ—³­Š‡¥ŠŽzäÓ¤n=1ø»q’6¢Cû¤èÊ•.±îÈgú¨¥ûÏ­¶Ó²Y•û=:Ó'S¤Ä¹£DL¾&w²²¥é?;"G®SÿÀ½4ݵsUïOujÙ{YVØ•Rô¶‰á]6“ÍTÚµ&êÎ`Bo2ʽ߀þWqš~l¼jËS‚Õõùò{>êºCâç´pJš:Ÿ•J•úš›7´°£^Ù…¥™^¢7ŸO§•˜ÏsSc¤ ÿµš5NÓXù‡*¤~5E³Ò5ÚB`ŒyùË¢)ga©¼qÆ"Ñ×·SËÉTg™üúÂ1Ùö¢¿ Ûóøtñ˜3¦.ºÿŒÌÌ3AW.DTrª*äAE¢eU\ˆ&ArŠ"¹EÊ •C"(¹ErŠ)—"äÈ®QE\*¹Â9DTÈ®«œŽ*"®QUEre\»eË”D\  ™Apˆ9EDQr*Š*(‹„pˆ¨‹”(¨*©ó|‰ÈˆïMÆÛ!m!!dHAaÅ\› …Ë…Ê …*©€Er&0)•ELå\‚¹ç.AL "áÂŠŠ€ª (.„QÈ¢ª"år™Ê‹G ª HÉIÈžÛw{¾:z͆‘@F`<©6š‚B"'ŽÅØFÐ\WzƒË —FŠooüÁ~¨e”_;SµM6Ï ãÅo¾çoÇï9ÿ ¢ ÂçüA4 E…9€‘Ͱ‹“­wGŸcŽÌ8Tج9¶þn/³¡…` ¶°9ÓfÍ @[Ïí³uô·õpR¢jް€Òé˜Ü÷ñ©5(0F¯Âï/­²¼÷cû@Æ`¿@€(¸×[\ˆqæÊIü|Þ{G6 %fŠzö³¾ó3Ðá·*étØ(œð¹&÷ 6xG»Oï1uÍ€€"‰Dµ§È(õèëz½CnëHA®ŽÁ­Y¼šû x°”Eßðñ’AÆ_ì/bDˆ" "›ªòàTÉ^(±§í%ù=Çm€Ù‚!ß`ßLÂåì¼Êü£,{.[1Öÿ-Ñâ71}ãaó?~Ѐ™0W•àM‰°‹K##)’"Æ×¿JGÐË`®|‡äýå­îD@ˆÙÉv9sÕðèr¦ë÷Àr …2ªèçw¿ßôz^–6á á}:ý%}>×k`Ú›oìuÃbN÷·[§ìcJyÙ¿K×çÈ)±²CÙªþcð@§zBENnþÌí¶­¸ }NÕ -"ï¯çËŽÔm.T´¹ú®L±þ0¦‹ˆ¯Z‹=˜ä­j9i{ÂŒ±qÖh,^ÙrÿNgé¡kE»yÞÆð1Ñ]Ž¥3¾Ó¬ø³€ÓÛÒ„•““S~Ö•º’¶•àÌK•U¥ÃÄ¡hµj| ÄÀå¶­û`<1ö[­¿f|^> I&âê»<$gÊ–´LD6X.¸À#Ñiìíp+ÎDM_¢ûŸ(?¥ËÑ·n¸EÊÃ{9m{2möÏù`„8d2 D’¶œÔ¬ì»€|ÃLE!›k–}®åÿæc»¨\ÇÀ€<ȃÖä.³KxIæîý-Rjêå€<ÿÔù;y½ÆÏ{M®Vçoèü´rL2UOGЋj|cìOv=–(Ôh5š{¤mí/U»‡)Zf›fiŸk@€!r«ò’B“[<¿{ú—y5а¬{§næW(ïòö¯p@ ò‚(@@€@€Ðk©éëbC¢Þ_çr_LúüÍÜ› q*w÷ÑX"¦á âˆyôßWP36#t}æ%ެtª>á-ó\Íöµfm€‘?pÃ~)‡«qiÀJ Åðîp€¿Ý··Ž×°Bð´{-&ÏjiXÖ»–êßô÷=DK=_Ïü- /3F¼ÿ{ÉØëz*@[vrÀ Bõ†ÛçoósÏ[¦n¦}hìT¶) bÆ:ž8g½YêQý:¡ÆSºÖÐÄ8Òßx¦€"ÛYšÌ^¤(i ² ]E6nÅ»C6ë¾µ ð¸¹ðq®­©ªÃŒ÷˜ˆfB ¼í–&fÀ!+¯"ÙõèGZïàЦúÜVvkëÖëv$ ŒÖg²"å9ßÜ(ÃQUΖK¯ÆdW±¼Vÿ{.U‹òªÞ{½—üÄE‰|¹«`~oñ,1|«r,[km §ëc’6.½}üYwyñ„ˆVu×ìvUs=ÞþþàW ;¶>ß6ݾáæ]6Fœ–ïéô¹æÓu 9ß[ì!íµÕcÛa8Vm”iŽpRë,šŒSpŠg0X^ñŠÆÝâŒfÃXŽn%Ue”r¦í>xCÇtâ{²ã:_⻌¿?‹óô|Y£SÄÐ@3L~\'áô%šHXQ×GÕÅ«޼å 6É~³ ¬áÿXEbæooýöùœ§t³ír€³—tó‹>lÜ»Ôû›É½Q™PÿëAcB€5{+ÙÇϘ[ãP@‹KœÔúòkѶÐ'¦«#êNÎë·±þy«ð÷Ûú^$ÜDW;•A„]Cu“•˜˜â\ ­¸ð_é¯ÈÏ]è”›mq„`"+ÂD€Váè.¥fýî7ø?ÀÙþÄ{ˆÄÆŒ/]¬k÷˜yÖvN¸‹ÓOöG®ùà!-n‡êÚ$ —!ƒlu¸v¿ËêcØ1ÝÝéoåops¥½ÎÎ!„a„BW9úQÙÍsˆg(æŠ?ÚŸ½zVžcÝŒ¥dÜ{ÝoGˆ8¯·›tp S¯ÜÔÓIJ¦Ú‰àÙLtu.¾~¤:ÿnìoëÿ"Œ»Gà”ɦZŠ‚‰¢W"¼\­u€!xõ³}>Á†ö.’Æ$ôBHá1Á8ÆÚL„:†·J6ò±ññå…ŸÁánÒ1µ""«µ‡ÅP4žaÞþŸ‹ekÅÕÉWâ`cˆˆ Å îÝ@ŠdL–—7–Ïì\"dìÿqIx¶ç¤œ3Ò]Ô~PbŽ ,V–ÂAçw—‡4—x€ˆ [c†­¢B¸\Cð·—pÅ#ëMºùEGvÀÉ1pë/¥¦úLªy“Îq"%¤a* †ÿ[Àl<)cÁëð0[Ø)ƶí.‰âáž%vÓB,,CÀ'— \.³ $¥ôñ~8ݯ11±´t}Òi‰a“­™í¸|O»s¿tŒ™£¹ å!=0R%¡ŸY™a+m–Ò`„¬÷\ÃÒâ£.‰Q~”Æc}%¯¬°HJ!Dv’±¬„#™H¥j|EדU1a#°/Æãé·3ÏènûÓªºëÇ0^éA&ßgÖ8Kn@Cj£1ÙP%d_m™× lzHÛúÛžãok_9+ñþ8Í|ŸR1"äRj\Óñã±:`A3ssH7ï—Þlˆ–vׂâЯÞÞh,£W™â“ƈ˜_W2~d§o`@ˆ7üS?‹_sxª´gCõF}!°;V¾+fÚ`šŒ. ©h˜êœ[y½Ô)ï?ç~ÕÀ)»öÚÖX)¼<+zö(môlðµ¨òL¯ÊCTðkòÇkÅWÛZÛ;xÞÚ=æÏcÄÚ»÷½Léf¹Ò„•«°†NînôË€ï½á£ë¦÷yuNíääóGËÌúykWÌk¡Ðij U²€ÆL`¦o É®¿Ø"Z¯õ)2×ÛðùºË],3é\þ–Ústt2±\ùйarý¶JѲÈÖw3ìÿÕÛWÊqÆž@ðá¿•w”?¢ÄœùDPçΈ¢Ó’Ž>'J Á6¬Ú,h3µx5–ÊŒµ" _%Òý5k‰âf:aÃû5+>“.¼æØÔï™z·6Nç@|,Vɑ쀯nòàÇW¡YŸf¥f€¿eáû¹Ÿ‰kO2œ8b§< ÙÜì7ÝY£†„ª‰:«·ïçß–µ'.¤qp~»J—i73¬X @Þó{Ê*ŠTC}M“6ñ¼ãït{íM|DrðbKõ |+ûþòËíhbÏ£– ª)”ðûê!{++¼¨f@:üKE<`.žˆr15½Nvú»Õõ³ÞÞÁßþßÝàpß g™!×^r3uù|Ÿ,0œßwЊþ.äŠp¡ LVks/data/tempb.RData0000644000176200001440000020205413265504513013633 0ustar liggesusersý7zXZi"Þ6!ÏXÌé ˜ïþ])TW"änRÊŸ’Øâ…§ "©Ù"ÝȵdׂžºèŠ,‹ÍÈócÄþ™…éÂß{u 6™åòÁ L·ÉÏ‹>ª•L³×˜’cÚ ëeÉÂØ>mç󰯂VïqN J”yš‰›€LHDì‚'Uy˜‘Þ‡XÐñœÓ‡Ö(kCÀ)úÚNØ”Žõ;2 à®\DήAV¡¡Ë$? )*˜—à qicLjãi‰SÂÄ݆'”uéÜ‚•ìKp²M†t²;¾Ø v4f£¶âµóPr±­¨Ï,.+_9´Ë†Á°§ÉÙ¿³ˆS‚˜ˆ#%ÖèK3€Úù <Ñ’T/n9o@”‚™ó ž+ðû©Z<'|ˆf¢;4M~¼‚ À¬Ž‰ˆ”:^ÕŠdTκYçÔTõ» ÷‰w :‰w%™Êòšgý›qê¸<<˜­3qŒ9®Ú_$< c­ä©”ÐJµSy›lsÅ}9àò„IF…8þ߯èÍ´.—ÓboHrN$v @m’ñšŒØâÒŠžèuDQ´l¼üMÝ˧Ünךmý(ˆ<èÞtú y[YÕŒ7P5Øà‚ô1ð}Êjº/¦i€TLèàŸˆGÄ“¡ÂÒvì¨l½cÆyJÓ2pàÆTrçhŠÚ¸7RRuo} Êê~ŽÙ"v[dpw6Äduy‡dÙ¿ ã—¦2ç¢Pù¸:9®Íµ|—& ]ü’¨®Cè^^ݧÝxúF3Ïü²š0í|£6çµÐÅÇ}ÁŽQlS“l™iñ'A Dn.Ÿ‰ñ[Bÿ~-íˆfn·Õ‹Lµ¦ŽßŒ´éŸÍéÒøUã™rÂ…[3µ¾‚ÔÃà& «Ó÷ˆWAê!3­HÜ=·†3ЛSf\ÉŠ·”lÄ/7ìÙhý‚Ù –Ÿ(:¦oE/wK÷éîÂ;é!k^|w–}™;r\ ö_fšœßÙ]Ü£¨Ð$<¼pºÎ.bÁ³sÙ£ëcí€R¦ˆÏ$Yõ@ü’L$ —ŽQ7G,4J"%mâ“ ©Üƒ7í®:?ùzÜP¼îºy®ƼÒ¨;Ë|wnŠù‡JñZ ÈØ+¤Í&’§ú¦€žÚo{K¸‡ù LI¦Èjñ RÞn:/2PCK’½Õ7Vó¡6ß¾~ M5LÚ6-oùÁÈЕg'a‹A5Ö±Û.f¢—ƒÎ8”î+kTÞôÍBMiüÿ˜_³ä3wóüƒG§1õí€U9Tf(D›cýìdË} ª|Ó„–bñDIôÔ:%ô@œ¹ÿâÿºÌ¶}ëˆàŽîºZeC¸¬Wj-ù D=ÄyZ`À¸þÜÝ…‹âSûÀ&¢§.µÚjõ\%kœùßBgœÈÚb°-³%å ì—/í½5åº#ÜÑÉ©©†OjÔî=‹”Š´¥sbûÇ{†Ýí¡ñ!Çs"‹È(Óñ?à¯`Ã}P<íA šŸò3»´Iò¢õÿ „]ÅÞB}ÝZº‚ŸÇ&;XŒß1h±>fï%#Â…”>uô'GɶY oä2dƒ¯ÓˆçT`yTYO 6&ª'´ü(AÕŽp¢,s^—]Ï¢‚‚‹í´ðÀq„AkßÙòWɹ†6- “fGù¼•³ …½¦¯®çºó;JnRTFçÙ ÆÅmƒßÑÕÖYøpߘ2;ÇÝÿy7D]T%ïìõ@“î®ü㻜™¶ó9îOMÞeú/´E«ŽJÚ–BÔð·jP¤¨a™ ÕØ`%w÷Ì&Ó\á½%7½íZ†ØõæGø­ìЉØR.ÿ[‰)FŽaa‰êQßð¯çšLÕ:Ý·0XZr/È¡òÙ¦Ò?sEý8_ˆjh¤Nêm‰@f—РĥR˜È`mLC­­#rg¢_w¤$‰½:оzø~5ˆPí„&Ö#×·ÕÅ¥@?Î]QÑì .d"ê·…XƒÂí ¡éÄ/th ‚B¶ËSïÿÖý/,· ÐÇT€ÀKë²4êþ´p¦©]k²R­'—Éâ'~pfáèñaþœµb:÷¾UEf²X[,‹Œ{PòÇ?k.q¦i²®uϸ¾@MZeHÒÏðÚ¨Z`“Cläßçh´<Ð3%3±¤Ü/5ÈxKš›I½s²Á„Méüà¸7›yÏjïªßdÞ³è1vRTM †?8ië`Àx¸­]¸ßFCÛ¬gìãÂù‡™`öåÐ ”vÓÀ”ÚÖ0RFõ \­þ„÷7FÁ½ ȩڸZêbøî͈Ñ)ÈNå$¾ºiéº3T:«(cÇEv;Ká&ònÿð¡n9ŸJwÇ¥Ó±º)åü­vßcxöÞŸ8½~PÍ„ ŠªX…ô^ÿŸ£ôª…”õK²D…²êxâ–[¨IvŠùÉKí5ZÌHÅÜÎ H 0 ÝçºA~ÙƒÆ?ÍïÁB4yµÁJØfÜ4Jë±øMÿ´ÄçÃ!RµÕ¢Äg;æ8»LÛngžmP§¡Ý/’Zq¬Y2Wõ3Çœ Òx$èÐèé{ì®QÆ%;@#ü2AÛ¶^I<„u^, p3uÌZ*i„ÊFB”´ûBÊk„ÌáBïˆòFD%„sýàuä÷O&Øb- ‰±‘l£ H*†}…]døzËëòÚ­o˜¨m™æÍ”4Æ÷¢ äR)n!´–ñ@ÒóA4“‘‰ÄR…Él€ µWM3Ñû¤hfk™ÜD¸þ3[ƒsƒÀI4„qœû¨üïS°Ñ&¸¹D¦rŠj»….BˆØØ5¸6šißËügRÄÁÊ›‘ Š)¢Æê Ä!|h(^}[ôiºéŽz Y{'¢Ê¥)ÒÛÿ~ê2¸>×rq“ºXq¾¤•:´˜‘ÚÆ_NbÊBëQ„Y¼ý·WŽâÿ¢>ÙG_*t!£"…²†*Á÷Ùø¿é‹‹§Ž™¸&”„Vsl¡)üò±ù« %/ £.D,\-äÓ# ³Ï «!¦5#}tW5ç…vˈ^ÉöHg r0á{,(‚õ×|Ä×nùÝ¢¶©ž'SÚ@óæ[ôؾ>«æŠìâ?¹‰FŸ§ÂŸ;?¡å}ŒôŸ[†Ü8RÅ9äKèGÎÚw2x;é¶÷áým½P% äÄCØê¢NÅ ýf¨“~Ýdé,È´SLx¦›´FÀ•D-Jù/±C&N¤L¨:_«£¾¢Ü]zÆàjDYRžE ýëÐ5¹«R+;þb—ÁÞ¯s(²ø~¹DêeovxìúF˜۴ݼ&è~`Co "®/p×ÉBá-ró.#_v8‘;ß*„ßžûî2~¤Xŧ²ïï­9Î2ÒšVƽÆï/TÈÀO»uÏâ~ûð7¶!cÂXKsŸé)®PÒáÇ#¬, êkûÓ¦V Èÿf×=[:ëŽFr;Ñªå©²×æ‰& |ã(¬T=Õ&ïõ£í +äi=Øl¾kWØ^•ãšóÐ]'h~ï`¸×-Pt 0¨8/¸h!Qç kÍTŽú2AÕ˜ùÑ ‚‹åkš–\~@”Sôš6•© ¤S•õ=„µÀ`‰ß¹jÊžä5¥4F®éÍgϤk¡Ó<ê í(ÄTOë¦V$6Ö÷öUt@õ&ð¯¼w8zkòJ)„÷,zqD©e7\?×,Sò^',ÊáÚšÝ1[ÅáåøJ48ggé?9D^+¥¥Ïý÷àb®âìÉ?»³ÃÂÞHÒª½ªz×oœ¡§~$3ËQåGÇ.(bÀI¨¯3MZ´7à-XNÂæP»d’Q/Ç…¯X/ËK»ÓÊPíóQ—E1ÁçX£|\ÎçÖH(ºŽ <ççS“¨=ÔŽü³õ"fá»<½OpcÃ;ÿ±©AMhÐúæ>[ºÚe[<Æ€7'à=õµhûE8¯3!#KˆÑ_l…'ĦH)ñ°´ŽT©ÃÓ©×`‰ÓÚm–ñ´$ZÿK*Tw^Î’¯ù).ý`'¤W’ðÛ nÑË)¸° ¬ázQK¾8—رin ªš¸ÒµVT=GŸøúÁùvŠ©¤g@S=ün„Ð`ÍoÈ+ÊqkÙ°RŒ¨ 0jNdÁ¯†â(kʰdüÛ0Qì/Ó¼UÄùlæ«5&ûZ®c´@¤V1r–¶;xÊým¯_J2¥ÎÕ˱±‚3°õÜz©™xïñ‚£T(Mw°oZØ…PY#å'eÁbïÃ$åÇ‹ë´eÁ”Xýï#7š` [Zßo´-Š–Æ•É!ƒ{ct/°“ÏX#( Kjz¼ÑRÕöo¹ñײø;Ί[X‡`Ž?˜$Qy°™Š|UP5Ê'm‡qýÖ/öüJ« ÍbøœÎ¶ýÄo#øã²5ÅE+Ü/}2p²*`Ï«›5òXÖ¶ƒð¶2-W7cŠQÀºF5î¥Ô†™`ìÛZÆ|wÈruB²&Â^n½¶¹q4¥AÏEDÕ[(vÕó¯Ë&–,ž×Oóü=¨d—MïØT ^p&’£öáêÓŒ#üˆâî ¼¼~ÖuŒ²N‰,f‘®]÷Ѹš„~îNÇ7\;ÎU‡%2bH—oós>jÑ¥j‚™$Ø8W|  12Š $£óú $¶ hÒjw•=—37ûPì[ê4õ¸+Üט ~%LÚ1ÎKi¦ TÚ‡›°,j€èqr}§¬™Þ3Æ£(v8lS£WŠ,Ɇ¼.#¿Í²úªÃù‹04­PÁm4•û¦2ÚpK/ÓÐ{r/"ç,è„>·:dçɳhÌ@âƒGþN4É>`þ2frd:õ<ÉYtL/—áÀÖê<Þ…n,@â|\'²Ž™@ôcÐÐ4~`[­C$•z!=|®N*hjRYŒ9X<'8‘1D°j`,uƒCÂÖÎâ²·vDøŸ«›*Tb 2£bö“+hœÞ‰h,þ(±’›³–£5÷`§ Xð„“`“xŽO`µ \ÛqF>® û^’»Ži½ô$uÁ Xãõø?6ÝÛ$ÖvpYÛ %½ÃÔùVS9ûHÆÀ+3„ b4ØøÏÉ!4h(‚(Af5n‹àOpQá}‘qÿ£l½[î"Ò{;¨Ñbls¶?ÃLYг ëø8ÅŠ8Tûçe0Œœ¿s\“X+‚–Ð5zßV®¨eMkޏŸÌ€?LÜà’šMaŠz0üŽS(°œA-Ê~ÛQ°}ë=À¿ñáf¦5¡^È‹­x~§Þ+´(5<®e^¸0m;«ý‹@s Ú‡ˆÈ ËŠÊ©C#O*㌶/â,‹Pž÷fá,W†Ylù¶uÝÎGV¼Šg—C¬ç¨ÑÛÈs[ú=ï1Ê»¡H&|:€N‘^\>SÑð†òoÂyc/³šæÃ2ý»™î‘·ÍÞ˜ ÏÝùk62gôBµÜ÷9{±Ín†i"øh•·0¼“Ù+*àÑ’@Ë­ìðJÂÑ¨È @IƒèW¤_w­ þ£rþµ†ÐÄm†Â´'…0¥Íû¡/¬ƒ·æ…=çZ3à²ax®ÁöòRÎ'˜aÔ'[RæXbXB5$vR¾µ=¾4Gwœt„úß}Ò‘ÝJk¬Z³TGbuYz%ðSqZ©?Ï›gcQiÞ¸†Àˆòö-§¼GÃKÁ‡äÌŠT¬œG¦OçÇ, $†p³Ž¿$ £½ÌdU¾â_A£ÈDiW6M™ÒùïþÅôåZÑ»{Éò*/ñDa¤Âu‹£Ž/¹Äíçžj7ó"8 yÀ’Õ³ƒ>±YÒyª2¯"–­Õš}å’^ß-„òÔô³KÓýí¨Š%(±´õØÝÛïUþd¶¹‚kMG(’ŸO‰Tß?êëdÔ¨–äÓóGJ’e$YvÀÜú«uqû²l î»TŸ~e nr7nQºÊHåÁŒ ñŒKÆ=Ô0ÐMê Í7.4î­ªiöÝòzÖ]pÅ@A`5Éö©ÿ‰8&%ê »¢Žš÷h’t>H¤=½µŒÅÀÕ$íÌ•ê6¥1œÀŸ”º) ŸD«¸÷dšôHíÄIqNµ×¾ÀކzJ¶%+ä8è¤ÞÚÕv!ç[c"ì«EÊË0I·øŽWh üæžAuŠÓ‡Q,$ëPkjÂqQâÙî‚ÎÖ» y^y&Ã}¨‘”6ÜXÙ‡ÈMo5~±²ŠôE[•MÍÞÞ2®UDšÑþ ‹mä¸ h÷×|Ê<¥ô~(øk‰Ù.®Ð…zT¦`xµ½^¬>Ap¥Á}’I”óA›.ÈxàÒ$j¶ 7ø ¼jê P*åµ¶nÁ:¶QüwÍ. .ŽÙ3y`$¿2`í…šuãê(¼gü—³¾yî-‡Í³ˆÅ‘qýaã§0Ç,ÝÓV€£‚… ˜[ÓœÒ$ÍuÛZtô wÕ[pjK—!Õ‹»4˜[ÔÜëmßqÊ(ÿœ#ÖÛmn0èºe:kÌÀüY-úM1{ìÐûèF?.Aí×!àœ‹EqwôXM½œ¬$ÜÀ¥ó¬cÛÂŽ̇ÅM³d-ðuXÇba»É\h(âáðA[8Ùu=R<ƒœNN¼ÀøZ¡¨ë¡þP6û¯Òolü36Í*SÐba6á¨Ý3¨Ÿµ‹ '¸SòÐÐDÖLF˜tÍt'€9|¡WŠIäybëpÒÒAÂC.Ö†°Áé÷êTˆJZL5hs‘÷©É郄&Р2ƒu?aO™›ñcÖÞ Iüä—è^þö$ªuU%ŒàÄ×NÒ!ÑrHÓ£Žƒx‘á̼ ÆŽq’ϲô„o¸KÓ ç¯MåIXZ7ù!b={÷Z¨R«:IëPH$ýɬ»‘4O§ÎEÞi±_Àé¤ÿ»+Ø¢ÝU4yŠ7·¢V½Ü¬ÆŽ(5ŸgîoŽX5ŒãýÆ#È †Çž­ÐÈö8Ò鶪>¿¡qÓwܕۥúÿ;Ä« Ûm¢ùh~-”µ@ÄßhäFÇQ#Lµ™£Ez¨ç·¿9úï§Uì*Ê l‚›ê¾®|„ö2épÐa†!IÄmˆ½Ø¸0ßÙahlF—}ÉÚvŽB´‘µ*]³ “ú*ßõ¨†QBʦ>¼"”KD¹Õîü§}¢Æ–lCÂã´&–ù¥g~ûæ½#Oêçh®²hH$íÓ²­ïZ=û¼eyžÇý+˜´q2nJ`R¼âU¦?½SG=³ºüoq×&ãRùlþ ¤ßó‘‚±;©¾Þ»ƒ‹ÅáâÁôØÉåÃ7¿I¬=Ÿª ˜)®0óArYÒŠQ–SÝÏL²žØžVfp>»–Rk#[.(B#ëW.‰pg Fs–Es€Âˆ îܾ*Çm"£'úÉÿäu¡ŸRjç+ã6‚²ÒçiëŽz¤ï³cÖ:"e:Áç½7λÄ:p”<=ßÀ;²û Úø¶Œs£ÌuÁÞ»éÀS (ÎÊþ¦ÇZMû’I„…0¤ñ ÜN‹œ„}e³ïœ‰Ë–ó¦…´É»Ä6ñù±™ýkˆ‚[f1qÀô§†þ•Ѧƒ•‹O»iy°:¾3ì®Ù â¶|£[Os5¾¹­+]s®IÞl€¦N¾ z >-¤L]QÉ«Lx˜‚ÚCcvé™s9üÕUmç5ðÒzÞ!ë³X‚#UÏBÏ;è™ù¨Û–xЦ´s£Ã0 ™~!âS;C²líB \í„=žŽ|KC¸ºÁÙeLsƒ´ÛÌj*¯|ƒXÊ6HPOßpõñ7ö«Äú#ÖíÊÏ–©ÓôËwIî{¬ ð(=ínâ—cæ`sÉÎþIeˆ­àöÙãò$•ؾ°Âýy2§*[®¹>5hÜáè >´Î]o4Ô;ÖüJâih`à¸èN=º\áPé©Ý–NIfZ!ó¥˜Õ¤¡¯l»>ŽëúÝ;£æµ2-ðȬ[{§ÍóV7v»&d«ÈdÜ ÷’{ù;Iú=õ ²ÛáÜ Ïm3_p…´˜‹H¥ûPæoy¨T®¡lR=½Fdù«šõÉÑÕ°¸²ÊkQ¾±4ÆpÕ˜c5¥÷TË/¿4TêZù&10í®ÿj1Ì{ìóvnÐcÀ"£óh‘Ö; å=øÛh’‚U¼Y5x ‚Ûøc‹G­pêËI¸#TÖ…*ë¬Â´â¤R‘µÞÃûÉðÉç(÷ø»~›$îÁE|·µÜï7tæM¶áƒ|6Ó×K^S‡L`ð¯¨0*Þ¥‹óŠdµT,ò+”Ôc!r^ˆ+hËÑrÔ2$°“3LÛj3óMŠZÊÍ¡ïxLj[ÚÈcœMµÎ(k²À-ahÒ»nu<{ª¸†îÁf’@ßþ AlF7ÎÄ‘œ5 KˆºG8œMnæ"k½*X¾-±Ú-«ú©ÒOÃÏyväÞ^úÀ\-B„õÙabÇF:<×µ“ègã <[-úû¸Õ7©bNÃ3̓š*éeÚËmR[Êá0§C0/òQV_ å‘D¹0W5“–¦Ä©ï6óÎô”ÚÜì|céÐÏ@ó‡gí…OèÁp Ž$ 3äD!pYk;¹ûS&WÜR:ô|HL‘¤Üj¾ /õâ{•†1öÖ¸óÞÌv ÊÎÐ5šÞ×^`ì(üÜä +Íî̱l(b| ”éæj(²<›GáWÄ…Ùr3¨Ìs ㉫³™Kgï±gR]8Ôþ¾t¥‰ªE¦Æ ¾½œTÜÞ}_¿Y-¸E] ¨‘ecÙþ³pœ)ߨ ÔX¬¥’†m¼Y¥ã_¼rË`¥}X¹G0çü€º0p=Œ/çÅ2hZ¡ÆAü’ iŒqd^*NÙ´Ýk+‚6ÈU’†-]Î;$½7¼e;ÃNÞîs“ü§“ã)‚£5wÈöÔä²ÑH-~NÌAmV*N«æ&aØh$öÒ ëpœ^°[XæzØ®AƳ`é[²ªƒÿƒ3’Î&egØkؾRu–—EiŒTiD ޤãò J(ݵ_ÓYTt YPî«z×VI,zB¥M” c˜tÒûìz„gCèi5‹@®T·ˆ¿¶ËJ¹¿í­R"õÀZ¥YŠ:G})8Mµâs£/ÒVŸ©q¿[Ôà“¥;Ý9°|ücnʱ2Mñή ɰL–…´‰P_¨•¯ãa %‹ûž¢èQˆZX|øëzû%í¹'“#r˜ ›KYÓüßfï»U‚±\ˆÏZ+Aez–EHô‰°ÐÿÚ +þ¿Xlöx’ÿŽ9œ,%”× ûé/!‘Wæ8aÓë 4ñqÀ_ø¿ã ²yxK+v<ó ‹¾­©ÞˆõÄA;éz·¾"—ÉÌÓ}3÷¨Q9o»ÙUצɾô¤z¹¬Ñã‡XM,³XD;Ó¡@wÌ^× 6,¢µ5(¥JÓ8›±†qükù¿ä9Uõde—tŸt>_tà2{«9vÑp¥Ês+N”ÞÀmÁàÏ´<±¸uÅð!öl± Ó±…âÉÍGwÂ\ú8ée²·?QØx9nb£DEÆäÖ> ª¸“eÓUzQ'ÿëæ~ëßWVõ†Â_öl! ¼Xú·þ§Z“£œ‹*¼íÌùÑ©ô>\^vÑKyì±½4´bÊù3Žz†ÇLÛêkj^Ñ FÿUÂY'¨"¡Ñ'¯P2,HU¨y"U]G •Ÿ‘ žlÖÄ83¯dÓD#[B2žÔ?¹SR‰y"‘¡½Ø²Ô~|€4øb½hì¡i t™²¿J]²œ2ð1ŒÄ\ÿ:XkX/wcȰ¥Ž–®ÿŠ:êû ̆÷_†Ê Â(› M0TVìÁº–(^'ŠÀݯb.éŸÅ¹qw+¢ˆ:têÈv8Ïnû߈ ò°S2 ùOÖâ”Îw"Ä\mÊwd¦ÙT&fªä Š ª1LœO\µºÒ@¸Cà—¦l¹ ¤©9t«‘;ÔK¼5ÞD°zCÉ.áIÏpzõI¶ãƒlãr³ºk÷"Eåˆù¾+7øy¸l¼+￰ଧØð½ô·€Ïâ 8¨’aä~Óðk wª—r©’Œ“Á[|OA>@…¼×_·× ÌßM@›ì@òù Äî–±ùé$E1÷MüÊ®ÌsU!gVÜ$þnÕ£¾[¡W+ ðɬ„ÊOdžÖ{ù¹¼ÎƒÐtÃJs+ó‰»Þ[á?— fï¨f¡»ÚŽêéà‹„g:´Ä“Ø7D)Ý0ÎÀuû=ÅôcçáX‰!ÈS‡qÅ“™æ9ý*LL0˜ñ¸ŽõÍ#Šå´º˜tÝ\“–îË;šÇ)N(S@T’FÈüøØ[Üa¨E@á›×'ַ׊nÀn¹ãú(ƒiXui÷ëþSqÒ®ÜWBÎ4X¶¿7'£§Ä èÄfÊŽvÎ Äú6|v\.¾Ù𬠵 ú”x`(mjÁ¸…ÌLÅÌõ)Ž„ôº­³ÎÛ«¹KzÎa \kB÷4Æ(§YA–7Vºn`ô*fç/`î°”ÝÇÃÁe›„ÅzÓ aìŸ|€ˆ¡õ¶TÊeºé@x X‰tÞV¬­ oCVvÛqÔ6†h­ÙÄñÉ:7µN˜¼=÷…WÌúdÅ@›ŽÑýˆ+øuQ9Õäëò&j.'º¢<}®€Ì%¢R 7Êø ‰·CÍê÷Ï+˜W4É1Ìùk à!]î{T$Á8º0„Ÿ„Hž™ñ=û˜/¿|ûeÁèv„K¯ŒõH'@H ÈO1ÂP•í ÝökÆmÔùFö) å¦ÅqhÒþ%.¤Àá¡1 äÿXÂê°%<ª‡™¥²= N‡jÐ˲:QƒZè±/+SÛ°RWÐ+¢,%‹³ RT±b:—¸‰f4·ð3o†„;P2¥4+ç™TðĆ»ôf‰ b÷Ìáš®ió¹òC|RŠÜâþÿhwÃw©sý<²ØÏ[ôÌìU¢q§<ò&RÍã"vE<—¯§Ÿ¦Öw lv:  žŒ“qfãœM]¥ÎM¤{H¥ØPm8HÝ›Úó3‡¹±>$*æ E0ˆé]1KÇ„˜è·ªÞ®'-u"›usbByÑ.$¯²¨þAÇ[  ·Î’EWðè9‰9ÊMÿA®ømË•Du=wvñ¿Û$«BéÁ$¹¤}¥]Ò[U’:¢ïº’ò3döÆ88nÍF,ÙôåvÙM¿}FiH2¸óZ]h'üæM4„t„éŽ/Ç`ГäÔ(óùî †ß®¸Ý-)ÁÍÓ<-ä=›õ¼rÏôƒÊÕ"Ý )ÛQH­ ½Å3dåÿâr›ÞZšaãk†!(ä¶±, õäÎ’÷S 8Ô5¯ZT© yåDŽ—ÔÈÀinJâ¦*üø˜ñYÂk/®Mmš7é5ÈÕ:­þ»hû¾d5BùËP°¬œa b$pNRÁçePÔò÷6Ên¾FÅë·w¹ ©Ÿí@ J£OÖJä›}®›Z#ëÌ ‚c‘ôTr`ÝæR#q!}13Å'‰qÙF©nóÞô¶^½”§r2B8àg"åg"ã)— Òˆ¯eMÑyÙíÞ§²}ãsLîñz¨¿èÖÚ†‚-÷BbètA µLö«¯ù.gJ¶ÅøF¿¬~p0à/yE‹éqÒfóù’»ukGdZ"V®šl5h´Hø^2:M]SÅÞî^UAC¡bÿ½ 1 “Xë|…¹Ý<”L1\«¯œûÀ¢1WõZÌñÕàŠ6EL€½Sc`h"Ç @@sgüC¤ô\â9ÕÅÆ§ña›ƒG` ÄJ}>Òl x·­ÿ¢áKA’ •’lΜ‰¦öÅ0X_càäRŸTýiÖœ¥„,)–_Æ]x$ùÁRÍs§\ÅȨd¤ÃœIÑ!ºu ô1ÊpŽ,¹s½}ãõ˜_®ò˜ÙpøJ¹,ïApZ¹eÔLQ м5¢SK˜oˆ,Ì µµ+ì­Ñ ˆ ]—ƒÎ/¡Hí M’õŒ¤E"£(  »ÚÄWáç‰ÕAM!‘9Bu‡ð‰¯˜U  ýHDq`ÒVvÚq”wL~ˆóoû%s!>@š±O[ºñ·ˆzMŠT¹é*bK곟½]ÕÙûÅü\<ÞO¢‘åkÒ¡žb…hÕž}á3o7¥½êíàGldžW"•Á.d•ùšaÆEÿ[ïùvгŸ üÂð€ë2{n{•Ðδ:cüÓ^mcÈ­¹$ÿˆ¼V=ãd/^Œo߃ߡÝNñöäÃ3ùÛÝŸ e:®Áɨ¬«œ )IjRM ”Vç† 1Ýã|þäìéu°c™¦J¹ÊœSê‰XÍz”^S]­™%æ;|XÓ!Ù]–צàÀûE/UL20yÍŽ3®YÚžÄú­¸—üilp’.ì»Ë#¢eÀµè„„HrŸ®W±k³ ‰l"/)°ScïŸnáãv¤šØ`1•Æ’õZÿi†CV%ª8u›ª·¡ÅJë±¥ 7+VÃWÅc –ÈêTP¼¼ÉõŽ©}—!hÍ«ÜmÀVì®äŠyøt©aÁ*UþåÏ7|8:ve[GrÅn¦A¼(ÉïˆjfÊ@óñ¹5Þ7‘9ÞŽýÂ]J°†ô AºÙXÎ:õ`³’Lï_Ëí§Ë ¶õ¢g?Y» ;²så–p4M ‚µs æP4M¡3»¶4º‘–Ò/ \Ó!‡-(j†Ú Gõ] E\hÊ„¬ÑžZ™KógͮӤ\5CÌ—ð{¹‚z~•?Iî·Á¤-æÒŸJçCtìuÑëäˆ<»øžâ,×yÕ‡¹0ì^ø6ºœ?ÄnÙ\éÚf󿽆Ah'‘,rCDqŸ/—'›æSg\`º†Óø‰öô(¥4Lõ)½ŠŒØáM̃ˆØš-¼çÚP„Å>$PœLÂu °5'9ì× DðÞwÞáhµfüTç5Ë× vž²\y4&#t⊋Æw¼…Í‚Ø!/$· ªKñåbX«o·=ò`ìǬµq&q«˜ØæÌÉë\ÃB“ €¾^mùóººKaw•£f­oa“)á—)÷x£‚äp”Ã÷)`&½AhØã´o_Šôµ§' à!9½ê§cP9ö!ˆ¡Âj”þîJH%®›j~“F"CŒ-Fâ‚zFžsÃ0{ÖZé,¿\¬5ÆQ ßÔ/n‰ébøŽëqÕ¶’–³¾Öø›¹oS® J¤PþǪtȧElö€y»¼S„'áö¿„º¤)É¥>òMÖr?U4¸‰Å䑸|Â3¨Z‹°¹©D DHu£Â}T\sœõ]o'îO’È\5î„ZòÛª A…Çš¡YÖ7K¹bD”•Œî’Ž ¿ÉKw#t’ª Lg»±rÓcÄŒü{pDw©GÃ)35Yšv@ŽíM’ÚƒmHÈqZ4lˆtÍÈá–rÞýºKø«vEuÄײ¡z_œzö®HD[êRU!TŸ® Ýdñ]V¡nÖ£§Ø†nR O0@ëµ7¹Ï·Ý‡Ù™}ãS˜Ä»¯¯ÿP4i‘ßÁ#i™’.(•å)]zg1ÆÎPb'Ý¥–ÄÙZÞjóW%E7Û‰ ‰LwôŒj§?º8 K’k À9‚Þ죵+ˆá{Ù³ƒ»¶`ð­–Rè¬}ã)‰óüºØ£øq½£|ÞËdbJZ,5ÛÓéSý™r(H€Ìˆùüù¯¾²ÄEm>Èä6o§È¡så†YJµÃÀH©½TÄÖùœ°¬2™Ñg(Å–£2“Qæòîsäð)ûv;¨œE€ ªöÍne{¼È¾N÷JX¿0ê ËÈîò¼6I/¡û/Š­. ž£èm?OV¨ÔÍÊ3Üž¬¢m~Üâqn0x*5—®]4_ç’ÑsëC®SŒ!{»4*¹‘ÙÐâÓnwÆ<³dQ9æ¦>æÉïn%ÀçC{™m¿{îŒ\›88Ÿ}J<²ªœ©¬=ö-Ü¥s©Å/·=MfåkáHÙÉ}*ê˜ÈRáŒ@¾Ôû§]ùd°†eosKÌ™i(”β{ä=åÞp3µ\'{…L׉\Ghp—ÎÙÿêÛTÌÎÁ@Z¸°,!èHÁóé¡ü¢°Ž# |VC'3úÇZÄ…èĉ~;±gñ BEÊä“TÔ;ÒÍÔ®þHgsÒ ø¦«cýŠð'ŽDÎõL¯‹ãÿ(Ø æÃhú9SD.¸ét“Ƽ¶‚¯­áë¥g”³?oí$Ø6Ã»ë‹Æ¨ žx*ŒN‹÷Ãb1M‘©Õø^q îB` 1ªÌ¸3‘Š7%ÔO Y€ÌA©ù1[×¼£N¯ VF¾Ò¸&`˜8âš“vÖÆW %€W±G–Ànq€ȘÔ㙫3AÓ0ݪœM>Çcß$PÔ¶ëËw‘¦*¯-•OãÕnÅÃ]Óæw(;î¤'i¡TÓC1ÍËXŠÅžw_=vw3ñA±B÷¬)‘"U‚.:è+²2xŸväYöqx‘2]¶t†Mp¬ï¢Lhzə㭠Â]° ú“eÛì#˜q6܉Îë^&vßš·ÓDUªH’}ÃßñB©›ÔðM×õàTl±ß}ÑTak¥wh`XØÖÄ»UiRa"µ”¥#Õ½ûRçY#½‚ý/òŸN™¤5ãJmK°y«³*©KçqtÕì/Üê²dm,ÎÚË%n¼8 ÞV‹Ùåª+ˆYnqmÀFåžïï{” |KÏ„¨¢gŽíX¶)É}^î‚)JXæÛ„;g:YP:¢Œq†fš+Ëìe¶f¦Œ ’—*gY•J6†PS5·¨Ðµxò$ xÛ½ÛyݹÖ}"pÅW´›MVöõ/ÌmõWº£Ý`E n¬Ùæg¼“„kt PÜ&c¬sŽ}rË/Ęâ©Qvù%6kÁN•^Ž éªøzsZ sd¿©ZÎLÛ¾Ùe:& 3A—¢….LTˆD=‹ËöQÁ½=—ž#çeªf7Í-‘w+£¨õÎÖІž;ž¢ú¯]ÌgDA‚T#\h &6í¥ã‘ád- ÁáÍ>(—\°*¯% ÙS® VÁÝ7uõ›Ï CûöA’ÝO/œÍ"ŒlÕ2<¥Œ]F 8Ó=ᔯSBö–"êVFX‹g°naM" ósn¼çJ1½BJF¸ýÜPcŒ;aXn£Œ¨ÌGék{ ?Yètvd+…Céˆø·¤ÑÑ eºflvÃñì=}~ðõ_ê%6›ø»7¤– ¤4žçhµ¼øÓA­3v&¿ñ«n/ÓDܱ’I7~-yýôÉçÉì ÀφxÈRã®Çæ' í5…ÑÙg¹aú£p¡šÍ‰¼w±UÕM&‹ãƒ}Rð×)åXþ ßú½”¦0ÅSiý‰R~ AC)Tˆ0I۽Ϙ­¼³H4ƒCk2>Pôïn`Ó1{2רm(HiO=Å>Ÿ’qiõóÛŒÓøé.€2×ö“9²HʤÍÙ $Â`Çe|6!üù¦ú$Ðõ$_ä´ïñ7:ƒ™{Ïb^â|d)GK §…Êá_•KªÄ¡»þ.Ê®‹rÎ/+hü¤!ù3|0„¼X-ñ\†æwÚ¬–Û–˜cC²S¼”áCõvüïIzèNëÒ5<âÈ!B¤e&ˆd…'7ÂäãIç#?RfýZ €ðÝÓhÒ§ÒyüLñL«©R#óðªiYûC5¿Ì˜­d@}óÔyƲCv)ã¦ð¨WÚMªcfç Ö‹À·m&ˆ++í=°c®ÁjßàH[ô†=(Ò³+þ2Ø) õ˜QSðaDO…CúWÅÞìv#ŽÀKX?VñúB¤I.G2’3)ŸåÆú:^ š3j/¶õÞ-´(ʃL¦÷>oˆ;U”tòé(1Ä/ÒN5æŸ|ÑZ4 óé¤P혴úYhsR¿x ¸¸ëIZȳÕlò"U‹óéš5Ó!TÞFN~—M¤d`ÌÑ‚¢‰9Ñ+%àè´ªZT ?Ûk%Йh3 ©©#èÿ?¶IäJO´agŠi®}ɦ³éšL8à–åñëáá£jó(ÐQ:Êú¯­%há>imÁë5mU–‹aï›O'8n‚øw½Ó ~$>Ê]p“ûb5DÐsûgë7Ñÿ«pÕë0w·Ÿ „‚À>®Hq¶¶.¾(Ú,‹V¥HPžä2$ñÈ Ìú‰¿¬SøÒñ%l„0AÊ® ¼–¾÷^÷ñ‰ëÑE¶Ã5ŸD¥íw²Z`twŸ Í¿Ôø·(/Y ¸Þ4PN«cÕ³0 ²×Q›W˜âq8jf½|òõn´»}½$ydÁYë§1k¦ðdæt‘ýNÐ{‘¨å-+¼¤Uï `c)ò]^«c¹qœõ“̾X‡¥¶;´zeg`NŽô ‡¼,‘ud âTõ­ ¡Ü_ôYá\hÿäÂñÍ3Ú&ájý "ÀPÞM mtH0ºÂ‰µ»ú÷Å"ÿŽà[iŒoÌÞ19¨À|úaH€‹0Þš ¿šl·ŽÇ\ª|(³Ù”ñÕF1ÞH«)¬òÃ!%_é¡>H+víX÷•Û½«Â´ÃYêêÜ:6am‡W‡0ÎNán­VŸ2;o¯ÑWª#¡Ívu!,묲7ˬ~H¶-bnÀEvòØæ•LpS{s©‚g @Å‘tr}²hëãTDÕõ‡œËÕ¨©µ>~mKuó)á9Þ¿6oÁ’*±2þ,=IìY¾x,a()ãþ”ºôSΛBDƒk:…+w©\ºwFZ“Dn@YSLÈ¢ÎÄE®ïñ|b£NRꮇ£‰É7òpרƒËnã0 LvfŠÀ/H‰‰x¡"Ø~™Ï¡nfú 3ËU~Æð±—·ª£vs®Ošu²£úžf]v‰š‚…þQ(?vÁïI°¼&«ÈÜ‚ðŠ%lâlŸ5—Cø÷£tØq ‰,JV½’÷« í4pÅY»Û‹zKë1Ál·•}R°…•!ë‹p`Ÿ™åÓsÍôp꣟-O¢'Rk×p ¶¢^®žOW î_¨`¦¢q2®{Ð Á µYw™5l¤™ñ Äu¼Ò<†ü˜a{š‹ý?èÊœþ’ŠèÌKAF§/0Ê~ã„Uÿ÷÷mÍÌW q ž dîÔçx‹ƒ(1a BuÈê-ÌeTŸ±zðqÖh&ooÎöjÚRSø=Ÿ@cRYïý]Gï,ü¶(+ß;×dÄ_ý`  xnm5\0àýxÞOâÆR<|ŒžGs#8Ä«²òMÕ-hYhV(& ÀVÓ§²ìÈŒl¸vÿë‡mK=׸$uj‚ðNãöÐØ]D%8ç’Óãq¼\aöJÀm<¡õûqý<öí-#¶7Rý²}¼`™,nñ;ø>×ãȶÅL&Úàµ'F¿õ¡ŸrcMX}ް»L4 ;Çh«ñ|OI²ý&A‰F’ÖÆÊÆôÉI³½bëâ ”W<ÓL„áשׂmxö3xñ ˆ“@´§CÑКÚw‹Ñ Éñžáœ òÝk†ÓÜÕQì1Exÿ´)ñL,;¶5Ülj¯ãöÕ_tů»Ëñ‡²C³ßOóáÕDÑÿꈺ÷4hWLá,±|¿:»cÞa~½(h*ɽsÔÒŒ»Wx½#,N°23¦jæ¬|ž´¡ù_öÚ¦ªUÉ=!û+öËÒ`šû iwGÜLœö4+žËœ`aç_ð¤ðìšlFÓÀhUc€Ä·WG‚ßr¯Þ*“³±è˜\¾žIÆ[u­œß‹IËãê6©XS‡U%”‹~w –D˜•Yûª– L«òà!®}ò¯ë¯®ùôHIo¢j’Oç]Ù$çæHÈO0ËŒo‹¡1‚ S þSò{çàï,¾ÂN è1sçÈ‚˜ÌI)Q†jŒ˜“Îx¯¸ÛH Ðç‘iöÇlBà®ØÊy]-^K¬#P‘ÁŽÅ朜HzkáØ8$o+¿*çP‚ù¹SÂÎo,b8c˜ÿŠú6ÖÇ>®Y†;Å]0÷⊀¸ îÇã3ôÆ`§ƒ£TS<5‡º(g™—éîå–uxÕÉÕˆ÷SŽtƒ+YÓňˆàe»0ÔÀ`’Y—¢ 3¼)Ô©F˜&ïpÆ[9„åë0ÔštØ©™@Âôdç‚gçÆ´šI(ZöV!ØaŠ#<$zÏûæŠ3Ù ô“«nJ;¢ $â_÷L¨ì¼¡¦ i]þÚ¥09¯çe’¨InŽj"0w%v#ÀG£|Q_Ë @Q]h6ÁSu=•P½]Nóá…ÜÖ‚•fI—v¤èW5wwTh°m® …ÆÖú©oqÿC&ÍÌ\C{9”¼âèàßZÂÑæ2ù/t'ËíB³ p—xjËr®‚Y.?X¦(ï P*M`§@RE¡îœ—>…}ìp²±ëF—ÛÎNÉ…M;lµM—â¤\««ßÉý'suË.)Á YÅf_$Í¿m§4|3¸]m+oƒK‘5Híd³íÑ,>Uøw»|BvÞ˜â҂Гô„¦Í…¯]–Ж½ºž$kZ/ä[¹6U8Ôïòè*úGˆ¿ªT&Xç%• ´É,=+¶t—sVú)Í—+G¦È#àR4¬?I8ñKIK´¿F@«èÖïJéÖ©cë±G{bg¹é ½*:縭x´|/@õæs‰¼[B˜±ÚÓn/UnA¢;ïe8§fÄ„˜¶tK„wÇb’\°–2èlÚ¸§µ©[,åþ£ëJz,îjcñï¾€P‰¾z>]7œCùÇ´+Ñ@˜N÷ Ô;»*¹n¹L¢Ånrƒ%Á®:qÉîù»ÝD2/9©êGo@žh0ØÊù¸óáhW“‘ÿ:èîÙZA·¬ïÒOæ·F¦¯T@PT‚àTnB!a¯Ô'Áº˜€1Š; L:$>]¹OhŽ©ßä¹35zM;&Þ‰«Ó€¶ô;!?ØTpþ6Õf²G(:~ cð—ÜîYñYb$eˆ4†ê¼.f Â(kÄèÐQ"¹`3ŽþLM¥ÒL”rdÿ>­Û/7 Nj¨r$åN}'uA´>B“wÑcGšïZ@•º§’<®&èVÊñT>ï÷”ygÛÒ 7beØßdÐqzy¾:r¾ÓfÜÎÛ×ýu÷§͹/™)”¢M¾r@¯°Bxã_ŽÍ/Rç®;fLÀ cý/Eóï3 ire‡÷ &áþÚ¾EXÇ©…Û6„!7d‚1<è)Ug¡&ΰç­:$vné<5êì‹€#ÃH¿_ºgç.>t™ýa« Ž&_%ðe4ÛBj6†‹¬v”…e8œIáê1yB|so÷7=w§(âc0h-Š îx‰í¥íÒËŒZô‚‚¶né~M‡/Ô[wHý–ÜÚ€…xdHå¢Å÷vGñöT½”IR0¶É‡mW)mf5*£*O>buÜG–óÌ„¤ÄÏ”'z>yáý¢to¡Yš÷"¿ªÚ~¡á¸xGuiƒ—N1Îá]Ë©`i 8Ê› ä >£å½ˆÐ`oÀ9ÙxÆŽ4›¤8 4!¢r/ƒÿ᩹åztÈès~/ˆŒàâ’e?ásò[àèû›Ô|–h™VžµÑWÍþZ ¿tÙ$èZqø5“$ò—áZ}¼Ra à°ü‰(#“sµ"N"Ö“OÅËÊÔ]¦à^¥'dz¬áà;ÔQý+=òx7gÙ&‹•¡žJÜN¿T®7ýˆÝ6;ºEÿ~Úç¬ùdÍdzŸr7Õ:‹¿ûï©Ò˜æÂdè`°ÂÝiÑ:5Ël-ûLÄá„«ú]Fæy´§D}i,ÍÜF#¹Ëk‰”H¦¿|óåñ*þHÌ* °^Z¦Ã„¦àjpù`l¾//öK‚Ù¶º'ßæ‡b±¼¯¨*e|ÎHÕ[T| (?çSç—WX'Ä(ÅD™Øm¹>ÜWÏOx$MKFãRNèÑâ¬MÏr «Rý0s¬|>TY´>JÊ­nûd§éQ="S½do1­x8äBÞt *ÔÒº‡$\]Y@ ÆÌɃá¤( ]°9Šó¦æ­Ì­½Ñ®SÏy$³uIùÐ&šôVu$ß~ü¤u³O¤$ÒþâwΠœ<¯~¯ž:@q†SgÕK«_&§Lc’VpËO”îZmÏ© Õ•rÙ%*•f,á-Ò•ýt<–i øâÓj®I¼ ³­Máaà&S9+Eh!\¾¬¿ÔuÊY'ƒfœÄò³äÈmW•ü®xOìzëõQy°ø9ˆ1EÚV‰khƒ/ŽzzŒSÃFK|ã8úxUÞÎ;Èß–>A”ðãßÚÙ‘FÌ™ÇJĈö$ç ã¶C¤tìŒJ?¤?ì.ðÎ=tC ]ùÙûj‚‰M8ø-™C{(’µV8 ~·G)°({×3îèk9ð¨¾oä–Þ-¨7±+ÆuH}a wQþ¹4›«œŒÈšSbÔFè;É"'ù-ðÉUÛ@ð óä$ëêýü ˜·#§<'èÞí]Hpû*ƒeÜ6Qº…®Qç Äݱ|‡§cÞQJY‘öÕô’(¦ZRÒs$ :w¹eb^9}3J,:ð¹ 6¦töÈÛó0’N¾‘ž§°$,Yf kmg6g¡xº~q•ã7 ­‘šL›¥n8±W¿~+‡2‰‚enï;x)?;kv¨~ÒÄÏï‡<±»é„±¼h¨EÕ,²U‘‡óØÆþ3ϵB†£êP4d ‚–™ƒ%¾odÊXQ ¾#vnæ1-ýþp·y×çc²hmKM Ú…•³3~0¾l2 ˆu¤0ñi Îûö Áî™Å¦òσDp ¯ÑÜ6•êÖ<R ñÔãû͘°±4Ðÿ¼ö%*ÿu.L^£Šî¶Â^S×>üO;¢¯s>¼Aé ¸Æ¼%™ömX‰6ð§ ©7?ït’oêBèS~0/9[>lr[²zŦŒ$âTlg–òÀ’¨t:ØÏ äJ?®À‹¶ÿ5b—oÊ®¥(pÊfÒÂü}Œl¿W Q#Õ›=)žgíor(»lvJIZmÙ£` t7Ò¸¡;dwbŠÀhÇ$£ß@gÖ'~5lð-6‚‘Â`Æû•éEßÒ¾ë!CÖ¡¿ãc×ëì‰÷0÷.À³9 IX#ˆå–‰ÿâ 3£Ê½z†Åh¢­öxÃ\ª1½vžÆ ²Å%€êz½#†4 m)’ƒW¼{Œtj »DS¤#*¿[)„äëÐû©M…&µM* Ê¥CïJ‹’yGR¬?W+YBɶ€LFÊɰ±¶ƒU¥O§a×V‹ž¶sõ"£R@‚÷‡[õ’µHÏÃ<$Þ^òõéqmþ¨÷²Y‹4‚ä¸qþÜ 2W¿×4’#;¼ZÏ*°Àú(íˆB ¤ëê0täϬ{ýz™.Øl%Èøqè g%’w†ƒòÇ­ùÃèýpø$ÕH{LïX#¾`˜ùO\‡r ¬‹ÞÐmçÊ uöZñ8ô‚ðÚÝù»ð$Æ¿$Vt/§Ä2\ý”p…b±zòûû`,ÿ_°àbÞ%CÎtœ ã ¿ôK¿Q»)/º-ïtç²ü€ûE÷Lµ`~ýo¥‘Œ-OæÜ¶R®ÏçйÓêšîXXcþý;‹„v›àU@á1NŒI€Ekö$Yþ¬ Ž×‡á–C³FY7Œ‰¡ï?ÐÉ ‰"ªžnUTÚ_;°½Ôž}3ùšŠºÒ=üŒXÜû Ÿª$¦»‹èÿÞxzÿã>9;Áá(7{Æ8jˆòé¼Û5èÎÕïÌç:76È^+aâƒû…“<·;ègÝ{Öµ´U~ë~Ò¸¿ë[ÙPܘº>p?¯F¦Ñ)I§ê„qëó`­Ëì¥A}Ë%,«a4—ˆRD¨ì±zÿŒ gįàƒnºøÃéa¡GÍ^ÛŠk ™\ñ 1¡#”YÝt{px»{¶lQeÉ­æk1ˆ}7AÔNѸç&ñ¸†Ä·ãºÚBuB¯»c]5¨Ÿß÷jBÀ€¬w-—‚y€»ó 01Ý-ÇOè)hy,Ùöhþ øh·a¯l.21ó¹¯cµ½"Žf…þ.X5ÐÁ¶:vƒãƒ„¸^³³A^Ëi€™ñ{qà.—™™À¹’DÖ vÔ~•Å[jÏ—†ÝØ¢± è%ò¢<è$£Gõ­¤9¶ÝÿÅàb Qï’­:ïß÷ª+| ÅÕÆ¥à9pjo·¶ô—#ö~: 7ÿ  õ4 æÃ$Hiå£R1¯fù^ ðàøIžzç¢bE;(‚*´@Ì›¬ý©¸`ÆÚ¬m<âŸÌê j°æý¶­¤Xwr@e³“Ùµ|ÊûjëÄ÷oTÉj[U5'Ö÷ð.s{|~³LÏÐU—m (T·¡ú¼xo1 ' æUgôrÓ|܉ù~<ùÔâ…ÎG-×m÷T‹ê³ò¯i¢;Ù’,Ë68ãwéªBñ"¶tSPÓ³]¡qöÒ 8ÀœÒÀ2GO~¨GÜõEÙzôíLKE›‰NµÖ¸iS_h\ÅÔ5Ÿç¢c£½9;ÝÉ÷›À¦\¯òuŸ’™Igeµ£È É!¢²dYß]c³“¾j}bß'ýYjN¤£³¡ä·W¨ŽC(Áí1pÐ5#þšu&X¤ðŸ´=á!Äu‡'jóZ²…+øpwåN±H'§ÜB[îžS–J{»óÀìE—»_# {í ëò”pVó…¿œf„|ZÑ›ópQ“/ȦízÀc%¶I„WZuET6µ ›¦ÀpOˆ¼r†“þ¿¦£ý‹[n]€ö$Fd¬aœÐÏ†ÊÆß¸+?‚ä’NÅ·§©ØòÑ.§¡Ç)‚m®ÆHEùºwŸê¯…KjRè·[q»AÉ’‰æÜýÆêùs7–Í%s¿ª{®‘ÛZs öÍ RÑD¶êŸô-Jã罓­@\^ð€–¬ÓÍçèo§ê·¥«0Ùh¡YÀbüÊü\Pw4NÁŠÑ9CËy5wcꟴ×£E\ŽÂF³kÀŽÌŸ•¨<(±G œý$'zª¼ú•,ànuu_,© xªí»x—T>S¤ oŒ…‚XŸmÄ}WòãUX²ÈÄt7–O D‚hCwáTÄaé<_[¾´¿‚ÀÕ³ûÖ/#l䊵 I¯µõ›Áuºþ)ÛÚTxɮþÛ9L‡zÇêpÄuŽm^þ–Ú™WÛæ˜Mxla" 7ÐÃIªZŽŽ|~U^œƒêd>yûƒMÄ á vi‚Î/hO½°YÆr¯ÉÄð8 éEïÆòï[?òxžÆo†€O®G¼ òÍfúMâ~{ß\ɉ]ÛÆÇT–ÊFžTƒnÒp—˜¤µ6ÌhÔçÎÊ=Í•Ö;†$@¥ê½¾¾ŸOæÕmÏd¥,lÐ]Ë~þÂÇïhJuH/¶MF;ýMÔ”ê(Q£Ñ.XøŸb-¥÷“*ÜļÙ*ó‚ å×ôN˜ÆüžN]L¶*¨‹Ké}ì<À\ä«°Nާ5“O›±8hk9ÛŸð³Ͳã2§©¦wÅZ处éîëInáªý¦nŽhç%ÇxÓÉ÷ÜØcŽO—·IfFJµøÒ}—êBÌ”nÔÙm×Úà®R;|NÈ vGàù’I“VêQ÷º*“ÇžÑ/Ú¦hòeq2ºù|°óº¬«åM"='GÑ^ÜžŸÿÎ.Ú¢‰@Þ×3­h*“~7l÷öuÕ–žÖÚTwÞ™8‚/ö!ŠKš¥8&å¤ä¼G ‹O± ñ̼'qÂÃÉ£ƒõí_tŽzx«qu\ï3<]9€HÐEbšDëà±J`(ÔøPÒ­¤…ðNßÓ*9…ûœ/s‘$ʥˬF‡¿©"Ë80uò§zúvvQñ8§î·NmÚ°/I̽¥¡Te¦ôxÁX3¾ðcÙûñWô¤vMÜ„½9Ÿ×b º¸Õ÷‚m¡“†¡Ç艺 Ðó-Uº-!=RDùÆ*:¹dՖŲÛúq ЫX4ƪK5Aòa#™9´9¯ò4jn ãÄTãÄðQÆ  föKWPBQ¦O˜‡¤Çátöbò3ëó† ;ɳF1>ÒrÃy¨^ÖÜzk% Ca*’pQ¹¯.ÚÀG  ù®Ú+™Ål碲­ÿÂÚ{!q${åóàÄÄ.¿¤?‹NmŸä®ú·_œù,ŸÕLãöà”û4ŽÑø|ïÐhûªóÓ‘uV“¢nÂéÂNÌÚäAZ±ò`í‡;¨¤*×Làßãèeßü!Ññx´_cÓ}ºµ—þÕa™×rE©j:Û|d›L âb§ü¯ãþ”àfîW¥=ÍYbÄÏvÌW¦X_§ ùóŽ·mü·’ì­S˜÷dÛc¨µåýÄ9€>›  q9{>$twá%úÝn0TÄíÉy:.è¸oÒ4cg9Ú7¡ê)$æÔkšw~»#Ÿú™$pvZÏä0ÉÊ,âs;à…+¡Ì`áeºz@¦‚aù0˜2HŠU0s¢Î‰Ž[‡ä¤ ò‡ ¥<² t[!ЕŠ;o3¸tfq VöUÖÝ$·Kù'ž߃…¬ÐŠzœýv¬aVѳC¿ôÔæÿ¡R¿}”¾béÁmCf"¹„`°—áÖ+<„ѹWZ•!=ŸG ÝRbjíŸÚïöß…ýámÐÄu¨æ¥i.ßf¿È!Ä*)íš¹j*jbSú9[眂¢gqÔZïænZ`š,‰¬äbÑ(Ó7HŠ û´§žÎaÐ?µ¼o¤…nN‰9­lù‡]LÒ›rpÖâÓ^4¿dÙR6sS«ì¥¨"ßÎo&®LÖ¬«N79§X‹ág þ6ü/¿ÔI¯;=›&H\Yd€&i™Dk,V¦I­ÜZ½U†ÒNÿ“­ ¢ßÞ‚éͬŽ´îÓ£Re*³D1®qÌÞQ32ÉÄ]÷rÙ›As#^(”`—Æ—>-(¼ íˆæÒtò ã«hÙ&™RÇYò = |¹“6ç÷2°+ðÔ>¯«l¯ÉIŒþèxŠ,­+ª§BxH¤¶áݳɠ=™ \ÙOÖïÆÖzŽKãhx7›ÎŠªÖª÷ˆ&Utˆ ðÄx–Ëä*êuÒËÐËÍ®‘õ¬£öè„2Q{øI pf›ûä_¦$á:¸†Ö—ˆ[î}žC†áÛSÂ…0»5›ª1îöë¬Dz#ºùC"`·¼8µ%""“„GþrJÔ(0Ïð¶›<ïLN”—¶»’:yìVMMšöîœ%ûÕÝo)fõDŒVÆz]=ÐY\‘ h¤Ñ^vG—ÏÒ¢¡«íõ¿ì¨3 þáß`P‘ k²Š`©-¤.ÄzýÈÑ ’Q—6Zí“­ìündèΦñƒP€ÛóQƒëÞ,“åû:±.!Ó] ‘½Mëž>àñPq\ÆäÜöS}]C)£§ÏE9ˆ^8õ¿³ÝVAmÊdÍ㢔ӟà1ÂyË,ÁiW½¨.ôÞÜyâ|ñª­ï1På`C„gtepkŠmí9#b²þ‘ -÷bV=龜ÂÔ²ü`p4g"ÃÕ`[Þ„Y|Ì¥èu'K¬ëæHRþDJ$·¶iEÖ†$`ŠÎÅjMõG°~²þ¿™1¬ý*‘ÌFÊ;m~Çþãò‰Gé4‰@¡É ŸfM¿}å ÂÚoñÓÊ–Õf+7D­Tö2–ià{«s€ŽÉ3äëyFz¼sXY¿dÎfûdÅQÒà™Ù‰Äç}ìįìoŒS?±\å_´½ŸAv’ÌКþ÷›C Cbw3&u5v§¯B%àä`滟Ü?—ªã€±cÕ‘·Äñß/O0ej’¯Xi‘ ý—ăª£²¤X‚è7Gøn{_å¡ë¸pázŸz ùy9Óí]åÂ!ÈÖ9¸º}ÞÙ~.úFÐk‚ŽiáþJª&p‡¿ÎøÌ íûÎ4ás³^nòZÕVg`}üg®u&ëR()×¾Þ„NÞÊг£¦.~‰Î œðb U®‰K¨¼·hë̘€ôÇQ_e¯ˆÓö•ç:R Õ¸ZØîß•Zað¬ÏKϱ5f¸%Þ- ¦l¥¬.B?RÊ&¤ÈÁ:VÝÌðÍzbH`l{¡<7õŒ 0Ò;S‘်µÔa.ë@P¶žo!›ù*QVêWÛ6¸Îÿ"oà‹È¥;_ýä"·ì5%â•&š|AwÆUéûÝŽh9Eúžä¾'½Æ ÃÐ 9‡"n¤ÝÈú›#ÈOÝ©+'mÀïÐÑoâL¢ç·9qc™`Xtq©¸5¯IÎõól§3׎¯)Î D,²Ê=œ"J@í›ù$;ˆ(>RÕÜj>Ó }ŠEÖ ©v§‰l&—iDÍÙß¾”ÑÜE„òÇ.àžZµÉaØ— &\YžZ..¯ý T[ȶr°Fª„ÁXéÙ°×WÙ_f õÛ6U$GE)„³’9[ ‚|’þD}Ñtáš”èoÞúô”ÿuü¹“eÛBOL‡ð´×D VYû+¢‡Ú€*ÏBº6ñ« %E“uFn–£±=à|¥lho@7f#ñ£+e‘½•Hê^2´;ˆÄÖŽïc@PÝÞ¨ö˜±%¨­·Ÿ&e¼†IpQ ßÝá0®|?³©ißíTQL.ÿˆ4ªË½Á÷ H`€ùðül<­æ@ñ‡=w‡fš<åj+J4Jgõ ’ÝH^Ý25UJ𷢉¹-}³úîf¬Œƒë™ùIR‡QÖK©Îž?«1*V˜¾Hç­ò¢ÈÚ)†0»Ô^¯ÁM¬~Å}«¼µ1&² ýò/åóG´–Oufüè[vÏ•D÷×_Å&O[ѳYñÛ~娺ç í–šyÍéô¸8¶&Š5Ž`Öߟ6¤·r2ûkj9Üp=) ªÏÔyÂÚå÷BW§!­7Åøc+Ó]—Ò*§\(ôIx—M±¥2Á¦íæ¶ÊGœÐdwDDݱæXÄ•Ï`Ê^<2ñ#h Y±÷NÃÉ™/ ¨¤¤º×—\«ô;ÂG¥iGq@õÖèÞŽÈi‹¶QøZ„=xÞêÍÿFz ÿÐ¥âý'·ÑÒ•Â œ¼ù ÐIät!z >þºAy÷÷xú&G|(äLT”VDI[.»“IªÛ Ä m†Ñ®ÞåPŸ%ö‹Î棾±Ü@ew&vxÖ…#LÔƒUðÂúÏ(ûÒdÌ^t¾F¨ó£§Øúj-W=d E³­:ùൂô©©4qáÓrl (4@® ÏŒZ¨ÔW8­g£aÙíwõ71¨¢BZíÞ“oÆÚã¿WÀÕø]³t´"#ÌkG‚™d!@dÝÚ>:ÞΑÐðNªÙ”wÙã5$9Äšº˜ýü‰\À‹{(»ú-iÈEäï2lI»vC:4ãÜ+ŸÄ@¸ÀJÍ·ßñÄÓõÕ#K÷64³Š +á”ü“¡o¿H çn£-.YìåÏ/ÊT¾ÈH„LÖ¥ÄêXáÄV-~9ö· ¦™Æ”nì.p3’"¬_ÐOŸr[N1ªwµH–Ò#FFJÑü)\³)㸭4AÒ‚DEöD¨×uMu~ð¾gvý´ˆñˆ ‹Ió`ƒƒoɈ ðFŠ&.©§õŸš‰Ž³1ŸÿKÖ•¦0­­ûB?¹1/ÄFâÀUËŽG‘JhÓO“b r)dÂKÇ¡UÕ  6 üyb—œ&@Õ8+—§pÝ©¨•z(³;EtZ"ÙâÑpÞ®ÌÿÂöSýÝHºg¥p=éƒr©Ògöœ§ƒüj[ãâÇï²in:—:àžÑ²{÷ȫޯ,˾ŽWÛñVv ‹…Q@œÈ^Í©}’J¼Â[­$qÃØ¦nAÙuL³YÜçúž|tÍEþ4hb1ø+Ö¿Õ!@Œ,Nˆ™j“$‰´ @f!c\³¬eNæ¼ýHN[Ú˦Æù÷Ñ<âu.öÓ©à‘;“kšÖ0…#S$EêÆê¯¬<›d¤b=ïyh…Õ„'4pÚd’¶°rV•»‹ÀGe .F!ýwÞIªºnÏ,•tÂíù´ª/?Á÷:ÝKŽ«4f\Ø_—E®•2±Ì–ó[N˱ô2Û•&ÎÑFÞÿ=0øõ”?ªe7'ljÌñfÀà*¥6Ê%OQfƒ÷”Uåhîp+G•Ù=9«B$íïë—n¶º°ÒbˆK{N£™áHnfq%N³°§5~òyÖ|n(⻋ŽmĪõš´ÏuuòqÛOØAe53À˜-!‘ßb+¶úƒïfÝ’7½<¯É$HIæÐüx°YXDùœokƃ¿ô¨Ö†¯s¬½©N\P¸½@6¾z¶‰un§ûÕœr‹äUÈ*Wbò¹ª{Òm.ëå®þw\Âó²)¶Óu$‚á»0žÜ,朘ӮûnW\rûÌxGvŸ¡²ƒ'D`v\A¥à#RÐF^7þ+ $7'¼Äpýo=w·¤äÇŸÓë¼h/OaBÎ šPG~ÚÒ8U½0äöEÈwÈoò„­`ºìN¸÷h‰àx% 1 ôP`ë0~~.^‡Ö깋¿¬†/J<ʨ¢LÅ)+€Ì‡ê;4Rü>Ý-ÿÙ<^¹¸ÃSÑÿ{8˜qVì¤^Ý…scDY+xÕÔm ·vÒPPy&3+Áþ®]Š4Ù œc˜ÝŠ >¼äËV$«É’n9µ a6F /Á¨Zç/÷5ä¬Ý#ƒæ $^(« H5fZÐò)Úøúñg™ fbš»×,jWÔA‹–2Wó(/p~îú3ß=‚´ºLµ½ãTÙ kG_Yæ·–m_±©úJp‡ êÕñÜ{Ö¿“30 89nè‰â×)cåÌ•»aá&©G‘F‰÷±¤Øè„Ï®†õJ“šoj„Ãlü¥‹Ýo¼/¿feo hDÉAü4{&L–6‹t¬ò¥=TIG’Õ×£fª#ëÈÚú5Ùi—nTÍl9^<;èN°ƒµÏ8µ…Bã3`YÓ ´v@ÜpZýbVð/—7Z–]1šŸŸñ¶ý¹»÷_½@ €œ=0ÖEC^ðós±]ReW ¡óÓmeú+àœ£·W}¢¿·? dïË~dƒé½7LÕ¦;{í B sÔ.OÓ™Ÿ¼~Ä9$^­ø‡ôG‹â«åÑ¥ñá-V{í<ÿón´.²ž‘"H €Â^ÞÛ'Äà¥áÍSZŽh(R_iಙºç‰Œzê,AMÍq-(TЋ;tÕÉç/ '4ŸÞr¼æ›íž.£@›§ Ãêz¸òe—· G!öJAßÙ2" ®¨ …¢ p¨¿4¶”p©ZF ª 5o™/˵±jºÚ0ŠÖDz_3¿ë=-ñ\ékKiǺ?eÏÛ$¥Ëæ4–Fp+0«UÚœ™´I„BBBŒ÷M¼†SAÕàÁº;=¨Q锸×9BK¦Á ã[åÍè·»F "éÍ»¡4BJ9ùÞœfX Ò‘Fø-aîW¢æÒ݉g€Œbë@W’2 ‘jqµ¥$Wtøb€Õ%)U÷,\»oöh1èÈݘCIÜ`ÕsM™]­e{~i{D2 ǤtVêÂYÜÂ{Ó­Ccê ·vâw Jè 4¬ ‘‡ª «NØiÜäNq“Êœb»“UPôã‹ä#¢¢Âr!ÙO¾žmsŒH»¤*{K›ðC$ÃJ1LGˆ>—j®@ŸÊ§"+å±gÆ?Ó”Œƒ•²†i"è% 3ós5hÍÐóìúô뮼õñKêbUÿL7zwû$–/iìë0:%j›3ÚOÃÒ¸;üIÈæü–[Iî²—.7NT$LáV¢w¯¡åvq<ßÙz§üÒ8€ðF9?þuš'£Q­~EgÂZ ]݉<¨ŽTœ?ŠP±³»ZóáÉ¢'9‘tbíOFe•Dú•!Q—îß©I*¿ßúò÷çw¾NÆDlÃÕ ª×8 NÃR¤Ê´L!ËkÝu÷²ÕYŠwn)ìwZ,7I±;æ˜z8tðˆZ9È ö™ÒÐ\‚u{Emo#% Šþ2q>a¸!hA:Ó ž…uÉþ—÷)ð^EÓÕeCRamøF%{…íGf³‚Afl,gyÐHAWšÚIŒ* ··š9õÉvߟR顪N:^Ø Y!6o>wL²ã™úÈOã[Ç µ+‹Koä½ÛÝ#õ!Ûº=âŒ$ÛD6þH>Αå™õRNàsßUì$è”4»þ?ÚR)¼íQÃ`Îiìöq°Óá{¬Œ†ðJ¬GÛǹvÊf¾øjÁ\âib'ù;ÂÚÓ·¢ þÛÆÖh7(|êàDæIÏcP¸N¸hÖ}@/ oò¶|_Íˆê ±šoÁ* µ4øäÖK™Cø[oõ tùcYüˆE€ÆF‹˜¸÷/ŒŒOùа$#ƒQÜýÔìÿYN¼ÂrFÌåYÇê1•¿”™½>†ÝŠÊ?ÁŒŽá/k¤ã$BkØ5©mD ½ŠçsŽq$)%#¡>D?”õ9ãÂ#o[µ‹)ãRHñzá›·ÞØRzÎ=#lvym%®º¸x±á˜ÖPrà ¦8!JÈ$G5ã)ŠÎ÷€F^+Ú~fZÝ$ôájõÜÔAÈÕøRj~ßX!¢ÍÏŒ#œSLµå­:0…ìµ-däî œãñB³6–)}¬ˆËƒÒ=è?îíž }(nZí eǃœ ÿ;WÃs¢ê‰ax1|ó¬S.ñkËrÕÊç¬ß®Ò½ääð#õƒBð™‘ÐÏÇÂŒ¾³y0ôþH¶¯£ÚÂ)¡‡R곍)-ÿt´8É%V[ZuŒHHN’€§{3ìðtB[ œÐΗÕýwž`×zºÜ ‹ADˆæ–²+¢l˜†¯@¾/ØÈÔŠ1Õ-—,s½“"ÍöåL%]}-À¶þf;fð60 .,G`Ã]1ÙCçù×Kf© Î à6ÆV¥•Y8£·%pC {°j–J {ÕóCõê2ðÔûs©—NRÿ:,öòVY+Ê<‘R™Y›€èƒk2|–aŠÉô³‰œ}‹ÞbÞ-®Š$­‚®l˜4šÌÓ*P±—’R«þì½ë}O†&{¸Nö¹ê¢÷sá—ƒ¦\ø¿xë~àyH8&Ÿ#à›C’n]h©‘½È;öï=@ç¹Lו‹m­-Üë297°—îÂÙWqvþr€Õ?¤æÃÁÒÙ=‰/ DÔ{7Fú¿qÝ묊mê)U°Á°Æ níª·—ß³/;Ã2h1¼îõ —Sáþäô+AÞÖŠþ0î"‰µÇœî—LyÛHÒ%Ćq\’qôKi?†ée!*½r{éPðe&AÚ0ƒÔ‘k)ÉËàwS›fõ,¤[ñ—-*dWk\Zᨃ€f›÷öó1sßXÚ¨ZLš7Žï+ÏÓA—ÙÎâQ4Mr‘ì©Ñc ×0^°ŸF¢9A´aÔöW©/~üZuøxÉ!&¹îH%Ò¿òA ;Aíd.gï‘ò/)±SàZ¤† o²a©g­ßK®š‰ï7(ª9ÞûÅøŒó`Åj*ìtè$Œ›¦GÃ#0çK]FøþX²få–ﮑÐà? ¬óäv+g¦CslœÎ0Ÿ„TGáp>sòj®>À¢¶®Ô¢/Wo÷àhzÇž}á·y\IÛ9xÁþã„§(Ø®èH:@WLÿÍnƒE¬/ŸMÔF·a'Æk ˜«±Zð¿brAç²…”òÉ*å÷ÌÁœy ð€K„\%º—_…›SEe¦ÉÊ ¯íꌅuÑñ+ËSûº9•B%ƒdeu0‘v©]~/Ð@Iœ$·€ÿsB-V„aîD‡Ø…/È4çŽé¹SfD×ÎÉn…9Ö÷kïÙÏŠ3xŸ(<ó‘§•™\ Mk¿>3®] ÞUšŠ´Ê=µ¨úÙQC„_b„t÷ºhϳd '-9øÍØ‚Ãúnû¤)ïa×égrö稛79açɾõòÒaK?³Ç¿¼‡kqP›ƒ‰Ïëk¼#§ AN¡ë9 ËõÉdü1ŒíñÅŸ×i_#5ûÀ€~µK<쉼7GÎpò'¬÷ZCÔúCBbá“f´{,°ð÷ób¥J²Éóþ]½ðbÖÌp7²ŠÓã%Ó•OœÓëd¡á ¸R+j¬5£Hÿœ§,Ô—\›Š`ç+*PYöF |Úª•1~YÓ6œñ$<–©Õ8´}ežƒõ2AÔ‡B‚þÖ½÷$ñÄ™*¡‘‘àYõzàM’%ŒÏ]˜,T¤ÿìÛ^ÈGȲf}À”$‰™‚ý>¢ÃÛF¸Ê©óq JÚ¶ß^í€UÎqçeßyF¶Àˆâ-”ѳ¶˜Î<é'hòD™ÐŸîÒg®¸ê ù‡| •"l)ª_a¯Ð˜ÁoBÝWãðU“NYûËmÝ]\e·9/65×é˜]R“Ä÷Äj`XqB‰&}(Ô¹)@ ÝÈú»¶œÚk²¿–€9Іû­Ì}Q:3uêƒʽ„ò*·­9ßN²*ä,{P»— •¢o S‰Ùy=Õžq‰ÏÆZÒqǘl/jÍ™”²XÙžÈ\ßnºHÆt÷}As¼ØÙ§w74ÈïTÇ'€F2r/oÑù¿ õØ5¸J}²•·N^80D°êŒ$¿:¾t©!ÝÒŸÿ_Ù²­àµ+UösCšãY~»çz«yÖw.v‚°t÷x 1»…²t# 69ÍMêæ£éœ¾$‘ùôG¯¹–7ûÝYY|˜ði¢Pl'4”ºÓx·MÓ`?ïs(ç/"øÒФW„‰¼HNB?Õ»3`YžjˆÊ"‰~ƒ¢¦ïÐ"Ç;è7ÆÛK\o’ ªÑh­Ç׬åó¦ë8cù%#4"N ¦ÇA ¶~¯9ù[|ž@yJîvàÖc`¼L:ßâøäÓ½M©Wm§CB"ºCE{β[°ZJáâp" 8ÄZ¤Ti8ÃwY×í< ­êÀ::Î/¹~1„M?Ú3ý¥(âI_d u¨¡ç Åšz·F;gÌ#¯ ÎÜ|ȳg)ûêisD™"_ËÁ ’ÖuaF¿ª•ÿí-,Àú;îkCÆéˆ²NÈñŒÆ$P.ú0…v3rÆ;šKFB+>7ÄÈI±Ÿ^Î>qÀÅt’+¶6'âÃ:Jb:‡äŽPœ2Õ°÷¢þ©@ŒHÜz¯}oS;€!îÞ› kêS6¾…¶¸$–[r£0‘øDMž‘þ|üñÒ;ˆX½¢Q‹éË‹V¦U4tÖƒüá·¬+ŠÙQ€xKµúdžJO‰;ÀŸ(¨>CîJÍ£iû¯Õ7¨öÛ£iä@}ѼÝ"é «É N@5õŸâÜökP3ʆ]5—úw¾ë'h²á•Àt”Qf^„=L/söäú4·ó.ñi¬ÝÄ+‰oG–'¼7급ÙÕ4JAvPD¢Âãè¥1àR‰®˜…±é, «¿7¥…g. suèCÈÉ©{.$*žÎõ•CQ®ìãüe¿æêN³/Äa´¨ÔùäðùžUÙ¾åI>Wœ,žiQÍýpöõ†2òiÓφ]µ~ëâïÝ¥FÏ9Nøù ¡«¬¾&rÃÇôÍXÑà Ò{(+V½übØ6³§Ž€_ðyµ%ØŒX÷¬X) ëu1Þªfhö× ê®v_¸¨‚¦€Ðèæ zÅÉõ›æÃP0H³Nc滾H×/¢,û€Ú@û&ðÜÁó$kõ_'`zA¼M¢h%n÷ÛcÛäÜ{~a»à&½&A¶xZ]Ì F'~пú–´+ôPD"I.¤d9†°`_õûÊ¢÷&güY8 hí¾Š•²eñæòF†“0Ö𢽠ñø€>£(nÌ ºzŸ¥<%~N ÍmL«‰ uG“7‡B¸zØ{õV¢J€\,Á×Ús0\@^ ‘<ð¤S¾|ét”u/'%~µìV$«õ-GnÕ»oЦ=¬bìÁ“^ÝZbXåïq·Ú½-ûCX7mÎ/ì¾€ÑaP"sz(\0O¹Ó¨f÷‡i ´3Ô™˜­j$„ªVÑ‚u!¦< «é@FAR*6žw|3õ³hÒ¹’ÁSz¼ÎR$¶KHúü{´ò[š×¼‰_Z¤¸“á4lj èë‡ t˜wQf­dì„ YúX™¶ü]ÙmóŠä+þN?唡fD€lÖ~‚~ê-m”2ga¿ù‚2Ÿ8´BÌ3*ÍF¸Ö’Îi4ƒ¶ùP2s}pÚÆiT?S`Ñ€¼#7ÿax…yÔêÚ-–Îoî ×§¼”µcÑV>­Ëõ©[ý³¹ÞEià(Wx•§Íj?$Õ”¸ø¼AJ:ƒ™÷âxi7{F1²¼‘;NçîdY¨ R¢j½QžCöñ–ß~,û韥\¦/G>V¦ÄÝ6-ÜÕ?å¤óhÂe ¦M•ä^QŠhÈ^U^ºñ­•lº"8p©¤áë±P@ÿ+¾Éj3ò»Èq«5`TywÈõù¤Á<åÙnƒ T´ƒZ¢ý?ç¨6¦Çq2%çãÂÂÅ'()E3)gN¼wÀ¥ÞXFF'Å÷ØŒ¦ú~‡Éšf5yM/{¦ü·Ýåº`‡P®A£µ/âî(=¨ðh˜b¢Øúz™ åÍÿîý/YÂô5>ü Ú ®$q %¬ß©SØ ›"[>c]áÓaÓjJ^ÿDÄ…Ù-—*‰_{ðÌ4É£«£Pò0?  „z@#á6ïWtÈ¿¾VˆH™NRºÊg¨o”VãD@‹[û§ä»ñ©¿JÞèüçõ‘Éþj×&ÞÓESÌvß˽k1XM.⺮{¹¯òà ¬Þº8±»·—šø…tê4ˆñ߼؅ìMò,¯+ÖR÷êT^m;àWš+Zrõp¬/e‹ µÿU£ÊGAÚyúTÈì‘°²Fœå¹JÞcG29™žjZî4¿câKô…PG»æòûC|•nÉQþMò,p§®Ã²,Û%²¨ ½Z»xknþÆ&0Í“öz…Â4—]²kBjøô˜©åÁ4„hûmªôKœS7˜Ò™ÏtŽÄhg6cXc‹H³ËŠG¬Ïq¶k´Ça'ÿ…R)‹RáqÔáñËôjúçåÂs]FÚ¹Vü`×q‡KtñéAMØ[w„‚Cs¶²Kk‹;`ä Õ“>¯ëshÈ?øõìÂÜÆ£‡‚šnlâPÞ쎣EQpˆ#7ý(é8¬Â†]LB·ežôt y®~B3›ÂÅ üknOq"@ØŒCJoЈt¾tsñwM!êZ-£g¤x4.#oã÷ù>]¹ûþ¯kå¯ôw=ää¢ða éÁ¨Ô2¸Š+O ²ÂŸ®Ì?-¤ñ\)²‹·òÒ‹‡Š€Pô­Pm`Öýâ]n­9^¹Õ‘TîªÛ?(lÀ6Fà%0Ý…Tú– þœ·9²h)àæöîeÿî·äsœçç[ì²€õ‚Û#n·×è¨ØÚF艹̃¹üAs±: šÿGªÿÇ ; ~ÊãwM=‰C¡@ãÜymD™îÆkm¨“ÀÇ,AUè—p TªÇÅÖq—ÅJúUäìmÂ&΢ ,FG™·É#Ϋ20el®¨þ‹´d¼°Bþ«¹¿§‹„> 0Û'Ûžÿ0iOâø6Ñ‘4×5 žÀ3èu¿'P£š¢«p;"2aâgȺ@¸;VxÉ<2cìlW•äOeåÒ¦ãI+¶Ê¯3’è®e·ˆEålTç±éÁohÕWŠKùÌ»ƒÙsœ—¤˜»l'quiÙŽŸHP˜£M¥¸Bö€> ñÅmŠšˆG[vÝ9^DÒ‘_æнi{ <ÝÒzRð…íŠÓíu òŠæ‘'.:XBåhÑrz"OðÚàÁ…Øôù¨ò>.:]ÎCMd,¯ËiŒ‘ 2a¦OWÍñíã¤Ðv㸨ãƒX§d))<.T²؈É>ã“fü—ÿ”p`†Y "Oyf6ziÏ¢V!Þ¸gæ $IcÚ|¬Ì†MK–Óñ§-Ñ•„ã¯Z>“0Þdñ-oÕôñDB(Gп’âÓ >]ÈÅ}tä‡&¥‹XlÖgá:Õ¢¤á™UÅI °ã›`ëcxY(Ü9`Ö‹’vMÂþù´ï_w—©Ìþºý¢×&svñi <1ž3u¢MF@]âýYÓÌh•®@÷«Š\¶ÛAŒb‰²mÞiàÆ·Dæ~ÒÐ=Æ£?þ™‹¼²o÷LÍXK×¢ÞgD²Ð ¨Un`šh9CáwòÜ'»x¬HgÔíuôsìK²•öI#*€ ¬ÁgÊõKü°–ÚŠfØ7GP–f2tµåwfRwE…ñ`à°¸¤²ï¬˜,HÏb_g18®”j¬6¨˜µKWHT—‰ þöâ!¹Þ4¤äá'Y•Fk9å»i‚ØL¤„v?ÞÔ»Ìß÷™…¡&àZÌÿœIvX´IÓìøÏ™ÇŠ£`)hý2p£kS"¹aé"ë ‚õŒIIÜ…4ÁõÉ Daá§øùG¹‘k.a!èäŠ÷B§¡rç¶¹EßFZÆ ¨Ïš›$.ŸV^»¦}lˆµ’b3î?  Âé„ÚÐG·&›U„ô öÆÞÜ2B aç§ÉºÛÁ —†Å/jÕÐWÏ\}ù³FŠøWxX~_bÛ™¬™bUÆÖ<ǵ\tHż,V )T(z-ÈÜ„Q9×¹áµËìÂÕ\vˆc”Pî+D˜Î -{ ÍÒ˃Fk)WA‘¬»Ïü×éŒ?‹òHåì„-ú½¼d}L@_V)È'ü-7¤þÏçS1â1P³ª èô›d¯6AL5¸‰»Áoó:¨Š€Ò²•R•ɯº9 —A<Ìå:þãV†B§ÌŸ£ÿ)ÜC È zòÏó}©%¾JeLÏ×%Ú®e_.6lO©jÍ÷QÈÛ²Œ¹¤ÙÔªnYPJËÆî~ÆUânt§ÇÍyVª˜øÌ÷§ºXsyX,Æû©n@³ÛˆKö*¡jg\4@u'ˆÜØà„Ê\K°^7åRS$måxßÞ(›x˜e¤2~µºoW{CÑ…v)·cé-º»{b7 ŠÔOIНúbõÑh±ÅÇh\G)Æ®)ØX»“ÏãÒÜZ^N9ümµïë!Ð0÷öŽøx䯖g¦”IE§M èÚ¤Ÿ¦Uoâtà !DHvã7K1Ù;ÈW¼MFíZ>5:»~ Zþ ?ÝbóÔ9‘)lÌœ¹æäÚßF!6K<«Ë¸ïÃ,PT±cs QX!kd;vº ±Å½õ¢¹òísìšDƒ%sM~+Õúû¼z]2f„CîÆe¼ƒyйr…,-ä>Ÿ#ÆD':û5Í '2ÙNᵡD°®#\w`ëA_ióIg(Rùƒ¾ŽbÖˆÝZ™Ü;d–¯ó?7°?ð>©«çIŽ„-†ÃäÐbÙ"%§X†âÙâY oµcne¥yÀ BHW´‘l+­Ï<Ú´ºzT Ŭ$ðcxË»ÇËÀúý]„*œPËP,°½š*èî&Å12xe<¯_š ´Áæ‡*Ië=Äs·Y5’2V°}¢ceRÂŒÖQÍpcV–Ö,«|ëõ0˜/lPâña;É{©qíém-„Z!‹Ò0eöxzÕ¼-fe”Ã{DÏ-0N+rô¡µ¬¡zÈ)šý0v[¼-ù’Aú*„Z]N`þôx+ñÔ”ï/û Ågç.2f~Oau,È{’úzbÛ—Úú_?…Çâ½»À 2p¤Í¦|;†@ìí-ì”uT)•Ö@M5 dèP(f5i j~Èæ'jÇñͽ°IkÞ#9axoÞÍ×w—ñ² ¾0ÒL6'^mö€Í`.¦ß¡ÝЬùÎíQ¸:`€ö~:¤>º}ÖbÂò£ŸŸ”¢"xÄçn_²4«s0„ÓzŠÎ‰°"9°€hþ-¾™|J£Z¨ª&-Î)´i2õ땞†v‰O›DÄ¡ÑM=¡Tö$ÙŠÆh,0îF«”mvJ:¤éù³ŒÓÎ`‹A>wfc\dЛaÙ_¸ÓÆ´£Ô»FaD:ä¸Ç3Lþ±M(3W¯ Ê&åGÞÌÿ£&—»R–~M/44ðä<ÓhK<÷ŸüpQ½”²>ïƒ ÎñÂ<ð:^Ù'z(cÅBÜá`ö6æ#µc‹ö´{®M6¡›&S«…‘Ådè¤æëlñî|6,qT6‚1¦mó7¯|Eü}°¼*ë 6´˜Ò¬F·‡ 1Y\Yþ,­ëx1¶—òèÌ«%ÎÊg9”¨X-©÷:&ž¡Ù5@ÏRF'}ö6ÓËnö7›í[Äóêìºî°x\ú¿+sK¥ ü16 •óWáú1T9© TH¶-Œ ñÖ?"p±ê™ˆU»sª0ó94@- USeq?<ÍVýËȾ–¡Hº  °Â óÚÝr_<é=T$Àò~ö_q,Àaø™ÕÖ¬(ÉPpª<5_–À^^ê©ñº 3¾õXê\Uÿvñ!ô4Nø$èóÀ%„™—ŽBZ£°©‚Lâ'µ|í¾¿ŽYÌTó‘¼?|€°@GÂ/ÓZÍg=ÿ¿çð¡ÍËYZ/ ó2…$ÞùWµk×j²ÝÑÜA¾ù» L‘ñƒDwhÖ–f뙄w°Åµ@ò"Éfz‡»7Ïih$1å9uóìe˜µàÅòÇŽÕþœ‡–ó8A°a»ˆ\­2°íyâáõ¦ O¦âÃÝŸ4ïÐ4 ËààäžÃ£ƒmq™±ú†±Ž÷¾?G*QD0'Ÿhkɨ´’ì šU1ÁvPŸç\ ýróX§g¯&­c}™¶_ޝôÀŽH´"Ð]tel`ë!,œj:ö›s ¿sò#iÜW†Ûô^„Wmܦm hà5øÞšÏTÉ.ÿ«âð7Íi,K·®ê 1Nò¶é®Wly1ܨÿèÝÈ^VzH™ÂZÔ”P­Ã“ãS“Ì…Wm£þ*ÈöK~7ÁØ×©@/Z‹rÓÔ¯:íð‡+ˆÐ£%”çW4ÔQ,ö)Ü[Ú-YÀ( i¹e€ë7OlWÝ¿jÊKØZ? i(9Å6L³d†„…«ˆ±²¸ëNµ/7I_ãù´õ=)Юþ‹*ÊÍÏP,äùüÊ.w0zº~/7‡×f%õ0Ó¹ê Á‹€Û*™¦  ªïc#TN¼¹ù^W6DºB$dÇ8yå êdxv¡ÀRaG»¢SŒ‰ÚöÃõî C\fÈ›F¤Á‰{ ûƒT?cSAÝÓÔUæ.~`¸ò·’·,û¥Æ.Ķ·dÑ)BÝO­y˜aÂÓ&°¡VV’ö†ŸX{9 s‹}·^&;R+Ì“žúWõPNøæ6cnMyöÆ6ÔÅV»‘t_Ù»õe›1Ì«É: ?Š]8Ó¸Ç … {rÁ¸ónKù˜@Èòûïn¿ši¢.þw̳;Xž7+³§Ÿà%|fï'1Â>qá¡ó¯áÖkÓõîªKì-–Ë ºÌw—ý̾ "“ßí'³¢âRªtË«Ù?8°ïa‘ó ˆšºmϬ½§ÿÐYÊN”S¨&Éåî@æÊ ¼+9RÐ!-ã¾äºõö%e(+·]Tmméª ûôz‹ ?,Ñ•D ܽeO°yºƒE`›n yÖÝ^%²,Ë­…ŽÁìTùö—©xJw(Œ,…¢.£ð½Q:’Cµ8¨°c·uÿÏû/ž£šU*ݱâ-k¶JûHíÒ×µæ,uñHuÛçJÝ„¦š´¯1©¹mú$ƒÏ…Ë{ÀÏd™SÜL·ØÉĉÏÐ5ÑJÁôo&€P&Q—bäöru¦FFˆub<ª†’°Ä]¶˜./ЂßRº¿Grè,YÈÿç1šÃ3H…¡a,<œ-ËGX%ÓK'É$}wÉ¿cx|!?Üúý:ÄÌ:@_ˆô‘ÀžÈQÿã ÜlEË·i›r7Ñoõˆ'eóP˜µæXÍ3uSŸoëáP·Þ_‡ˆÍ<ô‘„h$_ÕéEÜw Ñ3r ›ËË¢-2œbfç/~@<÷›[c(-1¯RÎÅ®Ðy#]UÎ=Ldsh×bT/ëåÝü›#Hpâ¦íÛQ­‡Ïäë=­Èêã.^ñ“¯sëဟGÔ¬ÞBJrb^£t"_·í¡…ÅèÕ xŽEþîç__º6è„ \ƒ{¦X×颙¸´‚N^)¾e&Ø&Wjë]9˜J¾$(÷ùRá)&Æ kw›t–›¼ÀzpÒ„•Z6Ø8ØË®ÙN“J4'¶è™÷^°u.w|2£5Ûb)\Ge. v–û,J¾¿°Ú,IùUL?œFŽnºONÃ#çÍ¿UªîUÔØŽ½Ð[Ú®èô{&åh“÷^©9ä-Nà)Ë-‘žÃHóàÉ­ö³Ö‰PP šŠM>¡UATÀšä1!$³1öe~Ë«*û7>]üŒÏ\”ìo¢èè±qa„­Ýà¿ ÖgJPWWË9~Ô“Þzë.qä-»n¼Ç**gÈýéaÍŒ ¬R-tpˆèˆ¡b»ÔÃ)”7·Ú8ž%Ϩâù…Ím7]³—‰B§½ƒæ©¶PüžAù³I~¦z¹‘{…OÈÕÝ»°)C÷_â‡6èKÑ×ñÉÀƒ§M$_K¾yzËÿ{´P ÜUÌ¥ŸgfVüN{ìÿ ~D€4pìqJÒo»xh´ê¢)ºõÓTþƒ‰°ÉÐ,OrÜýù‹d1,j•’2ˆ€w8€“DÛ$È/úhHŒ Z°ùD‰å$à ’¯“Qœä3¿´1Þ÷–Ï8·ðµ[•Ô¤ÔàZß46‚sN#Úï€ µGÁž~Œ [å>ÛD€ºÅ¡NoÕ'¡EŒbj£Ò-å…ÏÄŠÃ#ƒAk‘V֔ߞûƒŒ¢ÍE‚õY£ ÏçYVöÒ¿xDGnüG`ÀŒã[U”p,öÌ2XŸ¥lѱêâ$ÂΣO®.ßìø³qŠœ¸ví¢HX Ëo%ŠŸÓd†MÙ§ÈI¯n´¿Q™¡;¿ÿnpDH`×áuyÊ;¾nŽ“­85â¡X)†Ò­â;L÷Òã¬ÆñÌø²64!ó“'’+ºJ™8½Ðª™“’& Q//5½qôd—dk¦I¶ÚÄÈMÎfÖ†¤Yýž.|úêóVŒCIÖp ÔZÊJÄd£þŠM>x¼Ä¨^Ú¥ûwË‚RPQV«(f&ø†¥¿¯Ž¬¢¼‡:eú6­ï<™AÝ!)õ›%×b,!‘€ÚW&EÄ€Ðbà`¸5û£éÅ×BÁ•uÇòÄ’š©2‰l«ñúÆ"8Gb‰R?HJg/m¶§’SIîï4Ö/¸ç¾ó‘øé©r5‘ˆÀF°>º„?Ä BlJœPé-Ö×pŽŒÖånÙrZ ¹Hi_ál“ÒaMœ§Zä6EÕ&_…¿7>нÒ)µ‚toŒÇb˜s©ù†ÿ„.¦Lë<ý({üî„~‹øùY`-=Úv.=0¿m/ ✟m;­÷ Ì¥…ºØRP8MëÜ|ô5W Zõÿ-¡©Â®2½”™ôáÓ³§¼™yPîMÿ“ ‡ŒèÌÛâÈ7äÒÜ1`à“UêN5à¹U³s G{A‡'©%ŒÜË®Hƒ¦¿Ò¨Wi/¨"‹ó‘·éK°Š:˜ñ÷6öSÑXÿÅô*Y7~§„ «`p -‚A¿âKØ<âp²ÄtŸ\)èìfX—.Ó§ùŒˆ¾°›øáªžÎέ̜–!ßF¦äP7ñ;Ί)V9¾á+ ïvÿ`Mqõ1–΢<É6†²®ª¥F28:gaeפpµ+ ‹v’–¿wÒ+à™±Gm®úPu±¤”GDFÉ0_Yã_â§«s E³qØg‹Ï«XSþ…Ân;ª¡NWóš4¡bç ¥ƒMú¨nÊGY÷€-ö$Öy?«^ãâýb£;íxí.W[ö>õVäfÀ^‹¤¿ù2½!®Î¡c§ºàL‰”±Q@hOžŒ\öE¡Ø œÆ™|aÒcàqûëÙ-L!§€hÌ‚×ÔCZ=«Ãu-ÌüÃQU^™Úªó iqjAc(â°µgsŒß{ÿ÷|÷Î2ºªÅfaŒŒll.V¨ÍcÆ„‡öϘ}¤Š÷ø(k|·ýiÕéüv &oY›²JX~ÿåz¨ãZía"âqY™A(&¿Éäó¼Õ3®=Ò´¹D,Ô¡H_C•ן/ÓÍåf¨'x€qÙžÝ8û~î•|¹eͺ,vÿÝ´õ÷åÚשׂÐhב‚\ß2ˆ!J‡ÙSÈç̱ ®¬îM†š»6ë­9Ðö¢ªDÝqµç̽ã…äjQoôuLµÐ=d${*4 Ö2‘’®séP÷ÿCQ¦HõæO”oÞ1†²2ï’K¯=µ#¢l3_«_~·ÁÙ²Gä¹oÿ™Rj-w"&ý“:þ}*.ºøè…¢Ø…¡V uNØD´hN‘«ˆg«ˆ ¢ ÐHoP@´5­AZÒ\.%ƒ0hH¦»xžtº6€-ø“ÄÁó]RÂâšÓ/þÕLèÝÀù ¹vZzjLÕ.lÛ–·¥ ê3©Ñ”Èc ‘Í5€$<ì§ÐÐw‹¥‚ì´˜Ôÿ×½â¯i–ø'¼ßypþ”r‘ÒŽ÷Kÿ{ìJ·¡\®sŽ q=àQþ‰\in茦\ Rý—c`1ÉW#4Ýܽ^"þ„š«µ-Ö¨'ü7ÂËU”0'aXâz¶[œ6º"oxIÎošÍ!·I‡pªdðű@9dAOEc…kãP˜¾Ž®ôÙ32Ë“§âä«–…¶qá{BI{înôÐ7ãŠúâú «Ö¹p¿Ù¼•„äóÁEf.¢¥Ñ8ý mƒV0MõèÂÜeŒdt k´ïΰà|ec«ëÆ–…‚îÌ«÷×rG’„Ãf4*»ÁA¡s2°þ®[¶hÀbšZAæ¾™X­zØ?oêô:žôC‡Éòw²ÂT.R%ù×±(Å ”ŠÑ¹Ã×—zIŸï×§a™oï;dÅòaøá2» ´[U­9 °›q4æÖê3Çl›pêu»GLMÜPx=dÑ ²/=@+ónVÀ‘M%nñ+=ô"pÉ-òl(:Ÿ‡¿ì]fû¯rÛ3Ã&U‚JU¹©Ðe³?4‰RY5VpõT u¦ÅJ^§Ì›Tpåe؇˕#h4;MÜâMüvCà™BBôEèW-ÐÕ5—±„#Úcc]ó!ó¦-+f)^|ö°ã9”Æ#º™¿Î ¹àq¹Ã¤Ãþ²–[YïˆYEå†~nÈtÀžRˆ$bZn²còθœÛxÇNÙSç P<¾`BB§!Œ·1…ô&õÈŒ|áÊÛ,yW$F™ƒj¹%ùð;¨[52Þ¯cžë"÷?¨™aÐúÜL'ˆ˜»‚࿲&±S”z´†£ ᦦãUÇ ¸=Sù<Þ˜‰õ‰éUÅS À“CË0BÓ'‡ œ'ÜÚ³fULµFxxeQql'ó\µ’Ü.yói 0>°ÙÐ×tÍa oˆÿ|ôˆe5«5öcÏeT‚¼`‘»‚7\™9þp ý³ôû°\¢ÂŽôy÷óH•Zæ Þ¾­L)§0Ù¤õ®m8Tˆ2¤ ë×Ï‘roJÐ{?÷¡»îëñ¢ÈÓÅ{Sqi‹u¿ÏP·ùP ™V»>¹|Û.§•×\·,H‘” à‚ ½Qc;ç_R©h¶¤Šç‘ǰ™SMÍhˆvlܱŽV­K@#ÍÑÈ ›=”)JËË®œ;4}]{snêìˆô‰ø®¨Ûôf‘Ðæ<ü™­µâ†ìçtŠöq.Ês•Ænüó›²*E .h9< #•µgù†‘β.¶Éò&Å‚MVH´p6~I<Ńˆgé_”Ù›‚¸óøky¶ydyÈNã¹>Äòu_¡P5.Êä g‡Ý¡˜GÎ ;œ‘¼ ¯N‹§å‡0€QP_§QM«Ç~âsŸ ûz¹,‰»ô_Ý¿Ëû©™s±JaйVU¡Þæ.÷”HÏ@ºMž4d_>Û‰Ö"?8½‰÷¾‘'À£ÿLÜèò0Šø8&{;ü’U@Mh¸—\X•ãö†hÞ—ÙÛ]<83[ê8(¤µ_m•B^––“’ĽZ#'Ö×ù§—N8™^rx0@Je¯Îtw2´јÌAñt|«ó éBÑÅ~Î3P:«Øš+*ºYGYói½ÅN-ÃÁ‚žC‰=MJNŽûæðdJsçñæ‹Å-ñ\`—¯.æ”ÄÅ÷l÷åÇ0‹zqî°Zóäð÷£1ž^Þ>pÉßZTø´àSƒšÝÙá·<É™3K™ÿo¹¼X&ÐY‘Zøð¤Œv]ϸJÁ¬˜ýXÑ;¯h‘èÁ8¾yöV¤:[sA2fÔG«6»ƒ…¤M«jžÅޏvÚ`€ŠÇ!’5ûñ7´Š"P(ÒâæåÛ$4= ‹ÐoíßZ&ÜÏWóqÃǤUZT<Îsˆ+=ÍM³‚&&)~RÖ“Õ+aàMÕX¥$lroÍašZw|Ÿ°˜›Oê:XZj 'aïÈšÒ°ˆòêÔ§5†¾™[É) Ã-ôle®Önõéœekhåu(ú]|­Ý S³f‡Eòÿ\<ëí FLàa»VvJå®[‘qæ§È„mdÜ&bBngZ!ìy0I.´û V˜oΚsµa¾eBqª—µ)›æ5j»ÐWEŸï÷ÉŒ:Ú+¤ˆ£êÿ¶ýf-ú=ŸF[cú-¹K¿(G?8„dÿœÛ_°É  eGÚùñ $îám6°wON7×ÄÜâND¼lû:ŠCKšÃF]”¤aN©×¦|¼ª¦à3£K B¬=°Šéìõ6%—½Ú"¡–|é¹AÝqw¬RVÎÀˆê”±õ~ÛX‰Ë’0wˆ·Wfï»ìîß»ñaø|3ÍxéA¼„—³‰®M!ÀŽ~;8[iGeïeÙ˜Tz»ÕQ' ‘SâùÝ+ªtÑJ½q°6Ec£NÊÞ²6&÷JÉ€ËåFòV¢&|»ˆ!Ï©<à“hÝÏl˜Oz͈Är$”;|–Šè¹MîÓ²ƒ*£(².Ñ|.ÙŽ½Ô¢ÜÆÙå÷ø=°ôåÚs¦´M3Ž©OPUõ#PG÷,pÄ–òƒiÊžíETIk]€ÉÌንü52 ý€ùЋgR¡‚{í7Í`O²üò•pȘÃJ#ˆ§Ç·@ý°«(ˆ=CkÎK•ýñ…Xz6 †çÍ GŸ Õ ¼ž¸ù翸z:G+#¶õ64£}9Ÿæ(¨;2HL{ò­åd@(‘ÊbS9ª;‰ƒjÑÚR0ŒL„PÍ‘—w°8‹‚ÐmM:­ÎúLŠ?ÓØ[ʼnUBÿV»œþ>R`Z¥ÉÏô{z ñÈ«k_Œ©yà.1êÀ9)çÎ>_ÛÄ+‡ÌóŒZ´Rƒ4µ•[1g¶|¸æ‹peoô£¸ ôìë 4tzU¸ÁݒǶ¨zp¬•öhõºn<ïOÅmóìq‰’ïAë*úw¾Þc¿x@i¡iiø€›Lk«t<„–Ëê³nïßE}„óœW×2ø-ª}ê½Hd«f¿B4; L™ƒ·¥ˆê^ØëimXjëg—»Sޱ!{éÂŒÖ6…ç¶Å„½$» Ÿ¸> 9 ë–p, 1=9àñ£ëäÅdœÖq\žk®3ªë53+§YËxM­7ù™³H$š;š RfF_HÔ`x†½ŸMÇuhXgÄÕºFÌd¥W‡žì-™Bd0rÑí„nÜÃÑJKvT8bR3À° J™;iÆ›ØõÒj¤åÔÄpÂ~­g ™Yçï—ÇrFxGÜ sÝlª!×-dfŒGU~BŽÏb™K½›WÌù_܆[b,†)c丼RXª}ú@ôïÆÈŽõÅMXŒŽ‚:sYRàd]B5Q+~)úÜ U ƒë·ÌiŠ¿ý›šX>`KˆY&ë[?Òw€táÈ·gˆùyiÚlû9þÂ?‘˜²ØÓȨ(º^VòY”fæøc”cT‘å;“ ™ÎÌ úÆ^5®ýWâ!=h•´ £“Í”4=~x&º¾ªû4…-„9“­UÅËCÞÖ“†,?ï<â¹ûô/ªEç Õ3kvxS”ᎈD28ÁÜR“±¸Õ! q¶¬S[|Q]›ûÁRßÉ_›]d€à÷Ûu‡^;ÖÕ—òÎI/ O©þã&ôÅCö9l|o}Ó‘»®Çv!p 6—-jþõÑò#zýðºÐçì¨þå‡M™ÒvU"9Y'¾´Pÿ¡Sý‚&øˆÛ’gDg—gëTÀVƒ&kpe¥µ©…$~CŒiæÇ3®¬û[౯À_yøTjØø‹ÿ@©JîÉXSG-Ü{HE”­ú8ìYL~‰ µÞm‰ÔZõ‹Ö¾Þ¾ß‘VâÑ›hv…ï._ˈmqšÔ¾+´o£‹w;7+4¿™ «ŒÆXìÈh”A&Î YÊç—‡? –°lÿýÞ~ªU:k©ç¯ ®9f"^ÑDZd¿¯¼‚ß~·EÃd¼3 k÷ yާ¤rŸã³üž”!ª|£ í¾%¥ô›MÝïêóvžNæ—¼¥«Íê(KP¼,íŽ&]†›çt5!•ËD ;êá6·]«Ò.bH«J‰íeU|£ é˜×þæERg³‰ÏºzâAðPÆúÖh&wÄâ=“2=ää,LWœ¯¡2ù¢%µŽ¶YØNO!‚YÚ£—,=G¬ 8€=@2–79ÎYÄEàÓ+óAˆŠüÞ(¹¡6ÌÐס¹Èx¨jdWÖŽµ ’T&ç@ó3¼BŠcE=$>æ³Õ*RRõ}âq 5«Õú72t€ ïF0·â9ý+Ò‰ßcìO³Ø·1”˜‘ÏE¨´c9ÌSTO…ž30‹ÃfuD™g(ûÄUzƒŒÉ%µ§º! 2w[âéWs§Lÿ#sÌИk²0Û6Y±þÚŽŸÄgسâeÀöŽÊçæñq¾v<1^rÕT2¼Å~æãTý5ŠK¹-iq\O”“h-N+šš§CðË®^7£ óý[–y¥øœ9þÛˆ_z<iÁðÞj®Ô¨}b>2ô˜÷njä~ìËwS•7Ü×R8C0*8œÈy¢¹í WÀýËá!–ýFE|¿|Ÿ%ämÃr¶Ë?‡±v.}oø›B ÙñJÔB°ªýq n5q­2ÅOœ‚%ö,¢È"•yÈN7`Ä=˺iú¿áK ‰*4œ:™_m7ûüÂ:“aŒc¿ÙÐÇðº©$½ïGcÁË»;ë:ŽÔù€Þ”ýêm™v "oJ/qÕü5ºr(òòfðê°ò96•˜ù&™“AHÝ„0DO¬©nz2›šŸs,}•ŽÅa0R˜u;_ç½#¿Èÿ²•X\Nþ Õ;Œn ‹ö=­øqÏbêW÷ëÆÚÊc`¥í›äÞtü–v|¦ÓTmœ fÜ5µÍ °¾Rç¹P´þ[B­ÚÓjP=Äâï²Ñfo Éyx^O÷Kê¡êæ.çÊnÚBA¶#Á_s€l“%Œˆ²’·ØyÚ´+ª‚í)Lº¨åv¤#œl‚m¹ V‚sÀY}å‹è„„@ ÿm9í õ¨'ü>C!ý‚c1ä•vá¾$­ª‘uCÞÛŽ¤-¹7@Iñ•ç½”µšq…kmÂ2¡éâ~3ÉôM"È øõ*­¤/GùF×ð>.ØjömyâÂ64ô­ð“Ý3º6nœé·|6¾U*Ú‡pgÍccàÍ\Qv%L ·ƒ ]æ.ÅCGÕ÷”zÀh‘œŽõt3»Áô>9²sê£mÌԎЦf)Î.“Šk*è.ÑÔ",„àªÊ87Œ™ÿ4Æí÷J ãˆÉËä$èíJt_z¶µ¿šfPe>½µÔ†_C¯â\‹»Ï@e›ûÅÀÕù6ïñìŸBgêK½-4œûÑMWíçvN T½…$öü8:®Úu‚þÖ©˜å½fˆñ…IÖa=öT,œ@åÆž#¶&Ç×ÂÎr"¿¯Ëí¿U+÷ÁqÀ¸vHoÑâÿ«µ¾}V¼/ù}2wqÄY•êïð>D.D9YF_è"9ü½èÞ½F©wqv;I<J½Wÿø$¼¢s9°cÃÍ‹·ç—%ÀbŒSR‘taèŒK?Î|b¡‚ét߉,p%9´Ý@v«„…´4Ñ`7öN”:€=djËÁ9‡3–„uO‚ëÐXÃ-RËùw,©ß¤Š®SÇL7wÏý/7äHeÃÀd£È*ö\»ØÌШ=¥îî.`Gãö_$ïjš1½ÃM¤ÚZãC9í$¹7ZXϳø¤ÆH1Y/;§w‘Ñœ;WM6µôpŸâ:m„N($¸Vî]úº(s©Ñh ›l£Ë<œµ9ùñNºXø jnÈ 4FË}ëoCäd3'5Í÷£¾ÔlŒŠ_/ ¶”…‡WI-u.+ˆ‡VYŸ·Ñ.Û¬.óË4½º-ÖÐèáƒ^± Ï‡–‰½°7‚Pø ãC2˱†K±¹µ{ax5Ä…¡ÍÓxlžÆýû`Y =,øø#IWeÃ÷˜Ÿ—üeä¿pPüÒ¶¨€p½# ¦v‘,î›èW¤–¤´£GMÛ¾z8ÁÎ Y¦¾y àü;µåQó®Ÿì–Ô-ÙÝ™%äpÅ‚{¬S¥­6À‹â€•ˆ}nnT&¥âiÈtŒY\ÅŸÚ·×P‹IÔ‚ïPü”^ÇEp<òéÀ©ãÊÇ Á¸°W¥+„⯊³üXy”h×Õ² ßÜ·b66@^œ$ŒxÑÎx 7ÃÝ…‡è¨¯,×¢þ”'‹ÍJKèvf4ê9³®ŠÝÏp_‹«ã«„=K bš0ݦþ*5Ä+6hûGCþßô|7V…‡È¡^°¯ Pt°£‰Û ðþ™vC-…íöžvYÀ©ëÙF¡TéùŸ:£! SåF'ƒ ÕÒ/¿Â¨ˆ(¬Þ¶D øü9þ_“·1˜—7fE¦%‡ë‰ÞQ6QÅašÍ—ß^Êö¶„»öÝÖ˹C˜SEwĈ<À„×ð¥ôx4á®sZ b‘dºs<ìŠ1Ìêð°¥Ì³HìÜlµ~w dÍuaÚ;‹`y›òœé|Ûÿ|‹ñüásê²èÙ o î æBk›—VÇkoæ°exÇ¥^ÇÂêDh ÓttX>›ýÒ;ç9¥†zˆ*Ðô?C{k$¹kPoц.Žà?£ÍåÔþ? Û¹ÔJ§€{?ƒc vú(û¾ž¹d¡€Ðò-½0J@¤lñ¿º>l¥7RæÕù˜/7ÞÄóƒsŠbgÌŒœé¹}•Ìô:}B%PO/!sŠ Ös0fiü[õ¥£Ž¤´¨¨ ‚ly¯ÎÇǃÛ{V»³’³ýõ" Çå>†+Y&—dzG±oéÝ8¤$?îËÏ(Ècs‹þÒKðþƒ“(”zc‘Ul€G¦Ú¤îVŒ™(ÒÃ?ÛÈžƒô£ì½xÁ!¡\€ÝTaŠDGô>œöÕn†á n#‹%0JN‚A„Izû†‚[åv’t–W󽔜¤·‘w:zÍÔ"oG/!~¢KF¸ã)Ê´ §×£}-juQúK8§Gäcƒª ܆kTß6sŠãe¬Ì€òÌ%ûLí®'0!a¶2‡?}IÛGU¶&…øê?ö?S ?v… Esïð¹¤õƒ€5ül KªÜ}Pɶ˜@’‚á%`¾GK”Éßó£”%¶5f/.ì&'-vò0ÒsçZ¨¥Þ¤"5L4ñ:¶K[ž›¨”r<4’LEݦ̟c›êâ+1Ed6V‘µØM‰+¯$™†ÿ9ËÏß·Ò²R®‡/ÏQ×êMãƒ"·”’~lçÙÓÉŒÐçç^F€apªôòJõ¼åˆƒ}-ƺx5ØC*Qá{^Vyî–©P£'ÿ¤ÈKÛkåA`¢ø" ¥¥ÚÎ#H»¦¯‡6Ô?wä°Öw ©2Pþ¾¨'x붪ͳ0Qñ6vNªÿ8Š«Ù ¶1ÈŠp­ CWõ%¹8ýÆLö5ý]?ÛY5ÿ™ddRT#®è(k+ºAIÑC®-9¢Ð??_Kc-kš]BÕw£„IzÂdÜ¡ð1_Q<Ç áR·>W•ýò1C®Wb÷?F‚òAh­sä>ÜiðÜ—Ž›Ò#ìH}+%Ò$æ79ñEdd›ÒÀéµÈwËäuH·ôM78Ëc¦Øú·TâÄ0¥ˆ]Iwt…e‚TãM¶ðRmªyë³8|&psU±Yˆ”@`頜ӘØ÷½¯  ñ#–~Ç( ó.MõÑþÅ/$ÕÒ/º;ÿÏ5‡Â¢BÁ¿—‹@SnxF‰Ì4‚©r¨£s’/óD]•·•P7týËÑØ»;EÕÃü˜§6å$UpW“oÃÿÿwdg¢x [fßÓÌåÁ¿ÓÏx Âd “$1i4ìwôn©àž±óÜ—Î/b< µFOjO)Wh®OfàT°÷%}2%H’„&Ãö`në‹Êâ…ƒ)¼ÿƒu-r=‡”P±·^ˆå80»ý¹/–fTŠ Èà7ÂëLj»CO4‚à°0Ó{ð*¼ÖÚ•ì[öÂq¥ÈÓëL“0–NEë™®º}µÃgRÌo£^}ý·Ä5øçŽRV<,—o½Ö*_Ÿ¹l¾Ã„SŒ<þ®ÔÿŽ15ò"aÅ.Iûѧeêüˆ£,6bõТZU\eî•äù…‘‚BÙ:¿F \ãíAæî#ÀÐÉÄHàly+WYß…ôí«‰><Žï^˜ùß0N¹¨GÍaD¹“R*,Æè×~Oî,ÛZ!ç Š¢ ¬q~¸¯”úŠ{anßsAÛDÜdªö„Žíä8N¥t×¼óç%`[,¶‰Eh„HÍ‚òÖ„/sî@S˜— ÿod}…ôÁTx´h°\ ~5"'oÛ²‚©¶—’æšÞ*$ÿÑQnëd7ÝÙyu”Ü;†eù’í"4›5ÎCùªøÈ‰©Ð6†ØkL…,ó缞͉:ø»°d ÚØÌÕÐixÄù3¹º*—åæ-¥"‹Æ ¢¯&ÁwšD·¡hh6ëéÊõˆ‰ Ø€hÀøb¥v=û,æ-Ò¡»:K­goÌù o ëÿ¦^PFžÃÊÆ%ŒPg{æ7•doâƒp*”&ñ'ˆÉ6Âö×ëJ‘ Œ9•“²×²¢?[gô"léã–¥y(*€iË.ùóF“ºõ—$§QMƒ­ñ÷z¢A`˜1®ÃÏÿèi‚Òy„Ë9!Aò'™Y‹[ƒ|–€¼!XeaöÁ°,œV-*ü—ÉÆ{ºu@b‹fÐS$™'C` )ú!:¼ñÐ,×3$$¤b£²m6u ©­‚ºËë;-HlA°%|¢à«jV¨@C(:&'æ^tІÔÁ-,xÙ•%<¿S¥±õOµŸÛh9&ÊÌ›/TÄè,W‚WÐ$ì̪KSÿ‡Ø”Š/ÊeãÖhÕšoÍ­G¦ã*¿—6p?5E8Ÿ²å86¬ê%’ûô‘µùCn_t-ý‘Ø5éò°ˆÐÜ iž^obƒòü³çºèùˆæÌ(ŽÜ•¤' ’dž"¤.¨‘{¾ïës ­À§·4ÀÄ€–¡ÍNWÁ9‚ÞfŠwôöhð*Ū†Œµ¢ ´$1ìÍÿ¨Îᵓ Ë]äFIâ6ê¡ vj;n&‰™âY»ûLÛ [§àüÍßeÌþ§>e3†n;ËÖ'Ñ,^»PI†¸M]ÁamÙ'v¥?|—+oJἜ$¯/6ëÞ†jµ°bgÔ¢ ìYLr‚³PK ­«S÷iQý†®Xô\IÚ*ÓN&^k‰OÝ~ýxû£®Ió%œÂ]‡'Ñ! è=ÔãÃ+A1tn¬QVck~m2P«…F€…Õ΄#zEþ ‘ Ÿ£¿—Ì¡|‚ åÆŽM†£ÀSúâߌ\V¾ÎVØž(gÃ÷Çjô aà>vg¹0©Ÿ¡N‡]œù¿Íýƒ½Ïƒo—¡(;ˆï=TnkÜ÷–™VªæKP”€Í+Ènjsì îÓÏ·z¶ÀîNØIfôØLf3%pŸöd/CjØHByGìÏÖ• ißwÅÏÀ@÷KÝpÉŸ×Ì9½vzÉu½vhCŒ,¥÷¤ý}í`®§‰—O£Y³ŸÓî{Ï)üËw4é'N´%ìé-ˇž>ë{‰KÒ¤+x Xo.+·ñ + j<óI–ÜúJOÉ«{Â;Hwvƒ!< Où…èvEÿAms»/ižÅ¶œ Ûu“èWÕUòÕz<-œS+Ž6|«‚]^ƼàT¹zx Ožb³ªÌ“ÄbÛÅ."ôæk¦°öm:š?l»¯ +oNôJÍqVy¢¥·Ji(Pqèì¾ß°_+"=Ãi˜=Ö2X‚$ô#.|Ê›—9ñxÉö»î©Eë«eã-2º:¥Q+V®|S“B€®ú¸¿²CgíïõİhiÍáN,‘wOÈ B}aù3´fë$Ãñ o¡ì·›fP‹nÙ9–Lçšþºß¦³Üµ†Nf‡_ ›Î”ïKì;AHîõÿËËmE×Ó <ˆâ‹!©g6C1BäkTò$•½e’šíþz¬Æ$å€w@I³C9Yhëcı&#yN2ºCvR€6PU@-•ܽFs®s`!ý”—‰Ø©ÇïáUê”EW-÷<Àü»Ç!Nâ?>hó9Žôk ô Öµ£¹[q¤Ø\:í9¹Ce×ñ¹Tô÷nª6Gc¯À63ä\4J]bqóKòˆ8n w;óð[òIÍÛŒnºWõ8<ó¡ñm]È%},ʘúþ}’õtðhæû.ð/òÁ\G9(ðçTjªÚTDî]EöñØ=lbÝ’ð§œG‹Û9àúO:yx$’L]ª´ËÈN§FFVÈ&WÛW:BUXÓðNIÃÜ꽩 C[vÖ‰Yó?Bt¢Np”ˆVïRjÝW‹’Ò HsZ,Y15Vœj ìRDlþyŒûû´ˆ;ô㦭 ê±¼÷- :,»òsú?B`E„?™xñüSTág2ëÅD¸#™XÁ»£ã_7©iât[MÝÏ;аn£l’sÜ[’©¸x¥œÞѯ‘1žÎ(,0M·ÉTs1“ü* ˆoý{R<ž® Ÿ.:x}²½ÖˆìKËtµHȘI¦”¼ák¿¿Ó˜YÁO„v`Ù¢|Ær:…Ð J†D¼éò.›˜ûí®{bù5]³Çê{½OÑ;]䢜áx{é¤&nÑRöt L=‹ÿnèÉè½à¹k%VÍÆiõ«ÚÇ{ jÍ:wŒ8ûmˆ2´ƒ•”©ålB’<†g…qü£[Ë’KKÍW—éÌ8E»Ê¸õÛ”ß{úÀ ªÄ‡ÎªøOõÔUõæŒ×NÞfõqë¢Ð”HÓ`Þš4p$Œ`F‰ ¸. pôãT˰,xÚ°ãöw.o¯6¬+3ï„â¾Ï¥†Øèíþ’Úõ×IgG¢½ÄR’sвp©mž —[ì}ó5yìiÍu—Ø=)î.›Æ«ðÐ/,¬Ü'S¤ýÊr|ýýñƒüðšÚ»HŸ/îÈØœw†Û×ÁÇ"š´ü¢R VoþnÒ삽¡­m±°2Q÷ª2ÕêâÌ¿mû®¡P#—Ñôn‚Vóq‚©“Æ)Ó‡˜¬ŸXµä7_(ÑZR'®ALˆTœ$(˜àŒß­˜h±ÇèŽç qäðŽ+ôÉÄÂL˜"ÇiÔå.Ùeþï`_螣ô7ެ?ÙÞó$~ç?V'ìwÖ{Rìè6ÀÃòY§êø*Ûp2?6k ¡:g¸DT-8zûÞðÃ7 SŠƒGÐkÍQ \’V}F8#þ‹ÜÞq~òOlkæª÷–ölT#ïÈ|©)²µžW•Á0¥ ^0ƒçÛ§…ø)T_œ{äô(¼&ÍEZp ò1Cø]OdðKº1Îá¦T³Ö!@øÿTŠóÚ½WfïÛ©üÛšël¬À¶1´þQÅ£DgmÖÆ(’_˜(ânQõ÷P“³¸úµ]ÊVXg»ó·T[6ƒ¼Îá¦&€r 6Á\Ý.º¤‘¶˜Ç÷’Æá(t&µÎÖ±O„ìïµ~V8¶³ƒëU—Ÿ ~Ld4ú>?Û( å»þe=®Mêx†º¤ž.O/R¹V‹¥.Ž“J›öÚ/ê\°´Ü8È— üµiG5+ù QAä8»¡sàè/3]âJÀp‹ù­5/êâe–ð®gì|¬¹}Ys}¨¶¡º{"ÅZJ6м±0‰T¿cŒëB7¥Ü)þH‘Ìëd d¨{Ó¢TP»ÑþÁ:m=6ÿ òAçÞ³;Z´Õ¡­î·ÛÑíàû4¡ÔºëNi†~0ㆠÙïDÚ-Áé» ÄæÔªL•ËàvYÒ´Ÿ>Ždc`ë 5*vàOĺ{¸¸ãáiYü´ƒËéŒ@2ÄîNV½ ÐÏÏýkpa!¤_‰d±.—'Í@e³ï/&U™48ß\ɹ!(†9™ŽÁtÒºŽª?’pUl9,çf¢š§š‘x&UTÁ]U‡œd7\™ž2½èý$fÑp3Ê,}|¨pâR<ìMv´jy+ïZ™ðô|xò*¸N‡ÑcOu9Çãqî¾ÕлlˆæhÆJ¼ƒÖÉ”BéJ—äéB<  |ÜØq£øº"ía’/¤Ó©ÙS™÷Z>÷áô[¬©k¦zÌðœ«ÃS[-a\·8lý}䛇4ÿl oÔå¶Ë¶)) <7ǃP ×y1ÂáqWkÜ€_¿à]ˆZ˜÷™7æ ¥ª|V ÁÚ›•úà0m׿¼[k#pŠ+¼C…\G:‡_( ö±qn×;8q*‡ž×°ÆüY(ž_Ãåjúü‰M ¥o›x@¶vÛõ„§“EçLzÄK.Ä×fÓõÎ!ð"s×°Óçšåö¦çyÐ#|¶xg¡^µOl\ØÖÖ·ðÞF/`å`^ì uÁÊ(ì~?;ʃd˜ÔÁªžGýËê¹›fiZ9'ß"錘”ñÈÓB³!-³•¦Ÿ ,ý³}C`Föx*丶çØv¿8yò—àJ‘ÓnD–û´TÕ†´o-Êܼý¿À?U©ËÃÀk-™bi±›°º“Qêsü*™0BÑu××½ÁËðH{#àßaZþˆ&Oó 5©~ Ū ½µ+¹ŠÌ[q0ºÃ÷ê‰ ,Æ™Ô`-;£@NV¤ûŸÖ! 7f°«ý¦ç®§ Õ jÙFV3w !é˜ ^îÉüÝ-Kù˶1û@ + “Î=_vWa6}€O öΧMÂqï¾É÷ƒ¦º7G„XÄ—¬fçi¾IvËlTúÓ>P‰6Ýw™÷Pû~òêK×V¥4 'Ó¬©“7ŒxVé<ÈÄ+„Ű´ÚtyÒŒNR#‡ÃGMz €ixˆöÂ(ÚÝ"ƒns…[âЂÛÌ.²ünÊ)(@W5%>˜'¨Éñ»÷™4¹5Íè[ã©N“Ú$h:«‡}Mg‘óFnåë/î[Bð{ÛØ¶Rn‹qÊ'ÌNêN€T¢F¯¢ÞÌ2’YZådÈND²~Ú"šjd¬n ¶5³cø{O×v+ÊŠŸ.7ÊE-ï+-AÓö?„—ð,yt´3è‹+XØžÒŒHœ–ÕŽ=‚1a8§‚Tus±ð’McÓZŒ_M/~ÿàKnƲ¾qìG‡Ü &!iÊ8'Úº™°óžôîÅŸ±[Á àî±ßk í-ž°ˆcÙ¶aœ¹·ß¤*_803x~t‰à|ðÓh¤i^òÒÉñô×~m½¼É ÓwéIþˆ·‘{ÐúÀ›6öþD¬ý.ÜôGª%ò{#à@¿T§~½ØŒ›­ ¯C½Ö-†beX•‡ëc,U$¯2™k‡´JfO…£:÷H¦-Íô‡µßg1ðB>#§ž;ÐR$oÄŠ—oð‚H…h]p6Ç,U$m0äiú:ºJü•\` ¯TUf_{åTúˆ†è.ƒœhŠ¡$¸sUèà›%Nè4NŠ«òû`qÌI¤(–æžÚ/¨Pá…ßʃ½CúõOÛ¡NhY‘‰0ìß féÂ~Ô8ã—Ò"îʸ}×G6%.›³þ<&ÜžÕ‚k\á=¤eŸ˜¦Þ)‰}ÆF^ïÒgÁÓZtÿ–²©ÁbBE£Nï`xTf´ýÉͬO‚í‹ ¿ŽÇ¶hkQˆ†\“„q±$äf.ÌK€¾«ßŒQr;¥Ð”,êËÊ‹«£ª?0¡·£‰u6äïßÛ?EçHœoùì»ÉäØ×N/Ȫ(¿ñ‹] !±:X×|Œ#YK\öFÐɼ~ªà¯ñÖ¢ñÀ.×}NÀ[ëè7¾®RˆµjÏŒüåFEt<foçYGT¯…‹þ.í,oáÌdeïmÄÂ{4ñÈ;‘bÙv¸èL"¯ž77;e‹{ô³ùFiªéÁºÕ?sÇ£Ñõáò–gÕ8(Øk–=²9õ÷F×p§ƒx²ÂÕ€L†vßq FÃ:î¸/«~™u]aZý»xõ¤“ÌúUU[ K6T¬ëÆ­/a¦$¸é,ÎZ2¨øÖ´à_6¯Þ{¯õǸ*/Á,õ}鲜 öol‘vû/7D\è$´ÉÒ°\ˆ{ÇРj£T 5ËvïÅêá ÙædvG€Qí‡Ro…0s©K‡GÊ$Š«+¿ö J H‰?oŽÝß\ŒP´ MD/ã;nˆ×·:"eP¢4™²?Ì?ƾ4]¥ H›@TÕðúªëúxåéD……ÀHÅkMŸ ºÁùʹÀ]ЇŸ$Ú71Yû;[:¨[?rGDø},Óé0&¦ÍåsåþšaÞI1´­òYºn3é¡jÕVF;¨¯0ÙÅEÒªw rd2Ó»ãF!8þr:MÖ[Ū¬ÅãÄ=äÌ9zÕàÓ¤p'hÞ*OÄTÍ®ÖBÿ ëݳ.‰"ÞXÆí„:XÐÉEp xûvÂVÈ•`µNóÆë {ÆŒŒrô»!ýåÊäƒcmý…U&Ú —ŸTЧŠ#Ó¤$@ä˜Ó0ˆ.`4]¨¸•ÐMã/ú“¿¶”õ; êå:úc<†»X©bºœ."†6¬ÅçÂŽ8§„QÔ$µl\-¼±0Ø%›T{ÜYs±œ(JÍ)å¼-ÃÐËf1rO^NA ²x#h©þæ;®¾qlÌT¿í)‰b'[ÍLÔ÷9_±îÆŽÑŽbwraü€J‘µ!P£axP$d‹Ð)ޱ˜Ê4@SR\Üðfº ÒŠw¼LØ&¶«åi :çþýcî÷WTQªË0©›!oçÎÇ@á5=¤hbreã£Yi`ƒÎèцAQ·;ý¡= [lèëÒô®Í4ú$„!±ù´¹ïºÌ¨MAÒU]gäq¢„„ÇÝ-·öZÙ¡…,ãBÔÿ¥Ø2@@Üš¸—Ø*5¿øË°ÏÉXœ¬{ýa¾]Då‡è ‘›„ë©Ìí;’ýô§œ$>P%ÌQ2îÑWæLò‰ëïW1﯒'=•⦯Ày=ˆéÍcb†»165%ŸŠ¯2Ü×iI• =9bOóïîÿ^² ÏšuÖâ§_˯ÙÃÆöb€×pÉ q‚÷å1}5Ò˜=`_Üv¾hôp 4% i{b2ßÇçÃ1^‹ËWj€¦²j®À£³dD'ï8 tîoÕ!j0;5oV"ÜÍ? ñdÈ—CC2ÐJ†ÿ{Be†3ì#éÙ¦õ«ùår¶/We§©{"ãGw…m²T¬Ìwírò6¹ö‰Ë†réDáT~‹Eô„»¨ÓÔÉ +9|:ã`›°ãÜêù ÕpWGË%“ïw9Ñ3/¦ÉJ4ÒþêM¶]R#¬Ãmõ­Ôx|Fh™ž~ OÇ.K]Ξøm.šÜ¾I°¢ bþ¨¿ºñ;¨ÐU¶6 _ñÐë6Ê Øa5ÜJ HÐP üÁ²²©±ñƹ"Ä0=RÞüH¢7®åò,SWÝù9rM‘#ü‘Æw6ç….ôª*†B³»•©çΗÁÂ݉ š¸°ldoé04S×CUQ‘—W9©|ÒÔüáßèHÅ<%˜úÿàŠûaúÿà¿©¥s_SCíêÛe•‡òJÖ°¼“*•?wû,ŤÁf¬çM&wAœ›”D™Ñ‘ûLÊU³( ÝTîÊÈÊȼߨã—Ù­éÉÙ‰-w»œÁáÛ ›™feÃç±iLn«”4úåƒÄKÀÚ»ßñ*Ë.ƒÔŸŠF“e æ:Ù%i“¤zuÆq¼ÂÔ)}èÁßö¹‚×½ÎÅÍñ^ ®Ä‡_ ”ñÇ:MˆÜFx!J'g®ŸI2ÒœÜ{ó»ÒÞ”ªŒïù}£g¼óõêæf÷Z…­)Æàn Âéu\‰Ÿ3èbžç’ecŸƒÿ9É¡¦}ñ£Ý)ÙMK–ö4„"ŠBåê4j¢Faz;½êz¯#©ÏßÈdñ¾Y”Èñg´â©ò[ÞãŠDry•FÚԔ듎¨x¥ä1ØódûÇ =®ž§¦ˆ4d0T&z†Þ”jÿ(PÕw`*û—ó-…\?S"@œÜ”†³å\Œò<@Ñu5AûüÏlp%QÒòc 7šº˜ËŽ9¸Îù?tL‡›•Ú+(ôôð.ˆ«:`ÌÄ|¼1Ö6¶.÷f,ê~\ c2'Ó. +Ã$|“îêcðlìD¾‡txD‹Ï§GP~èùäéGy¹üç4Çw«ösi$ŸCÓ6FÓƒæRg“`ú[‚*+Ä„‘ZŒ9úK‚«Yþ¾ ²P%ÛÔ9,‘™e^›ÂþAéd0Ë>–, sH/Js.áeô«ô‹$FH£ú£k¤ê]ª9'(è¬B¯!ÿIš¡DtuÖF •úH¢øÁæiä¤:Pë"Ë­¾x'³¯7c3Þ-Šþ‰ã5ÛÔ[?ÁaÜS¬RCªX ð 7kJìÖ5óí„;±|DAÅÌ>0X…Ä—l›lU5—ðµ.°&¨ù¾»²mã›81ºº?E1Ü0PPlÑ)«pû·û\ ¶vyH*gsrª¨6YY ÚŸ¼ôK5®e#RÚÙø"ýE"’pßIÄ<„âø'ˆärh“LëaŠÖaßRZʃ€×–¼žÀqÅUUš†÷ÖS@Ÿï˜2œ%ª ~äóÂÃæjäAÁT~3«ï±]C?Fo9Zzbhăè6?(Å)ûš‡#?Hkùm¬ ¯´O¾´%Åʰ'á$ùÿ2ô²qÅ„’AZ:Ž÷±ÀcÉàžqPf{^7£ù¯á ààD†ö×6³?ì•o˜ä+Öö\LÒ‡2Ãm5G#ý=‰Ö}fÛ½×Èñ†=ÍĈ&;AvÃân¹eØÞ⛋ÕGmg]gϬ“öð@Åeb&Bê7‘#¬û‡ùµXŠ×dšs’ÊGåÍ ú̲geÌdv\|ÇÚ7aÔàp¶{ØØ¬ü"K±•FþOjV±é«Iåºt^¿"­58œâÄÀ„hgW%$ÜY*Îô3^ŒéYJŨÓŠ4¼©[B1zK‰¯žÂâåþ·:ô£˜ñÞBEƒó?æ%>f:XĪf@p£¼ûiݳz$óÙ‰~Ͻ4³ÉQ¢¼’)Ï¡¢=›È@.0nxãòw_…¤¸4“*ûìío¯¶¨PÅZ8yá­K€‹:(eZ¥<) /Ý–sgC:°`·Ëñ5þ  é¦öØ{ˆStR¹ Ëu€U،ۋiûTîkݘàTyž`­0«ý $;1åYWˆîápç\vΦ·°y×ê‡û/˜Wx¶ A Ex8“mÄ¥ŒbýéGü&ëÔIl³†„¬!Õ=ãø]ó]Î0²²è ö ü¥àd¾l­ “ÞüÙ믞ƒkÔ(>ʲÓunÂ[ñ‚s;ê¹ 0Ûí|‰AQ¦aœO˜íIGìÒ–maïx½Z>EGi¾·4´U'P‘:¼ˆËbç å²â©§k§}Ze=øŸ¢5Å-þmuo·ªeh¸Í,:î =Z0WTZa}Ë¡[ O¾Ø°¥ ¦Óßrµ/Æ¿ÑxŒ×É9WMÞþý”úuÏYj‹ß¢€‹Ýò0±—8]f,#Е·U5Cjju&¢‚Í\ÕÓúÞ® kJ¥Mä/½·¼^6Æ Ý)ÚKn÷ˆŽrnv7,>ßæ0#q<ùª@¾.±ö~Œþ¾‘yt9‹¯"ØÎ?fbЉõ*§oÌæ·Îvg9u4“þ©´FÕY„3™¨$ J£ÖøÊ1”¯é9ɇ‚BŸK ³%Ò$!dD"s½¯D z¾Íöš¥ÙÎE€<Ðdà뺲OÊ”ÿ‰-”|Í.¬Æ§`®—X4îiÉH¯öò'SZi–ˆøpµ{C‘”f”­"ÞØ“*iøBÆä4¨+HGJ.$Eˆ…þ=: %3ÉRóî?Ù—²Œ|]„˜ØÿÙ^&¤^Ìɰ*Â<Á¾à¿þ¿NeÙ×5¢ÏèÛš¡ƒèBÿ®ƒ±eú1ÿ]—û§ÿò{K¬i´lžë Ò#Óã²öŒ™õ·"5(žF©õµKÝŒ_ævœò/Õójd`ÉQ2¦ÖXÏ÷bwGAlÁ¿–‹Â–Å]‘JdÏÇâw°¨¼¥(àD©É½„üÊçx ɹµ ts(ý6]$LÜЗêbÛ±šP‡>£دÌCiöf:©t9ë(O÷jÒžò±8bp ?D³?V‚X¯púøŠ/ -Gßo5ÆÓ(ùš%„€-LHûöÙR©‹ ©+iü®CéŒÁÿµÕçq½Ãù*œåÐØ4‡±Z›E¦ÛoS¦lvP¥º¿*!TØÉRĺâ"Àåª&§lèN’î(IßZ•У¯žI5ħy÷n ¦å…Ñßž½s8wªãmª-”Ö*ê°V˜rO'Ì—h¹;Â#§Q2¥­Ø3¾JW „«×äZ—*—?ÑnJ±%4ä(Ð×Anã{γUΟýp¦ÏnH»Lœx:ˆÍñ)ùV¢ÙD CIçï’—·4 /<‹màtLQ‚qÄ7xÔüuƒ^´~O;Õ†¾·4@;@ïÂśɆOýL7 cLGÜîZ6V8²8Õʈn)·€¿äÞ1Û_$ɵ<&±ÆÖ[óG¼MÎdR2óu*)*ÙI*I_{WÀt½<í¨<=ˆ€„Í) ªqªßŤ¸è(z$2û4$‰M™GK푘é«âí¥.mkeÄÿÖäÁ7~–Œœt¤À‡…{3è3¶z†½4^KÅ%wæ_Aú’ P_>ê$r—rˤ¨æŸy#GÔÃi&›èú‰¦a‰´m~15µ j²¸«à2'hH+xÍ™ó¯ö½'p"LÀ&v,C*ž ¯Å þ k‰/ÕÐj }äj6¢î/^‚D,/O@ ªdÝ%~Ø]ñTæÞ±Ž'´pÏÙ?ŒšP³p­¤Ïü?êbÀ½h¿?<`S¯ JNžtµQ´ÇÖP„Ô³@é½": ŠÙÇÌlä‚'À¬q!ת©‚wéÌf`· d±Ø`¹ PS°ÄD¹ÿ¯K„ÒŸÓ„‘cçá:fƒîˆî´ë®XC÷E%‰çÕN²P<«5wÁMج‚ûŒ·…†1ã¼gF.„TÿÀ nPRb—b'…Eq£$C2·× X?’kÔgRt_Ýnç$þØ©½U­¶Šà=æyKÀb%‰›ïÒöUѦµÛ[èØ• Ñ@°Jñuqoƒ0ý©¶5ìðõþë3g¢Äù$IÉAÆ=R)pbÿó_7p*,IØàSÔ ¹´' ’K±Žpjp¿%‘õ€'©Ÿ wœzz¢ÜÍìâX“ÕÎÛ4ûCÚ#õ ÈNY©|cAÿ?öˆºZ‚vö!~¸³! š¹ó>áí7¨Ù_C ŠÄ4¢*åŒX±µ.û-H ¿‚ vHŽ+´ ÿïmXÚ­ØO)Z÷²ŽãŸ!n³f {é‰>ÅS<´©ÖÙøWë!O%õaã€4[ ³¹ ’ð~ìRï*ç¶Y™OBk®HáJ[RV ü –,0Ÿï ~?Ùëþ×Ó ªxf/­´ÒF U™)­îS*æMŸ¿©I}‚ "wºZÐÂÃùsèR~TòŠ3ß>•óºfÚ=aÍçG¸¥Í!-N³êkQ ÜPÎdòeÑÓ£¬Ø\ñ=¦7ËW7C,Ù]°"ÀŸ ÑŸÉ秘åûpб½ÈÕÝÓŸîÆ’oÙ1¡ aêõ–”üºù=?!lÃi$`ávk¿æÇÑ>€$Pü#碼WÑdJú·UįmE5PÈ‘³ŠçàçTéäÞ«_Bá[ïF•9Œ‘Z¢2A½MÁ`ÀþÝΘ¦ƒýÅœ;ç€8ËŒàáä5³ŸË›JÖW¿ Ëº£PÒÞÙ~8ýa‹ÿ­"€c<Gk«ÁÀ˜ ±µ Ó§kÚ‚NU‹R¹×ïŸW+ÔŠma×ÞÝGúUZë=98£ÔØ®v­L}™Œ¼ú´¡¼ýa$޶/Üõ׊¶Œe  aÎ5ÂgÄ9SOR @Qáû¢'î½I±‚H8ßBµðÕ!Tà“ê>2?tU°xrã#Þ!µÈ„#ÇvmãÕ“ÛSÀtËêK XA->øy1ºÅ_:‰¦cö.þû¥,:¯\7¥)%t³(ÔÛ ùà¤^:^†ô°m«†wˆôer[}Û`á®0]܃0ÙçÉ#ɼ¾®õÖ&9žs¢w^ï^êªxv¯ s•-¨JO–ÀŸA"æNáÀØ3{¼sŒHkvîÆæµ·BÐy»·ŒßQ¶óÞ¦¯{Ïk]ÏJ7èñ¶/U§ïéÈè± Ч«ÍÿðñÉ15°e®3EÆœ–ß"n-nØÃ~èF1á§«K`GÙêOZ½9XÃQè ·nŠF(¬»mCúÃ=i #`ߎ™–ghàcqŽ“>ñó Øµ³@øn\¢©*aAþ*,`n€@²èÝxÃ÷~æÙúÑÏ+Þ eœZl7–J³‰õ‰ýÚº6P´•A^‘M@²ÓO &ö¯Í·q·I ×À_ Ï?¨t7;ÌkO=c? Ñ‹W¥Ôhñkm[Éã+×í-Af`?Ž áCYDÌk¶É%žƒRæ5©gÄu4n-® oMk_ ÅïÚm|GãµóŠ”Td~~acçûÖi4ùÀÅiC”¥¤´Í¬/(ÞY —–ØòÈósƒ;ÈRIpZcœÕ6ÑÀWÄSÿ¡¯Ï—ã©»¼ª4Œ¹}Êõ½´ó<5(ÄE•É=d‘yqübŒ-ïÚ§«‚ÆfU! ¸Ñ5‡Â𢻼øÇsïAü9ÕrƒxÅô%â-üŒ²æÌ[½žWÎoÄ"|¤‰ ¶J¬E :üR¹ÜÜóà }l˜øfû:€¤¡¥~Y;ÕÊvËÅMÈ9nóƒ~7­Ïõا®Ø¶£ùM!ñ£ÌUfNtâÅ1w» Ƚ.k£Ôç×cߣ–êµ¥‰NnfG%E¯¢ˆýÏ-¡$É0j>èÇ%Ö*1«—ϱµ „Â$ž-™ÆÁWñ!W•2¡Ú˹ErMíÜA5®7$æ7¯.Ãôâ´{FÐ'=“×]½Á¸c8’uI3ÇDx>¹“nÆÕÎÕµ}ejÜN<¼zÑ~“ZkÍË™£ à±QOS[D >Èâ\ë÷ˆ¬_Œaî$&?ìúˆÅGV3DU™çEˆÇö>Õä25Ø®‹Ï¯ùû;¯ßs‹w+ /¼áUå´7hESž=|aÞFŽ¥SÈ· ˦èP.bjeReAO^A¥m„p› &ôËçžÅP¿*"4²§üÒ¨Yí}°U±]’U!Óm' ñÖòÓ¸^‹1§öyvMÅ$JŽcíK\úÄ8[,yQZPóÅŸ‰;XñáVoæY·6~cÏÙ¬y ±8AÖV>HsØ’ùìh{L«è\­iT¥ÿÑjLZò'ãÔ•8 \ùßæØ¬UV™e?Í6»àáFÍñ&êQ×2?ªÌøsmg ì¡)š®À :¨ÞY1]Äîú¶Ö~rŽ ¹ÊWÒ1ÂBÝE°ª6w‡7'dö˜4hÌܾ¬j -#Ò³E¥l×e›Œ]z|V¢z¯¯¯W%ËKEuå-‚ëGÁƒ¬ÍˆÅto€Ì {KN×XZèA)cŸPC3<ò`¨H‘"\Òv½¦Ë^÷ VÏqês¥ÅmyÍf´Âú!EZüTg*kLµ4’pu’It´LO|ˆA½(½X‰m½Æ{"ölv •/þvÍÜ›æX & ¾®Sv¬øfe_ËØ£vÞ|ô<c·}S£p¶ö¨Ä25 ÃøÙgy@€Z ÔøÐe¶"â/êTúÖb×~€ýE/©A Ô3÷…PìÅι~˜ô ìZÎuq.UÆNº›Ü-‡]l({G jÑ×ÛÊLK ª{‡¯l…¨U1稪ço¿Ç…êßd[‰á6Èß)û§Ñ.þf½×hhf#¨¢­½ZP²¥¯õæµ—/U«W;<çUam‘ ¶AIcd«ü;˜•æhmXRvY² Õ5áƒ1“©«ô)÷« ZÛB†Õõ¼©þ/Ï¢u[Tâ€7 BéN†À‡ó 2‰æ´ù9•}ÂQŽL/ÅA¾¦ m^ØÚBÏÆŠL¯³ ƒAÛgá9ì†ië$¥¹ÏÆ«P6sµáÄ‘ :š4FßÄ ÌXÐïÚ3þëA¡BÛ »W¤õU^̺¬$32qÿ›Jqd4Ú–™µhŠÍap.ð!ô¤÷ˆmyIù0‰ Áò’töí÷"³vî$ü·ëñØ SÄ oÚä`H£9Í*¬®¤!|cDÐmÌ혆x;b1²1ì•0ü]è皃žkþ1_“çêæ]ü’”ŽlLçú©l®¬à“]o³j[° ù£&Q6…m{>Ç£ž:ãÊ,bÒ‹;D /N^ ¾v¬—²[¡eãþå®ÅwOÊ`j¤¾¾Ñy¤§‘ Ü¥¢Ýôt…g£jN3`ÅÅw0¾ð_gÞ=PáÈÉûHf§IÁE ÈQŽ=Ãl«ÏÂÔ×$ºTåŠôÂc¼i:<^%{f”~é±äcg£tàè‰mD¦;Þvj‰£“gUTB·ZË–i~‚®&”»#é‰NŸÀeÜx þê ÝokÙj/q¼€eK ³æa3§HˆÛõ¬nÛ3\UÛŸàF'’{…!„¬¬‰ ¯=<Þ"ËFn'±çkÇ> ®éä¦h“'QÎ{AE…À€Kx]™ (1ˆ¥“Ö!œ’DÚI©°¶ÔàmÈ´(ŸŒ@·Pqöp…èð`øðÙÞîHR¹2qe˜÷ä¿×Ë^Iê H£è·¬Â’=E.a>65I³tg;J°¬>+§ÚSF¯S\ÇäÚ«‡Za ‹ ~ ¯äËþ#ËŠ<¥biMÌûÃܶ±O·Ž}…û‰ðDìf1úË_ã¡ôÛ(@}•˜7ñX1•À0 ÏÀpøuÃôáEüµ°…¿ÜDkЏP¤<=.Áí¥“vÆýNþn råŒï |…«M>_¤ä5Ælµ®‚î¿}áÿ\xóÛË3²6$¡aV6aî¾u;%MÚ×Â&êL®-‹°];+–ÐLM†}‰è Uø„ybæ%R' j †¶taŒða)Ü”@§½6R‚¨ÀôîÒ)±€ÏK>LRD¼3ÒÅÄtîÀÊÀ„ ²¥¨Ò® V8à-_y«ä_pÒPœd€Yï³òFû-±Dì.ÄÿŽa(4ôVaÇS”³Ðëã×#̳é€ ÏæfþZðô]#Ö#ÕSGé`gX³5༑’µ ™ÖÜ×wƒã“Ta<š<æ~œ¶â£kƒÏ/ æ5„œ%'VUŠ?lá$_o¤õ™† É ŸHéi¸ÊÏ[ GînõD¶ÍmZ™I9äêºÆço·ë¦:PgÊYÜDÿâ9÷ë9í¤ö|Æ»¬Ú‰jâE¥’ÑMüdKÖø)²öêP#\F2Ãa‚ º¦„–CÚŹüÃq¾6ÖÉÌþèý`¡F¿äɃ³Ø—œÀÇW Çù¾(`´“èy›j!J,à6ZÉhòôPŸJÁûDÙTCjŠ­cÅ2ø%‹ ë4$†ÙÙ¨‡j{´z‹w‚»Fò€Nr±$­…/¿ÏˆmØÈý®æ†Xa>ÜÏ âÒ“ùW”zIö‹-Ñ &œ±£Që¼ð Z>]VÆ×«Ê|Z\ÇuXŽ0kªžãe“¢ÏÍe®eÅ-ç×±2ÝjXî{öaµÞ&#ïÐØcy á ;=<¢y¿>S@çzoX¡š\6°‰]3„¸=5û:Zvàæ»ùFâÎí< À»ÀIJ®rëigíËKZó4ܾ3ÖðRiiªßdû Cfü|íAm!ú6X`’Ŭ¨ZX„nzÝqÂKz„iÝqÛDž‘®Zƈ|¶Ô\'J:¯rÅkME’&!fœm#K‹8M»gŽi}ñ;¤|DßõÕÞWa)ˆÉÀìWצa³hhë'ò{ý¨?ãžÑ2·ÎÁ¸^ލel^=ÂK ç"Ý_–ræÿ¿#þž‡âðœt:9Âg9˜¢¯Úû‹Î]uNVdº¸AÍj™âncRÕÅz¢š¿ÐÜÎýL¡MîY‘ü‹,à¢Ë,îjKL6D˜òU›çØá© ŸEo°k¶ª)·o  Ñ-~E†&7ƒ0”™WÁ“êÊ_"ÏtwÃÛE¦ý#ñïŠ<ûÜÔµ¥Ù¡¹©`¶à+7º½·mD1ýd¦½Áw‘NÝÊËñólÈWûˆ A:âŸXªËŽ£tO39µÆºe¾kÊ[‡‚'¾·—éì[Xm{|Õ×ëÞ2‘ëü7&âEYÛ ÛÍcâÙÇ2|v[s€Ÿ.8y£I€íú|ƒ,3aœnMĶX2ðlo €Œé••é¦bg¬N¦Ô s—OtÇ+–"¸“„^Ošÿ,ص óõ¨9SæåvN‘ŠQâ´M€ü8†1‘í,a…~6@©‚‡úU‰Ò´ßþ£ºYX˨žy#ùNï½ã‚øeU9—ô?ø2=.Â#&ßÑWùf?¡˜ygÝÔ‰¥¥K„m}´C$ŠAŽF™k${ºW½p^½Àp$e­GV„§ôq3ÔyεÀû¦E›ÕúÔùÝÀ«7çäd3{îÛ}Þ[wZMƾìNÂ$úÂÏ4QOF¯I´ÏkÆHÂ]ˆn7òˆkÉÈLgUD€º“ª½0䆅ÄÛ¯g5ÍÁÌ˺J¹¥ãاl3ŸÜ„ÒO™¢¸å¡ùÏiÙ÷¸Ò„S+è×|׃¨–l*y«€ê—îkW}u?%e™o5QÙž—HÔ¼Äöž?"˜µæœF߯S<àö»pп ó„Cè^i¼1z_|¾å¼7\Jò¸IBõpñ *¯ý‡é0ª0ÕåPg5ØÏˆ0à'6ñöj -êé$%$ðí³Vuä ¤uµ±G‡¦{ÃôwwÏQ·©‹ËBv€^Yý…L·?›j™Ázæìµ{e/7•Sbsßî£dLìùÑJ{Õ0í XâõïP:>7>Š4¼ Hã_ë÷5~b„ R4Öåhòî4ÿ¢Š ^yW˜ºY|…^´w;ª80ºãÅ9Zü ›¦ È£/03gÅ Ú«ÆJÌ»?–øº.±Þ'ÑŸw@ådˆÿìjö©‡P{ˆâV ¶ñÌÝ—yušÒƒ§ÈI Oø.ïgüߎþXš9nžÞmÒ…m7$ P³Z˜ˆà?âzfµp𭫇ËðR÷j ¹¬Oì¶Øqæ °î¾Èá\ÏsŠÓ©\YüªÙO/-üÇ_±V‰„Qn˜Of{ÉxŽP!{0&WÚkæõæ–ÿë€ÑÕÉuv5 —gêÜA°4C;9OÒq‘ Ý (;’GüSS8Q?Ôô.…*íGl%Œ)» WŒ+êÙ£æLbÒèºY‹_DL¾aØ»(.SÿW>ó§ú!Š5+ÍJü6­0ýÒ¥Nä¥9.›¾í“#°Ï§'ß¹f9ê¡Á2i†.k83& Ttäε ýÍ¥I‡”ÍzÝMâ‡àé@¬VÊûŒ€çÍb»ùG*Ðÿ ˜ý¸† ^Ò ŒYn$þèõ, ñhŒè»LszÙšˆŸ2NÄR³Hy¦ËF¥Äæ2IÓÄG¸Gþuà0qæY‡å`–ZÚÂcƒYÞ«nwéIþkÒ†š4ŸÔ&ÞVMÛyi6Ñ€Icް7½Ðìgƒô¶&~;céÐïRßäê‡1©Ó=ZfáÄ>mÕ”a]Èžv0Îè+e©­íiS. .»4½‚ä‚M ¡Žv˜ï¥Ú*V‰cÏðËZbÃåUsl‘ÑÔ•€!~¶YÅ׌Þx'ézÜ;ý˜WŸÜÎÙüÈ7™pÐybƒx´ÌùÎYì´¢ˆþÌ"7Ú÷2’þv¾ÖЦ?[Ëþ³uùÊa„ˆÅšcVG†æ>¨ÎƒŒmŽ?ïpdº£ì¢`éžGÚ$`PÚÆurÜÈsŸÃîëõ%wÇ&‡z ]3m¹•I3×ê8%U»E·§Uþõü·+ ¤dƒþ¦dYð-Õ/ˆ IÒ¤ ßó¢è_¶£G#¼šÛ°KÒõå)?'š `«‹ e¨Á ÄI¹c΄*såÉFä‰Ð&7¶'®“Èr˜A “C£ŠºŽ²S½ o¡K%òr9¾OpÇ>$‡<µÁ¼æ+OZÎ Ëÿš¢çSvût»^Åëz «Ñœ+ Saôc«·0†mÀÑ]×Ñ^¯³"*â2i¯¤î—Z‹7 Ùµ?/³/ÌÃï²ÃÁY9‰›8ðqâKÑOkêe., i ˜d6çˆu`I3©Êö­mÒg:|†Ü ?Ù2¼”]ñëHY=²û¸cqšh'‘ž‰ø„`e^Úþr¿ò›lÛ1_,jà™¤~_k÷©ËCJOî à"SÆßÞ@`MKÝî]PÐÊä;VjªƒÝÍT¡îc]8ò áu=ú½m´¹’:' °F¦Ð;¾;Çí¾’‡4@þì~^Zõ×{mA¨òŠÍlž™CÌ )ñ‰íÂ=|“삃ì¦êÓ1Î=zú–µu)Ž? ˱—td¾>ú%,÷…H’+˜Ú€ŠÆ“™äÖf€Ã˲ë–µ­šg™˜7š÷u2a¤ns‰ƒZ% ÝGþÊâ"°ÒpÔøBæ¨_ú2zò$¶=ŒÂ†ÛŒûP„Èœ‡š.ëÊB½RͶ,¬!pæp8c´J®cØG<öûHa¶¯OÄ `r·á·gj€"ôñï¨õÂÐ~áQ´Éÿäà,Øþ1WbÝí¸ˆñ‘ìdT²5:̵ÅT*!êræ»ÃJûK ËÑÆì26‰*mŸ:cãºwnÄá e-&„i "ŽÅ¼hÓƒUF~Œ±kháÐX”äßÕtÍléP.fjÒ¡ë`ËCæ¾zB E™8ÍMØàèA» ežÆU¬!•ÑW†¡¤¸ÒzÂã,,ƒæë%åⶃKï+?:(ãÌí¿w(˾î'Oʉ¼ÉEöJxTïË'¯NÚ™ØcÖÍqÖh,»Ð ¶š¥dÙ½0* Qw|\¢Ø¯ Ö &ÑkšRqσ‡ùkz=/Q8[^ݾóq»”YŠ•oÒ£íænˆÇ†Ÿ­ày[ ‡ÑY«>Û¿ J¡ÐY#°H’>A¢Ò ³×ßµ3ïu ähËì|ŠU]ÿ&͈jpuIä LÑÚT‚iÔ»œ,ƺè:nk å&0%îÙ/û­.æoB9ŸÅ.~¥ÁõsH¡Çj­”x¸:F«ï!ÚøÙwþïÇÍ/[L!3Ý•P7ŸBwãò9.Æ ø’dùà%*Ê70ˆ|ìd—Õذkù܉ŸP¼÷)*ä p€²—#ì3mÓ¿~Ê™£8—°²0UtÂxž©ÿU •U}ø7GMcÈ6ä]Ì "@°<›ºÞ³n4ÆqvyjNhç£Ân›\\ŸÛÅ ð…:ú˜¨'^Z-™Ç©Í„šºJhËL§$Õ'.ïP&ÏÜÌ~:g¹^Š_=ߢå¢EpöåÔ/Á9ÈšŠëN·¡’.ܱŠt¿Ôò¿€—1çï5꘨oi¤eÊ‘‰oÒ…­ ¡9Ø 5];~ ùŠ)¯9ZŠoñôŠˆ=,ñªï}ëßÊÙ5Œ[Фï6ZÒ~€¡pn.j “ŸÊrðÅqJß¹7Qàö~¨,nÓlVa7"­ žŽ¹ ä*AV«~ÛZ‡ N<Øâ›&ÆsX*f6þË4ÚèlÉð¦”\ÁéYÑ/Ž>¦þ)]9¿ÁUtáŽd¾ªÞ—‹ÜÃ,»ÒД·5·â&ÇrY§ä†Â¤OäxqNnÇnw ç/ªá_pU^èøÎ–˜tkÉMÝÒ„åúÌÿ2káÖbÍ_ÎÝûNåAR»Eõ*»OcÉò7ê~®?ÐfXAŠoÍ®ò¯¸¢º…³m^Z=f9ªÁú’ÞéWjÖa3žÆÊù¢gJšæ»V¦?0©r˜¦YøÛÓô¥¹£HSÖ-^µÑdÆ{†Bk¨wtëÑõÒ]¸´wì7& 8ù ݨ²k5ºˆL"ZxétÉÄ Ânðç•ìµ…< ú½3*å²px¢“qlhT&èÒÕR‹P  D]Øz?Ï9ûÔ—‹¤ˆŽQ{€ÕqMXF³léÒ^êœd¹ÞN͆!° jºû#†£â¡“AhjpnüÀ‡ÌåŒqÊ÷õDêX¤1ke¶¹é;½Î[¤\’>æ&šÇÇÇžxWºW BÒƒ]µ(V¸–ê’U&º Ú¿o¶Hé$Nýt5GE­?þ†€(ùSÞ͇)&ÌÇÛðKwé“’mNDuf÷§ óºÐþåÇb5/³TI­Ã?ÞI`3Ê6„¢×•q>åz3‚É,.òè%‹ôÊ¢>¥¨ÏÊO¨B°˜TÁµCyÑcK €Òz gC#}£¿Q!C˜¼2ç>1Eö±)¬ôˆ^È>ÌåLi«n·mcÕ3Ù˜èÛ ¨)–gQÄÝÁ>¹CHáA ^áÆôcn—‡ùÇ?a›È–P]ðJÁf43ûëvXN0âà– Ö³¸æP³ö[â–$ Õ{yOõª'*¦êK¯ÑMž.?Þ¬pQÙ¾ÌJTÕ—ÏM+Ôþ³èŸ£´V6âax"÷ñ‘(/ Ê=cðí%¡æ½g²ñ£úw§%X¥–ñtÌ¢… ì# Uw*"»ó*:!¬àÒs‰4&*->àA?Ï¢˜ã5¬™ ²0½ÔJUïzN7ø:ŠjáxgÚîíŸ@›1?WŸÛV’ël“†!rØÀÞçæè³Å—íaŸiO;i^ã¿>»¯ÃÅ‹|€Cœ;r«‰3|KØÞ žÓÒ¦“hãøì\eÅ8ãAõ[?›{Ú®Í+ùà ¢ûtÑ‚}‹µL>> Cí¥‘gˆ"1z\–•:ÎñÇ_ñ??޽dG ^Œ$êá$[ø© ²~ëÊÙ.ZÀñ,ˆ¸#ëî#Íÿ0ÆôXÐ*Zƒ/»!ñ¹ž?¸‘Ï›0u²~-ùäQø ­£ÞÚGDÿ· D+Nh€u6\æ!’³y®²“a_øiì2݃±Ø_ M£žÆÐ gƒ{8³i×Ý¿Döeä¼§+Ã’¾)QO>ûOnº¯3¢ ‡w¶/[™å‘ú ¾ñ|ÔèŠBH{d˜ÝÄÕ?ÙÛÿ„žã 3Ä}éÅÚÆ—±ŸÍº›ãç7Rzé¡þ~ ý0MN'm›žyëÜr'5C`;,‰¤…_qâ*¤mæwR%³!•sçψéÖý<Ýõ^Ø©_ÃФÂPܘÕî%9Ÿm”]î @ħ‹@ Šhmõæ.{ž?Øa!ôR#à¦Èy:ƒû´>ä$'ŽCþRO.\Ÿs­»¶[ßAzˆ—Mz+*NnýäFj´ö²Î^¿©æË)·ëÑv¥-7*!Y×±îr"±Ó39ãîiž—•1óh àêé3ä wƒMX),Q*öÜúÔâ褱ùz[YÇúíLP='€}¾ái¯9b>€ùœ9I‘·Ìr0ÃÌÏÒ†íX5y"k;$¡¸ª› ­—F˜æÖ¶ `9âÔb,±uO®µý ×/fØs¼ªÆE¾šõK“j¿”ö¥íaÔTxs“@mXF9Ý<\nõÖ A6õœ¶µ<® 0ŽwÓcÄ- P“îöŽô»áÐ}ôo§§Ç~õÓÚÌ«ZçÊLâ)w_ %%±âñÎD¥gˆÖD/W,ÿÔýN¯<ÅÛ>º‘ì<è ásc£j´ÙýŒ×ô,©kàÕ‘Õv}Oi›3K¦ƒóó·›ÕÉœ€bu{þ{L7i~üÃ…ýq‹®ÔNµgsË@Ý[N¾]÷’ˆJñöñ‡óeޤ¸&I}Wêf,GxÎ99ç[76ªI•¬«qV[»9£îïg;‡¿SêD¡n»þH|'yæ"ES¯èíØ#ÿÄë~ºþ^œ‹CŸ)rÿ8Byy¦r–ƒlsÉqÚAäÛ¯j²ë\s¨añ¯]L`Û²_+ÖsüÙ‡‚!¯0x/ÒVl§âX*-]’³PƒÐò˰o'ò Ï&¢˜¢L] öÐJ³ÁÝÖ5Öŵڎj5ºÅšƒ¬5¼'Ä’'qò à9… 2è„~öéÐÕÎg@dé4Š×ŸÑÀµ½AßÖOÉÝ˺¦ê*ÚüÑ+±Æ|Ø[‚ÛfS÷Vû…@x"êØ­ûÃß#ÆNÁ×ËõÇ—ÍñX\N™aIyÔ^íê€f2 ZYØ™`É…fp#x§Ã&èON þ=óî¨þÍ¢7âÃ/-.½å5ÜêÉé¶S'9W~tc‘ >ÙHø~EÑ‚za£Q&¹÷̘Öo"lHw’ù­ýîòzáä—Y:¸'‚Î=´Ym¾ŽmÕ=æœp­S(j^»WVnˆ,ÈKõ:äüõk(6ýE/‡-SÓ›(0¯£ $è];µh®¼8tÚŽHZ`ô"íCÝÉZF»RœÿR¨v¯)\´æßïEvs“«ö‚ðå0s?Þ¢»?—ÒȈ¾9 ÌÌ¿þÿx¸Ômi”žHcÜH[ UzsÍáu=õŒR=½=´SÆ ¹Çy> Ür~C×´ÎQ?þ©©u[͸`}4 ­½ ÝDöØŒôÀäTË‚FuPý†"`×—òh…ˆÁº%°SŸF>0 ‹YZks/data/hsct.RData0000644000176200001440000060716413265504400013473 0ustar liggesusersý7zXZi"Þ6!ÏXÌã9‰ïÿ])TW"änRÊŸ’Øâ…§ "©Ù"Ýȵe|ÊÄÝá|¬ÀE®žåwpƒëd ’€6#»‘r@Á!Ä:ÁU¹° ºÙ9F²…ÛK=üâ[.‡VÛÍ?Ã4‰Ö µ7îŰ<œxf©Ú÷ñä™@K<uˆ„Ñyõ¡ÞN[د wóÆF¯¡Óµ?¦„÷ ý™íÀGÆ u"gæ¸$ïÕ8ÂóÞùœÎé|1Èþ'GmƵ-³sò½Õ8H1qH_¹SbÎß)"ùhŽÙùeìÚV©4в÷'á 26•$Š7…ª²¦Ýøý&·¹Û as°öM>I`CªOÖôäEè׃f]Ù¯„Óº(‡ò5#¥2¼‘Î<+|ª‘p_=£»ÎƒLy‘É=Egë̃^£Ÿ{µÀ-åZÓö±¿î\eTÿËr^Õ‘X!걩¸\_¹xN¦‡ûØs–”èìÊÕô:n 7 ˜Ê“, ‚@ì¹~jxCÏTÁä÷;Ýôti›¸šíííË_.N‘[§wéeŽi¿Eq žõáq{+ëMèF¤•´ÅÑN’¼æÙ…û+åža”­—£ä!¼f€lÄÐØ–ãå‡ÝL§!ZØÇ…Ç–Bסw‚£6wpg=8¢Ñ—éwÈ?|…„â4E™u¬9:“«K¢… [bºöŒsWW÷¤' ”ÄØ•º²|ÿŒƒ¦¥¸ß'j2IrFÊ D?uX,ÞÕ2& \{¦ÝøªxXÒÚ.£zº…r „7÷XÍ 8T`oSÓ½›ÃDýz¹y‚Ó¢“;Øá§8ל9ÅPžÐ;Ï €ëÂý._dÿ¨;/‘‘¡gÃ.﹇ÞómÁ74×—LùF%ýÁ{Ôf<¥º4\¢•ÆYÐ÷y[:½¦‡‹5zZÀÜmlå1ò:ßMšd¡~1}†+GvÿË©ËppgöÃù5 æú³qRv-ëìÈòe²ZÁ’ãÌðíI »¦BSuÉBÿ„Gô’ä#"¢|¢f®HŠA¯ýÉAïÈ¥¯õ%úïYa—Y(É`¥µ5Éâ7•‘Jæ–`Ì%~N¹e†Óþ¿‹ µPáSå|F§¨v7Î×"˜™×âOwy!ÁœÐMí ü«˜ ³9GÌ¿ºÕ³ 1Å›ïØìöàÛ<Ìï­G“÷«,PÉ?¬ÓoU6ÞɃØÖ]Hè”N7nºÄ”Ã;S]ã÷õx´Lë2ky2ÏvÊ!,[F†H.ò?_‹ßu@³“eøC9¼¥ŠøŽÂýn Äo²a¹ö>®m?â<õ0ÓÆÀÞVõ™‹­J°Ü2çɺ×\µ8A½À5A{ SO/š0 §âÉ%­ª/ßµ(xþØ•¸½[m0aœºizŽYdÐn 9dÇ TÖ·—ùòø-¡ÛŒ‹»‘³Ò6CÜ•B/óÇnJÚýá¡ë(vŒ—“¸à¥‘½J¬ì5Þ`ƒeÃãx½wˆSµÊæ‘Õ´Cù4 1ÀuVj†»—X,Û²zÆD»—QQ-³%˜Âìì¤zõŠpj 0¤ @È •ÿÒÁ}ý F‹+:M)Ê0·>:¡ß Jôóýâw2UXŠWJå´9äo"b/¦M2]~SÙPÇá:’!·×z^Á®áíè™*Ô}F7cR«ËI{/qž}„å‹»ýLØöN`’O¸`¥,î1Ø'®ÒpU2~ÁDçw[ì [rg$T|¢ò£pæˆ*&æ„­'-€OÝøÂP*^Ù™ž‚¦3H4L”øi*eÌI}Ç3 ÿÕŒ*t¾3ÎF¯×I„5õ6ü9%šeqÒøœšÎ‚%ÒAU°ã:ör‘ö]•Qlð¦9aW}Ã|` ·ÃBŽèÄÛGà{~Yï£|®§²j AåÜL@‚‹ÜÖÜÕwí߾ɑ ¡]_Ø%ÓØATÙò¹®äiOëȽãx»ÜªOw¬9DÈV»GФ½­¿[”¢!‰±ÜH’”  ñ¦sÛtš*ÿ9b'y1êRýbVt¿þLßL±Ü™Ü½A÷̧ñlh9´b 2þé>Œi lº2‰•Õ†Òšò}=‡QjêÙO]D †¸_.8ä0¿}›±µð8ŽE ¦Ê“û¿ÕÊ~ØÄdkÃÍ5n¶œˆ(\5-˜S`¸oT%|Ö\SS¸ˆ”yݦ#o>t8᪌(¤YcÇÝsWF®¶ûJ²Ì=ç qfx'²V„ææøA:1_j†íê›w,Ôvg6²›h(ý 8ï8Í& (or?*ÌÊî´¤IõS,cl Õ|¾ ä›çØHcû-Ϲ" ÞúòûC"ÛvÌö4Ìé"*ì£Q1s!]ÚdÎA3©Zpsx»2ÄøËùº8‚’¦9Ó¶ÎÙëó’Œÿ–ý¼1–q5ùmU™<~T§ªd !$¾xÌú\ÙÓShÒæ*6"-£·•cã…J3>µ}nÑÊKÒØwÎ{éV¿XÒài/­ß¿Û/´zÊåf›Ð`Õ¹@ƒ­“ï몙n môûH²§ŽüÕj›Vd*ç*E“ˆ?$ôW®ý}7ñ “ë‚´ÄåšîæÀ‰þ'+K)=ìø¨m—ykVÿDxìÌó¶„%ñHvfÿ[žJCÈ«Ò=Ìp,L]PPÌA1³ÀSB‚‰ØêÚŸÕ|³ã¨oˆšâLCïr†OTc©<ÕjV0òÌG+ ¶TAÇÓÂN1ƒwjX(ø 'ŠwVµ|¶(kæI¬TµÙ±Þߨw®©k<Þ³bÁ›9suÙHˆHçfÎS¬ ©€ÉB[ëÈ—wµC%üÕÆ}ž'¦¢ù 7*Õ;[Ìîv;¾£¸öšŽ_ˆÃQL¤…ªKºN±c¯%Ĭ[„d3«Ög 2"#•>*|S´rÜø¶ºPÊjïn×)ÏÿS}½Ô$þYÔ˺‘¬ŠÆ¨0qòh0Ó•|óä.Nò¯þ1MÅlŽºà¥Puš–e¤ÑU%ܕڗƒÔÒ1zÇú4]Çù´~ xI‹šPf…š•||=}>m·y⾦§œ×Ú‘;Áÿ·ÆÅÎ8 íøø­}·øÿÒÞsº¢%\’׿iê4ʨgTN”€¥ñpüh¤x<—ØW;ŽHîØ2êV%¾….¸”ØèžÇ•“†28N’gÜ“Sq¤^ÚeåÂú–ŸB¨J¸1 Ó¥r spéQ’òÎwî†$®Ó´hg×ºÐæ.°»¤²Éû«œ%s¶’,0­04߉³¬!üãtÃÐp±Á2ñšÀQ¸ˆæ%i†;€¼ÉEãD ÿZyžcGGr6á¾+\Šxˆ•á¼ÁÇ=¦?¨£<;NbÊÕë±åÍ‚ Gvëß]% µãW0©|Xa^pnƒ{~xÁ*Ï0«ÔÇOÎMäÈÒNOçsF±µÊ½½¾}âe}í;¦§q6Ü«‰øóßpê¶–)¯°ä(ð=¨eÖ.µ®»¶zN„{ ;bëe|žIŽïw%ÿ#uŒöŽúctŒÉsÂÓ9YLIÂÊG¾æà¦/¿¿j7‰ÎîpaíÖÀ41™H2æ¶õ Æ\í‹$¯Î¤Ó³ "å>‡ÄZá3ôºR+%jþ ¼-Òªxñ(y¡‹ìOÚ_ÄÀ  ¤í’¤1_…i}gÙ†ÑÀàðôšILÍ­¼8ãJèNRJ?­òÖ=Ii¡B  þH:fvY·¯ÁðÜ U©¯3à©f[ _Zætå¿ct@d&d2¬”µ|岆§Ý5ÔßëºA}´\¤’rð„»@ÐiózГhݸfd±ßšôA½áM|Pìr!Sf˜ÚÞ‚F5àŽ~¾)ß–âý0ÊʼnuV—tZNÛÄ…|™1sùÐNr†ÜÞhô“æÖiû·Ø´T‘Á¼ñ(L5à 9¿¶üÖàÉÇíaLÕ[†ˆææË.É$³¯Ù„ß,ØAqZÏý¤Q—c0tâ*—uSà7óž U)ðB†‹s=õh¦òEÌöj.Þ_4ÒJ®»;¦I:Þ£òÃúiÞWÏ`9Y;¥¾:9Ç‚:7ͺ¢~­ñF¦V>Q!91P™Ø“u{¦–²ny3…¶ŒÚÙŸ5®Ì"TJ¯@²ZD*‰4)a—à·šJâYÔçHŽn¬½ö„¦½°+—»?L°ÆV¯‘u©i(¥¶x³Y×vô¯¾f¥pÑò^g %¥XÁ×$†E¦ŽÿŠ×­» g‚‰<¨Š7^ªò7Wd.I×y’ÎŽw‘ekòª“¾õl*-Ú5~™Å¼ê2“Ÿ‹G‡F;„£ÜslÉÐ7Q̽ðÈ€s²Mßbÿ=6Uyö( Ì‹¦¿ í õ:µÒ¤.ƒ'ò) ‰°Ï1Þ« !àÂq» øûŠ’„Ÿ©#g7Á–‚4øNªœ†¿FÁ¬‘Ò}.¨¿%7ÃøØZÀ1$j¥Ix“¯59¬“1©2ª‰f0ƒ»H?þ†>ªUÄ.¦—ñª™Ññ/“]mwùâÇÉb?3 Rœ=å„<´)¿{ p$ÛS£]ü2\{MÄ‘~÷§äŸUïþÌ ‰²¶È‡–m/ìÄYÒ†`šª¢=¹Þ¨ÙÓÍ;!ÉÜ‹MYGàŒ™J‰} 1?:odÃ…‡åuáÀõ2wŸáJažœè½,²Kd•[®Ú@ƒ6Ð^3dÚÁ|(S…ñJVß¹ÿ(”^6>¤Mä3*Ýsñ¤^8J¾;!n“GáäĈ“ÍV=ñµýŠÄýÖØ&€”w«)ˆÐô׿gÀN-.ëæ`]ØËyz†@îL᳸ ¨”û¦Ü§Qe#gçò¦`í IÓ>ì |3¨›lõS;}Øõ£œ½ìÞrïi‘VÌÒ<‰õ±ù±nˆoœ¾–:"­Ìˆ÷ô¶àÉk†ÍgFôs“T{àD8ƒ a¹ì²9îÑ×!€ú†”´´Ò~'æx›ˆöÆœ#Õ ëçI貂¿ªÑ®, tm‰;ãóV.Q)®˜<b䱞E†*¾hìÆV×—3i`¸Dus)ÊBè74ËmgžÛ­nbhN¸:}αcïì$e‰ý|¤ÙjXúx¯½Û¦öx±°S¡TS.Xâe¶¼%IOU>ð}«‰KgЧa|×è§l2Ûþá º‡ãÒlçÁË©~$ïweŸµo¥?(' ÂA†åVGÏTçEmÇ®Ί1ñt˜¤¸‚Ôȉ90‹¡?+Q^«¼^/£&0vYÔè*|Çüí?…º]-Aê wÚæŒc¢$“¯2<.cQ’Š\vîÂ;«,±–@ø vÃ~>¦Âܦ`†êË·¸bT&„ò¢kƲ±õÆä£¨nrïχB:¬UâB1"F²£fÍ»ÀO¯^‘þ7½C†`V€ŸB(%M»Œþ§à›4{9/X°CD@}` ñŒé]˜e=¶È]µs“»AQÛv«ãéDw­²gFðýÂÆë†wl¡Öç‚êl¤.ö*¥õÅR²¡¬.@pfŠ—îûþ‹K î¯ˆ]”ݺd¼ØDÛÆRaèï7írŠ™8°j\¼êÞR¾j¸°H6¼Bêö`›%¼tL/W‚ĕA'KÃÅ,«m³Ìö Ñ…ë÷(¨Yuª>{É)YÀåˆY+&äeam±Xpa"xc¢ÝëfaÖ M§f9 ýÅnhˆHsæç¾È=ýð.w ½ÞÈ.Èüï‘¡¶1¶bB³Æ‚XISèn5™¥ñ9_ÌNÜyìÔ"»yþ^’¶3XqÆoß­¨ôã0Iõ¡?´Q»QÚ]FzÕÄÏûÊ2™21 ;%ñ|f~)Õ_¤Öêf¯å:#Ú@Z5}ÒTÊrÏ÷ô|C¢/Þû⺘ü‹þ‰þòÂì.Å¥‰Uký«<²”‰ÅÂÅ„’A'GûOÌþîvÕÇÓô5”Фm¯ÊN>õSÑ Ë'Œ$N–p’!miÈÊž5¡Ø`ZjÓßIæÈɘIÎ7Èaqe'îóÄ»i?],RŠö·³{ß¼EX{¿œ÷ ê2êÖE®`ºQ,ªP¬¡2MeiÔô,Ï PQ¬PEá'ÒÁD]X[wûG ¬·n­¾“Q “mïÖËG¼ÐÓŸ¬Ö,¯"Q!TÚóÞ—SL™Ú{V¨ 8Ã,ÛYSÏäþy‰Ÿ‰¿-—æµ"Ž%’’èóÃ!–€uœâJч¯œIé²'–Nø¤a‘÷E§±q¹<13SUœxÀKø6$y5[]]/Ãq¥ÆÇýÛôïTCs̉”Š#ò-=Ö! ´µŒ¬ÅaPÏLVÝÕãɸZU«ÚWøx0 Þ‡±&dw¬‰KÖ±ký©ñaïËÈVÙk#ÛVáúÖ$n=©ÀÓž7vbŠª‰wðD >:o0VàdròßÁ^³ã-9 gq÷0‘-%n¡+‡_±ÃdW08ƒBZÓ~œ™ {æó`Öï±Ý+eù=h ã{™I8üàP‰IÏ6ó$3ès¹‘°¿)W,žoS)‰/ûemCCŠ&GP=pl«þ³ ­ 7Κ”„ÞêøÞSð_Ë–$ B–ì Vs9ºŽ•}6%©Ÿ]ùñÛ#8£óy¤ä‰ÜÓWܱvP AÌ¢¹¼@ã@pì°S¸Jµš’Úeá –ç7›Ü’!qL›!Xˆà}i¬®•îC'^(å~ŽƒlsÜ]ãÓjRù(vˆÖ²ŸÂ*[^èC;hUwÝ»|ÕlûÍ…Úþíò¥¢ØÚ±«úàj ´ÐíûŽæD2-§©4|Gp…Ÿ=UHÉ ÙÓ,5ÖÉÄñM=.ûh¾Ð?½zOò©xliIRX“”¼çaù2ƒë÷ó* ®L:—`Õí«ÌÞ`ŠKŸ0ëg¤%˜THΜXÁ(S#²lZ4’ðÁ šƒûËî›…ª~_œçÈ 7ÖÐÍê3~åÕ Åþ,Ï—ýÕ/<Ü‘Ë.Jñ…¤Cï²ê|˜uHI q¹‡­`i¯¨¬TkW§Ø/]iQvØCyÔ—ˆP3a3Õ öÍð ÔÀ¡ò¾ÝÎdŠdåÞw9~£ÍÚƒhˆlòØpÖýÅÛ³r¼K¯§-þsjOzÀ¤.™ÎÃÂ="Ñ ïnå)ÜU¢Ûå6ÂN>¨“y"×ê§µŽ·¿õúŠÚˆ s:Ú°•ʈay6ÏaùxÍÔ~’‰Ü¢LeÛˆèî…ƒ!Í@|n¥•ssdXÞÇ7¼[”¡ÆO#­¹sÕÔ3‡.²Í;¨¬æZÈî¨^Ò†ØûÌ]?ÛÄ»ù—ÌCÇ1°bÁ Ì{ƒfqòhć0sÌ|¹á´/Å{øDTìÎV-ÎǬÂvR5‹ì£i Sü%ÏŸúÀ2EÔÔ­ø®z´ 0ŸgˆcAœ¢Ö\±™ì÷²ÀÌ=Ó\ÄöU3¿—¯®·ò†ªÉU>o”®Î|5·Y’qµïÊû$áN%2ŸÍ0Ð:ä3Ò‹<ÊUFó€ÝÑ1'Fàe¤1>ü¡Ã[ÛÀJÖoA©þööøÔøf0I°J _67–Æ»ÝàjÁ»@]ãÀ"ìøè«Þ~(1·{6²÷lTð"7iQê“{ÍmsÞBm­ØŸ—m憈©X B–JøƒFïÚuþv®n]b¥¥Üh³ëÉ#ò£—¬Ùò!棔2<(]¡az*|«?Õ(ýýlÍW댯›³ØãÚ.×—ü.š¨;ÎõáÛìãþ²„†¾ï«G €Ï‹ù†é3¢©¹—?´þúÌu'•þã&àÝQ¹H@ãµd£€{Øy 5;Ò²NªÅ7CÔI [ç¦kxHÊIýìØQ”0'0´óàìP+ÐX†/z¯áãìQ-k‡N·%ÎÆ35b͠ƾd‹ñ‹—¸rq‰} OHfE  W~ /—…ú”7b|¬ D2ÏŒºfžÏ7ò¾6teÓ±:³Ž/!uƒp›®”‘…6w„דÂ×ÉÎ<ܬ Ç$øÃ ™²|:°o©©òîçX¥!j¶I nÉQ·õ™m/Ië0”RÏJêG ÷å±EƵT"jP1Mšº/ž¦,›ƒ5ñÛ¸G:‘Å€³uãç”Õ!ƒV€iŒèêzÐ(ñl1ÔåÂKß«Q>;7´^æLÿç·Ó(&ñÑÏ¿:¿Ev\ 꺺ì­4»‘³¶%~€óm¹‘à& d´¢Íè†É®ßä#é±ÄÜ·à¦Q}ä2u,T=øiv­ÿKAvH0Í’Qµ'+ïTŒéIø.Ä%’êVoxPN„9ñj°MùQ%ò['ùvêÅüæ¿…ážN ß$¦Åoצ6ö"ÃÖ I 1ë“{lC¤Ä×Y9kÓM#…¥îÕõn‘ ÷9#&¾ JBzâ9Þ6Åøgÿ‰qVmÎ gt÷åÊbr¸3cÀÙ°”'Åȼ£ïXþrå9ëöÃ~Á)C€EEžM¥<ü‡ŒÁãhuVƒAhœ¶)z‰Û<7®Qá¢{Îoc¹ú +Â’z„Òè§1jË‘Xà õæ”s3é«[¢Tg’Ÿx§!ø¹1 ª–½Ò>Ý„qa$!&Õ%>ÎdoÍT¢UŒ'9/(²¨;쀸]NÓž ¼Á½ÒNæudÝzçÏÇFK‹çj(Ý•%@YJŒeð§CÑt¶•hçS;Å”65]Ðñ,ºèÞp ¯ì§ÂKŽò%9äšœdyw“®Ó÷§üƒt%Á#shQÀéaàÀý¬L!%y_¯rŒ§Ó®%{8Z¼D…‚<œcö1âºæŸ˜ÑéŒZ“pÎdtÆPE†‰RÃ÷çmDíb!Ãù/,ôуó@¾ö€•[4HDìåXp.H D'8ñ×Áªø ËYh ÖLájê¥õ†¼ »ï·:=Sf™.ܪx€"X pù¨kåžØXéÇl×õwêø|Ò®Š|)råkÈÑÐ 5ëßÛ" ÐÁvŒÍuÌÕÈšˆrŸåíô Ž#DB:%,ð%`ð×Ò@RÌ8˜X½,æÜ<XTÌ÷¸déÛ18ŽèèNìlç3*Ãd3ñ+Z&ˆ™ÿdˆ’„DUµ9Ë@¤°üCÕCó`+»’‚¾¥mãqÀÍYRŸz«z‘¥Z¶rr¤—_ŠR¤¼U—¡X½+tÐW‹æ64‰ç¶!¥ÁBøC%ˆ ¦îþCi7Áíl5 ØJFʪŸš¤ )Š“2¼Y¿°õøEM¡»?LßlèµÔ°† Åìè/É$zšÄ:lF¯Ie€+9h'ÏYÖ« %hI&èìɬt"ùEËX€/¹êêQp×Ý£×V­!¤æ›ÞwñÓi”Áf øvŒ‡rDÉ÷‹‹w´|Än1ÎT:™U7»ƒ!­å ’ø _-0ÖëÇSkбà åu‰'·¬˜vÌZ ½=É:þ¹àc ÄÚvlv'0Œ6NG1•tø»Ù’µZ{ {p::m\¯¾¡Ì…í#MˆRb(k KÈ\aå“oQö^ {’b+¨%u"Ž[‚_ÐJ“À|3:ÍH ñÿªô ú=!ßÿ`Kxãu!ƒ/Æþ=κsn&`Œ{åÓ`€¼ºÎ¦ ß[3òãø;&H¸Á—iÉâÿ1TÚÌÐêKØ¥Ge¦ +=ôÆg»¹M›#l¤Bƒ(>f7$ôá[gz¯ikT. Ld~Än:.ìºlƒ‚p§}N;Úè‘Ö™M¿À2w  „ÅU‚Á§¾ß¨ý!—ĥŦæLPÜ#ZhÑG¦±) .r°6€¶?Â^œf×j 8F64¨¸³>ÝiYåÛ]Æ(м£7­èq”ów–G)j°À*¾~·L@~?]$rº^Û²×9æ„,í¨÷nnÆÁEɪ Þë«î6 cRMtŠMÞ‹õÞ×J6¾WLV"Dû)ÙHw‘óbEã°#¤}Šˆ]“ÍqýVvfŒÃ¯FzWÉݰï¯Í:< ”•FÒTzѨŸïñ™²c;]ç\;Á(•ròÝ}yŸ×.¹‚š&¹“Åâr£ÔÃ7dMÚ =ÎÜfûôô?ükvlã}n …†1'Þê0Ò×ÞÙkJ 5–yl±°À³4môñMU÷Í|´[,Ä$]À*\FXYxnó"‰5´¤ú×[k’|æ}×[d_¤`çH‡>v´]ÞM›¶mÿ 7k{¸¸OÓ œµºÍæSkfy›§†­À7Qô_U+ööª­›$‰iPö}J| n¿`6·Ùî9v¾Ó^PÝšw2ú˜Ârõ+k¡As é3>ÀR©}}~€.¾o uD°Â\D+w½´¹06©:6¼ÏÕÖñ» üu_Ë3¯æê8Õò3#0aÞOá“­ÊuÍ=^fÏ£¾:x ?PYh©#Ä>êzòµwݲ ‰}FZöø¿¡;gù~?Ÿó㿈ú˜ UÇ1 ˜- BŽAÚñ™uAn€SÓ†çÌîMf倨ɂœG%Ò÷A¥¶y £ xI‹YM±;Â¥Ýr›ÖÌn«žï”¡Æç ñ§{Q`ð îŽ´™2ýýbP­Èsòî¬Ê ÜËÌ‹×ÆSX!Z¡1ŒOÌwKJ(\ðT7ÍyÁ§!ðp¦ýÄ”Y–¸w5Ñϧ8‘!$Z®@HŒ²» ùeÓ”u\æJ!!«2[ þ,&\nö*pášuÎtað‘N§7Z7Ûç°€¹Øf8\wü¬éš¡#Ïz-ÿÚ:ŸúÇw ôÁÌå˜NèÎèd Ôú-[•ùÁ*3ÛÏcÿêÎG8Ú}`¿D×Êhܯ-×À_Íè‡"+ øcÄTäd¤XH{ò{ vŽyR'¼ÂÞÏ,AÅ6«ÝXÊŒlŸðs=FTðƒ,»4Ù]Æ^êðù.”?ùE°è>Àæ®Èk3ª«ž˜âe½uø2Vøe;ãv}ÓøÀèè7<ŒJÁØ/B·'U],SIMJ†‹ A{΄…Ä’Ø¡§ƒÞ¹„ð½Ýy$x¨ Ñ2žRÁ]šèú¼„4ª‹csÜH¢sÖW² ²„c†ˆçð+º§lÛv{ _N¾ì£ØªßFXciø³ÕŸdì(U‹ƒ¯¿ˆ/·œêÂdÞ"|±•ú~ÎX'…*Þ®zÈŠYتÆÌê~&•GXl˜Ç >Iá÷ÔR‡L\¯öÁìÒüÑN¤s­œàµÃù üŸÁÛ°±™«ÛÒw½–… “˯Ð8I„ÍÂÄÍÂ0¯£nw؆a…_Ý·@kÉMf?r3LÆ­ò ­êBc°%‘Ód­ ™¡yvòdsžÈ;¢1ƒëX)Ì[v¾Y‹£Ø!’N˜õ£üÿÁd?Uû`eO±•¡ CK]Æq¥NýŽCkÇOº ¬•F¶ˆFa£”ËV6qÏÇ6ñÌ=²ªC@›€Øœ À$ž)ÊH}àÒ›î3Qéun C~+>‚ó**ôpŠzÀ¯D ]Ç„ç#™+ûÒ‘GÒy¼ëä`£Õaù»A+§~¼eâbqåð••)õ#)ù Ý9éNÚRïãÃ% ò+0T÷3¤v¶ÇäÌ{À™Þ6¤¿/gÃZ4G›f ²“µ'ªÒéUæIŽ{izö‘QèFƒ¾íûšrN<"¿rØ$ÏwGj#Ð騿5¹êÀ2kWÒ‚ :'Ö¥J;ÃníŠq`:Açr5ȶ†Ö‰B8 ug=î˜ÒkÔ |¯˜Ôÿ+ÐV¾Kü 1šÚffÐ']y’Ð'7¡•¯o”nV’ü×®õÓòú:“@ùbKAÀZV–­Ê9ï÷ ÐÊÌŸ$]¡juùƒ ;øKeò\y—©+«jô±9lö¿·yd&<—åð¾ÖŠPÁø#TG9_µñþ€Ö#üöò†ê¿¦ÜkPß ™gìí:áäoç”Ú©rÆ›¯Ó§e'ˆ^báéÍËKn`?Òœ_Z|4ð~«%¥“dŸDd½ŽTñ«Æ´Áÿèð˜JŒÈɤ6ö×q}é-ÖáP2Ë£aà' Ô\›µ>éºIãŒX¥ôœ};mÎ1Ȥê-¯E°”–7ÊÂÉ)A;6á=j £æŸ{î³–ÊV„ÑÄUÏ%*§»oʵÚCzí±€Â»•àV¨MÝä$”í‰`v#j¹Ôñ•Cà?eœÊ/Æ*7ŸÁµ;{ þUІ,‚}÷’ÿ"á„Z/­šX§ˆ€K%©¨‚*jÌÅÎX®¥YI´kñ(ø¡q¼s®\‰viCÞ±%Œ¡x@V¹!žÜb}y"ûí& ì1_¾ZCâ$}Á‘ïÝÌÏ¥°¬ÀÌî«e–uâ‚Ïï÷öü?@²\вœ˜0¦YûDõp{“~”ô¸¼dû׋)\@h4­ñ™IJ.£g©Õ…/²¼'çk¦ñþµâCÂAü9)¦!a¿íοMºí"IŠkðÈÐ’ÝÆ–§‡,sâznŸT{,ÝcÐa}û‚|ºìíÑ{h°ã U`pWÈæa§´oÀ£[­ýq׿ ‚ qƒp„Y#½ý>»«æ‹ÃUéMã/­’ä¶ÏÊÉ5­)û<›=¿X+ú‚JhŽ›£uÀÐì0†­à6Ü^±¼Zwð–çÅ_ ë>3]kÏxÐêÂ…iÉŠþO}ÙOúN2Í ¹Á—O?r‹DõÜ»µÍY,©W7-®w¬”ÿÁü½SÆ®”¸¢÷Qs %©¥,½¯Ûï! [±_.··ðöM‡)F ¤ ð'5ÓpzB —à@‹J¬a} X‡îÕ‘xDå†íÎþ»dqߤ*²*Oè{ ¹éôhìïµ° e¶õI¾uñ!g| q}1H䓤œ\êD—|ð-÷dQÀ×(i é)4cŽ*fT+߃Íñ2±&W3MŠ@" `W¿'PZ®|PBqÑžÍÆš8ÒŠDRÐîö(÷ ›"Hø1á®IÔˆºA°OdÎ`ùàËìwžË“ 3ï]T|5žb¼¸ ‹Õ?J™¥¥Á¨+Å´òaé:VÜò_'¶Õd¡N>ž³€©p6,¼E ëñ©®J‘ý)äfu°-w:Z“(y‡Ýѯ`'JÄØœaêÑÔ]éЧV/ ?Ä+A…$˜êã&¸ß­3 G¥}þÜ0WæGŠ\@î§‚ÿØ.–K”ÎÚ߸ïQMÄ4{7Úwõåþoç{I$Ü?›¼ÑÑ[®<í¼I'Ú¸ªþÀ*;ª$o6&¨Œ²››"ÎìŸ÷ÉŒ –«{ýý¥»B­hβ°ðç…V¥8ã~$À»ìì N¦=dü¦eg¼ð%â¸LÏ­z)x]>¼°Å»]&Ä×"èç.›Òv«ëHöiíf Më^ˆý“ðë9¤üìˆìQ÷²‚ÆýÎãìÞÞà­t&?Š·ïX½€X9õ5:¾ø¦kØT,à·` ¼9{%:ë.\ÂΖ7ƒð%œôõ~ÅJ>EŒjãk)!§Œm> q$´ÁifAÚqkp#Úd;ÆZ³P —FŠ›òËÈ (ƒ0IÃZwYîÚD™å$ûdŸSdq ¶ƒ×f:J øÚ ´<¿ÙhK”Á·Ÿ¸”Ë0¥×,$»Èäíu “zå[ÓÂüÔtC‡· g-èyÂùƒKŠ;à®Âõ7 å¶:ZALBò u+K;Qã¢9):™ùÃX6q"¶7ÆÌm6ú?±Vûûj*H4nb(I”N”£—8AL7¸iù’M¯«ÑOŒïå™V¼÷Ò~åâ½Ûô,(ð#|‚/J_äÍÿ5ö²í]˜\:ï îñä]2vçÇ¡ÈÝštÿ‹W~ÝÛêÚ¦Aw@Çí[¶Ž’r)Þ‹)ѯF°qõ#²HœJÖ1Ú›ìOô“Ò,Ø"— ÷dQóIõw@¼Œ¬T¿é>½ÔbmWè(‚RÁ=vWü]L½:¨ÿNG0œ}&ºþ°ñ‚K`wãc|9òÃHÀ¾{‚û# ;½>ÍZM Y+9úÝ+ä» PÜŸNÉ~/’Ø#=lf’°¤s²&©f4ñáXýƒÈ¥ÔŠì+=)¯H"ݡއp¯¨ä@0hxøKõ/?Qð4¹™`óØ$*`}Ö¤³@àz€ºàg”3¿º‚´ã•¢©/R/ô-fªàÐÅöî©„µ|ÏGÊ×Rpо\@Û©ÊÞƒš“UÆ‚˜KJø™šB±—(j“xÆ(½Š3Û»ßÉ^©øm«V\|=ð~‘òÇi©§N¼ÏÕ©ùïhEdrzAN(]íS»¯ËbK÷W9Çþ„M%;¥­Em “*<ŸÉ»×÷)·d€Pƶ2¡ã~v&’­ƒ–óÀîtâ1i ° ½g\|DõÛk/Í<~ÌBGB iÒìMfh~PÞØ‹PBò•=7BL>%:§ÿŸ˜b• ò†ÑÇ{è£#=—‚ôÞ¾PNnàzL[î “îÏ*”ç• ƒ¾Œ—læNøO`®÷­ß”áÊ* ‹HΛ´ÜÒ³,Æëµïëµ'J$þ^©†|bÒ÷;èoðkƇËU%oÓG‚ÐJ ÍÍ澇<Á=9Ø 8I½îŸña=Öo!€ïR\ŸËÝŒË` Ç$4— ™ªg ix¶qø£+ü|¦¸ß3„Œt$Äk>–a.1m7Ýñv*sw7É?¤h¤J+x.±©ý¿´÷Mé-Â¿Þ ïó¦ZLÑ7€…xË”çë¶ÿÒ4n©'i¼ÇNvja«s]ärbiΠ;§… _È+¨àY"½© 4ßwOVø«(‘aRx¹ ó‚y˜C×¶.ÖW8Ö‚{Z>ͪ¼Pþá¾æÅ;ÑóZ b`»ÏÚ zN3RsÔÇYBÉ/á§É8yúTIþ2ÍýM©´Ãxܼ·…ߪ0ŒîIÏ;zÙƒµ ÖÙ¹ì­L1ÎQ0zR¶ßFêUè²ÏM„Tj»›ãܶâK`ËÈ”å´W|äÞ¬žá¦ëxÃ0OÙxOCæý-ÓI¶RR K®Qo3¾cÅÂc©µ3éäX6‚ —þÀ®Ø. ç–¬f¥t4ß©€É'wËsƒì­êÚ%Ébõ=¿àÖOrw½ä]Æ'¥RRg·EUÍ×r)ò×™GØðΣ‰h‹}£jÀ¨·ã+ PÚ`|$ ò‘ªw/òàíbA¥©Ð%5æ0£Ëf³CŸ­[:´¤'öµ Úù 5bȳ gÖ?˨l@gý/Lj—Êš•ô#Ò"‘Á,™ÃíwkAÛ‰„ÃKv›b˜~øª7̹%}–¸W}S¡£é»;"ù^ s?Ò„T brÌÑÓ5qêä`¡®1<«^‘et“(WKvR­•õ.4;3yJþ‡âCA’És«Bž: Ê"à¦xý#á:Æ^çýã9À§´ÇsìE®þ§û Tåü.¡±ºQ YO§D÷^B «ZÂø§Bv°=J‰¦b/È`+'™”Þògö}Ÿ*Jj}}ÇÕÌÕTªH"9²o¡`âÌâ²|²¾ñŠÄ‚YŸ +[~¸ú‚´Üâdþ…Ú趺ÛBc¼tûÃTEѸ2{Ä©DÀTõÙ¤#Ô…ÆÃ€t Qª"‡ò!1³î|©%£ª*S‰ùN`»ûÛ@•§Œ°6³¡d²4l¥s¢"¥üWíºšÀ„”Ù6‚.¹ö‡÷œ‹×™d‹”¶bOæn‰ˆ±<¿UùÐÌ È](Š èa2&ó<»ËwÂ@"ÊùïíªbQ•‚ž¢îdWdШ˜£ñ)­.™V8€!5+*úI4¬¯ùÜ›$™U‹³sß²+—›â”ìs¶ÈE'y­¦®bv‹6¶ANãÍÝxÃr^ÝO‡?6CsœP’VÀÒuÊióžý‰Iø:ÒD ”VJ5[©8V굎>pÂTª‘õÇØS—6yÒ¦ŠC¹Æ…™ö+ ü•EÚJr“)Ó·ü¾sFYè—®Ô:F@5šØ¾æÈˆ¿ÎE&Tjšßgï$Ì6c¦âNÉMÕÜ!õto–ÙIwÉÈô!À»í™d„ŸŠ»fÏ›$Uf{dAÖÙeB³º¡0ÆÂhztèL|Ñq‡5{y$/GY;B¦™%•â-Ù¼,C,Æ/nÿˆE˜@‚ÜÞ%àä©*‹ó¾mÑÔc—›nÑ•~‚¯>•ÅÆŽp´r©¡Ök‹AˆuÅj"¾H5è–°{7e­ˆFÓÞšß™x8MÁzúCÿ>d˜1¹ ŒùïÃas{¼È<Ì3PìZryúÅ —hMt牋tyè•Z=‚)ß©^¦ERV*Œˆ/ô1ŸµøèãÖMƒ(÷t«û0“iÙ •’nê­Sco{Œ;@&ä7ÌÖƒKþ8d·Ìr·èj]váꞯf쩜^ŒUæô‹› ›é†¢I1代]fµ·-™–CÞÑ=:Ëb¦Ðu¹brÙº("ãòjóg^»‰›Ž6žF$%N÷Y=G‚ißT(•¢Ð/¿ýMűÁ  É1qyÕó)`sœm˜MDgD€ uGʹBu¬¿¬›Ÿªšý)Ù &ñŠp9©ôÝW´ÙÈ5fþ´Ç/d¸yQ›‘)ÐÚ3´kÙáìÇÒ[矓¶¢¼mÝ'Â騬všcÌÌù+ÄâÈf."É+—Ü“#v:,òù ûŽ>h,hl(¶Ò&Ê×KüHE×b7ã¡-sãëšP çÚÐkb·µmÖO±øZ/ä×מÀ"r›å©MŸ6â#‚ob°ogo\yà„-fϺÍBÙhXØõgE7´9ÌcDþ˜wÓøÉ;rM~çÜÍõÐ+S‚<öbÆ8¶]f )'Zx¸ì’ ]Ç[ê¥ü“Æ¡³>ã›àx»þ÷ŸB4ÃJRñq.æÍ¥K¬( @ Û /´úeEÈvR&¢íkÔÂMx㋦S;{4Úk2/GDàÈÊÎsùLP.(3+Vä$2û¡ŽÜ¥Ù>xŽkÎ{_]O¬ÊWu ‹…ű¦ÛFØ~®NMyН™þ¬sGDGù_O `É¢‰Þá¤ã\u½—ÌjXC’º„î?‚Y*:æ]\㎿Ì"=}·$js7ª»Çiá…uØé‘¥îÚ¸ÃWg¤/ž7Úþjc-ï ˆÄðfò›Ïöâ㜾¨YvÑ8µ´`ÙHŠ›'Cl-yRz—ˆæ:Ä¡8º'KF‘Š,û]žd3ÂÁ€Å=(IýÉû£¶íQÞôzÙ ÀxrJºò— ôV^îþ#ÐPŠ1ø‡x ÀGç»ü ôìC¥%ˆì«þñP‚pèÐD~¥ç‡¼ÂÝ\aÊ@ôÕC›"Ù+oÚâ8GAž6òK´&ŸµÓµ®ê:uFfýÛñŽ6}R¿=ѽD nÖ°B„"ž,Фü† â1‡{y?#ã mÅﻚáíÆkŸKõDmEò+gåB@aY¤ÀëvotEÃ2j¶ö`ŠPˆ0p¦õÂè… ±ÝDNœ äÂJ¹ÜšÎ¥(¶]LUôƒ½1¥…õû裾'Œ/—®T¦"à ª'Á!œ]©=¥™7˜µ«qrT=n›3i”øîË–,|SvòÎÕäû¬¬—ѽáLçú­|´î–`g¿ZÕ²ý™ù˜M¶^ÄË6Åm8=žü®ô^fÝ5ƒ)›8Ù ;æ1¹4¼ÿ–c6f~laeþÕbq˜·DNáLïÍ7 oX].[’8Ø;Þ-„‚úþÐS滛þÈ6|0¾Kƒ;ÝÇ£ÿ}ãð†æ’€S¦† Ù&š7Üb`}`$€´‘Í ÅÄÉs°œÚ†ÔŒ1ÚYm½ÌCZ‹r´¾˜À™_²B±±¾C¯`O%ëÝùŸ'†-ŸŸiæŽB Ԛö(UGS-œa(éÐ$“¬Ñ#Ô|P-Nê<}ëƒTHóå蟔ûŽŽ2^¤Ü$¢}•ÊhKïfØ—ø•KV¿ù_ô ;áî&ùårªËæ ¨3ãm×úì/ZìSG¯ü|· YFŒ*ȧñå @O£°+à†‡O`ÎA“2w‰÷ªšNa;²¤dɤ´t £˜fêâ¥)²m¼= ·„ìkk ŠX1Ñml~jš®ð§æžŒÛ6¶“-£ç¦®»F²³ù×J]çÌìÕ`H>}8åþÇ'|Sc)9ÝfµËé}o4õ–YÔK¨ë‚P̧Sˆ^Y¬@KÀ³ ½å…U'pï’!,hû$œÃ†Íƒyœ³ˆ¹Õ¼'y]P4Še´.ëµ$…ƒIÐÌÌÝøâ_¼:Š€ˆY¼«†y}šs”Ë™ÉSÈðZoq¹|å—òÏß"?_" ÞìkŠ•úÀåÁ aí‰EýѰ눑ñE°4bãÕ—†özšð_¨f #T…5Åt”æiÜÁ#>ý˜N”Õ³@0›ä›x1Lô%­ÛfOëU¼¸\â¡r¶åž¡'߯›£¿_ˆ%È­CгGJIû%Ü1hŸþÍγ&¨EÈ€¿Y¨è½ä´Ï+ž ¿z©“.<È_Ðì£9`’žÛâÕ\§&¬8Áji¨½›quø§ÀŒGÒa£‡3¹’ÚW^²ÏôšƒàõQ˜Æ•s]j)úáF²8ZyõtÝŒZ;¿p›8…y-€õÃÕ.ìgÓñØ}‚«‡xÀÿÆ¥¼Ï¤;’C±3Ãü›…ØæèÙ³ßùl@¨ƒ§a-Öø4 È–*z@ª$]Xr<óŸÓ…a‚ÊÀiòv³jªë_QŸ!kòMfIWÞWF©9:b¼lޤt©Âjp:­ç¾sUt°òLàþŒsºa|·Žám‘»Ò³ŠÅÌUR`°ôÖÍ!š/’ÕÜÒ-(íØ§ŠÀ4-PÜšm± v;ÕÄ:…÷‘ð›{Ì·ÒºF|ëKÇZðæÅ¹æ2 宼dyÌZîÚC P‚kIË–ìNëýíá=“’בdò=€ëÙk)*8Ý`ãÁ$Íg¬ã\I9ê–T‡FÕ\ä']wŃu.ÅÛQXUb¼HØ]ÀâJJ}*]ïáÁ —×¾Åf[7y†FH8ÓÜÑ“¨bPÀY¡ #²Šû•ÄåÞ!œó_ÜtQRN²B„J†ÏœcS.7áe«n ¦:D–GÇìü¾ *d–QÌÊ eÀ=Y‹_ò§LPîÛ°z¯*T_N ¤ ê÷÷&îÈ×ÔF9Ãï¡  ºÊùvòûŽç¬TløªèKºd4À–äÑÉ—oòÀêgS&4k‹"$¾´ í ²ôkå¹µ-³­p$Ë0Ã*2ÔXý°xÇké÷+¦Õe¡ÜŸŠ±OâôŠš²uÁï“6ûáDÈpH‡ÇÄâbÄR3bµ|WT*Óª[X6,®Û6w¡:L~>Ö唟´š¦ÍÛ¦ •hàôú(]¯!j#³÷>+d2Õ¾\&öÚ¶©ç–liIE½mØè4\gt=9¾°ò{œÆ>/u=*5=]:¼3W¤‚[“Ûã5Jw­°ä‡Z3-öcåø;ì4ÀFX´Û !̛ŭï]ÀµÒ÷º°0ͨ2ÄÝ5hýŸá~”(-M%×—ü“\w·³f~†GäUËüu7è³b[ 5BkPáJîßsš7¤ãuü¯”áC½Ó‘V³ð4ñI°EûœÆ_.hú~Ôà9âão5I,ÿE½O¿;–Q¹ìû…”¯Ñó¶£õ£@ív÷—¦Jå…böcÙMi͇÷cèn³ç3ñßÙ»=³8 «÷ÛKÜÉ•+ÚñëYòg®OÒ¾^<¢™PÇ9ž©òé­ð( ›¹bŠÄ“N-t¢æì_ì¢ò.1Œµ˜kydiÚ.ð‰á>«L$Ãõš³`;b3ûìwყAxEÅ!!·»#žÉA‡1ˆ_ÁyײþŒ@#”LW“Å/ ˆÇ«¶j)½hÀÙܺ!ms®Þyý^êð;¾ÌÄA  L}°=A¶4tdãø3‡Z³ŒùCe/n`¶«î´CçˆC¿­à SgSi:âÑ?| ~¿Nø^üíõ„ËÈžÕxMdz_ÛØÒ0}oO@x…Á¡TØo¹ùì…4Ÿ¿â¨‰›˜½\fA¢…ÖÑÏNúsŸ˜o†$äU‰ºÝ$Ìîr3f´-Ìd`³•Dþ…Œ\ïjî2.ºÊœæåçŸùÍ*ÓÛJ¼g¾Ç:ªò&Ì‹J|œî£Ã,+|¤º!~¬éá2´Šená9rÆ¿U`/“¾&¢ÎDúµðM®ú¦ ©‰Ñ:ý”xߨv&ÊÿUÇ ö.%g'¡VÅuµ)þ¹'A¡€6¦ø«ù`žX­d¯4¨y™ÕýÂÖÔ5)I¤¥I1¡LŸ»¤~ÒöÈÉ>œƒêÚGFô+`Ån¥÷nƒçMg@¼Zô¼‡Fõ ¿¼”@r`ˆy~0Ú ÛuÏ¥H´py/²³¥1 m åx9ÑÑNÆËfH’å!Š:ÂÝÓýÂ8ƒL–MGo{âàžÚqÁL¶ªäŒ“iÇ\˜p@ø·ÜÁ:µ±¼íÉNN—2K›ÌÀ`¨Óz¶ñA’ÔEB ¡AÂ8`„úr}ÀLvD‘°Ìò¿´ý{-¦Ðé¿¢#„ëÔ“f³4øÌºªÐ¶â7 ܪ³ªÍOhÂD&gl üãKI‚Ч‹'CøÞÐum­è$üNxˆ|P=Ä層 ÔFZlË©$!$ä€IüL_Z˜ÓÏ×ÚÒ‚qò  ¦“Þ2v¢‚<7T^êwñµÛ5c°á¸ÔŒwÉå¥!!h“õ¬Bxÿž%%îeöóÀ[‚ a ¸\-¬ÕØŒí•pYòç?´12¨O¶qÀ4ú‰1ë‹r‚LÏœý¿$·DÅ D§åV,¦ô’H@™ËùG°]^¸Cµ§éitîbàt'BÄJl¾['Å?jÒ`¥ÏF>Ï ²÷{²°@M¶Ô Æ}]Èb{-Â¥·3=ÈuÄ$,v€¦œÛæQš;ñÌ{.–f:‘£yãI ÌGi*f‚Æ/ûЙxCx‚ˆÜ…NŸ0ßI²í±…ÀÛ÷k†Ò£¾oZ <ð“§Î‚_*b`Þ.[b³ R-O§éñ´DC(oêe®/‡…2íZ )Ÿí-ãˆ1’¤2!oÕ6³®5®g‰ÒíA•Bѽ^^Ç3he”ˆ‡×q¦÷Èú|’#^;­„îÚÆVÓfHË·Árª–¹Á_*îLÕÌ*êãÏYâ|&f©€è;ÙƒTݣ ”·€õ㌢£Œ¢a7V-v¿Àÿ§8ùu{ä°l[94$u7NòLžÀAaÓhGä0¨ln‚ðüµ¬²m^Šý\­¯Ú×k¼R8|„ˆ Í)x*ÿk/&j-l…ÿ’áóÈùè&šž§tÑ[ƒš¼Dr,¨°ñå-•Èã´” Ó¸ a˜Ìy4-´a”ô€ ‚}oã¶Øþâ8c›¾ÏÝ')h÷8ýnFïèz‡bÑ¿µníå´B[lxØi(Œ[¨¦Úƒøÿî§\ç³µ\RúÃcRi Ú;¡Kq¡Â‚iWBÔ}17“\*ñÌó8¼¬h euF\øv®›ö[š‹[.åYdj‘» ]HŸ±ŸíPJÞÂüëðþ]Z@ÿÌ‚è»æ› ^m…dɹþ½~Óu«G]ý´­ð¬¼¼ã3ø!;Øßé3òÎjäÂ¥ØÅ .««¿+(B¿Zå-ߎ`wDNz£WðœË¥ÉžÜ§y±ýjôåM÷ˆw)þXÿ`ÇI€dØ»z}ˆË0zÈ#ZãùÖ”Âí…$íÁëñ¨‹EI Üèà0 ±eØo]ù¿”7Ñ“™X£“› ¼¼=|¬?Å+ÜéÄJÃ$Š7Ž$Õ.ë6¯†»5ÐNgß8ÊÞ º÷h /jœêž<Ÿ{ñdðßë§‹é/• `=ºps¤Ÿod¦ùªC5©=„|Нtá¼Ì°‰®‘[=t«¹Ya³\޾¯ñÓ.¬¬b@uӚƊu ˆQEA pAE™hþÍ4΃R„ËI¢I9ý|%G:½ë&Voçø>+‡ØÕæëf*ð³ALÝû˜¾p*3qþ¸´]I­Ãwž5˜= ºj~ Y |ÖʉZ1º3!ýïßm"B²¾‰C)é'{F:½ÐÇuÎHjÜk¶ •!‚«0;a• Ž_Ñ¡åŒÉ·ö¼õˉŠ18ð{TßÉçj“róÑ(„†*U¬¸Â:áÕðg &%ƒ‘Zžo$*Ä .¦/Yž…«þo§FÍ„•ª´…v/Wy#òþª×k!Â]…ÞsD ÎGl¾¨½ÛðMH‡Æe2SX˜2òüбôM®>³yl¸],zmªFODðV ‹ºâ n~×™1”y¬Úãó½$çîÐ2ðùîMeàߺ¶e%jtþŸI“ {%…õ‰7Æoë¿Ö;D³X/ÑeÜ:,†DVÏC Q® [/É.RùRz q€Ú7X጑’°\ ü7K¼Rð¨=k~¬µqý£ÿIÛ•Ë:Ϭ$u¾n­¤¼+(S­9KÉáõ6iR8-Úp 2ÿÆ«çUÄ|FeRMÁA&’<˘:…Òy,±È¤tþ¡ßu/8ïØDÃRzT´ìÝB8ò·mN!(8R ×ó¦sü'×*}st”ì5¦ Ç[5* ¾×Íx醵ïÄŽKÖò£íÖgɵè‡Ü تã³óxøâRæŠõ¨è e‚Ú˹É7еԕÙ Z|G­*È2d$üÐuLèúô£j_"ÂÚN¤âîH°õ ¹³ÍκÑÞÅ®]ÍÎXÝRÛͨMˆAK¶Îý¿¾B†¡äÉ —áž?qÀÃÈF3ùK†ª“¥UIÜàË­ÁÜÞ¦ÄÚéŸ+0h#5ƒHú¦ç ™F?W/?çG>`º ^× ›J’ž]âãÜH¦%hK*On¬-Rþ|É ³½¹ 9.iIÅo×ì–œpR»9(îí£1‘àôÂ]?˜&ŒúÑïöÒ[º»‹s‘SwêKOH?1ªLÅk·¨[ðÓ”}L‰i§ë¾°ç”tN)V¶èn úÊ’»M…‰…õ9w ʃÁ5ë{‚ »±cÑ´4^~ÌæYK÷‹)àžÂÚxfþ±Ê-ÕzµŸO§ì¯Z÷•˜?Q~EžJ²?±)Øãd0—ãK¦ H>*ñ† œdJ-¤ÏÍ¥d&•i•r€C…zE}þù‚žà}ÜêAOØzka|ï¼M…}üT'²±ÇZX8téƒrKØC£¢ØmÖeË,"jqÎÿîÑŒ¥k·ýoû%ëOvORGE_¦j$È;>íøØ‡¨‰¥ÍtÒJÀ.q‰GVC7zËEè§\Mr¹í>ߘ‘eST:É®¡ÍÍÖß²—šêO}Å?NŸ›Ï¡ªÅ’/Š#î-»'ãó—Ž;!¿“G\¢(ÍZ½Òý Àòàd²òQº6+ÒŽšÕ"€±nc»€¦GoÁfK®‚ v_Ĩ€©Øñ‚D4Éo”±IèåXú©7üó.!»Þ~"Ñtv´»5Æ$þ`‘ÒðýkݘfRè)pKÝwåÕ\¸ŒêÌàŸ¹š?|;©-ÁgX‚ySß1PùǨ¾~ð<)¿#ôøÉ#Ø´¯þW¨£“yó êùìFÏ'K‰s»ä^œÌ{Üvègvl£].ÀlY9Å!n‚CùåJÅùöÿ¡jÌv› Ástþ ¼}ÅnVùŒ/&¦{ÕÓØÄ–¼ÉŸÏ~¢ýåîîÿæ®ô©T$à*°“G$ïrŸ\ÐxݲéÙ;_5†òj>«JÑltxcS”R*ëdfÉý:üƒ†•«°9¯*¶há ªy= &ä€] ÒϽnßÑ´Þû×·Xškƒ²`0 ¼·êÏ·³-²êþ<hÈÚ£§¢“3³QYÓý ›H]ïz¦ c¤Ñ‡çýmÀ9Fè%èQE†¯Žî‰×ðØì8÷k@+Å=ÎÓQŽÀ.ÐúÃ{û¿ôí,¸xLg)ã½ÎÜQzÌÍ•†F¹?>­ TÅØÏÁ(·ÖŒ‘úñ‘ ûÛà‡eÌïªsóÎc(±g$N—Çf ßŽð,¶¾£o£ÖÀìú#1&ÏPœóÜ»ƼžFô©ÚÅ·’l@ÅÈ2s6c<ê`ö4á5°·½N1š‹YZ<Çkº ÿ\Òöq¹5YjÔâw(LµJaÀ•J²2Íê½$·ú`¦Ä¨#òƒ¢F粌rÜZ.È$dœà2.Ô˜)8(Ô«A,:vÿmP¸ü§#¯ `:n.r“‰x7`tZ¹:ëþsV4?V(h† Ù­BÌ’®“0§¿kmCÜSržÞb6€rAZ%¥Ã…CäûWW3p×´lúùUÄœgwÆ*¼¨ÿ¨`ôkšcÀHUì[rÏû×ñâúœ¢ú,Ðnï(ó¿C­«D¦áÚ„T^Fƒ¹“ïïÉŽb¤oÄ™ră‡Ì0yA }.ïnMÏ”C59 m/Œ]žùeÿ©R¿TÃ&ùxJÁ~¢C;žÞÜ ®b’Ç0V‰>³1š­¿$Òn1Æ#('à '¨1²HÂ??Ö’i/7ªíwPÑ»äM0}ñvªÖÁ¾épãZVN)5kç ç¦'pØ•i{ØÕŒ;)7¡ŽÓuå6Ðë;Ê\J>r½qu˜R@Cp-ÃÖ1ÀâƒÓÊhÞó,ël‡ä9û`Ë1¨,Ì‚Üj¹\Ñ®³å¯EÊ 4ùËAš8Dh˜»‡æë/ú÷û{¸ñ.9tGêbà®åŠx÷`ȨŽj.tòÑÈsx‚Ô7²üN™Â‹Nª°Õ ƒA•z_ƒ‡r¯bš'cBRdY0Ž}bìf›ü¨ÙuQ6"â.å'-ÛæÛõükV{ VÛ*¨A “#û“Án£r2açć6à‹ç&VÊÛűðVYí÷aÚÁ"ž^žaÿJq~ŒÄi ,™"Ù¥Q‘L2€È*±«-hSÎ;q@[ î žß ¦V–a3ŸaQMÑŠvønî+a»åü{Ž\Å- ¸ZÎj*»·7¾nk¶±€©©â~pžEVña|`ŠéñÊe³£-‚’ÏÓ°×- N‚c_$w5ñs^O&ö¼$·GèE©¤9‰©“‚OÿXøol€=ƳÄgB45]}AG7ªvÅï3OnÆvÓšÔ§&ëîO¿ËnJ¸Ý­È…6¶U™àŒ‘çé{rÛôÝãT4µ„GŽ2ü§m \ò iýãBg1p¥óð¯Þ[U:“éq•[PŸNÈÞŒþƒñÐ…þê®Òï8]Çx6Y ¼£¿& }ékR¹B Yß,e™YÇ·ç¾»ëíÎ-ÙâŘdy<ôë ^žÌâ×h™8>%s¶KavŽ‹²hœVŽhÍ`“cÍ;iÏ%U?ÒF}qçÓ‡Ep"àâ“>@Ù`ñh^/°ð…âzä\´q< ¯ÐùÕ³½T«(äÓ‘ìª}LÒm±P‡ð¬\™:Øà×Ö>Ó" Ã@£aØ“åëçÄØ}jcó„=ÃnF†ÈYm¿)Âä]Â)ôÎlB3žÖDµ:ã±ìb¸§h§—O=nÒ¯{ÈѶ‘ÖC³…RyX)Jᘣb@]m ´ò1ZO—ýõ«,îÎŒ%e™P…×Ìï›ø¶k°æ†móÛšô¥¼ {õ´7`éSûw•‰Xj_êd’˜¶nº5àZÇu(yîZì?66eÝÜèÛàd¯4VÏB*×Úiö]äizíÐÂÅf—º%|_¤Ég¸±‰ÌMÿAŒÙ=‘šnÍøï±ö-Ðð–ñ?{Î!Æ¨Ä Ý¹¨1"£% avƒÉQ‡«™•dä ÑD†*1ßàðCîïüz‹}ÀUÄoIÃŽÌä±£²B׉âû²ÛkJŒqÄÔ£ž²iœHœ«`Ö€dú¡èllu¿FIwl©¬‹ RØTû¡•žvr›Æ÷ß¼-Ü@H©9NDëÅs¼ŸL±pZÅÛ_lX×ø§äZ}þ’ix§^´©8šT•z6»ó'ŸÝסD¨ˆ8¸Â…ž«L¨Wž’l¥¥æWñì± œ2c- ‰J"x<5É‹B=ÐÚÔ¿!Žs€rùšVlK~(@¨ýÍÔ?NâMSàÈê[o¼~Sg;¥²0Þým p§ŠvˆœY{K½A¨D5¶œ¨nKÕ i‚YÛ²*úòiq!ŽÙíÈà'ö=°Çxº=fMRy RäÆÙ[8bá…dïýóU—“ð·Ü("¨Zºˆº•Œ~dš/°´KMœqxYí¹$%‘aa]|eõC5¶¼b®àÝRjJºÈAf›ê x-rí‘1ùÆöe"ðêôžõ‚—þ> $p¾W‘3h¥VþF¾¥“mþYLDkày÷ƒ]¸¼ [uRëˆ Á*¼?µÿ“8z v<üsæ6s Àu|ňN?{‡7>CøŠsðš°t¼ò¡ÅGnD’pÁ Øþ´µK`NJ[–Øp³Úòšý3åHØN|Âë¦?¬ô½Åܵ_Æwr‹°)ˆ@tR…RÊ••¯­ Û04„žÄÈÊÿ}˜?3Rù„•9@,æ;¢9ò¹½ ò×ûèúDdbÂ×ÛÛ¯Á[xµ{L†TJªK]3®³îz† :ç\vu:ã'6`F;‹cEÛ0‘,=TñÞà7î{cåÌ0‡ }ž½%©‚¸©r¤F 4å¡\)01@ ó¡Z$>þŧ-Ê3¶`r;2Ll9qRkNµþYê¨e\(cXÀ…“J„ÈÎ| ÐNȹz´ÒaÛÛ\$€$+7Ùcÿ<~(záä§ÕÂlxµ%Âw;œt¯ë/º¸[\ieàdh!Ù¿˜â¿½Æ-r…Ôl 't/ìO_M|½D‚Û¼­Î>QC÷ãÞΔŽ9 +Ó²NqkÐ{ŽëÎmú|M.â­II¦§¨›Ö‰s«ì¡ÊG øè†>¤m+(Ý?ñ`AÍ–ñ0¾ÙzDÔÇœQáäp.±SØp]9ÏKæqægéǪ÷AøÚ¨‘÷i)Ãm—¦Ž =ÈBúÁÏ0¨ ØÅ¢)™€Pz´6VÅêØ£l埧\‘Ç aÛûÝm)1·/ëáR ΀~.TÌåA£cøÎÊ÷ª€…¤]Üîÿ9@6ÉÛȼ²#Ò#€±¡9’òœÁº¦Æà®!š°@ü ™”tì KPQ¤®Ú-Q´|<®›t (ü5½g.¾QG¬<bé#ç&g×Fµ‡Qo{RçŸý#ÕèþXN…‡4lN£Ùõk˜Ñy¿wF4 SÚ dëÕWjUÆYÞ· ì–hJ¨•Í‘·¥™Xµ0i”j=çþÜ˜Ö pHX>èG³Æ­“®ê§Ãjk„ÝÕî@>>tÇ·IÖûbÒ¬]ÆÄ¼0ùÈG°˜é©.A|hx‚0™ø¹íVM"UÙΰU|öh«b4mŽb22Ù.¤üF¿FØòY‹1ÇÚU< \ ‰áRÏõQy »Îx÷:y^"U>ÌACœÝ’YŒäUrKý|Çq”PŠçÔÚ¢÷b/Á.r"ü6÷+ ¼!0Þ3óE›ȤKö’ÏK‡‹XLÿ$â{ù…[0àöv…`ñ>{FS*ET‚i½m6mãògþ®ƒ¾ímë…~ò\´@>Œ\ý´.\gVãiQù´åŸ*íÒÍs†è WÚ,áúñdoF €£ö›I?0᳇ÅZÖ©xaž£Òàad{mOÊ [jgïµò¦ÔËqÓý!iùsu^¹~QÁý‘áßÙã¸aŸ¸~÷]7ñ®Iùëa³ úc‡½À;-/Ä\æv?¹òeg88çÅ€¸Þv{¬4„™ <¼=ál»Â[Ÿkö¿öD›ó\øÁ=†o’¤PÈÚa÷÷ŽßœËÅ»Ã!£z!N ¹®2éƒäšªñY[oö–Þ€YfdçÍT¦«t‚Žû9#wz¡T“¶ä–aðÔÅëw"×ðÐ`Â¥é¼yâÆi!:)ù“ìÌ›6Ï[Ôq kS–ž¤¾@KgvSGÞk ,QuˆlPôƒ˜ÅÂòЊþËžaþþ¢ÞB,}Н‹?¶ZƒÍß‚óê9_૊Òy©ó;Ú—ŒGôjh›=$èVGG¾· fîsEÏãy2b'CÝ|¬Žö¾ Âuù¨ÃØ4@¿»úv÷nõáâëÓ:þ¨$!7gÖGsªþ·¸ÌîÝöçñ«ªÐK;¿ sˆ j¢HdÀ¼båÄÁYÙ|ªâL 9‡¥¯¡ |VYxl½ÜèvïŸ;<ÛtÛ”A¥Ï+odrÎÊ¡œt'À´-¾+XûfÎ4ŒŠ«vçº %6ÝA=ì4e3\Ê‚´#ò"4±ölžÐ_>·à2»Å¸XÚ9µ’Òø6ªÍ…wDFœrÝ4ŒHèݵF&Ï@,ÆtÏ+¦ËD öýéd¡#üú;aKG}+¬Ñ2¶GÕ{Ê~*°ÃБNÒúdôÒ' ?Š®Wv‚½š9qƒcShù1ê.p rVv/Ï™NŠQ!•.f …†goYòlõ¶¢½Ió–ºç Pb·çÞóXÒ´H¬„ÈÎôoÎyBWŸdÉ‚7~ÖTÅyK”aºëqyˆØ¨–«`¼‡9}0ÏxN] Qéˆ,ÚX¦ É[epV×°Xü¦æs´ýR¬Bš½L–Àjƒéø¡»`*R‘þÂ…G3pzá⌦¼Â4û´Uj¶Vìá͈êBc“1Çj|)PbÔ2rš]«ÅÙ*UD8wNºVÑî@îKtÓ«±h–†|€ 0xt Di&ÿâÛïeÕ§![Å6{þ\ÖIÙ–éS’–ȵ謑5χéNi0hg{8dc“˜ø#©CíOùˆMbtÈL®Y`ª’CÓ¯‡EßÅ“È àS?Ê"GI¶®ù(|ýìí{ø†$H Þ+0ž^¬Gô¨TÔÎOJ¼Õ–`ê$žÙüáieÁCtDƒÅ|Çšª¢\MõÁäÆÑID„¦T,ªø4±Þ·¶/mÛ§×ê^Ö@úL¸õO`T¦JŠ™›ü:ËeEz™CªX(ðV«ÛͤP—&„OníË™<ç6„.Ü,£òÎÑEÀLH}D%”|‡þöpÌúàÙûöž#­QC?3Ï"Œ¼)¶a1ý]&DOµÄF„Ô G§XàÀ Æîʼ"­ú®>~2£Ú/–³mÞ“uÍæ(TóÄ– ÄSZ0™3mé)éM²F¬–ÜéÈÀf]— ›(P&{IÊóyØ“Á3›ïG$ürB‰ZAue`–\램XØ 9 ®¯É#+ûsôK…¶Sm«?ü‡A«ãïvšN±†á–q&VF¬Ó-å¼ke¸BbA¼ˆ© öÿéT*ºNÉ4@›twô0w±$zÓ>*%ZÕãîlNCŸvŒ.ÝŽdäaRû^;åÉD¹¢È¢ó°þÙ£ÆÄtºÒpCoLüI…Cå! 8ŒõZæÐGf{æùpz…L½.ôßìFîvP…ïÍ8*W¡3!´ªÕ¸§çÝÜçUùö³}·g8ÑõYËóNŸRú°‚3V2J'.`d“_÷WßÁ¡=A惆1HÈðB/bÏVèÈbqz„ã7FH{½mÞ4uë¯x/ŠºÐù *æ“4Óåµð5fq¹´‰Ôõ«Ý0&„hmb3Úv=0çàŠt…ÎL«jÃ꩚eÙpÖú—Íä%=†WÃ0*1 ‹L®BS«ßW·|¡9óöµ¨ ˇVDèzŒ¦ÂU^üâ„…+Œ€“Õ0ŬçàQ5Q|:«ÆvŸâ÷q²•[‰Ë½Ý¨É´"Ÿ”ÁH–®’>žüîÖ³âB¢a@KØâÁÈáyÎÎñ²\¯–©ÎÏdžŸà$ÆœÔE3ãj¢¼‡ƒ"„%v9»Eå[‹°<.?åø:¥ºYBÆ9 0 çËddÑ–´LR¯¿wõæ@¯¤ßTÜYW{ß¼hó1"ZàŒ„JÉ‘¥¨ˆj«ogk¬äÚ€Å_/ÜPÉKR¬ ¿Y—¥¤Éþ6Em Æ|¨ z¤²ÀM2Ì5÷ãDWF]ÆG”§ºøü17~|ÔÇ¿(¯W|À(€ªOæô?ÑËMÔ!oD¢ZÖnø_{º¨XCn<ùŸ%–‚G~g&t've<ÚŠ•¨Mé">íÝÎÝ\þ½“6+N2ˆªx‡ K;Õˆ6ôÊBpnoÁ¨R±‰Á×-½Ù£Ú7T±‘HüK~™,¿ëYò«â…ÞÔÒ½Š¿7™Û$*1oñä42% ¸nHâN¶oë#ÑFÿÄ&)›âÉ8.|ìDÉÚ+À4bøä¼­ÿf|™"éf Æ{¨^ëϽÄ-3ƒˆ©X ˆÍߌ…* ºõ S&t³ã@d¿1p>iJôtŒz·«5v€vTWçB…1DÚPš;,OýAÊÓ Y¶ûRÏ¡~Û´»•ËisÐÊî³ÙÛåˆ ï%³FÂ’jòtOíºŸÞFúT¡m©aQšr”ök­‡Xa¯SUA}»XŠÊyp»1c8€¥õQÓè~¿ýI;[ú\dn¨«(Ƥ ›6‚uÂQxáÍÄÝûÃK‡Õ‹ëðD‰à•dÞÅÒènÖªI’îȵøØC¶™‰N2U°¥Á^ öš—Mrç¬ó±¥éø2™¤1Vpfâgëÿ%"[m9Ðôͺ6± '6BüuÄø…?UßìÊP%FlLœõ“ŠDBÍ´º*œ>¹&ßgVáJ ˆr/—®nD…ÁeÚ¶:Ó—&ÛR«­)ÌÃÕ<=VÔj¾_eÚŸ2Hæ¯ð©?ƒ"¹6V Êoœ™V\Æ·à癦˜h­ÂuÖ HXj³TµTˆ¹I³û‹5ƒ/ë*ÏŠäRS n4˜ºN#ðÑ>U뀚 UØ€©ÆÿŽZ”õ­$Ö4zU*PV4îsï›ú%UÖúî¼ßÂŒŸÎlîÄEê q }–ðH‡T9¡ùÕ‚`­*h‚ ÞÆb ^=ýˆ«úÎqS"þH&ç/RG¹MJG Id70 ˜ç¡*æ 'Ƶv¨o­Gεõ…×C‚Bæj8æ/’!]††JöÄX9È›?OùÊ ß: ýÇþåÄ.~M€tt~¯ ÏHôU€=Z UpåçïˆæJ¡ÛsòQ•©.ð oÚ4Â~PJ3 _ñ}Ô ±¹×>ÄåQR°¦õ¾‰EoÉ 6#yzÆ.‘>Ï“²XJ/;¤LÕÀB[ãþè4x¼Ä0wªËņŸþk¦Ǫ(á–KnÕv)Žîäá3V‡Ã2»&vˆSš€/ÂxøƒL¾n©P Ëd1Ô†­š=Ëh¢]‘⡟²Øâý@?Àd0ÂÀRݬžªh=!Þà‹áDPâèȧSÔS`çî]ð3êÙÖ‰áYoDgé ø7‘³ˆ¿/»"Ò¹í5aOOÈüú!")¦C÷FUU5Ýù$6<)‘¸!¾N¹Ãñ>@Êô+ÞôÇ "ýœ…Hª¼ÔÉXÿEà†N,@oߨnØõy ™c_[Nø\©WweŠX¬®#JÚñÜÿpמrçï ÈÐzˆ¿¥®¸B7EMe¥]5'Çšq–*Ok{Fb<ˆ…Å’ß)À ßa•ŸËþæ4…z Ø%KDâ‰Ï¥Ž3Ö_£€†¸d]¸ߺ û&ºïX±X b¸^!n7ø¯2h†šp§§2ÁHtLOg2‚ø4­?‡Â¶jÈ+ò´žó]¢…QL¢–ƒƒÚuW ö†¡¦cé²ë„ºY®-6TYàŽ†óõ±ÃwÑcÁë3—‚«ÉÚêetþ y&$™SãIë¶Nê*}lGEb@&2Ê÷Öè³á²´Ò"Ä£ÓYÂ7….±¯¾ÙÿÁʨüŒ#ÐxY¡À,ý¦Ö?’|NŒr×þ­s¯a´äÆ’P@ª¿û™WP†(ôÒnò×§Õø˜†´sL}Ͼi2"ù뮣8ƒ_ç Wæ »dÅü§H©$yNL‘õZ^³‰ð ÝøºXRt—¿Âet«é$&³ÏˆlnÝ6Õ‚JDRÔÐ│8&ï{×g:>-¿`T7é´¤ì5Hj#ñ$¤àýý-L°À–çP:>žaPÒÀ'cîúI·IC§‘šŠ‘÷ñGW8x±¹ö^ˈ<øfÆlÐï¹íu%¼yøÏ  •}ÄË4‘+IeAµ¸—©ZVüÝâx.¦x=,|w2ÞLœÍ=îɼ[—ÂNö‘Õ •^îÖvùå¯XäX±–Fä+Y¯XìÖH/çÅâ Š[±MV$BŠ+½˜RcÖplkÈâ«]!¼PËÛ}Éuµÿx8]°­Ž†!p0«äÕ[¾&˜nÈê/ψü™åF©¬Èu¿H–ý”nÞä8w  ™é¹ýúª¿¿âÖþŸ’âj¸býÕöGÔ½·¥[0‡«d¯*Y­LöZí¾½Ãxé-†¹½ïô‚ø–ß8TV4²)0NF›ò3y Ç!ÔÍ‹½&ÜuüüN8‡é½ÛK…Xæêc­v~Y”ÆŸ÷^Ò£¡âJ·âë:ßXmÚñåîDï»–1]î‹»,‹Ø)[ÀúšÇôRYѰuR¦Nõ7ßÌö†‹ÅЗ»ò(Á!tæÜ¦¿/n†B‰%P+2Â1¿Yü ùgžüÞÇ a—vϵNÿÇaÛ6LÆ»žú^ÐmÞc§ïk:p/åjòªÎx'sf›si/hïkûõ¶ý1" OaßP¡ËY€°:ë½½XŶèú"BïdÂg¤úˆLF%Tæì˜^®â•Üù @Ò„úkÚÖaZð…ˆä:´9P?†ÏÇ—Ÿ¯m(ø±…\"v9Kùµ/P: ”Ùš®Áû¶7–aS„¼Å« Ù7—‘Åàë ÅPB0p5ýŒ‘ƒGì¼)XÐØU³]+ °y°ß×ïí L/3¢GA‰™=sS³Ø–kyɼg:¬?c‘)'$mÒôì©¢}Åo}g‹+ª¿:ÔïŽ[ƒá¦HŒ­Èï,Ï;ìØÍL¹ŒV2Ä®?PÉÔÒ‡‡˜ˆ? ÐÄ"ŒÜ\³¯³ ïsú­&<°cÖĤ}೟MF<¯ŸYé~9wãÊÐñÌ1ÜÃUœÆ0¬´?u•k §J!f7(æô.!3×V™t0ëÖÛ]7“”fÉ4“‡TgxŽv€†nrÔ'#'¬ºôûe…ó$Ì资 yèæ[{Þ^§Ü˜ayxêJ1MkM³)g•>›Ùºbš™¹÷¢±Ò{mjñOZm¥ñyiIúà ²zÄMºF4f÷8õ½X¢J N–è} ËìYrÇDZœQsð4|FMzħÙP9XE¬hOÒÌ}èN¨ª¤&º'y qŠk¡¾ ›%=IIr³Hûs£òm¹>ÄxX6dÿ[iP¸·|¾Øwf¤Zt¯‘¹ÚÛ*àYdõ±"g.×®„~NÜ„ÿ¯¯Ú§ì°æ+îí©•9Žc%Ïõ€pNíb6>+ä¯ö¬¶ .u<ºþÉ̲€ßI=½Ê8.Ì©tjÉ’i$P£È‘¡6_%I'W1“à xjObÉÊ-—‡º)ãå²ë®4DؤLZ1ãÚ@%§`èQÌÊ>”]/+wvo¨ÏÚbŒ¸›U-!1öÏ5›§þB~ÉÌè³—õùOü¡Ù„ìéO¤‡kkåß“73æ²~Dö Ûb%k« q–ØÙW1#vº>K<ÖÚΗ6Suû*­Rœo.›È¶bº.†—(W2Z'úÌï½zñ^€À•dSH… òðö°Ê‹¥‚ÉÄ»ŸEä3%õ¦¦ÝuŸ“ Õ¸0î•'Ÿ‚¹ #73dñ‰Ž¦)äºT9ÊI($¶æuOíÂ?=ÞD?}^*÷¿‚€ %,KÓQ={ÿ$a„®ÄåSó×…k“ñó7}å” Ô ™NeþËÕ€âPI&-Kø¼Æ&úŒ3 ÛÊ›wˈ2Ü Å )!k™¬Ô*˜PI,Ù‹^ä:—hu^òUÑf@ìRbÉ>´GkpÔ GL ýó+†¥Íõà=N£kÕõ5#Ëauó¿ìýJ·‡^3RPÁ'¿<ÚÙĵ»ËœQ–r´³Â·m²veQ‘î™q”Ê$c"r¶ ìí¹ÀH4…ÅYjÌ«àZ0tçЩ‰¼.sãÀš×s,ÈœË6>8hêš¶Ú{õuGi‡4"P:Ù-,foñ¤5«@½%š3°Õ‡Ñ ±.Mnžõ&P¬Êb<ä3†Vˆ¸\Ý<òZðºYËd²%fÖ¼­%yß ìÌ'Åù MÊGç27&Ø(|URLËŠÚU˜ç1‡œý—pœ~Ü–ÍZË—’p‚3òx‡’æì54†…šD6'‹\çÜèÄ-¡ìèç׊ÏímâùAbô/´ eoİ‘9r1Ô-šñ #_ËÄ<¡çEêßG-FßêÅ…“߸¢µ yŸ³d‡ÑÛºw˜ÒâÓÜ‹§ã;“ŸØ{­xŸÙòzÔè|ð²•L"øÊ ¥}v²­ä#~.fÒâÙ)(\¥e”ußuèi#”K=„î¹Ð–›„1ž›oû¼ì†M(i¾Ç¤£Moåæ>M.Îu'&ÂWèfÕ€JÞʇŀ@ñV‹èÕ 2@ƘžÂóŽÈ÷u©îõ°aelcHöÁN†ðL‹ ßÅïõÕ”,{j§VLFWØ­ëTßì A”vË̺6êÙ ¹¢¤FÔðg'œÜãð­·åô2^§,*U"©Oó“ ó’A„xÙ*K˜î4¤fÌÁÍ·õ¼ò¥¢x^1ñJÛ –FQ<Ó?^€Øã¼LÜCuÃãéN _nrά^¹ªVýd£iÁñÓ®Õ¢$&*Îuß÷´/GˆÈ‘‰T¤¨Y(|´h'—çó`J[¹)Ì(þך9¹ÜaàòA£TK¢ØgeDX.OÅ2/æºR1‘qUàãUÐß,)•iòÏ„íÇH»*‰þÔâò¶•ÛFñY‚n;’‹’|Õ£«„ûZRÍ?´t¢AÄ+M á8aÓbîh/Xè,ÅÃêéyY_Ÿ°D¥“Út ‰9Ÿ{Ê$¤[¹ÕÖõ hilyòêqŽc›]»lùŽáyÚÄ‘;q8tÉÎ@i4—U—ó8_Ö "uaC]G+¡BH(Ó³•°½Q“ŸÄ¥"ëò'“e‹6'ÆÁ½`{æÿÅ߈}¨òÎTÃz½¼û³\0Ìz—Úî s‚Z××›$¤©x¹¹än.?1|] šœãÆóƒ“ðØÖ#E^ÄÏíǦhg!”Sª ha MSlÜ™J¸ƒÒÞy&ͤï`ÔωÚXŽ”œÆ¢´g±ö>ϧš[Eè6†Öé:ûIq†{[ˆ?eY½Ô/Ù5ÐÚåÍ#c÷½zêc²vQxFÎG©"‡#wD#øü—ûaî®yt¹ó! PÎ~•RD£`.ôGë ›¦–?‰¡2f¾ûtI©Â}`µB.Kä#Emõ޵qHì¨=gîoáìõÜÕï¦â™’®B«B¸VÆn~š8Yué±Õ˜Äa>X’Œ+|\Ù‡Ð9X‰ÆyC+ô•[B^r|ë"³Ê"R'k‘B]EWXM€ó$ñƇ´0¨PѸœy´„ z÷«íLœeÙhýÁú·U'b„s™ïì·‰šý¤HÆ$¢lBΤÏ$뿤93œÅ’ÃI`‰¯a €P¬KûØ÷à ¶â‰¢Ã‡7iŽ÷ ðƒÒ믑~eø7Q³µZ5tyXÉ3EO’¯·G‹/žºŸ9ê÷ð|oòîàä³ rè{UvSil/¡Év$ÃyQa‹¾o(vûÞ߀šÕøî²Éù癕ž·XeÂÜCÒSVóÞ™ÕªÆQ2á” öf™[S—‰æ±Ñ|)㨠H©ÅÆ>u¨e®±fÐl5E§bÝ_²°<„l½T^œ„@GYÆ’›¬¢„Xƒ,£_6rä9 .v8N@g¨E±7?³ÍÒnàeât#[¤éÿù#‹ÊJ9GrÑœ$T1F8ááy´›EéÕ ‡ø`•sáôde¦ØéÑþdÌä´Œ"{±vï‘Çtým<Ÿ~fJáXç“Í£Åy òZkTmÚ-·”rHÛäJýY%ò…;Ê”h:DZ=/\ý.Þ|2©F‰NÆ5}6 ÷÷f$¨¼PEFm©Ä ɳ¶‚=L ®V³Y1Û÷j]H—" žZ öíZ‹ÊZùÄÉmóðIƒÏá·iCÿˆã{”‡öòp2èxŒ·¦l£’ËUރʆã%‡tUH‰arµè!;Êá#æ8º!x$ÈÉr2ÃÚã5™OUéLqÅK|ôÖ»Ò—Ä"§rDOà@/¨ö‰—ú¸ÚÕâÚw‡.„†Ê¬ v›J=âµáQnÓ•è(§ƒáA?¤ÈÚV¥Ö/¡fü)s¦X ùðõ-iø†Ï½á?iö{i&Ãæ )ö€„>Má0ß“©º8u¥ñ›X­çõw'I‹Äj?©úB»ÆïYÕúdœh“w­y,*g’ÂLž=IVæsº _¼ì!®ðãÇJëy?DôÚ*`{ÓjÏê›ø²:së4÷7~{;.ɮΣà7L°£;Ä èºDTSi\dšPä4ýlÝ¢( Q·mÙáœ5€j9"JÕÚNÖye·ð¿§TlBî‹VŒæ»Ie¦äÀ«úÏ—€~–áÂÒôMÈK‚6|î¯:ø[.ÎÿxƒÈ‹Ců`¡áŒt_3tþKƒ­P¾°A§ËðÐ…o~õ9%!Œ©¢ ÿçA9ž­ þ dp •Š_¼ø¦Ÿ© €žŸŽù"£t¤„ËôÁI!Š jµÛN V(8 ‘;¾Úü6a5{ ÅÀ £3ÞÖBî=$nµK$2lV¼ˆcw‡ý1oäêúK)F€ðæµ¶ï—›Š+`‚ oCíNŸ)2å‡Ò.…gP·?ìäj¢þ ¿Î匜ÝX²>·²ÓNWp0LJVh:Œ2|ÒN%KC[jøuÓê&–ââä‰T¢”ñhÕ+ãﯚKÚöB7:-¢(°-.ö„pÌ%ßêË_¹Ð´mË’ŽÙ/kõB ™«¡dÁ¡VË(÷Ðò· ÚV$YaA1ç×#>`{÷À—’¥¸ÇÊ¿ eg¨NCêD<<ÛÇß {só}ÄÊRèºG‘ºÌ „fšû}vÇÒ ¿“£¶Ì¥/îÁü~r"w(¸Ìœ°Fê“(¹O˜ÏÒ6&zbn1ùÇ¢¨k°óõZ¤_­+\s–¸ ‹åÝñ¸°IËãñƒ’³¿Ù¾3X®M†Ot!kKšÛڿ箥FÐ>{6Ù±}ƒˆªÚ2ÛIWÆ‹tÅ3qcVcDºJÛ'w;¾`_[ínV&Ô›þ.‹£ŠEGÈ÷Ï.ú&¡ùÁäRá jM!ûL‡Æ£’¼D¢‚–=ò!|)H'rBlÝ‹9Kz/‰yKê]x ,,^ FO¿+¯øO“Æ În oo3ó˜ÆV~ޱ)~0Vv† ÄÊ÷öÂàÞçÊ8G)_NÀyîuœ”uYàûh(ùOW–ÉD­Í¨µ6Þ‚¦Nµ@±¼wR‚JX4NåÃÛÀj§#©Šf¤aÂ)½ýúl YäPÞZÅ=æ¿Nõ_J†uâânKpù –—§aY»Äâ¾ä ÕÙÀÇf©È%² UÖV)‰¨7‡éíð´o¯RÙ¦ñ¬Uµ.ìºz¼ Y $C(uþsÑ ƒ˜>;à”õx¿­x"òa>\7„,žÞ“ûÏÇn<ýÛwAr ?ÆÆC3äÚåã’¹ †\(„)@Zµ—cV:ì!ä£ö"x' ”“)Ô‹TLgˆßèeŽò>s¥xºF wD$zéœìRí̘Ýl“ÿR\^we”fd”ܳ *ýÞªñgäL ã¯PïÊ8‰ï ÆÂ²ƒ’>G„€µ‘vW¬§qÑl"}%âÔƒÏfÄüÜ1¡keÑ;"Žjr³Î*€kr›z±ÙÂ˾5~”•™D‘KF âǼ"&­¢y(G,»§\¡ 'Ìù¼ù„ê<×S‚Z ÷Ox©Ÿëh}(Ťü›{‡É x± ½d3ѳÞ<ª;åHõÛ¾?#vË)XR’lÈv¹JоÞÜ?ã›Í¸A8_ЛËo–·*_BAîæ£ìHAk >ˆ¥=W|2ö§Õïô„f2&”§wQÛ`ÄÿÉÑ@»)ÌX=±ÇÆ´°‘‡8ái(Ž5_¦qŠÏBœæJU½Ó.ÞÁ[‡ gþB…t×µ:‘k 8@é§Õ¡êžxÌÓç^þ‹íÅš¸FÖ?„ùÝÍA9¸›ºHÖÝE@ïÑ&êY8 ‘åó¤Á=+=ô9¼ôð ´êñš¶šcÊ=j›<`Ôg¶i<ÇØÏxJá˜h š<íºmu„ìëe¼în»Ÿ^-9mõ^-MºEŽjîͱs¬Á,'›¥lº•>Yk…€`sÌ+,Šç¤ÄÇ>îÄŠŠabÓ‰«·5”ý%àþ(äÓoiwó§Î¼ò„æ‚o­™CBˆ£ãË8›Co†¾gîÄûã Ø;ÍR(NÔe ª‚ guíªïó:9¨ (üü”â×À`šP;ªÓþó™ üWø7Ìè 2U’|±±Úú™”8I°“³Ž_% èšCÄ[Ò ¥"Œ¡ÛVÝ:ä—›Öì­™ú9•šy()ÓþTGFͨĦ:IøIÜeb}½Z[i©oÄ«bWØx¸+Ò!“›jú™ÈºS‡qN2ˆùŸö&AH¥À¼Ã éœÚ„ó€‘+âe‡=£#¶”_RÉnÆFçͪdÖ*h/…]0Óõ‰‡¹èrøñ݈L×E¾xëìÎÕÿÏvCÚÙ0ç„Öõ˜ª´ ãÚF^։­ŽÁÝ^t¶y°¸ïôÖ“méELGUyC".ò#œî§$Ê'ÞT~Ãö´×¡°ÜB=»×IhŽ1ÊÒ]rŽ$A:M²à¨U|~äX¥¹ ¸µ@&i†ty`¤¢ËßP‹9˜XŸº!} œ¾ïªI칕$]ÔjÝ®SÅ0Ç”èŒJ‚_:²ÐZ@ýµ˜æ»;²§ðÝÁ3ÈÅr‚áê-.úœjü“¨Z‹ýŽ ½ã5ïæw[?>½ ²ÝØiM4Ç̽Wã“¥°£ÓÇÇ_Ä3?`fÌ¢‰øÿY·åH«&úN†7¥›?*ôn:‚Nó Ørô`Šª‹naé². ŠsÑr½ äøÇ¢bóõ ¾7×…¨¨‹“§(Nö¥dît‘f%¨ñÕwú÷çŠu£(ÿ¦"4†NmÅ+¦°¦´b«4´Ñé1¸hÂSδü9@ÖQçäsD5 ß-àØ›\`[ô6þàŸràšUo;îJ¶Ý¶9ë7¯§Ç!R`ÏÔ¥ RõóéØˆhÒúSª¹1ð’ãyeÊoÏÙII+”hÿÖûÛo¿ŸÐ(—Ã÷¢{4½Æ(ÝV¯×yä£Õpdç’t¶4‚À&‰sÛ厤5æL˜i)Ǭ÷çè·¤UÈï D.—©ÔUˆÅ»e4öÀu'­¾z;iµJÌ.mÆ‹¬Ún{e³³ cèGσ °rÚìH“¥Â(MÛßU1špH«ŠÐo“ê…ðA5Ó´NÊya¥4¶@É?dÖÅÙÖБUCyž­ÏògÁuí®}ä´6'Ým~O&25Õ_Ð Wa^‘ÀÕ »5kô 3Õ1€—«öa¸Z@Ü^ŸÎÀ7Í|eW7º.˜P©Ò#%çWBUÆ1¢oí´éî2„v‡F$§z|`ôDêßÌùäö”öêÊùàváÿ:n+²,?wºXà@NܳJEeÊ ³åˉÆõABG #Q{.Á î! YÅÒÏÄ<}Ö¡Ý‹kÚ¨¨ 8¹R/*e~÷Y!û/pŒ”š¿%— ta¢¿è¹ £;ʰéÛ, c_"ôÚ8Í¡ ½§§#/sPàzD-»:¹çnÀÏk«9Ï"³m×4[€ Å×Mž©ÓÊJî®)SRÔ(­³šý®6wØÝ:2i¿|Ë-™Y3‘º 3Û ‘jrÙª¦[ì›(©”ê “èNÞ9éðJŸgÈ=TüþÔí RÊ»=7‡u‚ÄõÔ,pÈ?a<îgüÕz>Jª…'‹7®`ìá{Á)AæNnU•;C}r’-.܉„Žø‰'¬Px6)-þD=Z9©$­tï óÿã  ‰5¥4”'lUª½¶ØN`+2Cö9Ðq¼;¡Z‰µö&è˜ »×\æÕ«àïô¯UôÙHUÕB¿•ÁÁ…sWÿ=k‚”ªÖÔÕ Å¹GpÝ“îïTPNÌrœÇw>1e¾E­7ò9då¦Ý¶o–…=á±àEÌ\ Ý£(ØKQ3ï-3'£}d0ô+ߨĝ2õ]õUЏÀYÞµ—–Ý “³ovŸ$à§îÅC~ºà…õùÙ¨.…&Rb[™ÅÛkÆãíh÷4¥õ–YÿË#Sñõöñ€X—c‰7{°¹áÓGíåÙ¢àªÌpü´ævoutx;_Ý!Ôªz92ÒÇK`þºó‹)`Tªå¡Ò²+æP;±Š¹…« {áƒ9™±`ZI6ugYÆÁ @×ÿø`ÂÀX‹|¨dJÝÃÑ^æ·%ê&¹2žvh]\—ƒ8“fnWܶ0®))ðÉÜ”Ò}is^>F³9+ۚΗ6·ô=‡Cë_ΉjŒnMSŒlN&òV§ë7òBaÞH¾v"NiaóÔ6¥¨~ÜôI›&ÝU‘’m¦™Ò[%ªK®éOE(Í ¶…6Öký¦âfJû…QÂ6&†ÛÉnÊ̓N âM®c6Š@MÑ”¥ÝPϲó¹úÆð|~ ‰Úuçž=?å Ú¨º3²ùÌÁ£CšƒÉ<ÚÓuRZHò4 ( ‡ ÿÿÂêkùSxÌ&›æ°¦µ“Tºßû÷ãic¬ûñ2²•¬™F}9¯š8\ùZ8t¸Ö-$€~óJúúnŒáä`øÈåŽÒÌ|õ+³±^†d[Õfb{ãtvüëÑ ñüÁm:‘&Ú9Ü€¤Œží¿—àš‹ ZìÆ/ Q¥(Yâ_ìô˺úŸí6‘Syív‚¸J7=kâÆ î”ŸáÌÒÐȯ®3£:ßóÇ wÞ'ž6쥇uñ™zuªhûå§Öøa}»o³¯¯ˆÊ(sI‘fþÜl²?`[h¦éºÚšW22>j–¦ã~ïZ¯tëf,È„¨ç‡7xΗ3 úp€Á²è9âInšoܱ.€4Åú2Þ„ÓâÓì‹ðRà" E9>»¦÷Õ6ÒaÈ_‰é ßb÷¢¬3Žs>ñë|E}ÍR9k”.aŽ ‰´¤Öþr Å™“JLO¹¨g$=@s‰DApMDƒu((U' ÆÖÕX½÷²îáïd>Ý%Jd}¦Ý¯X3Kâ±E7*U0"§|@”®êxK¶ãù](\`xßÂ}NHF„¦*:š–%®l¦ýšolPüu0zùZéÀ1_ß-¡~=‘,¶åV«rcö±;ÿza#.¿UüPÁ ·¦ThB?ÄeÂc.T‹HÏed~\78¿HRó©šL‹ø‹jHßñåáÀ´ôê×HWǽᡠŒ‹V®ICè[¿ä±R:Íp z=7«–Ú¸E£„™0G{I¹`ùКu†(Þ8ÜPµmí0¤KX¸ëÜÜ:Û]õEoçTºÎ'·KxÑ­³›BUŠ ­|…«EoɵAÅGEÃ\ÎéƒwÁ‹ù%8“o>檀B— `ԼƇËj\É]€”ó’¸Çñt¿ÙN‚ýß~¢±¾‚4Àf¯¹X0Ýö[¢TÖåf?Á5 ø2ßúb#õ1üœ^Nã¢x÷¼èÎüDoؘÐ+¨ºÆnZI‚®¶tÒý)^ ðöÊ,™`—K.ÄêçÎyŽvbY}Eœþ> Á}ÍÉ-Hv{§—«€Ï.D˜H‰Ô¼LõWÐåœÇü&Ú)˜q0ÅÚAL”Ã=Ú9éÈ›ý{VçvððùI°V€óÖHrÆ`y§}Þãÿ ¦Gg¹žqÊmm6wY“-dG"Ö¦I|.gèÍ;a6]÷Üs„î—PÈ7Š¿J$=qyMªq¡æ¯8S|'ð’!é™V Î{µ0°Ÿ*9³É¬sÃé·€ìÕ”Þ*8(,’=>ÄúY˜úàØ:I/Å YX ,`ò4½GjÃÛ­Gi¦ä÷k}Ëÿ+ß3Á·¤'Ÿ•.Åó“tÍúFAós Æ9"­LÖÙ{Zîs4 w?œ ¥¥ÏJÆ‚vfšUÈé‰ †"µû/í½0.»o¥5i%lÆNí6(§\؆Œ9pŠÜëéµ£W„ð̸Z´ABù"}ns[CX_t'd­Ø™ZXª!›F]á¡6þ™ÕÍÒÿ›ÚçC‘¯wœŸCêHê+é ý™´±lh[îÓÝÏÀgó¼HûˆXŽGoî¾w’V(è!OZ&ô7ƒÙ;ÛhÃÞ;» w¿ÇCµ8RÉí™ÿøªóÜØVC»Äã(F æË …éÈöC¢~m•òkœã}íoüý*’»“ŽÑmš$YÑ‹/e7þÍ`ǼVêÇzÅàu}ê1{pE/ÌúŽà‹prGIìùÅ =† èI»×dŸ×ŒjX*Àunµ wš3¤ Âd:b·‘W:Øu/¶ôíé- fã®i¼ 9'ú‰óc¦ÓÈ¥|pÊÚt ò€ê%jý‰¦*òÄ+ƒ×ÒÐ;Q„·röúò–Q€{¢rHE;ê},L*MwS¥µ,Ôaؽb«4JuÉ..MK€çÛO–‡¹ò‡ŽBBƒ;ÉA)ƒÝY¼®3ÊL)¬Þ/' ÎT\¥éw7Œƒ3¨pV?~ËjÆ!Ù¤€Å^X.¬7ã+X¤0 › Ò± 8‡+‘p†PÓt¹‡ß‚¡uM€k›*âë² †™öÔ…ËNt]Òž>€“¹lm˜†êÓ©«Iírôke_ž}{bò»|CC\èÀñýÛQ޲ª_•ó¿Ð®C4A9·U9“YŸÇ¾§%Aç®ÈµIs™ïäf·¼"Ân›ØøBùéë1ˆJDwÇPÐէȈïÑ'È~Э±=l@±Ô/`ã™'É[ÿò •Ÿ ìíÛZœ±@’ÉÄôóñ‹Wš‚šßðî+él‡]ÜrHÓ¼â/!Ò8F‘¤ò:i„…&kUÜ]5ZOb75ßs¦rcµŠßÇ`¿ÛHm<ËFæ4„°ˆSfJW…MꟲՀCIËèM¶j¥õ’-õñP¾´ŽWB~õ³`‘€ŠÖø“ÙT*j Úû—@°|Ã\Õ án¶É›DÂo’ì:ÃU–©¿ðþÝ™h²»A‰IC¹’xŠ€ÜçCÄpå@?ƒs©ŠùÕF¯˜Q- ]âo’\\Dq|U‚’›_Ü‚ d:Š’QµuSë÷¡ÙS´ÝçeX°å©Ó*Þ}ÁÔ1Š‘Né›Ð'8`ª% Ó}@ö„r@5Ò•f~Ç:­š„P†š”Í3ÊPÞ ñÏW–)ë,Eªî¤øZø’Dþ–Nô%—?Î]Ê ¼¥ëÙµ€ÃAû:¡±-$ôŒ¨çªÜˆæÞir=¦ÁE¾Á¬Ñ) ßx‰bï.28ªKžà[å[9“6ÊF2ƺÀ…Îß²ZK„¾_טl*sÓ¢ûC- ‚:"$¡öbý˜ÇØ…uГԆwÀã†êKvû䥴 äzü.¥Jë¦M¶4r‘‹·× Ð_ƒMä{ÙJ HÅÔÒ0È Ì™âUþ%npØV=ÍÈI±7¡H-çàsKùV‹WÀ·jf0Œµ¨é:—XyßmnoØ»¼BÚMžîJ×õT•Ðò°çsûÅ&„+4ö]\â…Ø´“¬A‘VÕUkÄc`bC"þ½ØKñjþÈêkó.EŽøÅ~·PÏ@å;fÚú¼áÝÙ«¦Œ[0ïsÍß,]4-Øn¿oi˜Ê¢:pЇ¶þáiyÛNÄPÀ¹9Ûc¡Þ,r{¸Šñ†·¥p4z=õ‚è”l“K'Ólµfh€;~–›OáœáMÝÎ8¦¹Ì½Þ­¨ÜÇ \†‰¹œH ÓD| G¬| (ó&1Ï£…÷1Æ*I‡2Îíü,¤ 1õu»ÿÒ½ ý#~6¯±Y/ >Ω^V´‰8Øù´3p’‡–!Ǩ IAŸ1€¼Cïqÿy…ä\ŠÌvèç{ÂÙ‚xØö™+!êmqÇóV›r)¿Î@Žl|é7¦´ ˆ³gý®5ç!ìÀ´Å ,Ū¸‹fP„¾bu΢?ÈnÞÕrÂÕžiÇÆJ¥Q19† –Jj¢‰Hކ¤*6…#hú¸[„@ {u: f ØDZ–þ­\)Ü|·:ZÀÇO/âÉŠ®çÆí zs|;^CD,v?qo„ÚzßÓòñ·ÊûŠ-üu[ Ìðêghvq“ÙæÍf|Šö‹µ#aªÄÐúsÿÓ ¥*•"•¤IE5¬Ü¨ù©âÜôŠmÈz‹L¨¦^>ìê&f­FuÚÞêõÅM-iàœ†bد†52Iä1‹…+`›ôsW¯¦¢hsn`쇥k Ê+ÉÜ‚‘erœïl¼¥”Ÿ=|æe¦=«èѯØ4ö¬1ÒÐÒóH¼C5+Ò‰Ú~ú“3.¥r„”êÈmÇqóqRx¹<®ÁŒ`†1@<ÍZ7üæÒEˆÎ¨f¯²¹ôy•DBÍ"‘F”Õotà¶&ÒΞ /b7j_%SG ì5egS&XˆÔS¢>¬Â¸#;;çºRù(æ =Tç0+çUË#Ó—zTÇÛsÆÜdÉ,å.²Ç†¢ä>«µà¸¤e@«IÉ"ذùu׿cׇ„®<ée¬¶¥'åÆEwì²ýPØÐØsG¹†jŽkj Ñ0ê1ت"J™Ü` ]Ž á­6¨Ë#UQ(Ü­¹õ(,Œù̶éGí–îa&Žvó½PÕ¦tÍ&¾Œ€o½á3t¥ØââR¤Ñ6 @º<†24kTƒÛÑêú|Á³ùÊ™¾öwLÚ:ñ&•Æ’J`à‹¼=ÂPh3[0†×UÓ]íq¸Jûfßvx¸š†8 g,3,!M´N 6TØîÙšʇ…ê‹ Å&§•RÇÉ¡ŠG Ý‘ºAe&€ÌâŠ79éõ)bj_å’ñ.h,ñ wÀÚ¤Þ‰’LÓh.GîiW=Øê#OXìVlÍäN#pí/ ò½”ÖYF *Ÿr2f ±ÜÑ«ñ£ÖiŸè§o»…˜ƒ€§äŸ2ÔV¹ôý¾±J*>¨€_Gd>€kŽö/ÓbºÝZ$á\¡Ñ&%‡Xõ-1+üb:²$ö~ÈoŠè‹òàŠàÅl)3yÄŠqêW{y»$š–‚Ä •,KDÌ€Q I»ÜÀPIOŠ:ʱ…È‚Š)b®j¨qÎw{`E”~…ÙÑÃSãÚm÷&‡‚jC$ØV­Q+RÌN¦Ç[Œßðsaî6±Dª= *íÞÅîÊB²v2µÁðaÕ1ZA&MäH²Mz¸4Õ/¡kŸð¸ +Õ™–N2Ú·Ž>Ü´ÙVŸÀ“%ºñлv]¦^'œç*@ïå­ž Zaà LuS®LxÝ„½È?ñô¢û×ËLq\pÖZ'ÍÆ„R‡þ!' Hërɶ®{EMÛæH8I›uTH) x`a91IÛôªlÌE³æåTÏ ´½îtkN¡´—•Ó3>þ²Œ£ÿ,ŸAU¢ öº¸Îfä°7ÖÏ7NZã¨r«W¸&òœ]Ääß"ä¤ë‡à‡©ãËNxfѤÒuŒ5&µÌ1wå¼?‹P¦TÒ­÷E<«"²Äà¡®E™&ÈG Øê†…–NüNÁç¹Çu{»¡4rµƒß%¦átN󚋚¼»v¨ï¯]ö·~©ÈŠL)§GW¬ï«YjÆÙ3šÑ[ÚX–5Ñ]2~þJ‹–L…> ÊFL»½ç{½paà¬O0¥ju§ÏD”Dõó'¼Ú”[â}ë]CÖà\8w¹K¶TùCÍ“ŒõGÖI£3¸tÿdWÑÛ;¶ë•Ïða¼1×{ZÙ«”l(Bš2æ÷•ö™0w£Úö—Új£¢ÐÍ\.¥8éOi`aµI"(Å/à0a±Š0D€è-ɯŒ{'´¡/©ÁûÑʰœDДû”™ÿ¨ ö!£°ù‘Çv±e¿‹cªE¾¹]»~”0Î>ßР×Yd¬ÉÁ§ÆÃ£Á®Ë@3I×ÔiZ$ÙfÈ;2 ›ÒºÉð 6q«éŒ,"‰ƒ¨•û8.ŠôYTØÐ&>W¢“|62gîÑšœË&tdüPÀ‹M¤ Õö{2?JR2›&e(%´a6ô%×Ô7\&ã›;/|­+bœþËé¡üŒ7I—ì‡0¾¿×/‹3”ÆUŸñ,”Ñåz#Ž—$lÕ`T¾åÐÌ.ŸU9øñâ7“KKcF¿8Ð{%×\¶ÆébôÓƒ’AºiHùá~vÕYneÃ"QK'›ù[ÅàÃÕs&#½°·@¨ä³wÆgeÇÔR9š©Q©8i,(b2˜¡‘ãŸ8Ì$êç6£´žÖ·…?N«q¥$Îâ™ã–ÌÅm‡Øý€¢Þ´›(N ›9Ôf©š)¢È‰‘ùbnÑhãW&D/:I3¤úû×—oâÝ\I_èëáý­É¥Öbë'{%Ž»_³GKñ¥v.g“uÇŠZ%|3HR4§às æ!zx;ßzž+×R™ŠTtn ç/gŒë³ûµIP£Îcîbœ¾¯t×'³öË@VÒŽúS1º%à|ÓòR±ÌðÃC-!â×ìåð 9d³O*†í]ä¸÷µŸò xW¤Hªé ¬¤Wj~›bxN¦\U%Ëß=„º#üX‡›]Ðn*¶O»Í:®›ôaÜÍÚŸÀ¾Jœ¾qš7™Ì·¸MŽùp]—â×Kæpe&EÌ ~â?Ž1Í÷ö½æ-RˆDŽCVF„Ÿö~ÉHŸ 'Ýè΢Ѹ‚èQÎ þž·})îâ–hòÓ¥—/'S$Ý·³ýQ"ÄeàÔi±y<Ë^øª±dn•¯’×øÁE‰¹ÝéZ´ùF"dÑ)÷߈âo[pþЂ֧ªN…¾R³[ä(¦…ñvË”–DvâPú^ºŸ“›3‘ö³sî¬F5ÖróLOzxßshßôùÍ´Ú:åØyØ*¤¢ 6]wÿ{9zBÁ–~)Ð$^í¬0¯Q% R|ï«Èÿ‰êÿš a[8ýé/²Ÿö£f« (M4]æË4*ƒêi„:t‡ýüÄÈè’ÕŸÿaGôáДLžBå·ÖõX~ZÖ›Œ$³ ¹´oåïÇë2ßãÁ¢3­ñ\ÉÔÀ¨Kx „!ë1SH² —bPLèÑ !½/5蘮ÖQYõËö‹gED% ÷î¡ÑI{Ü,³*›DŽÈoUj:a¿˜§ëLœ}ì\‡À/æé…´šîV gµáq`u“Ë ÜöŒÐÃqxsŧ%Ói†J*Åø]K¦~9 ‰ï¼R>©Ú=1ñve„åìv`çšxCE.V[ÔáØÀµóájgwze å j•-ù¨Òˆ1ž¹_3]›K>8LS­à¦ý¤ª›úµŸ;@!/£| ï.lW`ã«PëÎÅ«›¥”ôã£W‰?÷‡NÊZèHµ¢ZUŽ zœ°CÎÊE¨à¤°¼|^‰èUÅ:©"øošxõ÷˜zñ ‹S•9AÇ¢ÒY(oÜß^"§Qò¡ÖÖ.檼«¬£Y„Ï0¨ˆ~¿)“°Võ‡.rà³ÁÛ4óÇÁ£Ófí+º`¶Ñ[˜eéíÁo¾@Ö}ƒÛ`è\+3ÜÒVÉÛçiŸ)œ3Ùâ »R@Lpª„.VXcu§¹•c»üsôèáP`) u,ìuLl8´†0o÷ WâEïv4»P`ÈÍG8¿ýp£ÑOŠ·E0ŠLç© ˜¤l¸² £ó%&;Yè* ™þ°ÄéA+ùì=Bͤ&¹´!ßO(]23§ s'£ÄøDÏt0&Ú€2–V7£ey%òÜKŽ>±ÿD ¤$g+§a›`Rd•ñ‡o2Þ5NªŠÉ-RIXªù6ç^!8MVî‘¿Àý©Tíª|àÕ´·%Ï=AììÀÏÝHyì¥Nê±rzÙã „Ë5Ê¶ëÆž…°p*HÂŽ:샃)PÔÞSVŒÕRwMO¤íóÜ`àÖ¶´JÄ9â™|.'7.¬"Q¯ …Îõ·ÒôP¥+HÏV;é‹ è†‡týåSåf•ªÞ®ËmÓTödûo…¥NÚÁ7Ùz2ÁçÓ{ ã& ù#Y)кn°B-å @ºënÃix¥¿ÀeQlJ”dBT.Cjˆ±›ÕøùVVÅÚ&Š5×uœYÆv;½ˆd¤ÝAúÒ÷âÕðà”Å€ÿúãÄ Ë–¤s_3ftνæ®.äE'pB…¸Øû]Ä´¡GÍvcË!ð¼ð~’žC3n‰ÖÁå–£Ž4…î"8%,wÖ_â<âçUç°ÀÓô¿Èñ~Ó±—=†ðÉ<[vâôƒÚ«ð,·Éoqy@Æu¥‡HŠOêoÏŒ)±)ÊŽPX!ûsÄ·r2ÏPlÃ/4he„•Úí†Ôõ³©=¯ø¬ù’KðÒ½s_…¼p¥=£9Õ‘ç´ªɃ¤â¢i!%›ü^ó× ØnëÞcÂ!ÏoÏ:]Œ \‡‚Ÿß ±™ç#O$Š"O`†ág\.³•€ô–AtrR7éÂþIHÃC $–ÇÜÎLq^ºÙB§¹f¥}9ŸÊT TMZGtp`ž «3ì&ôNã ½?ÖOP‡ÃÏ­½P§ ¦éNQŽRþòXA Ci{ï6Õ1Û¡ªµzaM—g雷ý.t?Ã^¾@p¡¹ëjã².Ôˆ°ll×ã¥_MÖ~iNzûªô3žœ³sÊtMwªôe‰Bô‘¬!{AŒVb3u•”´Î>Û¸…·ÄGãzZå?Šš'‚,¡ÓÌÉ(ô$^u*¸[¦tÆßNã0Sýc|¹ÍnòPæñïü 1wÎ1‘v2 hâƒß'Š9Kl!‹nU4i JÒ>ÁMo Èéí%ÿÇ)N-añ²•eÉZ†}Ÿ’€Ï8·r‡¿c>EÝEXî;ÕÜošKŒ°W†fåŠ Ï‘¥z‰ö·;’¯“53±V- —};™‰ÊêøJáblëø½ðZž7?JpŸ‹Ím™]¬`âÛµtg.åZà›iÛY@¸ÛÑ+Ë­xñÓ6<²,ècÿ £à"ÎÇ,ôI‰Fî?o¶E‹Ãë<×À ¸>°lKwGè:«Q´WÌýâ7íÒ×´ÃbüP¸y¹Ò/Ï’!']ö¨”ÐåT ‘ó/1A»uÔg-£ù~‡LNiÉ "ÏnfÃ`X8s ’+~@7¿uÜyjÄ=í€_ô•aN6!Vðå´—v<'£ ErGè©äõùÛ€ÃßE?¸?ò7ÂvNÇ3yH…ž· cÐW7J³Ûãz{û¦Iâ§îÌúi‹¶í‘‘½+–H kÉy De¥|†¸Éo– >N(õ¦á,c…OnƒÀ¥üÖ{U]6ïpÞȦ\ñH4ÀUAy½è-­»Hcºt~ Ä#±Ú §˜«78Mý=»cÌ-ƒïÍœó¨ÞžpîRíÎ@©Ô®2U½ÇuGg~ç¬xݳ›Ðüˆ ð¸âÖê¬ZÀ͸Àšzê&Ú6Ðú+¬ŒÍ‰Âo’Ø¡U@öüL¤Y˜ìÕĪmŸ[:­röÞ£ÌåJÒ}cæ¯!H§šl1ÁO¥\Ùr;ÚO-›½Uu€·«ëWpÂ;_ž¬Ôüd¿ šì™¾Å¬¬WšL&¶ØêŒH*:ÙÃ6Á/ÊœWõ0ÝY—FRƒõÙ,´qrEÒ[Á?‘AE9ËPÏŠ*Û9wU†[ôìѤn‹B—ïï”(Z`ÌçYv:#CyÌ_ÁlÈ1bAOSzþŽ®uÐ)é¤çN†OF«l”Opg¿înÞtñìÝ ©{÷.âp qžw÷F•6r©»{¸cÑ[3…ïO¨WÂÖ¶yAECûëÂx?>Ä.%<ÙdíÃ8ÂO|U]¸rt~CÝõ²T[:Í|à»ÃAÈ®í|pRÃꔚð… öQ ëÕ’Z\/–@öfîð¹¸a±Ñ¿áô±Ì…_혢°³d8²Ð*Ÿ?·[L/Ü´“m5B–—߯šä‰­)æz€"ƒ¼-1G$¼t[éb›íºìN¹[»ø›~)Zrº”$7Õš?p¡h¥!=9õ}”¡,/ ¶ú™è‰æ·Qþ8;ð©n—Þ¤dªF>K§Hf0m‹TÍ“Ck$YÆâ}@ìÓå*Bô‹;I"'´â ((ŠÄ6EºÐݬ¥#„Ëì„ÁYÙžt è]aá‡ßE%ÆÕ›Ò†ù•‹<@Ì%oA›ÚÁ0¯hðÑUû€ìYÿ+°(Ú{ÕÝ€Ôâ{×R "ެ-š=‰›²q×¼™?”Ž88M½¨Ž!Âs$¡ÿìù$¡±ySâœzfAý P ´‚¹×3Ô,6˜H!™[Ñ[C1±˜ÊQ‚(Sи‘ú€ ÿy¹IPöÌ.ÙûÂ^÷ù¸ªI¶åe?) ô„8áðü‰Í€7$]ç$éìç¨Îðd t"Œ×¥X„FoRX𫜴»z*-µ"m'Y«ùhOPÍ€Ëý±¹êZ3Ag¡ÑñvŒQ¹Å5©É°1´»ÄÕz„ºh¶¯v¹’¥Òë?´JçéŽj+÷š.ý²”w-œ¹Š)?¯èw*±ðŽà^ Å@¿É§ç@ß]hˆÔ¿ª«ï›yv“>Q>mö¦BßЛð¥àR‚U•å8œI9è@§W 0ëÆ0ÒxÞM²Á.0ý~nj(¢°Šå`Þ|™yâìðMyšéz¯ æ½ëyÀƨ†vI Zé‘e¸{¬"Y7Ö5 Ôüþy.K`§Y2@t§‘Ë7Dsj2ikÑ_œ-ž²ÂwövåKŠHZ²E&öñV®Ýù{£;“Z½Ÿ·sÎÆó¶å[±à­˜/UÕ¿SA³';¿B¾¬ÜÌ€‘5a ×ÎnÙôÎ)Œß§ê¬©!l ÐänŸÆ×—yÍ;´pIUò«‘ƒ¥¸ÉÖÿ!²Œ¶únH.*o¯™6þÑ‚5ýt¿¦(™¶Ð&Õ£þ[q¡aªÛE”@KÄ?Ï+@ÒÌ6]?ú€« 0'[.°úÿ·ÖJÔ.ë–T} ÖCÞÀ‰XX¡W¯ XÄ6Õd’6œJÁA¾sž‹ÛÀî¼jìÆE˜Ûç“ÒU£Ï‡ŸxØ© â¼h”¨òŠ’1ûÃF"Í «h…IÔƒ¬+Î÷?8_7ARCümC5Ãçì”y4+3À÷’ßä­NùšNXÍ!ù ú"mœÖnl*Ø•äŽæuì=³Ã 2 ¥†1…߯Ýȫ薑-„ôþ.^©×öRý`[¤qØŸª ±±UuµdžÚOÃu;ñ0…ô@+­.þ$ÂqYù!i5ºù ÈJ&r¶%éˆ6³úò‡¼QÃ+/x·nxEÇp8€M†¶@ßU‰×Eƒ(¤<ŠQÎgVt'€H]c$¼o³3Øú?žê^*!‘q´ÏžvÉ⽑Xl?¾TH LŠx ?fg6-áMµø¡ ’Ž.]W˜†àXº« Á~üËܳ/£&"D³TÐF¦}c²Äé&y‰]œºånÚ÷‘ƯÕ9û÷×5jg@$]v§#loôY<´og¥Õ‚’]«2añAN–œnÇËî/SèV–(ïÿ@TEÑÞ­+àË1hÞ7© 3’g_Ü!4~$Ø^OÞ ƒx “x‡]g倠Aˆ/é-” Öé(*¬uIûÃGŇ&ou™$ŶìH†­<î,¿ïC®>Š¥øIµÝÐô¹…xl«{¦Š˜l)bQN~¹®ÅA$ªé[ã©p9§G1]±˜. R$"†ÍÓ´´Ü¤…?!! '? |†ouZYFê@þJ¨W0ºqÇøý¶ÁD¨¿9gn¨§êì‘̾ñÚppÆbxöv^\T]Žœ¨æìHÌУ×64_E§Y ÿ‚Ùg2‚wc:›êð9ÕÓV!• Ñ“¯ù?å8ë·um"¾aБÚ‚Ù4Ýš*Uííæžœ ÃÑý-ºVp© ®œê®åšL1Šü§üŽ Pã–7Bn˜òr3Ñç|3mÔ£ù‹AZTcÐÀªv $+‘ˆ_ш²,‡Ç›N­2(ÈD6¯Ëœâ*ö’ÿ8üOr—tW¯K ?ñª¥ïžö”ʼœÔÛc Œ¼çÀ^íÌìÈu{Lbž݆÷'jñüs‘„ª¡“Dò}×*ÍÒáéÑ$ ?ˆÜr@¤H¹‚B9ë7$Áðb4µ¿r‡¶ ³TÀe°Ü«æP‘’ À‡*arÇ`ˆÆ_±<¼êK2rj¶ìdH|‘{÷Ö:C˜ ùz]q‘ýøø˜Íÿ¦ÿãþ¦´i‘Ì‹‘8s×W,OžtgéCßகL‡“oÝ£ô+"Þ¡. m'¸A¥"¬Ž.æK,—ØCÁ8ß@|£ÐœË«D>”é®Ö¶ŒŽž;á­áñ+Ö©?ÙSUÙ¯è5¡TÇ$©Ñ-ƒGùìGÍR!õÕé%ÿ–H=}ݨ?†EÌ#讀„\SÏ—R×À>:+«Juб/=V7“c"¦>Xý­­Sx ·àLu5ïíÍÊÓu‹ÿ“c,eÓð;.ÁþYб֧­é3°¹åI"oVãif­› rá§;DÁ¼î«—¼zÏvaºç`íúÄ–¶'·½üs쎮‘J¿‘ÅŒ¤SSa»hjNë#Â'AL ÒúÏøKx.øg_Ù¬ h‰/Ý;õ쑆ÆÑ|løx„?ì¸x¶ï¡Íj °µÎø¤cxû›IQ2>¥þo'hç¯{k]zCE–r”޹Z?¢ž¨æ`¡|Ú9†_ä¾ i'fj¯oŒšöKÒ‡í Çcvªµ'+>ÿKûèók\À¼Êç£õɸålÝö*ñCQ<6é<`™©æò¦¹Ò#v„?œm+ÎEF6@øJ²òÜÓ%²¡æ^øñ®ûBŒÈ7<¾fÖ´¨ îî&6¶­›c"?´ <¶ßбùÜtõh"#ÁRäŒDøÀÕƒÅ"pðS´iC ½B2I›C¬´LÚkþ#ïîßëʾM¬®( 7ð€Ýã9 ·¨e1§+"¢K²Çœk&ü-‰ž&³(]š£´ðˆUÏÚ”@^èª4lDÈ÷a ’3þøÉL^“êþÕ›ûSä1o-ëgl(ž³¡ñˆ ÞI!Uü¹ŒáVòµFØ»×ݪ‹Éo¿˜À‡5¥JÛ{& t²*Z¤É¢gœ®Ü%¦Ó=h‚î¥gÉQë½#€/ÿôït] }>n'(.;I¸4UiÓ²5 ×&VvbÜvBhÑõtù[Œaqƒdƒ[ÑmuXUšh[³Ð­Zü2ãü4¿F?qKÇ¥íÚâò /Ä9®O¨ªaà*ôýýÅ ÿU’sTkHBŸ*åÕ1PÑLË'1u€;ÿõêØ&;íÔâuG °ˆƒ_x¾õ;¯w_œ2†æÃŠ/Ú¯wå½ ¹f£k„ÐgëNÂìø 5< Ó/àÿ—–'-tƒ²Ç†É´_@X¦ã©ûÂÞàÅâ:øQT1õT¹þWå3~%â`…•ñ”ÿ £ÀÐO%éÝ÷œS@e|ú%š"œì8½÷59Õ *—JbÉX š‹7FãNÄ÷"~ … À‰éý5&o‹"§yÝ|ùÞR`Bê=[o«^¼~ßHÀí²-ìÙ,!éšFJ´¢ùZÃlÕM›^çµÕSyô¤…›L`âSZ›@låJ3þŽnަçÍ=QÿöHŽ9ÀØÓ-ãC0ˆ±`‰Oã ë3 :ö.m…ãCƒ[Ó{¸C9뫉ŽÂ{TlÏáh×Äk6ÿŠ>H¬:5Çjaókê‹W‘Íú¤ÞheÚ–ù lÖ}ÑMÎóÕ{¾Áî`Ïz®QO5sûÂàœ*^÷³sÜãÊÆØ‡5' lÀ´“»•ĈλŠ>¦[jÕõ®,[Ĥ‚v«a0ñE̸*úmú}F>’ýzÕBK9‡¹¸{ìˆ;›š,4ûÐNÝM‚‡ŽC›Í)Ÿle´Q:*鋲ø‰UÑ&ì¶5¥×®Ð=Ô‚Ujnç)’)Têy¤o”ã¿%Ô™]í2Cÿq­W8ñx¾èʹ>„Ó3±Ç!™7Uò š.›ô¹òÍ ùÍ&Üé¦ËǘÈ:ò[˜ #Ûòá÷»oH’÷R¼›÷a<`!q1™øS¦:=jœ0;tiå2¼/ÕCl¬Þr¸Ÿ¤þ <‘FXš]nŽ: è~ÙGð ·¬ê‡Ù©ˆòÑ”½âú–6.ÍsRíZû]ÜÇ0—¿W˜ª†¤ËÜ}^îBÍÞ§ç²ü~¸*P™$Ö9ï×–o‘äÌ>³Ç{ðd± D­±]!j~i£‡^0¯¿Õ¢'':ØàÙ‰ ûÆ4Ð5} õ#Õ£/„nQ ´œ³9Ñ_⟨½w~VÍTÏ僦i!b_è þ0aØÜ®Â’Ë·wù™çüùNÕæ#meG~)~Ypœ!b.+[FJÔÎ8œzšLœö¼xÍø#nIübÊ“R23ÀüÉJ+Wu¡ª<ãxLíߢ'’\Š´“)0ð©•´=‡û¾”Aœ¸—-ïiWLzRá§ÇFË º—z{Èà"Ñ>úJ½J™¿“BÛß†Û®Îø®Ã¬k¾%¡?8/“€TBã1 "šdJÛ³õ€ˆñ9 qè·4é̾ ·°­MçR@CÀQ¡/“«ŸÐ¥žä™Þ÷74‹Ù<‹Íd£žóÊuaGRŸŒ<Â!8Ÿ#|D‡s¼(;òifoè‚ßè2y*0L²bRÐM&Å\è,Ï[™”®FÉ!šïoü¾,S龟M®wEÃÓž¤>7ŠÅàáÇY¡¸ˆz,tã@ªrt¸ð¦kk7ÈÉëÿ%¼Ð–NxzÄw¦QF'à¯TþbîL¹¥ ±K f1O€Í^7NoªýQ7&J«à$øÜJ¹i\—ALvŠdUYq—j¬d·%9ë®}V}õê?=ÃÛoâÿÀDûÖìa¨ŒÁƒÆ…½þnoj+ç¾”'fdgóŒga²Ÿ÷LÆ*ª÷ŽAÊì"È\ÃFŒ˜§+{àÏŒp[Ï® ÔtqÌAZˆgÔÌÌvˆæ¬ó°9 NY¾ñÂæ^ZW²~±"é·Õe{ øà„­3²2øÐ°,cåV¨ T Ç„+)ÐÞ:ÌÎ-UáUèf¯©™å2­-$®¼ÓEtÛÃçéŸY\P×a'2nÎ(ˆ_©Þ tÀ›݈&<˜U€C%ã‹8£ïn7ÀõúN“Â_‘ \¸/°óý¾„A¤y… 0ÿèŠÑ\L%±×·vó9.䇉êñXòZ—¢~’¯JØã#ߦ› ¨ôpwÀEŠÍùÊøV2| jÿHezT§¤ï-ë+pj‡´ ½{¡Ã*šgƒ^*`l×…X›Œaõ c„®£ZGg¨T˜ðó_+j'ËžÙ,'骈»'˜~cÍÒhÉ•²¨3´·ð} ’XcÈàÔÊÁ£bF}°²KÍ逃;è®o+õC‰5¶#Aç·P¶YCW§â.xÿU[rw á¡Üzÿt¾>GJùæym£-VÏyAît z÷ûìf«êäÈJç2HÝ8â;/¿q\P￵«£5¥{Æ™·7 80µ‚ãY³Lø‰ßÃÒc»Ï0—²GÛï¥Ñ¯y¼¢7ÜÆÇErÙm±·KqÞ2·¿j–KøC—X˜œy—Sæy§¦‡¯—fø…®¡ˆ;`d‰¶z„”•bÿµ©t32¾±ió-7 ]ÂnëiC°ëŽ:° ]Ç*ca'¯¹-x¿¨¶¾‰Ûyk³®ò~—~¢Õ°9c΄0¼æ}¼ðdŠ?¿êfÎ "æ¥+ DLË1~%,_@ýÑýy…¾b²a`¤å쑉gÞQ„wû¢lmÂ;á“ÂHG†ÜÔA(R|úÚþ mG`:¿Úû‹jc‡uõÉ&¾'»öꞬ<ç;nl®jT6S•ŒÒŒº:0̾TrБò‡‘úCÓC޶ywàâ=9º0|«Úcžù…$¶N©/MOƒßùþÏø\ˆ?ÄŠÛMkˆ¢†bA_÷¸&7ןˆ‹sÔÄÓ©ª¤2-ixqx‚Åø$›ŠôulºÒ¥} ‡SHîH)@Eë¿q®QþüLrH¯Ó“hûÀÈ&Üâ6zxž€Ð‚K³¿ã:.<Ç>¶¶à vž5Ï.ºØLOPÀ<’§aŸ]5³»üø„Ô¬ K1¾D­3´ìuN×DäÔ=¯)ìN,žlÝÅ"†œyE M¤?Ü~rÕ'U%r"ò¦è¼F§z`2ë`ãòC¢Ž&zSÍeX*}È1òX0µ¤Øï ˜”Ðf‚õ&bã¥6PBΓ#N>°sô\ªû5iQδ«{³¡,fkÛý+8’ætÿƒý\ä‚ùÝUmm΄ |¡![#C  KZ$ËÚÌZk· ;&EŠõ€'ûƘ¡ šÜØùn>ëòÈ| H>™j^[ÝúSÊÆˆ5éü;Ÿå;î:j¦•ä™ml0±hž!ä¦R=,·yhÌÌÍTC J#]×­w¿zÞÚÆLG«”oíÎ=‰?Á3åcòçRÍÌl–€œ6'@†ëèÝÇí²¿R-á`6xˆ Ô:î]¸ñ„r§©Õ§§S?r^ãë½M¬»É[TBüö—ô³=—ÞYÜpœAÆXÖÊ+,ÃTc5Ã)ÕˆÅßôo®†[ˆ”Þ¦“˜ÑSBYä»ýâ(Ö`> šr”tê•ö²8¥e^}·•Ê,zƒ—óº)ÌÝÿœÒóyaÞ¡—ªX)Ê¥·Ð@H„.Möm·Âë¾XípI,ÌnYpëôzì½ ±Sý]r¨;\Ä5ð¹z#f@zÜÀØåŽÇÄüU_pèLä˨xÖ-½7svW8!÷¡[ _ý×i–z³ çÁ±€…JAÓº…1j•WéCpýÃÅ8¯Öafx;ò* ï.eëØ›(¾ÌXù€F乆mYö¢d¢¿4g óÛ3¹ê¥ÊÚîÚOÍÈ÷Ð lmS*œºTÏÚtÓíÜÙ‚Ɇ'xûžCa?fÐ Ó¯¹wˆR%Öe0¾©ë¶ò´¼Ý,I-cí’´’£!ƒ–­[´ø¤€`i¥eÆ{i-} çOx7?oœË-:½í’Å©Çÿ¦i¼E}UfV½QÎãº7=2i~c‚°¡—4™3P~·$ϧ[¸Óáîa×EÖNÚʘ&h?|Á ý£àûêÒëF œÊ5q¤rÛp õ(ÍÃ4NÓ­ŽUùÁ¼µ£×?&a™4‚ãY–°óMô±öÊ{Rƒ]â;ɺ0öU5lyõ¡ÝƒÄCî)\HÝbu!ŒyÔ²g;Ò¶¹bCuš¿xrXr¥Epë¯ôv/jM£×H$HLc–Û·DºEêS@–z-ùо‘bξ˜áloz™:ˆ=šZUDŽî¶!ôIéȉU­ {fÈA1aÇÊ4¢&íçÿPßHv%_ÊÀ§¼'†UP×Ð]Ùò_y¼ˆÕ/Ù”)1Qœø[ºl‘v =\£o’_S¬ ­Âç§QúÓh†ZÑž¶Ès‰ÊPôû²~¤!%LϰÆÚ²¼A—Zso©0ã£'Ëù‚ Pii´VþíÞsÅŽÒñ:Må7dÉö¢\gôK C…Ÿa²ª1ñ!áÊm>˜¬§ç!†®­ï&aüÐ^_]ýâs ¶mcAgkùSß"@à Þ'mm(`Î'vªòWÕÌ7¤ §]Ž>ªYg aÚÈ<é¦)øÝUØÍÝ›çj³,('4 l¦‚á¤ï)-û_–G}Eå°ã€k–‰: "º|ã…3ü4r)rK|1)ŒÛ°‡* ';ŠìÂlðxçP¦‚³cb› ÒZÊj·´2™ýëºö%*€ žwýb“×öï¶ÚþÌEÌ=28_vÉè†__Ž-G z‹ë–¼éÏ£ •9 Ù"@2Š›’¾úÑy/Æ ¹žz*”ñö¹ˆ´ÿË%öàdùŒ-…ØÜ]v«–ôy•aŠ (r×ïÑŒ–yù.þ$g Wá^ûc ‡ÕÖåW”éýi¢§ªÓLUEê F—ƒ@×9ðw…{Ÿñeó–ÊúýR*]DÙŒ¼Ø°Ìñ°™ _Š –ÆÍþå_á뽩É㎠+?ÞïçhVŠÎ„TìŸ÷„µ; Ì$7 ÿ“ÞøÝÈR»UÖzg~£Éæ+ãD=ØÕm¦HÊ4…9»^Àì㲿JÂÊ¢‰“l›†ñ|d0ü-‘¶šEß(î G ±O*º2,j­Óõζ7§°IÈn¸t4Ãý½ÿÔ:}v1–³ÍÝ‘Éù0¢·1¶ïNˆÌc+©—ä“;IEIŠ˽†P¸–;Ó­ÕŒÞMÚnïæ(êIh¡šf…ôWš=êl}e²v!ÔäÊL’ûÁlÙÖnü"¿ö7²ÛVïR¥AÊ#gî¨ÞÊÒ1Á€õBmn[‹~—HX iX¡7•úŽø5™ÖuìmÖždMqëÙgWêø¾nz†v†‹MÐÓà”¾ú©Ç±èw[ yŒ‚§%_.Œ‡Á^ýv‹ÿ7$ÕTœ#5"e@ªüCqdúÆgY<ælM»ì•ƒ«ð*P¶!Ãß*æ¾ù™Ó“0©t§k¦-ü˜'s%¯‹ñÊZôé'‚·PTêÃ6Ã/á[œ-®[ˆ;aÞdøhpëÓ^¼Ì»éb`gtØDâòÕÕ8<ׯÖïù]&ß%³„›1Û¦oÙþq4V8pEw±‰:·BÕË;KZgFºûûpHO´›Ú"Cñk4\¾ØO²(/w¨¤W 9ü'9JX=F"O¨†`°8/bzÖÉ I¥CuCqùÈ™yjoÙP.÷¦œé(¨–˜MEÄàýÓ L)ȹ~´ çâà|$ö'IL«­¸èùߨ¥•!‘¤ì+Çð a­Õ&|ù¸KÕŽúËmf»‚ߎ"“u3g碞!jÍöIêk²ô´|cýŠï°ù‰7©žÆ,ôÔŽ§à'ë6DŽðå¹G/“eüVYïõ%'‹hü„À~(Ó!ÄD|Z»°žÕZ­ |[ÝÄ‚vM]Ä!A­Ü7Z³Òe#€Ò(?º¿UË̷ͯu¯ÖEiáØÁl*ѹHwÍŽþÅFê1.ΜIÁo²W]À …XÕ|ôX?€ÍFÝ¡cGÕRÆâ´:Z˜¥ÈbXм£þÒÖÝݸx®¹g*6u”vä“Ñ–D·-ì'e*tµáÁ…Æ»,rˆŸ>8¦KdÑ@/Ö€ >ŽMù…”×ûM[Ë?0Î7TÛZÕöÌ|̨Ì"7èï^¨v<@öC-°ãÏé®èèÈ‚?–sBŠd  Øç…^J›;Ê8Öƒp&2ÀÔøýìkMOéký¼ö]–¼Äð]`’´Î†`¤^uºøÕÞ\IÚÞƒz}È<`¿ !m¨ÎSʤµ ˜Rý! k¡¨Û µî®¸ϵQpÔÂe±e!Q6 ·D¸£ìÑö'ÙJ…•sÞ¿és"ë!£4áIŽLøMÝ„¿þ0ža_G|¨;vöŽxEhQù]—€;c$©Þ6AT¦jÕRExÑ/2o}>["f˜?§¥+Ç9˜0B~KƒB´§ºIøÐùŒ$tÒoµºh·BZbòóU|Œ…(è©›Èwû±WôüCÿ [,ºø§óf¯þçÔJóív¹+=ýA%ÛÕˆ_¶¢`6$eÅÝÚœ—ÇAÙá'Ò#з”î15 ¿dÒ…Ÿ1¼e)`õ˜|áZ›t*†zÓWkˆrm5dz,ò¸ äa$êØÞŒVÉ4Åö7Úå*Ýâ3“Û'{³SÉ*á­íïp7s|±þ‡]šÃU æ`$âõHÙ&:el7k,Õ¹ï¯^`à |`_÷/‘ÌœÐ2Ø`uF€  ï ö¾åúÃFÀ|ÒEÝìùЮ”û#ŽAµº|/•{¡§I л!Á§t„ƒ^–çœa-|âÜüØ—!ÛßG—í3J¹æ•+åµ ,¸×Hjë«úWW¬|ëü4–šÎµ2̳Bl¯Ó£/wñŠv5%!(_>\CwQ ò‡çÖÓ£ƒ$°ŒYd©]­ûñdD°¿ ¡4€iH½Zç÷çÈ}ã¨ã´Ëö=°2í¯ÜF؈Î[³5‡àüìwû¹ð¼ºç±¢‹Kµ†ù7¦™ÄárßoÃ^éü©v<%-@ó¸°‚™¼|*usCYÚ Œ«T5ça[P’ïtÔ>Ÿ pò®}S$f«±jSü+šK-ywǧ s©}¯ ÇB´¿P6uRTÒEa#ß™:JøÔŠ£†b/+GÊäUuý×5¦º¹Ò—À7WñsVlP³7ÈÒ¾nü€)Ê» X.õÐ2 YÄÖ%b™ÈRÛÎOo„йc÷˜YÁÄ¢”¸“¡²ÔÍªÑæ6?È ÉX¡úö£äƒGuåÀÙtœ·ii½Æ #Ù=sé6~jÛšgî^Éœºtš“¾0½/Ç@z5Ë@/Ïê(Xn:Ñ_b®=þ™ºžËjãzrµUi¸1Œ3|vúÍÆåm0ÙZpóÕ!òŽüÜÀ¾¿tû‚’ ÿBª§ùŸw’ÀF±=sÂ9ûýèܲ²9>χÓ%;àÅ—Ï>Q¥ÇbÌ‹ù_¾}R¹e‰PDå Fc+ÈßÒGª@ÔÊ9æÖp¦š€Šë´¥Í{|î´Ênåìî_måŠIÌ®ª·8¦T" ûIÚV‡–îþ•`görÀ;U Éß°t`ÔÆeÔ0ÞU`ºgô·Î Í… F6Aj}OïU…•s]¾ß¢²ÄÊ_b–äP¬•V*׿Rîxæí6Zý»G2‡7Ñwã"¢Úÿ{;¶Ú)y^‰¢@/¾-€ý‰:ħ1‚4W¢¯ÊÊÙH²Œ¬Y˜¹ö–{YSyëKï­^£®Õ$ 7—¬Ý}²IçòÂ_¶§Q“ù`êxõQ£ô qúZû;R€™¿¡\l%ë[áú—™0Û2Ûeg‹Wî[½ZÍÿ$—kÙ½ü¡9D+g–ê§7ÙøÐG3 6ö`ÚNþìÂÔË_¨…¿Æé¹×Wµ]É“#ºÒ1Êf“ ’ßB-ú —T·'­†JxÑÛ,l¯¤²²h/]Œµ”ƒŽ·£ÝG)æ™Ôˆ±+‚1⣴.ŽÖÚ0ÿáöÍÁJœØ)9ïR¼à1ò‰LäØcvcä¤ÊwJ¨¢‰»çÏ‚ÊÜÚìøõÁÌš±‰B~®Å¥¤õÑFçéôW0át[¸œè5”ÙÛ쟔WžL‡{d{Rr$â-_&š_4é@“ÍñÑÙà1ß¡!Pììâ`F‡®ù+Ë'ÏÁ`€þ% `p’0iR99’ú:á3ðrCœfçb*¼û²ùò©‘¿Þã …ŽX³ ¡áòbê± Ð#ìN Mc{1a7ÊÎ9ÚÕB0§ùŽþÏ8sfH+®Sóu¯BJß#©*.@ÙKr¯wKÓÓ–Âc²¥rs­Dd…;JÚ`¸ª1ÇòEKõcÌÜ:ðƒƒÝĦ”²ÝçÝ-7õœpJÍø ±ë›ß{nܯ´'6}²„ÄvÂ6 H]µê.WT{¹]|6^¶ºRfZÂÆß_­pdÇ€&uÀü?é§è+•I]!¸ù" "K¼°Žn¾unå<óà Ä hP¬R?ýPz"èºtËRÇ;F?ôäÖò÷AÅŒƒ‘Šñ\- Z D|çh!Ž5K?€´ûzYE¿ÙBįš¿euªðaÈ*L4'ACÖ®•¿`*Æ\ ঔ›a†ˆñ•Ð@ÌŸ°l§¸O=µ¢yqkó ãbŠ!Áð*à?ð╟£šiU¢9­Ñðãq¾M»è"À°)•×Þõ„/¤ ™ZB®º4&ì´Éúxø³€I/&v¬mGšÆd¦¿¬¾"”ñobÓ–!(ôg·Î^ üÉwñÀÊß¼áÜ'|Š-îd l  ûm;±àU–E¹OŽoo§¹ÔÑøW*çÜj¿XÄYÛö(Ù–7ƒžû2{ÀŸ ;°^Åk3“ik’¶÷¨U®t‡¹LÚÔE/Ã<ÐãöW,÷VL. l ·ÎÓ’«_ÌV®Ñ‡š ò4g«\Ûµ( ÛtS´HÄ_Ðö c ý…‚~€†ÛÇc{flÌÓ¶Í.#ql¸Þ%OJ:‘;çýŠ­Ëüõ4ÒKÐÙB€éNÞuÂIð?ÄËà™¥|PÁÂZ€V×>xüóçò×½&¨DœµhxoänÞ@”½O6!ÏØ´Ù'à:²L”3¹Lœ›#öD^{G”(ÉH– {ΠÅpû*ðâÕbÝ ë¸ñR•݉ûÁ¢÷ñHÛÚÆÿ3q|xðÿ»%|!Ý2?DVî:þ .¶ŽÌiXŽ<¶ÔüúÕäe=2©GÄzVéuyħ¨bï‹+5B;gW¨D¤Î…á „ry±±«Ä2ÍF×1²9¸9c¹a¿¨€ýŒä·ê–§iY×ó:kÄ‹À\¡Eé\Úë³)Ó­ºTintrayô€}_4ŠEÝDn¨þÃ'›@rGÛЊÀ­$¾žå3ëL†“IÒH•å™ç‘íjô[Žº&ÚmùíF"/"n¯1yÏ·ÄàN9i“ùˆ’ÆøÄZ긾Ÿ½¼1ºf_[êæš“0â©­ïgÉ1Ú”-Ö«ì½ÃßãÇÔëh^³= ¾våžðvf‚G¨æŠîUé¸î æ.[4‘—0Y½Ý~Òh´”Kqá‘çÈçªÀ†£kŸqFµeA‹~Æ`Ù©´–ÝNÈp½—É #g!”*3ý€æ%aoâjZØð ÉØr<°ýa[ó>>ßr¤’ì7kŽð <–/KBÓOwºÅPq,S/I²—ý{é*?˜‡¯O¤[³¹^5è]h&¡~U¤ê!g0ÆÝ§0;)s®ÆjZ*®ê»n6wTO‰¤ ¥¢·AvCX×pq@¤Ì YO.'±+U°Bƒ}{‘ƒÈŸü,Å›Š(Î+Ëda+½ùŠ-ºåÑ3Mï”éUo/ñ¨R7–èõb@îpá´6FŸqÐêßÔTÜÿúŒ™n—ÁáŽÓ (õ.Êùj?¯±þ¿G/:ÙÒóÉÈæ½yÇìFxRÉDXCx0â?å#‡,Yc±Ã1ÂDÂfðß`ÞhyåÂ2á{ëµÊÌÀ’ Z”j(Z 3xh$Èó½Z‹üªÔß!áÛÃk ‡~ÉZó3ïÃ5ˆjx£‡¾^ži˜)¶?´i‚þýí™a*ËX¼qÀ„DG¤ûëw›9‘ö «—\úÒCCHó⩵P½-¹£À_v *0ìñ1ÑÆ‚ÂËöÃp…‚ƒgj£tRºË¦R¹`1€õ5’י澲ÐÄYU\ɈŽ@Õá²JÛúJí-U¸üy ®œCÌõÜB¬kw±"®l€å˜qãÏ#Ðû¦@øn`ávÍB+ÛÑ UfFŒq”âï#9GXhl~ä8izäK¸óAL„§çE-ƒy\³—ÓÇx©ì©—ŠA{À ïn Þ¦ïÉn]â;º5ó5µß-:ä”/ç/¿E3gÄKîJh'¿[-¦ûøN¢e›HÇÍH^–üø}½]FºjTü—ïkèãÌA;$zþé׆„£ I×àq Þ¶ØØÅÝ:öñժˮééþc£¥º”* øÍfáR™ÉXÜB9Ÿè œ#sÔÀ[/ô _ð1áÌ9s[ØÑ1—K“>p;Á<åqF9›|pÁë+TÖÅ nÿmú`S·Y"¡eŽ& A®Yìôa_šæ½:´lÑ{j#oZƒíÓ’öšìœ 3q„P’ÀúÎÃæûW®Elß-Ëz-"9ýš›•xŽKIƒêG¸?€:–`¯ö‹.³^D}PBÐY/oî'V§þ"ÖßHÃO²c=ozÈSÂð#A{Ì…y \»ØBx©H“)ÀcËeB@ôø/;•s:-:_|ê¥üº6Š1;ëƒlæwˆ8[Ã$ÆgqmãXãµæudK‘IPHòø¡W8W?ë×{;Ú2ÿÙ*ÎÓ§ µ¯YÒÅ…WuHŽ„£Ô‰Å#ø>Éu6ÑB4×›BLii6þ⤠[Yåq%Ë[b úÚùÔg®Q³³«‚„ü,qA_¢Ñˆ7ý5‰D‡=q MŽþVw@ê‚i“J¶Nð°;NI3TT-ÚéšÂ}µ¿ ›Ñ" ’ÐØôñ5‚¼â½±œÑHýKgß¹¢‘`ó9"?de¦kÒˆƒm[+g¸VÔÆÁcÇh™ÑF…aþÝ¢±H«å¤ã´IJ˜«nõc »vÜòf°£ò19ë!z~*`­Š99T;ß~;/d©êGS†< TþäY¾""z V!iå$h4ç{žOAÖÅ%šðžô Ód%vWg5ñók9xöô®eÖ?çG'jxzx”È !*kuÈS"óøDƒ“$àE¢nKÜúaX‚ئ¥™L 5.È_½@qº§U4 ¥‘A–¸6@Ô¯…1+¥N&Òð·èÁ‡Ö>ÿcùêjÎîá=~pà~‘uAÿQÐDƒœ°!Oq"yÄÁ˜uY!ª¥ #Éæ®÷äÁ¤p÷_>Û·ù!D\DKÏÐB ÑG/ ^ OE<÷õõs°ÐãþL¶™¥a°Ív*èlCÌDSO…c³´Ñ1ˆû8†¤Ÿ&ªØy”-wEæô!¨Ï±å£Å!m@x¨fÙÒóeÂÃXÓº'ï8ñ¿Kjvvdů•“àÙöÎØÖ¢/é°;VÒRŠì¤¶¿üÏ¡†gvŠ!=úŒÉge`fÖjæ$JO’%í+R“'æ–Ù0&W=¿˜Œ^Ì6¥/ kþ´ÓŒ#޶yÊF£.Çv—„|J¶ÁæŠK¥H10<ð)À^÷š6c6U»ÖŸ¥åXJu1\$Éî8˜·bN©ª ¥О/eID ÖR€©Ü”þe-§TÉû$ʽ€ BÑKd´z¿ ;¬!t‰5g0}Mù9Ü#€Ã¶×ây d‹xÇfo¿uWN•Åë­9„RQ&ãý =¶(Ø»Ä)uCæ QÖÙé¤7ö¼õÔõZÑùãP‰e—ùB̆·.u>1¡?ا{‡ÁÂbÄ–ß”ÆÔŒ?æùøä.-QÉÐcé‘~⬚–VzîðøÂÕê=¬®@`KºÎI®”Oc¯n3-ITZH=½Ø€ÓIj÷8HÁgGñ¾âSœ[°ÈAnd;“˜b¡©çßp†óøŠíà&‚ õ¶jN«VÒêOiK>též´ïtµºÖ£qJʾ-ˆ±Žƒ]^ãèS9A:)ýSiÜîÀËý;:œQU¿`vý`SBïßÈ%’Ãg–Þ]WØ/¬b¢¬FÀM`”,;»s$'EÕ_^ͬ’ãHŽã8(ÿ—Lv<•ÆFa¤°KBSƒk6KÏÛfÿ­er9f^Ÿ)á"Å.;ʘ¦KHp’Öo¼zW±‘œ8(„Î>äï‡ô½ìQ˜&]0‹e1(üq¶ö ¶ÍþNäæÆÒ¯wÑè,{â>À¢Ÿ‚6fWK¬QÈè¨ù«M†5*§5nÓÇ0Fħ€›7YÚÜ'akÙŒvG[tj` ƒ°Jpƒv vn-_î¾yµËÞê| Äç=%¤ùØ$ø>vàG‚ *Ô_œ\‰Û\ͱ·ó…ÎÝê‡$H³ivd|=œâ—üÖý>F)䄼6#çê÷ºµQŒ œ4_*¨˜ÇŒ¹pþ¢®ókê»­Ÿƒ*MÇgb$G4G¨¡eLÃÔz¢A†ÑÁ…LÉœâ€<Œ7ä=¸ÿz-²ƒÌšÎ»i5Æ|ú†›—žpðd^[“ åýN·¥b Ã42Ó³«7ôÛEwn+ýŬ>Kl¥ÄéZó÷| Ùõ¹>ÁvY?ª³Ç åÀ‚5­PBëTá”û/·Ö1ù‰†¤/…ˆ@Á|iîõI1Déô‡OA Æs¨²oè?èbïIê;n6ëk5¬«¦\Š Zžp¬ßSn M’þ/`54V+Tà$J¬æ‹”åTN-€)S®„nCèãÊwåÊòJ<þn=ÃyO!Jøuè=Övþ³ˆÃ ˜õ®û'Ó÷í0 äËQG"3³o.åj¿\Ë|³³ùÖAù^{©nNÏé#d\ëBf¿îÚôÐŽ>õVRÀõ+Ú±z“ì$·›ˆ-{æ³;׈°é1êϽÁE A«Ç ø1Ëf†2Üûû)ˆ ÃR}æ|ø,[b{"áÀx(öÎseÕîå„ÇÃïuSH>¯šõÍíVìª,8–k Ì“I£¸Æ¸ßvèß¡®º!êJ­ ¨„Òè/Am›ÕBn©”LºÎÿ´tÊê$ 4î'Ð`_r]èmì\‹î^¯?´¡â¦àÆ.4Z÷Å›¶sÙñ|œÏÛFßqƒÖôËú‰u°®apØâî4ž?ç08ÇÔ6QýtœÈå@D«úkíjìŠY„ãjÏfÌ·k÷q©ÉS™øp{Fpj´ý¾µé”wF^‡†n;Îh˜~º.Nbãïbö\Y0ž£ ÒI†—æV­çŽ ç@HŽßýÅ’¾¤Íä$ºÒƒê§ÑúH9˜Ša!ùšÚ§8×ïÝý„‹ô_Ñ«ý/×(0i §†hYälJܥȔBÅZÊËm-ÊÀC†ƒ²÷î žSø‘½ Ô™õ&ï‘ÆŽ:MÕ sˆ8- ÝœV‚á ˆäßb¦u|×ßéœéáÁ.#–{™Ž¿59U¸Ã3ÛÞýãà—Ž§8,T}¿qGg^ñä`à ØÝoû‘{<¬u©IÝ®RHÀéeHM‘u„)±ÿ€y~¤ ¹jW77Q*=Eš'•Œ:h1rXŒ¬¦‡Q/¿§8â½Mþî¿ØæÈKéœ3‚–¼¨b0L€Ì‰‹FóÉY-ÿL6µps³ n*™iŠŸW­‡²íÏ„«Iˆé4·‰[´jHŒ£ßl¢ôšWwGê$=ïOutg— c‡ÛãqDsÿvêv™g ©ù9ÄÍÅÙÜ9ƒNß³’AMÎTHÞã)ÇÙܺÖ~F63Ëßàè7ÏùÛ2ýÖuï2•Ê:»1ë3ðÃæz·ƒ¡¡náÎdõ€ßßñã«ß(ËÜÎ_’ ·>ÿ"d[]²“êo[@æiÏE¾2ö@K– Ì·LÈnwWÅ©¥-»¨øP2^žSÝcpB¼ ‹ox»4À‹ÈÁwk ðbÇ^!,à:¹¬ k:9ÉÖÄïWìuXœü¹†°2¸Ï/’§‚SÒ¿¹$5Gß;ë»!>ïÜñP:¤hyâw‰ [¾UϹmñ g7Ò/+bÛëuõ‘k̓ûNtÔ™óîÍi> Œˆ‚ ØéšõqÜ5p³É÷? ‹¥ kɈ˜%aìIüÆ…°‰@&øfZ´lñƒ9fñ-är@Œxaš|JØh޵[—oÈh×LŠ”†ÿe)OB½Èqè±”øu—ªšK…™×F¸‘ibI¯¤út’†2~µ1X€‚ón'ï|k¹ûóºÝñ=PyàÙRÛè-•#g¬¡³ kZ¢ë<|r¶Ì»â ^L!Ü~Œ’”t ½øç„˜§w oœÔôæ\1Á‚ò¶rP%ß3(ØÐDË£Ÿ¼xàª&™Èæ‘ÒÄÞ!ê¢eý£Ï¾ø0|“¥o‘²Jmg²_…ŽÞÑ‹¢TFJB’¢èT…iè»@O‘]fúÇ6p¡I;¶ySw{®ºW"R`˜Ñ3`lßJ²ûù4Ä ÜÊ™¸ÝIï| —Œh`ÔßHg mfZ¿32Qz“~Y Xþ²6x—,;} ÇÁ‚Ž9Ž],Ž)P}l¯[Ï­¾*ýX]cæ#/U<ñYÕNîîeÆw;^iÈìç¬ÑéÒbÎÔ+’€“Ø«)A¦ -¯è8”ûŠ:©¸7bûó-ž°_@;Hr?¸Ov'è—ÁÛ~ú Ÿ†ªõâìXŸŽ™Ûóbõ†\õÝÙ&ì3Ÿ-n‘|—i‡Ø“§]¶%£­ÜN¼t€º½Þ4÷ÐŒ=j·tLˆôÆÞþOÒ4ÎlvIàŠ¹ª¡ÿ (±“ʾ&ÖÎñ`¸:¶‰Îu¾ÐGÏ;PìWMQÇ?í²‰ùÒZ4½ËUh î÷ÜÌ–präó’'þ´‰ß¤ssE¤b÷Èélßÿãeêü2ETÏR®uY¬A±ò®½)ø•<LüB3'wˆ#ãC,ާM¨‘°ÈÔ0±ûÿ/¯$ƒNCx ¯áFŸ5n0ú5ÿUSÉÀw5”ˆ•°ãeãwhºë¤‚v7¤½«DOËE8 omn>t±gH8©/r°ç=îð¶Ó8w{†gºn¯a y¥å‘ðë¤Ø°”bÓüªÃ{+Üý¥§“\à:V¾°€Ÿ“( h‚lbt­R_2¥±¬1-£1áÆXE'úeœ+1O‰*)±œ¹p°¦¯1]l«æˆôÀ%®¢æÏyÕÔk&Q"3#ÿ‘8ÙQK öemŒ*7IÍšՎº–}Ž^x]/ g®y\»Í9I¤¾" •›IÓA@˧Û%ÓÍõ_–"°Í_œjXWÖgU½#r°­UÁ1ïa ­Ý1Æïç›>GÚgõho—œ`2W8us‰åи`vܺÖpþÖT+êüâÍÊ­ÕAæOI¢E¨ŽtêÜêù&4`™¾•ÿï}àx\ú ÔúIÔ8s² šµ±ÂµJv|Ùo”<»”Wô"ïõ4Ýz§º>¢*jºEµ|ç«2G7¸ª{DbÐëy÷ÆßýìÊ*¸eÂ~ Pé¹R— œ18„_ÅçÇ‚jEóÉéñqˆj¢Wl´ö'Ï,~n¸Š”‘^]lC˜ÅЀ|¹J­­Ö,•[qPKm­e™^#³LêÖõauúg$™™°ÃÜ z›T H)8‘›ÜÖÈ«"„¹KØoíöÿûuÂ_ÂsJj<ü ë#çeeVs€%ÃÔ@½{Œ4lˆ&)ŽÌùÒ.úYP¼"ãý 1BY‚×uaÅj!ôpáñ ËÇbÃá.‚Ÿ«öRÊ}±©ÃÅ›ç1Á·z¬à¡§´IGEw¤>ït"ýþb]7=Ë<ù4)1F’ä_›¾bâ»ðg l(רƒ^Ñ9«d ›Õ!5Ü¿—YÑuõÁºîÕí‘íäÖ\.ö6p9oFÒqÃŽ“DçCàœ_®bÔÔbEŽ–¸í=ã]ZHÁԈȣ¢ §"Ô5Br" tnT©‚^P¡ ødèQ”«$Žh`ÔEŒ`¯ÝXdŒ÷µS“yH…Ä2‡!àÙÀÉw‰DlÀÒ)ÆGp•>å%›ÝK’ú¸–½¤7õYS[Ž!RÉᩦ&„Ÿ ‚NB5|13Yj\'Ñ$ûVð*ÂQÇè a‰KMð!ïá‹’I‰=“ØwkØú0¨:63Ì\{?8•¯•ݽ¦øïbŒïRŒý‰®`Ñ•½à6á²g· !Oº[e©èËY2•ø0ÿ¿Ì{ÅY­Ã¦*ü°¨Äã•Ï0ûbxén¤4Û ÒdUµ€¤Qi“H…1Í÷.‰™l`Mµ”"84 É´ö&tløäÀ÷VÑžàu]"¾Ê®™Z!„€ztÈzÊ©®_ÆÇæ­O X®¸ß`w¹À}q½Ê³tÈAhåÜ©Á$ÛÔîx.¯–Üê{«Ã( c‚à,»úÁÌ+mx9I' °Åç&š?Ò ½ªD‚ƒ»æñ*+:¶¬ÌŒëp\zû5²UZ÷ãeÍ&Œ(¸Á©“£ƒq¢àô¡SÅÊ>ùu•}€Ã EÆTM«™VÅ]ƒ$4›¼<ü{úî·‚8(cgÕQ¹¸¨ð˜ÎÌuo.w÷}0ã|}]bðOe*ˆW±‰’9d§y^™, “«!ÌMêë¿ÇLÏåœ*œ™Ä©4’{¼œ¡éœ-Ǭ¨ËʆߡtMu˜qŽ`}k< ]S^åÕÖ¥3ü¿Ä |ÐK’À0øÎ`T<ȲyªDÀŠõ~*’ÁÚ6"G|åéÐ~ ôûDyÄ002©§h«PÃNß)Œ{÷V?9R¨4iò‡WG¶WŠÎ×lã˜&èßêKNþj1Çž6”ĤåïB˜Íj1.Mɉ"¥,t<)ãŒiŽÿ*!o‘èMBqW©MX) ôR9™jȸaSëXò)Fúr%ï ¨ø¨T/b€#Ô^S‚èD‡Ï¸{&Œ×ÀÞܻؓJÞ¦á}õ°3Á€AO©}°cZÜ¿±²§Ô䌗J¯HBÔØ ¤×øéŽ‚ ».+õ(¨rñ`â"Û¯žôάku'tÕÜ©Á^Azû¢§Iq×8ueꢵ5XhG â’ÿWÞ]-A+ÿÃyÌk õÁj¾ÊWÿo>'4©à õîÚ`ŽóM&,­·]ƒ«µ XÚiÈT·ba°7–{c òV„áï´q˜,gMÅ#1Ž/Tœ±ÿëÌÀÎd’íÛæðˆ¤‘È)6¦býðذÃõëÜ *5ŽUæÇÊŸMŸ ¢:¬®,(Þhˆ\ëÕK€ ÉqÀù6JeGŠðH+ƒûR+`v}m‰Ü:aÃÏžšæ6û§Û&ò{¾F¬7ÿ#jýT½w¤|[”W ’w*Lâ€9ÉÝ®Ay›¿æóãÔ&ڨͳÌ9¡†9_ü¶_;ûƒXâ‘Ä¿{"wÃkAaWt¤“!Ž"_=å¿Òʦ+мæ :ýæÿçÞ+k¬,É»\å;>ur´Üž't}%C ~3#½…qÜÚ‹ùžòV‘3¤1ËîrÎÓ4ŸÁúÍ$ØêDI#ø´]àsy@sô Yv†ö(÷™ ‰¹ÿŒíxžWi'βðè§×iÕ#×"`'þ=MÐ jÊÈ%sXÝ_‹˜÷–eî{°lýÜ]¯vXª7uÑMx⣰Î×}z§$‰0Zzª¬nGôX§Q<¤`÷ýóó‹»Ç&}ÓÎÚU4®7åF¥ƒÛŠuTt”bä(Ù›“>|wðä‚ ©ïq®ë¢ sõÓçP ¡#o¸np•U¾/’qp×2ŠCy‡w*å°ô`}Vr-ÃýØ ‘ ?^¾ÿ¹KcÜ™ûÃðDÛ­Èÿb4Ã’Z*€m qµNˆ~XF{zÜ^…˜–´kûåÕ›ˆäW¦˜¶Jo96ŸÕm>gÚ_ÝñûÀ’o õuw$©ú¸mQtKã+å{r«5VÅæ#Ax-JhR—œ PZ 7D›,)ÑTÂGºÙŽÇJ½,Y ž˜ ½GYlz¡öõš8Ôûê9ýŽò¬ MFð໤-‘p02ܱ-£D¶9Ú¹ÜRAÃK2¾ÌMC¼ÿÁ[™éþŸ Ú„˜ý´Óy"…OeÔžª ³(!^‘7*lšÓ+U;ÂW›k‹¾“ír­ó£€pXï]އøn‰ÕýØCÍh^”€]ô6sr<ätÑl(¿ñIÉ\¸4 å´DÆJyåJ{O±¦J¡~+V⥞¬LAý°ïå “9ëÁNv\ŠÓŠHBMÈ«f —ulÊ5¤)m§Aö½5ôb#½ñ´'ËÙòŽs9F‡cÚ÷YŸ}ѲIÕüÕn|r²Ù¡¾KÅga¥#Ý _xÐ[úrrg³ ­æ¿ Œ•&`¦eú"—Ç„<Ên&<ûo ¦Z©°Yc5i63½™5§õg·O¥àú©ÆêW\Qcç‹^s¾YÄ5ì„ÊÎþW¸†@´þj(É`‹Þ™‡ Ò pªü³š¥öÏ­~‚DبÛã>ÙÂVë\'h†” Ó¦ô®ç#">þ<îß¼ì! T"ꊒ¯£óE&ûµ|xß™¿€Æ› ‰IŠ‘©£ú ­>ÀóÚ\É¢\?;«ªòl¶Ç¥øø¡AÞÝ€ß¸Ý /8@ñó®7Òt]rW$Œ’Ìê·Vr~€ü·ï¹ç[÷víC¥Àº•ÿq{”RD4² Ñedÿã/T¡™7¶Ä¡$”Þét¹q¡•ËÈV“¯YKöŽm1ÃH—„|®Ëëc"osV_ÖêŽá|õSŸü{p 4Ü aB“¡ãzšÊ‰ƒ%Åøƒþ(èïÃ5©{Ž—À×Óäkms”^æewøË¨“ÎuÚRÃXšÔ%r‘:Ž’7IFC4IßÛWû•& ˜.èFi,úG6*`v~8tc /¸;KRœB“Í+T4¼N£}Èæ4öùjßHŒ&ñܘbl0ÂJf ±ñlÃZñ#û­¿ØÓ—Òµû¦É_þ¡˜$îðÖÀÐ-‡V>l}w&mMAü¥¥ÿDZûvgX<ÿ®Þ§Yçxè[3°±}¬^Hµ±WF=Ž=Â{Œ|ç.3Ñ?ó’×Ù“8hŒxR‡½» r^‰ L(/‡h€£âpP½X) #Ü“+ Q¾¶¤=2æì‚²‰z_Sf…/ßGÆÀ(j r [œøÙ)d@»sä- ¤+,‰ž4U«¦õ&~þׇó]ç1» mo‰†™ =U6©ÒÃ\ÁN'*ØcœtñyÂ!ÑÄ«±¾K@ó¦lý?|èî³_Ñõ"·6Ø;ypg: aHwå*€š(¨lBPóÑ)Š æÛ† ?‘W=V0ËKþ»îÌ|ïÁ§üÓ=E*}áOî1ÿwÊ wŒ#äI-h|8\Pyï¿uAGÁ!ùÈ9 à<‹.¾VflkõG»ïs˜yDT²YŸ¯ >ô”Ìv.[ÓÙbæa)(™ªöìðUì["g¤lcS?™+2N$“оÛÊk@Ö½€#ø„r.®Ã¡ôF2X¼Tpnåà{;UÊaŒÞƒÑæÌø¨ ƒ vœãËg–Öšì¨p\±}U¨í°6×]~ºç ÿÊ, &JäÏ.k ˜µÎ3µõ8ŒDmS’§b Ò ÊUüŸŠé9ÒeÚ>’ù<Òs®é_EøÜßË Áº¯Ž:e\N„Ù“Ð €´ÙÀ ~Ëí=ÕmŽÿ„ã[/<-êÞ…rô3ñEÝ÷äjzÊ»¬ð¾ì¢È³Î l:NÇ5­:¡¶ ¾˜ç9¢¹¾K’Òè0. › gj\ÙH¶"埪r) CÇ1uÑñk߉šK8ÈI€ONtùòÌèRRµY ÖÁ]PE~^I¢;]-úÇtüR&ψ³'Zõ £R CZD*gDˆ ã½õÅ$þç±{s*Pã%å¦Å^_Èø}‡QÔx^€¨ê`ü$‡œ©ØìsfJ%(mÏñÜý%œòþ&±¼ÖÕôC‰4©|Œ ,j· þ¶¬"ßÓ‰pË¢GNújgo>Nl´ôVHYXçï°x {Ö%c¬á¤3™f‰Ý4 5Þp͵øôZÞï“ýØs—Œà>h„‰òhY,¦Ø¸“zâÖžöê/æ÷`Tî^z&¦JT§?NÐøõê A#»•4”/úÔ²œs–äüÎ%°ùÞ–£U¯ãµÔD!(+`I¥uÉÁù"G`=]þ£àâÆ¶|ìýÏSk¤¾t~ÖºŽ­°|6£+³)híñOÑg¦vž$שËM"Ò?½ áóq²¡‘Ås“ò÷ŸMP?‹ÜÃ;Óï"eùRg¬ùJq°•^ ÔƒQ&?þÁ§fÂrÒRÂO‘ÙŽXéáŽæjµvòµ ì„ôinu}º#dؤDÙÊMNês$­Ï•Ní[û: Óë|ê^ Ho¸ñÞIz=Ç6†Ò’¶C‹¼,v:ukâ‡z;—•AµtrFô˜sö3•­Ú¨³;J¸$!i¶ŽeýÝ&=cý ý“lÊ"”“y9Á ·¸_ñéÕõ—$|µðzȦF¡]ý2ñRϼF|¢wXÀRôµÐFˆž¿•©ìw’W)êêé’aÀPVÈÕ 0åÉ·Ú˜ÞŸÅ–Æ¥¥WT÷¡Wî4ùŒÈc”ÂD´^>þ°Ñ*[¹cä¿%ÓÂDøÓ¹¤c±KÓ”ØIŒø¯ß'EÖù'»²:eY.Ú ›Dìúk\mü²Ìc¦Ï|KíFY]0õ/ôËÚ& ÄÜqîÇîå{æ8¨ ÿ-žØ›¡¢ë‡áwJÎ\ãed/ùâ9LAz üò;p2ÁPc#öÿ@$~Ó€î˳TxwæÙ…¦èZ!±óº€ª}Îâ×£DÖpW®Š[Ö󠢉«íØîiñ ÆBxKRÁ¶O]GØR² B[´ÕD?=:{ÛÈZ3¼ólšQU;/nÆh‡ky¨¥ô°m…Ë6ŽKéé‡Î¹XÛœ€k7I{b½–ÒÎk&; ŠùŒ ‡{ÿ¥ŠÊìàâ ôKèðýøåaeä*ýÆIýÓ ¡Pvá^X ¶|®g'ÍaG¬ks„MΑ®œìjPZ¾LéŸñ¿àÐrÉÜÆ"õs‚ £ë2H\X9X -vµWA¨ù\1ps×þ<ø}Ø’ŠiŒ­ ):l/N#Ä·:ù+)Ï¥‡m_~7À´­®¦ÛC/fE°ÞcTçzOAz¢b“Šh[MAŠZ8T"?:J¥SI·=ùÄ;¾>¤àrŽþ¿ =4GÇSÏ2„(°ç”™ëŠã³En[”¡ê„\»4 ˜è]mo¨Ù‡»  MŸeú”aP”Ÿ©üýC µᢡöy*,®iPƒóR:áÉ>¢ó¨ºáƒ8®Â0|ñRºÈšÙŒ£[µPob»µßÕX™õ‚[âAZâ‚(7–ç\7øØ¹F,±¨sØ…ô #Ýa´¿CTCpÒ6./f)÷FTg‘ÉH˜ýJбyâë²–0ÎZýîŸ%ÅãК»ÉÝmôúô|`E`È/[L>ý^HÁÁc…ª4nE‘çy›)àë¨r˜I…qTÑàô:Ç#ŠDd´X:þÛñµä]{Ð.>„è™I9Ú äפ”?‰¨`Ø@LpaÆÖ/¾,èqO7°¶®Ò¸® Â<$g·«Êÿ–¬ÍTGoª@ÃD*SøYmñÚ#Dxú9I-”»ôÛ“¦ÈJ¹Ã~¾LÎKá Ä©FØÍˉ]`$ûÊ¥äú‰ú"ÖŸ†Îö„{à§æ>j§ŽQ/ –1£ƒyp@««j¨]sh›;@8mIRˆT<¤¿ý—™*h[N>”H\I÷8$Ks!2^÷]\ð·fÿ}‹ZWƒÆZB.< ÁcÉGÂNÓ°9¾ù[_ÒÜ fŸdO•²Vÿ˜kæˆ _¶þ”¦’V¹–¹´ŠKïs6ND‚(¤ä7¾‘«—cïþà©ø¼Ôh4ôE4^èNõËïsÝü«ø|x¨B…ÅÀ£:;"¹F‹Á×ÄÇŠ®¾áqÙ6)" ¯-¶‹vxsµüé2ð1›¡†XovAÁEí¢«ºý J¥ñmÇjLÚzµ'WÑXŠùw*)’ÕÚÌïoIâùà8²0b? çRd-2XŸÒ=Có:ë3Þ>ê>Ìéļäö¸ZÝA_z„Ú“ ‹Âu D^ýÄ­¨ë˜û]J‹˜FᢆCÓDêï%‡#Oá’:°fDe!œ¢~,æÁ Y·M0ðü<’4Ô=0)ÐÙÂȺŒÞ*Îhv—³y/3”ð2a‹}T½›GY%öza9½ »PÜ÷بkH n¶,ñ´¡ú„îÐò úRÞ…—Š‹´!íKþŠ`Ò¢$‡êbüƒYm»ÿð=_¾Ú†è¸ }X`5 ~ç¤Én²mC^ó>÷ Sœ±«T[¶Š0T”W#™3Ü}:äýÛT°êQûò¡ˆQ{ßy´#{ÙÔÊu+çFeHì&­âlæ‚SpÉQ_Í?Ë4M&AµÉOÀ í¶`ª[Þg.–ú^#‰ã:¯¬ZÚ`N­¥o}‘õ†c„»÷éØ>€*ç»ävç®À=lraµ¡'à7_ç¹±õ*i;€LÑܶ¡c›¸©” ñ.M{š¸œÈÙJ~V«?…ó2‰Œ4ÐHo|:@Z»yŽåZö‘¸zk¨ð±nŤ»ÎÖ$ÞÍ2ŽF˜?ߤf)–ßðK…׌.AVµ§B …w[uÀ"÷EQIoÖøнÔ“,`H˜ÄñRÑꂜ$¤_Ó_ÚÇM;úÞÈú‚†IK„½\ ªÜ‰bSS ¥ñIú¬éàK[ô“WÓÏ}K›Ójqš»LÐS~±³mð-œ7ðœ³á _yônã"ˆv¾­üo|¦ÿº?ÎKdäKXÍH,!éï¹5C£NÒëI e&®æ@ï®òL 9š ü¤SôÑè¼PwæNŸh(»)Ý#‰®iß‚öºJ`‹Ôï6çóÃ2e/–] E°ãòñ¹…Œ×¨˜ fO„¢Ô÷ëÊÝ]ë¹ ÇÕØ Jé7¿Æß÷—EuY·Ï¹öR&Xe]鈷ÆE'“w ª+¢ Rk0jæ\ï– õÂôDOL¸žKyF”Š«!M¸LþV|·[ó‰´ÆfyA4ðC4%ó¼Ó êY Œý^¡k:S“›)œ‹Ô­i‡«¸¥ñé¯[ þeÈ ã;¶ˆOÊd’ˆY a–¦4#uF¿]\Z„Ï ¡ÈÊÁ?ƶ5Á/òøÇÃò>ž(ƈ«E¾T£T‚í]¸yÝ*K EÓ¼ºE‰ž¢ƒ;T¾ðmÞ‡³Í ÇfPeÂ`'ArÄÅJ̳)“Aÿɓͧ"n§šçÊ$Tú"õ©¢ƒZ~Ø=¿ÿWû æX†g“ÕõÝ ýãÇœ¾?½Ýø³˜Á„‘õ“1=ñÜ¢)VÐPˆe(㢀=®"Úë%P`.Õ±%Ô:H²á3\þ‚Ã$ùÈ ÙèyÒZPRÄÂõ‡tÉ _ó¢ØP³š”õé ½Ã»'³ñÐ¥ú¹dJý êgÐÀÃd4sÚ”ˆX"Õ™›»—ÒWN¥ë*Êh&rð¡þºÚ+ÃÇ?¨›lKãá+tþ«#Ç™×€°¸mÇU)åOJïðŠ¢p1ÿ¡õ4¦m míNÇÚàù³æãMÂFεÿ#S7GÓ¢ižÂv¸ñqOû5‘N\Ó¨ZOñRhÇýò¯,, o9Âð E)/m²³T¢N°«õG|šg ‚P­”ÎÐ:bk’øG»Â•?×[@µ8¢ƒqYÏÕQR„ˆOïÇܺ9~9(MWTdsG bÌÄ FÑž!. øF˜—{$šŒýDcVÛÙI‘¦Œ þÞÞ)ÐÌÜv¸Ç¨Ûû÷4ÓÀÝa˜rûåpÅ9Ì$Ù˜5xjgÈÐ ·ª!ò¿=_‡ÙqF©K!ª4n;Eøs»ÞzÛî‰$~˜*fX»’m·û¨ÅšÈIÒ­ áH.´jeÉ­xoDŠF&|G†¦u”B, 2ZÁ›øÞQá|h©/ýöŸì¯Gnœí<‰»V;¿/«7„aÞ†¹*¦ÄI”WE†dšÀªò©:@¢Þ)„«‰Ö%z£fèdƒ=»éíDu7|LýÓ¦ÑcʆÂÛšvšÓþ®eËb[öÒ ¬‚[ÞëØ}›ýQþÊÊܾ¿RÒ8žŽ%>H>DÕZ+¥>apÔyîÁP¿[õÄV6žÕ Þ¿pZ]Z‘êX¼ªã"jL­ž™$I‰€$è1áÞÙ&í±?øÁ5F„‡‚ꢳFrÓ«Ÿ)«]A¢eVâÒí%ƒ¡\PÈ6™ÄDù¥ Á1¡K‹žAÉïwðÔB²Àú´Wï\Rçº&~ÕÚ£—•Q0k;n •©Gâ–¼5ÏàŽÄ'ãש.ͬbG>pËFæ­Bqsp.{w Yhš C" H=(¹-ø¯ij kÍ@Æ´7äþ€²—FÐh[įyfçæìñ©Lc˜\6ÍÍß0ˆkKŸ?­cknŠe¶õg$HÌâ¼v%À«ï.t'ª *™ŽRƒŒÖ®‡A|<6-xouá,Á—c×0hïðm!érkåcTJÚÙ!jøšÄoe£QÞ­vÑï×å/_÷(FS–ª/u9Ä3Ä?ºÈ[¨Ë ?„>4u¿·ítggN‹ ¾ãä û´Þ úÿÇ×|º_~¼\0Ìš oÃwÙ^ǯd5Ne“óõt¨öZë×$vÙѳ™éÆÊ«²{º[êZÚgÑÚ®c­E9°I33Pò‰ó45cŒd±ízÍg-3ENÅû¡è?@Æ.%¾âŽûpp©Eðw-[¶éV!.]øzdj‡’r9h·,÷AÉEˆž„ZÀÖò39D~Y¥zkH[œT€ £pÙÓ(²µ«|˜ts§_ò«Ð›õA|¿›Cpë Ù\pt«-9Ék<1aߨ¸÷Û9 -ÃL¶{=·ôò;–_n‚‡þ^ÕÏë!m4?˪i#$Ç W¸õÑ÷ø©%x^ð¸óóÌë¡l$‡¡Ô>IòH};í„ê]b^[GûÞ£ÆÉä ƒÝj†­è}0íšš¿Ÿ9ñοú™Èf›àE·_ â•È5k6šqtz˜¶àx ûñq„çîoÉm¡¿\[3»Ï+9WËct7Ü¿ËäI)9ÓÉû½·jÏÕOùO¡.ÆFø”²~u‘–¶/‡¾?V<ê±³œÓ1~³€‘í¤%h<ÄSïÝÊ”,ÆÊ#ù¥Nk?™,ú¯-%{8Z;®X_$3\”œIÞ¶&¡"õiï—”#{ræ3_ŽešÇìÜ©¶ÁÖ™:ê M>–ÆîŠþ6É€+t¬Î‰ò\Ö_æ-–'|Ó’ 3H‚·]¹ø ÛrgYÈ´ÃêÇèí¹Ú<„ƒ˜>9­1ÎVÿÞ®Ú×@Lœ‰†iijºF ÖLÌ—“€ûžÇÑÁè/€H/­Eºÿš‚.w†ZK•ê8©Øžxˆw÷}—CŒ©ÁÊØ SÀ#IJ/’övÜm{òŠJ‘ïï•TBú‰ðTj½¸ñŸv'ªrTÀ;W<\euWÌη“ë/•l‚ŸIñÉÕR?úÔž½PL$?(ÊùrcÒϲFjµ½fñg¼ v-ã’óæ8êMKª`š’7¼ŸŽêsy2Ï¿D\6B MPÇíž×‚ ýÿ¢¹ÿ–ßn½á £UdÄ~ôÿœëAkÛ´ƒ —Ø%-Qw nS‡ÍcÀè ò•áľð?cÛ&W½–©]¯½èl,Þ†‘–6½´Àß5ŒŠ<¸QFžRùÓlÁËCJ]+3ïn*Z‹¬CŽSnÐ}O€ :¯)ÉEŒ»Ó¢}ż›¦ùó#~Y$þ¼beö=ä€é”^Eï¿4¶Ë¬ÇMU±Äœw;O´,ª•Áž“MçIåòIv’K‡Ë¨$àmdû§û(›×iƒŠh¹¼­IçÆM­=þÃíAàHF©cáWJ¬©Ò¨ñõÜòç gÃNfYlòCÛAÊù\o Qã' V9;X8#Ì&BhL‰Î„Ü»¯Ïf7h É%ù>ZW¹ÖÜFW ~õíÇ'²kÌó„gÔŒ_t€Êù€Ï›y÷ÖBÿíþ*e"¡¦­åаQ4¬(á“4ÎÆŠÂ+ù¢ ¿Á:“ÈA^†W6ûªDým¡ŒâП¨~»oþ!S«Áð2žßuDœÐà@u§^Ëö„eÏx¸=mÚÛ"'E­˜Ê-&mª3mÄùªš]Ù$žÿ-uþÿ‡ç‹Uç@kÆÑìŽ~Ùi®ÜDCsíbYf“ösÛö¡—ÍTxjÈ ÷±Ñã‹©sÁGOHhLo yyK#© ض8"àöêãGòÿœpÙÝÌÁ+«Ñ1ôkøTÙ°Û<@.ê+–µØvÆŸ¡hV ï¬ÜkýÐ ÀË%WEŠã›”¶ësvÒ%Ñíu¹²ZÀË ¹ú\"-HåKÓµ3ˆãð2iÀ5BC ƺu˜™ùÐ~cHh£pþ"¹Ó2ÜešG,0Fâý6…*ØŒ.N|õ0Ã?Á¯·¶‰Ól}9d_ë+w–Æ6hu D¡B¬ Çn¯­”0rIî–ªo·÷ºô£Ìã‹ØcÈKQbñиdu¤Ä [`þÀ›ÅA¹°n°Ó®#Ê&Äk\\,,«3~é´š¿UIºJ€òŽÎíóáŠbK;Ïy @½ùÉâÓ“Ç]ʇ‡íŸ¼Ž»öÈ2‹?þ•þ(b,óy ˆBÊ㓤ù} iôuSi«&2àÞm©®ZÌ·šù'ï§q ôO½áŸ¼Â oô[ŽZµ‡õ–êßlÕÂWï)}dW$k¬gÞ/*éG·‡¹îveV¼¦Ž,g5»Fí³ØôvÅ#G´Ã^~o9#üF'ê5×"Æ!;„‹P†P† •JÂ.»zí"(߆žßo "ÍÀŽs†\ Á*&#[`%UcêJ4ŽeÄŠI™n6ù<[¼1ËBû§½¬tW"mæ4$°?ÒÐ*Øy’«|A½Ò›ªâÒ>2y÷êèŒyÕ•šUo‡¸Žžbu– ª6“G¯ïCâd˜­ß ˜Ú­Âšøø±j9så¿® ŒËiÎŒn.ú7Ðþ–¾ùáþ[þÂÿ€ÕJùÅ‹Ë%KÿMhéÀî- Üÿ"U™ýW¯HèEZžG¨Ç;ÝXÿKkX‚]Lj:¯è3I_ü»ë^öL·dŸÄLO„ÅSô¹Iý*™\B„®L!ï!ÄO›>ÄF}ÖHöœî-„;øëBòÿ¶kñF¬ûœ¯áoå—q]±ï¢µdéwX]æÕéŠ`b*!&fc>È¢gÒѧŒ_Mº£÷^ñOÖ[*š2”Œ‹`øöÖ ¿%Ó”«hÙ ï»óžÉt÷~ÒÉ.‹K¸TíÎíX¦ !:”•-ÁWDQ àÅÓ—f™'+ŒÁѱ?ÚDë2¬=úEk! âPŒ‡J¶¹…ÏAÈ m€­µ¾LªÇ¾Ý°Ès½qøê‹UYæ…Ý ^Ô>;)ÿi‰ÉàÊóà%°­0pZFôI>Ý ZÊ»™b‰ÇÒ\À+wpàsI‰‚™£M ]›èŽé_¿¯yÞþGãØFÇ7;Vý]©¾O#×ÁõSqô*?q¹AŸíØ8}{*«Èôƒ*×Ú)tK&D°n╜OÚx÷?ævßcyoTß§ydc[È-ÝÆ‹‹)íƒýÌzŸ@=ä[$}aŠœ–ľ7Xu1 aS»ƒƒ#ÿ–•eØáÇ)¥n#û5Óªù +µ‘ħÍi¸ÍH"ëUÒº]†¶g©`€ Iç†å€½üwe°4ã.ò—q£[ÍÞšü|YÕ¨+MÇÛfãL³»<÷Š`%Þ ÝäÛ¬šqVÙ»kãĈ>ˆ±{½j½Ško áÇŸW}5ÚŽžÂýûÑó‘àðrÏõÔ öãS‚ìZ‡{VL鞀3$tO,ôÖpIê¶óo󔮤ºÀ˜ëA×M¸›°ëˆ1ws÷ä æ“Íò3šìÏp¶Ÿ³˜Dú}zÜA[aòq/êÔ}æHdÿé]ÐÆ t1¸%CÈÔ|Ž®Ù{>;}´x‹”Šä„÷ÅØMãÜz ns'SÿwaÀLRÏÙi 2O§567">…¹ÁÅÒË!‡gg§¡„[yÌ6à*®b7åñ˜¡åùÇ”‰Ô]8B¦‘jKòÒ÷$͵÷ÌF,Ú´G23èâ–ƒª´4ø:ø…°àJÐjö£cs]>iH5¶–k€o»”×ÿ EYJ56,ü9oB±X'ë‚-Ô¿© ö…¨?63jï¡æ¸@>ö3ÅP²ŸB@a9UÅÔJ°È…%®f¡ëN‚qÌóÜqtª”rTwòIÀòj'Þ¨ÆO¢ïÇßÊL˜‰òlº¤ÃÔÒüö8’i­iÀ7„y5ºCmÜF~»ò†ñ8‘]ÛXÒP û§-µ@Ápø)Šá~Ê5³™*±ùëΫ T©º5Gòíîj.–B¢ÕÒãÓ]šGâ¶“Ç66[y1á›1¨`(dèáž^í¦†%fzSñ^‰ðÅTÇ#æ²úJ&×Ì5ÿªÎˆöŠaB‚õdÉ5݃Ì) G9í 1~¦¯ßÂðػгåUé¿ØƒŸ­'·B(õlE„v˜Ç|˜Âþk¤‚qrªJ«N¤Á"™_‚ãç¥ÐühPèHOŸ[ /ÄÛέpØ|¬µ5<£¿ðtZ³m5˜´žŠú ñU’šÊcÓ–XfÙÔNëñ±¿ÌuƒûèijÐrgôËDöÆ9㌵ŋÍá¡í’DF5Œ¿^„¡¤DÅÐè¿Jüâ)Y_7~f€øÕf ¯ÇÖºªÙ¥óÙí·;ÙÓ lA,­ÍT¿[mïb(„†N‚Y¨LÖ—6!:™±Þ‰Â™pÖ×,p‹Xò¨šèÛÛMvÄe£¯•fë3öë'NçT`»W jø€#ÝÄ¢«Y±‚¡£ˆ2ãµBÌ V0m{]£8;z;lõ˵U¦éöß!¢øŸžØâF.’ÊX8,’™š´ÂXªu „!5*[—lew`?3°eÎJª{u0TÝ,/Q§ÉÞOñ)°¬ Öëm‚Ò<9ëØMöb@ÚœB.WPÛsÈÝbuöG¬”AªÍhBI™ì‚³sÛù§ŠøYü!ð†ækO\ØZJWzùÉÚdšŽzލ({ÚuŽ®?ˆ·ÆŸ•ØK]}]äÈ‚µêŠ—ªÊ!©^•µo‰Y#ŠãsÙBkøGœéwÈbĦåJ wŸâ¨X•^ü.’5sˆÑHu†Éúz:\Lñþg•áw}º”؇žäüL´órÍÈvEmþÜznëŒÿÃaí?%èÃê#<7¸ÿ‡¿[Þ]ubRED:$ºšÌï>v»è¢ê `ýæÍ¥#õ½«~hêtøƒž­Y– îŽùôɦ¡!¥¹ù¨ 3|f@N¶Ì£‹=hdL ±C¸iKÿ t 3ø êsâ4ºÍŸNìÕ©p¾Ó>G¯Ø† Æû §æþÀqp ¬DŠ'Ò^ÄÜDÎi«Mû£4ø(³”N½Ü5þîUµÎ!%@,& pCÌÄO¢qGKB[´7dÓõ®öæÐêzÄK“¸QUmŸ+ŒŒH¿ˆç$U82„WCÑÙ2Tœ¤ó¨fĉú©Ó¤Ñ'kÃäm˜LÈí^lkbÒA´´9œ¯Ø@n=èûñíÉY©7Žƒb.¦ü{üN&3 ˜"ÿ^f~myŽ"8S.`|y_ÖûçHȯ¿¢ðâwòìþ¯B.ÈÖ”°qµ-á;(ãö>§ð€‘ò´ñ”ØLiUí*µ.|’Cîj"ÊdáêŽÌðÙcõ@ÊêH, '“›nesÊH¤Ô™l>Zêä%3ñ~ùƒØÜRbϹÒ.iu#áÎÞý@ÎÞφ_þ'û™u¼þ h–ÓÃ;¶ºŒ÷mòánËT½Ë)¬EðØžþÕzŒ&,-ЫB…ŠØl–íbq*|¶ÍωéŸôÇ¥ÈN· iß»!yîzNSæÜQ3\@GX iÆ„èU.‘>»UǬ¨§ë{îóóG² ËnÅœS"˜'4ì.¥¨5—üîÓÕÐ<®ôëa8x¶’$E/ôçÝ·¦tÜ=¢~Ý­|hÈ|)cž¤¦mËW ";è6]&SlÔ–™NCCŸ`¥·Ò¿ AdqU¿DÊŸgö»rOm™,—MG¯«Z–5“ŠYö`Æ÷PÍߢÆM’¾ˆ±¯„¢d`âC¥bð‹Q&láÆPÓ•§Ã$'7eÆúßå¶0ÚA\Ø[•T}Â`Ÿ«h±Ò›#q`"ÏÜs”~BÌ©¶×[ÌVtŠÂÚÐЇöÐ:=Rv“ yäC±@‹QjÊé–GÐÝÀ&ݯóYÑüå¬|t—a™Òsƒ‰Þýß7úµè$© h“¶?}Ûü-ö¯AZî!–® ;˜y¾Ÿbò_¥åÖÜž'ñ+šƒëu=ÇÐçAÔÛ•µ-i¢Â0.e 1àzaú: ’ |é¿áJ§³7,&…¶aÙs,q#žåo­Ú GÆõËÏýM2O”ù^tÂ1¬³Hƒ®¸::ID¥¾4Ú` •v·;—·’oÿÉݽ'âIF†&º…ß2Á–>­ÕÅÈafÃiÃŽ˜"a0Y·(k%`7+FÞp0 LdjýY¨6³¼ôòúXNh·FA<ùHs¾'Ž™èÇãÁƒžÉÜ/tšIÌ•ë#ñŸîåILݾœNq(C*ªØ‹](l“ÔÌþ0‡ÌDÓËÖ¤þ8Qý=™TrzpBýÎøÛ¿=´L{›^3Åw›¥=xÜuMú©žòçËtð´X»­Jo¼÷ñŠx žg+<³ˆ%[N¡–;®~L´PîÏö½/ì¸D§ßë³»°¬É¢Wrk¬@{zý´´Öõ ªvðû&JOôy‘#­ã¢|\Q487]—3µ{~V+È×ÙQ ¬nnɬ ‘Oç²Q·«0j5äýµ,ò<žÒZ6Ù¨²#iiSòRÃø¯éÄËã*9˜§>•÷% ©i™¯¼¾Gìmñ„Ô/þ}.a–Åìàf¿šnÞ¯¿ÕŒ:T¿¹ç”'8à!Z \ìÿkJ7yååh”qUÇpŠúÈ2 kÐm8» S¡´ç1ã,޵Éx¡7O)—Ff”y|hÂ'yU޳FG,¥¬"U£ÖÛ»ÁZöPÛŸ¡÷7¥l¶¬ò¼4¸?¸ØÚJuÒ3ÌÁŸS¯2ŽSߢœ¶Í¡Jçzm?Ï7=l‡ur6M!&÷XדñÜù®2­zG¯Rü·eA¾"Æÿ“GÄјÉÇ$ÃNca‹ÍG((c.Ó…Þ–àÐäλž}Ê{~™o¶ë•yÏå–Mæ›yÅɶbª½?mߎ,4‰¸G¹i£K}ãnˆGxN÷¢1Íó¬ò\?|—uC%oñ7s^•½¿8ÓXëtðh·‡Ëžoð¢g´³²Ç–ëJÕäÊ<ŠP6î>1^tT…ðîPⳤ”Ù“©²swކ°á[ÐÁ³Òˆ>ªõ$4ìÂÑPKùÒ+ÛòèZ¯Úþ×A›o¦á¶pç ®¥þO}þ¼ÒÚdĉÚ{:€Ú7W/8çémôÍÉŽ®$ÜHlõ,pÂ'#!¾Q›ÿÞ¨I­Ë«ÖøNÒKKþ•‹°/ µ>…uìï¼Lõ»”¹)m‚ù²Æ°ŸT[–Ã1Œ„ÿ|„ùê¤ ÿœ†_x-ð ºòBqCÑ e¬üzüê7Tšzk¹1Àa« ŸVµRÒÁ 5s©•..¶nC€çØ4?ˆw辈(†½Uu‘à¥ýÑš”Ñ(ÏÇÑ CŒöu÷Y»€ ˜ [Ö Ó1]ø3MÚð‡m”- 9‰WÖ6CtXUZþnQÈTêå ¾›£$ëBÂ!G±5$K/‚fÄÈëë”Û'é20¿•$ÏvAô0ÄöÈvÞRj±ÒËHxßàjÕGñ|ë>A1™ÄƒÍ’¸óúÙz+¸&Å9€ ‹ñ»X_»`(ø@H/ćAn³6ŸêðF÷0CÔ<«MÀ–Ot£S ¬(°¯ž=!opI£Š‡KhÆN²ê-2Ç¡¨@ÐTµniòöŸ¹H3¾S,Èëg…9·ü;qËäëȉž »¶ýÔ¡ô o0·>±ÁNÔȹC¥ó—ô’•ç=¨Ó Ë<—´4ÿäJ»ó<Æfvê¡… wÕ^ ·ïhÈzœ8­Žü—™ 驨6ãjX|2°Y6v0ð²þÝÛ™MAµÈq9¢?‚ƒZuÏ·& BÏŠHw{û\äDUA¯PSïò1‚ŽÇ¦À¹^|ôXãù*;\‡ªC‚¦ƒ§}:¢ž‹eΆ{€.òß\ÌVºÍK{N°…–è3C‚¼*Mè¼vP¦{ø!±ÐnÊ-‚pÍÏÛößÓ„¥ÆP¿q=粑‰b•Ò Mšar©c-È[•@S¡Q‡ ьıl ñHÈC\eCŸã¨”1ÏÈŒ‹.7ÚS‡nZàî&ÔýÜ*z·‚ÔM7P©õʰ^”‘`¬“(£ÝTñÙØŒdk56%÷ ÝBk‘0¯Aº¶ó°»G8­BöhÓ|IQ |‡a;·]mqyÑ–«T¨ÁâÖð€ì‡r„Kÿ…7>ªTíW±,j§§Á©ûÒ¸óÄzHD‚o‡½„…L&NF®ÿ*ÑÅQ‚M9Ì[‹¤"ó뛋zÿ|Ùu¦“ˆ‰DÜ@¯#sG ªÒ­"l‚\øgb^‚Q SNeÂÏvLF[oÌ­5§ä%êƒä©M HÒ‡‘¿‹•…2jÙ÷ rÄé-aøÜ&ÞF@ùÞW»9ÉxS|Ý\Ö'çùûù{_ìP»Ò…Ån®}V(b9;êÒ³ª ü•'õíˆ@Š/û#Zð‡Qh¬"ˆ{ JºksZ_§NN«°·y p’Ew4èügóÄ8Ô­´ÿ›D1¸Ûð¸$»7eÑ Íòåt™_A7êòm&ÈzTˆv0évÃY4¸JÙ¡g4iÔ fÎŽ±"™;&äO‹²rm'ä'î·îèAàbþ· ¶ øL³©ƒ,ȶŠ0î¸W;eüuo,Ý :¤Z¤þç&^ Fâ-—Ëâ7òØRòÐÓ%øðE¯¹vÚï6{ÊdQyñ' ;¸„õídɱrvrêCbÄŒ>Ûêö1Õ³i9›ýy¡>^IÔQg¢y‹uUHfí´Ë+P¼üVË,b$=xXôkçyvÒÍ´bÔwüØì¡3ÄI68¬Ê>a¾ÒúD&÷£+1;Óy¬Õ?O’´·n' w²:„Ñ´Éâ; 2h0c˜ü“·¾þÿçÑ),¶3r6H¸ Ýí•9ÆAC@gÞ ­É ¸:i¢Øš h‰/z’á>¡wÆa†¯ ýW5Z[‹0 Çó‡¿ýo‰¢éÀ.âi·ÄÖ¸-¤€\ÄÚK‹]á`Ç/ǘ>‚✦#'®]iXÆg1 %’Ä™`ðóð2T/.ä’ÔN¹•óW¦ô­ò!õ"dQ{-Tù¹á.8?mù™F+9Ê׉÷ý±«ñßA~Ü4ñ³«áž ±ï™ø»m†N:N¸1[XnÏÇ09¬èú£Ò‡-¿$;K-µ•q!2L[²“V€s=(Å¢Ò»yÂfX«¤¢K—Í‘Ôí(ÁCr‚¤=㺕æã0cXokçq¥fiÿªdP†Ç¶e‡r~eïmnð¿ÑûÈHá oÂ+–€ò!~¬Î¦ R¢.®~¿Ë'V´¶Ð] £äµ„ßËFjuÿÈi§»­YHà*·;Ñ{:"ä•£§TÑ%|c73ÂßN l ç©;¾$Cñ R¹ô*ç$jô?åì%l%KñÔL÷ù²#Cg÷ÑÌ!ày„aNЗ;[&4Wíù½¡²¸æP#ÆÅ,DÚ¢A–+ì'i«RétŠЃ퀴i°¤«°õpB°“å`üXü-ŠŸº§e@éŽÛNï:K Œ&í¯S|Â3fpo¨šo“ò»P±|ŸÀþÁJðϱ†ö‹‘ÓË`×ÈåH_!¢ss«#h•MÁüÐEÍ•ˆ&šéW޹ !ËÞ¡¹à„Ò¤Hl„hªxiv¦'Àd¤››¼r_’ÏD Lþ˜vdfÌô6X™ˆÒc, !ŽÆ=™4X¨0;JlæIJnU†‡ãŒ“Y”!aÆäèŠõÃnÀEÚî6 @Çzýí‘ÈcNã[‰Eª×ÿŠØaÉÜžÇkÅRŽìòJxJ “À­Áôàú-Ãè !²&Xzw­léô2 ˜ÓgéÇU„?NÁPA'ë+  ñ€w)༇˜®•Q§2[ÓR7,3KZMô2—ãî·@ùöN?nFéQê8Öt’o7ië ßRêJ zöë+‡ë mpÚÛÅCíùmàø†Ç?_ÜcG–…Iß=‡¬ž‘ñDJ$˪¬¨½ä-z+ûÃÏš†Ò†ßÙZÆt›;~ž0žªPw ý×þ?ͶîÕÓ ëP`É$j.Ç u?ñü ëÍìß ‰:o>âg4øR½ÛŠÀûskÀÿ«P¨dT9|̶ö4„âªAàŒˆ,gÔ‘è…JœÉ2@~N ÑèXs(ÀIûºNÁ—¢†*¦iý©6à‰-Ç*#ª4/«¯ Ÿ¼F3YÉðDƒ÷8C”€ÏÔ§§ 6H&úìb54¾Ü÷m'¾+{¢aÖÍ{¾ÆÕ(¨2ªL(KýReNÑX®×^¹L—ÍÌ;®ï]Â'ª\iºvê'™Ӷ{û†½ê–ïMù€û!»ðÝsA¡@*¡3š±È”.k?ˆ"âL *)òŠÃí i†¢Ð7Ê1•äsHoò· ûæëm­6âô½yJ¯&VmúETª¦ÔŸnUdá-ƒ2¥Í%Vj¹8w-ß2aý{»ÊÒ¾léã!‡.Ï“@xG+ ?r>|U/€ÄÿDBÅq^õÜNGö'æÇA ;’J%šD|{Ç në ,ËŸ­•*§ðcA8/“ÉL²ÈðÔ¹†á“ã@>•tO·TxŽ’âs¥uR—F¯a”;HÖ’ÊÁ„«ø¿z²õyõY"8UN_bv¶l\üféÖVPO»7%˜[} Zéï×ËÍÜÎÕPðݘ6"Yû/r«Áf¿r/åxö„°•î…‘s˜Iî(ñ"Hßÿ |u¢HcŒNœ€¡qÈd{Ùƒ'WÌ~.ÜÖô¡àšáh]ŸÉø ìïÝÁFž=Bë çwÉ":ÌtyR{ coh3/~gãg^õbVÏñö3]…ñ éGzõcòÖÕ¬¸„¸_NâÒÊ¿<\û⊂g^Ý*NlÃÒä íŒØØ ŸÌ^³k ®¿f¶–_¨á/׬»äÒÊ^HµÎ*aÆ—‹^öIáê …¾ènøè/ 2Þù ^EÜùöˆÏ¸RÅ1’€‘ Ý<Š)¡j©°¡}¬è4å'b°ÙµØdÒÀñ9ëÍQùêFý×”$ÿ©‰ÞLæ¼cÒq¤Åó—blld9ü•Vý»‹y×Î0/o¿D-¼H–#pw¸óT½&P„1{|@6Y¢JV1] G&&x»3Obž ;íPOé5hH›9ìÅ%º~Ò>¿kî…@|J!Û§z½23x¶X^ø\Òíy.e‡ê!W „I€pd íÔf?­GäŒõCIp˜FÎ_2`´…†cf #ç.<æxÅA’òŽÄRåÞÍñôãÐH)ˆøpve¹µ_ÆH¶îâÃ$gJg´Õu8œ ¶ û¬½|¡p*S7<È®¦æ Ü笒½@ê»ñÒôEP$ìg¢QÒš§ŸÈ.Õºû´)‡†FÛOˆ¶6œ«üeY>ži£ç "ðNŸÃü95ÍÙ?z®Z•È>N‡lsÁ<~Ásàa{«2¸ÖWÄFô¡ôôÍï*½i=ÒVî ÔG-SæQJa…þP\ƒÔ%ŽÝxªi:í  ÄV´Ý]#à.Þ²ˆåhÉ#êÉ8°OÔJ#´¯’žÜðF*Ú|j9t"q—‰Æm¹ˆžŸ‹G r~~â {0òúJŠÙt† _Ó~uvó#1¦ž’W”±Ñè8RÇ$þÕÌÑþåeº ¡-(ï³ãfa±B0­qٻʖÄ1‹¥8w7ÆW~æùÇY=ÐZNžVÑÿ7ÜffªñCZAv‚· ‚x •jålN5ÍÑí)òeÅN±€ïÜû'&(\ ‚tÙ4™‚Ïd‚LdJGt›ëÚ61x1Þ¤‡HžŠ9)¾îAü·m<ûF³DþfØòõ2š õ¼@²5ëöQ\úïB…‚i/DÏsuÉ–Á¨µé0÷Kˆo”v„´™ &‡K„o"› †ÊyB"¨±^àðå:ŸÇŽÛ{ùçÉø€Û·ÒØtÅ'wo“vM» ®tP¸½s.ë_M a}ŒøÒ$±)•g Y÷Ál˜ì•W#vÙÌ‹›Ü{g0r¥¢²_fQ«Úb›u-·tÞÑPÒþäè±KÉ÷ æ¨çÕ¢#%ü‡Èû$p«VRÆšogqÖÙäg‘abß&ßž>R¼¥5ú¿Ãńʼnt¼dй£G¿FŸÉO63±%©:úÝUžéiIW )êï%¼ 6Gq bd²ÔË‘sy*’6 ¥´²s˜=ß×9\ê(qAS&Ò¤x/å,ÕoÈÑŒO(1æÆe„©)¿¯ß¿‘x± >Ý9•± =ª& qñúÇWRxùBh^zˆe¹\,#Ù‚á¬k8 U »š~{}¨vá&OSÅ.Ìn{k4z:yð=>^3í»(Øúß;ʹ,øŸF+LŒ6Î1¦ôøá·Òȱ,Xj€)ÿuv"ŒR–|²iXBawD/B9‰é[€›¯{´ ®È›Ié³á`ÐGZî°_lhÐÏ- Q3þÓÉ>âóë+ˆ¾Yª@_Ö-â6Ï­aÒÔÉ®Qš£—-Ðð‰9åYµ& €ÌÇýK#Žh¸ŽN»>E·¿o]÷c2 ’è€(b I€°ï·¶a©{¥3f:ð¡Bµkì$7$¦É!}”Mw>K ‡áׯD=Jf˜¼éíé½hÉUœG'íƒiðo>$ÈÜYuÍׄ'ÂkêP$dDLK5¢@{ÏÌ’"ž8®êK®6%m?RRÊx­ r©:6(ă:ØmµIÖ­A±’÷c¢([ëîj‚O·=Ê»q`ûù²è=P˜!­p®’œòNSçË7‹€4qïÛ”÷îÒœ†ÙØœ‰îd|b»‰¹˜…Ó0–Ô5rPéí›AÁË.©–ôë¿ _Gjü•ß›S1qª²Î ^À— è; =™éFOÀÚ•Û}cÓcì¥6©~f;‹Ð)ÿÁÇ5X,¸ìÔ¸iZt^»‘Îá+v´ ão Ï(ãVÏÉit´!°®å@¼?ܽÇ?úÑŵc¯¯0‚eèÜ2f¯6öá¼DŠþùPÍMc¶ûNè¤2GbVÅZ_`xk^½… ŽQáUØð¬Èð +’,¦õ4Ʊèôp¨ÂÏ'bdñí+œüîRü-ƒfÝ J1´Z<•¤?K§èÛ«„wD ” ”ÈCóé¯û ÄM¯…çñœºfY’f=S Îl3ü0UsRD‚(_IÊ¢ý¨mƒ§t ßô˜îû¸EZë¸ÿ½¼Ì±=k.pû§³"¯êïky—žŠ…4ƒ› M‰HUM`&Bô8©2ùo¼GšãÖ²_„®º¯ssŸ Ã]¶XÚí77ðºÌºø»6¥"´$¡òpKXË}PàVv"«x»Ÿ(EçNðŠØ‹;r¡‘ný Ú Ù\ì×yÛX‹ÐŽ'á§T9.G ,->ªJä–©Á㺰ѧ4`¿qÛÖ«Î"•Û{L(ÅJ¦hMjˆDs1È ®Û±Ï¤8¨¯ÔL+é¶Þ»T:o³ØIÄçšî§Ã:p\¬Ý ’zŘ:i5Ú±þ`PbG¸dG³±±<¼jÞõbì&Ú2Ó¥ãõ`‹'m¬…OÿëIgdl÷[öæ¢_ô€³’B=GÕ³‹h"W} L1ãc‚P‡>x?$—ÃuÕR€lÏ‹$T,v‚pPœkÇ5\«­åÆYû@…x°µ± ñdšvE8åKÇ•–ʼneágû÷OÌþMD®c«§x5Yè÷ê·<ï!Ñžò±6÷)»RBÅ›õ- Ž%Õ“þÎܽY3šp 6ÔÐüÆ21Ń?ÕmÏû[WÊ®([¨¶9úWwyä¾ÚÏj;~N¯ÃZ&ŒûM°Û«¹2Ÿ`¾bU¸…ê´á¤ëÔîƒ~øü»óPe¼ÿÃÇ-9O «dÏØŽ råZßâk&/H͇.^zýpÜÞûÍäÖsž£&»sÍ÷¦ø*?¼û÷hàíï˜3×E¤¶î9#FY9W<—Êl¬VÇý#§P˜h…3#ì&u1þqqýÌÛL½óTkÚ¾+ûêjY`åM¯î!Ki?|uР[бÝkÔÙ¥dj~æZ³‘U$99æƒ Gå ?ÇTSÃÇw§½»ctæáhŠ:•l°diUëÅ­ÿžGµ/’Ó-ê!2ð:Æ lüx“Çnµ@¬â©Ò‹UvÀsŠD=䨩&«3fÄb 4&€pÿ\›F éüâÞÇéòþ ó‘‚+ ¸¿ËfÜÏ¥ 0ˆŒ¡££é*­GhSk$žÎ G`“…©â}ˆhnXdßw°DsWLØÍç/Írcn;N‰,ÞœA˜õ¸=o{ô}ÆjÕI\$¯óÎúøu²†þ5Ҁ㓮›¾ÎÚzþïš´…öfæê+s}vÉ‚©¸;lm”ôëþ¡VC:Kú|5Ó‘~ÐW4Õ÷ƒJ¼<áêq…2mS| W æ Ž¤Ûk¼ÁìvÉíȘÀP}Íø£ÓKȈ££Ë÷ë·ÃÊ…¦xõÏ«J¬ØÃÈÂã­Uâ2Zo–{ñRØx03•žPטdäˆÁÄÛì­V‰¯¯«!™® ÛÊ{â8tÞfþ6ÎZ|1Ž &,*í÷‘7f—£ ùáGM‘ 1‘ªoÉg¸^r–çdñ}î2JðChzœý†æöSDÂNK{’òÕ­qqÐ>?åÑœœÅÿ!|˜qس¢×4n³®$fCA—¶Kï•PÅöf͵V4i%­v@z"ÈG¿8n“÷³ÈJr5!~¯ùêpó¸”dw ÄF§Øßô B Å=êÍ›â`¤JõoN/Oî¸Úõ›.gšvç`h81† ô¬Â¯.¤oú<îœ÷ÐÒÒ<8N®´ÕóК³’åÜz¶/ÕÙ"#L/"~äÉ2i Ä›%Ñœ^º¨²þRãúË? Z´ØC:be > tMà«ÌØiµ]- —^&¸Ï+gÑ€t&#¯‰#NªéÅukºQܨÁ±A ™®ß‘b 9a–F»ngôúV^ÂÒ þ†ùFeå Á8³S–ªIäÒpXQ?=\=Ÿ2hLiŒò=õrºÙÄ»‰Ø2Æ8Ò·®Æ_ mõõ掄y®r Ó1ä­K×Ì"?è½– ê$Y®Dèä7ÅŒƒ”Ö"¡J † ²¡ÉÆ¥G,næ8Èõ(¶Ì. õ=ó÷PèboØöÐÚ€oB!À¶T‹ø’»<=؇[¸…æeó"HV1ÈŽ‚™¤.c˜Až¨zF¢<ç |]jÔÿÅ” ñsݲPôBMÐcåZI‚ôôË3ŠJ šH#8¡ [ÒsØ›nX0[Žñèr7`^]ãˆ:Á(àÈ7Ú}Qvü'!ÔD=ÝÎfyQöÖašt4Ëûµñˆ!z½ïòK¯ÉA"îÁ‹åå¡4…$¯G÷˜[åúÍj#¯Ò)Ê™97ÙcŠñÙÒä€u&Ý™1Ë"1nn ðø?7‚ŠÑçÌáaæ¯uÁCPú«övxWí›by1w¶dh<­]EôÞåË8†kaÈÞägu­ÂÒÛÁ]®«Ü8@y`¸€À —jÂÖçáz¼6þ´úÇQGÓ7Hë8Øú« S[.x›Âú®MVŽgéíXk‰<‚‘UPÉJÑdɳÓX¨Æ‘þI#†[ÑÕìM²ÑÁT>ÐÍ™DÙŽ¾Uo†¢qEòt{«ÈCc[&¥¤ Й"³5lZålúb3‰¹9%@ËkòXHZVeCHÅ}œÛ²C^p°“µ‰_-' V>5ÎñŒ,¸Mp½œÈÜviÎcO`ïŠçž«&â`ã«Ñšo`¬Å¿’$l™ÕJ¹zìªG  Ýð¸)æÒû~vïë3kìæÿyÜ\$Õ¼Œý‚4€ít ­® ǃ,™¸ÛÉ{´µ_]ÊZáHúôœ9}ÎÇâV›Ó¨ØD4„]Ãó¼ÆK¦&h d¿-•Ic´èW g³ñÎŒ’#¾ô“©v¸‚a¥ÐÞƒ׆ÀzØ$(tÔÒ ‡Œ’ä]û pã¶¹œÑº”ø5aþògt›†èÈàÄÏñzBñ5…̱ ÍÛ®q8“Ýx›LKÈäÄS „’@;Ťá'ëømöÈ9‹)ŸZm~5štp…kñ­½“é%•Tb™ôŸšF+jŽà (oaÙ¤Ä.ü-·H0÷³~N)ޞͷ°¡ea;n .Žú`‹±à„¥^X~œs„m^³’&2nÍà=ITos×Dìø±ÓV{‚ggJ³Ñ›¢pï¡ êC‰_I §%Z¹Ï,ס&l/ °J8G¯A û dœÅ`O±¡m,úª×Ïî— šÿ!®'ºPžîsL-¨î……©*Á|A£àº ²Øn&îŒP+oæk±ÌYqþ…‡¶hçÛ袲®ŒK¾¡ÐyD3ž¹ÍUyºbŸ¼‡n,jÛI>#Š‚ÑT¬?¦=3ë¿á-à®â)/mÓ*W-˜ðùsmtørÎÿyiªd³_¼gr„:Å2o›À¥Âב™†¯÷ZMö–Öë[GÔ–¦Å0DÖü´/ÍdÈD”OØå܃Þk½ˆgù}[@L¢=g’Ço 'àQ|OÔ¦?™AǵP°éŸ]i(ñ‰¬'¸_aG ‡r¤„ä òó¹\-µUÄDÖ6Àæ/?Ožæ8ê'Ž˜Ù&£ už·r;Á“6Ðm Óbuér]ÕÖ[#¤çÁ: ç! ƼÝ1”è±Ú[»úP üà€ÒÒKËåM2Ma áa²ç1ýºÚøàjŒi\ZKg\ý~‘Ž–1¦×=‚ÍŽ"¤åâ)µB,y/êNÇÔ<‚ù9䉅:"³W&°ÿr+Ùõ9yØ*(±<A(Å R4°BÝÊÐh =BÝCó™9¬û Ñ3­,µ¨7 ²€Ï6u½Šuyß—öDöG]ÇN9„¨ØVÈ„‹Ð¢çÁãT×'±yÓõHzŠMxíG B|YºÚ 3WÔîuqÝŽmPãd²´š°†L:-žGSŒTû§—d”j¬°Ž±hăFÐ\î¾ñè˜ôJöt¾€ø'£ÝMø©@}uÓ†¨ß Òblä2L$²¦ŽÜ:¤MÑüzjDúSÊ­ÖȪOwâûù¾ô0g’_”d¸þÂE@ب‰"TIÅ>,®a±£‚DnO:?F*ÆÄjøßÛÉ 4è7ú }îv(,w"5´û¥!0\ºF×Ý}ßkÿ;*2`„ó#/þk‚HÊÚôR|€¿$:ŒytÙ V©ð¹ÝÝãëWwTÈÇÇ{Úg#Õ×Jòð°óµ,ríÀL†e®Ü!Ð ÂH ©:ÂLTèªÿU]dH‹JejÔ¯·O zý¶»ª<ºá:.†¹]Õ&Çp­ŒÆk(·Tš Ôµ5ŠJ§²Õ–/.Òúó>÷‚ÑÖ‚’dŽÊ¢Ñ`nf‹÷nò[—3õÝ s“õV4ªSqÙÚ\Oî ˜i=Tw²{c˜ß¬#—ó÷È[¼ùÎìž‚b:kµ1µµ²‚=M0¨˜ígƒ¡_3;ÌGS£¬ÙgÎl^èŰ éðm$AgHUÅĩԟ3Ë­‡¨9#áÉ4b a‚²T¾‡4×!®a0Díy¥Ú-¢‚ß²çH½T¤þ¸} º šô˜ ˜œ€ádŸn-ÇÔƒJ9¥>ŽäÃ$•Þ~Ï/‰ Ž€•…¦=·€pø³´~t2 WotÕŽ@îï¸MJ~+S>5aáÚDNu‡pQ‰x)“,|B›îŒÜ›)8¾Ét™d%g¤Q Þ=n†Ïzœœ_òÛÌ)§úŸoyÖFÕ71íµ³âù‹5 ž*¿Þx$_àq6™ Ã¯ä¯pø9hðç9f@Âÿz­š²*ºÀ>A„{…f~°£.¸?9eIÙ6’Áš†Ô‚ñJ ¤Ã¥ßÛ·±õ«¬˜¦1´Ò–ß?I9®@é¶{ABw¬Y&‚â'@z$¨Ô­¦óÊ— Å™A+ËŒ·=¯ý,eJ½9í¡R_ÌÍÊðª ×!™˜ƒðj„`7¡<ŸË„=}Z˜ëé[‹š¾ žº>Ç1Ðçôq1i Å~¬'Á ¿O_hZ©g;§¬ñ[äùÓYBEØZ¤JÐuÖé M#°€e°Á‘ñ«“ÒôcNê=® $‚R\ÛXj*ê~‘sJC*›Ný«e—Eu8l[ÊŸÜñ)€ÀQcW^gÓ7vïøß`Æ¡6*%£…òZáÙ׿Dù‰rƒ×À›\µ¼ž{§ =^൱¸°Dó8û‡Sp³%àûùS&C]À%RýéŸF‡qè³4v™'|Z°ªŠÔ¹ÆLcÔŒ…Å*uÞª¡§(U©O4å£þp€²b¥¶I=è µ×m@}ã°ˆ‘ Qé©ä\+;Œ¹! ͬÚã8a{Zè{«ùÖÎæ ´« tzèÊôüÒé_­ÝéóQO¨¡²P¿Æ¾â[O.€ïìU©Ùhm·Ú2È¿{_„,÷- ®7/Ië·(,·¼Ï 6_žŠ¨Ãèòkýk”ÖLæÿØN¿f,gÛÿsŒs¢æ¢À/ YJ®'ŒË5Däùí_{Y7fs óÅžŸ÷üf°¹~¦vÖÜ«bûÝ: pÿŠZÆ›h.?÷‡§¾&@|ü““Gz¥ò¯ÏÆ‘ÑÎ{D¯~#~q^š}UÑ(Ï€Sž’¨ºØgb—H†ç‰Vm­*õjO56éQæÓzæÂÐ߯…žš1°æh†î˜ÊFði{{g†U ^@G´Ê¯0m¸åØš¶ä Ü«\º%_Ú€”SÅÓÑï‚ÇkK7Ï­[FðޝUÑ[‚WîägVc;œ>¸Â[w嬥uÂRùÍEr ÑfLøfœMc×”Ÿæ‰ I*›"íÞÌÅÔ•¬ñ¡Å|Ž] #4r%Õ_ÔDÖd³kñøÿrxp%zG@þ|9,=>,üøÎó¢oü‰?øPòså1»þc~ãRøó%Síø²x$YžWøÎzÇz³ ­m_ Ë"§Dàá1™õ¹°v„'‡N×aÇ ñ â"ëáHwÜ!¯£%µ øÂÛtõ”Ë+LVŸb/=ÁpŽÕeùdyárâwœÿi‘´È>€ô¼?Ï* Y/FÄÙvËsú8%ÎtÚ™ûrgßYqïÖÓëŽÂŸÉ¦L£±‚;ié.g M‚*g|[Œ É>œ »6 tûâ'F6ÄP¶·8iÕøN'ôýŒLÁÆ¿‰ùñd‡5Ë®¡Éè#¨Ôª9:¶W$WC 4œ ã4²*FEîΖ‘§£ï.€Agå ýlF‘ˆ, JxTý“Ê|•âî; c‹ÉÐÅ3·¶–Ä<èxûdƒ¼0CˆÍÏpƒTk²ÿµœ"ÇQ¼6ä[½^Êýn˜­6q4‚kF¾Að~s™>ÝES ¯I4Óé§JTTëÅD+?¸MŒ‹ÁèsœgæÁj‚‚mÚÓ±®ÛéϲWþÆÍJ‚íè{¤&‹§ q„±¸Ü.C“ñ6¶ü0Ym´$/”˜”æKù@jï+ÛB} ê=‘ˆÚ`9ŸºVÔõWÙ»ó®B£ûid‰ )HG0®ˆ«·±øËbHÖ×ðàcÄKžô7¿YX7(w!Dß?¹6À£¹í½;¿M–=ý©ÖÙ¨epõ‹›ÓÿêíÁVG¦Ÿ£íuK6oƒx C«iÇ¡8bÕ %?µã¨l¡åzYÊ;4soVž6ƒ(Wqe,Øþ¸§EOÝŒóÛŽpù>yÁ ç³jÃWOf°ÉŒ`SA”ŒâPâ¥àˆsv°¤<ý¾¨)ªü%;Âx×uøF<'5Üü†c‘å]zÇÌ™!ZõÙBl–Ú;*°sE’0ç`ê¿Ýªéo­ß×4njÇ_|¯°–u2q8ÔPg‘?×gt€Wl{Ë´³šœQˆ ÄaÒ(¡ $ Qy8Y´ËÿFºA?áýr3J·˜s \Ï4…mRÙ;q´Á¦2OBj°Bà×^¡˜eŽ#÷ ñÙ× ¨“\%]Jß™9¤_g'ºJÕ%±ÓÍpíÝeî·¦ëçîÔòß—~’·óÏ‘˜À ÂSÆV ‡´ úïWaŠ9yìo^éòÒ rñ¦‰·:¹K(PÅ&o0§ÂÆw‚üdÌòÙÛ‚µ’ôËö¨¡b”P²à8Öì”MCcþ 6w{‹å½’ ¾™½~4m ãNiÉåuæ=¨Šî[¦×š¤#öŸªOµ½aCÓ’“Dj:+F7Ý6$]º5’ |ѹè#_úHþ€qÐ+‘NÙzq¨è~¾c½}^³"mÎ/×`êvÌy9È«JÌ&µ,~Ò"7mk!YßðEü†#˜§jE¥XV×ÓxNzî"'®˜Žq•æ¶`_'}˜³Q[^r¼ } z‰&04Ó]5% ¬Îä&ÀÓiX±Üÿ÷DÓIRÆH;‹«]!íÒøßûZx’¨Ž¶Ýµ~SƒËǨ31´ëTŸâÌFÒX½w÷• ªJ…¥ùGàÀ ó­c%Z¤E¬ë­ ’&?X´ÔíŽL1’M*ÿ€#ËSC¤64ªôR_.8_v)Dè|=üœÍœxJòˆÜl¹/Jä=ŸhM†š¿J`!²¤Tx¦Š@Õ’?6Ò}ž´ŒÅÆ÷¸E Oœýubú›nÂT9<ô¹”¹å–Òñ‚©SSêŠÎÆzQaÂ?yiPßÝè†?ý@µÎLxCEX à_B]on–îmÕÜœyòIù Ïª½Î³¬µû´‡}¼öÅbæ¢ ãpX½ïWû6W»Öúãš9¬kÌz€J }'Ñ(;PÖ ÝJX…Rˆ.vèjå·ÿ FM½,ÌX5‹^¸…­'Q¥¨±<ÂãšÕŠ3MºúƲÉ×hpu%­³b×’…AÅnF«Ø;ª=g:Î^9l^¿NV+êY,ƒíïbv÷ÿ¸Z™ðê‚i¯ý§a7B·Gƒåê}©daÖb=á=UrPŸ•ŠÙMn´é½šážWÎ'Ü‹ËKrˆæ[8%¿Ÿ‡ßýF”µW®.SˆÇÛ“c{Á4Õç-§ØÔbÏàÿ% NWê¯ðð¿’šˆ M%®= V?RX}…OÙg*-ÁÑÃ0Û•ƒ îÔ8÷˽J‚ßÚžŸÃ$ˆ™>qt‚/¯¶NõlÎéŒÄ#ï™Ó¹vï„vµjósè忊çRÓÖ@ý䌷÷4}@'nœ~9B²ßVâáAB€#Ï[´œæEoÛpî0ž‡½ÎŠ¡J?¼$1BÁô7·ç;¶nÔÌ&Ôÿˆ–ÒâH ªK#ÔÝñŸÁåèü)ô€[Q3\ŒS¡†/é~Ð2›p¡žCñPŠ“5kÁérÎŒZ*фؼµÅŽ¿x(“^§`` !ƒ¯Ï S®†Ä3šŒQmÑÐq,Éu _ÔÈ$I¨÷]i\/wpÓq5jŸO‘:ÿ1X2K¢täNA ˆ$Vû ²|©è2÷-wÀ7•˜U¤±çÉw³ôÖ(¬ÊÑÖÿËba"©å ”_ö˜°ÔúÏ£µ©|ôUšáFòT_IïhÙ¦ÑLþ¬ÒÀ˜¢®DÇ…»‘r”W ÞƒP÷‹4šôñÊeùW[%”ÁðÉ[tú.LÆqƪÑÚl^ÿÅF–Í;O‘‰d›¢P9›æ]ÔGEÎ XÍç:·CÿiÞ}ÔW)b{¶îdb7uÛOë}7ʬz PTÍbÜW^J®Ýˆ1âsœ¾1ÿÜ€4Ôƒ†ðÜÿé×ìeæëjJWT*Ì?³Ó?¦yf4"Óa6*’8Îm“æŒeGX1,¡rÖ}èîÞGÊà•¿^4Œ¹Ûr­yMˆN,¾ƒ[reÂA¯ByTóRN Œ:¸ Ú¯—z·AÈ;fç(Öï¸Dņk”2á ×e’ƒe‡MµŽ ŠªQF§>mvŸ›L X“x Å?JHMÚo£ôH=ôH‹ÿø!ã¸ÕÆ1Y™Ù((†ï€àÿÌoGèµ€B˜ä#–œä¦NÁv\û.Ô<ó''…J[ý–fïÿrלÚÈ´:hp‰Žˆ Ç=ЦKû˜¸®!2ÛPŸž_øã† ŒIý\«»Ž²"<çô¤³}ÇŒM>9c“pÔÑGGâñÓ…7<á-µo¾n•z7·š´nÞ䔢ã›íㆦ-×±Üí…~`0Žäì©K=Ø‘Ôrÿ"²A«2ï¥Í{²í:3hÞ '§m6P¦ãzÀÆÐrñ§¹9½é8‘óL<—@`ö"9,€‹üm ©Íç¥.áßýJV˜ƒ »2P·'ÌÔ|Eªæž_7%y^æJ…9õn&æ·ÔìÏÇÐXNC[90±1Z÷áPëW7‚40cŒ$&Lö<3‹v‡ÚÚimî×%í\V­/`0iªË=öÔžâ ̽b Âe}nÛYh g—EP" W›•ì‚L¶÷6X+X*ø½ôéf€¬Ïèj…R0;ä»GÊ™¿ÜacõÍnC’9‰“vv°Pî‹>ÞGÝ¿rðƒs0RˆHÍ)8Q^LB ÍÚ7Å›Lήaý¨8.WvL ¨Q·"u‰uø2­Ћî9° ©,ôòËlkçháA÷Ga:°,YhËÖûòau•GjÛôdŽjÉ^,AŒ¯ EA± JšXÚhéU´óÀ]QÄŽ;%˜( ù„Žb»ƒ'½ÿ¼é»3Ù  •íï½êuw¿ËH9˜ÞÈQ«};¼_†<ªÜÙÜÁa#ºµXÅô+/-Nظ÷$-'9ï|~F¦Ç©o!ã\ '3U =Ü]mXs¼<·øQÕ³u¸œ´>†ê¨~Ê̾úŽØˆ“Jƒ"¢òi7Huj6¤j%²ùøã¨– ‹ÔO95¡mñ©ó=Ðô´&Ö=‡…=1HÚ¿ü&_„̇Ïñ[Ú—x¤:bô°þ;”ر=çt…5׌Rï"hÎ=ÉÒIm¾-xÂNƒÉÌÇ„Ér S‚|…ÚÙ ^:nGÉß<ý/—ŹeE£ºZès;×õ jã¾_lóå1…hàyèxš‘apÉ!þÍÆq–væ&‘Î aßßDæ'³Ñ*•ò¿<-h•0nçÙôMòƒ—¥+óÿ$T„„X ä—«Úeþ–­XÙX¾ï¬˜ëZ9ÄMg6'”G`^¢{ÓcÃ?LvAé @ÀîÜ` \tòÇׯê {½&fÑR•_Læ iE:ù¡´fï&/V7ø/ퟡ¢'¶™ÒþÏ“¨ðÀÀ:ü*Ϙ‚YÊqaµ‚bÒ¦,ÒØµº^ç”|‘¤Cá3%HH›–›Ö ÑïC¸8ûÉ[ V¸)âÑ­ýÜßû6ª1 ö'Ød©¿[Õ®»´f¿UÒoe\ó7›Ü£rZ<ÕeËý®žmŒ³­ +ŠƒT*7‰¼ß¨_8Ù‹ x÷EMODâ&‹×Ÿ3\T\4Fóð›çºMY J‡9t;ݬ$8E=p¿ ÒHø“¤w¥ )þß6)–-qùdI©Ed¯==¡N‚3›iÝ-5OƒWÝÑ1–g%½ç®Þ)|î-'” ¸$iÃ¥ólu1ûžÅ½Gr³¾ä½UÆD‰NgkjÒ–¯†!þŒZìhâ´g'_'gm;…€XæK'ž|“šÊ²=aL˜PËL¬¦gÕ‰‚ñ¬Ø—Œ…(ìÉÛ—$!µ¯B†„ç+š|&*yªm ÍîH7ËýwвÇ”æÓòùKžC-ÀöÊt*ÏûTËãSN®S·Ïù‡³]—o{°yzœôHÞxi õ±” 2‡¬Ãßé«J¬˜+¢ äbn6W5‘. cмÉý°•¾¿TÕtô`ÜØÊËq’ªÝ+\Y2¿+H¾³HUÊÆÒ’ ¿Ítœh/±iiUäiWn³¾ ²beìig<32ž0 ŒÏ’øÍÖâø¸‹¼gÊTP¢6”¤0ÖËJĈZ½,A,JG¸|Û6Š ¤ÂÈtfÃNÀëÙPˆ ±Š™˜>˜–d@-°t=idÈ,óýÊ“Y ³kæš—}ýÖ‘L’‹öGÿ#¡òMzÕù·!µéÐ>Û‰ŠÙP#kÔ,m¹Œ0e€¡N•-Ö‚AÕ ‰Þ" O”>¿Ò¿ëãý†ôÔø"ºÐp¤°‘=lÜݾá³ËÞ&BäY%½7#a ™Æ æJíèø¼"eù¥i›g§¶BcÌ gTvc-çØ;my\µÌµÃÐHk¢[cÌAYlEÝÖšñ—ð›hVF /)‹»y«*•¾u/'’Ý—¥eO T˜‘Ú/Ùþ‰3ëdïnrógЯ˜Aн'Gk4è.Vc€ €Â I@f7h ‡Ëf¶è6«òe®m¥óº˜D´ùõY2ízµÑ]^ ‘b×8²r´¶HAP¸Ã‘u$Ú=ÈBŸ`¶½ja0ZÉè·9Žêóe.€Þ^ˆqw{/Ûžs?ËÐd ÄoÖ¹}þ™R¬ŽÌ41cBxŸ¡íehP«°Þ:!Ó7£hYú¹áìŽu£a9wðõŽü;½å “_÷)šÌ´šòÝAô^=k˜‡ G9ãŽËyÀô,O!–0…né, ð?×_cgép}€ aO×q'Û± $\A´¨y‹Rtç 6Õ6É…„%¢tOñ·]kt XÅíÔF#w·o›Èƃuÿ´ºIOúgäC'ÉÈq`Ä»1˜ÊKÞGÚHl wŽ>)½w÷³°ó¾mˆNÍWbv‰ýÖX¡-„mR¸. HNˆ ¨“r)OtdÌ­¹u»ÉŒv*´Oþ`“*½N[›ÙÍs×2ÉËh˹ab]áÊGò‚z{m‹º(d:]Ž€ÚkdÝð“àGå­C&Ù% ïnãk¤!O5¤™y•ܑۄE?T©ºy­f¼shùÄHLýáËG²?Ø_Z Fôˆàœ§ìŠª—þ †þüîà—î“Û%†T0uRÃCEH„¡c!÷a[¨±õÅ_€ãiÕs÷ƉÁÄ;Ø4„ˆ´Îò\w—à ‚žtÒÔó ¸:y'CGÝ'-Oæ}°Øx¬Coõ‘Û$!.Ù9kké› \‹4p ˆÅÚ@þè_pµÃ¦˜!wüÆê òþB®7/(*ª¤³—Ù8Û1­¾Ÿf“ÃhÎîb³q €;%Â:ò6í“*û‘Þ•’Š” ë²}ÎbG }i—H¯vÌ~T?O0RAªñê äÿ ”Úä6sª9¹:8èH(Šç†Ðð…¡Ó¥U'òYJÄz¥àÔ72ª²µj†÷<¨’ѺËÚ‹§R@wȆ‡>Ýa‰`+ók™Î¶Ù⥀!V5gÉ£îrœe –·™ó| ‡n<@¥=52{¿£~¯íÜ%2-=’$F’ADÅ^WÕ¢«ëUëV…s8cÛ'kéÑÖ’AßYš“ *õ¸Ip0‚€Øm©Ò@Òî&l0’ú°,e×dgmÐ7ùŽŸ. ~³€zŸÁ¸Ë0?àîÇ ;»Ö‹¿rS¸ ôÅïìOНÂN¸ÍüÀ2ÒØ€Ñå1ÿÅnSBGh˜>ëxãk&­riü}±?+ÒßP}ß2|~îš¿¤ 5*³úômGâ` z©0:¯Ejzã9»(¿:}0€4>R‹ym”Ïi¯ÆpB·þ‘¦þ3FÈ| yéÚ§ÎÏC@]re‚“Ÿ«q¼Ûã4ÙxŽT~…8ê8H† üâtuîä•”±P‹qû‚º Æc¶gUÕÔÆÐ)ß®ÜxDÙ þ[9V­œìLj[pBö í6Sê®m 7YNQœÓWÆ5} +.?tµ'¡*<60 cÓS̨‚iÛ?K„ ¤á ³´£“&P?Âbë¢ç½þ ˆ½1´Þ”xŒiH< ÀÄØºÿÝ<¶Ä‡h‚-cYǣ>+^k¸¿{1Ü~ëÛ?Á³Ÿl86Á2<‘æa2øI{8båAõàò |óÞi§ºX8®Åyå4s$ÀeÒS•4¥_wPƒsz$/îV%¯ÝYoˆµfƒÿ¶ÚÄ6’ÕA¢åÚôÅMq#1Ô²Ú8mľ¹e¯ÃÇú=dt¬å!.Îãx€Ž}óùD l–•U‹yò5ÂMö·Ý3CšoÜaWü#z‰ñ xìíÂ߈È"£ç»{¾?•o9d’kè>‚ª‹9ìO¦çú4óáÛbÊR\º‡UP·|Î{Aü`¦g¢ÈAR&r®Þmø¯iö¯Uˆ#{í=ŽìÊED”Gè:ZÌH?b]z"£ǹ×ÊB;¾UTZš›Wêù…±aßED¿öNLâöÈÿ‡ÖáìwÏÂA¬‘‰ÚS­®”SÀϳFÉ«®Ç$ëTGC·…Z4¡ÁØn4Ž‹­7óèQïÊ+È”æ?‚DZ¯7‡f@R7uÑ~-ÐO_C²]LFò¥»?õOò=ÁÀ·p‡…¯â) ä«Ëcä]ó–™½¹ýÓàs«ÍâZÕ_‚çL°˜9“[Èà“Ý:­$©jþ¶¹©³Ó—/»N_K´O?C[ŠiQcýP˜ÞiX3ø¸51ZͬÅ5wB¿e'ueøäJG©zDšn ½–‹d¢˜ÁdìŽ/ÇŽâ†ö àžoçm0k{¥[²¢Tñ‰ÆÌª)ÓñXÓ³ûÖq±ÖGvHôX¯‘ØØ‹/éR9_'Ö•w’³U>vb±0 ‰Ú«r% íãcXbhŽ&HÇs'»Ñ=<õ°,Ѹiàxð®Ù2l_:Ë"fJÞïDÆNU.‹Œ£Í5^Å#õNtD‹luÒUeê¶}¨˜äZÚq0£²´˜‘XS~“ÄDFáÂy"îÜ"Z¶”ãÏiz¤CÚß¿6Ë’ôZ‘Ÿ“‡ÀÛJ[¾Qx5¹‘eÍH ;1ŽÕ `!/îúm©·º3 ÓPÆ£3GüÞ"4”Ô¡ù©ß½.!3Pí—vÿiûùaµað ŠHź”Qá\¼(I ßÛ\OMbz…r³Ýɀ瞈ַö 1AT± èÈ[d{Àv1(§ôgÕïÿTVÚ°T³2žUÚ®i'9:ZÁ»ãûÂùDø h¡€AKU†A×{ÍÝNܲƒ’¶Pç9njOÂ"¾Ü.šþš£‚ºµ)-·VÔ&ÏX}˜»áÙøÞ›ü9"‹®ö’åçh~(ô….BC~DiMw°}D®ä$7 áöæJ¹Šú¾¡º›Ál'ï-_ .1 ;Ïoüؾ V$VÔ~Œ#é ׊ËA_ìg먰I}9O›{Ö,*–À9ÃJIÍTR ÔÆuËh\æ§°¾\b0 ©9e ÉR {¦ ¯Ä~ßV® ä">²¡§¡lÞ›$)’‰©@^IlqsTh&(â]Ëa|øa8ÍÄŸ’â«W'‚Í©lpZ·Öä¦T2(ÞMZ[6Ùg Æ *±Z»×fò˜°©f‘!|ŒZtbóÎΡ^½Êñ^™ÄäÀ¡.ö:¦2¡º¡[+îgONåu (žÇÀn±y5}lÛ@¢«NŸ>&Œ–BMBx¬²?Zûïå‹¥BXzMF¯ÿí±@Ëå?­˜Ôà7þÐXD÷Ñ#âaÒÕî§©ãk[{ºêãxEá½TC¹‘wŸäJÚq{Éîa4”#VZN|nNýÔi`éìáƒlñ[±F[Etçk>`1e¥!°™ùúàU‡BAF¥÷§îí¿N.ÿ ë¬o,Åy¼e¾5K¨E‘ïhïQˆûqâ9§5Ú!PüªÜ’>:¿&Qîwh˜&Y‡)~¿ªÍeF!vv2ý5 Ë&}~f@ÕÍ`PÈË'%[Áo©n=%aLH@U-þþ“覭_~Ê‚Q¿¾.»Š¸ Ç-LFZºš««›‚™…ÑBA¼é©£¾­Y ö/`6×ÙÍ@v\Æ$œÍ4e%qS;UÜE>ÄÃ,T¬ŒÁÉF>2ÑhI1Ë2Ž@Ú„/iÏ^Y&Yã´6°{4¿ ¦ÁGà묂cý•´Íxjøy-u*5:ÞÔ­in)Àœ<¢—²ô‰õÛÃhYu¸.7áM÷€¼¦{«ñŸ345^ˆÞ"H÷OØÉ¸b´ l!:ÏZ€dhGϼ*BÂ)Ñ¥–Ü$¯Q}’îÓ-û7Í?›Z«¼$„AÌáÛp÷°Gu§ûÓ³/ƦjÂ)ù«Í´óòñ…òÕmƒkX×ïŠa@ëZ* âY…Ì¿³rfê6é’UÍFmZW ÕRN“á·þo$Ùƒ<8”æü¨hòkìÿê¯1,¹ÏH^à”¶7ˆîžÀ}TÑô3¸¯øDqË/~ÎUÿ.i³rNʨ%¢U‡Y€[h­­† ¬&„>]“*Ìèh^»@hÓÃ…f³QÞSšÈš}7Dn1“Ä|ùqºãÀa¨UìÇ·<ü¾óÊóU Ãmråv,¹Lh«nn!/"Ì_8ëL6ýYs“ÈBC»ŠA0ÿŸ´KFW§äÛÀÊjPµ~ÜÃÓûÖMé „c|§d&-·˜úììP< ²š‰W.O÷Èg•±NòRüË£m)+ À›þöS¾ƒÍ´ óþx×)}pá¿0Íx–3“þaÖÌMûÈ­5é¥ÿo’ñ—t'‘nËdd¬`§_wXJ^=Òvò?l³…K‘Á2r™Ä番húêq'ÚR¥ô±…‡ƒM–a~„ÅRÚ‰ƒ(ÁêâŠa­a⇮ypu|K„´2´|5ÖeÿÙî·£@¤aܹn¸²„&‚²ßˆ^£¼ákKѽ1ÖnÇìÿ™Ô¬6«×Â’ OÛ¸ùC¯Dc„#>àx´ëËUP“èX°Âñ~ }=¦ãœïºÌZ9•~w¶W?s­øÉë.µH0)-m‚ž†¶*¶šzSd'þõÊ»UÔ!ê‡d²´¼vÙã;È-VZU‡›Ã×2M³ž{tÒ'´?ó<Ù+öÍÝr¢¢Ôà¡ ìKàøµg0©£ ¨z§´FÄ+ßö­—ѼàX…šóC:bk ùiAÔñ,õ/pßSÌy™WËH®ëîa…‹Š;ìªÂI¦ûk'¦K‹”o`Ü~YŽ–lˆ…æo0ÝÑàÔƒ·\Þåv½õéŽð\³½8‰Vï¹³Ê!ã äéч¨¹%>ÝÀítC#,#žS¬]]ïSõ­(Ó&*èª5r¼ÚÙ‘ ±‹HPÇ ܨý>oZpA=$Ý{œ@*˜ÄvNÄ_•è‰F_j|FÀÔ#7ÀØ?™_ø‰""›³}íqW·Ulð0þ”F»çÀÚãçžÁѲ@AY1E½zVB¡x£4ùz½$Kð$ƒ6 ¦€ÄU¿$2ù@ÿ¤Ð¹/HôÃæ©×|aÕ˜5ìi“z—kÜ™NÒƒΗ(91¿]ÐÛå@z®‰ 5œÆ²§wiS¿È; ¤ÇÁ<²Æ’f³,ÙÃó ”ÖþœƒóåÕÀN›gÄÊ;x3۾ָ魙î?ác=Ò[óÁd½lÙÂOs;6 Ѧ3S>ÐgvÒ4Œ•tÂK*9»fˆ,çy<¿ŒsëD (²Ì"ÍkA#‚ûr=Ö‡–1"iÞ`©eNûžãæP¯Xšbûj“²MŠw _&4fª€¹ç忤·h†ÏuÛT×Éf½l>'š韌=„Ò_¢(<¹aþxÑ‚cqÃxª’K¦Ãä þýgâP˰»wñ±¥ý®à­Þëtg[iÿ\3¾7%‡É®µ¶Wqv×O\’yë0"0k‹¹O.TÇ÷xJŠ!!š™ÇTòv¶#q§t.¬2ê~D//?Ï»dƒ_-9iÛöKpâ0k;|*¼É6M1‘·™ìY·ú1ÛH”tM”»ñÝŮη—1 Z€v|ϼEá1.^èýǯ±ú°çg9ü2äh\¸t]›% ”¨¾~Œ19mnù1ÛcR%8Þ”ÿw  „ìdQ%°A® ò ®I¦¹>Ý ({Í[“Ÿ¡d!1š”·å ¬DUî‹ö1(J€,ÜEóT¦åkQ*¦œ<½ lØ.h¾Å:Êÿzý¡;e¤hËÄW ©zyÚU}£"&µ‡`•üH¼{˜*–žÎ‹ëÖ³®V$¡M”u'•L3œgï”ÿ(Œ Ò‰ÛEÁ[‡ž¶ÿåg5µÖ|³LºO=Ù~*÷ëßÞðê#ѵúiH‹÷çÕ'ï­¬)d°ÞÁÓb˜«ÎIp°vgyr7ŸÌ'×>fM‰ÕÂw—¨òï‚ukf¬l±~tVöF³é—·øÞ¬`ýçdÛÐIoÁœF2‹ï‰_—^®AÝW®Ô7ÚT;¡ÝÒ_´Òmi¥¥ô‰wtØÔ“7r EË¿n  X|ÞOú[únœ¹¸•æ¤ ª^Ž^¡Ž”¶%ÙQ7èô½E! áNóÝßó1x8!ÓlãJ+ƒ,JõÌÚ&GÓÓñ˜õþT©Ÿ"ËKcÝÝòeª‡5à\ –ôF¼~©šd2hÅ÷씬Y5ÞáëQRòÂDŠœa Š©tg…[2[mâC¸òOi’”>å´ÂiÕäÊh6gxçÈž}-ÅÞkâ^MÀcCåÝO”ˆ£µ˜ìgvÏ{ø?7N’Ã6~/¡ßT·¯SðüiTŸ¹Ò ¢yçÉNäÙ#} ˜Qwe@g“ñR”†‚ Ä zx|ºQõ‚Cv¾ ¦‹BÀL£»y˜RÊ¿QúzJi•›bd¼õ-ÐkÐjxï–8c½•ä æ¸´KŒÌ*C›ûÝ dç)ÁqtFÍV~ù ¬S;¹R—áMö»ôø~8ÚÊåL£ëC9ÇÒ”^aˆ™WÕ;ˆ*ŽèÏCóx±°ñÔñ¢öjwkfÑ<ã$äEþ¢¾îÙ€ñâÌ"­;ˆ5cQMI$P­Dþ¢¨l„Ža+ܵHȦÛhN0njËT—È.q«/hÃÔǦ»²„z¸ý€Ä€lVö˜£[’B÷ ±2TŒ­.[³¯Á±Ÿ¦8ÒjÔÊyY®üêÎôÙ¥¸y=kŽw´ú’TZØ$3Ë-ÊÙgÙg°ØŸQ(,Éù2!œé`ÈxZž¡ÃR¼hù±Û}ÑÅ©HêRmwn^íð´Y‚oX0½˜=¹Ý÷Â×ç’ýX•žßk‘HÒùÃr$UÛuf<£ŸTç‰djËõ Fh»#¤yýü´,ONhíÀùÎ6A"_G£0:…³5¶xÐw€7•á|Dö”ÍÝâÕ⼩"g[Ô–›íœH'¾Ì¦“_6Vó ¶ÏqN!.BG@5ªÙä®õ¶¹« äbǵ.\?¸ÐÛôí³Žü½ž§§Ø×Ÿ±•állhkɉ̷Ç<|V]¾¥d¢J ¿ É´À’%£Ù@îÕ*ç·çP|ص ~ñ““§µú{S™à5<dÑ?Cœ§úßm_øHàô”5vˆ„dÔ… Åõ«P:lÈoT’pd çÐ.1_±ä§È­¬ÚX–^Þæ¾æ„0Í(\>¨á·Ü@ŸÕ%\߉5®Ä\¬ áhb¬uB]@>ŽÈИžB¦\CÍ[¶ÐZ[d‡\©²›³çî ÇI!_‚ÃŒ'PáÅѽµÆ –Í[rJiq÷C0BêƒÛïßÜ(‹•GÕÿ+N·Ö÷SDT Nv1‰ƒ±øê~}ü˜iâçÓÇ‘}&>äiF–‹¿vWI—»ûl$ÑH9ªÉO\3‚-rÝk=l¡K!% !tš ‡%7w½ÑB§B&¤—aÙùåPýÌ›³óR R×é¡Tžìp0î¸Ùà hÆJœœq£¥X ß\/rß8GD/€©=†YðSe¿¬nm=¬eZWd/ vi¤¡Óô¤‡hA7 N7t Ö5îö£4ª€‚ aüqRæ’¦ÂZ.l…â §^¼,´“Þ8'pËG ·ýƒ] æ¿¶E ¨ŸÛX?nYÍzφ­NMa6B êß™akêÚ*OTZ߆•¥:”ufìT—3QÙFO<غÈދH¦þ1úDàÛ«/c7}Híu£R¦G¬Ãb™Ôy»‚ž¿ÿØ—ûò¹{7çA4fíÏSé?Öû„J:p<è§¶Ú^œo¾Êxýjœmã$°¬$‚1úÎã -q—¤ò§­HÜÜÏ'E34>‚]ÉHH+ìÇ®‹T«åÏXÛˆ"ö¢:vâ§)2“mïž®KÚÖf5‚ ˜y¨ëßvþfíѽ½O”úÏ¿µ¿°³^‹r+0Ãlb=• ›«×BÿÄìÑÂךÍtËÊ}wÝ9×ñ9 ÈÝÞ S¬€Øzù¹Ø÷¥‚ÁåÀÐ2´lE£:úüæu£Ë—šOy o¾p—Ýhþv"‡Î1è 2ÃÌJÞ€¼cĨöËJš®îmüø`¾ÚiÚ»lâêg¸Á˜©rœ,wHÀáãÁ– µŸ¥¢ÄaÙöÿqÛþk»C2Æ>.wè³Þ§7µêŒÝÔÀÍŒ3&o¹Ü÷ޏ‘'J½^ƒ×{Îß­mÝ¥C?VìYnÆmj½öÝ0:ü)nSˆyHI»çù/€¢YÿöNºÃu¤/Ôõ»â»#¦< òý%–z6ÖCUŽý§˜<î ÐbëEdûóÒ_Ü ×î·Sx„*q]ö_d!ñkvç4I”ë[—æøøiÙ65yº Hs»Ùy ®¾JuïpDc/!»ªtqpsÌÊÝh!^„ .KœBskc‘äß_‚ˆè‰;¸Fw©=AÖCîjÛ‚bÇo¤6©Ç^ƒü9Wé úYÚî¼ñÖëà¥KÑö‡òÍäCa ¡—=rΦIÎçù›¹6k²»pêÖøo 5û§h¦zÄVèÊ1ÒüS`šˆ¦¹×øçhÕŒÔÇ (Ý<½ð ‘—;˜µã˜ƒ×oä*¯…X¯ÔꢌYá*#1wD`5¸Ûº«ã¸:ßË9qsà Оw©»„áµ’ÄeOøò¢ÙÚÈ/¶DËJÞKÖ.ü§/º;§uëÙ½ÔCO<äq)Ù=¬£/‡Ÿ`Vx`ʨƒUŽgæf”…×$Úª—&îo?²V2^„©`áŒH…Ë ÐŒ‹”×-1Ãþ  [”æš|7±úœ Tþt,Ý’ '5ÅæýµLë#)L˜ÐèiX޵Ìx´ K^¶{™*°…ÍxÜòXïò/Û\8…,µ‘‘»Öd´Ñ7h€Ù¯èÿ>1õ<ü=ª÷ŠK Š…$`¡£ßÕãÉ ÞûÐΟ K6­\¥=û7Û.¦¼†«„™òš¾Á8’I‚€ C,»}òPW‘ʦ¼¦»k ÔMëÍQúSÒ±XË,z¦P]ñ÷s/ä‰NxONÿ—ú¶\Ä"‡—Û‡ëÍ;¿úDYý»iÔ1*’tÇ’«›­»æZ[û1…a~h¶„&®x“öîÙVŒOEáWt1βÞrÇbœ÷ÀÊÅ9¡MJ1÷?/Ú2ñ@ÉsÆY€ÞÈê”øJþ7<¬j_³×~yh­ÛkFHFg™)ìш¹›ôÉLF~šP¤Y T¢þóÄ<ÄÀ5Ô¸ú›Vƒ´2'ÖuöE(IûŠ9…HÕŠÏèHá [`"ÑdÎ,éÁŽ‚W,ïä?ohéÛgãžQ®hòÚSžØÖ˜ 4Q–{[ñ¥V¨Ä<ßÄ~Mï¤ð‹"Ù ÍÞÅ*ìÑÚy5Ͼã¼ÎÕ³¶ G,™²ª˜+í/û$Ÿ*½shké8Ôá:´ªHdürÕÅò}Ck˜iÝD$ªó¤å¦‘»­F‡#fòf·=ëý=™n,á.ôqØ“ÔvÌg¼G# ÂÊ4–®‡¶´™„kC”šm‡ZNë–·s.T¸Ð|͆§¼ÿF]dø»ˆ 7ÊÀÒ‚P%_¨*/@ª‡Ðæ=œÞŽ{C”¼™Ù5¶.–è‘I{ÅÒq(ôB'–W`_/竚œ Ð÷nT$‘i¥ AH˜xkI{D¨3äë™áÛŽ òò¾–jZÙ."˦)\se"M ÑÖ’µàö|ÿ꓃VMõ]‹‡ÊSe`ßÂýV NthVÂ༳¸/Ë דêWŒi0¹ßœß…lÖNLUþކHÐK´Iä ÄÇô½ÉµÐ;áT€Žã…áOè½¼wOvšuÕ4…&3nŒqHÁÜx§ÆÔ’§BVKèÏ/vypܲRëÓ!O$®Œ]¥¢„׿òřо߸øÖ±„ÊÍFa˜ELAd›cD¸ªÇýµítßqðcÈÒð»Ù'÷[Å9ç¾Ì£ï³*>&m(WT—*ˆ­'8¹³¯PáÐ[(¢ïß²kX«¼guŽ´¨zv;é,B$X‡¾**{JW”ö(·]µÞ>Ÿ6bÙ‘ô“7HÒï ý‹ 4}îbmÅVl^ÐÞhX$ÓQCäÚ¡ùiZ‘û­ì_ŽA¹ð˜ðÁíHWJE_`È.»}A¹ž9P÷žo[Êñ4¤y¤àΣÙð8jÙ,®C5±¬RgA*QþáœJòèÁÿãWøÏ9ìœÓ¹/gÿéo'1%ÑU¼† uRZˇgB^eY3íuGÄn –·/j<ÑD¦åÐaf©pQ‹Âv©Çu|JãËhs–@&ôÕeý¯”&ëà}·Î ïFЃºj‘ï”ÌB½kÔ²q ,ëþ»Ë¾ÏжûŸˆ T Öh¤TX…»éÙ—âxh3 9)ý1ƒt¿ =Ò{„-ZŽfÀ¡ª\Ë|ʘV͵¯.Œ{ŠU&t»(ŠeFVþù©Í B­ s•uTOïAx.ï>©Rè»Á‹¹§‰„aP"ô±·ãrÊuÎ$sú¼¤ŸÉÀ ,¾‘n”ȽºtËÙ!9ÔÊÏŸ¶çĪ& ªÙ…ûبüÃG}ŤNèyfè œhÖ>BÕ¤ùÑ@܃L«CŒû&'uE) |±}Îî~†kS6óàp[Gäßÿdý ÙñYC]É.ÉRQò}õ ̳U‹ËJ9Cm*³SVv,DàÌO G•2#vˆ%¨’èŠÙ d±«™…®\.ŸâMÿ©³QK“êÈñE“ÀçLå¶a´<3¿Ž\ëÏÆË ˜#hƒ€z|ê·ç3ÿmëHÎô±Ux;XçbA?®—ºÞk&gäžÌ)6í—,ÄÒ°á™8мµ²Á{å*q-lHy+§ù3ä7ËÞ'¾êq9ò Z‡Uõ¦~Ψsá‚ê àíûÐmW‡ÉïvFíBsÊóç:jh¡Ýrá(z ­ð¤Ê'Q`G¡8ÃûZ"¯ÙN–móÀQÎKÁ%xñ´ÞäðÄϱ˜ÂêNú{Šþt¥¸U³2.b=Yz9ó8¾JD‘$§žËNr¦ñªénãÎáj— ü4¬ÀŠ…º=Á1©ôŠìôvÜ.‹¡p ü‰ªà2 ƒ7Ä1–iÕŸSU¡Öècn jgƒ­>êI95tQÙþ+f Îe×ÑÀ‘]À ÄÁm1H×ÞRô¨ÆÒ9`J5?Ÿ×\,g–˜J$ Ä-6I¹çµ4ÆrÜ&"^"sžÍ`e§Úù23AHk>Û®À‚ÕEþäKãx•iE{p 50ÉÈÔÉ»µðs–ÞMéyüò*EW˜d}ˆÌø hå<0¡í¸tNðšâÉ ½59¨Óê “œ¢2 ‰–æit”X¢†qù®ŸÚŠß¯¯êߎ‡|«ÝØÙ@Ý–¡É÷:Õb«iÍÍm<÷nØ“÷ðâýP–ªÜ1—mèÛV!÷®4âãnZ¸´•Hpq ™sò…•îÕ£çÕV’Q4ÅÊnÙ…×p½BZ©Î_qÄü;t(^–®]ú¢¯‰˜j]…©)’¸ñðK:ªz%žmùÝ£äb ¤ãúV™ôvð–NAÊsu}:ýe¯ƒOsfb]×›çd1u4F üe=riýš°@ßH/ÜNŸÁSüOJÈÔu¨!-O  w¹Ö¾?#„”Ù õb<œ*[I ½è÷N@L-Ÿv\f\åo¾îëÝïTgÄaóÜè9ÀÓgµ:Á4•#£©³MèÝ©@ÿÓ«Ï”n;?öÿ»yiò õ;.ÖÂöA(>c/;sî¤2P˜«3¤’½y_9¹öÐÁ,jè«iËGW:â°L ìì\†‚àV=ËdÍàÞÓ';×»‰ •X%Ä…™º…ñrq`¡Åøì06Þ°Ç©áçà×KÚ"i- Dau]³†iÓh°\² Ý“3Ÿ"‹¯)·±8ÀRAâsŒ’“Ýâêe‚i¢uTµË™KÙÏF%ŽQ…ãUEÎÌø/‡-𾛄OTl‘)cŸ£y)Àõa´MY‡éc&ᮺÅëóÑÜ‹ËæÖ_å”Pp ¼õ}j7’é+fŸ.ÍH/[ûÆÌ%z#.¿„n¸eþPsI@›†ø2ŠœÿjA“Ó'wªîOj|z-r’‹ ³¿)¡BƒŠŠËEy¹ñµ¤a{ÄÛ³‰+ÕÏe;C™v¥ºI †3¿¸7(ë±õ[LEs&ȵ® ”2xQpzl~DÖ5'Þ›~¡ 2põøyiý Onn˜ÛœW6ª¯âj5WÔˆwëK”GPs„4§õ™ËÆ^Ò„ƒ“ßlµC {>ï#fâZ‰£D‚—Àt‚ !º“cQÐZlD@€‰7ÀŽáT{}‡'–ÌcjW@‚FeNÀLQ³/• X”``<[§»ÞUů%çqçMü®"¨+DgŒRê¡QÔ9Ío­Þ ¬¹ijÕî[NËK³¥k0Ë*Í…D›ÿ=ÄîxÍ/å„„u9><(Ö=ÉRsX(»lÁ½ïºŽ_oþô…•6aÕPi…jÚý]pWÚ>4¿C]¡¯ ".T5Îm‡JŽåXrËž’Ä8WHOP„$« †C…"0ô° (Ó!§ý3ÅCRð¶ïá¦Ð ‹˜–ôÇD 9qžÊe{Æ$e ùIÁ¡g—Ÿðš*„Å&NË_ô@~í göê\ÐM ÏvÓÚ×fò²ÏĆ.õôŸ²ÂØ{ôžB¾a •-ï¤Vy÷ðæY°£)[ÅkfÀ¡ƒÚ\[ox(^[HŠ”íŠ.¯ñ§$ž×ÀóE½›;~HP=Ę'o ï…rª¢h†¢§ÜñÙóƒQ¾ñ1´“$ƒ;zƒÇ):{1ªÀ×l#ˇ±÷ð‡¢¬Y£ë6­Ã½hJwøó&½F—»¼{0m×òHiˆ=TWØ Cêh{§{/ðEžj}r©tÙ ÷ ní@¡|O<ýº¬íH.¬Èj¢&éšó˜ûÒIp\ß+–çX™Ù¬†Ms÷í¡Ô=D[Ø ¨R„²é½©_~ÄO ŒÂÇFç«=L ΈÄažäT<"™9;Š+B4O ³3OÜ\9”Ø6Ô)¢1.Zú  ŸwƒÒŒ’FÛì7j¢ü,\ÁìØg޳pŒˆ{([V½&ÊQYÀˆ©Ë›3vµ¶h°á6>ÄÛ¼Ú¾›M£®ø¶_i,/8÷Ò=Ÿ†:îŒü»6“¬·(,Ÿ_8dObˆC’hp¾¡ºd($ˆ0¯S}¥¾LÛ÷j«Ø&x¶W8Ö<±åèÈ࣠‘RÙ¨ß墡Z`0r¦S%ý8†¥¶d_¥.Æ æ—G;ÌDK¾ž€ ;ï…F{æì©Èçú4Á°¹ææÖ k›ÙÂrOâŠö°>øÅ}Z¥ãÆBØ_Q[™* »PÉÓnØxɵµüÔঽP ‡j³Þ³æK´ß2甪ÞÕ´Â/ÓÒ‰ÝMö¢ö¹®ïfÿ[Q§›:ð‹#á®q(½ DzP»ôp¨¡=fÙéž2hš0‹öA®+áªlŽqùk¥jñ&‰.=’´FåÊåɈà.kð‘*ÇCð]Ê.;;„òÖ Õ¨,YBçrÞS޽º‰ JG-zÑ!˜Ä†ñ«À"*Ð¥'úu½}=.[ò÷>/‰RÇð¤¦ñqHµ+ ú“5¯íwÓæ~:5»Œý†J¾´³ð/–gï8„·—Õ3仄<Ø #æxcݬüÃë9 :‘í་å¾Ýë ·RJí¤,& ö~ãˤÀ—½Ï/å½§±é:bF‚BxῘ(ï¸Õuqwt…´ýœýè¿R‰ÒÞœn‡k¦-~ì’–ðψ]3·ÙR¼;ÿ}€9=a“¬L ÿ5ÈÜûìyýo÷¸êx@Š…N=›*še¬Ê(Ö›A™ ×^OøÉ–«€Ieþ¾³tƒ±àÂNâæ²Vï†/ãe…æq°Ù¯ 㳺u<¦ª“&Å,·äDnˆá©߇õ¬¦Ùõ.•¡ºEȼ£‚n@îrô;²*ñå͘ÍWA‰Ò¥2úsûÔ1$ú¿úÃÅjZ4C•Kè Õ eįQObˆÔ«™lTjã+ñþEG¼`™v‚<Á›Õâë‚ÕCÌÅ›ù0Ø$½O'r8ü/Mê©öWÑ´žèÜ 2<°BãGª!°¡nŸLÒ9kUYè©@u¤Áx/’5Hޜ婥¬*YmxsÔîf ÀÞ Hl…¡°ÍgÌ|øX¿$p©Å€“Ÿ6ÊjзWâÐË´(4¨« <µ^äf=jQè©øÀ}˜Ã-!2Ú­bµ1ÓÀ#îö¼$ia÷èú `í Ò0Š@:|Dƒ×húZ3HñN–Ðå„ÜÄû©«@O{IºÖ ýÄÝøˆÑù‹Ëß$KaŸgµE¬¦ ÙíAu{öÀRT®3vðC$?}Jú=\áwÀwòq®YsÇÁ< œ'~j¸ƒ’SŽs)_<4L"÷ηzÊx®Ÿo€`"ª;?8Å|ó¶€ ãŸŽÊ‚ HóÖäüÎÀ(@ÿk¼ÃNëÔQ†>âaÝ_![« ŠXËD¡!í‡é€ã;AE'tÖ~ÇPN”•ˆ“zp(µÎl€§±NqÊx«Š?«¨gÓäi\Rp¼Åˆ5p'ÿ59yçmɦÂ~%©˜'QT\“#”r¬j<Ñ5³&3ég*¼·‘èö0{;…ΠåEZêAX¶,HQ,ëm=“ú²æ¿å¬zeZ‘—Mšà3ÚYÅf‡_ý§ ÿÏÐkMÛ›TûTÒä¯H•%‘¤&È—ðƒ² ¬´ß?=:ï^S j-ög!ß~¼ÆëX§Å^å¯+! 8¾;×?Š£iÏ[†Nàa @ôŠB4S¤˜ßð¥ R »Óÿ•¤¦¾¾°Y€ˆn×4 =#–N”àáP¦Éסûwi)?z› A,1Þ+'‘ápf˜:G°:êÌ]²pvŠ6>†$6ŠüÅǸ¦W´Ù‰ä>ÛF¶­wìƒùËlL†ïüdu)æ‡Ç'"ƒý«î@=CÀ@NXî¯O ào¶°Æç%ž-K–:u³^WcÄøpÙÎYç/e„Qõì îzIÁúÒÓ\.‚wñf‡%ûñ¢ß>ò¡ÂSxŒâ}—U1Q¾Ë˜^rpÿ­ƒ+An«§ÆÊùkjoFœå¥*=vñ.³¹Åù%ç”h†'äèˆFp×Ïl±‚|Î;ó蔊‚¿ ®²ªNÝ'¾æţ*®‡ÀŒ+ê‘èÃù ä$6´¼/?ÏgX@ý×"q׳"b±Khz!˜c6GÝFŸEvºp¸¿nظËÄ pYèöÀ;¬)7âã¶´øí¤Nãîo삆õ‚mGPÚ?’µ’¾¤x'äÂ<'ÀIëõ•"¦Œ>“"Rº-¤õ:Bñ—ú$ÓÚA³ðfßËXŒ¡ÁeN±q 5™Râ¯O4LÙlžetÌ#%Ä:=¨Ôlœ_.<=SÒQþjäàÂ,1ÎFޱ(ÝúûË…Ù±Äü B®±V*dKqÅWhóIØe{óÀËH­8T¯~ lßpRtVï‰q—›aÞÂûhaíYÞo™ñœaù…P™ .Ò© “b–·Ý\¬sýåù;vsØÂÖ‘ÊÕ毆[‰6]{Ðó¸ÜQpå°T2f1¸p¼^u¤$ϧâò¥s''tâ‡Ïò?Š&©A½\±ÚM4<´ÈÊÇö•%¡{–]*zs¢ßXâ¾N„ QHCø)Ãz>/zñ-fæÿcöü97ëDV]µÍ;6ŠÐ›ÖÁ”hõ@ñ#º£¥YB‹Éi6]ø2œy½’¨½)Ú×C;™+нŠºÊÎ`%: _òEÚM£´(ôIzuŽƒc4¤2ÓpÂTÀcÀkDÖJ½e–¢çý¬êÌiÀýåúŠ(鼇p”Ãé4sEFɵòaúBh÷Û€ü_Õ“ÊýöÔ ¡¨i”to¿`þ¿WÍ?AèïÏÜ7:J80ç/Œ¾èÙ*‘¹-tý΀9Ù¨|\X·©£0tš¶ÕXÞéõjRu­,ÆT4üÑ:ÞäÐpáS4i£."ƒßΑ´{Ó1÷±à0Îöέ…”J H>57ê;¢,²ô€¨ˆnlÃSÀÄÍk€9kÜN!at²tÝzrÐÁ¿U­Œ))³ Læ'‚Æ¢5ðLhbq±²A&ÁV2¢¸ÝúäЇœuCƒVÑŧ¿†Jtê+aî<;pD\„_¸\‡°€áÓ×Ð]ÈŠá­þ¹#…¥ŸH!¶ÿY\e€œÕ~E˜FÖU¿:yu:¥ë†omžª’®³È8Vírõg Oœ{AØ[¤OÕ„#ÔÊ@íï(å,\Q­' 6ò_¦_Õ05áS‹¦‹(AµJòŽIjýý­!öŒòïu×8AçêÊÔ% Ï˜—ýÖ¦˜«¡¸nÜh¿°ŽwL@°|ªò}‹Ç¶‚THÝCû; 5³Š‡ÍѶPoqÝ2§JJ‹$:z¿*é¿’‹Î‚7¹58àZÍÒ³³á¬lÑÞCµÆQJRxŽÝíFo¯=ÿ’¤U'] ‡Í+&RNX”@> l渨^ºD" >}†vHð‘ôðú™M9‚/¥[É£;Oå±[T:;*4Ìð¢Þ-õ¬G:ÓéƒâQUd±`…£“áÃàžt±È”¥ç¥5¸¹G6ßIÿ)è&ðÁžÅÞoÒ‚íœ%Ün¬–GíœZåÆ<ŸI÷ â"«ð——/ Üc[øñzbž?’õa}º—Úchz”•ÌnnÕçZ¥!}‚b,öY ™7‘š&;’*@¦•&ù¤)»}Hg2]¾§5‚—]–;¢Ä±ÊGÆá®¦£ŒSÑekºŽæô?¡p8ÖÚª×9}ò$F€!ÝÓQ›y³!QÚæ†„ï+ćF¨Å:^ªn€$€öA“-5<‡^ »Õ5j“†ƒ…S¾ r`ð;#gB )&l@F¬QÁ€+Tg½y–Üû2Ú±Ÿ˜Ùª(âôA°á(l±t‡ÚškÕ¼ƒÔA,íŒuÐoÇfa*G¼¬1G0Ü/8ÊA@Ò*I6%¤Qã `¨µoÁµ°Ï”IVÀÛ¥tgŽ%¨“ªyžìfN‰l9†êyCyÜeÿ×S™ùr&va±þÓâÑ}%‚ êÞ¸àô.Ô¹)áØ¾`¦ g®œð»&ÚI!fþ,¢ÕYc®©ÍÈæóûÉtæ|Ÿ^3ž¸{ø«²ý#(UÎt þ‹&Q1ÔH/ÙþPÝã}ªè$ -Œ]„ÀÄO1‘ ãÿ2¯V¯i:æ˜#ð>¨?"dà¼ål 4„«0q¶SîÇ„:+¤wþcñÞ.[°Œ9½DýU^ߺ½Ÿa§Çå ¦›cýká·‡ò}²=WšÄߌÜðé Y GžžcN‹¤œåêu%J~¸@¬f=ÿÊñ$ÊḯƒM"Î.³oÕb8äf0¼2¸Q«C#nö…b´#‡'p§ÿà`Þ ¨ÓŠý²ê Ò¸‰¿•-ĹYHªMœ d2éÌ4¨†¹9øÀt…MߊYÓÅ;î!¥%¯‚ú},ýÞìï?`Ô¬·‡fä‘Çfobn½‘õ.*s[ÐÌíD¤£ë8Ûbãì‹r“3MüŠ¥&udÆ€d{24ÊW¯ŸÜ‚¾&·è鲡¬þ¬tùµ^µYáhåú(¯QS’Av3Ó§¿€–ÞWÌ쓆cÔÜA3¥7ÔÍÁB\„ÍÝ,æ:|CA›JjkÐ÷–±¤Ì½YW‰“Sô%»_² ü :üi¤\€xcÌè!-»ÈIÅ ë@4 4ð²“Ovoäc¾%3Šr›ÚÒ\§ <šÏÕÓÚ2õ!ݯˆ‚ºà²‚Bñ]ñãEDþO÷¢áTW ­äY¿f®Ò\-º…ƒÕŽë) ]<šA¦-Zòjï^,bŠ8³®¢ Tzü::C“*'z ß’_¦…&•E—JeŸŸ ìoNô:ŒúÒ cœqp%~IïñÚüÄ}qmÿ·fpm¾óg÷ÒæÑy0(Q|«‹¹ím9©’úÏ!³pàqùâßî‚[Å£š»­BŸ{}¾‘´Fwþ8€Z2aÔ±"pûV(§öjw–þÒ5ymÈ7£€ ø'”¿Î±˜Űu½¶ˆ8¦ÖQiö„¶±«Ûz†Ä¢××µïÙ·{¬)•S<®´Ygô.˜5Úžv€ªââø.!ÖÝ[Ûì6­ÿ4[íí<#×DÑÖªR¯(Cn¯ÈV¶Ø»ø·} çVà #,ˆÿ'íw†jAgÿ¯Å ëÚcDCeží’K]˜ÜÖÿ©ÒGþóØ¥éT­­f#«CäQ1«Ü“KÅféÁDG¯g^ƽ²zÎÉešD樉PÍÑÅVÀ~¹Nu Âk‡‰=G—øèEÜðkfá+è3\xâµÐ.4oãôƒH®K9Þ£¹0’S~ühº„êOd»À%B¢ŽPÓ©3&lßpz\ˆÆ]·»¤iGÚ<— .ºðƒz@î Ñð<ïC_»cŽ>—­Ü÷DE[6Â4ˆcG*<Ï‹OÞ¤OÈNÕ‚žjœ°s óÂ!vK<ð»~U>dí:õ8"Y©jÀœdÁ:û9&ÔKÖ­†åjwlVn³ì/=c´§TxcÏ&‰^M#Àêx (~%Ñ®û _Xü P›Ç÷2ÃŽØšÄæyG5ã1["7¼ï£òr¶ÀŽýâ‰txÕØa[Yàq!¬À[Ô—±7£½Šß@ßýmÞ˘¹õ)V°67<èWëÕ¨åñƒ†!ÃðÑåè›aÃë Jý¶ûþD]à•Z*·UMÍ=\óÌtûN øÂYŸÌ3=ƒ~Á^Y{ž íJ½CËræµq$´ðé×a xb2Þ`¥(96Hâ­¼ßO~(ÅÃ"-ØfD­­ØUpn(ï ±´€ÄYÙ‹Ê¥­Êƒ„!® ÛÐzWNÏë âÞiö$±ì2K!‰Ø€LöLvÊA`´î|Q§vb©gÐvfjTs_Ö3ÞÅû9ºÏì€ß1¯Þ59…v4+똵ÌkP+p– /ÌR¶‡R¯§ÿ“ÿ|2,Ïhñ脊ݨ8™‹ÊdR<=e®Ù»­7K²‚9­]ªn´/4}̓Ã|ˆØ^fÒIËþ·ÿÉmM°¹,1†” ëÞ7‘æÚ> ‘±Œè,_¿3q4-ö*c,áàä~wuƒHåh+¸!6b!áà6$B²Xo±‰Ø¸ÎÐ꽚âã¾£jÓ’µêh ò„t§Ž¿é`éÏZP¸ ¾< #C_'§2¶‡nAõ@²Gèû¬wäYïÒ´Á?·CÍåG¼»q'†xÜëBª¥;íbƒ\ºµ´Ë oÏö5¨ÈR\ÐØ™¸ÞW.zƒ[;=y öP)˜lÓžÅÙTHè±.Æðìq{‡¸ÁáF¾»i,M=õœ]!€ÆCxÍ,!Ä _žÜ,âW¡{ùÒwõ š]ñHa©P·aA¼þmUüdœ2¢ó-SM¶k¦Ç–|5Û¸‡ÆXzØq2´`O]ÂfÙE½ 0fÐ<㢠/»d®K„AÒÆênÂô÷û²F„I®\¹ùÙEÎÿŒèU>‘{øeIíùþ;DýCˆa5ß)Ó[6âÁîš‘ïÉxÇá­{s²Áç`ëqyŠ=¹»Š2ÒÍ7ÛÎ;GZú #ÓãðÏyî4«¹í¸_À±~—ùß(æ&¬Ôò|}¬›×ûs}­†J¨Ø’ž% $º(üÆ)±ZÎnð³œ{°âD¦mrDf–rƒ¦¸½ m§ i¶åžuab}NäÏMcp@ŒÀ:Ö(Ð`5Ûˆù’¢™û¶DÎ -Ô×ùÕä ÀcîJ,#q/x”üè½/š«½ìp´ù½'Û¸Kë¾pµi·ŠpÌ"»ððÆ6öHNØM —K¾[ïÔi³*ªðAL0N°µ¬1ÃzÍâg&ÿ¿„Oª6øê€þþ:êœß‚î?ôáoº6˰.’¨©cõØãjȽî>e9UÙd]¯©¤ÒH2@aÜ©UHJß'g„z.œgšÿÒG'ÇœjòciËÐ3~„,”ŠYP(¥r,[ѱR.|LÝ´Ní¨]ó‡&©3ˆøyà1`#ŒáÁ HkVÉu¦€ôõ¯7{8¤9O(Nq§w¨ƒÿŒÜcÚ:?È7Cx_Ø#€3dÙ 8LâÐö¡€8Þx øX¼úf쌪Ú''œNø‚×ms”øŸ®²Ûš¬œ£ÖØ|ÑŒ[½lÉ–D¥ÝÕöi8£rR(ÐöO Ît[½\íÝr¥ +1M¤1†mvOÏØÉìÚ'R0ºçWûÜDÐùJ¨orÎÏÓù÷âÂìT/ˆŒŸü9¨?ò¥r¯umþ¯VaÜ iËÑ#C”Êzþ¡R¹ËiõÑðéO¯7ô¸éèB—óÃH÷~å‹•¡K,‚´èüOøIMî-kÔ`ŸÝE¨Ä´rÁ£W 9T|îwЂӴŒM@¾‚LQ$Ý€BÖÄÂæñžXŒiÜÂì.~¼ãä*À+¨OÔ0TWøÃ:ÙZB l©‚ýýø:^í ² ó2y¾=ÝHoÓV/c7LBÁ˜V8_HJ)~Ì¡&ÐÊxþÚ-•-I(@à/¨2öX$šžøÒaª‹`Ï(ÎĈ߰ºXÙG°r Ôž$hu(ï×±’*Ä`1±(a¸ÈۧÛÖ–i˜GM§úû3~ ±¶W—û{w8—ƒ-álÑ»ÏI¤T‹ÒVà‡Tâ¸,"‹PâûOÊ©`i¸XöJl*‚mVŒîbIc§È }#a=…”£ÛGÿÓ/dù²¾>"<Ex‰Í›~ñÍømæQÃÍÞ¬Á({ï¶Ñ“‡‚£&ʵðê˜lü<"† 8äcº ÏÜTZ›+j[Á¤HW Ye¿¤Oöö¦˜»ÀrA#{pFÇéÌwÙ̉ˆýD²ªÍÖÁ9xÜÒk„\¨l ©z‹¤Ð׿ðÇyzŠ™„ò*NsDš˜Œ0Yö‡¹ý3Ò}=,ʶ~¼sÛx©dÔ Ý“ŠŠÊö|’É )†mÜxÉr)4×§]Ä=žé” äˆôXöïäæÛå"+¿DùõXý¼êëcÙH4€0rã¨L©Û/LdçÅ Pëüð¨ÕÐGm"V5D>:tlâ|ï̃Ìè ¡ö«3B9™.>ãÆð„eÖ/jMÉ‚½BqWÞh|nílŽ1‰?ªÉ+O“ü}G½Ø&µTÿór˜s©•25[þýCí¤ìÅœóŽæ3C÷Ã*ýœcU(¥»¿ÃŒLªuÐùX”W“m·Å†ü¾-ÕrOxڌ磻åѾ¯[ÿ´W ÆÅaKÕĺÃÓ¹yåÓÉuú½ØIµbØÆ|f¬_ œ¥Ä¦¤\¢™¶Òz¼7œ4'4‚lºÛµ™âÔMÅü–W-ëW;*ÚàŠ…{ÖEý:,4Bff¦ÇQqôSëÁò w6šPº­Ur 5á'÷²€éIâ\~uÉ<ÿ³𠀬dAÔ훥^^»cxaÄ1é/âëN–Í6AuÏo5-Ӕ®XÏLQº+–߬ìµ_z–Ö-èS趨þÈ«kè.£?8­™Ót÷w¶ÉÁºKŸì÷ï~w Â¤“hhMpUI~¢ÉCSöqÇtaÖp%RÒüWI BЦ,ocXñ®å1ærïiüÍößê}Ç»wû3¢Ü7€ŒP20¢b]"|@e_b•Œýe/×þä_•V6­GÆáîD¸ÿz¡Ã-`Æ —xQÎ;©×=ltÒçﲤë*+ã”´îX@)vÙy‚mFxì\çTâo·šú=“ÉÿÈXÏ&b˜q°3äÚEä²±ØAì“ã/¦5Šò<=ÒižG¶0\…fî”Ã…*1Œ÷fòaÚ'=Õ9ÆK¦¼’½±Ë*Ù¯]ßn+¬¦¬„´úà8ËÔýºÌ$~áð.úxl àLòçŒDŸID®–MuOŸ‘)œÒÑs›Y3*¤h›’#0„qºÏŽb[»LîH”ÎOŠð2fà>bï?9û)J'#ã~™Ç;7@@T2Òh‹Àå@'_@šå.n*>è¢SI¦ƒÊ“/7Þ¦¬ÙLøÆm÷ ÿ£‚ò"WœUé@}ÞHrc,$¼ê_Ô;‚ß”™&¦µŠ%¢¾ÅÝjÀJWåaò«Îr††´æU'¤öÜÅ>±&ñé'Gt½L_¼ÚKmç§@¾Ín°ÁË‚ögçxùk3lÖ€Eìg²L­žV3V¢¨¿{ˆ&úá÷õƒïÞs|²žû«~ö ®sB <™‰Ÿ6&R’(vk¦²÷æY’¬Ù.S߇«ÉŒfSØyɆ›mÚz:®3 ª È•¥>‡ú¯Œ®ÂŒjƒ»þŠÇCïH‚ÊIHàbU‰|ô¶¯‡ÀHïÁ~d<æµÁ½úä‚ê‡UÛCE¨<Ðg6V~NÖM,ØVP‚‡G·¥ðN¯üj‰ÈÈ®nW…d¹µ“cÕ´J 3¯_Pe©+ vÕæpë}=½_­e5Ïo{ÜAIø7 køæ µió)¯nep:9—"ډ˧{Ìqø'Ák˜7éy“BÔL>}³¨´"äŠA×#‹c¾d¥E­:.°† !4Ýlp7é¡tåí÷s‡‡ð^ÜoìTeÁ6 $ÉGÿ›å¨¥..Hßj3ͰҫéËÊe a§òLr´¼$LRüÒ9ů^Œ ÷7Ö–Þ!«ˆ¢ÃA¢‘ÂÇŠxK£ Ø ²¹,K^ ÄàíÞS=ò²˜¬ekksˆYû”©+31&ÂmkxÉu!#$Ÿ¼F>Æ!‚tö]xzûìµ#«ó7›Bµ)e˜Q\ÁàÆL›¶OQ¤„€‰íÒbì/?UÁ¬hŸ»±ÇBºH^[(B£ÚÕüÕ…çþÔÊUk繬Š|.y»kJ²hÜæ'R~àãzuìþ<×·ÇIø­­Á8ø^ô“y–¥¹b{ifc颹7ðþì¥L}œÚÐý¢Í™&B±Õ¹-ßgßà_q„±JÂkÖ§¿Î>EuéE×–#Zêê‚‘ÇðÊ‚w¹kò ÉŽNuåEbl[½A†ô´ÉŒù v¿U£¶6Ý{Ãâçó>£§9ÇJgqŒO#p–Ã1ZÊaϯz¥<ˆ~?¤¼sBÓ˜iI{=Ôn[€Ê—„pLoÍ}^å•qÊˈâB¡þ[å›2Â^÷ò´ƒõJTQÆÅ‡Œ{Ùæv¢¶ù‘6‡tx¬d‚±³N5CÕu.B°9Ü*Ð\²éÔöÑÞ¸ }vcz‹FRÜšëÀEŒØÁäʲ•qËÐ$ y"Í÷JV%8J!'Þ÷ZyÐéê?77¾Gå ñ-Õ.ŠæÚ¾ñ-ÔÕ©·àbaÈŠHIŒ8pè¨{“ׯ€¬H{iÒ3Ï5bæ&Ì»{RJmÙ7%²¶È‡®AOÅ4Ù Ž•z¼þ(QŒ“68 z4 ácu¿^¢šã\–ˆ%}Øá,–?}öÓ,ø< ÐóÂlvÀ,‰"*€±ß[6K•Éoˆ[ɸíí{ £'\žqÙqD“k %?;DÀùýf 4´-µ!z¦ E. ¯‡>¥Þ[³WH ®uG Ë SðñÎ˱7÷òÒæ$Ä ¼0Œ÷cb§ÝŸX*é ¬šâäøJ›¡1×ô|Xi׆l Æi}Š O—Ô¿dm-öT†•ö ÷øöé,Âòlú5Öæ>^=aÏ}ߎÁb½ÇA¬lú†·ù5¢L°€(oƒ1p"”¬}ëë¼:ÕáO,¨8*‹E9Ñÿ8À•S¯å39”<Üv÷n®»IÊÀ‚6·?ZÎq«Ð¨¢fheÏ î»µ†×ªÎÆ߉KÐÔ-ªÄ'Æ œý¿R“n MȰٻk´º0¿ql­Òž,ÿÄ èÆ¶ØÚ˜q~]èÃ~„”ševU˜{Æ_uχª<èû¢)™Gé€h ÁJ!±Ÿç`H^¯*›v¢…ÝåâKü[ÂP_õTµðÓR¥ü>nmï¤OËLLØÏºá~¹oEç“ÚöI:«uwÂÃZÄÛ:»Ñ™6!‘®*cù}åLõ¿Zº€ ´ ùø9hœçÁf¯ùIÖ÷/|KŽšò–o­é…X³—ÝÞ‚l"zóîP M 0ØÛ< w’ý1½f+ذN,‹Àª~¯Ûò£<í1çŸSW¹}ÈQqt>-(Ybây-½ÍûŒä«O÷ÊÙ°TX1­Bö7²ÚÓƒí…ßóìÐd7YÀ^ÐfÁ/÷«¬ï8ILJ_VJÃêviol/ÿ7gr…?×Ã+~Ç…Šþk–rƒ¤4ž1ë!£=FtêHíe‚¢°eµºñâö:®G#报‚VÄ´vÞ=êVTý{Z p ²êI)§¯ÞiÐY³–Z+Z¡N„¶:a™µKWR-äÍ„ûïž ú%ÕH ‹æyÞ‘Õ»#!·*îZêîn_tÒ1cˆ¶ ~}(N8mýG”a IW÷ßz£3ˆ;X±°€€& ßžRZ Þ‰ÁÿÿœæÉ^NáÁ íÉxòkÉã§2F§g‹úï0¨‹Å-]¥³±" œD®Á‚:m“Ü» EUçém¯–£§|_EÉëßüÖ2=‰_^‡Ó2í $^Oß`Þ«%B·3bÁ¸Ìgæ·L.× ÐLS±ÓYÒORw>"5¬M#¤×¸8çúrà‹±›”4’Òi. #¬0ŽÜoØpÜø»µ±W‹.žÕ×j‚Mõœ°}>ËžƒUžÅº(~÷ ;òÙ™52 EìQÖ™$ hЬ5Ûìá–7êêR:¨0øÇàŽÖ´ÉX,A²”ŘˆQ#î–IZÐY㇕Rª3T¨™}™ì-p`XBÂf5ëÄÈurÖåfIâÉ`C,y®Ö3ÆNw;鬸¯1¼ãv Q*ñ;ô¤ž_ú:VëQnÁ#ŠÜ%âÅrykÏÆ¨.3—ˆj5ç̪E«‘í³CÛ`cÍ ½ßÍ[ÀÄ0¼Z™H÷&qKÒvÀs`äÞ†iÕa¨v¤nV–0¨ð½O¦ùëÿ­µVæÊxI$J§®9ß÷d¹31¸€ÍüQ®Â åÕÎ*1«hAG u ô˜fû(ÄBc<æV·Á%¥€°:òݤ5n  ±üpƒ#%]÷É;““.®õ¾Dæ*‘X(+ò¯Ž›©£G)û']HŸs4¾¤lûçÜ©àB²m{°úýÒÇtüHÓ9ÕãØΗ Vx(ÜCgE£ï @½ªD4 ½%Ï•ÇÆÖ÷Ö-ÓÊ›1"dT’ÅÇoåZ&GñJw‡w!MXÃaO¿áBÃ&³W1[|ojxIX.h«â™¹ bBxÏB?Tå©}Áiëa–ÆGÏPÑ.i}?ªép’¯)v{`®¼Sná2´ÐóNŒÇ\‹èßË#0Æ®ý¶Oc´à ü‰5´Æ6»oðŸ3cÁâpÆî§˜ìü„–VA×¥>0ß"5ÚTc(#º]#)—ãfŸú­Üh95‘åm<‹³å?{ã)Ïc‡ÐVw¯-˜_°îeDü‘~/±Í¯ÆQ!mudŠÄ'Ý2ˆÓîð§ãÿ—ߟå¯í  L´lç?óÀOÚOݎ•]¬Ü0Í4 ÎC99…=š3Hîc9ÕR_ÖŸo¢D ¹:$:³ ÈÑ Ó4Ëm m@dŸ£»9X`Í×(êŒ4?šæ˜ZÒÙËÛ®M[Üï}á}ðQ¼Dl f`è&”Ja¿’*Ò,#¿$p+ôÞ²¦XQ"õ ™D ç:PLXµ*&!ÏËnƒ&ò$ ¥Jàçus}M³&6' ªè~%HA9ð0¤PkMt—PZ4»¢:õAþ‚B{™ Œ,­A„buÃ]lîÆ Õö-'à(³;ãÎuƒh|ÖM?¤@‡Ñ†¤j[ÔJïÉ ÀÌës¾E=™ùe¿…IÃ06Ádˆî¹×fáy5Pš6F Ì-zœLÀ•/ŽRý+Dgìª Àù—?üø§±©ÿ´¡òð°²ïƒÊ¢ÑŒrF…«3¸øE{›‘ˆUþž­îFà!åóÝO4[Ñ+ áŠ9\>¨ä`¸'²©gèNB.Ðè*]¢ÒÒü‰3*O|Z8^$Y£«Ä_,o¡œÜ¨?ÇÍP­mî_[Pš¨vpoËîƒ =ËÛA†êWXµ¦oºK£YoÑS™”më(ÊPdÖ4À>¼ý=+ÑÏ™‰ðHÓ‡?„2<Ÿ·Êë=سA|¶6éÍžçž·”—ÙD3¡(€›9G8¦.š ¥ÖžŒ†é>”–þ¿ÄƒýoφjfþÒkÿ»áþ§ízoú;†ŠV4lö¬¿Ò…|)Été¢BkdKšˆ†‡ÒÃA¹(µ†ÖÓ&>GkQ]˜iñx¸Çëö°S6½]5ðï(ÞykDÄ ™Ä>ã‹ÌÆßîƒçŒD3¶B­C„N1—6ðƒhPQ«nYGn©D¡FEZ-ÍÚ_§î ilã•÷!¶Z#Å\¬8&Ø’=<ó%% v=·ÕuHЗ~SôrÃà‘â*@ŽL“.^O YV€Â2NXˆs‹²uÑ^SÍmeu(9û2F<¨7*{ß1¼a†ËnȘ;ŸŽ¾}n áÍc7¦„W)¶ïk.hæÈ^è—ùÕh°û‡£`EôK,ý%Ïýí_[ýÍæ±Ò*>!Ë´@G—¦íòÌ´BF¤eÓ=HúM¨+ázˆ£»aHÇ…1ÆFzÑmŽ€Î&ýÊ!:^»hlòeY¤aÓtŒ4ý¯ñOœFý=É›—le¯´XŽ7!¥ªþê-bGe6Úý¯â̈.9xM‰ooä fÅì꼨ù6J—ÙGR‘;œã ûþ¾ò…8~›X¨1¨×N«8¬ýâ ¾Nׇv2á a<^æ9ÿßÉ@ÛFRnº|2½¼×Ã,Þ|j ®l¢»“˜w[:Ô•×ÆÞxªÂÅ›åèW^m§ÈJ·`'wíñÚµÒ-/ÁµÄσ–d9H»#ß^êпaiM_Ë –´Ï¾œÆ}hUùKŽg|¤I WN‹® ZÉûa§€_Oë ¶þº¶ïï±L9xIeå]l!°aЖ}†$3LdU ‹qä"‡Ú &H©¬FýÞ½à%´€'·À.¡Ð|pÞìáßXõô‡<óù‘2¬vˆB TŠ‚æXÚ¹.¨éí„mV²v~?:®;e 6ŸLÜÇ %ó8¤)HØ€»B¿ b¼M^YöE(e,'£å+¥9„²Î7é¾dê…§¾â-þsԾ▲‰]Ø£²lBõ&tZ´ºÔÖs ©Ðhùšƒ!|¶éeµ.jÏXçB¶Ä¿#)úB`¤Hvݾrý$D¼×Yɰäh–ˆudOÑô€R#vÊQz(Ú (Y~‰M„™…‰~tQØm¦½ÓØ[À×þE?êk`w6,½…ü<„è·X:ÓiöAÞǽO„÷T¤w ²PÁÜðÌZ?TáGaôè=Ã@Ín»ü™h T1Uã`¼‚‚ˆºÙn E¨ñÚù³ÂGa\ëv»ê1¯•kdpq©:1…îQq?`h•EH¦.?Tæ¬ õÈæ<~íL*'.[´³V ÐzSvý°tÞH—òû¼f¤:7Õ›),çZ—€Ü¡É†•ªÞZo¸Es^MéÆBúÙ­eWméX• ¾ü"(yô „ûgKBäpÓ±[N¢?>[·¿ÔÔv÷²ýG9iÚ(yÛ-Tï-÷9M ̉!H`†¿–¦-–…«Å¤e~ÞÄ„ß@K¶PÊŠŸ¶Ø¬õuÖ»Âò$èŒi&ÛÖìîCuöABÿIšÆìÌg²¼P'Á%Ž’3ŠSÜê«jÕÚøéšw,!ßv"†o·n5ÿ´¼ÌL tÄDÌç­¾¾1MŒB“†Ó ¿žYZAi£FàúÜŒ—zL¢%÷&5ZQÜoâg€O¶0ñÇcçë¤Ï`IY˜¦0ƒçgäæÕ»®8Ž”¬š_ç;Qêa+€¢ÑЧøT( Fëe*¼qããá«U.à'NQÇ~º6LÚ(ºÚ¿dåÞV=œñ&•@ ²Ôž™A«Ï–bV_tœj5D[¿GÒ¹£P4ºöJ¼L.<ÌëLC4³¸®„’ÀhvðÆ×CœÿÐö„c7„ü—3¸à”ËÐ:êž7.sÌ,fóÅ*OÒ[0g¥™„÷ÚwÞ÷²pãGÌ>T lmt˜€¾Áø Ìù?sOÿD¹t#O ŸÈ`¯$ý\\ óƒêÛ Ÿ3BOGÇÒ·XÉïw8ù̯Ùç ^ÛÁ¦r@æ|oáÆ¾Ié0·ù_(ü£'½ŽËgtf>jTa%FBÐmëŽkÛ\w>Q,¾ê¬M³ ¡ÊÍ“álš Ì3gœQ2 v»r?™ÉCeöZåÌ®xñäŽg±ñ‹âà\‚´³“Rø±(†?ˆ­SGîð¸óÛå¾~øŠx2Ö}.DÈ£VíK±ÒC^Y>¿óÂsöºãg[¸óÖö¦Wpp¸9‡4b*ÅÁ8*¢?i9È!²•C¤œw¬´ºö_œÝLôÕ¦Äÿ¹Â\Å[,î—ü>ž¹ ªFÞ=ëá äT(÷:¯¼ÞFQɹ[ÿd§ñ[³Ó|o‡CáÐkt–™sæ9¸Ë>-÷/$Q¬r×ÇÄ.ËUF”ÿ€Û‰Xq쯞¯„Çsìt|’žVÁ¶áâIö„ÃImÊ9.¯þeZä´?5qÂÄeRWŽb(Ã*dƒûj€ŸÊÇe÷ÈéÝòò`/ïŠ8ñ2H½ì+úɱm*hE;©4¹›è.¾†§»¢_¦t¨&—‘ϲ̾nEZb­½T¡ƒ” 8ZíùâÔËÞ%¡}ë»qRLÞÆ¸Ê%ÇGÉGH«£è—NÙ,þü7,Ü\™™M¤ßº,M4‰y)è™—ªm7êé&§å«§Âϱ;G$Iyx#iRˆoͤ7E0B÷L™([{Ô´aÄDޱ—ä$öå}‡èâ¿@Ë\E û#¦„bp©7 ­SÒ’àï<ò%½ŸHÐRgk»yÛy"BSùÜÕvΧg…ˆ¢û´Ìàr·xûF ðÓ¨^Ø) ßN™×T^kQÑË*v¡¹æ»†{L5¡¥€‘1÷k›‹õj´„ºAJq©DÍBfÌÌ‹p¡¢ÝŸ30T_£Hœn“Ç€.q}ZaZ´1»4¾‹ÇŸ2º»|¡H{»ˆW}tϺϼW°¹&e¿’jžpÖeK<ñó—‡ñÓv;o|·z)Žnœ.è°ŒO ¤«y£¡Q 0&JnPkº nö묆Œž;Š)Ò^4l—ÈÀu_uòÚÚ]ˆÆmÔÃ'â©P‚ˆ¼ÿ‚ô'âEÍaCä½DK¨Õh3à±TžpÇä%Hç³ÿü?ƒt,šçgïÆ/$Õ¶mÔ¥c¿?vÂAGéH_J;oº¨Ôžd¸g-6˜tˆ˜ÆlÀŠ—!UÎñæA,T³j‹|,_än$YœÑ¼KM€ óø‰ÙheÚ³ÆÒæUµ0BïÍ$ƒ)'[.b#ƒùCªš'¹F •BJW®/T S‡ãÜw-f—ùh(¶ w-sÅ0¹Ô<´i*±.G+êÕÀ›—0ˆ]wG:ªµûB,Ý[ô·0Nó’P°­˜Qí]qIu·Ñ׃¬ãz5û³Ñ8sÅ<ì&<^FZ?ïúÌ̓Ôë„×[*)é‹+†@¶@ò±¤Øž·‡‹­ƒf¬‡ÇGc3'j«Ââ§!­”žµ¯ÁT§›OYk‹4ù•Ò?>`,jªrò‡<×®2úX#êsÄ“Uâ€Î7ù~ú7R¯{ €²ŽVZJ‹6yXv:þ$}x\áë9P,öÀ$ ~±„‡âÝ3¼k«x°¿"ºöðÕ®Ð+v晚£$h—'dTÇ9·c?6¹y}z¦âåù®BY¥Zâú÷ÓHSÏÔ¼\\¬CõÁ” œPÏ•ÎØ„S¦Ç ß}ûpiëCab¥Ñd– ð®) 6l+®Ãœ0r-Ó3Ãd•˜¯‡0uiÆZK>û(p5¯ˆFüPU èsì­ùyô<´C tÝ ï H'›â^:¹$ l¯2ïë3gkBýƒ°ÊßϤw…÷4¼×¡±máKP2-ÏQµVчhtÅÅ"rÁÆš™krÐÚpŽs¿vyC—Ô,¡1ô˜ÄQ81š(¼Ý²\º‰ÌtƒR'íq¨l–ÐbòÍ}ŽŒ½¹àŒ2÷ä·Q€–w–æp[âÏGЭòKsÃïV9à\²‡¯ 6E€gë+º†q?žoui…_Ýé-NÓkqˆnª¬'d“wlJ¢JÆyñoýKmD:ÃU|àWjœñ:¥Kò‹©i¾É»Ï§Û\e€Ìú§æ2Ô)§ñiÒø@&“gÇ›é€*š¤2žßÒî&óƒºãd¦l0¸‰ü¥/Ý ¼É;n¤uQkäF7Œ™8»¾ÁjãK¸\{š”vºv%ø íIª(§ùùSÆ…“îrö\âåƒVÌSwÁGY³£Ð‚!²={rOPZ”jq.J\â ¡‡&"ÿgĨu«Îå‰Ã7ŽÈÝ».6;3HnV s³']kP–¹BþUÆrH~÷”^/óþrß¹þÀ¼æèáІÜñáMÜ*ÖBSä ,‘2€ÛW×£ä«!ë{1âþ§çn†ÚŒBa)mE‰E¬¨f :£mï3.¾.“xÏ]½i¨¹œOš˜<ÜßÙ²)/Pz¢r5?¢•¥­0¯¢Ðô¿š5ÈïÇûæ3z}l¬>·•P4¬x«E«# VPyÚÈÐÑ £ýÚsaK·Q¾Â4ž5Ë!6¢<ð®hª2'="çÕ ò1'üÌ(ºR&Âÿ÷1•ÄöS8"pÒäûâaXæ9ç\ˆ8Ѳ¾ë«Ðûéð§3àe&o„z{M&i}Aîz}*t’ò¶¯¶À<3iûå˜ 2èµÝ/m~¹éË[J•6ð!Êê-³_‡­ØÏ‡5óÅE+ôrEéÑ ”C@•„é@$ͤ·ßÀEKºè{ŸÀ» rÐQ¨7yQÿ‡Ã –:b‚o]Ê¡¤wˆ~‰5ÌJ­QÁ ©#G:X[Ÿhé;zú)fôÄzúå3)KbX¤éÌ¢KµÝÞ9ü­¼­C~bfpFE‡ å±¹‘çë)Þ_Öû;H‰ýL¯±&û|2mhæøÕ¾pP©J‚4­]tÆBO^Y&‘í„ðâ6Vˆæ¯@&*lî3ÛŠu¢¥Ste= §.é»þöó+§Zg+Œy€+”e,F¾šQ†„v«‰ÕzEfØ=™š^ðP(ÕJ+›òo0QCøTLGÌ!` ˆ‹†&ŒFÛµŸ ÂTC²Ô©˜¨.„¼)ð‘ådê:¢ZWêñg˜¼Qî8K[?û6/Ý€FþT”idVk÷ulî<­ËÁjá­mÐÐùÖja?8ÔNSúóãÈÁ-þæ¿à«ÿíΕƒÊ}dHÌ:‰ÿ7*DÜù c:åÚÖ8<È à|T4Wº4· ¡|?@þOÃcÝ·üv[©²ÎàXvq_;'Û ±Û˜¸cOž³HΞú¶&ì—ÿßHÒó/”+ݺ&B´ËŒ.ýhµHíée³üDrCm:,µ/ß  ¨Ïè&] <ºˆ[_‹æ®ïГ@‡²ŽúÊu×oÎèï¯!­1¢xefZ²Eï1”±ªTKo^ê¡G=‰¢(`Û =5K 5ø8Ýq<¶šÔÍ÷[_’»·_±¾në­ì\Ô¤W8³õê/Ndo¢†³Ð2éDð+3nÒ˜3Ã_<ähà-ªÔö&S€Qºqû¾:rEEóô*\—ij0ôŠGÝ%’ãÙÛŽŽÉüÊh–"%È­÷9–kS´°bßr÷[4ðH¦í¬J¤¢.Ÿõ§¨l$¾$9ÿ6o/ /VÁ¡ñ7vv»Î<,¥d¾þ¼4„›Èé “§4Dt†kñÆIb‹j…Ûòe‰Æ½Ì® Ê ÿ?I[I-v­Y”9+š÷d.Œ ?$û-ók ÝAŠ»ßÉhÖ-0ÃF·ÿIÉ8æcûö ¤¹MëòíS©=G¤ƒ~O±É3Ãv¬ÿBøòg¹¼WÖµÎç«,½êZ~ކ'ÊI¢*HÉÔŸ’\§W%Á²~Z:å…Ü ÿ3ŠfÓÎ]óPkKyç&ÛÑ:£ŠáþÚ|É(Pö¶›vcø°¥âÿÞé>¥p@wné©îQw«ÖÏp¶¦©ŒZêaHQ§#Æ¥wýìúçv´lN›pºbVo„'ƒx4þÎ7p;¬ê-˜'}•ì¯?oÿ\é|î r±lð HseÔ`g`ØB/šÁõ.gä¬×‚,ÝîŸaìû†c"AH1f££oLbfMö é9’5þç/¢‡¥«‹Ô8Ï6ÙAûïDб¾R_yÄËÌ3'ä/:ýNr Gï3F&ÜxÝC—Ÿuá¿üň÷lTš"¼×á¼]opô|Ò`AwÓš:^딲Μ_?kh`!ÆqVÙ/…ˆ”>zãƒ,çû=ÅÌ«†Él庬¡ÁU¹¸îûo˜ Ñ¾Ý Úpà³óæ^¯ì²ÃÙ¸£u¬àÔv=#uô™Nó6|Wè®ðÃ:X@5 Eic*{©s“$tÖ÷ãâ)–1¹q)ë&Ùå†ÀÀ¶PXbÌEyäÂîyÅÍššŒ¸¤À>–Rï7RâdóÚÿþ¹ä(øÊšJÊðÑHñiàÃ(ÃÄÛ¿ˆ<Šó:²H}ÆÂz1f.÷F6 'ÚHO¹`ZˆìÓU"Î.Ú˘y™Œ˜? Ðv”[t`$9¤i4K}†ÛlXŒ¡Üz±PDbg2îœc—lkAªy07FÔ­ˆ¯±œ³Ûjå ²Uú×ozè>Ñ€ôÇr{èz/^ OgöîÞ,ÖâM!‘EP?S7ªL2E—õɤzCîÂ휠1ÔÔKÉ›ÿ1²p ÃÎCÒ’i€FüóƒÒ± 'Å-ëžv«Lß\eçåu~µÑ\ae×bE5BžjQv#ðšl¯fó¯Xã«Æ%´ƃñ\u1>Î{ rZ·Ò´ç¸Ü²3æ˜É›Ý;cxæ@©äÕN2ÏÀ`/ä+9ÀmÙ‹ ‚/+‘Äé<Çî²r‘x™ÓH Çö>ÿDºÄG´˜áHZ5ì°”étIáñµ_g›øKL³ñ›O¯÷÷´“…üû(:ã,¡‹yjNûµÝ†Ÿñ<J½Ç }†]0¾ŠM¾Y _¸Ž\UBk#Öeñs<•Ì+Š* Di5†Š@§·ï[ù´ìðjÖHHj| >–:nàN§[é{êŠÛQŒ&gæ {ÝZ¡  &àáŸ-+¿Ž)ÂEWñerݵŸãQVšW¹3î; ̬„‡ú¼ünYD9Ùa'¬¹{¡lÁIÑJJÀ -) ó`;Æ'þ†ùx‚üñ}6Ô£ïú¸íbµ$*xÇ4çõ¼Æ„jM¹èsãÎ=#¹¹ë·5uzëó7;ÿ#7Ý@™‘—ÿ¶}7v¦ÑDÂ#ÿ”"B„Êä6]ÞÓ^8“*ùöè¼A¹UŽR8ç<37Ñ¥.æâw~NQÉÌô²EÆ×¡Âç,•å‘ðì\lòù‚—*ULb{‰G®/|ílàí^« lÍ©XPŸÏ×øoSH= VÌóqæCÚŒ7ÜíæÙ~ôPŒ¼é$^…™0V=…ئ”=y~}jë•wFÞ9 Ûé¢l ¦ãJtοêÎ w`ß3FK-鿎JŠ#nb(G#L,ªU#A3na`Vùö½ ü%Š«±¾3\N©LÖˆŒö¿Y—ûÓIÕÆ1{ªŸR¬]£„ñXrmnÃöY©Uè¬Ðj¸µ³Eh' vwn«ÓƒëKð“ ³0Âg@s¬Zw`.^ñûí¼žsç2åV¦ƒ ¶Cj‰ápŽ›,oQ.½uà%œtôM»ËˆÀ\1ßÀq sÌ¢¬†ÆW8Í/ ‡ô¤ž\åŸË¥Mð!mD—ä'ó#Þݬ`Dì ÏŸcÝÀj÷8:a§†³ò4ò¹‘aÎBbCêge«VþšÓÊ}² äwë/ ’t0Ð*#1µ½¡;³‚)U}_/ú2Åã7ÚœÒuã=,E]ʽ÷£*e;/²[IZ¦«’ö¬Ãî…ìК/Ë“o¢4‚E0Ô³ì˜-¸¼cÑå|îÌË×Â1È€TF¦|ê,SÇ`ïn™@­—¯ u­OàÍ ¤”š9¡œž>› d¬PŠõÒACĆó‘ÒžšˆÒvžÅZÞ×¥ßG>u< & ÏqßI8ìç[]¾îE|äp¸¨i¡’ c"‹l{˜Óß ¨|2sÉCB8dSøq&/Îw $N°jÕÎéd{ç™h) ìï‹)ºÁ=þà7z_xêâBhó„Äc#¿îYŠ·œ—äÛÌG^-Æ›ÛËS[ ó%5ià4S„%"šºFë RdívË}¤Ï´´UWËþæH+?jÅŒÒKrM¾5«aôë®ô7Å= Näì×\êV?ÖEàž&ÜM½ªÓÔ®Ý(~æM½?¢Mé8_ù@i|=m‡€j³Ì5ô¿*æì‡zR0^Ñ+ 1ôA¤öl¥žSiž2ˆ¦näTÏÙ¨<Ÿa(²P[³}¸ ÜKRlžŽ Že\nŽñ¹¬N÷[&ë—GrõÔËhDy°]_l]mÍѸŒŒ¸p_zȪŸ±pg=Õœo2D…wí/9ÀŠö.³í%­ºd–† íõànµ“]ìNYÖäðÿ‚ õ.ô6möæ×öRÙŸÉÌY„­PÞ/Ï)ˆÙ‡† ·6²ãÁ…‚>6JZìj9ÉÃñPmÛã›Rªc–Q¹ÿÃÒR¡ÃÖ[ó0~É>’ôŸ6[të´?ôœeî¼HϹˆ:Ò¢FtjÞK­¯A :朸ÌùjF¿†àJ—‡t­»¯8O n[Ø®æ>f0øè´>†5v]¼vP kW°*÷á®P½AÙ©%™5æÖ:™ì±Ž‹+ãíÿ—PƒÔÞ úO=z ¬ b ô"’KW¬&m±Ô y 7lHšT¶y#4 N]öa$f¦ ͺ1<øÁœyºH21@‡ÓJ1Úî.8œCYˆ{#ÿͯÆñŠO¾Ù¨,=¼K$Sx2½1st¿6¹°q˜Û'Éïß:kOï¯'>ÌÉÇ6 0Ši¥f!@v¼hý“è©XÆø3×BƒFfj¹L|3•äNfukÙHÐûB=âZ¬ú¼®Q+žóaKSé«Ñ?€›žn!°cBPŒŸn]õC1Äöï!˜mÜ\௛¯£úÍ£{îýÍ Æê¶a‰( ×gJ])¥¢ËsÁc7MºÅz)ã.ÕmͰ¡>Û5Â4. ²ßóº‡z]R€)pb<ò´6U³c8ÐÆ #‚ÆôÞ4ÁÄð1ùczµÒQÓÐHl{´ó…”CS1~âáãú·pÊœý/Ð…S®*à$ϰ.Tt\›1§º.‘ßøüð[žýajhÝwõ"Š}F+d¨MRʨ°:;Ü ¯P>Üyß"¸XÌJ/ÙÕ¼Ñü¤ó©9ÂS­p gŸ¯Ï¡éHÖ©·é¤¥5mÐû!£þÑR<´zŠ!©Æ} ÊE6ÔÙ|Ê2ôäÈE#òª/ž6ÌfAæù6RPµkWq6 ð(¨òvËùgÂÊ;kÈŸÞi!V§t͆׺ƒe¸k;‚O›âÈô†PÌéóÝ;¾-rm¦Gº Íâ«C «Jˆ‹í·ƒ,8~ƒcãÙ ÚQ:y—ÒÒD†ÜÊZvaqõ6Àƒ¼€dÈú3L¿<es(Ò€(¿÷¾¨•¦›Ïð÷8d­N×…žÃfÏ¥ qjÞ•!¾¦¸C Í¢@x«ÊÛÊ.³òã%9Úi™é¾ÏRRŸ•ÔÁýï»Âxi×Å’ßà <=VÖc1’ÄÕ÷ñ. V鉫’vbat…Š)k_â«ÿ3:¼bI¶ŽVˆþD.-¨Üy×XlÁ¼EŒÛ`²zIP­’ˆ3 w‡–¿†‰@xbX³«UØXÈ}Ïxæ›rxéV\®ÃéÔ©¾Ü‚þ‰Rñ]ó€/û[ðY¢§$(…ð/q%ôgŠx{8ò?¿ë²á-â¸#DTûäÇž-B'‚×@³y„?~A\®~û%£ôBŽÀ%¢æÐB”ÇùI)lþÙ¢`¥É ¶›ª´jK¯U•eĵÖòLäpwöJFòF‘D@ê¿ðlpÌì""Ïx@/IÍ,ÓÏf[¥n(Õm ým« ݷ줲»¸h¾{–c_- ²¸TH'ìB,_”ãs¢CéÂ*À^ÑÚ;MJÚ‡nëWq õ OU­§dAÓ1ü?wÕ^É7ì¥À’GùËê\û3]›µy¹(üW…ø-ͳl¨ü¯ôbšôûLÝ#q]Sµ²½£Ðžh=ç«·Ôàôk°ã[ƒ&iš"7O&Â@_ÁYGº¬Z–˜ÅfH\xÚ’“(ãv &°£;4 âñ[ñYZuÎõšR+%ö{öWSöz¥Ä£Ð3¤áºZéÅɳåªP’ÝÞ­) J¡¼Ø›†È—x¾ÛS¼õó0)/i+Æ4>ðh Sg_Ðr NßäÿÄ!hÖ›pd?“Ls`„þSÕ’Û›N•VÏ£åùóÙ\Êc?ùª~§a[ b²fKtb-W§(“=h·_nGcýÂ`çöœ@ž¶2R£“¯¬âŒOÙD7•÷½3ì­£óúž3ȬuŠº%àVvcÔ¼C›;ÇA Âiyn îŒÔD0³y( (¸c•‹Žø›,ZšéÞ@¶iÁåVÓì¾Û^¨6O—qyUÆR0øé%/µ+uZÉ¥ÁÄpãøçÒuúÞÔ: ÌÛ Ýõ¼"ë˜Ò†Ê23ê fÕêr½ Y. 7¢Z †My"ã4=Õú(÷§v¹å3©'¸š¿aDÃ%!âóÃÐY–ÈŒ@äØÕ7Èaˆ…ðüJä³ÕŒ0K\ŸÁÝ*GK©ÝnÔåû„ „)‹(ÉÖAŽ*{ªÉL.œ)z-…8þ)ÒÏ á'ÿ´2%!kÜ.ÈPó­Úåáï†5SæN;˜‡¯íŒ™êÚ¼Õ7=x­Wëu_”¨ù,¿™J˜ …ÿú¹Žïെ±çžŽÓLŸÖaL”êåg¦dÉ žPHÌÒ´äod–šH—7…)->ù[uäëè[¤úÍüvzÑÚÐn` U,ùôŒÙ™&/Øs7!XÜk±Üµ6BÈȇÊŽ<°iïŽï…)Ã-Ùžë IЬçm%\ेé¾í¹ƒƒ:p6äÃÈOé h1¬_šZAáàéÖ,Ï:M¶*„Ç0Þ»Ò…N±1Ï&¼¤!¤m2ºEðµ‘hÏ|@»ÇºgX+ï!]/ŽánÆnÌîw³…åQ%Nœ=–›CfþSI½É]â úÔ+V‡#¨6€/÷ qõ¹:S"墨9²SpãO‡=]áVp±dK5ˆæqž‡rÛ1à`‰Šq"M&2wæBo0Ž$Ç ³” W*ž¸LAD '{Æù|òžØdcøa=FÈVÏçW£ªgÖÇÏcw]üi> 3?49Ýc4Ûyãüƒ(ûÓò}-—v­ýR5à³·¾ïL"¤@ËèbñÊÈ~Uú×Áã´}~ñÒeEi&¬&NÆN³Tê%Õı[~Yc“CSФ͹Y°ù3obB¢É/ƒä·¨`e8©@åýYI6VFª/¨÷ÁÝ>àçLê‹.U£¢fÏA­IÜm?ßÏHT\ék]Ÿ?ð´Ìyïþ¨çZéDX¥rþÿ²9—\?õë @‚ªÆÞX>¢ ì”0Gmq¾ß¿íønu,?M&wÛ¿z×¼†®Ñ¶K‚Z·¹Ë`Íͽ û…-05V†Bp'­ù†A³*Ã-Ä›O§È±8[údþŽË³T _*²66ÆÚÙ®ßÔ÷7 éÂ}ÅWh×ͺ´bÔ ªÖø4žM審©¼DìŒÉgý¨>ã K]Z®W>ÊqÖGD´·Œ­úM´ Œ˜î T!ÄÓQ› ¡©Ô©ðC8³BíVk)›ª¬rDI¥v¬é …†ïÔˆ¾_ ñ?­ïu¢/½9³Íó5ñ )•­[G!’|@Y”´Þ]‚XPŽüµ§M³iܦÙX¶À»·Ë_1ŸÊ©ýJñ¦¤¸bokþ–ÙÀÃ'‰¸0VU«n·Bö-,÷%M¹ÁÛ7/¯ñ K :Ã!w ­‰‘ùÙîå8Zwj‡K3%ºtœ\~/ªŒoûE¥ž_‹c¦lPÅ£ÓV: ã*\«ò%;XVÔÄÌ|²gÙÑ̤‰kš6Êf‹lÊ´û¤S럶-žà^Âýmý܆ü¤ôœœHná‹äÉê¡û6ŒçºèöûÉ"úÖJ‘<¶SäÑyó„61LØ_‘jÿp›– 7Ûq¡^j³Ü1á3/íXY´ 3 ?ö[qÞæ wƒ™§i›C >Lͦæ1ô^–á"Â@Ò”æËw ~œö¦ˆˆPžì–TÈA:ÓÊA«Fì¥'‹Rh»>Űõ‰M2˜ýùöp ˜ò—§®9B`hÊïäè\J ˆUN³¸ sé+nQžÇrcºÌþÒÒðÜÈŽ‚Ÿ Š¡R u²DbÕ×`ûÍ››œSu ¾ú]]ñ³”¢kïä­9ßVL®ܵөýd ÏþBNK‚;ƒ €¤UL¥x)‘ð9¶§k2?cÒºìo$D^hÙ-±Š€0Á уäà,‰Éf†_~ÖŸèÆ±>kœ˜ìÁÚuáO÷÷ ®¡¹è0¬*ªÜÐ#?¬HTT6ÛÅB¨ã`Hï.›óYHG€ø7Ÿ¢%5ë8,:Ò…ÝÜŠ‘+r?·Âô7’®òq£_ïdS‚¡òÊ`ñ*ï?Ó&‰óåü[°ÇÝ~SµEs¯n$ÑÙ†Ç._}ƒKÅÁ­&3¶ÜrL÷’èoK ¡‹s(q´„Ú_+6#„ §˜Ã#ù¥'Ŷ¯Øéùs·Fäÿ%%°5¤¡ ˆs¹€/¥kW›ÃGÈ[­Y‘üösUô®Ê–_þþ&ÚÕzŸ›#û’ÃÓŒòŒ »a ”=¾‹ÁþŽ8”:¬ù¿šÿŠÎ¸â¶{ˬçW†zr(‹Y.„êNo'z±êÑm?1‡TC°(ôòû;µ$uß¼‘r ÐP…<^É=O«Ö²[ Ä‹ÿpS» ª€˜V«2 DÚ*|s:2éÜy”íÎÒgî»*—l1j;ÆSxÿ?ë RK¥¶3=ýNôÙûÙÈBj/Yá¿ýÛDzÆÎ4‘-~þqÁc¶\Ðx;y Ë篻7W×Ôìu‰þúákgù!” `U0s¼ð[Ë)k<jXãœ; Ó–F–(Ý©)&ì™Ì)â»0Kž;êwy©8°zƒÀ³¥¢:ÿ (Oä^¤Q;1níÌÆ»¹œ„ùý¾!Iò¥–r¡x‡Z|%Þ™à‘ ó‹c~ªþ7}:Þ•ná¹Å«Ñ¨Øò%Àmt {a€L”ÁËGœVŸ&¦[¬§n:¯küRöb×ΆÆz`Q•Uhœ³Î´Ì¤ÌŽj»†jÐ*OƘ‡4*uaT·`R„³}€}ÐŒ?KtH/Xfü@¶˜¸}íŠÐ¬ˆ^Å#4Ôw€¨¡8ñ”¼~dˆ&¾òªúÈðÆÜæÕ›xÚøã禙pבÜ/c³åuîÅÔl˦4=Oc¤Õ¢ÏÅ\.·€ ‡†8ðé o®ùª7JêpJàšè• Ý/Ó+)‰OY3—`€’C`šÿª6lï¤//)…æÉä½èå£MØž#·9Ýi{}L^)ÇJ)ýÙX œòÉÊ#Zª!— Q1ÝŒ8  ç?«|豫õ^ëS3¢ôUAN“üг½°Œ¶×{d¦têΆð·Dˆé GÃ-çøú|ÓÙCQºæE¸Ã?õw*m7MMÜ—{5ÕÂЧwÆð÷6h$°ÿNÒœŸµb]ðŸY>5[cìxóöŒð»ŽeA( í'Dw7¤”¨øt~ow9u%ërÜã¦á¡oLyZF'(»!1#ºh–É B˜Š#8.WŸƒž4GC©ÉŠ–¡˜”¦ -i¥8-oÀð É6Ë¢2ΓÞdÈ%â Æ±`òÃæa0™8ìTR™ï,M‹"º¼·›lÑà­÷‚’-N{ïøKt<àÂñ5^“WÈØ3b´V˜L0‹ËsÀóÓxí!"¡°XIGf–X«D«pCCñG2W/éb`—Þ÷exH,È]fC·¨  ä6ÆßOé%ujh6 8%7‘¸Yç¡ð áäÊûŽ%Zͺúcõ¢›o#5…MS 4DBw¶á@g3ÓÕÞé„¡m "L`cí¶“©¤ »˜¿m‰ãIkª+Ð{Ù°$Ú1Ê«“`8 ¸ß0Xgß9¿WƒÒM˕ݞHéä1PÝø [¢yºJS·ŒN ‚¤iúo–fôƒý€“³ª¹ñ?rŒï6"ìÀÂY„î¯tÀ¨ŒÿWU<?Åñ?<ñMýäæ=¾|`5,êº Ý0þ–öA¸ˆ|ó3°7K4ôC,/R>ÇÁiÂó;­[¾†åò“!Èx•µÏÃ#ép{'ÞâÕ–T(ñ¹»‚«þ½.•bTºr¡ ØÁ²Ü”cêm`ÌÓtÙ!{ŒÞÚËï.Ì+Ç7³ÊÝJåås›p´é6|N2÷*fþËË'm°Y!XæŒ ‰e3F¹¹îŸÿ²Â³Öì#ŸÇ|[@Kz%Ú1øWVFç´Vý+Lð=÷©ÒLùƒAϤø²ÿ¸\Iç ‹¡“(o YÁÛ¨šy­¤ +}ÿ\È*Xa´„ÑÂáhB,&‰À(ìrQœœ|þµg~pr R[Ù–Ü…—¶ø:'(¶‡C{‘³-®å59ÛŒËÉ %ºwânq?ÍN0;©jEálõÖdYÕ² ߇7Àè 5aw©(gª¦¯²¡{ûýUìm¹´ðÞ‹ °ÀÛÔPä±°MõºïšÀÖ µ}¤ÕžÕ×ÿ]J$É¢°Œ):l OÆf³f£©½èÒ^U?cY¢L˜/‰4Eá|…мœ"¬Ìÿüþ•øX—l`ý¯êWÃQ!I7Tö×›°Aì‡.89 ¨‹#á;«:Ûà’0ÈìH™FŠWM_ê&Þ'ÕÄó|òΧ­âÖa: ŒÚ·C­¸T¶Íü7±Õ.$í΀¸] Ä1‚“ZÜl†HÉÓ ƒçø²d„I;Õ´ŠI>ÿyO^àŽ) )éU¦¸tAæ ²mŸÔ‘²^&Œ ‹)K¾‰ubÚ½û¬‘2 ôvZM”zŸkÓãÖ´÷sÍêÛU*æð-ÿt§É¶©å”cž+Nm_Ý JH¼¼MÐp¬Î©gBÂtØ…&­†ƒ×Áý影ʶ5„åùiý!ê«ðAÙŸ³Rá–Ú:xÑŽ²™‚6 VHL}í£(]¯HÙŽ¬8ºeè ÿÚ²ZIæ¤ì‰Ý«¹ái2Ò¾~»®Ic}ùϳ—Àm:ÌOi*–®7·7ü II¢è-†N“55­(u²k5OhJ2+1ܹŖ¼Œ%Óß\.«Î…8>Ét6Þ=üùq+½z%}-ím¤Œ+*ƒb˜ jqïy¹¦gÐ.mWO¿ (TVÈšú}|ô.õ ¼è? $jñ!æí? ²%x_iSDìH†ÑÑ¡«€/¡ÂHÃñjÏ­;«c’wLåÌAz‘?=†]ÙG³;"c¼'çÒSùðjäP¯/NÑÐmOØ«ƒRN$îÿ§À¬»&ý$Ê^y.ùÔæèÙ"0¤[!;4ζeÏKÖ¬ñòïŒÎ’syþqBWûJ 4Ùç9%jÍ£†«ÌÙÚ éŽڅ4_ñ³áÎNm MdaÈ Ceúô©â¾/¼Þ[•iÛZ²ç’ª­AêŽ4¿ÄS'š¹|Ú…©4«톅"Z€=-Áp(¦øõMKGzŸWj0 `až“)‘]&ÈÄ®¡ôöñq7 ŸßÎ-`Tµjø›´ÕcUP¼¸ôÞZø Ûí°óÆ—z:]i=àÀì=Ÿ}~D0æ[§^Ü.¢Ròtë¯Å´g „$;oI^ßÅ•†Ë V¶J²¼8P×íôVáTIO; …3,÷ÿ?ƨí÷г‚þ"ôتÚYòó5úÊǯg³€Ó†AWS«Þu [“1=õïº{‚ 9‹}qäˆÀË¢›Ĕެ)O>ü„yHÏ·z®)Á‚Ø‚‰Å~saš>Mu?å>Qb bR3FÊ”â+ZS¦×m-ÿðyÒÔ†A\¹4ÉÐ/ ˆP÷m$¼ñ„"Wcéuë œ ]Z©2{»jK½‰êÿI½9 IæDÉôÀ.ÌNˆ.É#[º»1@¨¯IŠûÐE²ºƒÈ)§VÇ$[b÷m1:¯E¿²röôßì H±‰öÕ.$ô¨–ž7ü“×K¡@t䊽 <€{H«2‡ÀÏóÿ¬÷‘“o’`fŽ}\l6:ÖO¹6ÎYC3àîT1Ý›šÍ$. 7“&Ó òš½$⹞PPK'A"Ù Y—NLŠM+vªTv3y’}»V<ÿPš’¯­“[|×§÷è2©Á’Gžâ-ŸÿSÔ‡ÚnŽd>üœ2ŠåJgñNÐ}ˆ1^aÆ‹­uˆŸXŒÐ[èj^¬^Æ „X1ËtêЭöªNÎ$8v‰›Õ,ˆ‘[s‡ 甓WŒ¼AA»”x"‰³MÄ H5}œd—G†T‹éé¬I·OЀœY’öo]è†ÉFÖ7Tr@MçäSUÃÊ æ×þåÝÌ]é8ä|ÕAî“(áÀŽ9>(þ“»]Ít €7%Aù»]+…3.*]õõ:ƒw«s!ò[d³HáýLtï±ÅÛÃ3q;nZæ`pûе+2ãâ&U§»A»¼v¤é CM ½DÕ«[–;À™…¶”®á–¥æm`(&çVN4$ %ò!|Ï9$¬€¹£%QV^pèNÑûØ3|Q€½D\æ¯ýÕ;7eÚüs3ê® R•йճ I©ž9ä‘‚ Ÿ&V˜u«ûÂLt:擌ghއž)!ÊØjnªüŸ¥}ò…bq§_×ÂÜB9„ ÷Qš”°HŒŠÿ­íÂYÞ[*d@^&úOHY‡úy ó°LÖMm×VIߨ6¿wößëR”7“·nkK£‘TgQ jð5¦—Pé· µj“ åûNœÌá’MiU7´ÀS58ZKRf‰á©2x?ß–K¢mæ)å[+ÝÎL±Ã’^ÚÕ˜žL7mØ›xº= ª«Ö÷”“fB2Rœ5íp'<Åܺ8³(%Âay¡Âp¦ZÈy>È«÷Ò£‰Â[€‚1¾È\AÀôk]RšvNý6€}NQÅ8Ê0,r`§‚„˜ö1­ÌÂËC„Ü.IÛrnãŸÿm?áϺ÷ôJúfÇc ZŸvc‚¹â\éš dåãì–[ÝÜOf@-Ò-U÷CªCÖ+;/àP€ñ]B°1¤–Ly(ÆsÚœ%)ê†Â¥ÿÿvˆ.CvFúcðÁCÊWCª‘M£D¥ºªSnñÑU—ÿvG=fCâJoPv вˆ®„ÌMÜ~ÉÞ0ãuËï_€*í¹ý'OewÞŒ pfEÄdÀg«â‰TQrct@£‹MAñ©<ý]žu§„Ç›Ýâ’ä½–§Wx[óBª/žxË}|$Ššàç²;'2|‚$®¨üC0é–y©s°”1àæT¾xÈ?²°&àͱ[œ åt÷]Îk'¿+oª©âM/ìÜ)ЀOèâà©â#Б©ŽX:P*˜ê¤WpÔWÌ< ŒÀ–eôHtÐÈ™ŠÁ‘3Ö—/³ÍŒÂÏqè$&ªbëg ÜjkŠ‘c ÿ5ÎC7ÎcííMÕ»²»‚_íìÛVÏùXYiv§ŽÊîT9Ç.FZÞ’wŸ`5ÕkðQ€ê Mƒyûgs¹w‡ôñ Ê5ž ߉@'Vò¨|ÄJÉo+nÄ0ÉÜIK‚È<´>Ò†|3% |••MW)(6[­Êf¨¦MÆ_eú½´hFùS|ÿsÿYŠ<“Îi•ÿ •YÏ´"ùEwù ßá¡Þ •n°xUÐê¢kíÇT6C#F£—F“–AeH<¹HT«]I¾OFrÈuÆ›ò}ò…\0o¥µ]„"+ŸY»‡‡ŒcÆÎððÇXÚÅÛ\r¡á»€B°{Ý% CìBípCŸÄß|Kìk©kÓ±°Çó9Kpœ†ñÙ´¿÷àΔ:6ÿ/W™æ}…~¥¹ŽBü?Åæ«’\Küë› ’¨Ÿò$Éã­Qv±ùŒÎ +Ñù&e=(¬+·47­Ý;‰~ªÏÇ”„^€ hÝôHÈÆd‡'Éž#N2y„ ,/ê Ñiê¹ÄéwºEf BeÜ"ͦ=",¹Ï\1â¡ ç6²¾i€ƒä‚¡ìõì… IᑳP¸ ÞÃ4ƒKùZBLítf¹«ÇGetQößLNj×ëL,œåIÂý\óÜŠà£3¶Cy±äaõ0IҞ݌´NÏ<ñÕ1ˆ½áOq„˜‚«éslû¨Ê—³ÔÌØD¸ª, ^Õ`ÎÜ©8A.z[$°'Rš<Óøb“ ñ5ÚníçÆ¨êÊÊY4‚«±ùqª»~¦fDüd¾¤Å$i¿S–œµ$Œâ|`Mê§©QÖ¢¨ï¡€i:ᮬÌ0Pd•ª ŠþÍPÄ+; §ªdÚ^:‚)ý£Æ<©RõkŽö_LÆ_k ¿„ͦîr’–5>-É Eà¦ÅrŽIó)w¸€j–9’ÈŽÕ]M#áPg’îò³€ [?aKž³Y ¬§ýa¡Ç¿àª/Ö2t´SÜb\€#.fËÅ!ijªO]E«§aÞYÍ'\4ôÿ<¶€ÕV­ IâáÞ0ÐMq™åÙrvÙøò †ëÉÉ"–>;×t \ZÄoßý:w?ࣾ.ÜrZ Wv-R@ÔWYÎ[(—JÊÑðûahG¢â ¥ŽÍ,°qÜ¿®Êó¬p›¯6v5¸Â„Z·ËdñìØâÞ^àö~(3Tàܼœ¢¹îЭ ‰3½Í €³„#c2ïjÿAlŠî‚º)ªÑ?\Ê+ZdÓa>dØñŸåQñ+WÈšnÙÕoce³(0ߨkµžÃ¹)èvÒæ¼ƒõáú#éè¢"D«©pˆˆd×geì°dî5ͽ¿ í•(0˜º)ŒÑÒ<Ý3R¦E‹H5xéÛù¢d+þÃÔÖS-l ¿‡)Õi[—+šTalΞS=|DHIR?ŸÝ^·—¢«1Ãrîõaœ…]q0¸µˆ? mN}>eŠÜ+o7ÒëjP¨x({t–â@X¿aU0ÂÀÉþG#aÞX…}7­,+‘J0ƒó¿RǵӂÃD»ÿÔýäþÉãoJA×Ч‘"—?^3>6I§>#­›ÑµzEÖãî"IZ© Å¿è.Й¾Y·Ñ‰øôÒ>†¤T¶ ž°¢ èïŸg2*‡‰/Ø“`ù¨ÚovTÖBý¹\Õ-c·pEòÝkp3)î_up`f\õh7de*N=ödé­PJ&JøG'/–ƒu59 æåòLZWE ¸òœˆä=c-,&õ•ó§_·ë+øŒº#cñ˜ô„µÝ¨9\úkjŸ~²ñ4»OnRà—9¤Ò<âò·¢9ñ¦K$Yšˆi=8Wv4«Ð£A56=pÖ Þv„‚€iñгýíFOÙ_n½ Òî”põ¡'j 2®žŠ`ȸúy,k-S­¹ÚÿVFí1 Q’Æg%'ë–˜kµÛ/MŸ?1ÔÞ°XWÞÄÞBÌ——\Ϥ d4›Üàôß·ˆä{š×¼Õ±uúŽS×ç¬þ$["Ò`æç° ‰4<±™ŸÏ?“Öj8øµ=0º™;P"Z,½Çw ¥së:Mx›êt¬%Œ*uÇá%Võ`ÝÇ´ÊÍßä¾CL‰‹uˆí!%^8»#vÛZK‡ùçT0’ÞoŠš÷ß.)ŒÏÒ5¯Ò‚å{ôoN$ß:L3?^?¥øÉ?~îÔ¢º†½»]m¬…þŽÅR©g[®„±aEõÅã"aLyÐmU|ÁPäÖ“ ïÉÎ:ÚžÕçpd‘ʾ%ãv³Ön•VÒ¸4}ì%x]M‘~îS*·Pöšg"aâ’߬¡ªìÉ2}Ì’¸SiÊ8Úb³†̰êœEVÑïžËº.„ó"aæ¢þ¾myÅáy7ñÔl¯ .‹”úQPU_Ƥyiàq>‹ |±\:ÊaV¨r8ú¬­$ÙÐBMuª?ƒ’¹ŒåÕÊ!Ÿ_&¿@Ëoµ;(X/XˆÛv)nC9$ˆ¤Å½¹N±»þ‰Þr¢!eúÌjåmÃÒ§¼öy¼Ü޶ìA…Òø24qC«XA\©ÜnØãFN+ì€Ho¸¢ Põ'7°À¤Û“Ó*²å4ÿÊ Û[ =ùÓN¿Òõì 0™þºGÝà{ µ½k$^†h/[ /Ü&¤mç¯ôÑx.:.Ș‡ï”¡ÄLýŠõäŠ ¥,´gF˜- ýÀNlÀþoi¹Û`(ªÿ'‡ B—†Ž‘5;bo£DÙÉÆØÒ—àŸ3`þËû›Z>z¤ÁãW@u¯ ->l¼qõE’uÖë<â'‘Lm•:ƒOTEÚZ¥M=Ìÿ³KÄ߀ÓG&g|ÌEf1y{y6]ÊaoÞqID«Ù…‡à·ü=goêžÆáýßSLÅçI)…ß;B«ì¨Èy?õ6+¦ñj‘[A1_lÈÀå̶ãe¥•*UÖÂ¥–L_ã dõ9:+«GÞØy°;¤((ó:mæt£G7It"Ó¢[4 Ó¯o®„-–‡Ñ3Š¿X¸òKÿºa,ßßþ¥ Q€¬+xþ¹GûX3yUGà RÞ×)8áÖ£°Bé+> ŸsàåÓ,)‡¨ÞCJŸ¬BI=Ǥe늾Jz{š>pêråU ªå¡ÎdÌ!X@ÿK“?'S—©Ã´ù¬ 6“>B'¯f”63—¨¢»êÉX1z8@cÈc·ì”v=…ù»h×cõcÌ ¦eI—”5@+äˆþ2Äh û%B1¼Ü4¸ß‰ü íH:Pkk”š¨O9™—¾Ç/¨ûj…x’›$¨¯zªÃé"’ßÈ¡» £6½cGÆß Ï’6Y~gd¹‰tãjÒ*%Ü<äñZ ½œhb|ñÈÒ5b‰öÝOͽbü'.ƒ.h/i²rê¬ja6À!òM'GÄeäe S4¦!² È5ò«,ùåA0EÂYÏ<Ñ)‘2¨EG{àýüfQ.[ïþ|& 1–ZæËt yÉÿrÆ ì¤É€ðÍ:éQ¥oÓxؽ(˜vËL1™/¼3ƒp¬‚£bAÝÆ(öÙ=ã@ûI|+Œçñºwl5XBˆTóŠÒÖ B¥bÛƒ×C棭åÜôÖÛQÄá» ¯|«eü´a†aÞ{¡"§—4_D×÷F»Åñ%€\¨eÃx‘£7ý•BQ_ôã´SP ‚‹Él=#!šÄV6±C QECá‰]OuZß­³X³Ñù«iY•§Nq¶ùÎŽY· ï Ð|é¬ð9t‘‰DÁ§v÷óþhx&µöø¹mÝôÛ4Ûö¨{Ï5›E îN†; _’ˆ ˜–M²UùòšÕÐ++Ô(”ìÞŠAh µ0øX5›rX‰çÆ¢ÿ„ÍÃoñh¹ÍΨ‹Å°¦ÜJû2ØC¡c÷>µ6Ò\Ç©Æáã:æ'¯Ã`³VÌÙ¦ËЖ!±çb.€¡[ŠdSµçŽÖÃ]ÿH+‘’d$]Ö§ä1²ó*÷ß³ËÖaé}Ôqª^¯Û¯ÉÒhî:PØÏ­5T/ýBÕ¥ŠÍ ´PëªiÌ$e×äehã¢m9BFJÒ-#‹©üJ~äg!gµV{a;£Ã1³¯×ÞÀ:Åí¹éÚ  &/NnÒFÞþLóXzÕ½úhiFá' ¬¯æGtHç­°ßH*KΜ.Pè|o@S_Z²zÂ_Gc¸°Û­utùþ3'­nŸ8NÑÿž³‚‹iÕ¢üý äª%C?Zk §ðzÚ…P P¼ï“m™æ «­w7½k“OP3Ûwf«¨ç‚ÝêI¢Ôþ c Þ!ú‘Ž@ñeq5¸ÝÄEü./ír÷í‹–à+áû`^ÃÌàð‹»Éë8n{1Ë„k¶W‰FÉ–«®Áöš“»ˆoøïßdäg€ #°¯"ÿY¿mClé>…Μ‚… þ‘2/Üæ&$¼I›ä‹øBR¥¡äèMïÜ~–AÇÁ,ÏÆ:b‰ y|¯BB¢%vžùyˆ§ðm?(Rs­6I}7ÀÌs.õÚ hÓ£èÕvV{×¹ˆ”ªMR™†d¨¾9zƃwa0|ì§1ðj¶Užú3r`--Ÿ—ØÐ#W”}Š¿ÆGQú‹¦Ë½O Qû­p¨Zœ"f/K GßÕCŒ5˜#èܾƸyÙ£\çZ9BÓgް$ðEõü0†ïj/‚c¤ò.›ª¥Ž_I…‘Ê3}R+Ö^]#G£õš0k=ì±ú ÌËj\…‚‘! ?òƒ’ïXój¶g NßLÙJ¸Ò“¯-'Hêo Ëæ~ÉÜöàšM³8"±¤ž]j曺=‚pÄÅŸS¯xÿ}'aMc ‹Ã7ƒ±k`úw°_x’°¸Z%é]hÖ(ÒUÜ’Ð2Þ‘ö&"ÂRûlB•«½¢¤å¥ç¶cezQ.¡}í’ÿ剛~W]l¢tWVqT>¦â½tìiJ ºÀÏû kY î+ì„D5 £cÁŽ{¦ëUÉ®Èe©±Â“-=ùI‰˜>{^±#Û5#§hRt4àº+à%&8£Ršy›–P2@;>yaÕ\ˆË ÿ£ò‹ælȶ`"‡¦†‘¿7\hášç½u0fPáß:ûi$ò™CÄ ˜”ª+sÁÁ•¡dSz„)T=prïcUÎMÚ‹åSã•e²²‰ÍÍÕŒ¢¼°•†ê´hƒíc#Yl!a¿ÔLJ=Úµú1ÈÅÓxŒSŽ•:Ȧ<{0 ã®þ?€ÒÝ’Ëì.é9ë¸ÈM€}WÅÚ{—O¿=acE Œ×ÏÇ,Žy;@lîjˆÜ>CSÅÈj}Xa `ª¥0ÝìHE†&‰·~òAÙ-BþR´gù{挆D¶ÙewǹÕ”9v »þ‹S9d¨;Ù”+˜Œ•›¥«4Y *K"Ì#×áÀÑæ8ìÃÉeñ€?SÄôj­$üß' 3ùy~Së&ÁÇÚÞº@ù­ÆøU"ÑÀ”ÙµôzZº-5=¦tì¯|‹è¾¬¤dáz’ 0SKr„„ÿpê±î1eT- ÈmwÕ“_eJ$~ÈžÂLôаW3È‹ó†Ñ ·BsƒýËC¹%éÿãOé]'@–@°2&h±NÐ|ˆ¸t唘$ð¹óè§ìõ÷f¨G´fŽì¥ (¼É5dMúïoS8ÜAË„Q©{Òè‘t|ÆáƒÓë;F;1-´ÓÇÐî°âb@!Æ–êC[´¤d?Ÿ"0ÅËk$ÙÛhG)¨Å¼eè,Øòœ\œ…îË‚ØÞk«¼Î#‹UV²ISÏFS<%‘œ V𠱈Ý¡¥7ã^Aó<Ëò}‚97Zc4kß½å P©ðxáÓ×ÌÉ5ñ~Öê`Bí½<zûËØiÈýÚÉÜC-@3 ÌÈ[ó5õ;‚¾îÛª„¹bœ°‘¥ Q]}¾ë±€—ü /&ª³ñ[ð¶Áâ÷ÇÜÐ×ÿï¡—EÉL\!̘ö…Õ`3*ˆ›]ÔËöÝànoii¾µ ü¸‡X2í ÐþüL[Ô¾~¿Ë0 ¾›Œd=&¶ Ó­E©Bkùvã«Ð¸ e¯Ø$Dz†ÎZÖ¤RHtc Íç…žwÆõc¯E‹îÆ„Ài÷têö\Z7´nò áÃñí¶RÚxSà"°¦rÆ´ ¸8{-ù`¸A ü·›=‹œF}"¨yËŽ R\g3Xîë¨AHUÏV®Ï(£/™¬AùælBGJÿñ7ŽzÀ.u,•¾1· y‚«Q¹.œ ‘ þ¾tbóe±˜„è‡o{ 5®+µOu³²þ½w1)’÷Ù.Õ4ø5’¥Þ ’t‚J·­êÞUAfîŒd”föPîß]Ê 0–@â<,)”]Z¡Â/ÂÈС™n ñ·‰ÉÉ ¦$fN .Êÿku=™ú¸ÙO½Ñûõá>=bÿ=JËØ†NNûRòAL$þÑ@á©aá~Äð5}Ç’a*ZpÔÇbÆ×€*OšSzSÛÛj¥Ïµ+HøÖ¦Âó´ˆkÙ08¯8ýjS?)ò@{l~·G–.¤¸<®Ò#†©gï¶Â¦t7r;d}üÆnO8 ?6sĸ>–÷"|ðÝb´(y„K°‚íàÃk3ŒF ²b{ãÚgêy‰\ƒu©¸Ž«Wq¹Q¦ÄlCº;’dõ–uôBqòÈ“Žî£ÁŽ] ¼dÒ)ËŒë|êÂ0|¦Ý7°É_ÑÐéKóÊ­3]:ð½ûë…–Á¸× õ^&ÂÎ> v\VØ"kB„J~ë?¨g¦˜,"£_XððJÔ*Q ¥D8qE™Àc¹ôíÚ&€W–tê÷^æR…üæç‘bðþ3§"=¿t²æt8ŠWÓEBAÕ8¨’œÁ&£áPhÑ "ûZÇC½$ b]nèÕò_é@xGˆ¢dFØUAÄ>7# þÓäz ˆfÏhT“AGÆs¢Ô7OÈ6ÔÔMŒÁ”εEàà6³Ù ®9·Q"нlD+îÈUª}(Êì¶XI>A–=„ŸbÓÚØøfê4A:D@êê«44‘15Í7ù'®~·hž*—B }Ô‹ƒ†Ç£«P›pPTˆ&-\zG-âGãNàADÔâÕƒ"{Ò+JðvBåIêE¿©û~|¦Êž–Þ_üß$*År‡èuš;xœmÄ¿nÍþk>O¨­Ü®Q¹ØÓ?ϪÍN °ÐN^<(âýúó5ÕÙÑdTÉ"Á·ÕxƒÑB;5à¿}lª¸~™Õÿ'ê¿×’ÐñoZ°>™ïh“þ0å[PCœÛFýpÝš—ÿš7ýÑM/&‡ “õYÙâ'†R)$!4* x’wyÊÇ z¸¦öžÖïB¿RWdÎòR¢œZ§öOçAAQBöˆžˆOYNßkkÒVK,¬ 6 üÙ¡L0-ÑZÃ:Bíwë×ýé«#pI?¾~øÓ#Lþ_fyÌ‚•ä|n'ں͕ÑÀÏM†²Pôb(G³C›rÙ:勊Fà$¯Fövè­Ièda­.‰GÇ+ÍÈVDs3G:t(ùÚ ßû·ÇR5ŸH%DL1»R÷;ÖÁŠoA’¾ªòúŠ÷“Ÿ†¸U}g‹+:­;%»åÀBÊ5äpsàÙAâg¬³‰4¡¤àCÓ ×ÏÎ%3̪»Œâ~•)½üL¥v-ZwNÿ¾ÖÏžWµŽß}Å.Çf˜Í±óŠH#z¾xçJÑî?O­J.ìÞ•Žjyv³lSÝ«\òâ÷Ìžó¢ÍöânsAÓI½}gMc§\ßG0 ÂÒ5Íï>:Ÿp?u7¼Ø~ÇGvBM¨E¯øgò0â«©‚QÇH¬bÏxH«Ïl’DjD®IjßuØoW`„ú˜:¨„¸9qÈcq‹³ÎäŽÕP¥’¥ô‘bç5’1¤°ù¸$3¼=SUØÀK¬ÀÉäí0íð ¡›@Ä1¼œÖŒ˜Pú~ª_]&/Ÿô{ÒE³¿C^ %pC\®öÙExkÉp‡ÙI]ë›M—íÏ1¹ºº¨~õÛ@_bu¦2<‡›AÏ%•ïÛ\3ôбKÞ—[¡4Ðò°u¸Å«-èô|-1PUëiub¬¸LE4?RÀ­†í™m’Ó\ßd'™cuMî”ç$MnÆŠ)GYW˜ekÆ¥ ka¼nòqµó¹ÒˆÖŠ•a• žm†N0çS'l[ôÎó¢»OÀð6Àî¾ý=NÕ¸úÓ]bGÄ‹oæLw„À£‚’IÞ!óŽOcS®¸®Û~a6å~V°(WQ8îBôu:-°@…|w°ê üNé¨*¢ÓÆ'1Ú[p•™g•è›?IooÎèE»äÉÄã¸]Ó³¸¢x‚’o‡WÑaÏÿ.ÊME3zcˬ9ø„c¦£b”Ö*pU—Uôê+—3iïe3ƒoê=+âcÖu/&³”æg+Ì«ÕÖN’z÷«‚š~'W@낼`V !Šœ×ïcì#‡‘ÉEXžcï?£Q¹u®„Ÿ\¶ÙèGA`5†ŽïN'‡šÑ¸žóÏ1b6lÍS[[æ7àiè±v<,¨Iðm™/&Bš­|e Ç’Ï®h è8“Yµ´KàÃÿ:¦Û,ï&“ðòàšÙg-«¸~ƒ‚óÚH[ÂÜp“UàÁ"åàEYuÁÊyŠç†˜À«4ã3‘;óÀ ¨)w« _{ÖÒb²ÛVæ;êè³°g5oË‚†ûÙq-’5•vÏÅtãÞ½£pѨ˜¶Zi%¾Î.Ïþ%1â [¬ÍdÙ5>-Œ™Ä­<è…TÑ|éŽX r˜ÝœºØû¼ä*`>ž;Ng«W.ìg¬Lh£‹æPŠ!¤¨ΠVæbf¬}]|§jÉáÍóßõÚ:zš1ÐãV–·ay³¯ìĤèÔQØÓMZ£µîòQ`WׯHCíôWíèÍ!æÉ.øzì i…§µt¿Ò æ*¬1åÍ.;xõ×*]æ5Ç(HïÔû𻜻—[<é ýÒÁ]’fA6(×o_oËàÚM›nFlTù]àB¸—MCèR¾w¡ ZZkû8j­…Ç„û½9v‰W£õ`‰ î-DŽ»rêV¦Iž ^èàñ>á*ê*´bw˜¢•铞ü?LÇÜ=’ÛÝX v'üAOmÌq`S~똆•E¼&Ó0w’OÉ 2”—™[&.)É9!²å[ß|ž˜™2ªªÌsHK0ûµé—irÀóqzïØf("T?óÅYÜD\k5¡Ï&®d9 ¾" ?³ûq„Ë®8¸FHuÿ©§»†‰®üa(*ߘø‚dà[‚r­òDýŽ?sº¦¯óóËzýÓ—,Ùîn¢­/¿J­z…d𡊤`—Ù‚pJSÅjW”Tü»KzÑ-²+.¶ûsxínö ö³˜ÀpL>ºNwr|ŒÎÛ€v,ø®$‡}ÐÂ?,îÏ]å*¦ÍmT¸džÅ£¶ Ä&1}»^wg£›ˆ“íé Ã&º€]{;&¬pVäo³¥ÆÔqhzhÆtTʬâ©SÖ´L}À?F/ÃK¾ƒ>xH{P©j>-I¾(±!Faô]z”§‡Ðc1‘5ꋊ÷S¯ZÐÚevcàaDï ³_$Ãa2¼ Û€ÉÚãŒüÑiWêYm‹½Ë{_'‚@;[Qàß vEÞ¤Ä"G&ôÆ‹ÞÕ´ISúZ‰ÉíŒZ6\“5¥îŸV.¾Ì#WTËž“GH ' ˆÂ'ééœÛvžGÕm“«A¸Å—TV‘KÛLâ¯ïDs ¥`—ç_­ÅfTö‹²n˜†2NÝãðèo47î ž BN3+7&üµî­nQÂb*áÙ;È-NõE®¼¶ëHÐÑ(Ex¥ `ÍÆ1ÏQªÜ|¼qå²Ý‡’Ù¾¶¶„pCL‡Pr7pÔÛ`š‹XlÀtþ¥Õ´/Dx‚©ÒæMžšë7-)êåIê 0©f/{\¦3R€¸¾ÄžŒù“u'r>mÁDÈŽ2.dÛJ£ô­ \^⎱£|Œ˜ Êò`ô‚?ÅÄJ-ÛTÏî¸ œj*C ƒb®è0VÛ‘pÀŸÍ”#°yš6‰€Pi-tüp ’mœÝà j=¶›†u¥Iý0Qå_][†Õ+úÁ îà³ÊÞg›!œí{XÏêÓª›'*Á¶ö„{®r_—WX«Ò^±~ÁòCãÕ“˜e¥†/ÞcÉç„ÎJ©ìƒ§–®T‰s]·(?zO`X®õ9h°v=Œ`SEœ×°$d¾ ?‹ˆ— %É^jpaƒ8å>üÂô•ºÞPÓˆÉ`§`䆤V|‘ 7©(.0Øü®…á‡ñ”Ï¥ÞÇå(´nÈÿ (åÓˆ þžk™ Ë™+¢„!Üó·n©¤Í¥8ÅŠþó8k•ÅÊÊØÎZ^ÔsÃC‹eKÔ£M˜… Õ¤€hÑÁ‘V¢þK¶; åt,V™œ/“ ºx먵÷—ܶm91„Ö³ÜÙ<5 7ëÝ«!Ý[O­‡’jR/V÷àAÎrù&@~'Y¢à˜i{åVâ€åÒH’:ãÿØs I\Ã\¯J.5‡zküÍmÑ¶í ª7%³¯§bÍIÔÄ—Ô`µ¹ôCIô!ÃÓÏ,;ñÝÛßðñÑçîîð”þÞD#×8(]ÆÙG ‹˜LaYjº@Öi¢ªMÚÌ'dùOY—Ò1iÔ¥£–ó±*±oÍÝÔÄ9¶Ô<’Ï(qƃÂCv'–}*—r`ɪÝ×Úv² ˆ¨T7ûFE$G<åÛd•¥Á'ø]ßã ÖUà(ý¹Š½I\Sút£ã„Þä)q±¼IÞ0_È¡•­-Õ$š¿\Ÿm²ÿnJ<×Úf­'izº?ßwĽþÔYŠWOh™b™Ý)in 0•hãMH²•¿>ˆGwJî|ÜíÂãŒÃåó•Ô͘¶Ó-©¨X«‡˜(M©KAiP¯Æ3)U}rM«‘*Úd#BµRÃN>µrRµ:x•Cw™9¾üó½ovè ËǪ`äÚ_Yg¡ÜÐh²Ò1b¿aÉÁoTìÂñå˜ÔÌî «Ã8_øMÅ”Ú÷¥XpUd<ÅŸ‹ê @8@RŽûô£ç^9ìçÈðjΆœTÈÑ8 :àù°# æº>C”6ÎæÏaJšnæþµjÈðÑ·“u¸:F#± yˆuNWœßx–a.¤Ìº] düýâDkZqJ z$õ6¢é¯&E¡fÏ01äÑ„ (¨Ä|g‡ª1×r‹cG7…î–¬»üŦؼ@m°–e˜—¸?S¶D%¸v×ò­¡ÂŸé²sJýoxf@·"9s: :¤Û74½‡ ½ fâl€Ór›Øí˜S!¨94h¾_.®i5 L¬‰õ#c,1õòA*éGGšÅ÷’œC¨¤`š§ê|וD1u>½Z¹ö|"°a< x¢Z»×Ú×±š¦&a—º9sÃÔ2W¯`}}VÝhÖ™‹îÀíÄòeòu^·ô„ǛݢrWøë•ÿÍ•Ú(¬õ×9ãĘäÞeo¿jN…µb\0]w|^ÌQ~`YaÆ5Ú øïbÞ,ìðøß†À¹ïÓŽd‡0¬Ä„}š–*LkYŠ¢Î4W5Àúäí0B§!à!_¯}+ÌUº\Ò ™Ž@ âCÚ€û}¯ÅfÙé&’ÐîS*Äßk½Ø…<Ø TZt’öµÀé²fc+F·|¿A(ZÅúƒ ^F¤àbîÉri!€ö —M;WÏ*úê<Î8gNcúf]_ŒBFBkœ³ZŒHP—WÀ›­Ì…‘Z-+ÅZškÒÕÕˆÑl˪’_9Z×w|jC™­QOÈ ¿t÷&xÌHýýuüj$¸°Ò,­EŠK"Aâ§Eî[áBNêJ³ì€,Äâ×}óŠúpb¸£k“率!ÀÕð·=…ÚÛÄŽš»#qêýÁ’C±muîG% ®rË»@ +m+aJ3âõ6Põ¤²QÈÎ †Ñ ó¯*ß0¢/ši¡T03ŸÎéWR˜/uÒÈdâ…dÅÂþáô#¼5ŸP Î¥Ìup£:ƒí„ê¾K˜.¤òз'Ú€kYÜÜTˆ•ÄžÇùù[Œ© Ü;çº&£?Ó ì0m2Ö}11B³âÕGbèuãà‘R}KCBË!ÊÁÓJÊ%}v}%™¾ôÍbVñô¾Œ¬Td’`tùÇþ·µä\Ì⯩}éäï<]år¤B€x²ÓiªÂo%e«×‰­žò, ìÍ’ã_B\ÜÔÂ»Õæ@ -Ú’²QðNr#Œ×´;u±˜µãCK‹ôúAõGÙs© ¨fúû:ýØ`Ãb²Eôv<¿÷F¼¾šÐ<pË{_„;ÌRQ—¡¦š}YO÷žÒ߸ž½¶ÍW…WÖg¡Hh è4ÿÕßz½*JÜøÆdQ…Øë:e§êH•?Ô`WrÝh†ê‹¢Å¿}dú뛞%ú"…Ëéš”«4ïz¾p<»¥^ŠH”•êÎùýèª0˜õŧžÔ^ìůwÕöZ/±âÐ?íçO Cú¯ u×ÏÃÐÌTÜêì8» H .$TPË+ùôñ‡3WGÏõÙ´um™¡8¾Rí¤Wtý´ŒŒôˆŒ¡ýhË9é"³#™‰TÒœ¥ÎUÑnÆ Ã5cÒe w×xØâHlíæ4þCö xS’ùüxv'zDs—2Á¹ðµÿÆôÎ<Õ#¼Ø-ÖÀÖ(=¬DüÚžP9¿éÒùI¤ÓRÝ4úó‚m·x¿.}–A׎è?˜gh¡ªÇ)¶JB¨ZŽØÌ©¿¢|K]{£ªáü=Û±öứUY»0ŽoØLf@C ì'²¸Ø 5ÌÝ1¨#šÜǨ€ â„Få8äñ—½ÔïŸ8c~BÐEZCí" fï‘ ½y×Òy°_–ò½n¦ZyÃpòë§¶÷¥ZßO6 Éwlæv't0Ç/_ô•mà#¼ÚϘv…ÁÜú¥F….¦—9P4«{qþ–ÀGO´šï ÓZÙoâÁàJhRxIT»'Ïm¿Õ%ÃݰUƒƒåzÃU—Þ&¯Ì²õs-ퟞc|ŸÏRŸh6ÅæJö=ñ5Òw‰ŸUû®Õ|©,2T~ ÿê-¢xIŽ éyÌ#]K—ÏÐÝÅlê’&ÿpÒ&37}5¨uÜx VÞ3ãmW›§"€$‰E„|F¡ÅǵՒÈÞ±[€]÷Õ&½>§ô…!;¬Þ<’‡Å¦auD±¾‘(졹:id»Ÿ¥P þja£™îpŒeš¦«XqŒ J`Œ©ùëϲ6„}jšãµlÿ¥á$¦ð¹{-¡]â%Ò;çûá"/xýpã; Ò9Ћ”t>ОÒÍÿ~n†§ò¨è{arcŸ¡"ŸQHÀÓW û‰tmgÈ]ñÊɃXÄ ¼øºxs¤ùk´qIª€ª9ˆä·[$œªÕ1Ù]GwP¤~°i´Æ<`ö™„C¯»þ x^}qŽš`Ì…… Ëzn€EêWÔN9z™<önÿÃ1]o<ës €¡n´¢Õ¹Éè]‘i7Ð/Ûµù@¡o'^+ i«g¯ÏÙ±Oÿž³ûf!Þš¦“R¬/T¢ÀZ~ÑÉñ žS$¤åð%Bhby‚(°'òâà¯]‹Ï:…EÜQÜÝ[öî¯;â2's_ :mèÓ¢©!ÙbaEžÆ-§IpÍ]4Òå:YTÎéËÐs}66w-ÎÃÊ%Ô@ QPÉ*D=æÚÊ+6iyw/ðÚ-òΙr븇ã>½v÷ŒÇ•ÀeY§ÑˆÂì¿*òLÞ&{²; õ·ù>L7?ëãèm⨶ÔõQ²ê1¤ÍHÊRFWp¤RÕÿkeR+9ªfâÝ P‚j˨Ñ'hI²£ÇÊ äèu])jÇžÃwŽÞ_‡ËH’¦HÕrhSo9²“ÚRÄyöÑ•»ôîºBíjåЯ“|:4 ñ¸â•Dþ£M?Èj¯ô)5•FGnðž&Uñú’q’@¼¢ì°õr;’}´dŽ_f'ß8ô)G§ÿ_X,v?$I!Ä3†sv E%gáïŒ8"O*¿>-g%ºôæ·X³¡<f½¦ò–ìAŸû–¯Ó;7ÇaERnÐGÑ]ØL1|¶¹w3ÇNÜd}q ÿŸ\p€eÁs(Uå÷‚ª ˜kÇ„l†ÃtyÌ“¤c0OÎò?>ì Î5îØš‰ü4ÊøI%¢6K$Õ¯ ðñ5&¼6 É©ÆÈ0‚ø¤Dmºþ‹ÑP®ÂÞK%ïZ³4ÇÅqgåÜxO=ÙëÚK'²yÊ/ܦÔ:¦bÎÚAîåä6ö:t’~èj -¬KÓvyƒïß©¥ˆŸ‰Ä¦‡7×#”Bæ"6's¾ ïW]Ò{{GÛ²+ÒnŸy#â¬ÛÇZjx2Eÿù¦tŠ7¨ˆT~ÿð<9s~¸¬§šV;©àÔœyjhå/)®;åxµU¶Çb;éÁAD/mF°&¨Î®QK` ‰èi¢Ð—~asú_ \Á.Ò9ÿÍÃCŸ%½ìüë?“ º{Û$¡ÕѪ¸8cxþÚf%wý©gë’ú­`|éU¡P.E“Ò6áj‘à–Û$k")¥*¥}p ßòz¿tÍ·ð¥IÆ=y9í‡Ñ’"mØYmrjÓãÑBÐðƒö´|Qmçë O6d•=ÜkA<š£e»¢©*îïaçC( –N’Š4ºcþV³$†ª&† LÄG"Í?âWÒ×Dõ¡þµ&Ùûû²@RßY¤óioA˜Z~A­±dÒýVcÆ´ð´0DÆýÌQ¥µ ÄmìJ¤´Ð«äæE¸A4€+ í±Ú[&¯â•À‹¡77cõh_E{¥çÚ룃I‘‘ܵJ±g¶ˆ¦‡ª­*ÂOÊ Ü‡!DÞ§'箼üoýÆe4!É}8±¸9º×õ‰¡¯{Êø@ÏuU†¿Y‹%ôA÷¢ƒhR·‡àþady¿%Ažò Cúéc×;N¾z÷4` ~>â¡øÓl,:jž¹’û´d †¶fæi"@ìȯkX¢Ð'R°u¬&9é±ò†. ‚Ñ×v%4ï_ÄÈÛnF3†‡ÎïsŒFþKDó(YàV}m”;èÒ¿À×Áð1´í£K„Œ! C â‘|%¶-tZc·ó/wºõãŠÃ*¯ó±< ¨ãQ¤Mvš¸ósè«('$Õ4û¦:a}T®J ²a.ïùÎ{õ&$ÿó XBi¶è³“õ¦¿O‰RÕL ~â8-}ÕnáPÏ{8äCÔÒeØŒ¯•ïZÝÀ¾â±¸`Oðþ{Ö&TP»^zÂî{0oî­Î#œîn‚*Jµ•'X íÝõ!ÔÓüaû¦ÛM’iÉSæÄ7<Dô±¦« Ž–ÍŽÐÜ¢¦#°ABÁ²4+Ä ÚÃy¡‘B®Ÿ(ÿÁÛùbÇ`*9çI7%Ÿ¥Ðj÷ì7*¿xê„·íazùMÿG«/àïÅc¡’³,eSβôq˜H½)D&øÀ(]£˜Æ è¬H®;h+Nv™‘—«Ë$¤Ì<ˆtÝõ +¹ [x‡#-&j£«UcW¹[¢‰»™s%QdÆìûâÍ»A64êüŒ à· _!ûzs2RUv"Ø;M ‡Ã8XúwŽ©†O†oYqVîWUʵýÕ°¡ŒŸzS×ýT~«šîvE ²ûk¬~SÄ[ a(™ù-Õ£€ž®½ ‹DÈ—ñ*›ïJ4ÄϾn†Ì›M­˜o žÔÃçßLÓÜÄ,nscóKÓ¹emüoøspewø‘†rÿxHÁ?Ùý¬pëݸiò估ˤHx¿Œ2ªšãîþ„¨€ãPN³ìÂbâ±5CLÃÖÞ"ÿáÆg Pß7 ˆävM¯V}ÐÄÆ¡j³µIª$Ò<Œ vÜÃ,ï}P°/::p-õ Ê÷Y‘Mõ®‹ˆÐ¶|:“e.ÆÏT#û‰‹þLü{/$®ƒ’¦-"!Öp[Pˆn~!¨>»Æ£¥Þ]ÞB<ÖÞââÞliŸpåë•jÆ,øÐ Vå1¡AN±ÑÈP’–JAR`5™{7Ô72-Ä›lDµèQ-ž@²‘ ¸”F›Å¸Õw:Þ…9“l1(õ‡„׊#Œ×ôj÷`)e ÎÙŸáçd¸~ %CÜMe²(O•X &ÈÒíÝH¯à <@ňIQ›Sk‹× ˱)-Y¨6›Œ"­îÑ@æç¹oé](Zª¶wÑš$’p¾~}¨¿ÑôÍò¿¼¨)"H˜ ómWáM¨àߦT.Ót 5/„ÍÊmR‘ã£#D°À©“ë–܃Ô3Ï,”RÛhŽmüw¯ï–G=××âÿ…ýg” £Œc-õ/¦ÆV«5uy5t7bi Z¦Å«ú($ù|QøA‹#çw¡*»9ÌÑP&uƒZs»z{TCï,F`ÒJ¤5¶¬}v¢/Ë£ Yc ršv I‚(ýìuY6Ðë]1å”zÝJ÷²2±—ØÕ†a©+¯žVžßĶý\@¿é®·ÂL$o Bï‰{Dv½Yfƒi?·ŸÅDh¾ÈÐ>ƒ;‘±>sÊOá`Ü—¼Êhgª4K½.ž…cìÊÇBÄi0LVWÒÇÜ_®äÙe1 ú«¼zèEˆ©Ž#é-¨ðvœÿKý,é¿Î_à·È= —¦i<¸äƒZ‹ }?|]ó’9“”Öí¡jl!ƒ MN£Õݱ'ÃD*v2ÌÒ¥kzCOàÖÙ€O¬D¡†w’´‡Ñ¡BS3¤§O¶ŸdE›PàFð90’jD}¬¤Ögxž„ç£>%VUµ£}³ÈŠXé°;ãÝTTé$•&zö}z u6RšK(mÀm\Is¶{ñÕ¦Õ°1C¹o ×|d0«áIœhµöÚ9d— q|·¶î ²¨;àålwµ5µ¹¸ýhRÖÖ¾>N|¨ëŠ”iµwÂ%Â(_MžDyÔE=Ì4½5Iãâ;®œ,ùú‘tI ÂÒ§ «z®é`D;~9À´¯þLõâðgf;zÝõxF97`øÈ©Pdø¢‘½@/È=“GwúΪ&±?œð‘óââ+!»_yÃñÑ!ý^(@æ¯[òÂé6é6—±«ÞžÔ‹ÉÉY•®l‹mz"*rúŒ`;ÉSZ¾Ÿ}Ík€CæÊCÒr\þþÔ¬áÏZe©RúÓÕAe0<~àh8:úƒÇ1yæ¸{µ«ívçPa\;P[<ïT­˜–RÜ+S ØúF×\S/´û÷KÆpº¯¡ô(ÙÛÇâB ‹w5¶xº3Ç­ðeÞ©–âªÍ|U½×üÆ´vks-N(¸¯–úì¯ñ-G˱”_GÐ_‹¸ÏYþ­þ|]a±®Ö„Õ39˜ì åºÀ·f±žpœlX•·Ž§ý WVÙ,nwfÛH éf ¶¯zV¾ŸëÄol=Šzì´P8ä!UJ«ÄÂíw8EÏÿæª<.Èš¬„?šP—?SLnÞŒšàDZF¾.z¿‹Ÿî]Vˆµ:øÐň'Æz†Â¨*Ý¥8N´èËÆÖ#@_£ßùªs„A¸»J‡o»xT¹8?—È>æÈ·­ˆ™ õìížbÿ´†[WߣæÌ~bÊ•ÒÒÃŒÞðºgì¶ÅYàwhet ÃèÐ.·€ÑOISj‰xÓ9[+†Cpé2Aª&éÞžUâØji©ëî¬nRjá}·v™cÏKm¼­("¼Ñ?áK$ˆP‹÷XôaÖéù”ñªz©{CQ¼EÄekdû»<²¶Q¤H0Ÿ{(1w)1Ëè=Ñ|²yëjƒ¯$7 ´sK¯W (£ …ª2h;=”’ˆQM¹Ýÿc‚¯ŒtxþÀ5)„9Õ’òûI¬óš«óW¾|–ïûöŽóGߺÍçêÖßÿ]¶½QS ã2©G/IB ¦ Ò’G¬Ü- V݉T >Îûanû2 Œ–ÞË­Ö#?iêK@aÂМ¿¨x¡dÑgËzßÍ‘wb>I†3È ÓÞ Z—Õg”!³‰+ŸX¢Œ/V¢æ3á.ÎÀNQ5Ⱥ$ÂöV/í±_#·¬ÂùŒC¨Øˆ»°è¿ÏòlxˆEu¹N$Ó¶šú±!‰öû€†Ê›joÆ_)§Ù¥m%(÷êE†Œ‚^M…üµDðñ [A=®­B95§ÜÑwoþ¹Òì6eÕ2½¨ZŠ(` ïeßw(¶ú q}Ô‚E kÕ·Q˜EÅ0l—)ÉËïÁëØÒu2Ðq‰#ÎÔô͉jMþÿd Åæ´‰ÏÉÎÊqH=TÛëÖgRÕ†cuÎ=õÅÒ ²tëµV –Òøáºf¬htû/£Þì$³6,H G)ÌwÜäYܸWÑCì{ªÓ•ü.D:¯ü]ÚQŒpq„äÿÅÊÌX4-»nEq„¿R¨h“ŠÇ%m°sðC¦=K-ðž±÷üÀõjPA° ÐÁÔΞÏúah[N4%’¥Çs$tJx™âÞÀHqæ¿€Á´ ^”JØ0Uâ€÷Ö›_…¯YLÚOS¤ÞQÅÛ÷ô,ž¤u.ïc€„Fo}O­Æ®.ÎYæ@ÅӡɳnÀ„-G8?Ýi·ÈmÅ“­{I[ËX¸àÌòGàÜŒ`8·Õ*ƒê/L ø•\Yd ÿnCÒ¢ÁA†Â”¡LX[ÈëbÈýI ® øf“ókÇW)éæ3è¾WuŽ©fŒ•ÉðqBþ”*¹°øÖ·ç™ÔÉ«mo±•:ë´W|`ôÌI=ÚÞÚD{Þ×}†Ã~¯þ•ÿi¨z@ÈM§ó`Øï`î7ï¯tdˆ–µ \ØkïÓ§#³Ëˆm+ÓÜ“t‚K.k·å•(÷ίBÌ}Ï+†÷„¢l Â÷ÜGÍ‚z !Ú*™;eß±mÀ3ï²háb·;q?¢Øz’Ã0äÈÏмèy÷’¢ç#ò)yu3ЖZo‘È^T¿ž$Öt'…2yia* µŠ§Ý#åè´`i|+6è7@ºxB®NbñrˆÎÐw`Æ5©{œ‰èžýÐ{C9òØŠ  “—áyª¤Ðµˆþâ.C¹ŒtQ„õzÛÆBëf¾¶Lÿ¦°¾}º/¶ÛF˜¦K}ðÄ”ßTúÀ¯g€à•¿(ˆ°À¿ÞI ƒƒO؉–ßÞ,žSUëäÙŠ·¸Aë•߉FçIˆe‰æ ÿ(GÛ~ÕÅYŸQ=¬K~ï+Ãòyˆæ×6 ñÚ¢%Kj¤ŠKõïgè¥ ‘qé¸úçôÿ/ìUB:,j›PØvUvÛc29}*º~Y5tO‡è„ÉÊSUïàu£Õü³aCíj†€0DEd ÞP úÑÙi|Çd¨¾8—¤þag.ðôÍrºáG¨ppÌœ«Ëq§ÚÐAÄý»×[KJ8ŠÞþ"}£Ì³øï†Ö.Ù‡(£…%¸¶sB±Ü´Z(xå‹©ïÌ­¹C>i$p Â4ܱÌeÂOªÅŒK·•¥S¹Zz,JwB?¬'¾µ°+—¢Cáçz†áªœ‚†¥Ì†ÅH€/Ûõl{WÚXôL¸“’0:3rBt³%Õ17Ÿ[þ|WŽk©tYÔJޱa§Yuý;kšh_óù_µÒniÿMjÂJäËÔ3Þ—ÀúMì)N"õÓvZ·$Ù Iºcªµ$á½3²­àîzBÆÇÖZkÜò¬ñë~#Ÿ4©´Ñ ‘ˆF>$þi=]†Ég‹·8gP/ε¥£aKwƒÔè8ÙY±f™¦Rd„Y»#¹fÕÃd›’7N#sºN˶l§ö}èôðH(Ûà}™LrÁ $ j«}»BèZßÐCiT%P·©Œ@Å ‡ØIòs´‡rÐKæ µÔ%…m.„6÷î#ù‘p1½<á©2èØ„Ÿ•<«”îmÏâU8É|Îþûh¶àÃTüK~+mí™är^š‚ç> Ñl'’©[?°6Oµ:4†=Ê6o·ro÷RžŸkÞ„ÿ7y ]cj.¿¨ñ48gç^‹úâ¡é3µ*¼Tè8Ç9ªÐFÆš…0­™•VFå&±»}žæ@Æíè~o\xuí¦Þ86LLÎR„65`Ó •¹éaD[pqêÞh«”?hÆ([qE\­'«²2IJSìð”3©SB·Ã0ún›j@¡&¹ ^˜ÛàÞ9@Üßz*"¥xæ/ËHsy¤&ïSE}>T“üi¯ùå†ßqÌ ¶h5xo7üòËϵo­5n£ZþµDvN€Ùþm aD1ëºõóá‹c”J!Ÿø(џײ»¤4Cº³&ËqOî©|v=ˆ<Р0wn`NIƒ¸˜íÙÃë—v—Hðûîé†Î´V×½«Ã\b$Âv ?úQà 2°2‰o:¹ÏÆXéó #Ôýn¶Ò*ñ{T¶{ÔšÛòƒ ~œ þ`tJ¾)_Ÿ+7ü©]¨9$;V^J¿|‡=îLÿ}Á㞸#†,·rA†¬ÓœæÍ/L6àçÝJ, ÓRêè×u/Š]úlÞÂI(±7z¼=9÷â>N@ã2¹µmu<§48'ÃÃ×”g§Á’B³ÖußuuäÄ‹˜Òµª¸¯âÅ#MÛ^˜9¼ûÏFLE¨Ë—µDØŒ‚¶ËÑ_æ úJÅŒ®9ó# ÍIªâ¥üºmñëibn›[Dõ95?Óg/C‡¦Í«¾ f´kêeæcT°(¥8TMyíK)¹;¦;#Õ¶ø¬Þ§ˆVnžlÔÂáq‹mf_Ktyð7¾_Ší9x’*6ñ/Æ1ú_S}S?ðv¬Ò´b/øéˆÖdb͹ZVÚL‰Nƒ çrÈK=(¨¶§Ð\|V¼ˆ’·ËÆ$NKÆ–¤ëé·À-Ø"ƒbà¢æaZ)±»v¶¡Ðh Ô‰š`~#™–\Œ‹\JʸÚsuÀ c…™­ð+Rüp¢ˆú‚Lux³ÏàO"7õ·žc“#öc$Ñ·¢1Ï¡ü‡ÄÄw ̰¡„Ë›*Ê Ð×wgÆ!½ú‘f<²)ì!-ûƒf8žîY LÏk·}qR°€&2/” º‰$•ÒLF3,Ž$¾Ë*>caHŸx:¾ü|¬ç–'N›ÄûÂbÐÂJ 9øg˜«[–æÏïÄ_‹z¯€‹¯U—¹Øõ¹{½ðæ¿ò¨}1ò¸`IA '‰¤(JÙvVƒÁs½z'¸áf8´ãÄ5îXfw©(ã…±uCfw¾‚I-õý$ˆO ·c¢—ýcrd[ô€Î,Þ‚«XpîeuÓºñ¾~ç{õ¤îŽ|ÔFþTÿLý é])—Bä•‹&Ï1üð¾C!£˜;€lµÿÕÐ蘙cHÖ=e•t ïjïX tÂÀ`î2 [ëâH»Œg™ý¢äslZq\ )~t/68Ö“-º •¿“ =L´uåãæ×7¨È›xÆüfõ,i41Ù-ö>™“Æëu®„V$uuÓbüÕåÞ’+Ïs»qA\Íç {Æ}GÂB/‘ø#¥2hö¨BÄÅïŒáýGb¥Í58‡›¹B+”—8 ¨sDGrºjf4˜@B߃$”áúŸ õd?¥ÇyÅ¡ÇÖð4¬aꜺíòçäÊŽ%ŒÇfÇ}˜’§qòK‹Ï'RV}˜ÅÉŽ€„¦ –Õ 5)uu9È5ªªó£PP÷X‡E.Ç®lr ìG:àþ¸Rv×N‚÷XŠükì£Æ.MwQMͶ Ï~{ž9¡Ê‚8FMÄE„ÜO˜Ó&«~“hŠûs8üÇ÷eF!ßÞjŽÆÿ¸² /…1ƒ xÊMÀs…ÛµL³™­§°‚€.¶mûNF>î±BÇYý´:”À¦~'Ø ,ì˜ÙÌ€Acâ>¨‰"0QZ~Û£*¢ªÎlîúôü¦S¶¸ƒh惘àÂôVÞ(U´±ôƉѩ)mßc+á²ÙK>dÉ™¿×Ùã“¡^]h;qžèP‘L&g–íbx¨Kòǃ§ —ªÕb;í\õüŸ‚¢­`ÉïݺûíÙꪔ“‹ÝQãð‘ñÑûIƒ{ «ýAìdiŠ~È|A1±þg§Á_(™ÁH¸f »¿Ãø l\;rÞÈÂ}_竸g'©ø»zÒ–ÌŧÃÃ.óûän"gœoÓ`9:xƒ¦4Íw䲨¨Æ“¬´‚s¸~Ί„K÷«Df¼ÿF.Ò“b 7ó}U‰Mœ­¨F A&bÜ—…O:V3ëkD2¼p• hü­„7 ÉŒl•°"†’k»^n\y@ˆ"ø2Ûè½*ëpQ9"¤×–é×Ü̳‚pKH%¹Ÿònò…ú¤Ü„¾¼ÊsjêŒLâÐÓ5|<°Æ/Y)R›ÉÌ%(œ¹°qïQPÍÞŽÍö[ *æÁÀtí.Ñ2§°m3ó“4@S‹\ÿ¢Æ9DUx’TˆkZY"*Â-DzhÒœ•,LS«­¥žd6CïïW ¶:l×j^¶û®J,E&• Ž›ÂA?94¡t'é¿èqÞSЬù%7ƒKq;O„Фæ_r¬“NûÞÈz1ö ŽAËõtt9øâV¼ô` sÕ`ý˜B¾¶èàßXµ|"qá}':·Ð)ŽÛEÞΔk~[E)Ms|»qG  û”w¦Z¡v‘79Feâ{áÇo0?Ò¶XŠ„YJ¼,/ö¥—ž¦®÷qk1xpda½cŽÀº™«ßÜ+ÖÈ'÷AòÚrƒßýÆÛ å‰bï6`&v {”$m Æ1™Øë.‡™¶rÒ M“FÍmU;H©îÍ÷I Ç)ã»ZÍ<š«ÊÝQI—¬Œ¯‘êœn¤Oj?¢ÑÿoÛ²š ·‘1”ÍuÆ¡::cªxr0Ò‡éØVN˜Ätži¢8¾ùxfœ¾X}ÌO3 (fxØnNXÚ‡0¥2SÂP¶éÐòX>‚µÿã$¦­ÏuëóÔþ5 ¬ ªWÝ6¥6¦ð»æ ™Ðb5¬U)Æ/xÀ/ŠZóÚO‰gɵŸË¹ü{ÇöÄçšf  bòÒmvòKê '•ð/‰¹GÝY U¬ùóÀï:;Íù=¢árîœ(ß6$íí}ºyòË"ù!t^ƒÎ’P@«S `~TwÚ ©|Öe9Ä“¦QxaoÌ›¶>òŒ âN% ê—ã“(X.)‹}ò= ˆ ýÒ¸öoÌ«e´:ÚÀ]çò«à¼PìR ¶F áçÅà E±a[؃ޠ×|„ÒG7‘Û®é07ÁVꇢtZ€K¨ŸNRô%J=<ÆÐ«Ääž1b|dúæ0ºeèšeXõ“óí;£yTGHA±\ˆÈ#nøØ$â±Ø–fÿý²éÅçƒÐäVPÒÏhŒ¹ýÜå¹·ÀÜC Ç~%†íÔÛz¨{Oö;g}—;åd2ñ¼SBeûÕ꣣ý¬Só²Q«¤àXX¾ÑIjNV `" ÏF/¿c~X™¿Ò«­c4Ϲ>èÍAÁòmgn‡TŽ­ÐÕu±»#{rÔó: ‡(¹‹5ЫˆH*º?^Jæ}†3Pxå`•¢}Z)èøøÆAá¶ðPë`Ò†^Åö´Sây´ï™¡ÂaÙ·±¶fÊÒ¹¼~˜ãºŒ¾…ˆÚc]i$ÊòH”m3”!¶éâòv¬˜+–8 °/Œ²ÿÜDâÚúð‚¯MºˆhËL7¸µçX ™Lî3‰æ7üi‘àS™Nl¶ONÚU‚\&‚Û{ïžÕä@¬ÆÎR»8= ám?>Á9U….˜ÚO*MÇ"2Òß»<¸­@í KàH ‚¯ÉRnñX0b¢ãj¨-ýi]Ú‘Múí¤.øêåŽúˆuìGn1ì¨A xÉœžï C7²x¡”ËýH¶7ºÂóÉçì“©pÏîœËæ8bkFæA$ë"ƒ[žúŽd•4³Ýú/¤~¡2|œDã4£ïe¹™ã07à–àN˜ß ŒâB·Û&…üyÌ]+A‰Îy½SLGgêôûy\O –‚'šfò‡sëzȾôªf~3õðSÆÊ?”Wl‚‘zŽ&þ OŒJ)šŠ¼4µŠúm­Ð#µ$×Ü´Í €Ål²î«‰Å… Z9¼¨5D&¼bþ3Ú½ç훀ß‚poÏO Jh„±ã®o •p’VCv½q_Æš­4ÞTÇÄù™m¦P¹ØÁ¾lì"¸~ä¶.Lj Û“3nSð_lhìüþ¢_ ÀëHo¦¼¯´ê;BxoiºÌçÂ'×Zyﯿ[1‡6Ø0yÅÞôÕU_@ßnÔž6åž~˜g:a¯öV¸ +{EAÆ>‰ð{2Æ–°h½íÔ+5¨#jt3;³‘-ê“Òò,­¦h‚ÆÁɉUï¦f+v§Bõ©ÌȲúÐöÁðŒ’O/6Þ^»èF~¾Äy=ë`:¸^lN–>]ïïô¶Ç¯…=Ðþ#¦à¤—<˜PþÔ“¥£Zês’žÅœLˆÈ·‹ îùDAØÃÒÑ¢¦×“aéò"•ÏNGl‚náJaiH¶©µ`ˆ„wJKÐ[QaRƒcÚ[FlnnV‚‡nnoTÕÆTp³Ï'㺙aùƾAŠ×bŸ|Ë9;HîêÉ(Ø"dÍÓªfÓ–r7D©è ­½Í$©L®q/‡ÐjG CùìÎ/©öá5"ì×ì~n27Ô‡' 2îùæÅ§819̈ÔU¬Ây‰"¸]ÜþvôÝËýùèŸ<쥳‡ŸúlueS•»t&žü»ù™{Íî_a ¼{G]ˆ'(­¹¯°µ—NÏT§fÚY…MTqvûb0÷† 8åÜhÑÓýËlhÆþHyí±¤&Éá+v’Õëxí“kW[bá¡!%Ð3¢OÀèyr9²±Ä™ˆž÷ÿù7­ôˆ¬SgûC(!¥+­ÊD\ý§›&eúÞÃÎ9)Ì€ڕ ¯<¾H#8&o;ïyVÁô®íjøب,GK&“¡Qiz©dçþX ÄÖ$Þ7`€bÉÜÿ“·X«>¶/œv³DáEKÁìr…+ Ù±\Ç ¦\,â›}·­ëq|•Ç»kºàžSw€ÿ—ã—¸eèÿ;ƒÀ¨„õ¶ÆÀ¸Ì£èujÞØ0i¬èêéOëØq€y<Þ½Qu¶^AuÍd›¦EÇ@…±_ú¢|H­>\$/ éø69Dð)j—©… KŠ¿³£”ñ‡æ/QNÝ~¥ççˆv ðk´ù<ß-M¿Y=óÌ4ïE¿TœÍn÷šPúýd¶JóM³â?BÅáÕžŠäô‚˜Ìu~w¥ydE¸°Wøû"±CD«ø;ì©´/&ƒ.Öÿ[–¹ PñõJ½è³ôê¸ø”·Ö÷ø…<IJßÈáM$¤ÞC ;1,§ÅšH´ÿ…”G©§ù«N“Û3ã‚ÒÃèösƒÝªýµ &6IVÛìñÐ˔0Ò~ìú(ž5õ{V'• tIj˜nÚžë Ï›S8û°._DɦŽ‘ çç ¼Œõ8Ÿn_ï!| ˜ä44ÍÂñ”þœÉÿÙ ¬x,Â/SyÍÌc5-è¼@$Js;Ha±AvóL.B)^4YLúˆza€àÅB°Ï¥kT⨀†Á¥}ÀìÅ/túmak+ŸÍ‚€R“}7´Æ ¼rÓŸDB®êÎþ¯¢bÒXþéŽV{»áa¬Ý`ãÞ†KzAc0F%î$¦%tE›·Yì<÷èUŽ4qÅýØ®+lÒæ -Þ>ðÒÁØ÷ 6Ù *ÌRE…ÙÀäÖC“d;B3{°Vœ@Ôòƒdü:Éî#XAD•5~)=qM×Öt¶R+V~¼Væn®Ùwö\²¦uWÑ{¡S»Dì{AÇK¶ÈŒ¤¸;t³;í£ÞKˆ¯41a|pÊ­ _±/yÔ™€Ô5_kÁ˜ï§ˆ:”ù\> t¹´=„©Ûù4(*ópåCE&M ŠäŒšî2Ñýœ†`/Vû“;(Tú 1²xƧ@”t ЯØ®8H"øÅsÇâ2R²Ñ_½¾¨_Ã8wJ©Á~–Þ*ÂN“ëS-,lx˜x uº ©êÊ4PÑÉ …y6(„à)QÊVº u<†"£ü=}¿Ÿ±à"ß Ù–½(îhf(Õ\Ø!WJÁxؽÀ4Õ É¢Œ°ûƒà]‡ÖdD9‹=s}°{ØÚk¥ˆÿ‡,ß5éô@úçˆßŽ®¯–¶Z¨4CÖ×¹„Õ Âé?aÃçK\ŠExÝ*Aä9•®ËŠþ<°ž—µPL!Åü%ð$Š46ØNËìä³p&WHÊb«rѸ¹Rz]Ïj ¢1£q9@€»Î9(ü™¿9)¥5ò"Ï*aB[òVÉZ¸Û¼†ßKm˜Þ;jaãÞØ†0“V¬Im3h 'õÇZáÎ…#M¶MÆ´¾X‚ÎŽq9ë’[R…°ý“¶`zíE ÜXÙç6¼»Ó¨¼Æ¸_Äš îdÙ‚ÃWJ¶œ¤ï¥¹OPAJ« ;í·Ÿ½ÁÀƒ¨ˆjJaÍÿ,ž;ô‘¾Rêæ­RÒñ’3Ò"cÌD—Òpõi`¥æÛ™{ñnÂI»LݽŠ,)ä.`Ÿ…[1ßG|§%5:÷!Ù“øÄ¶uŪˆ ¾j6!çre˜Âk%¶«­¦g¸DH´2@E èË2n rÃUÛ/ð†„Þ9I4ºFÜÛéuFÙ\ó°Lûœ‘¼oQB²Ì¬õW¼WCÌêðü:K.gj¬%@!i‡ Ó"æ–zU¾» ±Á?­8¹gª 1©Ü¼DÖfú•ΰÙJœU{VLö„ÖO׎r¼åYãwþà]ñr&ñÉ-vǦ½çjl_,Hê™eÇÏ«˜F>²Å$xèa¢üäÔù{ËDØ•ïÅo0‰?F*¢¤¤yÅiÊk‰ð÷}™gÂ]4´/âÛ‰S'7À*²'Š…û‚ôª¬Á틃7G¦~¬„á¿.ÅÁÑÛQþ–@É4­¡ŽÏyU»³w:©Ô¯Š36¯¥iäþ 2 ”|¼²¤)’ž†WÑf‰Ô^á7LòiÔWÉQ»'º˜ úTrpŠÇhfÓ>PÚ]·ÞçîÒgÞpc­EëAáss¦ó£Ó‰‰¥{hïÃ~ü–|ŠÃj{lÔómAqE‡çî_©šÆš)П]*)¬OÓÙm&¶K뵪²å ôg0ÕRJ餅ÊâòS#vˆƒ†o”_V¹M€‹ŸiëcP0ì¦TÂŽ×%ÿGCå–,‡Eãšúž–§O&Uec!¶ š󲤜-çàK™Žß<ˆ‡>Ž6AÐSùß/PR£ˆs2ÞÉ6o˜—q}þOŒC„,.*³«—qfdÔÞºRX-Ð…©ýÈщA –?xÐ9Õ€ÎFe02Í9mVsâáÕ ²TrÏ\vþCž9Åex(¢9ñþÂϧ•‰koù" ÷àùq¾#÷e60†—‘P† ÜÞ·|]ù*ÿ˜Ó6yJ1Eú)›¦52³YjÕu7vÛë0ôåƒù˜‡Lõ™ȃÏÇ׈Ù?Q%ßüÛ9gúIlK¼îTÊ]Ú§¼e˜)ÅëK¤¾WY}W”{Õi|2‡pË}or4"딟×:´ä‚‚˜‰K¶\ašÈöœ‡Í5Å)ZÝ›¾s5Tü¯1\ºÚ&>’ó ÃOÖ¾‰5H “§6r,dínï§ÂAœM‘`‡,öÅy¾ r£ƒ´[g~ƒ\!ýŵïV=tš”n/!Ã& ˆŒ gÜdcºd•²Ãð?†Æ(^À¢)ÜÝa½¦9u,R‘‹‹²k¤E4àÙ‹ñ¸³síäPûÖÆóЗé Bç1V£IWí0ZYrìÐ u0ºúñ{Üï,åpƯbcÁOùï1})À”ß­" Ó¬GÕ ;?Ú¬ ñÈŽƒ^N1‰ýšfžN]š]×¶N¤8ìAÂÍUN&—ã ðœY#ê›¶¬w»ÞsavÿJf¿ÿº©j7Úev÷Ú:ßJ娰”}Åàÿ9ûÖs¨Þet)s½}£Èö¿“Újó­ƒÇ7ÁUܸ‚曪ºÀkÅ2yÕlõÁ „Ç)| ³àc¶Øz…Èy^€i®ï@zBŒl–\§äüðÞ‘Çþµ¸m2g§¬þáºàÀÏ!Òó#ù´h~£ub„”âï–ƒ\Ú:Ää¦çØ`M2ºV<¸å×åö=íßÉ åd Ë;ÉäÍU_,ôÑ+ÊåŸBu¬®\WÔá9ÍÑà6¼®‘â€ú°H–}V çø±ò¤e÷jÍÕC“7Ó¾"à­·œ«"­ÙÁB×ãƒß>í0”¤Sä}}Ð AÆ_ÙeLgL…-ê@\l¼4yo\Šš¼©¨(ç1?ë¾;œJ“ýH‰M7–ôÒ™U b‚ÊgzW<榟?2Œ^Ú&9PW?$¸½ªü“mL8]wD{›É’y}Z³˜êŠ1Ë0ÿ;º®Q È–^•½Ðª%Ý›{u±q‹>úýÇÉ¿ )Nš?¿š~£@Û´o'e{„ëIÍj+U¾WY¢íd¹±ƒªQz¦»:÷ A"z]SzÛÐÀg‘du@=©Å“Üm¸A+µrõ7Âlhë±znNºœr þ.2Þ)Du¨+°ù~ÜþN•)ÅÌþÐßþ“ÃH‰ë£Æ?ݤËL(ñN1¯CÐVxúS—»†¡Ê¢•#i÷[;Üà²â+ñÖ½ºi ¨V“ÆâѸ"µj”Žk Š•1—bÔ«ŽvcÅ4VÍMH’´QÅ;ÏÒl¡ÊIè²—2ã}l{W‡á–EŽÔ›"2ü¼·jvìtÀkç¹Tšù¦»,`qã'N.ëŒc]"¢Ò¤#–øhº~«ÔN,RŒUú]ÛÕaS5¼H3*ñFÒ×L0eR÷G:¼ÍSyç6ÄRC[ÆÊB2Ûö]Cô–z¾˜‘;wŸbžÞpËb_ñ6hÅú†r™?bo2$U@³$þjÂMSÉíÊ«ƒÆò|ƒl ŽÚ,?R0ÖaZ÷Êñzø}éÛóÁH4ݦŽMÅ † &¼&Ci9Ì3c³€&PzÒXzŸû‡#Ä2ŽH åª8G|Õ¸É>:q‘§6^cûJ¦ì­Šçør*Á¦~½cÊ÷º§¡Ÿbh™ç]Û+ú‰rXÙA4š7k•i]OYöõËß ¤ÁšÿµñÉk+Òcf¼70Þ]oÔÓ˜¬ ‰ËÜ5‡ 4Uù ¢Rýê É€º#1FjɸËÙêhhgwA,4øŽ<¤*Éûx_àH¥ ŽÂ@ØwCc·{¨¢uURaœ¿ã†Ã§Ò¤úJWK–P(õóe´ Kˆ@$x¾‹CÙsÆzæÂ^Ø Þ€¨<·!^ú!šÌ’SŸÕyÒö‹Úg%•àeWÖV¥"e¾¾P/5Ì;xàüa™«Èh±TC¨s0 ù¥Š±Yná%G‚ëôÂtM¦S -r½‡ÃÅg–ÁéIä;Às[b^/‹¨0ê½ïtÕ+D¯GbŸ¯¾ãdg8|À,U§4‹4àKå+skW+÷iòð€‰¡>nY»2¸–¼¥%4&]*¥Í3MB…po#2XBE9RÙ³PÚž1\bc'aÖ,^ø÷ËæòJ?ü²R:§›© 5‘ѯ”q0TÅhxdÁ§T}7Ø3K+ÛŠìý +–nñÊèNq³Õ‘$A 5ŒA ‹›Ê†ÂN(šð¥B„ÍO‘„—ln‹î…Á5‰Æ³ÉuŸqUx!l ìî§mÝë·£¤0‹²_ý“`ÙÇ´š#%?BL¹…‚£ÅJùÕl¬ÆÅbÔoEJü¨Æ ¨—,pO|ÙÛTßW„¥|$*ëý 7×öÒånæpeæÍÕõ}oÆçHá†Í«dþžæ<7S%'¥sHOæ}_EÝ*K&ȪøIò>Âb‰QˆXŒŽ¹rƒo<ãuú,cþY™É'¦£Í¼9 Ÿ©p$ƒ9¤ô GŸçƒ3»Gäß„Q«B÷¾ÀÄÁýñZ:°¨[ = ¹±rr ó&©×þÖÿÀL1~ØÞÐ}´±^_£4v ˜$4\› Húþÿ…ÇiE"hÛ%öMzûuÚòˆéÙŸÑ”iaÔ¶ÈE6yœ%—õ߯sN€*}Ä›áÜò,g‚À¯NßGã§™»*1y£iäÿ\è¹Péçé+/ÚÐÿ«ñ‡Hðž¬!l¼ã~Ô<ã øP¢ãën˶iô7{â0Ã!´dúu¨€øÖ]Ó3y“oŒ“1Á_œðº#ìÁó™£-¥õ²$©¿Cò’žl?á%ó1PG«¦µiýTtúv%°Ï§áI `6êZMbÿ£Nvóý0ˆÊ<¨ãë„D˜$Bá@\K:ªÁsB8)²ÈÎzBˆižìdçêpöHÜéV]uyŸÞP¹$m€a¶Õ€M"¯nMÃÛÒ”7<úË¿ˆ(þ Š2ËËn¸•—7¸}îɈ$èÓÑ>ÓÀÃpÜ’p3³ŒNfÉL¡ ˜O^b-ßH¢(ßï+ÙCéõ•×&tÕgxMºÔþ¡µˆl»žÒÔÆàÔçP8¬X,É*Ø/S‰€XµÃ6žu!€Å @wtÖ–UK—¢þzôµœŸäÛ©“‰”ºÅ¹×øïPµ¦H&co—9’œq44TÞ¹6Q€þmýªÑ±E~  1õNÙ ž‰ÕB‡%T¼þŸÉ™#“ýÛU©@§"Üë)Ö¾AË€õÏ}%ÉÒÇ4é\ÑïJû~ñÑA²ÿJ§NãÈ7‹…Q¢–)Oèµßórš¶xµñ†‡‡ë|‡„m,‡ÝÄÛ8H‚{ýÐÕMÙ¿—BL#@£W] &´%£ç>™Y OFå=¤¾ä챚¨Úº¾äQ+èÕ¹­óv¯ ïÜu”+Mdà£&™P:ïÚ§¯·Raú71§5¢èלˆ«ÝßC‚v3R} §[âKÓ7UÝæ©À›ÚõÎ),AA&ƒ“'S'ÃÑžLüÁòâü©É‰óæês»s,Ã*÷wûÕ@®M-~Yñjòg{Ê?méH 8‡9  TZ«(ɤϬ¿Fß@°È¯& ÍÓAFµ’+ª"~W¹Ml²Žõ2;— XÛ 2쿌³\R +µõÁ§-t`ô†Xi‡=4ì+ ›.õæT†BÅ’×ÿ;­¶xKZ$¥u][¿ÂèÒwæ#ØÅ¯a×aߥƒÓ‹•OÁ8(¬¶^¬ø™ñƒ—™5,ÚçÑ&[¸ ØÃéÔv []…¯ ÔÏ¿wè]`¤E™„u4»$\¨ûè%+ð(.8I2â€ìW ª½}uà•Q·ºO€ˆ0Qïiªzé¬ÝÒ‡a&ëè]<ÄÏž?2–—¯È»U±$ÇgÑÁU «ãÿÁ!Wç!lVŽúÑçLvfǃ<È<ÆV>àŒB°m¥œ9ÿn£%S•bXíüôèÄ ›´Rêkƒºøµä9ÁW¯aŒkR[Ç£]s¢¢ð&óà‹íJ¸Œ«_Š^³fÇu\ÂA,Ð’¨!ý ϧܺÁÏ‘9ŽÏñ6ЩÕÖD(é-’FÁ}ÿˆiÓŽ›‘Å4äyTS-B+Q$ÜV‰IEë e1\b¾³¬2MïR²”gúwð܆.[z)OZ‘lªú\¾]µ%®;5©, ‚a€‡4ä z ß‘`¨˜T;;F¡/ý‰<úï‰Ý‡7}ÚK8\¿QHê6‡; ÜhKjÅb€rM¬³õ[†Õ( â¬0‘¨ÙD‹a9²÷ý"ÆÃEì’y1ç=«ÉkoŽ7#skêÿå”ת(¼´M*žV‰Ô`ˆí»©Ü.åúèÓÓe°œÐEI­Î{ÀÎð÷Yè&Cùƒ‹D '’•«„^O%ä­û†ðxÝÜjÏÔêv¯ÁDg‰eùg^©©C¡JÒ5ÔsQVÝwìÌ;“'k ^ì®iJ˜}aû oƒÏýN*• 3o%ïë’Šªö4«Çû"!‘íÄ) J-·)#ùZþ¯ í‚b2¶ù@ÓœUºý|X…¡ trÖ©“ Ž…dÖ ž§RÇÜ€}dI;®ò~àRù[<¨B±ò«Ð¯c RÐr)8Y|”ŸÌéôõ®^{9Ô™L=Æ_·Ú}mÊ­Ë•JדðA(1Óï9)B…E£|e›}÷5- `€Œ_̱Ë@ ÏySÂË!Ø`.ü¡ó1_GU ›7Ýmn+4veY®¯vĨ÷ê¿X‘8OîríöÚ'ßøã þX"dÓñZ §Ýÿ*RsËäŸ×Ÿ :ÍepŽg.Žb,ÌXÈÿ¾U"–-grrgdÓƾXšÈpßàX¼™ß£Ìå @~UôÌ[]9–{V§‹¸x±òWœOö~ˆ·c<éj ŽX²W‹S† ç¼Ð‡É®\Wì®'¬ `Ä JUõÒæ‰(^Ô5.§ƒ-2"ºEŒ×¬b˜õKÕc›ÀDŠ]EGýúµ€­~§,Q`[挲;~žPp=}pŸ‹9{JÔ†%m˘¿Ã°ƒÂup£:ÃíiLÆ_x3w^'_xïdß8ã8AÁâhIo´dÛøNÚCêD°âdM=!ЛÐ 2Y d0…;aò  ;QŸ’•%4‘Ë>ÄÆKúmiVCÎ÷NjHº³ŒæþNdÌYð§´@OóÂÌB…º5à0yƒ]ÛÔÌ„êI(¸€\_¯LW¾b׊Ù=?”!A@Fž«Bf+2ÎÝùê-à>~ûséeÇÁÙ ™ÿ–¨IŒ=uà¶ÆŠ‹#ë‰: Vü¡ÕAY©ŒF7RjçX6¾¶4Wó³½ÀbÑ%g:LåY TÞšxµ<ýƒBû´2Ñ5e'}8ñÓoä?ޏg£|·pöóß„š¯l‰ðjPÄdj >¶L^ü|TMìaOÊÖR·ÓœÊåý©íàŠ–g},Í¥*©~•ñuá£]s~á ƒ(}Õìíº·¿ JeÆk· +Lh4Šz6Ñï]a.J¶øw‡RÒ%B6zp ˜Px ê™m;ø•¹ãÖá:Ï‘øQOØ·GüÑG"Ñ-­…S‡›ªHr¸¯¥GËÖ‹O[a§ hkb •"¡ Š!„®é43âu(éÝþ [©;ƸÏ2Í·ù˜æ²À±OìÅqý–P ;ØA¨^•uV$¤Ñ_ÛýkwÀå‰ËwÓϤžïÐtCƒ»Œ¨Òƒâ°}\‘(ÜÀi”½˜DÁ¤(AÕ9ä-Ó÷SØÚy›‘°ŠQÐ#wF]´G—øMZ9¨Ï/8ãÓòÅ’÷$VáH€jo–ùºŒÆÀ—¼Ò??óŠ7Þz!¹  °z»r‡áF-\QÛ{7’Ý`‰É,æ–³tî›,/ÒBi8öRåÃn²úg6éV\×~5à²``äÀ‡$_Z¹ÃOE·”4Ûúë­Í²ÇNq‹4-˜o·á­êü0\»… !û|%/ ^j«U¬MQ,IÄ)Wk³õë›êg¼Bâ‹Á(©8éhõ˜=ܾæÇcy´8$’GIïòÚBƒƒÆã;F –ê 4waüé@”n¿wßÈnï­˜ —ÜÑ%A¤ ÔsÞ.¸ÌbÃCä½ s+SÝo—°$›4®·bBt{ŠHï¡%ºÔV×ÈüøQµ¸ ÒÊýŸiE“žb]Ck¿¦ÝÜ—i¯ÈÐtÙ ‚Ö3i1úb²ALÔùHªÑq~M(?#FHaº Â3šçÔ>ÅAöÈ~[0}§8; YaÃ1ò×è^^¼•J4ìÌÍvÔ±7rÞY£KˆÑ˜9qœ™{ë1S]‡3–8†%%·l´$ÈŸ@……þÞ,ìï¦ýŠ3ÕùÒú ¶=ðò~ÇtÞÄž‘Ä¡uï6ö¨3Ó!¾ŽÃ•*•Ô;6gÛˆÀú।}ƒÞüN± pt2Ÿiõ—*—Pë9r¥f·*Á”ˬr4Ìcƒ’ØbÑ"bÁÔØ±»øMßá¢OÕ[xMÐÉõî êoßùµÅ´Ê.w]Å?‹=ÞMq ¦ˆ$ ÿÁ²áž¥·bĉ};І‡ÜiÚ"UkÖA¿Œ(M\N”|—¬.aI‰bZsãXZÈrWðA5íHA9§-oͨóCg´X¨%€Ê¿ì†»JdÏf‘Mc‹Ø6!çܼ€—y駬µ’õîùÈ–ý4Œ Ø1Ýjß®Æož§¦üÒoÍ#ÉïP‰à~ˆ.ÎÛõ•ÓZ|·]g§ M¾>/ö\||ÃÞ‹ÅŽÇ8Ö•¥Ü«Úþl{TN8ñ”•,ðBB1œliÞJåL}ô½@RÃT‹÷SÓÈ?ž>óý8  B¨7b’‹5‹Ç8# ßü(äB´Bßüˆ>·MðS·FC>&*ÆvâFvòæ°T«¼ÖHŠ}_ë`·9‚£­Ç=öÓÊíÎꑳô³°sÑJÁ148p$Ò›1ß>þT’AÏâNŒŸ?Ù;¿:œì<õh nø(¢j,îNÐ#‰:2éak8`”O%GuiÞEìgŽßù±Ÿ–¼•/AÑq01EÕ²>–@"`@ŠÅ] òm·×æ^Hiy 曋–Ýfp-Ôø‘™(nH!Óê 1õ½[ä5¨åË[ÏšMÒ‹ÖÁuä½EjM`½+Bæ~Äz¢]d%’¼üë5.PÖ(šÿ•‹p[h®Ûs·(™ì(f±ËX.q‹ÖUcFªFñmœQ54”I•ïƒ}ŠsÇ'áD´àÛ+M’öôEÅ¿!YZÈÊKI¬¹Qìi…¢žÊYgÞuð?0ßÐëÇŽkÏ&›õ%ÀÄAœy¥kåí¾ÁcðÑ%$ ’QÀ±`M9‚Q€@Lú9®&ªÀBrªcH–(gA,TÍO¿7˜j&’ÏÁ öƒ!üŠßbµ}‡qã/cÜÌvû¨½0ÛQ]ëŽÝšmšíú ¬Œ@BQ#¡ïÎÐUAqŠÒªÈL뉆—†I—~Îy»ñÕb zÃ@Ç$öŽÌûÔ¡µê.T£w–Ç qoãPš>NÄÃx½Þ]^¶Ó¦tå`ƒ¤±5þøyë9€ZËgt&•3{!;z÷‡¨ºZ¥lÄAÚÚ½ü$yÄ{¯ @¢¾„±ÉÜyz ¬³£~Vb.Ô ÞVÝZoÙ=qŒ»²WÊhe¸{L.Yý²Ú‰y‹ÌQ°‰Çî]i­î~‡òÕ¾›_»âÙ×€Cú{|q³¬+µàÏK®î?][¬%£dïœS#¿[k™½°-Ö["ö¨Ö¥„Oàœ4¸ÿ¾ŽÜï ö’vE°&\gGíÏâÏ”T·wyãSºAËÊ/,ôSèÀQC𣞜¦Ô°‹†ÖàæmBb.o&¥?ãŸíý,gƒ÷Z+ÊÆòôP]kïÞn.šN!+%®÷ 0òƒã]á jRFñV]³.ðÄÂ,;šæ~¼a–[4”fŸ[™¿§€®ïŠH«ÒÐÙøM˜a´lV²£aÔ¦$ן B5so†.%92ø.GL‡R©Ë,™WŽÞ{E±t¨šÏ-dPD?€ ”öYìqZàW 0)= Íe¡«ùáÔozIÀg¡¨Õgã³»c¦‰þŸ‘ÊV~vÏ™Û>ìð ´q*xx &?œ'ýD¡Ô–õl^¾Œÿù ”_©‡±±zõróÛ7¡×Ô†,´‘C‰W:˜›ËÀW÷ÔCoáU¾Kæúo: ýÔœDÀÈ(÷Ùd@u¿^1 ×"-¶A²úÊ=YüÆœ(tm”—‡Å!ÛOy'³B&P€Ž•õV°ž%ø'+4í²ËÀa¯EÒ°ìmOvÛšPÛ·@'iÞKë#1Vód0:<_‹‘šæÎ GŒ×øàDàÒ¿¤ù5¹ê§;Ö`I€U,¥°qMÂË,¨O)3d õ’k˜ÑêÙKåÎ1ívÈ¥Ú”Á“ÿ> kªw~Ôo^Š]·f𽄇|Á6L5ºÔ7¤GL;ÙË´òy‰EGÍõILãoÉø¨Gí:Ò%©=:„;÷ gÉõGÚà›=ÝÑžB<ïâ“X*çðmYŒfwßHu WЀb+LÉȾr¤ín£S¨…ß]—¸€ óŸéÁ®dÛn4jØø&r4'áß R2¿P´ÇT¥%¼}ûQ®:#A´Îj ùgü¼»±(‚0¡çåæÊ­­…oZˆ)^µŽg)˜øÙ¡ù:bm¶/vkù…â[J}à;Á2ÃIˆ«À‰ºÛ|H(î .q„O<wó7Q BÐZ§½/9ˆ¡¯‘WÔ\pÆ"g@a9OÀ‹àÀ­LðªoFâáû‚ý?Èh¬‡ž4Jä1GqúêOƒ×uƒÓ•|ý.X•Ul¿Í7)Ô¾Ð(K7'Í`Ô¾"\bUIKlFŽˆD@ÙÔÇŒLç–m&âÛ{ÐÞî/ΙL¦¤¬ g]LA~CŠœžs'i ­Ÿ¬’ž¨•ºï'Fü±¼²‚zëHÔÙôÌ+Áë ü†æ{ zF›?gP qç"ˆp“<ýŸãÅ]Iìyöô`;#.„‰&¦1;án~–81è5%îºRB½ÉÂÉ w§‘Å0( U¸'6k}Éè& Ôuö0*‰\•¤P£—3êÎâ ñoÌ4:X ”Џ¢‹÷½€Ṯ½QµÆ%Aµ—ðœw÷È}lÿ+ƒ£ÊlFù±›âþ,kçCº8¡ æ«öÄéÿP.V`}¤l° µ,Œ¦ƒý£vaY4K#ÿ-˜:{)©qÁ<_ì™t^}<ÝCJ íŠb%/ÖóÅ({_U –‘•ÊXŒÝxßÐN‡ÁS’‹qUw° áG7•ã3®xË#–Í ®~ tdº×ádâ§ò ç©}E¦59Cv«ˆëEáe?:!ö €i¡t¹‘»'°.g:¼Çµñ©šÁÌ+z´›H½ƒð•}º}¼6êòYdEˆ—«´¢ w’–›C+…8ߺEÒ) 쮈PüBûp3S:–ÆÃ}µ‚õ¨Zmwä%°h©¼+=ZV1@†›¹ºÕÜ¢zà4§>` Ûp#ßÝ|-ßLª¤Ð‘ˆAØ–RLð5:åÑÈfážRA/cç‹’ 9]m²n­ÌˆmB 6˜sv«î—Š¡1~m¿GŒòE“‰Â úÞû˜¥þ¤ËL:ÔˆYABáúoòÞ^zñÞàȵ‚m‹ûëëÊÊ9ÜÌ‹bLG¤i2Õ½Tl’²¨ZÏHˆÍù Jùb¹‚ ÷„dYpeϹۇ0j yx0¿÷ý‹ÓwaûhNÚHqø;§6 }›u϶œ¡ä¸éZ“` ¨/2›ÎÆMì•9…‰?Í̖ۙ­1z+´â¡¹fká Ê‚èòµ=AùÔÞ¬ñ½I#"ÎOuêRi‘‡žxýöàÐ×cø'¾ÕîŒMÆÏ—sýªTßG† ââLvÈ1ý§3íhn :ª~ÌÃ'8óÝÇö»„øeŠ­) °íµ7z½P)Lövƒ˜T¨Qv²®#*;¥7 ÍI…Ot.çµ>ÎpürÎbcĀݫRWeÜv>·&ÍMlþ:Éþ Ã|Ußæ ÕÀ¬[ô/ùcúvÅš]M™1ޝ¼ÓK‹ó8!µQ ý¿Æôâæ„(ªfKæ¿Y)%ð1 ›—Ã{² ƪkð;€ç`jH¦ÑÉ‹>*–ˆáÒWYÖÂrö"é°)´õ¡JÑæõI}'g`a|È—b=﷑¼TŸEJ€ì[Á?·*!3¨àÞÿÌmkÝVÈs¿Sºc%h_RÙhX>¡)ÀßiÝÖîWÜ7ÛÜÀˆ¡R_ÜØDßñJ4â¬9:vx}ûÙ¿}o—/Zž¨ß¥æ£$þÎJÑaSøúON9 ÑÖò!Åbm”ü´§ŠÜfÞ°‘ƒƒ4º‘¹onŒó”ÁxÒ¯þUr6’CümßB؈†#ONi23²3Å©ÞÃiÄU¬]i±™†'î¿$SBšG‰1^°Åt¤¨8ÄõoXcŒZ?®í0l¼ ©TEÕ³ªùã‰/Gì’GáPØøLÏÎäb†"F­ôFj¬ôn±;‘Ò×äâØ<¡YÄí­ÇüU(2—É­Z[%ʨÍ@l"DÕ\~X•ü•öžÈÈa#lmßµxӮɌ¸eÞ oLYtíö €Æ©o/ë,Étøªãm rë™ c ì ÜaÇñú§]ë3âʲ)Ñò»uúíbAo×HI-Y22¢º{‡ÖJ½ÕtwúçQ¥Š:}Û¸§V—Ômo)‘ª‡¯9ñÉÇŸgع¾2pìKð‡c#?«ÎMÆÏÁˆ‚µ>:tb‘SC“¯hÆíLԨš“*;¯‹EÅ•ƒq„ç÷çÙ6Ràþ?_ï_ §Žª]-ZDç` ÀæÉ2ÚÛšÎ׋½G´üÀäËàK™È•Ô à±’€[q#«eÈ Kɨ`œá´ &¶órYÀÄ_¬ÌÎð¹b>Çn©ä?lóZóÖx6(0ÆxB–áðiXEŠ– rÃdR`Ë'ã&Bzb ­hK"šÎ›5“dñÙ@ÄÉZ3óˆ yæ?k Ü`ÑR¡ É ?ü¼J¸º½ä{Ђǡ #n9°iÿ ÜÙb®õŠ-†>ÝUcÒÚ­€M÷LGãêõaý<5¨ÅqÚ-YÒ2tÏgjègZ'b¡/s! ÷\ÿR4»K‘f^á± â6Ö¬O}às¼Ï¿ØÅ˜†¬ïàòuÑŽLYÕ‰]À t,µKÒzÂX"Æ.€T) ¶Èwº>²t¡`R;ÑÃA”"ð¶qôj× ^˜J–:µõ¹êšF?³uìiˆrÈŽø´êÌÏÞr®f;ÎË·‚ŠJšŠð (“fx"”v_.$±%YSÜÔ€Zß×ñ‹™–I ¯¼¡Ö*y• Ò+-c”.WÙT?ý^)ÞCŸükzÍMޏYãƒ[nh¹»gíyïX§þÃcG´B¤Õž#x«b¯ƒ”¯·mñfØEH¶‘ÓA Lryc8ºBYÒ ›Jí°¦…Ô=V~BÐ0Ý@ís‰ÞIN Œb˜ø…} ’'üÃp’áåxz¡l‘ôàÝ@°æ®6—ZsT'óñ’çÆ©Ä'/í6ÐŽ\…£ª6Ü«z.ÅW–¬ï¿§:”94éyÉA÷ÛA¡aS@ì\ ß¹“ë»}C· ¬Í‡@t@ЈðúâóSʦ¨8Q+èʃ5ãHŒ«œ¤àý · ÆbíŽ ê”>ÉÜ0Ò<íd³i ë~ ÇÍ îêÚYBÚ×oEÞA Õ«JLÏ;´Ï‹Nk+n~%+Ù×IÆ&¶”D´¥°žòƒÊY¿„®›×xª0 UËiTªÿ>²H³ŒÀ7‚½è›ËŒ#Ð+Îø²NÂ-¡&à g”úaÔNnx}¶¸ˆ‚÷ÇëË'å˯/Ÿn ®!Á¶–¦ŽsEÞñ‰zHc$Ļӓò†®çû檅ë#R¸Éhñ7TeÔ8ê"!ox:,ýÜOrChgËÌÑÎØ^h1&ò”IñîVîÌÕƒŽˆX«K¯§a¯âÅ„ýôg^„×Ré'kîÁs1Eç²Ñ7$»ù~D­¾Üš¢I”¯9õZ泎¯FõÈáÂè‡sŠd> À=:$Å­6#4EÛhôMý •LÒI§ûÉñ·2]Æ ã-¢…ñY²[t.MX*ZæR.%É Ûp¯;qîyD²dM"ñæTb%ÐÛôŠ äÈá6(¾£'ÛI±j¤áœg…,Òâá6& FUx“`£U°Él¼åPÑieCcüåzg€(~Øûrž#úÁ LC,E.ó²}ÜIš)éEsæøõó%puÉ5úÜÏ*ô9ì¨0Ïrkà’œ$¨|R] (`lV þ€²óNâìï · 2ÌßÊû/Qö‘¨Jª( doüD"3‚šëû½­ï·SG&&}Ê^¿Q6 |‘D¦d é»Î¸½N5n ®è àŒÁy¢?>–U¥×§æÐ €|pš)Y›C”5 f:*øý: G–™TËk3V›Îê‹°-›c èÄr鉦æH/²Œ™ ¯Ô'd‰Ê†±%ÝSqP±¨39~o?±bJ¯³4öÊKÌÚ<Èíøýˆ¢·Q â«MÞ@ÝÜ z¨!ìí6Ÿâ÷ÝÂŽ“\E“ÊÉ„a?5TEûžo`òma<ˆÑâiüÓ`Âhx¼‘¡ž*ÿRÓ’×ù6…M¶PS, Çôî€8/EDÆ3!ÃÄ &,¼ëýŠ Ms–P A>©îªHøN`¸-Qh¿%Ó¬yDÖ4µ†&]&•Âî•vüã–è°Ü.j96ˆ•#Kä±XgY_3;~„¦šr-ð.û™¡!w#)@“@åy˜ùÌ{ÑGoJ(õ§°dÈI+€{‡Ä)É'<ß5å⣘Ðù•¡ÜÊ¡öŠ7DI•k™$,Ò…õª @ÑvRp2kf“õoð{0ÙC¦b¨CZÀ- Un+"vdÏa3 |?dô«4í&t¡Ê½»›mãÞNÈkÁÍ"‘uHë%ŠêÀb3`„Án{éùݯ«G“W'dê…Ý«›ÛìŒnfúúà¸Ëw&ðІv&g0º<,c¯™àæ@©‘å`Ìvuž|F§m>Q)y ¼>à¸p\8êm1í râÓ)ko†=áîÜ*«+¥MÊäY] çå‚qZ«çHN%/}‹l;褪¿Ò3zw‚‹òŠÚËñ’w”ƒCïódE]ùÀ$Þ²13"Ñ )é’OМEÞn­¤ë7…GOÛ—¢Ì­±Ç´Ív™²¼·sZjÂËÓäA'Ô80?%¬E^éßÍ{´ì ¡vÚîÔ³â 4ÒlúN6lÛêL¸´×‚7 ! ¼=]ÓØ3eÝÄç¹.8؈Èb'²Š®ƒbÐ)ò¥wºÿ÷tØ ¯DÁ¸+§¾ØesÂwÀ½»5áê3ƒR?‰ßpDlÿRã7|ê¿è#°jŽP\e x)³Á9ž¤ ÛcZŽTi€vmøb%µ–rø¡L=z›ù'æ •& g’‹ë–ˆUwNÀ2¢¬öÛ¼º|oÿ­Ž<´Vœ%ôKÝ*8Dåu;i¯ã>]+¯%Ê÷ævreN)(4ÎS8ODíÇ“©»Eë?ëÜÿBûPW\‰¯<‚7,j]êš`žý&Ò2ë…ßœe[2£Í,Éê‹Ïc‡”Iî·ÿ<^Þªˆ‰>“MÃh]¦´zÓcêPyð7ï²³;T<¬ŸNþË®o9î§«¢0jÍ0;©Bt) &³ê¹Ç’ÝæFª©í–‚|¼˜•×1Ú’ä6„ùò PÀm#€3ôP, ÚB€ž:hFÜA³¤’?ÀÿŽÈ…A!CÉ1PÅìBWô’ÔâMÔÒS"LI‰l¿|Ùè¢CQ[íÚ{Ýûºb—›ÊRm÷æ"ó dœ2šs’ßÅJÑ r2y¯~v‚Ï¢;×$ ¾¶kh|iŸeºAyT_Âkå½qZœÉÀ±ÜNʨòXÍæƒž'×M è*6<Åúe—CÁ=UÌD  ѧ@ÝÓÅcàA¦7&HÈ’OÌw‘”@Ó^gKgÛóQW0ù¨1‹°ç9N£-âð× ü]ˆÌk^ 9h%Bö GiCdqŒËrg›®½G“¶Xå‚ÒÌØ!ðîÍ’s»°³ rK×S}µª]†ºI !¡TP¡fø`P7†9ûš,õ”Ò²‘6ñ¸ Z¶îÉJ3”X6Œ'½1Ãʺ7¨=¶í…,¥m®’ŠðÉ#í}oXõæ$ ÝÜqÓBUœ"ôiǧ©e´t«¯ûZnÂï¶™'¸WEÅø|å‡%…3yvI²ªE1´.Ì5TÍÒ NÏäc˜ÞªTNÅi¬‰ÛjrÁ=îzv¬O§ÔcŒ§Žì‰Ú4†E3ÌÚDG/Éä2»+ñ~ÙtÍ”åY µ¤øÕóiå1˜ËÉLØè©ÂiÇu]ó,`¼S26]K'Ÿ~Ñì(’ZÑ—’Øú8†£²\jß7`€ã 'I°Ê|謹Œ¤)nƒ°³¹‹l²c¶±˜ødÈ[†¥¸¶áÖ#³¢2ÌSOã•Ñø‹2̸ Öê-N]Pb3ÙnEÞOºp¡q—a¬¾¯j`åˆGxP‚- ²û”˜im⼊ì‡áÕâºL÷ üiyŠÈQwÜr|÷ðãV[«JvðµJì=ÕQòàÊ=Ü~e滑'[øéè ÒCÄdì%\C·[ÝûÌFÒöD!(nÊò1|¡W©k8çßé‹Å`nѲÙo»bíSZÈ#Éš•‘÷“HÃ{}¤àÿ3Í[ ²P° LXW¦>ñÖ´sûg!±Ø* ck†"ÇòŠ­—®ó/LV9¡`BvTÍÚÏ+·P úšA-â¸b¤ý5{WÈÛ–ßý±nà¹öÜ €Øí“.ù!^h¯ÞšYæ·€ò*Ä¿U2Ĉ•œîƒG¨F›§©+5ͨŸú»–¬è°zR›EÜ—“ƒ>$"5JÝÕÀÚ}&ê$ß°¸NÖ!æhµ}œ¤Ôn@n3Ò {BqøäåIÇ=j]æÿ¸¡‚œzS~$°\?F_Þ²[ )Ùè!RNÍå/„[UÙð¶K/céÖ  eÄ,‚õ#“£Ìxàj:{¿C;‹aÇt:*XrtÄ+ —åÕÃ"ª9{Â.k¨Š!WL9ÕR•é¨%š*x7$DÊ£ŸŸ)ä3×ÉFdÞÙe†?Ú„gÊð&˜a:²B|0êŠÉèx½ÿ‰Ä9”–zNœ1ÿWp¦Oè¹ÃË(hmwUzÈéõ:œk£!§T_-ð}ÎMjȳ“°¸ \ຑȷqï›iÖ²(À²ˆ‡p˜!!%¤Hô§ªÉ,'lÅ㑺…EßÎØkq£ª˜Üh¼NJ/ÎÅ„à‰ypÝ;Þ›E€uØ~‚¿¹"Å4«µàÔTÐÈ)¢†‘9m“x67¦ùÿ¾ SD–×ôãØív–± ÿüGi‰P!S š JÏ•’bîœC°>‘1»ïDV/>•¯‹k(ÓYuq†æ\¹Úø`ìbS'öÏÃûÊê“ Š€{^IX”‘r„¾g¬jX}ÿܽ=ëCÝ…*š;Gd ÷Õ(²ùÙ\ð1à%—¶~1”àlšû¨Áuͯ.l1ãQ媞gœJ¼¶¼é¢£Ýå«ñ5fǦïyšœ×ý&…©ùÅË:–t‹Î½ý•pœ2VC“kÆE[ÄÚ°zù—oœ2š°·Ò˜1ü‹ ÙfïгµwÄ]’jg+Ë­Öúéõ̼T ºê+_’èêÓ[b †CÛKHðÈ’w´á]&@õ?xŒã ƒ¤Š>ÒÍuÎÍß…ÝéU‚WYF*ǘÃ$5‚8Ž*Ç"&çävU­÷• 'Ñ‘ë%.…FÉ–„ï‡lò‹×J”)p¼þ”›Ù¢¤*\ ö‡tÃü‘lyìDZT0ggymì‘KPÁ¦ï JJt(èV6]­Ã´OJŒ‚ rfèoÑü¼5U´w9N§*ÛœrMic“ËõŸôŽ’£„LœÉ1óœÅ.„Ò6°)À7Õ)3ƒùdßZ­ R÷à s—Çz‡Á’ßú#ü¯Âü¶|©áU€(£Vlœ”¼À g§>* 4èá9,=w–¸`ø;Þ§uôNÌ ô«ƒ“tnsB¸Ÿè[t±Áz¨ÇŽj A/-ø²Üt•s’dGk›·zgBÅã4BBzÑ™Œêg•¤#Ú¤ÅU…-còœ ÝYŠ-q: °õr{ͽô‚U¤×ú¼ë ÛòÈ+áÍÅmÕBðÉ7ZœÄN±w5¹zÔ!p T˜Ö `ò$ëš–”‹Ì”iÂFiƒ*ð¼HÄ’@vþÿµd4 "šïù)<ö÷ø†ap ÙU ï%£‘øey:«_?¿T«Qà.ÿ™I:Išjž­X|%_ðv=óc‚S˜§ÌðÈ^y÷»£O §5bYV¯…¾UúE`cêPÔ:Ê•;øž‡ðÜm ùû*›÷B—¤—Õ`¬6¸ÙÜÐ_?¹œí}ñÆÚµîàh5ì+~™B¨Lµ¦ø÷R^ËmÛ•Žô[¬6&:  a|mN +Á+´Þ+Á鹃 ¯ÿn4îž«iHªI~’ðÌ[¯èßPÄ‹é_:‘‘v›'ÞWvpwTûiÔorw1ÃXºöÄVÆd3£pI&PiÔ‘ï”éU zrºGþÌ~bÓ68‰“E5 ¦ß Úï`Kj®Ó¸INR,ûÎŒ¹ý]þ7Ì98ù2#L†—6 ú j~þ&_?)Ù)T_hkžó¸Òy'°$vJÁ¿+'4x{¸£9H”Uáoaߣ¡¢›}b>‰~³ö/„öHæV»ïíàêRçFÄȄ¥eä­¶Pd?ã…ÄïHÝÀ82iRm¢`ž'¾OÙþdƆP¸©ëØQ´É óù›hi¸ZTº1<µÄô5Í»ÐìúN¦¸z?µ¥Üšµ‡à±Õ`é²¹ x÷ü1uT -j&s6²§K­ +®1|7ê …œ9­¥ïä¬N¬…£ƒ1K’ÁNÔQÅÔ]œ|ÌËh il®¿ô Œ–LSjÍ‹3Ív3 U:K€7~õt£ås÷ü M¬–ra«î b­g–üõ:r,ïç¨^Íím+úâ‰Õá>8qÏí¦ö ¬F¬í0Qa_ðŸ©,¿l-z Ìv¨‰Èxò¼ÓQdx–Á¥7,ûÈÝ;å­›à`ëv‹º[â Üû‰&É: «~0TB÷óÇÎÜý8‘q €te¶SÿcE.²>™­a×ÖÒ^=Íž,ëý{å!~hùÕ é@âË™×á"äc=vmbhcb©Z;”Á•üÂ>´È¢;OjŒ_Ï ½Õûñ0é=LŽiÜšÖÈîï²'é>ÃWЋ·¯îäö•°xðß óªháî|Ûç§hu]‘«¡«ïfRBõAy0Èð~ËjíÒ:;¤ûh…K Z‡£œ±¸;,uÚƒ¦˜Û››3Pzû1'¨¿!¦S…ëj´*G)³¿¡âĵLØF‘¡:¯¯ë¬ýˆÛc¿ôïÂòÑ9ÿSŠZ-š4‹CÑìF¤zɳÄk+£Þ G]@¶±¢ò`ûÆ'bCØÈÃÝ.µ›G™•‹b=`¬:ñ7ÑÝ©è¬*3·ºšˆ‡Õ¦Z› hÇ;1$¬Üð•ÀIO'…eä[ђ訅E ™t‚`Ò~²ü„‹Þíž™SŽ´EgîíÌÁ_à[ÁÇ °ÓbH,ùŽøéMFÕ¬¹cMýnW—ƒæ¸o,·Þ5 ôjØŸWííË;7Û$@˜³Çk0©Ä„za•ïXŠ2,Wl¾5S€’¤çaUÁ@³?:ó)s°©¹(bÌyè}g²±m *¹îòS!’;ŠQådžY­œ ·”Õ1¤ÿçæ€çùäëþÚߺN›Ÿ"=Ñ&÷³ |tV ù\Ÿ^\'Y]yjÏöv×Qׯ^cÛq¿ L"kHC[¬ª“<–Ý mÝùã?œë]Žžš® ÑA¶»Õ¢ñ&–¦ã?’òΫ÷#»ÙM { zôëT8D&þLØ(Al¨½nIâr#Tó•*ûÕ?b×c*´‹¥wèÓÃ(}`P:{fž>30ŠÅàQUSúÊ3KÖË-âß^DmùcÞ]‚êGÆ·ríüÊÉüUBõ;ñ¡šÔ!ƒôðD!çúˆ¾bâåÑÈëÔÝ¢izmuêga.mù‹ÁjlÞ˜ôF~½›TÖ%«aP n8 vßê¿Ks# 'È©l¹½é—Ó)Ôk¸@mkO'x2Á¶ èþ·ÒÀ>õ^à<Òìɵèkué^&I½+¹àîW^~gŠ8gv iEâfGòHö"û(Ñ÷’_Mý>SªíÿÛ•aÌ"tKc:6Ÿ5{Ä\‹ßW¡tc]óë‚°š)Ó¼—ŽÀ蟅Ân…ïEÊôó³‘om#d¢%O Gà—Ç$h†¶&:¹QãÊ»Q}–êÙ”¨á£%ì¹)v^×´¸KزÌÓÞ¾ ªÞ‚Ì?ú]±´Þ5aŸ9½úBÊû…jÃǨô+æÞS ÆŒsncõP8gú¤à%}n(Ûƒi&qÙøê|êE^ƨw¼ör†t%ÿTä–ÇEön£Ìª¿`Ö^/|'9Ùox²”Xý榜áËNë¢2Ú.c#ÏϤ_‰ØJ!ºL'¢á÷Vhé@@í8½">-WŽ=¥¹<_rÈЩ‰`Vá2 ò*æÎnnDåEo«/iÊ—ý  åQ<µG`˜‰6Üšv [_ÕrðW“Š@¡XEz{K(éw‚ò=À¯ÞT(‹ü‡”àýúrÑ)G”¶¾Ùðýéçui ÒÉB—¬G*vŒbÓb$ÒþºN,çœJ©`HOªTŠ™+¥;cTJ3Ü m>EˆÍäòú§(h¨eÊ^‰­|Á¹í¼‚ʘQÖ~`:4#ÎßéXôçÝÅó&Æ«–ò «i¤'zmµ8 °%J-¦±ñ\8 ÕËKÒF:ì¦t÷šò*2Ͼæ}qÑ0k¡°ÓÒš^x—ž|`¯7'æZ#åµì°ƒqÒ"míýö"fîrµeŸ ¸çl/3…áͶ@и÷Ðm4Û§)pê×™õ66»“:þÞË/z¼¼ÓÒ:ºó:y˜É¥B°ÚI4_Ÿø¶Ù©õ¹ÊË•²VÄUg òšyõNÏlÞ„§3î 9@^º÷Dü=Wû•Jc‹\¾Gµ¸ê|öÅ Dâ¦:ð½J1—©Ýà¸|…–nÙ¥£¥é¥°ðR7nf¼¡Ö¯Goú~=£Î¯ê ÔâHWÁrŽãÔOõÆ>Ç\#¿ó^M oë‹ÄšZ‹%×ÜŠ¦2û]Oyqk©qÖ‹áPÅÔÜUKÏôPvd¸«¤ fz*²IV™Ö‹ÿ;í”û>WŒîï–‹,oyÎŒûµž¿äžs܉R\C%WH™ú9æP¸ –ÿ¼:‡–N[‘¯­è÷ÛòM,÷î!ØØÚQúok„PE_HO/]¹X”W<»¸, ˆ875µ«¸ª»ˆÿ÷ewñ$Py>>’p!`®-¤9€á+DÔµüDÏYúÇkY ÑŒãé]—_¶>Ëc^¶w èçË.wE:À†¥zcgT“‘ÿÓ£|Ï.{‘Œß=]Ùºÿƒ°¡½àP‹aýâ2Áàˆ[èÕ*>` QGœ¤4éI­Ì¬zÅ©ØÚ(ƒœÑ¦Ø#×\‘t[ uõ î Eû"ÁA®ä"e^™¢e¤W\å¾v.û`¸cE€gœ¢””;ãUé1Õ†¢e‹æO<"}½ù”Üûýà•Ó˜’צá&Pc Ðè‚Ð%âÜ.Wã¤!4â÷NSÁÐ+Ыv æ=ûxòÁè­°í¹x>a¤ùE› {Xæ •ÕÆÂ†®ú™Dÿ¦ŸØœ©‡ûÿýw€€†­1ñÏ+ÛEGÅߺý3›WÚ²¦<º£•e”j-tÆ*ÇJB‰Â.‚Ù^ Z³éðƘáည¶€ˆ!Aí²Ë_]Ek¨»4½c_QÌY 3Ê×^XZ§SqgL#ƒâ”LEã䞺”ñ‹ôÜÃø\PÈÝüxTp›këïÞnÊ"UKáPྠúíùsÍX›·Ï«ë’X/ãÁÓ~m–e`2§[6øç¶¥Ò" n)ØRŸƒJ£W;¢°Ô”}/2#;Œ,Â2$Z”ôiÅ ƒÑñ74—-àãEo–‰vs¬¥ónÍ3ÛŠ¤ÂøI é9=ª¬tx² ˜X™öwm9üïñª%“I’XŽ{ˆ€™=ÃíVµ*@aœádã%–ý7k|_º|ˆî´&ûáî^|2ùþ“kÿ+÷î8)^I ™ rÐjÜ㳫¤l¿5‹3ær¬«!è„?£¶\‚D2›®ïíx[w³ãâõ­µ¶1pÅÙ&hºLïÀCä·Œ¤ qh¹zꉬç˜_>õztó#škhë”/€ÎãE™¦}ŽôUä"~šmì™›Räõ²ñŸè0ü(FÄÿTHø‡]¿"˜<êÁ‹@yã•FƒÓn³ýa–Éø&¥Û†'•õð¥ +äæç›bç訳ŽpþNÐÊç9×ì˜åÐØ)›Å ¹“®ÉêXc·ÌXñè)ˆP°Ä' `iFüc°xÇ^€µ[سl ÚCÀz$óyr¶TM~ñAqÁÓSéxdŠÍ°qèÂŒ{ £ñCÉä@Ru[ë,/퇆\ÞÍJr°Òò« pD3±Ãÿ]†#q€e ×£fƳï§fXW“Vù± ‘Û%HÑqÐb´š¥~ܯ‰$R²g(Ä¢3Í×Ñ PNöH§–î¯6÷©(+»â'týeVŽî‹£Wù,esNUa@!£P†U‹Ms#üÄæ&!àT#·Ìf¾0 Ê%Eÿ¨wÂÒz— Ž‚Æ&a}^‹ééx–¤Ê¯¨³ªæFh BjÂ}óRì6$+ŸS«8iA§ò]{+e.8E²v–¦´_Y®Ç¢ü"î+Ò¡éÏB­>®…Ž¥ªk¬ØvÉÐc²!§OæUÓ÷KN¦(HYø3«XR Ûý<È| Ø .wßÀ¸¼ÂýêG%õ®òqyÃàP;Ó!Þm ›{¤˜å."Ò¦…}ÄKðŸµ9˜Ž,q}î–{‡w@¦kn_˜#ïûõ¶Á­KûåGüvMn Û:ŠLn hê;Ãð?¼éϨÍûÿ— 1VþÚðÊ?³^Ô¸Š¾oÜìv ]åþ²|°" >Þù€GÁ(&VQIšˆ¾å=pè^@â¤1%ü¹ýîÀ‰;µA«©©—h$§ð-9*É_7úÜu WÖÌu“»:1¤ Eù±¥Ùÿ@¼á_^yæ 뻼©yñçº_»FŸÑTeàF¨O£Õß îN÷uÌí*$÷¸)‰NMH"Ó»`”¸^%tœ#é%*~¦i ‘NQG¿ƒ ?"wMÑ\.•‘í7èrN¹£×Û­¿÷÷`[9:[Fø™L\ s‘—T’Mð.9®/­&‚oU`öõÔÌ(—«¢/™€s5N}Ar@8©Ä­l·“zä.iª(ŠKÅ+›#NhÉ-ûaBH´þ$´‘SĽ»U[˜À([T!Ȧ©½=€&{üË…M¥˜ËÈÞŠŽ €×¤¤ããesûÈ…ÿ“ Æ¥¸ H~þ>5€¢X–bBsï.¸xù¹ æ2, ¦’òwÁYÃ-§¼:®Æš4ñçGG [W™•×¾;È Ç-c7€›þ¢s»B®3|Á9ª¹¦Fðz`1]c¢Vï_L…‚x[®óâç³í˜”%:“7e$i_¤¿±2Ù?!¬ ƒÉ¹©pš“^Ê©ì*òçÙQ+­Úëd9û¦†p ™:Š0 ¦_ŒãÛkÕȧõð¼?ýøæ×Û^dÊ_7m—y™†ü°g ‘,{ …æ\µ  4dÿb¶±Ú@Õ,ò&[Údn˜è”dÁ—ú¾<ÁJÀ*œÙ»mP=, OzÂpUŒD·²}ˆäFìA| æö'7á_ØÐç¾îй|ØŽï®0ÁüP›¤O—†.çG\O²v7®†P<†œÞ•AYË+IVÁ´Ð¼š\×þÌù':‰úâA`¢Q1³DàÅ`VCYGÐ0qÄŸ€Õ‹oŒ^ }¿ŠóJ XŠ<ÉÅí»cã. ykÖ¿A&ª=×ÀÑNusª"ûqÔ˜­OZÛÆ:‘ž°-4ÜSÑqpd1BËÕóòwJÂą́ôôÀ«eMvu¤i4zKø =/\h„å¥åÔ»5hH(ع=;   I¦—(pËË« ³±$?½ö¶TzÃAµ(LÛ¹{6£Í•3ˆâ¨º<5@©ÜËȬ ZWð|Xx{“æ‚—Í=§”ä*1lÁ™5/õÅR µTа8\òåôØŽèÛ.>…¢iu0Ü6=¡ß~))ÏÏßRvkÿ†@õlX«Õ»®}j­+¥ÛpÊŸ_ê!¯‰ðMf äß¡ÌÔúF3rë°oJð+Ðt1h¡’[Úx@.8P“ájE¡—ý>2s¯‡\áæ¹¥ ›ÄÓƒ¸ÛØÂÔ[öO1½¡Õ­;"Q-´"þô àëœkÈÌ]ŠC'Î^2aʱ€ô¹îˆó“Å=Ëþ$×Ö¨ Eéï-uÛŸª¶=Èrø¶ïÃâ ,®’úÑó<ãFé­1A—a–~8[(EYf£î+ñÐU”†}¾ÔYq¸e!Þ$W3”YÜ*Û?WáþWS³Í§iÆ¢èf¯ö;À£††eçï^QÄG”t-xN~jx’NÄ 5Ó4Îô‰Ü³^ûòûÂj…-ÅűýÓl–¨'ú}9­kâÝ´ƒ#‘£¥(Ýè¦4À¾Ð¤ æ¶—ü^²W¯‡¢OìXfü£¥ eð{U§¹(Æ7ÂJ*7ºw/FÔµN<®á"(j9²J™¿Ñ]¿«Û‰¤ñ(ƹ À-øRK™˜>©¬dÐïò$íRæt%(ÍéÑ ˜ÿ“ßøâÜQ`sSd;q4ùJ{=Yû&Yé¾ÞªÉðé§:U}¶ZeªdY}¹JËŒX%üy=/uà×Y{W ‘¬4Cc ·«`ææ#fl¤…‡H'¨ŒÍçº0r¸P¼’V&'££‰âcûH§nìZƒ•r§3wšÈ~AW¹6jlï¼ý ¤[?@òZ¤I>*oé \Õ’Àh?÷öçÖkÐô „Ÿ¯uP·Elõ¨fmÕeB×rár©¸!B!ÞáQØïÞ™Ô É0úPž$ÇCÖo ]м‰ cð>ô_Zº5ÂÛ#uçæ ä‚%ryqˆ9*øHÖHv±Žö´·ÇäH6ÿ·¯õbg¹²@AP…p„ŒõqØtÄ ~¸!—Jàá‡Õ bO]ÖU†)ò"Þ_Þpz,Ô—±¾}؉^ê²H 1žî, pÌtÀùødB™åí`ó»É vϪKóôÿ¼‡ %0pÇwA3jZ¼-mØÍ£¡—ì!;w˜^•iél<æxÌ´Ëó”,9dåÇXZ —‚Δ]VIë/bbhoä·²Þ‡7ͱ¦™2åˆ0_7Ð {he•îH}Ašqäß°-•þ§C]06ª¢È”Aå7ô° É‰þæ Þ§7ò؈ÏÎÆW§¸…RR>)éàlso"Y¿'0 U–@?xn®&t^ÙwIÂãîeØô¡£Ì0‚\YkxÊ?Î<£tÐxj €Ö(#Ú«ÍŸm(D# mu±4šj¾—2[¡H_òu§¢Á¶i¹ûHºÈÊHÙ¼°Cpü6ÙÀr׈ѭLÇš•·,9ÝË|§¾—¿Á+â›4ˆgÈÚ¢w†•±›ä?xìÍ¡~¾–ˆðý´·lz¬‹ )rÊ–Í–Úƒ“BIMw/{W ÷«| ^ fÕ~ÄüüÁ/!Ï ŸF÷Ða6¶–xAÿÿ{’€ 0ä –(½š¤!¿ ø9-ž;Z® z¹óòYæZDÄp¸u0ÜeBýìáÉŸd9†r‡®;ì@€7Ÿ¾‡‹„C0#ˆÖ³á îk¿û3à Ù'3²Âز̫Z»Ü ëC IÐÚ¸%逅tFáP=ydc+!^¡‡ir¦2¥^qB¥p¾þBÔZ%!Î(© ·Ö3 h2‰÷Ò¡0%߈_é±ù*Óɇ%£÷˜¿Ó591:TÁ?4\t#(…H²_×~r’ÑTˆ3g‚íd¾ð Ôåyˆ¢T?H)T+£½ƒ¿$P†N¨jÇ|(EWðâ[(ûccxô¬ú)EFçM1Õç4wÏSŸDec“avCظŒ<ÓocR­=Öu˜éóåÜ}pŒ¹ÃDPÆdJ½4úZYy%Þ&†ý.¹¹Ÿ¶EgòAxX( 0W¡¶B=?Å­OmâÈΣKy±¥‰yõñá2ü®ráï„Æ‘ˆTv,SuªÇ¡q© öh`%Û„~d- þËúÿÍ ¬¢}ÔK®ãˆ­êõ˜èb©ô° Ê÷ÙWÎW$œØ¯XžÒVÒŽ¿÷ ‹E@'æ÷5  ¢ëQƒkT>Âáû¢›\lJ˜w†“ÎäföâI2; Œrâ!ØkGd¢¬èçvˆó€µ‰ÿyæ Uã*p+FpäÛÝò/Œ:æ#Ì5l”7rïµÇ4z–—Q9™?½Ñ³”ìâ>Ž•¤-Ÿ) Å“• '…©âÍŒ•¼¶4n¬%‡GAÍß‘¢çÇ©m‰Þ?­“h ÄÊ«)·?l&v‚­«"yœˆ‡3°'ùî­n&kXä&i’1ôInžÅP¥;RíXå2G/ª¦†Ö™Ö, SdÿDü83°…‹‚¨¿ÿmý‹à+WZUîy¬f‹ÓïŽõ¿4…×°z“®ÿj|²A#«T2X£ïõå7W>AÊŽ‚š·RJÏdnv&®êfôµ"fê ¯Táj]e˜AúDÖNãA|¸#_Æ* À¥qší­ft­ƒ® /z²¨Ñöö[u݃ò.ή&`+H•øí£¨L®àÖ¢c kH2.‚ðyñ¨v“ÿvíïP«"æEë-8°/£=¤èC¥I^¨ü7Ìß_`ð„£F"ùZXÄM_!5âÍ8N^Ò³|ÊU´2BÖ´ÖIðL¸XÓ^¨H;qØË‡«}°úÁÈŠüð?¸€+ôzµƒ’´5=–‚Hä!ãŠâ6i3óÃhÉ;;yǵë üi,Ròjç0šã7:jèrn¹ÒÊã Û/ûªgüÛª-r6áÿjêÄõÛ'£‰/ܳšö_ÓÝ¢+Ýwאּô:ß»|ì¿ ømLÝäœÍÎ Y;r¥7 óTÍ´ÔÃ}ÄúöÂ_>…áÁWû¸/K@íÀì]ÛV'ŠR!âfæ^)ÕX‰3OS¦Tžj])3Ï„Šu®¾h»?;,6¾p±ð›še¨/öÑãñ@q«v¨?1^Ï#ƒh¥›aýFª³MS™/âq3Ì_Hô8EžòD.R›ੵ3IŒN.Så€=ÙÝŸS”T£Ð´ùz§;¶AyÝôI¸p4è2 »rýMàØsŸÙ>ôx664ÏN(u[›íâDwè_>½e³‡Bí׎Uë²êþ9ÀƒÉ"À7Â:¸–¡ŒuVu¢¦{©¶Kx]¯Ôvt^¡ ªð­ÎRd|‹ÜS¼·Íìªg~X|cN§JÀ8ƒòT/Ѧ~œ¡]°XÿdwÊ$ý!pX7Tˆ£šÒW$wÉ^8ó©`¨eÔ¾©–ƒÒÎZxæ1ËUžÃ¤š+¯FSnª‰cÎÇ7fÀœgíûÿQu–ÕþüŒ©2Á] a¹C»—;jItúâÍ—c‹UÊ+øj§Í:›úCðthû¤´X¬mŽRáœ{û€§Î¬oSçö #J"œ{ÿ¯lÉb>ªGó.RÜë at‹± »±KíúÑÀ®qaŠBúü 'a˜Ë=ò¡Ñd»ÅŸ$)§RœÂ®÷¯<©“@´7¯i#'@¶s¦9w ´>¾t7º«AÊÈN^P½µ1k¾PÐ#‰œf榺ÑPÏ3Ôý‚²µ&œ‘u¡ªR1un” ] 4²Âê×wÙ«aÍ\ñö͉]– °¡·˜äðÏh~À{4ôïJ‚]Ãz‰BߺþD“:ÚB@imÔLz(¡3SšµCÖ¦îÓA!óÖ 3%O=pB@>E¥åg@£_®_K*<×j˱HóYhÆïä‹zr>Ê ù ¤“ýîT}šk'tq÷Óö¹ÞoïŒA—\3£Œ”R‚%$d\“Ǻ¹Ü¥H$Š®CÊÕÐ Ç“ÃpÃ{HÉr×\ÿ“I) /Û`Ó%úÀ[dö‹×o%×Þ5ÅþÙΡ -péO n³X)#~;H­/4z è?Þì½"è4¿ã—>O´7ð A«Á¸Å³rËjã–Øg\pU2ý 6m_O¡Ä韺!A£øsGT˜N!c¿6s›ü$ºÞÞ1UZ¼?/²&‚Û鶆Å2‡u7šnÇq ÉŠxÇ#¨>ËnõI»ƒq%ƧÝÄGÁÃDìÄ‘W×)/yO·¾rÅVnb(èÔ(³äXÎ-?ªs¾”A/±cÓ3Æç ù¨ у‚l#„D»?uéÕýE+à ¥Ìoeª ‚s(¸ƒñòr ²'¦þ›÷Ê[¶-Œ»»ËU[¦ >V‘7Öx÷Ò²dtË™ævô+9K;2¢1© ¾lÔ¬u‘S ¿‰oVM%CWæ6uŽ…-P,Ûm²ž0AB7ˆÊ¥À ÁiòGMþ3ëL²–›ŽÁÁ¿üãmH‡Jñ³Bø¨2`[#3íбìp‡þ‚Ù’×âlØih­I3éj£ìVcÿSRãOŠÅæðÖ¿¯3”ZÇËæŸÐ w[=c]ôB¼Ï›Ÿ›çcÞ~ ₸Áàßði# ÑERÛñ$eì߬Á;Çð<‹Þ€Æ÷Ñ¡Œ¶ S(b$ÛÌQ5z©²9X“SÔSœ7vfÝä-R_œÍY¯Wzø¡ì7G4Ê/IžÃÿ‹ÂOØ)}Ôpü^5Ðg‘ÇÝ«|°K·>ýùÊoßî!Þ¸dÌ+~ôBveÛvÙß{A­Ôèl¦˜»™yß²T¬¥ƒˆ{hIRÿ#‰XSÁäÓFî|ëb€Î…³M”¿µ’>±íoö|òL¦Oþ„wiH3,ZeíÇ–Eñrzm‚ìRo^êwQ¡d1uUÞ—Ü8>wË,F¯ª¼5dj~ÈퟵÍãCÐê¦þò[¤°b¿ot‰BÎ(]ÄQP/ÍÂõÊ]ЫçØ\ÑÆÐô÷“·_0©Ø­%€Éb `ˆ\çÿiš‰ÇÔutÔWù†EѸÕï`rd¶ zQÈYö[ÈÊèPonuïMp¸t‰†"'þNï îà˜ˆv9tÿs¸åiùu´ºV‘ʶpQÐ0|”L8Ñpa¸«ÿ:M&ŒM0…4ž`¹ê~È%ˆÿâßÔ>X½炨ªÓþxÎñ#/÷¶Ï­š…ÑT¤]UÏ7 Äx(+½%Sàþòg™Öýó%¨8ÏL„Güg~ ÃFÝ;þ7Å\õ ü˜%:•y€)mY{º¥…è[å§æ ²".ð¡ÔG¨ÃÍáQD^ FocëF"¥ÔN•H«ÆÒÙ.£›¼8=\V‡Ëµ†åY÷æ^`Íe”~§ ~/§ì šÝ,º… !cfÞÅé>¶ýŽˆ’13ù(w*ì¥nœuArÆÅ=ý™z¯_kq[ è1Ôÿÿûvòú`£Ì3:¢Z}Š^ŠÛ×r-ð_ýMþÈ“¥ÏØcL™[È KRýš<ØŠqÇÆŽlÆóÚÍT>/³$¢:Í«‰Î@aÆÉ,fi^eŒ[—ÉÌBUF½†'Ôï"ûÒn% AxÎX‡Ô£’ÒB²5‡ ¶€㮹i3®’¸á¼ºlà OÂR³AÕíõ8¬¼ÆJñài±™Ü4ÒøN™ã[eÿߥ¨ÀzTÄÌ•®ï£]¦Rë=€Ë† @»ˆ5kJÙŽKxDû´JÎdû\¨.Å´Mã% §Ç:'"ßNHDÞÔµD§¬Y×9b8R؇v¤U|L·úiÞ³»Z×kr- Ò_‡©ð÷³àP#‹T0¶Iþ'E’uÞ“6Zç¥(Ê9Ú¯WKÝ4ÁX5Àl\ÕXÑ`÷« i5@p¶œë1þ4åƒeÔßtëi fˆ8]£˜Ußdʯk~ˆ° N À©‘@šê¥2@í/ ½šY™à%D{™h$ r©PµÃú+åBIC1çŠßñ ãª7Üd]æEÏN‡‹†[‚ÙNR¹N&Eïù:Ç>R^C: «önÓõ%G\Œåv'Ìóæ}Œ‘ôcb.qèXå±4~%‡Ö%'£çR;´_ªnh”Ò-çGæœàd‹Ò7á ­\†Iæ.$ëkë…°ùÀé»õŒ6–ô)vÝ wˆò-cÏ:é„Qˆy5ÀÕ½âTÒÃÉ8UÃXpMÉôÝ©Ur÷‹bÚš8‚åà~·|²ØçQâ}òúÈèä#àõRQ¹¡ šRO±!  8áV°=y?ïÆï¶f&ëߢ±hòã ™ICÕþC\ýïÓŸêvA.ú©§°V\ &rû£Ð©£±Ÿ*sTm¾ ‡ð#3aNŠÐäsÈ.ËŸ·Ì–eåI’=þ¤}'[£[6 âµR÷›môΚµ²Ï`W·y9PCyëà f¿ë²oí…?tô_å:0 €„¤–m¨fW™kƒì¸Iî‘=u›DJþè]Ö #“®G›Ç…ÄÂJá-ÍjT×`Å+ /c³s+0™BæÞ¼êÊÑÍd>*PúÔ*×MVr°Å.ÇL;vWÎéž…S9Ûìô~It“—9Üë-PAÚ:pÍa2S…O ýJø€©ÕN&­S ¸×Ðnˆ5ZôKú2s#å©à~ä‹&žÇUÏìLéGaÑÿôl‘v, 6ºÊ žs¿!H¡Z±sž‡ ÔUqb±X‘KÛ™bJ[DÎ .^¾ßúy´LA ¿˜¿ø·¡~zùÃÞàß!#ȵƒè]û¨5X–Kãf¤Ö-.‚0aLFí<j¿^²×ÂñÎE¾qÁ¼7* ²&9ñ ] PçŒé®œÊB–rµ’ý ’ãW—NÍRžÈ}M9v=$ý£N€9xi×%­©–tÿzn§'wÍpLÜ­%‘@‘¶¡wHKm\l>~vˆ~Û£k¯v®+ß#j”EuÆÑx)&¶ÄÒñ‰¯~¯R˜8<ê,7¥·0a¹Ù"Þ³òÖBCŒ—˜êã]ôS*.ô-eù–·ç²¡d,nƃ;¡*·ද«ÔF99ÆÛ.&‘ý26›ôœY´9ÄÀ'xMºÈT5”øöºÌ9&b&ZH…¸å@[-ëÉ ¹éA{ä(—üN™>73€Ñ]—DÖ–_uÆ|„GoffÂV7Rÿ±;rEn¸!/sîoµÊ”~ «»fXÒÊÞ»keä§R8´í)­m ¨:ß=¼MÙ6‹ž^¯¯’ð# 0Îþ>ÉP$7¼;7¡&ñ‚)7ÙoÒÊö‹â¹F¿c£r⟥Œ2‹C–Ó¦EæªꜦå–Ò¦~`ñ’»£mR¾éCNëu‘x QÅðÛvȰô·¹8£Vk/$(tjË“Ö)|_OcçÍ×]2v›Ž]y;IžçÅNØò‚»ª›ÞþKˆêG=G ×& ÿ‚§!®ðtîŽZ­÷d™[V­½;Ÿ¿ø) ïc×Ђå ý7=n}6ßêŒïòqŸ ¤À AœeCÚûß`ó\3ÕSXíôa7m€tÞÅó?­É‹è8 ŒÝ:LQq½÷…©¶ mQ{£ÿŠS¼#ÚÖÄiƒ›¿mÞÆ)[_ Bÿw¿ovÄ¥(¦Fø„ï¸Ò ãFmQAp›m4¤^ìÄúíÖ•ªCì±–0_ŠÙ+ŸÐ è9û-i‹»lßG³~šJByC"¸³™{S$ÔwÞvKýïÕOlƒÓmyÜTïU¤áË©±Ïn^ã |¢×1 ãŒëKá/ùú ˆ„S/ï’=étÿ¸r½Àì".­Q$”Ê8Í“îº~%øËt0›¸y>é-ÏÑ‘ÝÉé¨3„ƒb‚5°0ªÞDt¨ì<¥‚î¬T„ÝqДfÔ‹&Ýšˆ>ÆÇGæCVæ®v¼;ð¶djï”1·3b8mNƒ=1F"<Úä"íj$¾ùTþöñS ñw•–Ìx3ˆJø>Æœyô«[Ê¡¶*çÃ0¯¸€Ôˆ¬®WpÀ’y§_ÎõÊJÚz¿uLE³ÿF»‰ˆ¿8¨×hGŸÁ¸Ôþä:Û$+“<Ã…Cío²ÍZŒK„h1¶Pàè½*¶ê ‡úˆBð!²Ýz–è¶Î^­$â?[Y¤"ã·£ûöŽn þ¯J¼÷àøS%57ÂKá‘ÛÑN6¬g#ö‹Ãâ¹´i§Ùþ2b“¾h> ¸à-Òk—›qù­¿ÌYšƒÌðçª}…Û³OdÅa°BúšÕ§xú…Vðæ0æ=&ø÷¡h‹ÍN>ã]©¢<¡ìh…ˆ‚âx ¾ý'F÷IŠ.š!$æ½Ož¦!²bô¶ùäk8­¿ò òšþ yAiÁ DéÉ> 9¸¸”òÄŽW¸…ÆgSËÇkuÈšzºŽZüTø8³/Ä2V’¾BÖ£ï†M9*ªGæD©ø“Ðu« Ñ(õ'~Bh€Ä“rGUù÷‘F‰RjDô[—³®~^°Ó|»/a[NÁ'Õ .{q8ïÖ'ü]\[ܯ÷väËýèjX÷Pä’¾õÑŠŒ¬üâþf-ÆNÃO>H3ÌoÊÖ¥5í!¿®é…u7¸ÂÉçžQáF=¤ƒ Ó(ÈÙšXnÂá$ò1/#Ði!‰1”ð2£ùYxÿnM[# ™e6§]È€b›Óq~@ ÄN´¹LÑýécd*hyZY˜¬ÛÓ&‹ÈQT¶™hµÈÃt¸¸@I /_¸í¬…Sw¥ S½–t8Šýö.S,8]TQr¼œ:|–®7 L—âyæés:aÒÒâ˜u·» ÏÐO&]Ê:tÛ;¹°Ž…1îÿO"p³Í:+›fâf—Ã’á=o1-쓞õ4e×>¾>B¯Ç è›B:íË, àÊRX·»ï °»@çq‰ÂÒ–€±º º‰‰PՋŧ ‘HâD«ÐW¯¥…ܳQ´üI)s!I™°ijÌR'Q|¯}ƒ^Dü'–í—¨ 1¬ÜÕÂÚàdéuãÄÝü”¬/ìIP„å²o¹Všuñ‹}¡X•ÖÅ281íîõ5ŠWî`‰U?Æ«Aíi–ó^h k´ì£õ£Ä”­oQ;ÛÊ…´o}•ÄVÇg5åŽ$rSrů?Q r³H¿ŠdFýtèF¤ŸÿW!¡à-¶U[ÕVýÃî>ÅÆ {á]ÛV>ÛZÒŽáhu~Ù˜T ͺ%†l#=ÈÒZE¸µŽöàTÄ2pCfÐÝrﺬ²ûÍá¸M[úŸï®n\ËœV˜šÏ:~f à $¹=Z0¨Ô”:ÕÂ+¸“µ¶!óK“Õt¤{¯Äª8IéZWlúhZ œ€¢‘Ÿ»8êwOÎ2ì"Ca‚R!>«C49LÚâ½ :Íö£ŠN%Î3›f~Ó$ù¦ ëÉÖ‘^Ãt‚`öË|ÅžSOO.Of¼µãÅ4cqþš*Îó¥9T8¡îúá;ØYF¼¡× 7ná¬QY«VTÝx¤Ü~kUÄ+Ci2_5KábAxsAÈíöÑ*oêu'# CÑú’G׈ø€³»ÉG¸*2¾N Š:ïq®åWV`—muoï1ÀvŸfq•Üœly¶öËá:ød7æïÖ‚æŒZxwÑ•ae'—×jŒ÷NiØLÃr3~ &VªêE`çd‡ïÚÄXö£K\xY¤vín{Òû\nÃpKÒèŠ|B–~×sq®ÇLTÇ?H¨/¹±`¦ã›E!{Þ°H…œ2O…±ÎðÊJoiùI(”¼ÉY§æÐw–Äi‡Ï-„d4Å@u;%œE}%bËS?—vâ@.¤,W•Æ4¢Ò"f¡T­oã$Ùº¶Œ{ qСDÖ‡K ­¸ÏÌâê˘# ¬‚<ú0¼õ¹@’AZºáë+pÛUúy?O¶âMÄŒ ?À©ßø´Ÿ6ᇩ/át%iÖÉíw^:åN!lJbÑ!pgy@—Þó *!hÁ°ï²ÿü7¬1`~³·¶}ïaz[)ÖrR½µÒ 2°«¾š\S²ûóïZEoæ`ÌšZ/'ˆZzZ·¤$ÎÆ5ÒÝ»ñfOmÈÌS°/ä;žû)-†‰v¥8‚!è.æ7«ú.i)3ù>ÀJ/zØw wRg±vùa~„W'f£ úVð´ôª FxÍÑé½DbŒN´3ö¬Ï§U{J<Õ™ª]Ìõ«"TÿKdâ^;/×Igý=E„ð¢k´Ë‡.YÂkö½a”×ËÇÏ2Àõµq)ŸérJ¦ÏÎèdenÒQ<ÂXGÙþ§a¢©uš¦^wjÍY·G±TåêN·ÅÙ-ÔxŒÚ‡•íMr¥1ð^!â@øý?^$q¦zqÙÂúY³‹Kwàd>Z鯮ÁZ#$à1ûìLW–ËzÔδbìt¾Éêïx+!¼Ð&€ôu€¡íÓ†_ÓSSqK$Þä+NÙýÎwžüº¾,hà´òyâì"™þ!7Ö‘¨°?„ç”ÏâxéRx„)§ÓÉsWP`§•³k6¾eLô¾z]ØálÞ•0‰™ÄÜ%¡zf­%[vB$¸­.wPæ6€¥©d¶ì ÷FcX„QðÑIÀ–¹3´”²xf<$êaàùpl?cbã™E’öÏé+®©Ë§Y šãÖMæäþ­„ToŒÜ9©ÌcV¤ør(Íôë›A(u«}ï^ Šú¢½_¼Üѯö(À†SŽ=‚[*¤ú¾%s“!ÛO»D %‡ÑŒJR‹IÆïŸ?VÖìjd ‹Ý=2@Éñf%©P¥]DŸ”¿¥§~T-Ñê} úߎJ‚¹ÕtK*]MP*€£=˜_üº?-¤ˆ‰ó9£Až9{ToJ^C¢lPU¸ròeAÚÿ.Ñ{ÝŠÃdcßÌ«PC¤·è[Ì\3鹌Z¾¤ûîÁ•ߟ<ü€þÞ§`ïÿ øÊÔúÂ×ÿÁÿ·àì¿ÕØÌ|ª¨.[¸…¾ï¼”ðžÏ¸·ú1²4¾}C'ÓfÓF*¡1¥}™5,!˜¯'õߣóïǿڛ[û„†–4[¬ÉúœèW«KÛÅ—®M¶Ûÿò^²ÍÇ#Úß‚2XÞ{ôºÐôk¢%ÌŠ¯¦« -[J›¡«è´ÃtNhZ†‚,ÔšF})=­3qÆÛâP¡BïÍÄÉfÀþ½p¾u‰s !rÿ×€ùÝA~ Ýž»v"9ÉUîBR$Å“§Xß znmo´•Åjð-Š%&d1‹5ˆéòõp;l}uÕxCÝÎÃ;‡r_M/„qiq#þvÁS0ÐÛœŸ>¢B¸Ma&âZ²)Î)ÇI½ÌšÈäÙÇšDùž^¡˜Ìß·k¤ œñ‘.÷2 …"ÄÖÒN±‡’‡y Õ$ãÇ]ñtx ˆ*¯—LŒ^œð¾wùxƒ“ ªC<˜gxf©*@+•ªž's==¿ÿ«$GF«HKãŽJÅMÕ‚‹3–M¤{ze&Ñ`õ@ µÖ6 o=r¾¦!i\ÐWØcíbà6ŠÈQb§Ñ¬é]ŒÆ|sd¾öCÁ„/RlíÂ[Pe# ª?jº§2äuÛun—SûH*–|ŒØÚÁçsNVé<†/ ÂÑ!ô‘Šá±)š]ÿ»´XÒ7‚šV8ñý›µØU$"CÅÇ4¢’Ü*îö¦OÅ íe¥:ªÝ@„³tmÿ,v¾ÓÇ×ëâ\«´NX}Üý+8¿ùqõ°Å#Šéºi¥,Œü+Œ›*»Ôƒr¿61¬8¡¼QœèŽüù‰ô]ÓL›˜dÏœŽ4ÿ-ÇúÇ6ÊÌ­ððÇe—Â{ÃAŸ½þ ùìaµ­4ýåÕ©ê-¥6.‚­ ®a-î¦Íè.‘)·BÌ6€acô¶<Ûð KŠ+Mj\ÏÚýÞhÖópN"Vô„FŽ“àãY=îÌdk«)‡Zeϱœ:NÕFaAÅýÅ_‰;+®`‚ы smhé hŠÜDð²0 D…‡pd€v]¬ãyÄwó éEk‰ÁÍsÁýõD­žƒhVMÚ¼PÃPØ7Ù»‚cÈ„Ó8WäÀ§J ûè;ÚGÀ•6Á ˜Óà§Bl⾓BøœžšG°¿‚É"ÿœì_>-ý.OyOt®z¹™êÔ]šwP¾ÃðÂË«Œ †øã‡Šªlp¨–@æÎÍÃ8OaÜ6æ‹{ö¥B|¹CšY_öÀâ¨î`˜^9!«Ñ`†ÛéßãfõƈixFƒË±ôièPC~y¡óRâ‹hal*=ö8i¦nÊ«ÿkœo­½¶é0vî*6ÍEN‘;\¢veZkà”bŽÿÀÁ¾ù+(D»£áy=1,r k€D“on¶DͶ>ðî1¾è§å’’úÛâbCÈÿ¢~· ŽÃviÖë²`É´ Ô+h7åM¨»#Ý(%ßéjò'ncDU. /'ã +hÊԕƺ¾²Ð7~s·cÕ—)¸óVß´mûÜœJì•ÓVí}Ü6æ`2‹Ä><-š„»‚íüÚ8š“’N×ÁÅ>ê}åéá!çynòjÅ«QÙ«5£É³•Y¦qd½ß€ ŸEKØ€M¢ñ8ñ½·5ƒðb š>ÃŽ›{Ä “"8±2†¿…%#TÃG÷”‰UPíƒ$jwAFÅp"Ä7õ^l±â$³tt®tx#ç°n¼7Ðv@†³£RE(´x¼Nu»†ÏeEÍ7c"[z]æÚÀý¤Þ· W:½’+wHÝÏÒCE"¬FkÈ=ÚI#³b¡ÒYŒÆ3‰ôŸT¢{aÓQòB‘LÈ3aĞþvý"Gvê!Ç.JXÐõ§yO-5åÎˡΠý¨tLZ¢­e^þ#¡ÉþîýE¤Œ¼-ylø¾KÔÀj†]ìi˜‚Ý›[­3N†Ü’¶L²éŽzmËì0Ÿ"ï&Å;’¸¬ "zš¤Íì®°é†*¡Œ ÿIy{ Þ÷ëÊ&ð¶qÂêjòâé²OFq#^$Ù2}I,ù %1СЫò,”á•Oß©ÖFE'€àž¥MŽÌ™{–b#¥îαA•€-âÎø8³›û“à³Õ|o‹Š\½\ÀG7‹¨0S ÷?ÌV² gù¡? Bà²8<'û…ì×¹‘œ“ ¯ègÝ0,žg†é¬S”6’†…#‰¦Xqi]@µÉ…12ç,åþsê̥‘¬áËp@‚íC TšO\n7#}‘ÿ=¶\X‹uä\ë e©Ojµb•cŠ­p„É®Ã,=á]²¿\'ô/’ùüϺ‰?$‰@›É‘8w£N®¨„,®ò̉ÇÜS²ý_ÐH@\þ¥ƒl>ÈÜ‹~oÇ?âGÜÙt;'ã„‹Äæ=вòJ?É1nÝÿ' ’w ÈZ¢oâë EGÅ#³N¬'¬ —«%PØJTÞ;ºÐYhÏî,²<¿X7À(XE½F@’Ç{ך+>R-Z.o uÄÉÐ +¢_>ì»Û dé57™1˜8é~l#é/x‚Z4³9Óˆs~Kå;…õ\Êæ´^5$4Ãü-ëù;¹:)ùG6ó¡«² Š¼N}—jmpÑ5ùÆÛ:t64qŽŸ'Î&ŽƒnÎÿv¬ÑC›‘$¡AŠnÇ­ãÅ&hÜš%FÃFÒÒW¦g»ZŸ”; ðž˜nÄ®@Åäz‘È`U 4 °0§¨¼H!™ÇeÝ·Y@ùÄMJ_œ{6Ì”!Ç–@gÒ§†u_àíJ?žùÑ#]Ýâß"·9¼eÚœÝ;ˆMÇ\Fè•Lˆ»zß8DC“Rˆ†p^ÏÙ°Ä»²8º> VLH`ïyÿ5Ç¿gxZï$sm—Möåúö¸þ ‚'C´¢#¤}t·+œÔ…ëk¬€PÏ~äVb6sÏsy‡£/z;‘€y$‡€k yɳ&-™BÃŒ»èßÑÓ}å<¾.¦Ð¢ÿq¿GE9µ•Ȧ1H~4œOg6»+oÈãM³"™:ÇùbsEcÀcs•È „Á†ß¢@ÅRé;„ìrd‰¯êŸëÉЇåöTÊéÄ ™µ&Ð6NÌCûÂgˆš‘3l“YŇ‚Kh·´©CkБºã‰˜ á*D ¸oáˆnó †y_‡ùô¶yZ”“Pëœñf|y¨¿(¶íÌFDdAé»î/¹)bµAÉòl¹-íH¯ƒ>ò/ÇÅhÑoOr¸[­"` ߊÃï]ÿìJGwUÍ"ŠÜíaáîy ºÕ n/»iäÍdÑÄ N¬)‘%ÉÔê½iM†x„œÄŸöq¡øòG‹“ "{w|#`+JŒ½–9¨ˆ2¦7Ycl÷3÷Ò.µx¨`‡s6¥žz%â[Zl¡]ü NÏú¿'!$êÌÈPlF#Rëz›»{SJj€îù¯üq=]^='I®ö¸Ž}F¥B¼ .»£ËºkCD‚3ž%À¬Þ©óA´^•Dz’šÿ4²êJ‚ƒG=C¡ÒH¹ UqÝ» v8½[-[ƒp+ÆôþI•|»M]Lô§kZ=UˆwmËù£ü`9`ܺN|›çŠ«mûê¢\Uùá¨Ñ¸›Tц‹t\ŸERSf]Z΂r͵¹”d‘Èë"Ý02{—š mx¥¡ 1hŸ(¸E\³%ñY¾?Blo‰E›ôƒŽ‚ñPAÜÄšÅc6=»º ]ÃÇ44•f Q듲èlF€¡?l²¾LÔBžŒ»¡¸ìÐNä£Ý>Gg ²jÀo—8˜€ÇìJ›ZH–’¢\˜f‘(å·ÜËÌE;É["Ü:I6h±‘‡·÷«K{¨MòiæøžÂ%ÀÌyØÇ= R^Tßž<(ò¥Ðr(&|qaûýU@]ä|ù¹´Í­~^_¶Ç$ßqÒ—köûZÚz,•OdHj ô”UظL&7ìç×ù%(¤ÚñÉò=#Ÿ_Ù.'É»þFUݵړËSìYÛ›eYSBÓ€ÝÞò¡ _ ùl¶b­È$öŽðjÜkǦ&yJÝñ“'Ý;ŠÂÈ%a^ma4:È7óÃýÍîCùVãøC-|-ðrŒ(þØ4 Ü‚1qy©-ãÊùOOÿ!í"Hó^=Ã8‘º™Dè-0àùLm¼T¸Š}›`o4_cßš‰BF®™ªÎá ™—HR-°žC_«4…øÐ=Ÿ±mÖXx _%‡yz>Mk”ü‘ê¡#høÁß&v)`Âóýn¹6ÖLlÿ¤£%S]ín?ö“¯”BJb—¸ëií—ìÒº'ÖàžJ’ìÀa!™«ñ&[ìܺ¾=±ðk¹ÕY3ý"¼T3Ì¡•SÛi°…ÎÒÍzÈ£Çr¥àlKl$›Óe¨5æê»m­„¾­ipŽÿg$«Î Ô'ä:àb§hÉ£)ó°è&¸S×þɬŽÓÓ­Dÿ0yq¡3ùð`d=¥¦mv¤ÉûWõGY„zNÇZ3‹x@”âΪËSÃoëš©û,l®  ûÜ—"B{uÀxØŽ9ùóƒ;Ô ¯}œyMi`Z–zcF,s?A'IÕQ‘D~%ón<3æ×≈2QG‘TÞ¥ ¾,9ÉâdȽ½-ö 6åâd1É3ýS:ó*L y«Í4 o]ÅÇÕNÐõû™z€Ö%º§„o0ûÁ~ˆ¶kã’Gò†Tò½l¡ºL®MQyEèÅ€@Znê!qÅ G ’IVÝØþõ ¦YxÚêȪ֩áÀÏĺp&|L‘Q©:‚yrµ8å±>w³:s»IÇiØ4A8ÜÔžÖKuÙÛôʬ½ÕÆ@ξ•Šû¸¶wCÃ@Vñ4£ƒ„ßgmÕÙžd˜ZëMï4wä°Gù½=¦Œb4œ,å‹|üˆ-Žà'e&G?ýÏtõN¥²:í½IÄ/`yâ³ âÊZ#cѪ¬H×lßÐgzÊöÉ(›‘‹†‘ª*þvi¦ÛÌf7BÓúÝûa£“/ƒ“/ÖÜÖãmL̺2a"ÇCöJMq¨=¿Ð/ü]€½ƒæ°gÝðæ-W•Q‚wšp8Óæ^õ§ôŽ=¯¡Atëþå3„CÍÚôA©ÇþŠ·®£I0€:‚Æ¢ÁìÏBi8Ì UÝiþd¡– ¯{ ·ígaRöh×Ä4žD|•8{<&ÇoËÀ&E×’¨ †ûøL `d¿‚zÎ$ó ê% œFR$ŸÆ11øý±Ò§TO©_uÎá‘ìwáËB²Â1_ŸÞ{ Îâ,ؤ÷Y‰Ï= PÔØe  ñð•_mL¿p©n[î §Á—a_,’nåº –×4 µ êb5lŸ*Äüèc€‰%4UKC*çjÃ…Û\l¤2ÁhÕqóN»…¥ÀëÒ>ÅZD0"U}¯ë+ØmÏ.VØÓ8`ºœSÜ~vGྭyø9L{=§Œ›üª‚]œ>ž™5°Áók„ÉTÝøÙ‚ró‰ ¢0ê•]¾C°4m'”“Dé>èøÕ}F$À>À`wƒ&Ù Š©üç™7¤5 Í  o˜ó K,ïô‰}Œ<‹-…‹ºn5Z䦅‘æýöð÷Ûëߟûs¸ç嬑£ ?ð&£gòz3?âÙ«uðïûÅX×W v±ÃAÙ_TÕFë>{ 7(üS Öô@9ëöuƒÙ ŽÔ£1¾ÑýÓ Òùê7“µM8Ãç×ñòó<Mè};=j“WÜ3\¸µÛì„õâÁẛ 6Å:kœv^«¦Ì'íÑõµ4DŒ¶H¿8­­î'ìôq1oåZ¾É–Ñ8·úÆE”M$èæ xLÓnHÄܼšÞ³BÔÛv®us™åß3.­y;“RÓ{x£Ïö´d™›-'`6r&å½Ë` ³e=ɶws¸$“ì¤Ù&jò–fª£J8È”ƒLÌ[ ´œHæRð\ç[îlT8¦ÂKv‰Òq]Õ}‰ÝÀùJëÔéŠåïø)˽ ±x ÑÁÃ2FÓÕßP‹±@QHݶëq4Ÿ2ü™×þ›BšÀÒ(؈Ïe/Ÿ˜Ï…;µ¨`†£àQyvŽûK1*û Ú:K• Zˆ‰h6 üRÿØáÖ¼¯Äö›AE?*ÅDs´[èfXá…_SpäŠ[°p,Ÿ³Ð{ãX†ÓîŽ8¿}ëU“~=#Å«ü@Å9A0Äi:BɱÅVo´Ñ&R‡Ÿ)@Î1/żØÔ²L8mÌÊ¡ÅÙÕò: !æ\Þ Èye9fifr¶ÓêÎUÉåQÔ>ì—uV($õ{•šßÀøqè;A(êû·T=,ªOtÆü™‰—›535󻟋b(…Ʉ硵…µü}¬ÞrŒÞäçòŸÊãËßWi¯–Ü:?Ö§;żTär(sY˳\ê¿ôrˆû#錌£»Oò¨/Øuv‰¦@Y6JSÆXhi¡¤7$sOÑy4z4J~ÿôÎ=û@î°¾Þa`oß媛¯*ã)•/u-„ÂÛ@¯†mŸmw €9fŠWO°›Qÿi 0ÐñtÊB»²teð苨y™è±×W)Ðù[ç/“¢šÏ£Ø«æ5¯ rõ2‘‚ –®oÁÎQ*íá6u¤+òaKnì3éÕOhá;j£î® …ž·jxÀ#ÊAÅQ ä`¤9ðý?+ˆdÒøûTu ˆ0D]ôOQv‹v/G#p2æ dn6²æšlÁÿ!0‡pš%óGP¦Ÿñol)‘eÆŽ)äÞÖl<¦# ò,Õ›]|â®û…“2WÁ­­Ðƒã‹.QùŽ\Ä‘I øah; #!xðŽ{gV×¼Èîð9ÒÜ2&n½©‚¹jâ\þ˜qñýÖ×\ÙOìË7‰g ˜×|(º(¡Ok8 “+´•´g¸þzË øÝËøtöÞòI*cO%SÊn,)tóÂÝö"½ã.Ög f`2ë·ÇJLi£3¹ÎiÉì°ÕXÝSZKºQñÕ¢CŠó/”4­q·ìRË2ñ 5? ÁصµË²­tØbI‡-gD(»85ü°þ\Žê– 'î/þw‡ôÞ*/™¼9 'gå&6sïG¯[™Ä×JÍÜŽs–cÄDNÏsG¥DÑéQ2³!0üÍ‹¯;‰'!¸3£2£_£“Þtø{ž˜à2èôK8+ÒÜ6|É{©Ú¡ž£óE·:^)ø7 Þx 'ÍŒRö ÃkÐ/ÿF³r÷=6š.ôcoÙ÷ ¥;7ï³yða¾¼æHþI`*¸m2ã×éOOh¾SM„æ[už¶ªÈô>w&îOa·×äÃÆàUÞ¬y \«j?¨/[ª£·yG ¢ä+›“ðü{•Í UáÕY¨§r٧Ѻ3vÒ)Œ è†޹•wà»ãSMÀ•èËNG7æÔ>šø;OR÷ÅÔ—*h•¿Gþ/6€¹H.¦”-Û§û 0’æ0 rØ1ƒ¢„0H•°ëwq èÛ‚ß9;\øÆ5,©hV®%d“€èÞfµ7^­}bÁŸÚǹÞ®|ßÛJÝ©K3¢]Jý%Ð8kò(.~Š­ƒ÷'™„®|Üæª #õw¶©+j!' •ÜS¦\ÓlݹÒj…_9ÌÈøØO^õð=¯ Ç€ÄÀ=]bvÁ‰–iÊÕGè›Ç;”ïV^à2Õkk½G)*8þÙd<ÃÔëuq8·&f¦2ºß3 ì}ó´yNõýfá %3xè§'û ¥ä÷¶7ˆldÁI²vP££‡¯ ëbƒÑÚx{»x{\­¶ô´{xÓòKRe^Ì ¶PÖ‚ *q*ÈR ǯ›û¸Ï@¡áwªÙþÁB,»ùý$R£¸kA¿•Ož Öcd¾hÇW˜)A[nŸîûbš"ë4>ñ(HQúÄR‹8Þ1AénðÇ¡iÒ&¡T§ùü:º{±cÇŽ5]ý§1çù2ã:OñþÏÿ*NÄmù$ºéÇyÎD#·r;rÇ †¿¾.Ô£¥˜­Z<ÀÒ†ðŠkL¶ujý¥Ÿ¢ÿà™[¸3»^u޼¹* oJ†À;fòþlŠÇÒîïg·êUÜûõê=]òfï¼yZÈ/Øú7X“WE»X¿…}+Dô•T%Ðr”ÜW‚pa½l$*±ˆ¢µö&` 0•ÜõN•3œ ‘„û2PO‰»äB£k.—mi˜O4Ê…Þ6O´³$„¶Úp®„Ìx9Žk7©V–òGŸ¾î±Gä0ðp˜Ì]&È,Mz6b%ÿ²ô—Y5?C¸<¬º²+7Ié.ü®tÁ…{C–;¼‡‘üˆiúŠØ©DgaÜ_*ïút5'?9ÕÌÝ24› Z¦¾UG ¯Ó=Ø>±Ê«ped«gN-&Üà·ëjæ*£"Mäùwƒ”éÓ?ô¢Ä^=Ww=‘àìàóbjºrØÿ1„Úd±éRL4Èlé2×R”yÈ6Mè$rpåbí5“œÎ숋-Êt`ÜKßLcPB76æ/™±í™yÿæ a’`o¤Ý9ˆÒ/õ{EaÔËô ;rTž_u„®ªG; ÷0[Ó¸K‡À˜´À(õíXQtü’ êÀ…/[Ý–¤yÇÁ ·vɰ…]u²pB(<'š…¶gøtTFÌÂ6÷G%Þÿ+A®;·R~ñ#»jñü ËÀƒ#G?¡ ºØQ•Üm<(М ÛñBÊ—â›>0 ‹YZks/data/grevillea.RData0000644000176200001440000000441313216675765014513 0ustar liggesusers‹eW XGŽ(ƒ€²à($(¢Ã%‡¢®v?ÑCtŠåÖ€*èŒ °d?1j¼åüÄk1)¼Åq‚£ˆ1(«ˆ«De·fºº;fûûàuU½÷ªÞÿþ÷ª'Ä{ž›á¯2º?½ÐÊo>Ïèã©•ÿÍ"~ã§.},®Ëé>|àûqyiÎÂ6@+oV;t<û}Ú‘!š{¥dQ‚ áÛÄô:÷£€¦–•U1/~´_)O/_××*úãã<ÁåYÈ¿ÎïïtµT){ExO÷©¯j“H<1ÜÞ[4»³{mzžämÁÓŸ};¡¥ ³¶¶¶f#0S[[[¹ü(½Qª¦ûJ n‚d.¸E¯±ú3£2›g/Ä3Þ¹ëö`ÆcYÅÓ2k -o"ÝZûÿvS´<#öŸÁ„ËÂ>ì.M»²Áìï-Ì—˜Ý31ýxj— ™×ïS>®.Àì£ìWOóDyðÙê;I¿—•Õ¼?!èÃÐ>Ç%µ˜}›–¶ñrUx\¯ÍŠ99ãûÑSpW7f9~`VÝfl•œŠÁ°´hPÛCq\ü© UŸJ–Ã_´£þÒÅȇ¡È¤!w¿ ³;¹uöõ3f窂ʷÈâ†ËoÎæËÇcFmþøtÓ 2î<ŒjîØáõgÈúæŸìCBÅõ†›"Æ$Ó¨ýÎÀìáÔPŸ±EIñ™¶N0×Þ#âüåÜÜüjc /~‰üefMÍ h)†´vq†KÚþ…¯=Í6C–dbvÿü‘®Nf]õýù&†­xõË 1óDíB:7fpye@ÛwÅ1ʼnåú#Ã.Ù„'6­…[£ ï|–yÊ0 š6kÑḐ×=\—‰™Çoo´ Ë!úƒ‚L[¯b&“úåýû¼¿Ée:†.o¥ýÞÀl ëØgénrñãwóøFï­0fhïM‚'ÅI9N“r^µú¹s'†VÎ?Û¸.@Zt3<¿5ÛíH2À í{ÂyXL%Í»·Á¦HÿÞWdœŸüEyf¦jêwž¬›»çâWV ¸|2]ÚïgL;W· ÓÐËPäI#—7¼>åŸ?ÆÔÔÔf!ñ¿º1òÉõb_S:>òÌXÌt½ã6f;ëìJ¼ u9½377Ãs½X—Ö´^ù:tîînÛÙÆËP¾@!Å‘—侺{"ÂÚÚ³OfˆcÓÎõÖ}ƒþë~ Üw †™»oÌÒ~Á¼\Ÿà?13¶ÃBÃf4Šx îKñóÁŒ›ï/ m²´½„ŸJcÇÉñ„O¯)ïꪜ„ž"ãsÑf0ƒ¢ö»Õã1°!çÔN<#øäšq‰ ÌÔh¿;¿p‘Ö­7ض͉dà í=†™Úçh¾KZßÜýY©tÅ)å:Ì´ŽÚe%%çê´ò‹œ Ö!å+åòÄÐO¼Ã×ÕvŠŸÎñNs»˜yGã¸}¡ÏÚd3©Ÿ‚½eS2Ä>˶þ“}g&Üìa® }âíÊaõ 1âúí½'öUX˜Ÿ~Ñûcö!dÙž`WKÕ %Úï~ͽr±#ÂÜÚï~Ìœ÷ö)"øÕÇYåX’>rÔÉ$\CÂË[kjÅ~Äß#|øõó¾+³îÝ'xßìiâ!ö}E1éŽù¨·86žcíoD‰ä'?Q1 * Ññ1+É»Õë'C´ïœŽ~\bÂRÞe\ô*ªÒÿ?ÏÁ%R‚ks/data/plate.RData0000644000176200001440000016013513756343312013637 0ustar liggesusers‹Œ}wx“e÷?{ï½WK ´”î•´9Ín›Õ=’´MdA%¾ Š€‚Š  ŠŠ( ¢€¬ÞlA2DÊ"Êþ%÷çP.àú^×Ï?Þ¼OIžç¹×™Ÿó9]~lƒüÕªU«Q­VµšÕjÔ üßZ5ÿS=p]?ðY{ø`ï(_µj5[.‚ÿÚ¤Zµv“+©óêÚu;›W~pòFöÄÖ?©~)iW¿¡z¢ê¯ã£û+{OPÝî|µmÔ•Tãä¶žvQ½ø¦þm/Sƒ¿çüøúÄAÔô£e¯}Un¢æK_^z(sµÌzúçý–S«œu_¶k½ˆZßšUó¯÷P» !kÞêBžïlúg#¨Óµ;a¹Ÿž ®mN]ì¤í;éJàªÕ}ÿÁ5u{s±õOõqêþö¿-:|µ†z›¨•>‡Bî<]œö׳Գõ×#/¬êO=ÿnwÇflKáW=µèf}ê5¯þ of¬õ¿aÕPô¹tMá_åóÀ嚭»)ö•7ž¾üî—}I•5úÛ¯RòIU£Åá‘íîPúWže[þ£ôâ·Ýun̦ôê¯$Yv,¥´Õ¡ÇûuXDiy?ù÷ª¯‘ò¿š¨ åüÛÏͺ?—”OµÍ?ÕÞGÊì?sÚû));ÞÎ÷)Ɖ[ }›¥û§þ0\A Ç3=õ >¢ÔÕrPê‹™ÛtɦÔÜÙú‘ÿ;B©QÙ£ÝGVPʃ£”_&ýË¥|¸efq¯”fÁÿ9>šœÎnDËO;ÿýá(Ñö.kí§ÛìÚ5å ÑÝJã¤ïÞ uîÓê×8Aêqï4Ùt’Ô46íè¤rRvú¬^L8©‡uï;&f©ÿ'ŒÔsƒ?K%õA¼úÎËw¾ ù€4aÛ‡ßZNšÜ¯6 _OšiW?¹ýÇÒlnzóiCi¼uþ·Ï§U½'_“¶Ýw½Kv$mƯ}%øzÃHû~pZÒ®û柂U?’Ž65¾ÔëUÒíôÝöó:Ò¼Eï¤û#¸|M(Cq;°3/R†)¾Öó ÊÈU|Øq”QÞbèÜWm”1l}¿?s‘¾z`83·“¾ÞOŸ÷­¥dOé‹> LÔ^Ò´é×%ªßI¿,¸¡~àmèÈÐ*°MûÿB†®¿áuû Ÿ¼ÙˆŒ5î~;ðë@FÛÍàA ãä§’;O¬ |®Ìn<9“Œÿñßo‡[ŸL¼gþõÝ |b^M;k®\s¾™Cû½8ÞBæðÍ?lþ.™#´oý¯NO2 n¯P2¯_±x÷f2oP.ü³ò ²tü¤OàÉdéöqçßç$˸¿å£—#j42“åhàµÍ ˉ‘ÿ;|ì²Æá¼X_Þ¨í ÙÊø=hÌüA¶/Èf Œª0ðÞ€ž½2ð>݃“l *‚ŽlŸ´•»nà¾G¶?H¶³¼O®Å´•l7ñ{{c^÷à0ÛæsMPà“½7ÖÑn=8ˆ ÈØ-Q&²›ÝG3Löò/ŠòÒó=Z.ÙBmÂäÕd<€«È>! -ŒÏ“Ý w§“} æý±}$Ò¶7çä3ÿŠ´©DšµI@`‘HÓ§©H‹w¸<ËDZpùþ[,ÒÚ†Žr¾ö·Hk˜¶°óBy»dOùUµPþT m…ò¬Ô¿Bùs^àà† åÎ;Á)”›¤žÊ•ò½…r™ÜB¹ðÅàÎÊ×R¢_G¡/ωPŽ Š‡l¡ØMiõ„Ò§špéõþBY룅2ïÍãsoD¥YÊs¡L ˆ'Ó:¡ìòŶØÝùBÙ4 %«+„âºÔ»Bñ44„b­œO¡X!õP,òM(fò÷q„"Moøl@# EdP< Ew¼§¢ÍùgÞ½.u¥þŠêñúÎI‘zEî3‘ú»ÔÃ"õ`pº2Eê¹Dê‡rDê»Ò~©Sƒ_ï#RŸ›6âÄ€å"uàGրĩž€¼·]¤fJù(R#NžÌ Hm)÷³H jÅ~Š” AóÈ-R~¨™³»DÊúK·´÷JDÊ—»ƒ RÞoï‹8°U¤¼º}扔Ñëƒ[¤ é×á½McDJá÷):ÌSJûÀ6lýµH¾½#hȈ俤#’Ϥä7+ËÉGòƒ ]$ÿУ\"yGAPЊäõ·R"’—a=’Waý’7üTD"¹RÊM‘¼"!¸"ùý¡AA$’_=:)pÄDòó;Žxæ5‘<,`éG‰ä2iψäìà°¿SKý&’{ͤ–"¹¡Ô3"éõÃÇ®¿|W$ž‘zQ$¬ª// ÿÃßã÷NGûD|Z`ySßq+>M×ôu»ïÛ5xœrEÌ[R^‹èòœŠ~+¤\QÛ¥½#¢LÒî‘ûåy?ÔuÛ/zÿ å©èeľ Š¡PÑóȺ;ë-BI=+B®Kù'BÚöÜzâD÷׃b×!º5½Ó¼äFŽè´]ñšååM¢ÃÅ}¡æ²PÑNmXÑjÜo¢]¯·û¬è-ÚUK¿Ðmþ2ÑVQ<áÈÛ«Dë9í›ï.­L÷^zQé-ö¿ÖèÝwãE³¼WuF“»»¶‹Æ÷>õñˆ]¢AiP]<#êöœ|}Çï?‹ZoÙÍЯòþ‰Ö»/RVþ§™üë1öÊ?.-j3½X¨úßøsÊ­OóTgó->Ÿ2Tõ÷…œzÛ7PÝŸrlôªõ“©f}ù\j¸è¯W›ýÓ‘š÷ûpæˆîŸR«×Ÿ¾Üàl"µ=÷EÝQ¨ci?Wé ¾¦N¿ä¼¤s¨¨kÒˆüDÝ^ÞV¼~Ë7Ô]]#QÑs8u?U°ððŽ;Ôãå–©éz ñk:kÎ í»68±ZÙªEí 5ªîË×Ô3)°ìw¾¡žC?„)›,8 û检!Ù‹ÂíAñ“Há¯Eõê 9Üë8äsïìo,z7 –QÄ{A±3’"S»¨®›úÖ<=`AU=—¯)ŠåoÔ}¹_©ŸöW¿½wŽýèkŠ: ;3êó Y׉¢„ÜïU÷ákê÷‰<­ Þ¶EïÇ|Åôù#à h(fôLÌ;ð#b¶Ÿ*pŠ9'¶.ÆÛ:¨–Ÿ¢Ø¦·iu†bþ‘r§êy|M±%Г±gaĽµ ­î·Н&ÿ{ÂÞȬ°’Ç~LÖ³—zÍø£Yw­+MPM$ëò  _°' Ï¬Ï, ²Nd»ãE)ȺóoÝæ(ʬ»/³Fàí»N£Ìîø}f‹3¯$"e&5 ô”Ù­ïÉÀ QfVPmפÌgàoeJ³!ð÷¥> Ìw`f4H)«M@m-šEYµ¤¢¬œí‡’_* ¬7BRÄÅ¿(ë` AJY[EÐÑ¡¬ûØÙ©°×³Ÿ·eÃþÊ^І[)û y~)ûÛ‘W¡gmái?° Ù"`GÛ21/6·”Kd›ˆùµ½m NÙ¦ÌÓ”€}2)(°¦’í­€4¯ó/Ù>•ò=`¿L L`àsͲ  &Û²àvü~7ü+Û×ðŸmß—ˆl›á·Ú.î>á/²†o»pÿ–F¶kðíMáçØn´Bô¦€Ý{ÙÞþ·½Mæþ«ŸÜ!{4ÖÇÞè« دAwÖFö&µƒ(p?¼ÛUv…¡Lþ»08ä: ƒòÚ` ˜·o®UÒ_†x¹^ ‘ö’0Ĭs}Caèùièt [ CÛ€ùõù5ahýf¨&íG¡ÿO®ÐŸ–v½Ðï„üר Ï ýfi¯ ýÒ®ú÷¤&ô³!õS³U?pŸ¡Oõ£skµ{ó¡?øzÃß„A ýeh<þÃ…¡iü¡ÅÞ³B6;¨Ð„¾úCÿ!ô„~zÀ<œ2UèGòýúO:¶B?4h>ä }žœw¡'é7 }ŒŒs}ûûQ"ô5‚Ãv‹Œÿ¤üß¶¹RçÕoDÆÆ ™ÕMdLlñ§È˜˜•uDF‰&` ïÍ¥¿ tg^:B'ÍHŸÐÍ•þÐþP(ã…Ný¯k&ã)B{TúUBû]Ð|*ÚU´‘ŽZk`¹Þº 4¿´a…ÐÂï4¡™÷ñËBƒý)4ïÕnu¡™ð#Ó„f¨Œ3MÅàAšléw M?|ªƒÒ°aS¡vÃÕ…:5à6©_ê>Áã__¨[Hù"Ôu3’/Çt7hVt óB{°_hi·` KÐ;ÁiU z~Ѭ€„ôlÀ ²CP!ö9`ÇRì AQPhdÀÑB5ãP¹a÷©ZÈ8H_€yH }™ÎvWÚ9ìG¶—ŸØçú›Ð×úYR ýëÒOú7dÜAè_–rYè_ÂüèyýôÝeHèCä9WƒÌ‘ñwP-e‹Œ‚âÊ"2¾‘ç[dLÀ~Ì €¿í‘Q öªn'ìaÝô°¶™ºiïe]è†J¿&ð »XgÇyÐ9¥Üº¦X?] ŒC+ÍÄ¡ ;_[ 7’Ðv’òJh;ÀnÓl ëõÆ}5ƒo}vô\ºÐÔ•“ÐÔ}­^;Y½BÆ „ZÌ‚öJ1ð û™6Ã~&œîá‚&Â>¥g`/ÓРYlƒùR—òF¨†Ͳ‘~CÆDúéÀr·Ý"Ò÷ÂÏHß ãe"½ŽŒGŠ´‹Ò¯i R? e¾”kBqrFaÃþIøkæ‘j’D‘²ëŸR÷¥-aÏ¿+’÷c¿$C‹äFÒÄâàÿ‰—¥Ý ½írã®H˜Þ/0àEB­€úÆ"~"ü˜¸Ïd¼BÄÅxbKáÄœ–zDÄ8dü@Do¼¥"zÎM4·{GxRÀ^ Ÿš¢_wØÉQü~}íx~Ä>Ï}lÁã–$zÿ?¯÷SÁ0O¤èå †¹Z‹ðûÃöw½vV„/ƒ_Ыiµ` W„‡ýþ2äW/¹ŒÃD¯–°ó{™d|EôªXî¿W‹ð„€Zzj±¯'햃밆ðæBþ… 8‚!"üäVïwe^Øõ}Â?é3IêeÑ'GúÕ¢Ïbï}‚Ò/â èÓLÆED#ȳÈú°ãûÖf»¾F0Ü/úþ(ã¢ïZü®ïìÀ(2Þ}ÛK»1`÷ã|ô•ñ9Ñ·æ+òzÐák%"L2Þ!úœ€ŸÛ‡ý×>ÛdàSô'í@1 ~qd`Ò• /‰È±Ø_‘dÜGD> ¿-²æ-¢þò€€¿,z_4EïµIÁÀ—è8èu=è^½6Bßõz ç,üt0|1H„…¿þùS"\n_“èUSÆIDø—2n) ªŸ."\†ÝâD¯u5}ümÑ›ÇÝ›ý¹>Ëê·ø ±4¨nõ"Â0{¿>."B‚a­,Ñç瀙¾~žè33`}&wïÁŸ‹pb?G¶ nÀïCïDxeÜ]ô¹Psk׋>Ó¡WzO”ñdÑ»™ŒVÉ[¾}^‚ßqKÆÉEÄÁ0B¬ˆŠóÑZÚÅ¢û¥}îC.õÙ$íGÑg"æ+b[»7Gm(ú6„ßyFÆYEäzÄ!ú> øDߺ2®'¢fÄÖ²Dì$55x;ƒˆŠ“ÿ‰èºÒîÑ7TÁ)b¾Â>‰½Óq”˜áØo192î(bØÏ1Se¼\Ĕ˸ªˆi#O.¢÷àœDóy‰npcgÍýþ’ñ'ÑïâÑ¥]+¢Ÿå냆ºˆÞ‡õÞ\&¥ˆ^¼mwí’ñ,ˆù*¿_5ß|øâÑéðw£y~û ¤·è7ãèÝ2høŠ¾ð Dß¡°§"Ý€®""4 µÿ‰>aŸöN†ýþôl˜ò?ô_Ø-¡sƒæ¦Q„&ÃÈýDȾi/îš(B{¿äÕq¢û‘lëÐ!ïˆn†„çžj)º5tT›xP-º¾¸²YH…è2ãfIÖÌÏEçõ©ý¿˜¯Þ·=]¯¨h÷Á²±ë—o­–O¶1ü g$ü-gûàqŽœqR3Kî'r”ñ'rN‘óCÎ\ï#ç8ÄÅï¶ë¥[ä|'ÖÊ!ç*øÓÎró3$çwÈ/8?’z‰œÿ“z–œËåþ#çtŒÇ¹8è&Ž&çdÄ¡oK½BÎçù> øy£u.”œœË°.΃r?’ó>âî®ÎŸ'’\ Üß5\ÚäúPêUrý"å¹nH{‹rš܃3ÿRN)Ç)G‡8N ò29£ÇÍ™¿9gCÀ¡µ”sññÜšÈævU 5Ê%Œ#· þ\îdÄr?B;wÖ#÷â© KyM0ž¼XÄ+òr‚ḉ”7qþ¼…Aó°åmA o}À׌òþ~Gù‘ð§ó9®Ÿ÷ò}ù¥¿CùÒ¢ü!Òï ü·§Îß®ƒ¯SþA37” ·¨ ~nAQ`µs¨àÅÀí‡/¡>Kç(¸4K¬T†¼ia6âæ…c¥=A…fù¹[TøîWøkÀ<½‹ŠZ ?SÔëU”+_uîøšŠ5ÁaWPçoŠ+‚*º¸Ý·_QqÆ[ü?äg‹çʸ2Ÿ”ñ*¾ï—Ýâ«Ë¨¤æ«Ä‹÷,™!å3•| ý*9Œý\Úÿ^Šø1•&à9¥åðïK‹þ ˜JÇð÷_“r“J?@¼§tîW*¤þ¤ÒƒÒ>¡Ò«ÒN%w5é»qÐÜoAî0äUÜ*įÜF¼¿Û†}ïË¿›¼!øÂäž%íBr/zˆÜk0¿îƒX'÷^Ì›{Ö˽+à~¼Gžú3ž¾Ò^'óìéüŽgxÐ}lEž©Ò"Ï é‡’g'öç$Æãùy5o]̯· ò1ÞDä ½V|ÏËóå…y÷ŽÁ¾ðŽ«<ˆä†yóÎØøï{òÎDÜÅûF`z|DÞ·p^¼+¤CÞo1/Þÿ0þ²úXŸ²H<¯¬XŽ£j?ñ5•y±_Ëü8GeƒO)ò(…±ÿ! 0žÂIò܈ÂexÿŸà?Õ’çUÅ!®Qd ª«IUû™÷U¨¤ýJäù+ü›ÏbVLÅ~­x[þž*>àß}…sSQ ùTqçºâöoÅUÄs}õ°¿}m€ñ…CŸúâ ·}šÀU­öäËB>Öçœ÷‚üðM‘‚Œ|ˆ¯ï+œßÎàpSÈ÷sðñ ä; 9仆ýﯹàoûÈßòÉßGîSò+qŽüœ >žëŒûûŸÃ~öû±.þyî…?MÊYáï¿ûþEþÓwJê#áûßó­–rGøfa|ä|Ÿšÿ=y&_â¾FˆkU\_qû¸b‡Ôÿ¢b‰”ߢb"üì ”Ï¢Bõ®À¸DEcäsËïà>å‡o-ßû»|öKùøåÏ#RÞÏ/7ãyåR^‰ò)ïEy¹N¢¼&ò™eä~e{,û~uÙRħʦÂß+-qN¢Ì‡¸JYî[¦Â¼•Aˆ²ŽRÿ‰²¦øô"Ž-¼[°ß½ñç4Ä{¼ÿCÖ ¹+¼~¹„÷iŒ×Ûþö±ðš1?^¥ÜWÂq{[ >ê­yóœßœáÙ¸üg«ð,ûZx&c¾=ƒ¥Þ5Æë‰rVxZJ=.Ü?O3á^¿Ð=óîÖá»›aKA\­ô“‹Åæ¥c ïJûC>–fÁß*탸[ÉE¼Gɇ˜¯’ÙxŸ’0O%±K6a]JÖâ=K8®Yò²Üç¢d”<×¢Ä+õ‡(ÉÂ|–Äa=JØO/þñ»â$ŽG?‹ûÃ}нXÇbäT±QÚE¢X)å…(ŽÀûw„<+ŽÀ|‡ðg;þ~cÌwÑ=ì“¢¿à/¸‹Á,ÚØ$øDÑ×À}.õ£(úï]ô>Ö¿häbÑ>ÈÕ¢}Ø¿EÛ0ßEk¤üE˰ŸŠòsWB^-Ç>(úŸ÷ ßg-ÎGÑWrÈÑ*üå•€ÓØ2"·òæ+S[ý°­òæ¾Ïæßë¼TTû`Ï«§>‹u#Gk¼µqôÀˆ¶¢YJáÊgï]-¶öÓ½ÞûÑbÕ§» u²E‹MÏŒ}}ƒhý]+ß¹7zŠv©á½Êþ™'ÚyÞØscS²h7ò£C¸Ì‰Ž÷Ã-ys—èÔ¸Mœ1â¼èT_ùÊõ½D—öA÷·‘èž(qt"d»ñ“ w­"Ì+ý:ÑÛ*q_"ââ9Q!ˆ;ô»Š}]]âßD4ç•£û—¹·ˆ¾€< Ï_,ìoÛuV@+b‡býb%<¡®ˆ•°Ãþ"®ò.n6ôhüGRˆ„52ï&Ï#–Üz>¥-â©-¤öÐ^é‚|³ðÎhÈ9§*¨Æ#…S8±S9çìü¾³ä°ãþþ®×ÎÌŽ»È ±ýS¥mj9^²5€=™½ùÊl3üª¬[ÈgO@YÃd²úÁËü>àæÏSfü*ëiÿõ3䣭3e—¬/CZŸƒ_h„ÇE5ùi âÒd)I¾h1™ÿ‚ÿf^ˆ÷1“Ì/‘é:ô´i‰”Ãdš¼Ÿé圠D¦ä•w¤ãEÆM7fëG¾@ÆÅr¾É8YÊ'2¦#Ïkl ¿Îp q ÃW𣠙ȟêÿ Âk“^ n¡þÑ|qa(ô}aº”ƒTèδðeØù…s`âQáع…ÿÈ8µe¿L}\ÔyÈ¢i°Š>fOÈýHE¿Á¿*nÿ¢8~Kq¦<¿T<þZñ|ø‹ÅìGŸ?SÒvuI"ìù’ ¼ÉLøA%ŸÃO(<¡’¤¼¢ÒNø^i²Ô#Uãçk*5Àþ.}vM)òVTú=ûwá/¸£€/t›1>w)ûoc`'¹ß‚ÿí^÷sƒä¾ ûÊÓãõ„Àð(à¯yìx_Ç}ZœŸ ~¶Ï {ÓןíËaXGßxø½¾70¯¾%˜ßøó>Æ-ø~åç_‘vÂÃçã:`?â=ý­ð‡ÀÏóÇÁO÷ìS¿ ~¥ßuöÀ¿ö—ÂõW`ÿøýØŸþAÀÝúŸf{tÛ™caÇùÁÎ÷³~ö»$ÎQøíÐ{þLøa~ÿ»rÔ¯’ë/ü)°'ü °üñ°7ü1ü»Þl¿†À¾ôwæû´A¾×ßñVÈk-ø~à9…ï>ôºïŽ=ø á»)ã–Âwv˜ïOèß%Èaß%èOߟ°}çqß9äG|°ýü;Û˧a/ø~•û¼Jîóµðý »Õ÷ÆãÛ+ñÖ·v‚oìOß&à×|ë`·úVóûñû>ø|É}"|‹å>¾÷`ÏùÀžó̓ýà›Ã÷v‘ïu×kl÷OçïOƒþó½‚õô½ûÆ÷ ÛÿÓ`×ù¦Coú^åç¾Éã~ëå[(Ï¡ð±è[ »Í·\Ú7UóâïËë…÷óÇȸe`ýaùSØo!è?¿zÕŸ ;ÛŸËû¯ãò#Þ(üOAúËã!üˆçÿö—&AûŸAÜÄ?òÎ÷ÃÎ'¿~¢_Éç%zÊßý2>ïþfìÕÅùà}F¾»·¾¿q®ý5õ§H¾«ì÷1îÅwòÞwÏñayqòÿ±ó.ü e¼Fønîõ‡]ï;{Ï·çË·NÚ;÷#®_òº¡7¼~é?w(üSïXÈ}ï xï4ࣽo@nxçÉóòð>¸&ï¾ßJŽw턜÷ž€|ò^ƒ)«yYÆò¹,qÅ2È*³r¼©„ãW¥ÒOxßzúºlÞ·l=ôLÙ6Ø e8oTv†ïsvCÙÈ×òz˜ïòVXïòÈÉòh|¯X¶ û«Œë"ʶcþËöá=ËŽC®—Á?¤²øý«óó#.ZÞZÊ_*ï‚û–CÎRyOèûòpÔE”÷’çúá{^‡=Tv•ãˆ'å¿Sü0*ã¸iÙ׈ó–-áøàÛ˜ÿ²—÷(xÉãëËóB寡^¥|öO9ü%*_Àïù.¿ÿbœÏòyþ?Ƽ”/CÝRùrŽ?~ ¼XùW˜ÏrÖ³åë`/”oÄþ+¯„Q¾r¡|Î}ù·ü;ä9©|/ô|ù~œŸòC/g}_~Ÿw÷ë Ä«ÊÿƒýQ~ û¨ü.æ·ü>ÇÁj๵°ÿ+êHùOõ!—*¢¨¢)ÛAÍ1o-‘©hËvünªèˆyªèŒyªè yZÑëR {¢‚ó%½ù{‘ŸQ¨«€ Š$ÈåŠìç %?‡8ž§Å¾¯àsXa‘ú˜*²ùï.쯊BÌG…—íA?äXÅàg+†c¿VŒ†®Ï÷™Äï?™Ÿ;vwÅ«°›*fA^WÌF=GÅÛØ×ïð÷ßçç,…PñäQÅçˆßU¬âû¯†}U±ûªbâ|°žùy›xÞo ÿn¯ß¶Güû×T±ç»bÇIwòüp¡â[~¿oy^¿å÷ü–Ç·ƒßg'ï›8‡;xÿìÀ>¯ØÉë°õZßò{îÂþ­ØÍﻇÿþìÌŠ·¬ØyTq€çÿ'÷a~îQÞGÇx?žàßÿýTq~UÅYÈ‹Šßùyp|÷ǃ/ðº\ä÷ý“çùäFÅ_8O—áT\aá ÏËßÐ,G+þåñÞä÷¿}VqãÃÕpN}5ð¾¾Ú˜_Ž×c=ßçχ:òq¾ÎׂÿÞŠý‹6x?Ÿ;_'¼ŸÏ›¯;û ¡XO_æÁ×zÐ×‡í Ø_äëÇvK,ö—/ qq_*?7ß[}äãsç3@>ùÌÈ{ùl&ŸúÌ—ÃãÌc{¨ûÆWÊqo/Û?>þÀï7XÚ翈Ç3’?v‡Ï§ï9~¯ÿa_ø^‚<÷M…÷MÇúù^{Ä/Y;~Ú»ìû÷DvسكÉ^ÜHöoW,l­‘²aÿÛ#^c›…¸“ ùha;Ü ½z6àöDY¿*ì…¨s±?‡x¬})âºö}ÀÚÿžÑÑxXG6ìhÇ$Y +Ï_±WÖ[ Ç_È79Ûì &Ú…3øç †'OìE8ç¢^È)v^u{ÂÕv¶+ZÖo — x!×ð löYá,뮄ë)Ô=¹fâù®52#\—7Ëi„k9©É/˜ÿ9CË™ ¿%g#êurþ’u·"· ü˜Üdä•rQg*r‡À¿È}ø¿Üwî*àMs¿yÄzp-rßž7÷)ÄAsSƒa?¿Èm…uÉùï_ßÝiÇDÎqY7-r6ɺ<‘óÉ£÷˾"ëÀEö»2^$²Ège·’q.‘µõ:Y£ƒ€·÷DVêÔ`!¡È¼„ t™ˆƒ‰L»¬_™­d½NÕýùZXCþϺ ø]ëj૬ë/XÂú-ÖÉú3æÑz66X¨#¬×oÊ Â°2E¦qÅÌì“Ì|à(3ýÀeÀüdNG¾0s=êw2Ï!žU'x9Jd…_—¥¾.«~vÇÕù\<9ïdݪȽ„óW ñ˼æÀ ç…×–—|ež8´¼Ç#ò^.8ï}àÏòÖ`åA‰¼ãÀGæý w~oùQÈäg·•?8X¦ÒZäOúä(ëDþfàÌó!>šÿ7âþ­á¯$á9ƒ,X¿»`üý‚kðß áœ¶.´0 ù…B ç-祥ŸBî_X€xAažSX€y+ÌE¢Ðÿ³0þp!â|¢°#üî†ü¾·O(à8vÁÈ[üÍï{qþ‚s¨#(ø8éˆ+T}6CÜ¥° ü¿ÆÃläùX…8HÁ2Ä >ò_¼‹8MÁl¼oÁtàû fÁ.˜}Rðš\ŸªyÈ¿ œ`þøýùëGÈ_ \\þ\YÏ,òGOšÏ8Á|ßÓ†ŒäË"?ñ˜ü†È¯ä×:ïàNóö N.ožÄ‰¼§ çòÀGæ5]¹& )E^mçyuƒeñãDî)Y·-rOg™ûðö¹Ë×Ï}ZÆçEnŠÄ­‰ÜêÈ·åìB|$gêsò³ÎÉ1'TÖÏ>qsB[Ìé,ëi„ë6â0®#²Ž]¸VÉxºpq^ëÿûs ð©®5R.?çwsÌD®¿YÌõ<;€k²í h³šŸ“íÇ`¹L²íE•í'à l?ñ÷÷˛ƾÏu9dÙnÈsMö–ÐÃöè}{ì+û³E¢Câi²¿»Ù>]âgÉ> ö°ý%Øö Ðçv®s²o@}’ý(ì;û ð8:KxÕ8ùšj¼§ÃǸÀÙR¾’c;êê%°›Á?vv‡¿ë$YÏAÎAamrNÂßz“ëéëP¿åÜ#õ99¯I}M®&°3\=‡uáü“+ñ × Ô»»ÖK9F® ȧä4€™Óu^9O¡~>g쿜…²„r¾BÎ÷RžRÎeÄ=r[!ΔëvåŽoAî[¨·Ïýóš{UÖÝP^Kðä1n1¯÷Í›çå} >’¼ãØ/ùMa×åÇǘŸ+ëªæÛ=~´r€ÜßÁßtŸƒé©Îqûf°—=]`‡zâà§yô°Ç<ýaÏzfâ~ž¥ð›<‚?Á~óüÍ¿»ÿÔÛþ“7ö½×¿Ù[ ÿÃËøï4ŽO} Óû#Æëýñ¥²†gYg晴Sý†ÂÞˆ:Š|~T#ÄûŽAOäYgXÛÍ›DäXàƒ#[ÀÞŠøu²}vÏ þƒè½7÷ν^@}N¯ZÀ …çÊzO¶:XþÜW„¯^¾×Ø5½»dváCdJ„mD>‹íñu_ÞB÷Qtê~þ¦1ççpêÑâì‰Ö»‡PÃÄ·òZ­£OI»˜z ÌÍ»=àmê‘Õå_e£nÔC{êPÈ«K¨û¿ß¿2=ê]ê>óøÅ£Õ_£îšÊcÖ-6êvñ“}7ª¥nƒÁ'ÓuWèµð{S×’àkަ.öæ^øÖC]Ü2/M]š¾=òþg©óÅÿ‚¢.©Ï›üãÝÔ…¦ÚÜÜA]FK;˜ºLŒùfõ°7©Ëç!å_æU£®Jìÿ®SÂÞzëÞÏÔµõ—Ö°P׬ÍÏ­»Zƒº†,o·¿ô8uì óãÔ¦ÔzØþ:µØª^]ÝFMŽi2þY>—ŽŠ:öõ»¿PÓø¹kÔæLj>s‘][ù45(¹‘SìøƒjÖίȹ¡úw—¬ãPi{gtÝm*lxd|îÚ‘•—«…™ÑùóÊ;ªàðŸ5¾¹Òñ¥¡ºÊ[W#Ã?JJuå0úˆç¯£ôÑžg/ˆš69ÑðýZƒG›Oгf¨ÒkŽÍ¬û¥þó_Ц²žZ´ê¾åtúýö¢×g7)-¸¶îÝ®¢ñ¼Q¿]h$Zg)ÚŽŽ V^‹NGa·»Œ:ªVïž¾½q|7ÑlîÜ=)÷‰úïÙV¼~«¨ñûôöKZ_¯üÛúìùŸ¨*OÔigø§¬­êèE·»a/‡êú¾gû&†Qõߦ*³êÅÄøÌNMvµ=´ºp5 y>0²[ÔõÔêkøé>A½i7Ýc[žkEÝO¶úïë©§)ä]ÄzÚ³‚ŠˆÂû£.¹ÏxĹúæ Þ4u9sþhœþwBx]’òqâ…|V A¶2ü7i©àñIÿõ·ô)âÔêïQw¬ “éN"n—±ñnC/àÀ õ¯1l…|3•IùCôzf 9•@YBÖƒQ¶ñ ÛûÀŸÛ§!ûpYÿAöÆà ±¯nÛ‘Ž|¾c3ãÑ[!?霞—ømÇsrâ§ÏY…ÏÜ®àQÊ}üAy­0_y¯Èú=ÊûöE¾]ÖUQþðBåÿ4ß‹©y *àøI¡z¢ðsèŸÂK˜"%ä}ÇÑ‹xžŠî÷Pœ€õ,F!þ¥bÖŸÅ×𜒞Ðs%fÄ%J†ÃN)Ù=Vrú¿” —J×]ºñ…ÒŸ¿q[o ýù$OÉ®Ç|çè¹tM‘H›-ù¢DÚðm¤!>(Ò>|*"m1êßÒæA¥½¼RÚ›°ÓÓfÂÏHˇ]žfE]fšxÐ4Æ‘¤q=ZZ*ê¢ÒÐWifø3ié’¯J¤ÅîO/’HëºÍ4ä‘yxŸtàVÓR‚eâ^‘–ÿm®ˆ´aˆO¤U N0Mþ¼†HËåúE;êÛÒrŠ4ä™DZâ%iNYwøìî4;ê–Ó"Qÿ–Öõ–i5P×¥¼Œú9%㬔‡™G…ëq”/Éún¡|uÛÊñð+•ãá(GßD9,è6<øDœCY^¥yh¥~‰ö§P"-”à5ÊPè}e(âGÊŽ¨SöâßqžZÙ|(J®ËS¶„?¬¸‡ø‘âö…âê´·€‡U@NÅeàƒ—e=¼P\‘¼Bñ'æGñ+æCñ pÊŠãÌs~ â€ÔÓBñƯØ?^±EÚ¿B±‘¿¿ q$Å—à1Q,—òK(柦x u`Š7°³Pï­xyHÅDÄ'˜wæYÔK*ž–_ÁõòŠ-¨OVœÆ9TÖi/ î6íÉG%Ò>_ž¶ øÏ4®3NûUòrˆ´s,w®òy½ø]âß"½ði÷Q‡ûàšùEz äíÓë§Ü‘^8¿tô'Eú3ª&¨KUéð|¡ŽR¥EüH¥–uaB5q2Õqà¨â.ÔWò‰r¢¾š†ð¿¿€u¤eÀãÑ1©·„º pjðç õhÉ[%Ô#P'«þöŸúuÈgõ›’oO¨gK{R¨9Ρž y¥þqHõY¨.Ô‹S¿‡ú@õ7ˆª¿D|Xý âê÷oTÏ –}é„ú Ô5«Yî«—"žªÞ|¾ú+Èõ ÄGÕKçRy¤^ ñEBý©ôóï!ù…zN¾ø”PO“¸@¡ž9¦ = Žú5×c«³y\È!µr]­žSÊ|Éà[R#¯'Ô½eàQ¨»Êm¡æ¸—º¾¬¿êÚˆK©Q$è.â tü tøHú<ô-êMiô(mÎ…*ã¤ÏÁ¯AóÀOASÇ¡ÿá}i¼ôw=?Œ/B…à±¢lðePz°ü·‹ ¥äQ"íAAM·P1ß”êWð¨ö|(_Õ&iÿê¡„jêæU™¨ßUõEœIÕqçô[¯é»¡§Ó?DHúHÄ­Ò³PŸ‘Þõ©é P?œvqÆ´÷%ߎHC$ _¥ú0?ßvRYSØSePOUÖq޲úÀ–ÕC©¬.ìÆ²ºÈŸ”ÕáïÕa|pôTVvVY5øõÞûÀ±yï"Ïꎟ¼·`Gyÿe¿þ:ü@ïØuÞKx¾—ósÞ3\7u’ë¥~áï]èý ycïÆ“0ÎÁûòUÞÝ\µù)ïv䥼[Ÿñ äE½œ×÷nò©Êns÷Ã9s#>ì~S¸í> |uÕç%ànÜç€ËrÃ?î__p®É½¸%÷à¹ÜÌâžSxšá÷ž~À9y²³òXä{ ûÀ“9ëa»Ì£@]ƒGƒýæQ'æICœÃ“¹ç‰ÎÞÓ çÒÓqvOâÞ.ü섇ñužD®'°qAâð?>¼6#^ï1b\ž|ä»<ŒËòÉ}õä<+!¯Ü èUwÞ×mAÜÛíÄ9sÛ1n7òúƒçºx?w&ì·<n;ä‚ÛÁóýà3óæ.Ï]Šys—"n]µîE°ßÜy°Üy˜çªÏ|Øíî"~¿"®£(âõ÷Êø¦pÀ¹tÀyu?Ë¿¾@¸§èž ¹â~ö±›yHÜ,ßÜàŸîßùï×Ctßày¯†: γð4F<ßÓužn°K=Q§' û٣¾ôèQÿá±CO?¶^åHÄuËqÞÊôð»ÊÀ¤2ÆÓ•MC޹얻.ûxƒ²KŒÃ©‹øgy'à0Ê£‡¥Gžº<‡q+ðßÊÇ0g:ãmAž•É8š­ˆW—d<ÌYÆÝaœCKà|*º0 –ñ YŒ)ž«bðEÏ@~T€Çó!¾‚y=+çV±–ñßóï.q>»ã‡;Àïô%ã=|¹œ¿ßûb~|;9ÿ~ þ«¿-ã çüÙŒçóëGÝ›ð³¼ð'ÃÞò÷`|d]èßuƾҷûØÇ¼–>¶/|£ |ž ìc_$γ¯+ögÅ}ì׊Ÿ!—*¶rÝ×gÈU¼û°b"äc…r°‚qÀ ØŸí¨Åu\±_Ë+¥þåï£Î¥|2×…=¹Pžß•Ãåua‡–ý\lÙ>Øeœ§/ûþ[ÙT®û ¹T†ú9Qfäú¯äGêg\‹²®|ÿ\¯…úká݃ø™wæÙû>×Mæ:®ÁGÞ"à€½äñ¼©˜7o×u‚?ämÆõ]uàÇxî°^¸ýï9Ëõ^?±|Þ»Ò³õ`žå,ÿßÇwøÙ¥ÿaœ¥¿ã½K`<¥ŸA•¾ ùRú<:ùßSðÜÒN\÷uvTÉOS%ë¸lôI ê:DÉ0è½Ö¥u¢øQÒò¾øê£ù}¾Å¿cß³~.þ祘ýâbÔ]ŠâŸñüâïa—ÿ„}U »AŽq/Æ8ŠßÂ~+~•?_Æ>-~‹÷Ö½øMàÚ‹§â<O¹)‰ý[üæ·xÆU<œßk ש=Çï;óRì…>/f=Xœ=YœÁïýSÜ“ÿÎuCÅÕðþE—pŠ€3EG0¯Egp®ŠCUý;ó„m|P·†÷+BÜR}¸TѬWÑOxDÑÈ—¢‘Èo=ÏÏ{šëã|\?9(Š,˜¯"-ÆQdÇþ,"®_‹ãû„ãuÂ<µÁþ.êÍßç:À¢j8OEMä¿?ÿÉòH^ ‘ySòE‹ÌY8w™áXO+ø9…Õsdm ;À²QòH ˬ·xHa^;Ü<\ò¹ sømMÐCÂô"òÖ&…ämÆßá'ç#žbdÖpïaø¼MÌ×Võþ³±^ó'°ó]0?ßÅ~(x“qàíÈ—‰‚O1¿\‡X€ú«À'öKÁJÌoÁ9àù {@?F"NX˜ó]XýR8 ò±ðØM…âù…+!× ?Çú®†}Y¸¿÷ ã-~gœÄyì§Âÿpÿ¢ZE¸þùQQ¤áý3œ÷éh^ÿÉ\7þ*בƒ?A½Ãû“ë#Š>Πh=äyp’¢èè¹"Áû”냋öq]æ –'êByÿ÷d9?²8ß/fûª8ù½bîWum†ÝZ¬a9Àu ÅÈßâÈÇb ôJq>×·2Zq>Ë /ôlqôJñS8ÇÅnþŒñ³Ü/Ïu±SXn¼†8YñlØ“Åo³¼š/¿ÿß‚ã:⌎kˆ9ÎÉÁq1ÇQð8Ž"Žç8Šx®câŽ=ð§ß"ŽéØ…÷pT‚'Úñ5òDŽOPWéøySÇ;À«9f?Ñ1ñ0Ç Ì»#|X%ðŽ$È[G2æË ÞGx¾à=Ž0àÄ!ÀM9:ò}à<ÛïJ|…°ß/¢ý p<ösÀ‹ÙCœÍ~ñûì'û!è)û‰»ªšG¾ö½8‡öƒ’'^؃oÏþ+ÖÛ~~…£9x°ýÀoè(@]¨c⫎a/9ö#Žæ¸%ãUÏãkál¼›3<ÝN#xæœýÁç|q?çüÁ@ˆp‚ïI8w!.èüó弟«-ø]‘2O$\äÇ]yˆ‹¹Ê¯q= ýæš;×$ÈKó_ºÆcÝ]~ÉS[õÞ|-\$ÎC¸Ê/q1¯«q2Wâ¿®dð“¹BÁOèj¾ ç5?êN„s7~ç\Üšó-Äݤ¹çÎqÀé9Ÿ—çëaÝaà<]áwxPMžäø&ÏBæ§ùøWÏçÈz6ã9?¥Š·æ"×['GÞjwA}?y»GíA¾Ë«á:+ã3lü™#Ïíú\“÷)à„¼C€CöNäz ×àzÁK@Þò®àû?‡!Ïð§<¨Ã#omÔƒzÛ¡^À ^6òFs<~,yM\¯d‘ïñðý,ü^Ì›7—ãE…ÌÃ3Œë”^DüËû6ò°ÞUü÷­È{O`½§åº>¼ÿi®w⺌²ÚÌËÓ’ãeÝ8îÿ³,ã.cuY>ûÅ~æÝÊõ?Ï`Q?WU/Ú~¿ä„¿×6âßãü 3èakÈ è#õP®…?œë9îïyî…½ë翳ìO€Ýîg^ì7ìÈQgø7þÐ/þ6ìç·‚Þó7çúÉïÛˆë`ëÃô×P' û£jšpm3Øþæ\oÛv”¿߯?ü4Âw_žã'ë 7Àï𠮟ÝÂu­ê\wÀ¿ñ}Ë×ûa_ù~‚Ÿã;ŒõóýÌ|5Ç`_øŽqïè}ß~®“ÝõóíÇ|ùöqÝê>®óýzÙ÷#×£îÇx}`×ûöÃ.ó€ÝàÛ‡¸]Õýâ8Èσ÷`½âûŸsšëpÏðøçzØß¹¾öñùµðWƒõ×@<Ï_zÒ_~’¿&Æï¯Åß«÷ñ׿uªÃëWö¿æÕ_Ÿ÷WæjÈ÷m„÷ó7áºÙfð—ü-¹~¶ìp>Omx0¿…¿3×e‡Àö÷âý÷Ø9«â½h‰<¶³:×;ü…>޽À;V^;óÚ´§…ãeà¥À8T|ݼVöÓðËìó‚4ƒiÂn“øa¯ƒü«mú³ØÆ€Óþˆxp-²wÁ®Ê6!?•5ùØÌŸ±Þ™$ϾÈì¿Â:ZžÛªûðuÀOD¾(³ìÌÈ;er+Ë)ùEÖ2ØÙí‹Ï~ü"Ùwd~ða}®Eö‡À#f' ^$ë¤ì³&²šbffiÏÿ~£5¼È–52Ñ/,a²^áÉ÷íü¾åô)±L Ó^ö{ÿ_ 3×…˜ëoÚ4¸ÓÙÏN÷Àî2vƒouÕL;×úN2O[¥O¸_Å[%ž¶êïië%^ƒÒ$ü ›ÒB$.”»À¢€>GJ®GRªÁï¨4@_(ã¤ÝIŠë²î‚Ÿ/T{#õôKU£?W øŒ(e Ô—UïÁ×”üì±ä™à÷H®&qß”´ x²¤Ñr()ñÚDôï¡Äu裑Ø|% oIÞ8Jh‡þsñà§x#ðÙσ&^Îi’ð\ Ç©5½ñÞššÀ_ªï©z?úï©wÿT½ãW Þ5÷P•õO>'8/ ò0¤ñ!/§™&ó¢¤9,q¤­þgÚ(༴CPϪ]%q ¤k|´.}‡t ¥½Mº‹2/K=×Îð1çùwMö)c?ì\=÷çÒ÷•çŸôŒ×ç ¾Y?¼¯ó»T=7Eú—¤½†|ö}àÀµ>ôÑÖNOÞpÒ´F_4õVô½PD? u(ú{Ñ*‰;"Š…¦oûüÚøU¤ê |zúì‹ôê˜ß4Æ%*Q_CŠ;ÒŸ%Åëèû§PË>”ú6xVSkݵ!OìϪüÃ@®stp]f)×ñäzZæu¬x‘ùOfqç»\¿È¼¢c¼¾nÌK¢À÷}yÒnyÈ3’ÇüIœ·`»±âÆYÁþFÅ"ä+8Taá:N®#¨h“ò‹꫹îy×yäºeÄK¨¼%ÛÉ¿sz _¿Äv´“óNâ¾\SY?ø;eŒç+ËfžÆ——é¸ë’Ëš±½zóë=ýâÝ»™ùÝö)ÆáÝÀùìo€“÷®fþN®óõ~?±ŠÇ“ûzz7=¸/ûKßK½þÐîþžíáclsÝ„÷ç³÷¡NÝ»‡ñòì7y"_V•7gÞ_¯àï}Èyú×ùó9öƒ±ÆçßëEË[Ä~‘ù?¯ƒ¿_€óï-€?â-dÿŠ÷‰7uäÞlþ}>ãø]ìg>ø´ŸêÕ°_¦Àzz“Ù/U°ŸÀü‡^%ö©WÁ~¤’ýA-Ï·‘ýÄLöc÷WÌ+ã9½“oàçuM®ÔÛ’ýü3\×ðüeÏ[ðw=ÈÑ'þ¡‡ëÕ=°Ü™/ÉóåÅú”2?j)xÛ¨q*MÄü–Übœ*ûñ%S±ž%6Ì_I;ÌC1óIÀ¾,æ÷-Zù*B^Šò¡o‹ˆ?#ᵇ-¼ÿ²p3üìBæ.Dœ› µÐ …p Sø{|ßBô»£ÂxÂqП…ÀRátÆó¾É¼ÈÜ?±ðcÈ…BäY¨pä{!ûÛ…ßãùEõàÿ!^JEƒ±¯‹– n¦ö8·Àþ+N…¿[\†sX<ë]¼ûªøo|¿„ëNJˆy‘û#NR2ò¯d5äP â¿õ<*e~Ò§!JgÁï-E6*eüŠ›yÜjœ#7ó™0~€ÜåØ÷nÆ+»Çaün®Ÿr¿ ¾÷T¬»{óÏ’ïYµŸó»H|åçcóßD]Tþ÷CŠo}ö3å_E<¢ %òËð;©ÀÄ/Æy*xã)`¾²ôÑ ‚ËÓ…à‹¤BäE©pó[Žõ)ü“×£öQö{çé‹¶ ~‡óNT\‡qÚ½p¿bž×âÇô7_Sñ|ÈëbÄ縉ù°JZ1µ‰ùÎ<¼®c°¿J^À¾-Y?OÖñR>óÁä¿€þ³ùo¢ïp>x)ÿ0úöæ_”qsÊ¿½QÐë[ 9Q‹s[0|4ï0ÏûwgWpÞ™ÿïÉý6û¸ØËòÃÎ<ëjðÇbž‹{°Üiý^\ŸEßAï¿OEÌ+_Âòy] ¡Bà7©ð,—‡2ï`!ö}!pQT¨äsgfyŒ:d*œ…sVø-ËýÐ#E¬ÏŠ8Î\´ Ï-úžÿþóžãßKqso ׯóâçz‹×ùï¬çŠYN”pf‰óS2z²|bT²ëP" KŸ¦Ò¶|Õ°‡JŸ‚Y:r¥ô <¯ôð£|àUy•OÑ÷Äþ¢äõvüT{à~mçÑ/Ô¶ qÛRä³l/oi+ŽÆ8Õ_ [=Ô1goï…“Eö ÄY²ç N“½øŽì/¯Èþ¸Ñì‹Àg_Ÿ·­'òµ¶Bð7ØžFm⌶Ïà‡Û¾G~Ùv x|{5Ù÷WØ{¢¾Ù®ÁýíøÙö—W±¯D^Æ~u'Ô GoΣÙQŸïà8€cpÈŽõÀi;~?êãù£"Æ×Ep¾”ñEÜW¶|#¢¨â\EèðDÞ½ê>Ì÷\„s#Š’w(R .U”Žy-ŠC¼´(qÊ"â¼¾ñ"+ãlÈ—9—,²0~ “ùp]ü=Ø¢(qŒó+2óïRÏ‚øcÇ[‹¸opQ¬SáuÎ[ÿÆuþû°®…‚óÜЃ¢p1æ¥p&òÀ…¯3¯À›g!ç[ ¹®‡ùžœ¯ö¸Q]äw o#.Zx•yÎ2ÁQà w3Àþ\+퇼¸…à»…«øùà% Kj" ?Àz.FµðÄ« ßæñ1ß_áÛˆòy(|óWø2ð…70Œß¿ŒßëÿâSàºÂb~^.ß÷ÑïSiìðÒ)wþ×TÊ<®¥=àO–2Ž®´=ì°ÒŽËë{¾¦Ò6W¥-`¯”vcÔNЃ¥u!/K®±]±öpÉnØ%{ᇔü(åmÕýùšJþ‚ß\ÚˆùX#`'—²ýRú÷Õ˜Ârñî»±ò¸ôWøÑ¥Ü‡ÄÍãq§<ÊëÇ×äæ~ín7ó³‚þqOåþ3Ùί#¹!×ÈÍühî­Èï¸À¼¸`þסÜÇù=~ŸÝÌ‹ë©ÅyGäQÈÓšó ]áyÂ8ŽßyÂäü=|ô¡&ÆS“‡ùM=µ¹G;èÿówà—yø÷lo¿ÂöøsðYP)óÔ–ò8KÑŠJBÿ–¢^.°nð§Kþd>ÞЧ%{1®’Ä‹\SÉMèÅRæo*Ã:”fq½áPÞ¯±Ÿñû‹ÇàÏ•^ä¾*쯕gžÝšðgÜ}`‡¸°ãÝc¡ÇÝà= ø£Üoå à&ÏGì§8…§ç˳‚óèäýŽ“ûë0n?p`o¹Á«úľyð}O£Çòظ&O/þÔòsØWž2~¯aXÏDæ7žÆyû9àp\“‡qÕ¿w³Šç…ó÷¥ñا¥á,™ïúñùâk*å8CéXžçOY>}ènŠû¸£ùüÛ¸Ïp<ßÍ< îÏ!ïÜGyÞ˜ŸÐÓv­‡Ï'œùŸÑ_‚oúÉåÍ—ù1Êû÷ñz~rÕߟ¥¼iÜ/lÁˆ`aÊûñù¼wÁ—’·üòyïAæmõq”Çñ¡|äY)?Ô0 _£|ä·(?ýÉócÑÿ-¿òOû¯üæÜ›àUɽ*ûKRîyø»¹¿#Ï’{Š?O£OwîÉëE¹gùwç‘§Éï4åþ‡Ï«%ëâ>×”×]ö¤<%ø òò×É{:ØfzB`¾‚ ݧPÞŽ`¹™‡ònÁ?{<^Q%·Ù¾pÛ#ñ5¹™Ï½ËAài©”ý»Rô¡Ò_è§Gy¯ð­D×C½h´8¯è ¨WŒ G]_L…~äÿÿ"b‚ß+68ÃXî7»uÃqíÁËÛyþXêîbÆ"³<.ÑÿCž8úcôgˆþý†cêÀï‰ n3¦~\Ì~ÔÃÆ.’ûAÄ5ó)┨ˎG¾NįGÿÊ+x›ç›ø ê<EÞ<‰ûÏ&¹Ðo8iðŠI—HF\@$«Qgš<ýX“×ÝÔ>"ù4ê…S꡾>¥—ì¯ R²Q?™ò4p‰)à)¿"ožÚQž‘šüxêKÀk¤r¯¢üEEŠlh-å\‡Îõ³ŠŸÛPÖáúù®Wî<z¡\‚õQD´ò׳_’Æë’–Âu‰Õ!r¿u²“¬ß$ûH‰$ûÁ6IJ þ ûAèû?ÌÿÔFúÕäHF?GôoÉü19†à\8F€×Êñlðø«È1òÏñ5x.¿¡¿…}Ƀ~•Î\è;çsˆ»9?‚ßàÜU7èP“ó"ø1\m¯'„\:ä\ƒ!G\oIܹ¾”<~äúQòë/éGSN#‚ý–3ñÀœE²ï(åüý’s[öõ¦\¶gr㡟r5rÃQnÑôÏF4ZE¹£Ì]*yü(÷ìæ¼¦àßÊÓ°| ~¼ùè‡ñx¼®ªßÁ!ÎîcÞß=\'µƒyPÅZ®ƒZÃ}²–3Oï‡Ð»‹¸þi!ó¾˼½¯­˜Ë|Âs˜gw6óär¿ ö?*˜7ªb÷ñšÁ|¹¯0Ïë+üž/qÞs2ó÷>yU1žùa™ï½‚ûÛU ¯+U áß`žØ§‘[®©Â =[ÁùÁŠæ±urß°,Äé+ŒüžjæMe^ÚD쿊èË æµªèÉÏïÆãoÇ<»­ù}šósp>·çCï ®_ÎTþì¢òUü<ð RùZæ[ç}TU§÷-¿ï~OöCʹ_fùYعå\OW|߃þOÆFÀ/Šù*]„ý\ʼɥ¿±Zý¢Ü¿DÃþNù£u|Mî‰È‡VùßðïQ‡Mî[ìgw÷ä{qÎRî ZúÎCé_œWºÂvr}¾__ìw$Î_UŸ•aÜçr ί›÷¿ûm¶·"^ïæup¯Ä:¸™_Úý3óÖÏý;ã«s>³9ü+OøgžÖìgwGžÒÓŠí¾NÜ/³#䚇ý1÷«ótg??„ýú(9·ö)®É“Ìür°Ó=…ì @¼Ý3šý籞ž©Ì?÷ÇÃEs\˜ÜµxÞ8àörÿÐÇòf¾_ú/÷EÿFrs|ß=þŠ{ö•{+óäç÷­ÇþH{ž§ÇÇùÇü-¿w%û_ó|/‡Ÿâù„ûÝ|Œõò|Äûi1Ïë»Ü¿tû‡sù{ópž< °O=óy^y?xæb\ž¹Ì·Çr÷q¦ íg\ÁÆ ¼ýä]ÌþÊç\ïÏüàÞŸÀö ÷*ÆQVû‘¶f^ÿ¸oY?Æmã-¬ˆ§•åAþ—•1ßÀīʞ…?Y…ó˜Ã}Þâ~o1>zänÙø3e0Þã#þÝR~ÞÒGûöxw2^`㸬÷8ûYðÌ7y\ a7—ugÞ„~ˆß”¥!Ï[…Cñò{ bÞîïS6ù 9¿^ö.÷§xì½øšÊV2Þ{-㸿QößçÎGÙ)®¯¾¹Tv‹åd-Ö-X>s_äòî7ÉòšõTy"ì¡r%÷ÅPsÿ#Î{9ï“r®_./åþnÆñ¸¹.›ñHåŒ7/·c¾Êyï—Qå¿÷g\øXæux™ûëÎgžÈxŸqaïzÆ™`Ÿ¼Ÿû1œ>óP”5äÏöŒŠd|¾–û‡d✕b>ªú¸7T¹PÆüåeOqÿ¼‡0þqï‡ý-Fò>Çûr÷÷x‘ûñNã:û™üœ·ø¼Ãuïó:Â8þ|^VñsVó~XËûdÝ£uëUûíNïA®ç8Îøxý+ü÷Å•­ãs¼‘÷ífÈé²m8'e;ù½¿ƒž,ûžë~D\¡ ¸j*;Àóu€û¾äÿ‰Ï÷¡.;Êó†:]*û…yJŽ3NëWþÞIî÷ršÏÛAe¿ó|žƒ_Qv^Æ?ŸÏz>Ç›à— îoÌvJÙþÛCe;ù½wsýÅ\_±ÏßOˆ£—å}sœçéW–KÞóèó'Þë<÷Kù“ßûßï¿ß?¼/Øî(» =RvŸû„Tg{‰û{”sŸêræ})oü ?3ŸWÖûå-¹_Jkìërî'_Ævf$ÛqÐcå©lÿ=Ö'§êœÕàùoÌý|Zð¹iËûžû†”uc=Íýcˆ¯³y]J¹>eÏÿó¼ßÞxÔÞªê¯ÓŸícη—g³Ý¬a\a"÷ù¿úÙÜâþ:·Èy–òjÌ7ÁyõrðMæ—û\7æyoÎ}}ZA–wäþ2Yþõd;<‚׋ûž”3?D¹‚í_Žÿ•ë°Ê3˜7ÃùW®GÞ ÜŠsQîd{>ÿ½ˆŸ[Äã/å~Ûe̯áã÷|ð9€õà ~þ¶¿¹O]9ÇaËGò}Gs¿™±ü{îk^Îörù$è¿òI¼_^àycüjù‹å/òþD½<•Oç~9œ*çü9÷u(ŸÆ}€ë›TU‡u vŸ÷4ÇUÏ0>óÆ#þɸ?¶»½ÿ`¿{9®é½ÍròïÛE÷yÿÕàúª:Üç¨ï—fзeÌ#]Öë\Öžë¬:ð÷ÛsŸ&öûʺ²öç—ù;ó7á:Á_¹îã2Yž{¿•çõá:|Ëzü¶#«'ó½Á}C^…Ä}æÈ7…yS8Þëãº6ûë¾ÜgÄùï+å~&.Œ£ªÏ¢ïëKÂzW}r¾ÍÅýñ?|áÜׄýr¯—¯÷Eaü±¯6÷¥9Ëwxþîðønñ|s¿1ßU>Ï׸¯!ðç‰ñ¾[ØßþšÜçç_Ø>®#õí‡<}|?UÍG#È;sÈ?ËQ{ìK'ì{¼‡¿;Î×g’?Œç­ö¡ýÜŸØ/þæ]ŠÄxý±°üIØ~ÆGú•8¿~¶WüÜ/ÀÏ}ü&œ¿rÚŸÍ}TmŸ~ð¹<®É_‚ù÷s’¼1äŸð Ÿ¥¼~òœqÿ;ß×,/7qŸ)Îcû~â}Âu¾_Xž\àóýÿýä¿ô¡¿ö­¿%ô†¿ ô˜¿'÷í+?Ž£/ÿ`Où³aø=ÐOþÁ8÷þ±à¯xl<®…¨ô÷þ}(×éOågÞ&"×)>^û ¾µ×åöåç%q¿ZøvüyÜW–y•ýÅ\‡ëÎÊ?€ë{æÉÅ‹P/Yüón,~©˜ûysþ§¸SQü#ól þ&Š1ÏÇQàÆŠO .¹øóݯI üVQÒù¬’:ø}IWàK¸¯UIuîsï]Œ¸§(ÞÏ|#;PZ¼¼9Å0Þb?pnÅéRï='®E±’ùDÏÅ}0ïÅ<®Ô=‡3QøVŠ{ðs™ï«˜y@‹Û2oI[æ?aþžâNÌGÒ ü€EÌ—\tx°âšX×â:ø{q]žwàÉEÑ%ÔÅ]aÜà?Œ§ûÏ+úø´¢ËxnÑmžø)¢8øÀb“y¡‹ŸB}oñXæašÊëû_Jqu桪Áãf>›âZ؇ÅõyÞÛ OWÜù`À7ïŸÎÌCÕ÷Kóµ/˜_žÇ¨Ï-nÈøÇ+Ìs”ù˜ÐçO}õˆ½òàZ1^±h6ê‹‹^a|'÷y.Ì|86y>þ×¢Èοgï¢L¬s‘‹ñ“…ð™Ø'Eé8ÇE™àÏ)bÞÄ¢§™GêUŒ»ˆùŠŠöð{¢߃ù}r¢Ty¾ŒÌ“Cq:óø¤ñ>|°¾¨ÿÅjæÙR3¯Nóû´EÝp÷—*º œdó]N²èoìÿ¢xŸ¢ÿ˜‡êÏëƒ}|—yºGÁdìû‚)Üi2Æ]À<‘?o-˜‚}^Àü%“¹¯Õd쟇×x~Õ÷_âþP¨+À‹‚×  fA®¼†u*˜ œnÁ î£|š(˜.çãáûOç~VÞ"xÚ Fñ{°-ðñûº±^E8Ϲ|_ÎI•ûQ™ Ï ´˜ß÷£JÃ<0¯RA*äQA:äbçºÀÄüX¼? rqî Šq. š?À»¯-’u,B{ 8y üçº=²o“Ȉî=y ¡ï€ñë߯¾4$È:aa8~ãl|šrð^æ6à0Žþ=–Rô#°ìÁü[SQo= ùùx­³tÀ7dý ½=¼EÙÇÀ‹m‹~Ã6u¶£Rq »JöÍö1èëh_„÷±ï/’£vàŸ-‚®½HûüLiGÐO2½:øûÓþõŶ8‘6+HW^M¤½†~™éÁ×’^ø–´ß€óICžP.GßÅWØŠ/evÕ¼ðµHeޱԲΩêß™o@d¼)ë±EF=ðMé€]{ë¦<¾f'ìMÂäÌê©o õ€Eá—WÐ7à‡à}_uÿç@U¼Xi$®D(7î'Þ—ßO¤4Ÿzò(ðÚ'ÝEä}˜¯äîà)OV _Cò—èŸ&ëÀDÊZìÿTÔ{ŠÔ{àËWÌGý†’ùÃyßTÙãYÑ/.+x±¬"¹(kú©e½‹¾,Y›ä>£¬?%~‰²g’Íq­lŽGf~.ûÇ +da\„åGy°ÈÂ8Ër}Éò*úâX^/eúáXJÐïÍ¢—}OÉ¢F¼À’‚x®%FâÃÈÒuü–Ðàò KWô±²´û8HÜA––èŸc©Çϯº7ó É#Iæ¿·3ŸƒŸmÞ ¿Î¼ïef^ óðe˜gJÀ™Ga}Íà¹&3úkY/Ï!™Ã€/27—¸02‘L¿!N`ú6(¶ÊÈô ð•¦O¥ž Ó"ÙG•LŒS6Iš¸P2ÁÞ#“ þŸ)ûÉŽ>I¦n²¯™ZK|™ê¢‘ñOôI2†ßmü.øÚuɸûÀ¸ý‰ŒÜGÛøp£FØ©dŽ8­Ñø‘±<FãœõoïÒÿ Þýðè÷Éú.ÒÜš~Ëî`MÒsý²þœý’Ÿô3d`ÒîT?}©ôå¨+Õ«—k¶îºLú:RPÆÄ“2É~›”QKâ«Î™<¤ ñ—¤oˆ¶>âÜš;[‚9Òœÿ ¨H³}µ4°gIc‹&^òî¦+æKSø9õU)I }GêyàAQ3ŠZ¼ƒº›ä3$u"Æ­¶bžÔŒË~œÿ¤ê½¿C]­ö΃®ìJºnè¿¥k|(Òþ\²ö4óÔ”ýyI?—tYˆ+ë¢>UGÒþ ÝÈEÝ|ÄUtЇòñydŠß"ûR‚Sö¥„yØ gÀ;’èiRâñÊ`#%ù±/’~•ö%Ï•xXJyç$uŒä"E âKŠ•ˆG*Iœ+)Ç!N¦Ü)û|PZsôÛLËG=1óò<ÁÓ“ž,ë)}7øgTè?Eª8ßÔr†r$ßÑ ì7:󦮃þ•j¯ãH‰³%õ ôUŸ XiÖè<5UÏçy#ÝZÙ?ˆt¯A®ë|軦KÞT×ó£=‹ý¦Ý û†v6âvÚAÁe©GZøƒ¤mù¡ùçT³zC3JâtIcŹ×ôD_NuPÆÜ$õ¨CV—A©»ÍS'Ñɤ à#zõø(¼XDÔ8V•}mHe@?Ïôg¡ÒÇJ^%J'ûÕ’ªð骅Rß“ª+p©éïAŸ¦£_ ¥ªÝn”¶uáiˆç¦u—x^Rò¼+Ï£ž]ù7p­iY?Ji ßK›¸cÚ<ÈÑô†ˆ?¦?‡ß«8®¬šTM‰š`þÉ!í•À:CNÓ΀vŽ™EtxuÙà‹Ô…²©“ù|·Ž^]üNêTæ3bþ u‰ìSEêLàÜÕ#e¿ë€žOì;õ½½Nö ÝX"ø³IµRò{‘*땾Nö±¦t~7m»´ë)Í =¨„ýKJŸäï"ÅÕ ½¨ƒ_VÌE|^ñ%ä¾bΟâÌ¿‚‚ôQÉ”ú6ôHª\v}àñnE_è{Å6ìGecäí•yÒÿ¨:Äý É<2½¹IŒ÷U·ÂøÕvØ+êéÒ>«ú=_“šóCê[Зš~/‰ HSy©™ƒ¸³†q%šbÉÓKjæWRXØ´zžòÁw âº)UxP)ÝëÍ)ícØGÊ‹È×=6>áäø£3v·S¾¬Î ÙÇQ8èçãt£Ü9~ˆsœ´#„sª2HÔ œ¯ñ÷Ø/p®„?áܼsxæœ{ëw hÕ°ÿ çeôårÕA_-W[ô‘sõBß/Wò£ü”|-\¨Ïv9ç'D¸Jåy®àqsM‘8uᚺ×bÔ¸V¡O™ksqpÀÂõ=âT.Èáb>æœZR~ˆœŽà Î釺ˆúnæÁÏ‚~w9Ì3Ÿ3}µr>y4åüÏs~^:ç&ð§;÷^»ëßÂyBò\çYw"œ7À×ì‚=%\Mé_ýà:0_r½…«‡ä÷®Äù\±ˆ›¸RÐÊ¥AÍeF]†+õì.·”§ÂåCA×0é¯?|®…ëYÔï»&€'Ð5|z®©àÓsÍö­p½Þi×[ˆËºÞC¼Ïõ1x¡]Ÿòz¯äÏUÌc¼æ±ñ­áñmC݇뇋AE)\‡çp1ï»ëø9Õ/Éi4ãZˆœAz‚Q"'ñ˜ úåä#n‘óÆ3ñ®œW¥½'rÞ/ßÿ¹~˱OœŸK;,°ÏÁÏïü ýËœ!^æ||ÊÎwÐÐ9ñç[à+tÎBý¸sú(:_‘ò-pžPÇãœ"ùË„óEð>Æ£üð}^Cœ×ù¼´w„sˆÔ—ÂY†õwæô 6èNø°[œû'å@$óX÷•}ÏŸ’WE8ØϺ3߯9â Nî7⸉¸¾ãœ´‡…ãpÐM+ŽèÛì@]‡p,¢ãMi/Ç‹ÌK9çÊÑ~¹Ã‰þbŽTð3::âÚþóWîAÿ2;ó£Ú—ÌŒM'òXàZØß¹}:žc#õ½°ÃÎö\é_»ZúŸÂ%õ’°·C?z{mÄ;mw¥ôp>àÕvBÞ¹Ò§ á EÑÕ}Gð„óWðI:Àùs®o„sò ÎÒ~ìôïs>ÿh\”¯ë‹x°3|Nèwá žu6ƒÂ¾¼£öàµ/bÞù9Òžö)Ÿö1àc·F=šÝ#í awBŸÙПötìo{ô‘±wC>ÐÞ&0m[w {=ÈÏÇ÷_ Û¯Á2Éa[ƒ¼­÷…m"âƒ6opû6CÀüŸ2UØT?¶nˆóÚƒÏ?ûŠŒ+nÞ¤Zµv“…eä›%rÉ’‚¾ˆ–è—g‰CÞÀRKÆU„ùxGÍПÓÒLÚßÂ’„º8K+ègKøÜ€‚œ),-4)pÿ&ȇ˜Ïñ÷ɸ’°€w³j¼|xžä׿ŸÀ[a~ü,fêÍS‘3s%sú˜£o6ÝF=¤¹xSÌi’T˜î3Ïj8òæ8צ3Èw˜[Ã1+¶h¶0— Ÿ«Y~¾æþØwf)ž óKè3lžØmýæyèCkþü¸fÔà óg²N·jœ†Ià•1̃Ü2€ÿSÁï)Œìãÿ¤½,Œ ¿6þtP„õÂDàŸ6½;Æt ëby6¨€…™û8™G"_`. 6ÎüU˜ÁÛä9šç3v‡y*ίy5âúæÝAõÔ@˜7J¾3aFý‹0¯=ˆÿ÷øãüÀ3$ ‘Ð{ú»àß54@CCä+ ƒæí`aè$ër…žûöê×K?]è_…<ÖÁ¾ÐE_?zD?ö¥þÈ/½|>úሃëÑòa›ûQêãÀ­¯‹s«ox~Æ/ÒŸK¤¿*2ÖÂ®ÉØ-ùgEÆÌ û0Jd8Ñ9£ó“Ñv²ŽûÈê>A¾W‡ºF¡{ë¢ËAþF—;E×v›. v ®1ú+隢þU[ ù«†ú^­çUëĹÔr>U >T¡¹{IsòMs6˜–± ÍÈÍî€öû±Ð|;C³ùJÍìG xb…¦ο¦%ö¡¦òê-Èg¨?B uÖUm@¼_ {HÍýéâùô#ò´ò”À³*(Lò² Õv—ªµÀ¨>Ç>W5 гšBÕ|Èé¯ÃîHŸ)ëµEúTΣ0ŸwÚ}¼GÚôïVþy«¼»S¹yå^îW=‘ûIsÿc%÷CRö†}¤¬ ;MYzðÿöYˆºŸ>¿Ãÿ~Šhü^„ x¼ˆqÀ•ELG^#â į"~•}”)²‡Ô«i”çŒ"ʾ·9õÇ}e:î!_UÇtôYî\¸Œ.sS†Ôo{Xtk|bã¬vˆn·Ví™±¢ûVäÓzLó·xq–V„h^kôî»ñ"ävÛZ÷µº öLO· BÞ+œë¥{¡ÞOôú9àèŠÞË‚aÛpÑçôÇxUÆEDÄ?èC¹ú¥¯ë5 û¼÷áêw ù¢èìãè¥È{Å´EŸÞ˜‘°Ëc˜w?–û,ÇÊòq§ˆ=L;d‰¸7LjoÕ0(¨EÜ ø%ñIðãÁ‹$âKч8žqTñœ'Œg-~5òœñ'Ñ¿9a$ôub›€»pì‘XÎ×Ü'+É>ÊI_CN%7ÀùM.Åz$ç)’ÿCŸá”ñ˜ÏÇ÷‘vââÚ¥Ègh!Φ]„¸ˆöÙw›´ʾ¼¤ý|ÞÚO¥œ"í 9^Ò~.ëòIË] ì‘À8xÞÖ€w\ó!{øã5/ Ï ™/ý Ò yà=Áªé ¾OõŸÈ³j8­‰í$ž y2M"òsš:wAêŸ0>õûÏ@ꩲï8©+¤½DjÔë¤î.ýR×ÀºÐ=àrÕ ÀwªF¿R·/ºü¤Þƒ¼°úöŸ¦xÜ5‘gP‡EêS˜GõiœkõßÌo_ñpM(ç›ñHUNi3/þà¹5ÓÀ§ aOÍ'2¾^%oøš4Ke¿ñÀ>AžH³õïš³ˆk;bÞµèïBZ%ž¯5Aßió¤|y(ÇòXÞ Ï…vâ¤Ú©Áp’ž´_ÕNFþW;FÆI;\ò¨–óZ+âÌÚ,ä´6é‡6ï«õJþ؇ÏÅ5i{‚?X‡uÖ†±Ük"㤹 ¹©ùý4ŸJÜÉÃyÁ5i>Àù×|Ìñc®ßÔž÷£«m5iîK;ˆ´¡8/ZðÖZŒo’Ö'ýÈ€üî@û–ì7÷ð½q›RƇ¼¦–óšëÖ{=°7z½\tøé'Qød¨çnÉÓI¡'Ç ¹ZwTÿ—?¡Õçz›>ôQÈš:Y?̤âß¶¿ó9…œjvoÄt i¶àæŽ>©ÇïÃOmîs„z¼õC‘&ãõèûï’?>}†º󉫻ÐQw Îm·ï‘è–3¶×É]?S·˜-§Óï·§®è¿C]¾PœŠ*ù:«>y®eD.uü~vÝW¼iÔ±¹¶òéIÔ>M×2ÇPHm?(H¸£È¦Öÿ n<Ìe¥Ö •ÙVB-?ŒzúçÑÔ¢­ö|ù/Fj6ò®ièÆu ¨qÃwVOž~‘σ¾ixÌ»¾ö­?©Ñ¾üµá·:Pý 8o * Ïê–lþc΄•ëçöž_SIu·£^íŽRÍß·mŸ›EÕJÿWuw£þËAÿQݼ\úm£‘Ÿ©þÉÌéÜÇ›ª«md¼_õG‡/f–üOõÓ› —ÿ4kså§…¼úÁ–ÊŸ*¯dŒ9šSyjâö›Ÿ«¼ò¾”«•ÿ6h9)y»¶òÞ‚ñWÏý^OT[ÿ¬–£ôxíÞ ¢.÷¯m°÷ù_âžî,šô)Ô¿;R+š=;ÂóÖàÆ¢EÌ•­ÖÓE«>—¿\vj®h}:þ…MÓ7‹¶³Ü+6Ðk¢}ýW¿_5aþ;°j¿Dý.í ŠF¼¢çȾ3½}bÚà|Å̺T¤3' Í•ïQì`äc™w1ö²ÄËQì_³qàâöÈû^ÆI)žíÞø·¡ß’!÷T¨Jع”ðƒ´Ã(‘õ{âÍÎÁ%IøÅÿŽP҇ț%} ¹‘t|\Éoß‘¢ÚýòËÆÒtÕî׿5îZùâ>•'í‹þuÁ/•פ›î®¼q÷«==ëVþ{¶Q骖›*¯ÿøÑèn½Zyu}Áñ€&ª¼dþßÛW\ªü¯ókîÕ[)ªí¾0jâó¥•÷»/[3qŽ:‡»&†E®õ:J>-Ñðj`9šÍMw¼ÿÅðìë¢Ù€ç¿ÔÔm$Z¼¬ºµù€h1lè ³ççŠVúÕþ¸Þ_¢å¶?Ï¿¿c€hýIµô ݈Ö/tª¥™|R´];õ?ïGE»Vk"ísßkÏØßàV?Ññðo±_—^–îxãÆ¹Þ¢ó³ˆ/w}mà:å‹nû³·¶Šn•Ï¥uÚ\]ôi*\ÑãÕ×vݯ%BæˆCÿèÖ‹”£n»‰ž*ÄóÂ#~ÞRÊž¾¶áÓGçøT¯å¨;꼜è¸gïpäGû ¯\ô;cwˆ8D$ðq¢ï—È“ôs"®Ðσ|Ot"p©ÑÄm£‡"½ó½õD1e&bîÍÅ"æ üý˜ÝAñ>AÄšGˆýñ´ØJÄcã÷q—‡íïzí¬ˆÞMÄð÷x+øÛú"OükÂðô%¬b8¾Ÿ¨“ý¹D¢ñÎÄÅÈ;&®CŸºÄƒÈc&Mn8éE'É=7Læx`òiÄS>Dü3eââ©AXë+‹Dê,ÉC)#$®O(Áƒ/”‰ÒÊ+2_Toâk‘Öøç4ô)iG×K;€üxúDäT½€³VõE_õ âEÔù= AÝÅ!ÞCOcý¨yMúqqumô/VwAÜQ/ù9…º ýÕsÇUz!˜ˆêJÄÕ𯅦ø5JôuÖ°¿4Ï"Þ§Y‰ïk6àßµ5Á£¨MA\T¾C­A∄vpåÚåÀ÷kKüОåxt¬“. qOâ°„úPè¶H?RèPo+2"0þŒaRVÍ7_‹Œ/'Í8Œ|…¾òÈzæ…׿¶~C0ÌÓCèW"Ž«ç:ýïÈOò¿7ì’~“0l@^À°q?Ã{¨0ÌB>ÐüŸ0¶FþŘܾ<ÍÂ8ùãx<Ǹù&ãûXS]ÄÉLÕdÿ8aÊŽÄ” õ’0uľ5傇Óô:êLßqž…û%šîJ\•0{Q_cÞ|¾…ûXê >hIEžŠóqUóÈq7‘Úõ)ç‘çIíǼ”­!‡R9•Z¼L*ó̧®Á~TÀÏ «n'OËø‚P€K(^GþVqùD¥çV9Mþ@(ÿ`J-ÇÕ]À«¤Â>I¯Žünzä‹ÓÓ‚Û¯¶H纚ô?ðéÿ¡Ÿ¹jŒÄí ÕZðfâÍ‚Rñ}€ø:½|2]DœWÝ}´ÕO!_¯ù¬žüŒú0䘦êt4jàW4“пA³}&´Õ°O´±8ÏZä––ûfkÁÇ%´Ü¿^û·ŒC ]oÔáé†J{H膡ϣî Ô‡è®!/¤Û y«›Åuˆ§‰Œpi—‹Œ˜ï ½ÄˉŒºË8º­èW®›‹s¤ˆ¼žŽåžîkÎ ýŒúXÝ}èÝaà tÿaÞ2Ú£Î"#ø¡ŒBä{2Æ!®1ï™±yëŒÃÈ[füƒ¸¸¾•´7…>}¼õÙØú§ïÕ¿>£ú à/Õß—ñaà~åÖK†Ï6,–¼Ä ¿ ¿"ße¸‹|ºá}Ô—ÖC.8nm¸„úcmœ?#×¹ã‘Ç2"q-Œ^àOŒc oŒ3o1îÄø?Ì»q.ðdÆÅÒÆ­RT;¾F®1^À8¿#®nüyBSÔ…˜Ú#ïkŠÁ|¯@ï Åx)G ŒÛ‘4¾/ãf¸òßøò@Æß¡¿ÿAÿ˜¿ ¼ô „‘÷ñyà:Œ{!÷Œ' J¿GÿŽÊxý`Ûqþ»ƒa›0ÎÄ>4ŽŽÀ88ãDì+ã›°CŒËQÿdܤUÆcÒÿÆUÀ¯§ _n´I¿W ÷4zÁƒküõ7ÆýÒÐ ÌøxM­pÞL‘8§&-ö»)òÌÔ õj¦NÀ‹˜Âa¯˜À§˜gè7Sì/SWàxL€0É0Zàø*S7Ôa™B×7;ÇtúÒô;òêæ̳Ù#óÂÌöù*ê†,uŸ´tD<ÄÒø M ‹x â~q ¸–9ЖÐÖf¨Ÿ³fÏge»Èú-pLÖ­Èã[×@~eÂ/™zàŸ2'I?Zdî+sƒŒ#ˆL>ÿYñÈOf1Þ&ó_Ø·YyÈ{gqÝ[v䑲Ÿ—†ó˜]r!kŒ÷?ÌŸƒWLè÷Kÿ^ènRÿçÓ§g¥—}»„>89}òÔúŠ 9]Wè‡/¦<—žý‰Œ}ЛCƒéÓ§DÆDÔ×f C^7ù¡;„::ÝLäYuo˸¼Ð-‚¾Ñ} {C·çU÷>Û9KÁ3­û¼Øº5èg¤{òC7 øÝ«l•¬þÅB*ó=B{}´Ÿáœk‡"§}vŸv<ô“u:B›|¤æpušç¡‡4ÏŸ¨õÖ<#qÈB3 ï¡é‰}¬i9­iœ”†y§Õ÷ÑG}ƒóð—°OÔ§ƒNÕB}v—ú{à&Ô[eW¡^ ûFý%xÔŸ!¿¬^?C=‹—ñHøÁ>Y¿Ï™ÝQæ¯D¶ vh¶øÌìÞàáÎn|Qv}è‘l%ð2Ù\'™=]úç"ûàÔ²'@Že—בmÃ9ÉN‘q3‘ ÜMö~Þ›èç”ý xµ³ÿ~ÁVûÂÖŒñUý 'lÝð|[+§¶†Àóf3„­pp¶8¶nÀÇØú¡´-ùy[ì>›ñ€¥ðmÃÐßÇ6Iæ#„mìfÛÈ9›rØfýeCa{óaóðû¦s_­6Ü×*úÞ {ÙÖz?û8÷»~þhv:ê̳3Qo™ý<ðÙëËüsv,øÙ³³d¼Zd†̶BŽf?ù—ü™Èï¯ÈVd¾3; z(‹ùÕ³.be­Âø³‘Õ ¸ÁÌ 8÷™ÇdüRd. Š—‘9vg&÷K·î„žµîƒ~µ~ |’•ën­g±œ>±\‡¾°v_€5~©u¬'Ö— ¿­e˜kà(,¿ßbùáÓ€BÈñEп–"ø¡–8²o‘g7?b¾.ÃoÂü%pîæýÀq˜¯Ïhé"ël„% ö†%uÍ–1°{,³€û±|ûÑò-øà-S1–×˱¬ƒ½nùŽßý„å2ðçÖš˜w˯–w°-ñÀ[üûþÀ¡Z"¡'ÌG Í3a/˜™ÿÞì‘ñEaÿ°0b=» uÁ¦a³¦ì;ã5üݸxãw˜o#úìY/°' ÿƒØë \¸áðª†wPlÐa¿éYŽé/À/ÓWÂî×?Ëx®hØ‘§qþ3Þþ5#þRÛe]«Ð-‡üÓyêµlе ˜£ÓŽ í.ØÚç°ÿµÃÑ—ù3¡Es¡9¼f p"šxÈM_XšŽÀƒª/'¢> ;Z½”ãÓ%¾ù¡]ðùêÞ-³d:`Àþµ„Ÿeù™ñ˜' ß-§d½«°°œ³v@¿;«úÖjC|Ú ¿Õ:zǺøxëû°Ç­KQÏ`åú ëeè?ë?ð“¬÷à÷g6„>Îl…¸Ofè¹Ì^2þ/2c$Hd&CŸd¦b¿d–o9C浟ˆ'eÎB?†Ì7QGŸ¹¸èÌí˜×Ìÿ€³Ëê3+üY±ô±fާ˜Ð7G˜J€û1þ ?Çø6Æcä~x†Ëð³ !g Óà?Q×ÁöOU\?Ù‹¾+Ɉû'›¤~ dð(Q2ì,JîÊýîû"¯’¬@ýdr¶´c)¹}Ú“G ®6yú¤¼Š> ©ÀË–:]îRpÝœ‚ëS¯—‘º 8ŠÔ5÷JŠ®èƒ àúeÅ?È›+ÑGƒÒZ .3íMà&Ò;H=G飤<¡ôïe_ R!.Iª÷¤HÔy —øõÀ'úŒ×Óàaè ÔGÒ5ÿ$µEÚ'UëÓ-ÇøÉ†»VÑ=zºûÏjwÆÉ¯Eq/X÷KÑc#ÎSˆëõÞSg¾$Bvîß¹vÖ!ÚñžÐBÈ‘Ð!gC7!úôJèu艟=Ãá¯öÌ”u^¢çdØ?=7gÚ“åO×=„e‚/"l8ì¶°à] c}Æõa;`Ï…}ƒøCØ.Ù'C„ý‰¸Ux#œ£°;òiñuàSæ¡DØ~è©0씞wOì9vF(xrE(êÌDÈ¿s!¯ËüŒ‰ÉëÿÎâ¢ÇÎy·çWˆ#êmýûd’è¡ë.t½²DèßF¼7éUÑ£ç`nTƒÀ|ÃNî>þM÷®÷>õñÓ¢ÛÔ«.9ž{¸N¸]>D½UÇú†d}³Xtø&g„å5íÄjð‡²ç=ë¥õÁY_À{ŸxúR?¸V½õ=zô…¾ä‰¼çB_ã‘þZ®EÆŸ2ß%2ö@~f,‡|Θ\xÆØà¯¹¨oÈÈ‚\ËHÃ~Ȉ>.£'âíÁ#‘Ñõeu%ÎýaÜ×Bwý’“зºƒÀùé*ø'²¾XèžBŸú^ ]Cè[íaÔÁißE¼C;q'­øamÄ»58NÌý5mP³U¿9¬.D¾A?SÝq;úñxÚûÑÛ°¨qbjx«êà¡Uß5  ÕHä;TõL_Žó’î€ý“öâœwiͯP~ ¿A©åþ8Œ[UðþM=&óØ"uâÚ©u°RÖa]S£ž"%ó›|u„ÉÏJ¼ŠHn.ëðER¥Œ“W­ _‹$®›JJ>3‘ã$‰ï‚×&ÑŒ|IÂYG+x¿&¨`߯Ÿ¾ŸƒçÆÇƒï'n'ìÒ8iÎE‰8ÎëÄîÀ<ÇŽGü)6 u*1—¹¿Òrôù‰É@Ü#úêõ¢×À>ŒvÁ>‰¯´è7¼JýºBOGUF ’."Šíü¨f°óû®‡]Ý÷uàÀûVÀ?íñhœ+™í†ä4ø=Ék‘?K‰Å:¥ÜľK݆ú-ø…âì:刋¤MB<)m7úú¦Ê>9"ý>ÞO•ƒx‚JÀ.¡HÈ_z~}ý®î W]9Z×ýñ+q9Í«ÈKh6W®¹ŠüŒ6¸p-úq ­ûSû®äåºÚˆ éÂaG謈ké|x?ÝDÄËt³aè¾å¸ÀIøÁŸ÷¾èW$"¯ÃˆüQ‹ÈeäÄ û6—q±‡¿ÃõãëQys-ì’j—eïÖ¬µÞkLh‹úÀ·ˆÆá×ý:Ó šEá·Èƒýßj¸Ì¯WÝŸ¯E«Ù#kÕœ_)ÚœC<¨í5øíûB¾uP̺°pèEÑ‘æ/KÝ÷ªèÔöö€·ÍMDçç$ðGtY²}n–¯Rt‹¾ë3Í2E·ý=cš_Ù*ºÇÚ7¿¹êãªçu~óÖË;—ˆNÝ®F†”$:áK+>_úÐörõæ¢ÃäWÛo¹1)s“hwXâ DÛU›zœ=ÑF´­øh›™ Z¶)ImßÚøßuÑvWèµð{‹öMe½NÕóøZt¨Û$²ÅòDÇŒOßÞ(:GœP­?ü°~‘¯Eçëôãxp-ºìGܽ«L/gˆn2üĸ:ÿŽú¥.C&59p¤PtYºø×o’>Oî:z@/ºLŠødßê¢ËÓ¿])¶ˆ.lGt©³²¢NªèZwÀŸ‘g|¢kÿ%Ïí=¿Vtݯzêú±“¢[B³ÌœÎD·)ÁiµŠn;$ÎAto³¨EÁÂ#¢{iâÌzún¢û,ÈÏî;‡üüÃ_ME·”Uï2öiHauÃò†"¤øø¡“Ö:"ÄýYçÞ¿<-BNßþ2¿áâë_ü%z|9²ê£éí—ˆƒý0%î Ñ}û©ZƒG[žX_žÑm+äv÷nðO»Çwz¡ÍŽÖ¢»¢ÙýÔñ¢»qbäøW‹î^ßó’NtŸ(ql¢ûëèÝýÀy°zŠè¾z«û_¨çèQCâMD°wz$ˆžÝéÑÞô÷ WÀ^4­øìW¡¢Ç$[IÓgŠ‹pÎzìüúê¦ëú-üu‚èq£Éì¿R¢DHͱ·æþ'BÚŽ»AåïŠ.©'J„D¼¸7òf‰QA~†X_=õYlˆñhŽlY’#BFþYãÚÍ—EˆLsGˆ¥‰¾6Çšˆ-þu&å4òËâé‡Êž!ßmÀvXâȡ钧A„Hœ·õ#~ú<Ÿã3¡œÝŒ|H(×¹…^F<§g}i!ˆÏ÷ÔÁî™ùÜóiä­z¾xPOÎçöd±ç_aMWëý¦A)ŒyÃÊá׆F½Q׿„Í‚¿¶DâÎã5Eó¢…ABØ ØMa‡‘ß c»%ì0êXö"Þ¶šŸ÷ã+f£®*ìeØßaÀá‹0üÔ°RäAÂìˆs†™7 S¢.,, ~lXâ{a©ð܈·†ù¡GÀk ØÑˆ„·C=z¸yÂðàyíEÈ÷Z,ùsDoö{O@¢÷FYÇ ú„"ŸÔg puŠ(Dü;b¦ô¯ªÎ _‹öG"nÉü·‘Ÿ#Ÿ‰>I¢/ãúº¡7û2¾#ª5Îw”q–(îóõ9Ö)Šë‰úuAüª_>ÆÙoâýxœýþ¾#º ®£ûÁ.Š¶Â‰f¿&úèýhÔsˆè Ычà×Dß¿Óx–˜ìߘ,ì·ìó˜)°kbÞ‡=³öaÌŸà_ˆ­½4Xh"b{"®þ ïdl1ü£Ø °3bßA~*v5ò걊k„8eœñâ87ò§qSáoÄqýaÜ6ØÍqì·jˆûÇ·E<7>þN¼ñÊxà›Eü‡ˆÄ3¿ëû/¡âÿ èK%4È[' Bü>á°g>ïCÂ7ˆG&ìGÜ%áöWÂ-Ø_‰œ×IìŠ|rbêÁHåÂK"±€û_ŽÀ8'#®—8vqâìÇDôy‰[ à$r½RâÄ'¯A¾&þ |Sâ}Ä7’êâÜ&µ@¾#‰ù‚“:!Î’Ô qʤžˆ÷%…!“¹‰uJê ~¤È›$Ä7ŸòüTÅIbÿAAœ^žŠ{SÆ‹)î ðÅÁo¦¸À³Çý„:š¸ƒ§Aq€óÛ mÜZðúÄ­OJÜðSÅ}‰:‘8ÆÇWÞ4õž¯¯x¼ õGñfðÎÅ÷‘õ®_øÏ¸»²Ž™â®_wøÿ8à¨(î=)O(Žû*ÄyÿŒÓ‚*®x³â˺ÔÀø£=œglŽÜgÃñ•˜Ï¥<¢è9Š\¨l#µÿÅtD½EôyÔçE¾ô豈Eð·ý°îÔ/xÖ¨¡¨ÿè»q™Èo;ø)ž¶PŸ]à©ëÝGÚñqü¸¦Þ­e\z¡žÂïK{˜Â+À?öÞ3ŒñÍ=Û¡n'ôYi×Vݯg3Œ«g-ð…Þù! ½Þ£ÐŸ³ÔÛM¡À¯Qèçàí ‹ú…Ðáà; õ‚×-Ô.ýòªûó5õ„¿La&ð'†;Ðï¸ß¿O<âZ•èwyëØwÖ? r‘¢^oQÔ~Ô3õ›†¸X¿Û¨K‹öŸ}|N1éÀ=ÇÌyŠÁ¹§Øhðèž(ãõû6æ'ö4øùù\<9ïËÐw¸Oo™ç¥>Óð÷>°¨Ï=ôOˆzŸˆ9à]‹¬‡:·Hþ{ä?¨¯êë[ßí’¿¢¢€;‚}BQkÑ5j-pÑQkÀŸÕëúOìÇýÖú>ºø×¢­ÀG@PôûÀÇHuÒœb ÀǼ‹çÅ\~:6Dæ)ú„bKQûx®b·Ê8ÕüÄ Î@1Û¤G1u¥?OÑÜo²_gœÛ¨$웨¨{é›/ãÔ"íê³4h†¨w¤ÄPøNÜ'|“´C(l/ÎuÏ_Ñ:ñl E]+…Â^ c¦<½ u;Ÿ+ý“AÝV´Þ=x‘’ºm\ùê¿KÎQ·=[ß»ë9O]{ãœvZ×ê‡m÷SûA²~œÚt–~µxçJô¤ï©IrÐ-)§&ÕG¥ÏIÛN ƒÿ¼bÕ?w(sR“ƒUóP#,`Å·¾NÕ§h\(ªvvÓ¢u]-T}•æÒú‚T½WÜŽj½¯Rµw¾›¨¥¢j_K¿CuÏýÇýüTÕ½ˆM»?V¢ºûo§¹MÇ/¤ê N²:÷åªÕDú!Ts2pÞ5núÎi«O5üá2¾Û@ՙ׳:øeU÷÷'`)Sµ6'­µßh¦ºÛL,¯yôÕí2z)®¾úÛ}$NHuÅ3®$t¡ê²nësÃF,U]úæ½’{k¨þŒÜ´äâñ8Õ……!ƒ[ïÿQu+[Ö›¨n¾þ»þjn{ªöò+Þ´˜uTKßõ½êkŸ¢zgϽÑ3¦ÕïT?4¦úvÔÖ»ú|ƒ …Û¨ÑÔ±53Í?z­µ i—¿\všZ%=›×ûËÿQ[×ßcž)uPýcûù uzö×qYãÖPç1QWV8}OÆù‘O¡äc¨ßLé >¼”IÒŽ§”ÛX×T™–lEŠÖ苬8þNåSˆ—+ß@æ4øÚÒ¾“v¥·@_Œ4Ä)ÏEZ;Ô+¦i¡7Óc§ƒšTÕÁÛ¦zr]uýM( }¥h,óçmŸ¥ºê8Ôy2Ÿô§ פß3ÑOÐW4ü–4õƒÔrFõ>êUè?NéÀíRúk¨ÓHO‘v<¥!_GiË%NÒŸ¢´Pð⦵EÿåMð¤*w£ž\9õšÊ>¨7U@йÌï6ü™ŠlÈ Eø¬‡RaÇRê§­ƒ ˜µ!·SÏ1Ü)ÈiEÔç¦7A©—aW(ª£“¢êMP¤è ž8E'Ôg(¤¸L!ÅE/%eê'”= ”PG¢„]_5ßΩÒ ç i3 ùgoÔ9k‚/Ïñ+êP÷äXˆz3ÇÿP¿ëÈÏ #RæÉÑüöQeß'qhdÿ õ.öÅèûbŸˆý`ïºF»u*ö.è'c»…z,Û/ÏC6ô½&Û'࿳͖~.ÙF¢îÚæù¨ªñÙËÁÓhu¦öi°Óìo£žØ>öЏ<²¯€¶%í|²}oÿ öž} Æo_;ȾVâ ȾûÖ¾Kò‘ýGÈwûQØ=ö³—Fö«àuµÿ{Ð~|ÅöïqNí;%Ù÷£^Ö~zÑ~\ú©d?%ý²C}µý'ðÚŸÔþ+úíØ¹±ý$êžíg%ŽìQe¿}d¿¾?GµkÁˆ19êá<8Z!¿çèþQGOØ¡î_ìˆG?&xúÈ¡F½¬Ã„:MG&ìj‡ v¡qCr¡ÜáÆ>u€Œý±C`:†Co:žEÓñô«ã%éã5èwÇÛ˜gÇû8ŸŽÏQ׿ØÀãØ†ú1Ç™!Çð;~ƒÜtü…õwÜC}™³.ä ³9ν³#êÚœaØÎ8èeg:x9&Ø_ÎRY/AÎ!X_ç8ðb:§À¾u΄Üt¾9æ? 9W!ŸêܻϹõÙÎÀ;ë<…ú|çØÑΫàÉvÞ¼¹ª£nÎU|¨®Æ¨§uµDݽ«-xš]áØW®hð‡ºbÁ{éR¡½‹y4\6ìoWø\åGB®§¤E®qà}t½¹àzv˜k6êŽ]‹¥ÿO®Ï$®Š\k°O]< ®½àMp¥ë ø\B>¹þ•þ8¹îCžäÔA=dNÈãî_ÓYâÝ('¼’9að»rB¥ßN9½ ‡r€× œØm9„zÉœ4Ôõå ¯åQÿ™ã¼”“úÇ7ê•s|79ƒP™3öJÎ(Ôý匆ÜÌ#qüOä³sFâ¼åŒÏj΋àΙ&q”óúúå,€ÞÈyuý9Èú Êùˆ¿Çuè9ËáW欄™ó%xÄsVñs?G^Î'²n%ð{èóœ…ðrÞÄ~Ë™!ó<”ó’ä#¤œIÀ䌓ñÊ:ÍœþÐG9åXÇ7öQ÷õÎÉ…ÝŸã_lNÿÞùž£‡ÏQ£®?8}ÊIåõì#€”Óç!§+Ïw[øy9-àwä4‚_“SçÈuþ§ë:ô«ë"ê8]'aºJ\¹vIæ%ƒã •°73~„¾Ï8 ÿ8ãìëŒSð3XŽdü 9­¯¾ =÷=зö?陇]ÏûLþe}η^ƒûémˆ»è=ˆƒèŸ‚Ÿ¦ ¹§żëÏ ¾¢g^ý‹|¶¯ô³p.õ3à7êgCOé€OAÿÿ3è)ý§¨«Õ½­g¾|ýzØ›úuGIú°ÏõÌ/®ß†}¢ßŠúuýž/ß:ÿÛ¤? žvýð‹èÁËGú«ØoúaÏêïA¯éï¾4Ô‡^7´Ä¹3p¿C{ðÏÂÐÏÀÐ~µ!zÊ`G]²!ógðÀ~2xd]ƒ7Ø0<͆I°w ¯AïæCžcý ŸJܾ„Ü0°~7lü2|¾|†ƒ8o†à“6œÆy4\…]f¸u62ß”±÷ èx1r˨A}·Ñ ùcÌDܘ ü”1ý7~™Ç ã(È'ã$Èaã îK0v«ñÄK°þÆ•Ò/%ãF™G'ãnÔO÷ÁN5‚·ŒŒ!ŒÇa?ÏÉz>2þ {ÝxçÍļ*¦–°ÿLí·1u‚6õ‚Ü0% þܤ@ß“ }ÊMVÈ5S¾Œï“ }œÈ„z<2 …½a½oz¼å¦WeÏfêûÍ.ÄëÌìw𠱿BɃøÄ>0‚oÀ<ã3?Ç÷{qG3úÅ‘y!ÖÃÌ}aͧ4o>ϼû{üûÅ|ö†ù&ì8K̳u%d ‡¾²ô‚…ûX4x?K.x­,°Ë-£àÿZ˜§È2 |î–7À·bùq9ËFìsËð YØ?¶üÉ÷ù<ô–ÿ O¬u!¬¨#kâ®Ö$Ø7Vð’u®d5É~1dEYÁFÖRÄ…­åàY²?‰u$âšÖÀ/Z‡WÝ: ûÓú<ÎÚaõ$ëlØËÖňkYÃÏ·.‘ø|²~#ñ\dåx’õ{ðYYB¿u?ä‹õ4â‰Ö+ˆïZo·(ù²Þ‡=–Ù¼™]ÓÚê eFÀnÎŒE\7S»63vS¦ üF™ù’¿—2ŸBß̧¹Îðe2¿@ætð e.ŸTæOæVØá™{_È< Þ¥Ì˰/2ï€ß(«Ì7VÙ!ºÁÀkêôÐú(ôеž®ó Ý`^´SwFÚ8Ú•×GÚ™Xwí(ØÃÚBØ“Z䯶/ö‘–ûËkø~šŸeÞ‡4Á¢ùPò’æeði†ÂÖ8±ï4øWšÉëNšîƒšvðK4õ˜«øøÕÿ/H}ë«èk¡^?L½<êÙàOS`ž©î;ÐQâ²Îê¯I×þ´.†çIC éRpžuÀ‰“ÎûAÇv‹î;î²xWÝ>ø)º#àÉè ^¢ æ»Êøš¯ÏA/êÑ/’ôÜI?ñuýô·Ð…¾Æzæ7Ñ_ƒ?¬ÿý íqn ñþ†$Äù qІ©èïd8Îzòì"coèE£ã2C¼Ä8GÖ{ôâoÆ»°ßMua‡™:B~›Â°Ž&…ì“F¦QÐs¦¹À›~yÃ'çÙŽ÷ÕžîuÄItŸc>tp?ÏîžÄmPF ÈãŒPìGGéÈeÐ(c(ðÁ/ÁNËxò;ã}¶/WƒO/ãü§Œ°sõÕ _õõØnëÍv¢úBŸý¯/ÿ–~8ö¡žqÒú 褿 ½`è ½gpl˜ ûÑ‘á3؆_á§8dì‰~ÅF#ä†ud”a¿jdœ^ã&œ3ã䓌Wáw›H™Â·1¥¡ßŠÉóh9fbÿÏôä˜éKþÝãëõ@Ž,EÞNwú2£>äkF”;¨(#ë1|G¿ Ï©o QŸ‚x™¾¼6úÙØ'ú­8§úk°³ Éàw4 ‘yÝ€ý½høxqc[ø-Fìnã«°Œ;74^‚|7q_cÓôA3m’}ÉÜ ~€Ù Þ7óó°ƒÍ{ñž–HØq–rÈ]Ë|)߫惯É2òÊbméÂ}·þ†ýoþ~³™çà þ`2« Ǫs$ÓaäwL«à_›^ÏœéiðÜ™ °ÿLÙ°kL„¾'&ô£)]ÖÖýVL*øƒ& ì_“òÜ4ö¸éiÄÉLã‡3M…=nâ~"¦%„Lk¡ÇM?‚ŸÐtþ¾¹xÌÝð;s ì;³qSsY÷°pžÌsÑÿ˼|HæJŒ×¼v‰YÞFAæK˜K#ø“–^ÐÇ-䀥ûвþåÈk]ðrZÃáçZU°[¬¨{"ëh‰o!ëgØ/Ö èÛ–Ùu™ýÍüà\äƒkÊÜ{)󉿡¬÷Y øYqXǬpä_²¢‘—Ëjˆ÷Îü ùüÌe²®2ù}2“Ñï)³!ücë>Ĭ3¡×­ÙXGkcèMË9ı,Kwµ<%qÝd‰†_i¾8‘y ö§yÖÏ\€¼‹¹;ìÓUijÿO¹Ì}Ê´ÿH|$iQ¯IÚIÈ¿j3$n•4œŸÐlÅûkæÃ®Ñàþðtú:ø÷ÔßÀ¿RODœE=þ­Ú9¡6"þ«N–uÈ¤Ž”x#R·Å~T×o!BÜ’¾–qЇýqpM´ó@³`ßpŠDãàê~‰ Á/LjØi”€¾Ô…ëLîÉzTRí‚ÜR͇®O3©Š¿VÅ Þ«j‡óž~žû<­¿`ú‹°gÒQEé]—J» ½œÆ8ˆ´uCiB§ ‡?•1­üv%ðT¤D5)9ï§t Nªì þXŰ›àC!ÅëœGšÿMÁùiEÈýÔKð S¿Aœ+u—”:ç:µqÒÔðS¦6ƒ\H_¥|…¾D) Jy q¿”\øÇ)Q˜ðìSò1ÈŸä ðÏ’—"œ dJCž&éŽÄmQÒÓIŸÁŸNzvL¦RRkà ÁßN‰«å:>‰ÿIî%Žã6qNìÓ¸Ä9âÐ/œb¿€Þ‰UC¾Å n€b²WŠé~|Ñס£·§ͼ“Ñà¢è°GûmD\¾ú•St=èÓèšx~¿óÈô[»ŠùÚªpÚ3¿|jÇÚÕoUžì0æì…OkU^éE¥§Æ{•×§J|oåSÎþÝdÒÊ›Sƒå6+ïM¾¾ã÷£›*ïML|‹šr^Ò9T¢î³Ÿ6šºd’¨;¡pÛˆ¯ž ƒÙæ\­h–²ÉzØÔK4]08ÉSY"šlξق¤]óÏ÷Íl…Ã_;‘#ZÀ£ÒüæÖéµ>õˆV‘†Ž†É!¢Ísß›7©!ÚŒ“òG´k ñvͤý+Úoïß%ð Ñ~ êú;Öð‡f¾è´é©&‘-Vˆ.M†,8óý3¢Û¸¿[ÿÔˆn«Q7ÛÃúŠU7Aôh^®îGK²f~ÞH„2?· ø¸ÐÞè«Ñs>úôä:¡0Ô±‰°«à±¿,ëÄDïñÒî½_ypÑg?pl}vGÁøÇ>—€[ï³}A"ÖG#’ùúC|ßþ௎z¼éýÞ’üŒ"º¼2ÑŨ¯ŽFKD¯@Ÿ€˜Æà%ˆáúã˜)¨ˆyïsøÌØè \ˆÝ"óƒ"vê´ãLàcˆ³¢ÏEÜ¿ÌßÍ“B·¸dÝPð‰ëú¢o‹öóà0o¼v*pæÚ ðŸk¹Ÿ®æ[ð3hй@kÔéw¡ÞF=ßS÷¿ ]D=?-Ÿ•‚ÿ€z¡NGÅý¡Òÿ/z4ž›ö ð¤Ê߱ߔðß(6ƒG@¡n7õ#Ü/å¶1YËõ=SP¿š¸}þ•õýUóÈ×"A‹÷ŒßŒúÕxúÆÄí;Î \iìZԱŪ±Þ1•࿊qc¿EïGß’húÊô{Vö{®z_‹¨aRO‹¾+ñœÈ˨c@6>[ôÙ¼xŸº¨'êmDýIøNÔõ…ÍÆü÷ä~!‡Ÿ;sjˆQô8þ‘î.ô_éz½ýéj­nŠÎ)e36¶ÈÞC}B»8_­k¢v‹šï–š“cEãíé}Ë’õ¯o^úYñÃ: ¾õª mÿM“e¢æ$o©¼[{bÛL®¼úr÷ïø*÷Ç?7ãû£JÕ±%­¯—6ÌR]ùyÉ%}.©þ ¢•R´TmÎ…o=¹/PÍÃTŽ;aT'í :Á5šö8{¢õî!ÔL;ñ|Ê…ãÔÒ)qûÔ¦¶¬ç vµ>x1ü©‡})Sžýr |˜©Ñˆ¦~ ¿E‘ˆüâkèq%ó+íÀÃ(G_¦°o”|åØ7iˆ¥íƒÿ›ž„8Qús²Ž”Ò7˼{Õûð5¥ÿÿP•Ü‹êEôñUý&믈ÚÃ.¡ÞÐ+”ÊöŽñHÊC¼Ž†!I¨;#Z…<ý ûA]~†ÚŒ|™ý›âu¸Ÿ³úˆ÷©‡"^¢.⾟6àˆÔZøcêdà0Õ}‘{<ÓñŸàr¶ Ž³_Mý}Ê1êxy‚ÎÏÊ‚ê2n|ÔöaÔuB˻댥îÍ¡·{T kí}½5õØ,å|ÕýøšBþuÔÓ ÿ.,ñ„žßþîyñ­žkÿë |@ØÀa„ý€xBxˆÌTÝŸ¯)ìòá³e}#õBõŒ|T¯EÀ×ö¶ÀnìS óXÑDŠж_ÖY‚ïW$^$cß>¨7|¢~®ïLðíõŸ¯ˆbÞ†¨RðåE½‡ºÚ¨C¨ÇŒº^º~°E?ÔM<”·\GÑoø‹ú½ƒºÑ~{À/ÓïOðD׿`tGðµD÷__´u“Ñ,¯£ßD]@ô7ૈ> ^”˜zÐg1Ì{ã_g ×åĬG=o̯àm‚zôØî¨Õ7-6 }âbŸÏM,÷‘Š]‰zÈØm¨ßŒ=„þ±7ñïqIÀíÇ•Éç=Y_ˆ<ŒèËõ}wƒ÷-ª.ì‘(êÔ¢ɸ¥ˆzõ™QÑ$ê*êoú)d>@ô{®ßÔõûõÑQ7þ¦èBÔëF¿:‡è °[¢¹ß]L3Ô­Ä´OXŒõ 1ÅXŸ˜w`ßÄ,Æ8b˜—.¶Þ €± ‰ß±±bçãy±»`ïÄUCj\ _)â¾BÝÆãóÅ×"î#ð‡ÆÁ¿q§ÀÏüñÝÀ㯅¾Ž ;2þ9îÓÂ|Šñ߇#þ x%ãïc%tiB&úó$pOÂ<ÔÛ$ì¯UbuÔ…%*À«”h@"ò."¸?‘ÔöIÒ@ðû%}:õÇÏÙÞ óhœ/ó‰CæÿÀ3fé‰þqæÓò<ÿ}9ê‚-?Á®³Ü‘µö‹5öuêq­¯ ®ß*`OZOÀž±Þ’q‚ª÷ák‘Ù|™™¸æÿÀ–¹û6ó8êˆ2ïK\‘È2Jûø!¿®EÖÔÉg‡]5 û4 qs‘µý÷²¶£N:ë"x:³› ~8y…ªûòµÈ€ºÅìçq¾³?dž¥g±³õàïÌCwv=ÔKeæ÷ÝühÝoÕû‚\dÀnÎòbŸfMDý̿õȚ û.Ëÿ&Ë„¾SY Á;“U¼H™·Ñ—7óê±³˜¿%««Œÿ=ì~ ¤ù]âSÚ«‘OVGB}‡û¶Dï§™|¤æ}àÏ4[§ÕœD}ƒæ6âPÚÎÜ÷[œ†v8ì íGÜ'ƒóiÚŸѵ’ñó‡q0\“®!üpç;´?Ðþ‚¼¬v ô½v,ꆴVè?mÄWµ}ùû ÀQhÙîÑ¢nŒ´¨“|b45'TŸD¾FÍùuôººìâü½‹ü¡$Q5à%UeÀE§¯B^1½šüï‰çh¸¯¹¦ï«ùy!ÍfÄÅ4€×Ü—¼7¤mŠø­¶ìY­ |ïÚ2àµCðžZðÖ<ÄEáúaÿ’,Î;öE|^ ½Nš{ˆûh3þW°?4\ÿ 1!Ž©a\´zâzê¥Þ{œßåásÁÇGÚ£øö:ÖAW÷Óõ|H¤³"/£ã÷Ô=¼‹n&ð«ºÙü»7Ó >S— ü¶. y•ó„eœÌ?.õ!ºlÎÚñ>:â­:½ôçI×8óÇ÷gUþáœôïÈš yFÈ:ˆóØ‹€S²D~'³>pø™éÈg> ÿ$ósàr3/Â.ÏêÉñmîs”uBÖçT=—¯)‹ñÆYß¿™%Èú Ï“ð¿€ÝŸµq¿¬÷awf}ˆølÖRäs³>A¾*ë+ô-ÏÚˆ¼WÖ÷À“dý;:õ}”Íx†ì]ÀÛÚ ïd‹‚=ù8ÉÖ ñÃìïP¿—]"õ e]A|8ë[ôËÈZ„¾^YcÇÊÊ‚ÜÉ E~=ó&ÎyæFäù2ÿ¼x¦}”2ë!>hÝü„uòúVð7VÎc[ÿfy ïo1I~a²Ô’¸…‡ëŒk2x¬y òŽæ ôý0EÖ4~Š©οü‹däù72ŽÅÈuP†/§7ŒÃ¾4h×6t‚üÒsýW¨Ð?|‡¾'ðçÿAŽe¼ùË(ƒ]®»¼îâźŒo|¸]ürSW믽?I{ñXí*øÕډ܈qÖšKÀIiæÂÕXáÇjÂÑOBSysõq¬¯z÷zÌ_}0¯–úØG–Æ2®DÔ‘…óf––—Ÿ¶p=™e“Ô»d9 |†å/à¬Õ‘ß°r¾ÝZŽz+÷‹³Î¼³¾ƒzë"ä—­K o¬7k] <ŽõM䕬¯âçÏÞÞª…ßiM€<±öD¤µã~¸NÍ‚þƒd ¿ÓÒûϼ ÷7OžÈì¾Øœ„ø¶ý.ÉôêŽL«Q—fbœ‘©îküïgLfœç{ Ÿcc_zÓ´½Iù ýäuôÏáè3 ô'zqˆ øA”aG=CF=ä%t{PŸ ›9¦ãü‡îEô•Ñ­>I7ñyÝ3Èoê¬È7éz0~¶.ðæÚ‹Àïè¸G×y,ûQ7…åó2àdt{§Ñýº˜ >ߺ­À©é†!ï©kÏ}ùNpÁ%l? à¾zQll.ESÄýÅZcÿ«÷Ç¡\ƒº›ŒûTíçÈ=ÀAEB~;q9Š<ƒ}Ð}9) <`Ýú6úM쟘fÜר@Úƒ³¸—Î÷ǬÄüÄ|¼LÌzè™RÌ6ÈïðQÌ·ÐK1»€+¯0Åœ@\#æ:ê/b룾"¶ ê”b ÀÅ— χy\S,ãAâò€KŒ[‰~¬›" xÅ*ô1z¼Þ«ãnÔs¶wÛc ¤6?@.µª÷EÝQ¨ù+_w«{\G×åÿÛ¢Ã×TÏ»ho»jK©æÇAz#UûIæ«îÇת{A¶·ŒPÕ¿ßÛÒŒUý3EÚá¾÷àZuõßÓûö4k¢:g({«Ûk­U—K}ZùÛ cój^;Xù÷uS‹·§þWy«ç‘Ãï6xJT‹˜ßvÈÏ{EmØI¢ÁDø‡Ö°Äý|S4 ²¯tùW´xï½êkýgDëdåµ½oý$ÚþõFv‰5EÅ–o2g×výy¦bžè2|]G‚¯…ùÿžÜo¿À~‹3Á~Š[uŠ^-~&pãñwP/“Ðò/á êËû§œøê5÷Ib*J …=›4 vf2ã]“áçQJðõ){`·¤b):ç¤8‰}¦-)¹^=M ¼QÚjàÒ«£:Ý »<ýuàÎÒ× AϨ>APê†zJ$ý8¢õÀת0žN‰qª‡0~nú€© ?ª3Ñwè‡Fãù>ð‰ÐwžT;gVyWÅa>UÀå¨fƒAõôÍÿkïJà›(Þv ½¡´@éÝ&mÓ´iÒ¤é]vR(m“6-”Š@‘û’KnäPûAD9”C¹ÌÈ¡€(7‚  ‡ÊñÛr _2Ï4Y[QP@ðÃßïµ¼ÙÙÝÙÙÝÙ™wž÷y„шS à&Â@†«¼®²zØÚálžsÿ¹^³¶;ø#´÷Ò>þÕøÍ<.þ ç¹Sóx”šó·Çàq¦þì;Bã:rþµð,ªÖ#¡jgewèVÄÊÌÍ;ð U /„Æx‚g&Ú|ªòˆ·ÈG·Dî…÷!j.ê%ï :ÉTvº5²ëП‘·Äþwâñ\×1~p­ñ< ¾ò0âù¸1>ã„x)æÍñUÔ| œ¶f>ð š È“Ð´Å¸OÃñjàÔ—0ÏUÃñÔŸá» ~í£ž‰ñ€zÑòõ -Ï/ÓÆøE›|¶–ëüi•˜Ÿjƒ0¿ŠÇº'‰_€ñq|wàxâ5XwÒœnZ³ñ~MG<9òžÔ71þ¼Ÿ‚Z‚ú¨›#þ ß^¥òI+1NM*eøMÛïðIÒ{À»%"^OÕ˜ï&¼Žïr‚Æ…¿÷Ü'Z òêâ߯ú]¼ý—†ã4C€?Qó<3µyqo#MÅóâU“˜ÞƒõøÜ'ªÎì='*Ö¥ø:Qþ©r6ÖÙ”àU&±·€{‹Ï+‰-fß/ë†<>ÅzôïŠÎ,?ßz>î“Î#ŸƒyOÔ;ÐM—¾;ü\$‚?ŸÒ™ˆs„ýˆu™Ðá²·¯K&!™Œ§Ê‘—UwÓÓ$uf¸•|3¦&©uë§>gÚž˜÷SgâíÁxëˆ×3ÐÇó˜=nÛgÂ^â|nÌäWýÚ§–}2.lh.Ü>Âø[…ËW®nl¶ñKlÁså0)¦æ²ˉCc¦Ã.Üy½àÒ±9FázÙŠöU“Ým?áfžEÐ-Ô:þà¾pÙ‰Í3„íN…¾s¶Ì6Žqo0Dµ¿±S}÷òoýNÇ*…N¿#¾ŸöÖ²CÄaÈ×φ· öЉ#v Š7l^g®“qm)üxhÿÚ/Œ1­þ´ÝK/}k½ïå¾éçao›Rö¦2Ì—M7‹‡~ã£ó¦;Ï °R{>NqøqÕ*Û—ß”/¢U·î<8ø‹2êlZ;käüÓ´úè·š]Y?—Ö46Ì™Òm'­ùÕ =‹ú‡RïÜðj!£>¢Þ²ˆ÷'­nG½ß 8xƒzÏ] þWÊÆËÖï'÷©Kî79‘Sê8­Õ–Õ¥µM·BÙ¸Ïz{·œÕÿÖ¤Iy»YÛ±B;[¯÷̶̑=&ËMeÇ>løÜØ0ÓÎûïØÀq¶I3šºº®Ü>.ëu?ÑúvÖ¬k}¸O«¯ÇzžW‚&;ú§´&ç5«ål:YŸú¾ÜŒ?ð£4XXu£û¥íÖãp^`êçy±zÚª3´n{çÿL§´öd¬ÖjÍù¨7ð²Öúóë6•jY¼Ùt9‡é4šN¯?Ù$pÒ«åíP±¢aŽS,„ƒ4dïêªß*µ÷ipÉío;ÖIC‚ü–PvšwÁ:PpXë˜)#¶Ð`‰¼j»/ hд!³–¿–Oƒ ÿIן³gÒÀ4ð´lzo媩¯Ñ€Ðañ/yåÝÔI¹Ô_q¥ãúœtê‡üJZ÷™Ôðž·¼¨Ÿpc~–×íàTê÷‰ò›çB²¨ßT©¾­ä$õ{wr¯Ð¥†Š|Ϧ+,D+-LçÓ¢_?B™~¼)m7af¡éÔº^)?=m:Z•é@š¾Ò,^pí …iæ™N[&µ¾E}„=¯ûÍ.Y/,sRž_òÌ"Óæw™®³éÝÒ7÷+7µ¦Ø6¼¥u½¿ÜÞn1vÚk¾s„-òO/wM|UØsûê”Q¿¾-|ßþùS ŸΜ~3xéŒ+û'+Oª÷'×?Œëp€zÛã~ÖT•­3?úÔk[í£‡UÔëÙ³M_›äAkœ°µxÃZc—¹\ö3­±¦©Ïº/;ÓÇŸŸveÊÔ+ÎÜ*½·ÙžøÔË´¿çÐôÀJÏ7ç{5ݲȕV?mºmÑe¡v k¬U-©E,¬—­ZPÇÿ±üZZõ¹²4÷àÔy—Eþ²*uKaóB깓é¡V<¿-‘¼jôรá{"ƒ¾‰hyVx Œ?¤M/‘ôtzÏùÂ6óÏ0—ëÙEßF’ÐVà¯™ÇøEIp;Œƒº(-‚ž$`â þÍì'õEêºXä¸e¤ö1|?}ú¶°IMs«-]m$žŠnk´vÓHµƒa‹®–¨ˆ=ø’…[ñ3?Öéó*Î_­óæ EŒ_–„4:˜öÝ¡$ú›$,ãÝ…]¾‰'-Æ}RO çù³/9qnÙŒcäà/$Ñ=ØxÕz>­ˆS(8¯Mìj¬)ÁÛG”ã‡SÎG{*½°nP!^QÞo›(k†–¦c#›Íèkˆ¶>Ü7íMvn¥¯ÙG_ز_DÀNáhYQ´”*Âùk½ò¿î+”¼¿`™ï¾V­;æÏ{³åßÍÊ|+È»j”ys#>®j”²F1À¿4âñ±FU'/¸|rÁ1ࢠ¶ ¥`Æ%2¾kRÀu¨ f‚G¥`ò£ †°uxRи¨‚g±ŽP ®_ ~º q­7à§óo·”ù@ù×Û/°Çý,¨‚ùC~)/½’ó¢üýÀ³çó¼ãüíˆwæoÄ:Rþ2ÄÝòÁ Mò'Ÿ“? yùý¡Óßü>ùŸžoDÜ>?ŸáöH>Aœ!_^—ü ÄOó=°>`¼†<9ã9Ä£XG'ƭȯ5B÷•çc¾oNƒ_DØØ xscsà—Ùà‡2jo0Ê‘—bŒF|À†u0c-Ä/*ä!Rÿzmvß^ÙúEš»3÷·hÝ/²;¹Jëž¾4fÝ€QÔÏ¡{¶}Y#ê×áÃQÙK6Q¿íà9ö°w ")‹D6[¸i½ù¯àM}·?Y'`û÷ò-aÔ{5ð‚^)»Œ×ÆÏ¤^ Á[[üaÔÛßã°ùKB}"µC;yÆP%øž}’ìÞÛöAõ¾Þ|ÅÀÛ—¨÷ú:¢cêí=›šzðŸ{½6¾û«¿øÒGÀ[#xØ„Åu'SÏ1»»žJ=BÕ7õîEÝú³Àuù*Ú¢lLà#àÒ<°®ŸÙý¸b×Iê±¼Å5,£¸àÖÔ=:Õh³PÇQŸS‡è#–Ìt=vô¾/>™jºr”åg™ÎÕxïå)ß0Ûç0c•6×ô©ÓÀ3[n*|9¯Ç¶•?L©ñÊÑÍáËÛ=p„Pv ïE…xMïàMè×4üø†Ÿ!žÐð{èÔ5OmxúYÎБÍòå5àÍÊ_|V{ðÒg€>nÖ4ð¦gÍïiÖ'à»ÍÚ ]á,ðjÒ¬Ø<›f•°¸%ÍæzÙuY|˜f#ÞF³OÉæz3ÙÏA·7»×gœt6ÖiöFèK³`úï4ûÄý²o#®“ãÊðÔÖvá>Í ®9G ½ºœ$®ãšƒõ[šÓ:›9ƒÁo›3q œ7€wÏY…zçlãú­ÈÛ¥9¥l–êkÿ]œ¸žëY!\÷Õw®E?×¥Ÿ|±þðÔêwAÇP¿<¾ú³Ðy0¸A·ÔÀõ  Yз5< |Œ¡?tÖ £€ƒ1`>H _e8…xP®üÜZg œPnð@¹ÅÀåŽ=w.âX¹ÀÒÜÓÀ_å—€æÉ1¾Ë˃ÞX^;ð4çu‡ÞcÞXè7ç-B¼,ï3Ü¿¼ï€Ê»=®ïQ9 C¼(ñ ò9ÏBâäI%öÅzHâ X/JìŠuÒÄŽÈsMlÃâ¥ÖãrŸ$Þ2q3pIu±î—4qª¤kXoç8áÊõú<É5Àw—tŒç»ðùur¬G%{!~Ÿ„qLò ôßÉÐã"É<®’|ù))ÁX×Hyyy)ãHÙ8N*çãKo.I¯1I]ÀÔXÇJóÀºjZGà˜Ó€ç"é<ß:ú6$ý4¾cõüÐÿ×kžžzÓ˪‚GL€^/:#ž!ìü.¬Ü'Ä yáÄùxÂM¬ß 'ÀÛ#ðö† .蟩÷âNqÑ<.LHm|WIOŽÝŽuªŠëfåå…B¦+JŽƒÔXÏ4ˆ›)ø¾ ™Ès80ëZÂ,ðuU¼Îr³ß À'Õm2§¤Ç¯ßyË7’‰“I#š ”¡¤ö¥Óƒ¦n|†Ôê0àHQ¯\âƒ|âñÂÉÏ䇉×˰<‰Ôxqãë/´¿@/ò®~Àä2})ñšúÛ÷›¦Ô%žñX§r©c–G'&'äAìjIÔ2ÝhÒoŸá˜é7Ì·¨}Ío;/¿Ií5ÐC´ŸmN¹RGžÿUåÀÖ¬-*ê4\©|£¯u.üɧàÒwÔu.pÉîg€ï«>ƒã©Wè§?o¥55˜ÿ{gãêS‡Ú{%,´»EëÌžùƒzM+Z÷Ãûgõ\LýÍ¿SçàŠzÖù{iz£ØÏ7þdºös\êÉØçL·²T$yéUÓmýñka.Pû¢þcwún¢\ËQyË«eiQù}¬t<>ϵÞo7äSwáE°ºw´„Ér¨ÛÝÚ4–÷¤n[t0¦nÃÛYT¨§«ï¦Ž‘CÌã;ËpÊŸºâ;îæƒ|W¶¬¢§.‰,:w÷~=îÅ]´ÚGÕÜ&Ü¢Õ2ŸmŸ))¥ÕÜ›õý±ÅyZ­ZþšœëÙ´ª…%Öm"­ÖzxŸ¦?ö¢ÎÚ}9§óçS—‹ò uù:£®½Ç­Ü>ŽºÉðÝv3Oú³?]KÝy<ǽùêc½AÝ'±ç…º/nOèP‡ºƒ_•º_¾>jÎÄJÏ{Õ—,Z­_—Ï.PRçðˆéÓoK»A?ÕE°\nPy{Ùâr?vh3•º]œX£Äo'u½T–æN]ã®¶mûå{ÔeîÇ]S?I7_t]œ7ÍžžP…:³´Ô¹ÓùV#w¤Î­¼nž‚Rçeg Ùû—S—<Œ—]N`íZ¼h§ãÖ©ë;3“»»øR×lÂI]Üù¹šžºÞÎÖ­9ë—©Û×ç»·†î•ûGÅéúR÷=ˆ³¸ËâÎÔÜcXþ«t=™{n<[V“zì´Td õ¬bé@¯SÏF¬ÿ¢žï| ›z“zþ†õ³êIkÆœêü?ZÝÝ®ðä!)õ<ÚØî¥:ê¹Öòxv¤žãglµŠzj2õØÁâ‡ÔcläÔz#2©gŒÏÔçëuo÷ÿn­ã>ðœy õíÍâeÔÏCõ|t'êwºï<(œw飥å:9ÖóZõu^ùxºB’Oƒ0^l2ù ¡õÙú ½øÍç7÷õ½»žÍ¨ÕÒVÍPÉn®csŒ¸©ä$Ó§¡Ì«¨ä4ðRoÆ÷@¥àû¥<îC¥\ïIª@¼WººØÒ/ÙtÜöw t#¥+¡-åùÒåÀ÷K—@ß]ú.ò#¥X裸ý™ö¹Þ ’O'¯tûòG*að‡T*yºÇ’ÞQÁ«nô ’e¯WÛ&J–"ŸB2óTI¿u–D=*©>ò×ÏÏRIÞIYæ ñv«Ù5Ð|ÝÈs¦’êXD±¸•øn³S‰ûNR‰tþékã¨ÄÿRÎîg~6oÇüH¸¼^·ÜfTâm‘ý©MÃv³¸# ÛÂðG4l¬¥{t§a},a ë’šÒùÃÙ4¬ôfCßB~p¨z8!Û  WA7Éš‡9¼Z2ž'% ± zY$ªòœ¢¯*DÞ¿|,øÒ¢}·ŠoŒæ<ð11œÇç%)ò—R@çƒ(—©ä| J#òÅTÁ¨‡êEÌTÈC q]Á»7†ã¬€—Uí^ µ|ÿêÈTÆæ±DS|"š`àO4ë€ÛÒPä%jŽb]0>ynñÏ#>ßñ›ø1XOŽ_õÚøRÄ´Õùúo#àT´‹ëÕrÜe‚øú„ÖÈÃO€þ$It`8RÛ¼ >I\…øP…|2*ß=Sù÷Б—"Ÿ8šëoF·@ÞKôpôƒÑ3 ¿>(#~"F‰ü¦˜ä;Gƒÿ‘F_þNôOÀSD¿ ¼Dô\è°EσSô,ä}Dça<]Ûè Ä'¢ã¡ ø­ù/tƒ¢s¡§YQwGuyÕqQª]<÷¹Ò}WM>˜ê8òÅUÐ÷ *à ©j>ò¸T󠳤šû§šÄðT5‘͇mÇÈóÛ–¢=T{€+Rq¼S×ÝŒKÇóÇõÞã¾Âs·:îj®ëÅñT ~NªnÝ)õ<Ïê ·LÕ§YjœðÞh|ðh¢  ¤ázWš^ÈÃ×L€^æ3è‹i®CŸ(¼4Öëá>›ÍÛ©VÎúCªíÊú7ª¥xßÀŸ_éy²¶ÇT¶¾an7<ߪ‰ˆÇ©Æãû¦üGÕ«¼}GBŸL5‚ÅsÌgR`ù-Tõ2ô‡TÈ× ª<ècªG¦J†Î®*ñ`U pXª$è]©b Û¨Šàç DLÅuŒT~ WNUuðÜ*˧ÄsnËåϽrô‘”ý o¥Ä:U¡?¦¬ÅpŸ•ûÓ,ŠFƒ×ÊgB§Z¾“ñ£Rùè®ÊwB—U¾ý€|òåï1\-•†}4xÜi4׋î÷4z5æ?Ñg GžJk}¸OcÞ?€Â÷]1•ç‡^/HlOè¥W¼îS%t‹l¿Ã§±Ÿ!.»úˆ±|U1ךçÛºpŠ.èGz´GÌ;ÐÉŒ¾Í®£Ó›̨â[´»â8ôã*Ö‡ódÐ5ôÆAW+‘ç÷&ºBÇ7‘ë´'¦a\˜Øýeâäë%ždë.4© ã+é£â=Nj ž¤~,ï†&M1é0‹ËÐääe&?÷?ùè–%—Aÿ-¥ãm¦)\8e6ž“ð1Ñ” oLSv£M9ÀÖ­×É}sy|R¾Âü2e"ãÝ¡)à©£).,¿›&_Bÿ™|å’w ßHM^Žøsò¬§$OÁ¸7¹;óÑd5tÍ’«#þž\ñï$ðÜѤÅl]&õF¹$þ&@¿:ÉßýDÄYiâzèK'‚Zb!t˵ÐwM¬ä„}àíH¿MN‹&$C‡2!zŠ ~h¯èWäKÆâ½>G>§põ$ž,®KI4trItŽI=†×¥¤ž{2 ë dŽO°®HÉÌÈ–CuOè²Øz+ÕqýY]'è"êú">¯ÆxÌ©n<ëW­õä>ÕæÇœ¨®3Ö/tÍð=Õåa@—Éðˆ¶ýáSÖMH úgrû‘ðü“©à¿!/ã{D`œBšà} ôÄý›ð3âÂÇ @…)˜× “0âí[éùLudy45ãœÔiÈ?M}“å©ÑTèaдPôÿimðJãïWÚÔ'uo hºtìÒ¡EÓÉHÓÁ[HëUÅ:N=%úëz<~RoÚ³Þûx7ôëB[¬³í0«pÄ©pUÞ°†’*N.¦9mß!ާ‡v4|fX'Ž!Ày:¼¥OR÷ïG€ÑÞ<ë:¿½5±¯›¹vKwb§ŽÜ#{û„pDz|Yçávß/Jµ]„Ûý^O¸Õ—éô · w-\[ÂâÂ5Ìó…’/ŽP2éå WG³ù¬pµ[ýñkáJø…=§Yûír_¸Êqv¿O&”4[b™0 ¥ý¾ãõᡤV7žMr…Ҥݦϛx%]üú}åö±PÚ–Å{„’Ù¡çµÓZ ¥[o8ù¤– ¥1 ÷"”a‚Kµ$çQæ–n ¶Èë„ÇÞH¸ÉÂI…ßò»žzé\Šp{øKÂᎅ ¦Î2b×®|=týG¾µæû,ð=Ú]:¢¹õ½(÷I•oªX„§ˆSPHÃ/Ú/'N’Zæ™lù}ªtߪfÿ6oÏ¡¤j[&TEªI7­wmÜ…T³„Wþ78O;ÜlûuE-k»ý¸»û­âï ?În‘>¢«ðÍöÞ[ý¾úaÞæ%åü_•êi‘ñfû™À;د~ÆþÌí!ïÈ~q•Ö\!X|ƒ8(Ì»õ}8Ä‚/ëyYß+ÃHŒ |Ò g¸\jHÁº¶º~Ô°ßÃW˜ï¾CœÂ°ºÏ†7¡siyú.ÔÐzï5ô! ÞxõÈˤú],?‚ê¡Ó@õð]Ñ?÷Rÿ,øôM0^Õóq’¾'æúÙøÎëWá}ÓJõûnƒê÷`S¿‹áÑ©þÆ+ú³àSЗ ÞkpÑP:¤äYšë÷Ûˆy…z$Ô€¼9ª/Eûèbž¥_ÌÏ×x8ÎSNõ2äçs¾„Êí?ß%ÃŒã ã3ÃW,o„¾Á<Ãp|†‹˜_®ƒï#·*tms½ðýÍå|Y¹IàgÈM¿H®ú¡¹Ð— ¹Ýï Íí†qqîxô¹0¾ËÝÅò4ÌÑÿæ…¢ßÌ C»çuV^g¶žGó>cúˆæ¿Àä•1@=5ÖÆüÞ˜Rc ø¢Œ©˜gl=ØÆKŸ â½F5ãm°mß5ªo¹¹ÚÝ8z³ÆFŒŸ_F¼À8ó)ãG˜ÿ‡b\ilÞ-c<›ÙŽŸÁI9˜Gùü¡R}ÕœWâtßó 'Eóúa¼“ç‹q\îZ|ßry¼#·%ƹáü}âßEÃZ|w s1?3L`yµÔ0ñ6þÜØò™Ò¹Þç.àÆ5FàÐ5{€‹Oáñ¡…ˆ{i¡³M´ÝÀIh ~ê„åà¯Lï IüyI.àâ:u•Î×8$Ï£RI‘§„¾1‰ ¾×]‹¿²B}NÅà»jà—‘¢h„ustmøw®“¨ÈaãV¢¨‡uaE6pOжlÁž(za\Ql¡Å#Š~ˆã)0~ ŠUˆ{)º!RÑyJŠÙà1Vp~sÅ(\Ÿ"ùÇ ®W©hŠï³õøm_S€'ˆ°ÞµÊ¢ày#ЬÇðü…˜UૈY‡|ʘ€P`žAœ^áû¡Ϧ¢ø7ÅÀ *‚Á'«ˆƒN©¢»ˆñ#³yW1žà¡Š.B^‹ü<î‡|<ôuäõ—%çí'GŒÈ×!Ÿ[Þx:ùxð8Éç³y=‘÷îQ^xyx+äÃq~ù|à2ä\7FîÈú"çùªQ˜“¨Ùh—¨ÕÀ#DíEÞ[Ô8à¢8TTà £úáùB¿G¢â+Œây9QM‘OÅÖˆ ò“=ÁÏ–ÌñÉ?#6¥9âè)òSO‚$íyð¤ýñòôà-«ÈR^ouWðÐÆ]Åqâ0¾&¢ÚÎøvˆ*yZÊñÈ_ŠmˆöPp>ö˜YÈcŽŸ(‘wÿE”yë².È#‹œÇï;Ö)+µ#ç!\(£¸Ì ´® |Ù:Ä#ˆîc¼÷º¡xŸtà›ÐE#?OŒü]]à¢trðÅèxûéx~²n ç™y|9º•ÀÏêÖbý‚óòTæé™Æõ†‚¿[×y²ºFàÐÕ¿œ.úºpð,è"ÐèòÄuí±®¢¸nÏ+=Œ<)ÎSc[çA»‘ˆwÀçÞø-)×é‘ÌÜaZ%a§-´±q$LÊæ9$øD2³­çO{«“`®¿Èõ§ý‘IêþÞ‚:û×>¦©s°—aÎÚQ¤Ö%èù8ÿè\M;”øtpn•÷§Ä‡î²,\’Z¡¾C‡T#µ¸Ýä*©µwÜ«ß'ÚÑl=˜Ô³΋Öæq»7õíÀò¢©¿\6à¢1’^œ÷ü´+SiШ«žÃJFÒàäy„@¿ƒ†ÖéÛyôâC44ëÍ¡ƒåË£ṳ́µÜBÀI¥çïÁóY#{2ž *Û>Xyâ ѳüGª¨ ^ÖX|/©ò}ðJ©Š·Q‡†ÆuºLÜuð™©[°@U¯ÞRSøTM3Î˶yµñ¯G¬}¼°ZžW®]Àò ¨v;ÆÇÚ À&ÔÄø0!qЄælÁ&¼Ä·_ü%±çƒÅºMêÄÞ[štqªdèÑä Œw‡¦(Ç”2q£”­ˆ[¦¾|Cš|Åi|}'½:{i:Æ!4ýûîZÇ›YÃ0ÉòeMÚp%K¢ ³ÀÿšyœñäÓÌÀ«fz°ù˜uîÓóÁÇÖ@ƒö¨ÿ9ÃÅÓúÍ0oËø™­3ÒŒ—°Î’Áq0ºËw냿L× 'é8Æ?H¤ß§$ÜxÍðàÃÁ÷KÂÁ‡HÂeX÷”–ð<Žè7¤ÇÀ‡"ýßkéð˜I¹.º”óYIÙó`­/÷‰ä{s¯Óh‘L’j»ôhF$1ÓM—3’°Sà¡ ûA½æ¹ÿÝ&’*¬(ÝG$|œ+ü¡ÄäÜ0xž‘ú!?Vú2Wä;Ëzcü$<‰‚|jy ôZä¯c^!߈~A¾ó,ùðvDÛ#Ï]~ónù[lÄz=Ü'òBŒ£Ž‚?#êàªe?B—BÖ z¦‘+wÉÇëÏB"œç¡Kï0¼¾µ_/çÕv…Þ9uE\’:Ï`ùÙ´Zu–`I¤ß\¾þ*uÜö™°gOgêÀÒ3Ÿ3Ýùé¦ÝÆOÛšJ7”Ez?OLg/·=àº|¥ék‡“êÉó2…ͧ¶%*:ekr_8¹ÑòxT~Ù2ä™×Víþ÷Ö Ëƒ ü þ#ášï±£‰;Þ®Ÿªw§nF¸ÍÒH£ˆƒÀò¥‰ãM6Ž U›<ër{4qæq×æ×Ž¿hLÜZºvjëyޏï–çè˜V?˜Õ¸á”¦{*HðÌC´ókå#IH;K·¡&!ËY¾2 9»Ç’xOB¥à‹mÜÞ_=¾ ëg “Pú\+ÕÙ;$¬šC²Ðï†}N•4Ø@$Q ggŸ¥=ê®#’«èç¤éx^¥í‘Ï&}ÏôÆíáöìûNÂÁ/B¨ø<•ûæï òè‹ «^È¿?Fè”…§‚+\Ã÷_””ë—…»?^ƒá'ìì=ìì|GšObÿOm8¦|ÿø8ãœ.ÊøùÞyö0ŽùÔî¹ÝþãÇø¯¿¾ô§öÔîdž۱ÿî§læß±{9/Óà~ínÇæ¿ç?µßo—6OíÑosǧöÔž$nÇþû×ëñ_¬7¯c§öäÙÓ{÷øµ?—ëÿGã×>ü©=µÇÕ,ÿ3ÿçö¤¯·û½/_ýq5^¿%ÿ–ñó/{öGÇ俨hü÷ »ßîÿwïßô¿`üZV>ÆëRô¸¯ß˜?2¾m¬Åø¿Wý™ñ2Þ¯ñýf߯ñýžÿm¼Åø¿'ˆÿöîƒ4~ÌÀ{5^~ÜS{üŒß›žÚ¿oÃíØ±ÿ5ã×¥ú+ãå”÷j¼|ÜÃ4~ޤÇÅx}ÿMãuHxҌכßS{jÿnÇþû×ëñÔžÚS{2m¸ûoâãf¼^£ž$ãunô_3~]/ýÛÆëÑñÿ“ñkn÷¨íaŸ÷ߺ®ÇÕx{txXÆßç©Ý¿ñ¶ð8¯c¿ûµ‡ymüØýÿ©ñã û§ÆãüOgð_/7富—{ý^—öQÙýÖïëÿLEã¿÷¾ãe{ý[ÆÏ¿âq3^¯åÊøùÞú§ÆóæŸ/3ÿ^Œ—w/ÆËÎý#»—zU¨Ÿ8çÁëŸ?FÍ¿k|Ïû1¾ÏýßOø;Æ÷­õ°ŸgÈß1¾ï‹Úøq=,ãÇø$¯«ÓS{¸ÆÛ9ãQ?ßÚ¿c|ßèrã¾ña?¶â¯Œ—‹ÿíaàðêÿ׌_—Ç“`¼®¯þ™ñ2olû«óðíÒGev>¾íåju¾]<ækXÑøïY÷j¼|Ú£4~ÎÔÇÍx½šyÊ‚B«cTèŠcs‘#:±X䈷4 ±¸9¶*tMDލX¡Rt€&â+m".ÖD\¬HÜ:EŠŒl‘S¿±ÈÉË9:‘S(ÞÒ4׿ÛŠed+2 m޲~®È)o),²9Å¢}Ì­S$rlu3;Fk­3 ÅÅ …b§IŽÍŸ§HÔ:EJر5|}¨9…"GY?_䈷X\…1CäØ.»~n±­Ýê狞³ckÄúùÅF±SguÄm©ÀìØY$zÄ,NžØi,rl—mvŒâØÚÍì4mQŠ]ls6ݳc«NVÂÖnfÇVk³c+–cÕÍ≜ ±“)vleޱØVLßPÔˆúæ¢ÇßìØj`vl50dŠNjvšäÛ¥­XžNì˜k`½Ûyù¢7+¯°‚S(rlÍ›'~Fó ‹ÅûÛ¶u¢{jv2Å[lÕ1;sDN¡h‹R\L)*–¡È3ŠÛõ3”……6GôôˆÝ@ikQc¶¨ÍŽè¤Ù¢wΘ£øÓÐÚÙÅ·ÄâäÙœb]ȱ½sfÇV|ƒè‰7;¶¦2;Fq1Û«™oPŠ·ˆš*ß(.fõb:ÑSeqŠDNFc‘cë-N¾ÈÑ79¶Xœ ±“#ršˆ ŸTt³ÌNSÑI‹ó V'OQ_'rlXP(®h¡"O'rlO¢Ùi&Ú"jªñ£\ ~zçˆ.ÛìØö)Ì}K ÅÏ[a}QóšÛ ¶8¶}ê+EÅrÅû䊷ä‹Þ‹ó»-EŽ­O,,=ñ…⯦Å)9¶ON¡øÃRØT|ž¦ŠÂú6Gôùh"~šäÿÞ±Ýú&ù¢¶nš+*fvlu3;¶¸Ù±µh3hˆcvÌ÷ç÷c+§vÝÚô)ZY\Û´ëÛ³·ù_·+ïѦ{‡òâŽüÇ*ÝzöèÄÿíØ­MßòŸ-e+ìïÒ»ç€ØòcX°ÂÃÍÿ»sçìt¼S¡*®íÛômÛ±7;”¥:v·ÿr›‹ù¶ôks/data/air.RData0000644000176200001440000047041413265504312013303 0ustar liggesusersBZh91AY&SYOé4 éÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿá ÿ|B¥|@ ¢Q$úÙ•˜!­a(Q$ŒQ@5BGÓ(”C#†I“PTV¤¤i¢ª”djUTV&”ÐÒ”d*) 0* F(2¨) ŒJ1R ŒZJ À)CU!Š’1TQ aR†# ©ˆÔ F&Š‚µPb2UPb*£•)Xš)C”ˆb£1 #˜dИ¨Äa41€ˆ ˆ"'Ó¢-4H¢€H k h4 @+@ÐP ¡@@>´P (¢Š€  PP(H P R€ ªR” • UT©U%IJ€ªPêÊ€@  *… P *"•J¤I ¨ PH @©TÀ  @ ˆ@€D$AD!š0€€„_ ÷°©@ ‚JEHZi¦€ “&&€L˜FÐ €&†˜šžMB &F€M4!¦šL4Ó@& ‰¤ôdh OA=LÔÀ©í2dž¨EOÉ@@ÈÐÈ€M”ô2jA<•=©èɪ4Ð ¨1ÒhPЃG¤ÈÛT0€)æ¨B SÒdɔɚ˜Ôz%<Ú¦&“&i¤= ¡é2€ê Ð44FŒ™2`†€hi  O%IH’z“fT?Ÿ¨›5@iÔ ÐÕ2h JSj”šƒM”jdl©½Q‘êz'©…4l"6F¦›ÛTÉ‘ú¡úƒ#Å=Êy2ž§è£Òf)§‘êš=M¨ÓFÔõ)Æ›S'‘©”ØÔžM¦„ž“É=#ÚOSÚ›B iŸêªŸr§ž”ýHRú§õªqýO9Íø.=6üó×§†±¹Í~{æäòo1k›Ìߟk<†!–Ye†6Æ?{pòh„E$—S“¤*™)éäxð]ù¸3[4Ÿó³fÙ¶Ãú¼6­UømZÛi‡m_ˆÊ²üN­Y©‹úÑÿ=òÐ>÷)ô¿ÄÅ?E)óRÞºR¶(kBŒÜœÒŸ3l¹¤ùÚ;'Û{‚{ñh~)ãéôîçðƒ×¿¶ï¿CïÞϳēÝ'Éî'»–zñÛÇ'/|==Ýžžù[qì“Æx>(;Ò{éNÉé…2(ïzò;o1yp^*„‡§—¹fNÞ®À||AõëUœO ܼ|®…ÕÝósDzEæÔpH/í!|à÷{½Qw·Ê;<üÙ÷~//I•æT¸^GÝ=ô(¾yNB~‡=ËßZŽ^?…øvQãÚîüW߯yå³|öœe´þ“?ßE®êõ¿ªú']ÛÕÁ~"'…ïdŸ›œ*(½¢ú·\}ëx—¨^%™½ìh¤‘ž‡ëý`*}6A^m«»žÛ“ê¹Ë׎_?¯>Ž'x÷gèó¿9õë(+ò¥{+ú+ÆFúþ\|Ÿ“ÆCä‡ãüý ×€þ⮼¾”¾|å|úñÛá‰8úßjù>•ãÝöNx߉¶rñ²¤Úm66VÆÉ°Ù-‰mFÔÙm-•m¬¹¥8àã”x„ öÛ((¼LŸY$îåGž^ïhú>_|¯—¯%EG ×ö˜,+ O`ÚT$3Cí§xaÃL~ñB®kŒÌ?pêГú«*‰+ ¤+c£C³*¥ËRÄ"ª$C¾ÜDÃ$¶÷×—Ãn{€ñžõä¸F·—ƒâlbsª|)ECÊóa¾~zóÖôûo̯ŽDyO<¢ù6[Ûw³yÞnWe¯[Ã×›Ñ๠™Uýzðñdt¿ÛŠ¿*|K„Zodϼ<×kÏ#R˜8î;‡¿ ú_${ÄÏ2°Lõ‡ÅçCåMí<|ùݳê%©1zÁwžyäÕàó!öã`óÞÕyðù¶W‘¾vrƒõ<<ò?ôCeTFW.3„Q²$.È»3´HØLÊ0p¡“–.ëLŠ®ž¦{ä®,Z-$BVd¥– &YPÍÞ'žy½7‘á¦>ü÷õáîzõç¬JžŸM*{À—š`¡8¸µ9kÂå`*BlðûÙæ—•ï>^ÿ¿£ßÌvWÈð£ÚÒ¼=÷±·šG5¾{Á57¶Þ¾qéóÂ;¹VUàQ^{ï쇉⠹޽´†õ^®ÓãááPù£J·Ï輇¼ápÝòzs¨ù=…‹‰éÂü,gÛÔªˆHLTÅ (="^>æ~ÿo¿Æ÷sÆÞñåâ¨(oúß{¿R{ñ{ÊPÃ÷Ïfò%5çÛ¼ï¾õ¤ ö/ßn¡½îÁŸö¢ ´¬§ãá‡Ëú³ÇݯȘýé]æïì{³–eÃÚWF]_råC«/†ží¼f Û*ˆ(|‚ ð(Ì{ÞŸŽöyùLÄZEÅ—!€¨Ôb@d @Ј@D<ˆQõF_±ô^zõuO #÷Ÿ+)åëg±¶‡Æ2 ªwÅw§­/Žú±„ O4r´_Ò¥ QzEŽëc¼MjVëBŸyõ=ýüÕäùçÑ,ndñØ]å—ɯŸ°GFR´3¢­Jþ¨ßÏγlÊ»*¥ŠV„ëÃL227‰³(ÙVøÙ•/î~)é~ô1¤” ޶7²hHûJÔ=gi½)Ý]&ïM8´h€ß•æpcAéž"-Ðí )žõ²w…K˜MPØ­±Î±{;P²ÂÏPmYÑmÒDðå.È…Ò-×½²_'}Ô«+ʪù“CCêT'é-wðRš/Uްës‡!~Ž8ÓbhøGGþÁÑ÷„(:¢‡¤1æÙ^ü¬çë¶'£`ü¯¡˜ˆ+gëv¨Ñ9çÅy‚hÐ:ûõÇùªS¬E“ö¡Ñgãû¿/³’«4—”‰×›²MÕË€±»ÏÁoq7·9úM$s¦Q‘˜e^ýU!*Y©mtÜ$t®ùùíÆP[¨FPÈ¥i&ae•!±š¼ËÙûAç†ÉŽŠ6Ò¢¸ °EfEåš\$`F<|¯<òÖ•‹”vÇž{§žy½o¦K„¯%Å£:ïlþ—=Z§g•²«*ªß =7"žö2õ“Û÷‚CÑå‹8Á™‘6M¶M¶Exd@ª®’¡zÌÒBÇoÌöaeÓ•®Š ‡u_ß„üÁùÀ ‚Š,:?è®ÁD+,3,Ø êè  œˆ¶ ÿßYÆþv«uFTdTUÚ’ì-@GfIò»L[~ÃvÒe„sUËÓ™+ÂÿŽwÔ×Btcl{¿ð¿‘îñ÷5ÿvYåQø÷4sÝ ;ÊÇÈHB*ÊHfJa,Û:‡u «gGe_5Át€æ¢á‘aœC2Ëjaކ–™RLa@Ãw™RÒîU)¡{wÙš>ëµ½Wׄi™çõ?äK¦Õ厙‹®&UØu˜ž2Þȹfm¤vK\ Ò˜M— áRU°’+µÌE…?w€Á/›¼¼wã\8à“æç‰{a­o<¼`Ý£‰˜UEH@É*Öãhx]TKä•~Ò“Äo½Ÿ·5…+¦öáõ·åÃl³¬Úy º°¦,JîŸJíµD§wF]“dÛdÛdµŠfIUh ËÕYY¿S½Tîzá×õ„꥿mR•C7Iþ!b©™Q~õ<2D¾Í;«‰í"&UC¢:éf95kÍQµŒC"²£[+¨ ¡P,þ˜¨†X—†…hò§øÿ©‰³¯Ð´xÒ5eÐ2åG™Û»¡2“ý0¿¹Þ¬J>œ'¹÷¼‹×†…ŒÿBø¹Å¿…˜Ô}ß“œ/£ÄÒx—šã´6ò ŸêݸT051Hª¢ÛI~·ŒLLÃ&"V¥zUÕ½'Q%ÒU»îÚ'»Åßœ&ΊJ£ƒj8ÎÈúsŽ&ÖÒ IËkuµ<…Žþ:ö2îg¯mòÏçÃØ¿“ dÕlA4ƒ4®Z[±À0X­l{•Ä0eTŘ¡‡:éÐtÝ_.Ù^Oj‡xÈ!OÝ5˜‡A4yqãV¥“>öLÝC§·DS\ø×ñäùs€k4Up¬ü œ7s6¬ìœ1õº4õ©€¬ÊSV™~ÇܺqVêW]jpË,’û2: WGjvâåéUª]‘Y݃ª?¸æjk(0b¹¶vA¦8{èö7±Êî©Ç§Òí0U—–h_‰€d]Å¢K¥K…xpn9øªVméÒÖÙXSm²#*²ÐD ƒvAP²³²( \È”™¸öró5(Á± ­€É"w™JkØ8!)D‰ˆTºd@³mNÐõbâRdÙ6Ù6ÙÙ^Ð*¢û˜§¡Ëot˜‡_ª‘±qˆ°‚C:dÐŒd1 Í—‰Ff Á»ùîb3,ù7áÝ7%]7Vø®”&g‡ºjLKßQÄùá°©œéþ¢‚i¥'«²ãx‰ûűúPê#ZZ×±™tÒÇòº_s(jmmŽ]+ú¿³ñÔRÚòcŸàÈx¤é‹:•ª²@¡k-R™‘ÑXÊy±®.*>·"òÈGQU@RV±î]&¨È.‰a”¼ð9ÏÊ * Wt »«B0,¶ñGJÖžœF {Te©…#!W1{GÐÏßǨUÑ|@vD Å{W\bæì$fG—u"ÆnYË êÂ~æôJ_±±ŒÏ®žoò+ŠB¨21}–~Ž©»Ú$î¿´Tœf9¬qì{ì°ž Úr5E%§9½*+|’ DÄ{Ü!~mûfø¶krÁO(VŸIáRE6c-_tÂKûC2nÛ[.ºJEÃeíoìßAVc! ¶b2®ü¶©Ñ,Óm#ºÒLÍ-þÑì±J«ËùÀË3µ¨f—´r²'jDPãdÙ –± !³ªì ÀHeF¥‘)~çt*§¢ß~²©HøÔ*‚޲ñ:HB¨ìõ¶³—u ê»&ɶɶÈxІdùgV%Ô+C$9te©Ë Td™ÅqÆK>wì3­åøÎÿNº^¸Ev-§­º§˜Um¿¤× °. õÒT: Ÿ¶Æ¶=jÞ0ì“ÑÛ$6ª]Ë7×lõMùUü{ú(+7csމ‚^‘98Õë „L c5“DŒLK‘°qÕù’ÒøwŠ˜©™¿A¥ âT’\n³: "²ÙT˜ì Õ!x›J&Ò{Îò™ ;.Ô-‡<^äWhûˆÏ[I)£~v¿W_joÑʽ 4h ü67lT3­*:”m„u©÷ìä *ÅEQôïW–ï7'«~u 9Î9ò£ ;î~­„Ä)–Ö+ï³\ª¹Ì{M¢Ëˆ{Ìs.ÅÝÕö1fó_àѕļ/E¹À+3ÉÿÔb1ºÇ;i›#ê,ªt[—ÚQîZÒZÆŒ0U+¨šÞó®7ÉH(ü€¼¸`³æT¤æR9QúF¸økŸc~°|ó|Û+k£×~¼tFÄ÷Ì\47Ò¢Þ.­ib'KÕ¶*²Ü JB-ó^3çìš½7¥4*äK.üB΂xž,ô«wù+>!)üðÉß§ TíÝÞÕ¹ˆv\›Eì¤Ôn¼Ê”MK#¼®YâZTe3y¹JºßÌPJ”›Wosê¤ÔiñFTŸ;X®´ŒX:«R¢l›l›lœ+)v…5S.Óå"mÒ}“ŽÝj”§ÆÛÿ=P®¶_wzS˜PŠàuœLÀÂuPýp)n¥2+é˜#[i^jè7Û®;óŒ&³½êdë‡uUµiý×ÔĽ»$C;ÂVBa2*DIéçW“"Ö]?¢=BƒØµY]/ZÛ—§gvãF¬“â¹Þdh€‡l_ÞZOSg6äŠ>“Lä±Y.¢))æNXÏjðv¦³Gß8)ҮȪ^vÅîIèÁ_s•ªNR‘óìYyãáo Hžã¿]6ltt‡âv4Œ… ô…­±zÏx¤dº‚àDâ´pƒFc¬5s2ÞŽ!C1™2™r'Xóù½AÉwKŸ³òÏ˨t T˼ÓÒ;Y›*d gÇœë½:GoisæfÖÕòµ‰.´4È¢• ¨õ9o½ÇÆ”IÅ -ÏÂL†XÏ—û ðx¤ìϺ…eQ ÷¢wùšøâ³J£Ž':Å.8í‹ÑUnªÐR2¢“˜‘ñ¼¢ð¿å_{L”ìÓÛRú.Ð –¤sô‹Úõ ÚŠŠ¦ÂeÃ’Cp—7e>¦ ¼+èè>Vä2K.ô ŸªŸÍë­ñE@Ý{ç¥vgÁµ_¿\­C&dƒ Ê„uñÓÞéI…dÉds£ÂaæÒÕÚ{Xôøß7­Âí‹Wé9º†¨®¥ë_K^Ü3J¥¢ª&ɶɶȩm³ã˜Æ`ÊF¤6˜cçÙ¾œã— Ò¡LžŠŒ»yO¦ ± ]Ý4ΖŽÉ(;3èúé•V' ÉUMÝ ƒnöbWŒùû|Ö%®&1Û£qŒ<ïšUÃõáÌ4þQI'ÕW 20¾%¯ ¥ù­P :Ž[Ø ÖRXÅM7Á¾9®kýs ¥¯S¤XAZÏJ݃#ºâfœ¥œ=Ù@”€fOôžœ×þÜ73S°/ò,«¼ÆT ôÍSG*…à4ìó<:«èÙz‡Õåï8û¤Xö ˜l™{_¹oÞ/sÚ¦ðÂ[8Z¶éfžûPO¿Ÿ'Qtz}Ué"ù.*cß']ñ7'ÍB˜ÍÚbU“d@¨”¦¾ð¤ˆ†ï2вòú…ZWmŸ‹W†3Îslü-x(½ñõt|£óº›–Õ}þ¤Lrè ²Uö—@èEM“m“myÒ™†Iõ‹á’Ò›écãtŸ¯NÓ¿1ÌïF=²ÙÞgǹ\ 6é‡]•}ŠÓÞVvr˜8äOâ’èìyó[¥ÆêüAÀÅPüzÆXbðm8¤ÌÀ2ÈQe0Û¾#펳D;ï”fçUt«æ¿'¦ýy”  ø­¡Ð'ò3‘®RÙW.«ƒu³½v©72|-+¦­Óɉ•x ÍnT‡·TWÒ«Ý~ÈI·­qÔÀ\M_iõlä£`¾Rtf&·ÃHБrP0ªSß ÈHÙ×i—ô`ÖûæÎh v€€+QÓ¿ùµ²a‰*\$‡ë±yw‡c„odŒÈæ:|ÉCñœBëëfãî¦åŸ??o¼øW¿ÚOW7—Ó/™¼®ìXŽÎtXòƒt«ÃÄÚ²ºüñ? …²ì:«Õ¬ßvfâÏg>=b‚ë”-Öô¦<œ!­qÂt%ýQãÁà€dpÚ¤±aâ‚n[[=.Š9šÞæo ¬/wzÍI? ëç28ÅáÛÏO)焟Òþgz° d.žŸrR jÃ.ô˜•èÖŠWBÕþ^¹þFºbN]ÒBTôÅ4Åól†—pãåâÄmÉ*‹ãÅ‹¦Ù™¢þ¯ÞõA¬]ÇéÄ×—ÏÐ-6|?²vò˜ã‘{²cO5§ÅC DÙ[z˜XeVR·NâVs8ƒízWJ¸¶ïé}íÒè߯æY4!UaU8âvM“m“m“ ÔýãrrØÏïÖÑþõü‡æí›òòÀEߦzÊÌØ)”¤†î…îÀÕXs) ˜¢=Ù‡[k6Ñæ´¬ JñèÞŽß­Ôa>–c+³GGQÜQ¸‚æ¢éLø+4D† 0RLA#OÜý2•ÓdUìT´Ñc¾«GªB~–"á20Ü•Km¸Æ"Ë1€ûUdó0Ÿ»R°ïÙö9 Ýî«s]-W“¯Ùy`Ú0õRõÖÞþ½|Ê{Éãc™ü ÇO;”õŸ±áÃ?S˜þ_Ûÿ0BA΋ϟª`×_Œá—6|7.£,nëq!ä å5ðå¨bÝ{‚¤$˜Cò£º¹$÷¦ùgãÃÅoÔÅ¿¯Ù“ÙÚ†¥IÞã|*å6ÿ9tÃÑù}‹öxO+¤Ë *$ñž×Šà|oC.SÍË{*ßµá_ÂtžÝÀ\<…gå†Ast\&f&K;“².Hµ{\8U«sÍXÚñ£µ2õ > t8VVÍÕžŒ^rž}sÅãÊÜ%%ãNiÃG–‡B1¾à¶1–/§eá+Bܰð;ß…Ÿ¯Rç4ÂnÊMºðmÖ' Œ‰¢ÌâÆ&qëÞ¢¯.ø¡nόΙA®˜ÄZÊõ{]yÐ ÷¯8ÂõT¾¢l—.UY&!áU9 †ÄF€àÉ…FWÑj³–«ÆC&é{øFøEÆ:åúâRéU™idM“m“m‘¨\+ö[‹Õ"Ú;ÁñÕÈëÍÛÛ-#·×¼iQñ¢z`C3ÎÙѪŸ©„ Yî7\o‰`·Ö÷wϬÊöõ@ª–= SÙ÷ìš5µ°¨Zˆ-!"“¤˜\\R ÂÔ3•ʃC[[æÞ˜Ðºv]#…"Ökâš÷ŠÐ_kvU-\Õð‹y4•ñ {¿ÎÅr¦œ$°­›K¸ÖVÕÛ¸á+·8ƹàQZ7é{ªö™Æ÷g0%Ší»Z±ÊMâ—¹'oc0¤FÖ&‡“—Ñz4£`† ÙWÐ1böçÆ%“EpèQ?&e1lùÌ!,œo…J0â¿YÍÛµy1™¤™yEÆlCåÆ|gƒ#0âè @àˆjÜBº(¨nq¥úšìíãMÛöe­¬ïOÙ0§½ášž7I‘ïë×)·&½±"Y`Øx¯g¦(1ø-ZÎA9ô:yGzô·±Áfùˆu:o[hê:ž<,Ú³p²%×Ê„‹ÓOÛ°š€Ùd¡p ž jï7£,Ûï3åéPð—XP³Ö“Wyã¿%ëöÛü••OÅüg áw¨€û©Ò¢³ðì½n'©wgdµ –ëéÖŠFS¾BÃkw ³bÜkçÐNlýMyMo]ñ· èc,Ç 6M’›Ëåg4-: o\ö\$gŒBýµ•éú5á“•/ŽO‡eŸJfÂ{.çSªDÙ6Ù6Ù7Ë>ÊŠî¡qx¨$¥£ö-)F]X õï–UÂÄøË«V¡Ú;aWpêð›½ÀqY¦N¯0ÈÙ£XÛÚL&±ˆO6’ŒIe$^ !Ĭ;dN|K·‰ÜÕàäe—ÄH˜dmެÝr¤ <˜fû£¸×Y ®ìÅÛ¾¶ ¦ õMåëÕs¤ECùŸpivksû~¶ódªíñÎým¾½Q†¶m<ÉÉÔÜhc}8ýImÇßã—j7åëi+ŽˆÿÄOS“ר CÅE¯’8ô½W×w1Äb}ÿ0W÷ÿgƳG €ã7‰Töæµ)wªeÂ1ÃbpÊ04 Aøm§PÌA8®Ïðö»ôwsÕî6«{SÖÿqâõ[…ÕêÀôÑØ}k.és´oLŒ"Ⱥ½Èù»v:~Ý<“÷e}ñôÕi5uÒgÄ ÅøÍl9·¡¬sÕª¤§õ:¼®¤ì]Ä|ÊôbxƒÆ Ü™ÜÕ”ÅĦŠ.ªÁ±åC[sg§ÛŸz._ $“™E°jÔ0pn[Yê±·TÙ| J¿ƒ»ª«MÓè—ûK¬—´ø~Žft—\ycÑucÜÅBndøn\àê=ËD×Òß ií‰2°1ÎRKÔL¾¤ õÕ!{5„ç•è©Vºí¼2š&#uø—oÔ_ä5ÚµÓ¾[8´Ï›âZWÑ ™©Nv(›…µÕ9p_¦B£ÊZŠÍÊ>5ºû7éžVÞWÖ½Vyfá]¢edn$fÙ$ñY¸C8”`É#2V¿¢“^ÙmÿPF ,ìwÛh„Ýôõ,ÝåÔ?à´M“m“m‘—š°Öâ¹Az4b"ŠaÇ…­Ä#IZ3Ÿ¢%à 4Jn¯>'y‰lÒ«Ê2ÝvtwIâû†Ym¿®X½ø×Dl;n2׊‚’¹V|¤m\MÂB ‡´Lª}lk…é °l Í 4Љ}Π¸ÿ.œ®íß^ªÍA¦²yI€?n8*3É / 5ùU¼³²ßuâXpZ¨Un®P³{âìò’/³´ÂÏi8'œ‚T(¹ù]_6""œÅ[B•ÍyfâíÊfÝÑ©@)¬jðÕîwú©ð;»ö=ONÒÚ{®­Ý`Í~-UXS“}vPœ–û˜p=ZËÏW-†½R¿ƒ-ÚýoìgÁïØáa(2ìñt»W_D¦†³žz ^Õ©HiwORÛª¥ÂÍx©Š@EÎÌaÉAéeD+@&ïÓ 5¦¥CJågB-ƒJKqcÄmêÿ}/[jÙÂy‘M¦¨Á\=M1“ÙŠÀA+6¬Ú£ÃÝnMìA·ŸÍÄËð—sì¤RsÉXšŒ0z¥ &Àœ½|+Ò-<-â3MÆzK¨mÒ7Üãù1EæJEþÄ"å§”oÉ•,Wiš¡9Qu`€\É„z^-˜,Zô0¤P 'g6"’hDàâmìpoîn¤|z—¨á¿$tû?ZüòbïÙ§µ 5KDÎÐéó3xo0È6M¶M¶Nâæ@Q•ë„+nM# ýº}n6(ÚK+AbŠY/5 (AkQ Á.. sZÖ‘ÿ{Z|¥6ñ¨néÃa˜&SßÄ ÒàZ„|þ›Ï·)º­ØudÞcÏÝB™xºüŸ7|UÈn¼ãø'z¦ªN¬-ÿA‡ìaäkŒ”÷~³bÂl)ßlÙk¯§‚U%´96á]éˆ)·+£™AËc0½‘JÊ陉ÆäíµÇݽÈ릩Ýêòð]Œ®]ß§ø!/‚ûùR1™rݲ_¦|—{>§ ý¼þ˸·~?³ü+Õ¡æ6xŒ1 iωÏxTL¨½˜ôb!£ë>%ìk&Ád|§¨HtÂl¤1Ö¾ëøÿ¬Fe~ Ë Ñ<œâýC nîRöÊsAæódÆçkäªGZeæÓàu(ù•éÞ©°àz ŒS«F¿O”Xï^™½ Ÿ'×Ì /©; ]N_äðÎ #.>ŸÍÛʧqìÝ‘ÝZƒHH Bo°,–PØÁjkMTÆ0kÉt[Ù%°~cñâMÖÂØ|J:ÇàÎñ£e.W¢Ïè&·z›'¼'›W’Ábóã —:õE°£0¯<ØÐ­©«Óc.¹\Æ!1*Ê*üÂ|”ÈZ‹¥%¼<:"¬-ŸðÞ¥ ͆ÍÉfzþ^ÆSwb5  ’p$¤ðjï=/k‡µiè &pDm²"®ÊÑŽÑa1Ë ®ÖßCŒËBäD(ãý&Ðäq’ÅþîòA[‚›q¹î¹ o ½¨‹)C&ÂÄG0Úýkt"åMÓ=5VúŽœä&þýó‹šØôÞoZ³Œ-~õZ¥¹a`1ëÅF à¼!±ÐæÆc¡F½ôåZÔvD’’pY½€´ˆ*ÂÞû¢‡b3Q½ÚRc&Çÿ+ßPíYÎ#ñ›.Daöúêê:}š˜*׫}&òj±¿çø}=[s°8¼–³€…Ö ÍQ‹(¹ 5r * cÇIjá{h Nš`4‚/¨ø£’Ùb_ä6ž4 DµŽ£ßÏÓ’š §"IgG™¥WÒ—5ÜG« <ùOeäñ/S…{íÕPñ‹-ûGÉqcÛðxpÖ:R;j>O•=Ž#oü­(ßjv]]?{¨$\^Þðïß±Ù…ž<ħ|?Z„tÅ Œ¯0±a‡…¸€­;kÉk1W"رB¼Yb€¸Ð@õQàP6íº­0iFmÕV÷à|õÈ®5iHÞ¼Åð’¼+[ÀTÊBFQg ú‡E­ïÀñÈÖúP°hð‹DÛz†H©[v´ + -j×­²ÅÙX9dÍ9=õ„TÏfÃ_‡%}Ðp¸ÓíÎ^𠪎ËÓxˆ¬o1cq8a“ÌÚ„2ìzO¾àe%ˆ’ÆGVT€d¥ 1JѲ"®Åb׳Keâ!—3©§”XƒÚYõÔí±EŽïbs®×gU¯+ŒC ­M-%È`™©¥Ù=²É߈ç“dÛdÛdfõ‹ÙQP²(MÍ¢¹AÔkLaúÚ}k€‹èótc\@iy:¾J²)õÃ:NYiw™Qf¼¾Ÿ³þÿ?»ƒò>xFÐæJÞ³rÖ$¡Ð®DÓ 7öy>CÚ“¥8 lGjÔëQSx"!`D:Ÿx”Ôµ =½}w‡ž „ž+äI©&H™)bµ §½^Ÿ¦ý¯kÛ‹t¹èç¡x¯Õ¬Ù¤ž°æ¢­˜£w™ÑØ)l·P'¸%þ­åDK͹š¨}‡öp f¶åQÓ!¿Ý_î¥A›TÏä7Ó»ñ½¿ÝÅÙŽ3» ,nãòðá—¾Oƒ_OYf°ÙFsTèÝòæ|±—+,e™Ož.ÍÒïe]1u_1‡íô9Ë%v~+{V;V–ŸíYe¸”Þ ­Ô n¸ÔuÐ]Ûù&¾Ôdú`ú9áø³GÑð¯|²HÝÒÕ‡!ÜÙ"ÞÉ‚øó|Ÿ¦è: ýŒQ®B>ó1Ç‹Ït-Ú7OÛŽÇØ­7ö8N³ì×]ÒhÒuY*3(k1ÿÚš‘ùœÞ¹¯Û_§Öe¥u·k¦Ç-ù6&ww>Rמ”¢Ïz¼ˆ B)Ò+µ`™XiÅ_AîG_=2µèŠ™¹A‚I¥@(Ž«¢„a&Zì4’•Íé/’ni-ØTâ*ap‘Œ”ì]±U›¿"ü8!Èhøª' iÀÌŸwò¸2JôáІážÕBñLÅʤÆ(P¡`¶é =Õ<¿Fásù¯Ÿ´åãÌA%¢íÓ–Ÿ…‹m»ÈA#=2EY§#Ô®œá‹õVŽå\•QXnÜf[–Õ,T.ë}“´½Öƒ9P¯¿fþ"±°Åá5õ·×Kï€W¤1¯E»{n)å‰~e¡7rø–PصAÓù¾Ô›Ä®úL?04•Aî·V15ÑÜ´2¹\ZE˜qeˆ|x/84’‹Í¾|Ÿ¸ì¼iuÑŸy§¸[A8¯+K3ZØÒj©E½Êh¶5‹·îÍ1„žÒP«T‚2RUò8(Þ-(ĽHç Þë\*…az€CÓª‹KË«œsvìÖµ ¸m|³îßw×?勌Z©'úÑ—Gß®-½œÇzðŠp§µh߯^È·”¿fEðô*4°ña ‡dGÙæ-žrÝ»«™‘˜" ?¡·=¸ãÍ(†7Ý_ífÖ…” „I…§Ó5 ‚¡²° U›•‘”ø•aP¶Éühð£f og¥_shz¸K—lË·¯”(} nõ˹Õëum†™2uWrx_î¦Á»º=¹tç¾R ô‰ýËnž»—|»Pjå%íœXër$í¥Ki@ ¥ óíN,Z$ËÕGr0×Añ™äY/ô>dþf!`m’¼’N€£¹Ãå-M„\8ÂíË`²Ü=øqPªO;Ré¤ÒÇkĉØá¡Ö+☄å}·ØOS¬î]ï,€Ï1+ÛèJ¢?ªV‡B& V£$ÐÔ:N-¸Â0m^À¹NRÙ3#b…‰£'¯™rºø_I 4@J¥X˜q?¹<)³ô=ã$¿xËÅÍ2OŒ ¸¼ž°Ž²¬È˜¼®e%/õ£ÛüÆ/‹LW®úç¶Ûñƒ'ÈÀŒ¬Zoï㙬í‘Ã@ºdö³.ÝÚ¸g Á™@´6qhÐèá^[…í*ÏÛ Æq„ÒIj…m]õ±ê1x/#© lÌ–ŽTPM—T]TÁj.Çwz? ‹èøíÔÚ!ÙyÇéÞ!²Âíhcl_€œyc‡æ0£ïjª?ªþ¨?f¼îU]Ç×èŽX™¶!šùvìÝ+¥Ù¹­˜áëé÷Ã}0õê‘Ù-FG‰†Ý+¹ZMmY±DXábbÄñ`½-{EUPžµÐ¢.SäÂémªßK¾²ºuë®8XRF!äçá/øY;ӕݘvÒì;`&0IÂkUŒ—ê³¾xSÒáQK(*®hüÄ}»Û!…óË“ØìÃæÝ%»&ŠÜ㹊Z Œ¨A½ÅM Ò&½Žjù9m7Æ/^s<-OÍL™LõmðK„èóxjÍ yÆØ«ê¹›Oºñ~l' â—/¼»ª¿nd”¹~ëûÒúùÏ’žß %g…eƒm¿Ÿáòñ¿íjjí <\ÈE<÷põ-ŠÄ§†û#=ÄèZÔZ¥e4,ÿ±ƒ IJ”'IlQÿÔ`Üw·ï:fýÜŸ‡å|ë6¯÷]~/÷+·®D‹Å*¿"Šë?Z°X½$W¸¯½ysaÑyÕ媧§É ­ƒi*ïý³Šùâ$YE´9*$©n«ß¾$©òL>/-kZÌXQ¶È£™s"÷Qk[I…V.âjÊ ríºw8ß¶²ç…éµRµ¬×r›‹Wg*[¥¼ÉtaÉ¡×)£¦¹$IlE ¯åTpdn¡k\½_]Kšàu^©k½DÆè!¥µ - yD€@TIV–Ì‰Ž¢³­‘ð‡¶5˜µJÙ{A^Ô¿Õ{2€SL@˜—Ùg©‡F”„y‡ “>!Öz][Å: gdå”>ÐL;ÒÐDd˜hqæh¥BLêŸrùµ+]†ê/~îø$>S÷ü¦ïi6¯ƒî‹[…ɘd9P‰²K²9vžpé4­(Ö  få| ÄÓŽ¹š¤Îu‚&¼áÞÂvÔÙTïàŸ Û–ï}¹æ#4[ªerìœx` # \S@úבÑp#‡¾h“y>å=¨+¹ê‹ÕöÓ*ÁX´ #ã*If‹ ~'ÄJ¸Ÿ•¥+Ȧ–/÷³ìèŸn_Òº¹l*¾n s’ðzØÔX8¯&X s‚áHŠO߯œÍ“|ÀÐ)JβÙeÉ9@šà s‰¢M¦J÷½k½n­ºdÿS.ਠ%†/#Åù<ŒÚ¦QîküùÕÉ­è‹”øô§B­Ltmö÷FÓ]g#%.kT³.Ži­`·³7—~•’9—|lMÐ"±†Ñ šPžàO¿ a•¥×õê{¦ÆOZEÄŸŒCƒ…?ßt¨Í• \,6¬P£gÿ•`X„ǨíÉa µ ¼¤  SãÌ1ãÝHlõ„ˆ¯)è¿!—ìퟰъ×ò#Ñ^$‚¿¤ý§H Ù@¿ÙU[jOm*ÚôØUYÌ–é{@ØX`…3 8ÔÐ4xÕ£Z@w~ž{gåó4úŸ„ƒÖ¾JE¿Zæd}NMÐWR-šGö–[ø´üšò_5ú¯ê©ì»‡ÙâžZr7SÒÚ!`séêo¹ƒÁ.yFõØljzéB¯Àà`+ —ä»$Åâò½OeocŸ‚ÖL‘³¦ˆtÇ} )ö×wïß*Ž¿O:ðšVÉŠ¨_% MS;özWIþüø÷X®zŒëä™~y¦õ'x@U%¸TiQÄ/˜ziHš †_ %¼å‚Ô³<†baìP<ÿBuÑ5Ë­é NnVMB%J(MTn5ˆ 1dŽ«}Ym‘ªÕ¼Ùí0·ìÇÌkðÆQ;>ëÂ=š§Ç Çé–å«S>Ýïz)œ}ñ!¨Z±I×Ͷ£áïˆe]¯,Ý@|쨛#-‚ûbšP3‹EJzÃ&s„Vʶ­ô°LSbSè«× 96ç×­À<“e$ê‡ëc‚’¦³5¤ ËäS6ô7è¿æsß}¦²T㪷Þòq„²'¨V@Ü4Œž¶d¾Â˜×`JJ¡RÌ£–¤Y‚DÈ~d[‰xÁUøkB°e'XȬëSopVÅ‘‰8…QÙ๎cFçyÛsbX_q»Ûò?ñ¶M"w"ä#;‘cî–WXR©Š8'«•L€÷™.RwÍŸBŠ65ì3둯´†a[žncxÜL6å 6ЊõÆûß:– z–šQ^eV=A¶Iø-ocov÷q™p· ìîH¨Âüqe÷¾1_%͉vg‡Ýcwªc«}Ó1½m¿ÒE-k§f¶ Èò¬Ía©}þÇ•mš¶M’ƒ»"éà{9ºŽtóþ‹ì\¿¤ËB¤Ô%Ì•¯%²³\a´ÙQ8qžfaÊá5ôÒ±.Ž–ÉPÊ­ˆ†Ì2Ã$ÐuL]3€óVµVÓ ñË,5Ì.?=?£’0ƒVð0×½†t ×ûÞúUiM$B|b;ðÚ»´"fªöFæŒ!ïZó¬)W+Þ!rz,ˆÔ“Ì GÚ\õo!<÷Jýˆ>Â7Žxžž¾~m0n[§Úw¤¤úO¿ÄN±”®_YÃydÜf’çâ5\xÕb0r”n½&°ÚŪ7X{Fw”¬dÃ>_‰é‡L+*zDïC¦L$bÈÑhHHõ4ÇèŒÞê£m‘%emb"¼$qûÁ*’ÒÂUY¯£Iy|oüvóÅŒ-âç«Ìã1ý±”[ÃtyÚ–7šv¼TW2ܼS¢ª·ÇŒ$ßÖ Êïvz0Š \`zñ\_"©‘œ.²’ïúÏ«p>Öx‰¶ž| o•ôn{2 ”¬Ï´fh¼”2œíceQ£ªÞ¹ ®ø¼ ^%Ñ’° Ô…ý·g`*\¢£^†0?i1²ž×Ø…‰†{v÷¾ùmG;ñ]u|Îûê %­ÏiK¶~….É…Kãz=næÕÒ}pÁ78×RT` —9¬“žD ¶×Ñ€’ ‰§^Ÿ#šû?¹Ž¼dÛU4z?XÇo”×_ ¾Ü.Ó½TÜ5'@j8¡¯e*j†£±P0h°–íyœò¡g’BÔ['Gäí,UÔ–w“O—$ö΢Åê,ük¤·fT¸–x'¹€,}Ækžïä@ÅŽ[)¸§Úu ,%åÇT±0Vˆc¥TãCP˜Úö'Ù,{3‡qV²zÀ‘¢kµ„¨A#øHnô¾çÝÇÈßÏ[•°BÛÓÄ?|¹w4~(|'©¶û¯~˜.ÊfÆ>ÍovH¿>Ùɼý\‘`BD Óy®¨ø´T(áò”²n* r>x¿áKG—+Ž…Iä©óJ/,¯'(Dš®E(sRH<7ßð7ͪim—ZÀ+ š·vTˆf!{Ad ¿N•î[oï+SÆu&ÄŒ¢ÐÚ¼ ä“®ÈÁ‰d-òÓ¬ñËj½ò‡ÂòÎM B²XºÌFQíÜ´E߸™”n(!_*lêÂ5 Ö)þçÏ~=­7v‡NëñnÈ¿+wë=U5ˆaôôøšWGÆþ–&ô!Û x©fÃU˜óñÆN³ñÛï¨úOÔ¼Ô…Aìm Òôüöå¥4á°dmKáö›à}ÖøšC Cm‘a—Ï¿›¶¢¦±U3åÙRN]Ö×¼:M¶™‚ª>5± ØÓE$¤æ‰¸!7ŸQ[¡^\™éŽ›¯M' -ç-ILÌ„›ÅVîÂ3j€l2ýÂW¥$ç¸ËÆw üçîÒ%Ý3méÙÕÓ^°¶¡øXüâŽ6˜èZPŽXDã4õšñ˾aoæ)÷ÊZñ°/ŸXXMˆH-EjëC¢Áº4ÁÊ4—'èÿ o“\z)Øš£ ›"¹M7zñ€½ðP“O®wB”{wˆ>‹ô»—ND"Ê2å¼™ÿ>¯Oo³ŒUƒy²šñÞ•Üý`QnªPññ0 Ë®ÒäÑ|¿sŸsž×ye€]ˆ-ÓÃ<…v}Ý}o„ºïe‹½Œäñ/³RJö±‘x1OH;H3Z)/ßcDg°ùÑ-GM¢ÜèÏlÑ[n£n«¹§K’ôìi÷ƒ+’Tß°‰Ü„#ßÚ£kcÖ+Ô5?1yYh}ÔÀx0ÕÝêÓÇËæêo¯ô䪺¿&~×¶ Sx¶?µã¹¤Ü=<¡ÕÓ©ªœ›^窎8ú%¤^!H7°xàrÉÅÎÈà!T„t¢iŒÃ‘S6û%á¤çظ€Ç›(b •uþ¥ÄÇÞ,ª3$ Lä_ ©]3Ýõr“{åô|é+NÆÛ«M)´þ>íô7³¯‘¡ùZÖÏföyF Hfñµ\8”ü‘p€’ã Ý )ÿáy1c<Úhb$’ÝfÕ· qª^æh¡ë¡©ˆÔ¥YžÍ”×”­TëßÄbÃ=µGɉãŒs!fhqÚú¼ïÒÛŒÝa¯ëÐÎQ±Ï?/{.€£·û•xöùc}dc¨Âb’Æò!ʰ¥ º ¼@Ì|~[ås …ZºzVŠ”kÓœï »¾£+_—Iï©óå²Z¥Šômúð¼ãƒ ÁSÓ*ë ·[M3aån¢îvæß‘‚ílÛ(T11N+G£Šë1éàc3<ÏEH„é¿F‚³»?8ˆíò RÊ@n«‰óÓí{µl!:.Ëo/6•¢ÕxK3*(GÕ¤.>È êç¿k›Ç¨¿'zе¤eš©\¼# æÀÏ£·”ŒÕü;﬋ñsvÊ-¼³ó¾2‡¸ö;è2%Ïgë쓬&ó¼ë8v`’¾Â7Ñã£îð¼§mn³®â4²8EåèuÑß(jã-šÙ÷†€’ø¬‡·ûýŒiyÃѺsÃh¿~nY£O d'©«2„ËÊo …í³][fëŸåqîs@ñÀ¥i.J¹‹ò«-W`0µlË4§ý³uvj¹˜¨®Ó8¿½@¬®‡uÊ„åŒ8µß~ëóþñJ¾ ÐS,{3Šh\_=üšß˜æìCÌetøm2ÇÄœVDœj¦Ð—Š×pénZâšA±éõq4Í\ñ ÀÍùâjÙö„ÁÈ?‘­Êñ¶Øði¼&v©kp_$ÛûÕìHgS^‹ç4‹ƒ+hè*RÔuâ¹~ò ÈnÂRÉ)UÅðÓº³ºœUD¸"ž¦nÔŒÄ(<µÚË#lU^-kÙ ~l¿aƸcÕsä,ïpÕ*ˆ(£#ÉMÀà„œ®ÁÖ~Ü?¤Ç¼ÚuöOï–¸ü"‡xÒª:œW3u´„ R¸-›TïÏéû’AÑ,Ü¡´A%Š˜Ùl{QhΪ³%ä'±òÏ¥Ž¬r5ˆVØž‹—¥¾No=™!”³¥Di ¸–yT ¾‘¥¶uø~/ø¹ç?ž9;éêú_íb¸vʲ£-ª¬B;F¦´Y ±?•““³¢ÞkœÑ`üÿî¾8c7‡ÕÁðÝ1,=s8Þ\yå< ÃÕKºæ¥0\Né:f½aÞj]Ûc³]ÂB½¶n`¬²ŒŽ•"n®†šÏùÇÕ|Lª×èÝÇN^‘×" ?EÔ‰bõO k¹ä“b”h‰²*«Ãl ·¨âRYfÐGu{¦˜Æ®Þpª^ïû½ób¾×Å×\ç\uQÕºÍR§jÞ·aéÔ{Ý—íW‡fL%(rÚÕRΙ­ÑBS'­³ «÷`ÿ…Áš•Áååÿ•zF!útxn¡»ÒQBdtYø6Áâ3Ã>3ƒçÎ/ˆb`h.PïFH%¹ªõ`¨M®ï6ù¦bª·_ŽYsÚÇ‚—v€/Œj>’ŠÔ/>€c82xoÔéŸ^ÖòK ˜Ømcá„6éÉq@u·E´&µJ–ãkŒ+¯·à½–Ƈ‰!VY¥8)bqt®„ÅB õ¨>S8†Q"~“#„·4μ^¸xÑ$ˆ+TnÖDßVt–¤QôÏ©Ns•LtÈ`”x ;>óYQº÷"²ÛߢۧµB,Ùžkù1qp׫ìò7Ÿn}oÙ—ü½_W½Ã¿Ð{v ç¿.“(XXCø+ÈÄ™®»I©À"àã2€}ÔðòöHv$ØñeÔuQt¶R–6¶™M9¢¤`úò€¡oIA \²­C©UŒâ“†Æ7¼aYBiz›È—í[û(µ38½‰Ì®9UuM²ö{^ìC·“s~ »·Â‰¯Kn0+ S®ø«] *7û„OSÊôïIª ¿ðòVï„0 ·›8åò»ýyç„ÔøÂ9Ï­qÀ\¼ÑZ/nØ-üëne×ÄV»úÙÑÓÐ.ôðSq]n''\´ˆ0ÞnYI2‚bö“V‚`ÒÕ¦øü©£(P›%ÕAˆ5YµÃ•^mc…@Gü 8¸D¤M‘CVn“/‰eÙwËÆ -Òg,–ÉèÎ-Ù½ËÂÍêïƒìÈÙüÏÞ/ô@ëâzc è‹äD[¯’È>]ÚÅÎxe˜IÊR>,8ƒdÑEA¤’á»éøýŒ&#Õ|/­òéh:°·Q¯¼ž÷ѯ{Vóqá½·Šâé]QPpZRÕ¥ÑËÌ· ²Oðø¤òv8þüoV§",§ìµž2ž.T£ú ŠkèÉŠÅP´ §% °Ü2dæI™^¸žÎ½¹?`Ñ—ÎI…Ì‹¢Eà+$^0_Ô¸|½H q^½s½’aJÌ:.$öÝts[ Ü2íéÝ÷Z‡·R׃‹bvp¸[ªç °Á>NS‹-ªÛFy³¸Ög=u1í¾ÿ&S¡â‹¨²Ñ\µïö¯’Þ;ÍÆ8Ìßtý²õ¥Ž~KùJ÷*5r…9ÆeÕ—$` ÇN@xÖ¶†Ÿ¹]ïÖë´æ.z£ø#2kLiO¨5£)måL¹°g©a–øÓA&”UJUȼ6­—æÌNkdkÔ°Ú ˜8NnQä”& ûö~®lºðrahC±ë+Ö­úoÉù§‘‘Míƒi)™>{=·yw_G ߨP!§b˜–7¨oý¶|üRŽ÷ÓÌ)!Ôc…yV¢rÌéJÉw\8±HmÂÛ1ºæÙfêወÚñTÞ×™ …ŸìŸ-ú”ö©àÓFm&qBuE7:¶|Qœ ÍrººœwÓá‘©i:”Þ¾OX^1wînþlmª"XB¯R–¬Å•%ì÷U;ñý|ìj¢¤ÑZê¦8µš¦ਥù»wnºqÇÚlí ׳[2#ˆÝä"øæ%]‘ $A–W”ÜÏ'nø"díÖé›”¸…T°q\¥€ÄZ n’ùÀÉM8Gˆ–xyº/’Áü*[·ÖÙ1òh<†Ò?µo|Óp>ã3óÕ§~ËH»çE¤u¶Táéè¼Ê¸‡ìÀ©¸ÂÖÛ"¥BzK¡L<©3…¦£G >gÐ;|'éÎ}k|ë )†_Úìé9‰²ŽÉÀn©W/tì¥bñ‹¡ÌãowÇ÷øñÃ¥Iž¢$b¬²CdYðæ‹l‰PÙŽ;¦ã… RŸ…ºÓ/Çi½<¾ïB àT"„ú8hçëKŒ@”à~¾TÏ»”À=9Ô%å³.’žÔNmäåú=>1úC ìz¯(£êUšÔ¢g:mC2º's®”XÖ0Aî+ÎÂ…Ÿ äö øDÌ_Œ0 {°k°-j!é¸ s54M§”˜j}›C<*Ëçs(@賃\¶—ÈÍò‡dnkí>\™"]ëj¤2fÑiñjý¬v®zx÷%6þµrXÛTp¾’|ãZÝȨ&ˆ¦œá²H~D»QÄ1L,H+Ù$‡Øl_„6Lyë¡ÿ  kWýT*½àÙðíÐyÐ@óÒHß"š63F¤mϨ]£U 2!>¥ª.Xtü}§Ë8 +ÚNTø|þÿàm4ëÁ•Û ¿ä“ê„~òô¬q†¯×·cÜ®9ëO'ÌÂâóý%•ÿOg§S\ð­«_pŽ+»_³Ì“ ·Þª9¤OºÑ™ÛóM5úò»âª_Ú$\‡^y”PHñ{Vf½Y½È­H%röúdénßÓ·Ü«¨ÜV¾ZœÅˆ¼Ê9‘4Ó&½ËÒg¬¡]N{ÓVù«‰í·­üñÌÕíA2L?ešW ºTÔŸ†£ ‰¤øVr®äÞ—žÑÞGEŽ ²Û^õø§´¼ó×[â ºÂ-…ÞY£º°ã$,L- ‚üêÎû–Äj \Th*¥ÓGËhœ{È÷†áVˆvPñÂÂoÚ;B¸Ò£R=ÓÝ®JìÖ½›u÷ŸjÄèÜ &n^UÓvô„œVq‰EH³öÚÓZjÞ”ã,ÅŠŠc–nsfæ&`3ŒÌ´D >ʼç~óç[Û3ØUãxËùnÞÞ9çñåÄÀ¤ªÅAr’ªdRT†ÓÂ-‹Šý(òÁ«›bqDEhÔU>yÅT%#]Æ;9c){2ák•¾†Ãh­¶Ëyš-Î;¬½yXœÅ>NTœR†n†ò í–ÕMWlÁµ=4<0š±µô”&-½…†Ã:X©sS?á½Ö7‡?·ç`­ò*»¹TÉR+Vˆ¦¾®kPQ„Sœµ@ŶcÂyÉ_t¼–kukÆp³u#'…1í(õP4 _6,  ×1Kh™¿C39ò uïŠ R×ádÅŠ©w_2gËÈÝúw5”֜ᒚ†»ÚjIçk,BuÈ&3¢kCpb²g&èõ'âÖv —Fi ]ÙdŽÒ*¢ÂÏûKÌ£€†j R•9\Ó¦ì»?×Füvïå¨BeâlŠÜë]ϽÈSDRè u°¯€ÂK~èšîZ)Ñm¤›¸i%XÞ<¨ÿ”a ŸrM­–­6–¬1ÜLvñ,±“DâW(6òˆ`pKÈÎé­Å’†öã Ò£‹KüëŽzüøÖùøúF#:o‘œFS »tòy¶Xà™–`Ñópõ ÍTÁY)Cù ¡óË!¾©Oñý–øÏ¾ùf÷Åüž°ª‚2놆eî¹–šñð1 ¨µLÅ%á#m‘!eSžYçà´‘®ú«®¦²E³;+znÈ4ï «WZÎ`ÈgÊÊ›‡‰pP‚ÉïÐbC’ëï¥ð>þ˜Aé”ÞŒù#ƸNGx*.Uw¹†ŠYd‘&ßî QŒ ¯Ý&R+¸—àÀ“–Û*HðPW2æAå§{ tRØPÐ*Q@‰è˜g¥l{Ûh©v8h"Lиˆ™K>¾™ÖzîbÒY%&¯昬ZO®û Ô××”Æê‡ŒBH±;s•Ð\„7Qû ¨C'™Wi2&À'^Ù=íO‘¥Ì»£ ö=\›•¸\KÄÌ´1Bðº¶Ù ‰ ˱Þ!5t5Ù,² ÛvÞÜl*LÎ~À¾>;a pOŽýÝ=nwìZo™3åº?Lö6ëÕ¦0­.ö®Ø}Ÿ¶œÛÈx¸çÜ×rýQgê5»D<6KŽN¹BhßOð,ÒhûÒªñúXþŒ»I¡Í¼yÜ®¥â|l§È4‰5ÏqLª"@¬îCùn[aÅ<š ¯&pd鮓’U ÝLŠîée‰‘ç"þmF$Ü‘qb(¼%OA6›R•1wFã$Í\¬däëFR†ë >E}0Æ»ÎÐ7M%ºc÷QºÁFÞou ¦1m­×£XS£L Ñl™E:w*—}¿o©Á|pÛų´aLl8,±ÍŒ Ox›½ÅËTƒˆ’‘oïï$åþáÆò›SŽS¿-·¶#xŠñ´¿¿'I¢ì“Xù¾¯èõw¬ÞºoеNõÏzŒþíÖ'SëÂoßJeP3÷Æþ"ÜjSSæùìͱEo¾w|ÿ±]twNRyiÜ[¾»× •ÀÈz`5:œRꪎ©(z3ÎÇ·î°Ù¬³ýæù‚›„ÁÀ|â†N6šFµËUôŒc3(˜LL‰Æ%¿{ôÞ!úo9L{²)æ4¼iïÍ]†&qŽx®éÓÇ (/ N5)ß§x[üþQ=wÑÊ&Éáeaae.Qü#©)‘åª]åÝåáwì`çè÷UŸÇ|ñïü\wžçQÖÖu¤ ãÚñœÎ¥²“Ó¦þQ%}Úè¯bRª"]ºnã8!ùÓåWñ,ÉOcqû¤Mî‡{—\f{ÿú«Üµ24;L¼V¶ º'c½`Tt^ô5(é®@Ƈ_eO‡j‘×ÒþÂõµZ·LwI{µw uêöµú2´ÃᲡq@Ÿ'"©¤ ,y)¨²’ÚÚÜ­t¬mF‚9ɸ©x'J/0xVøêϨœ_KÒ%¶òþþxÔse¦"2æïÚ{ô_ͩ≩¿­Ës/ô>–ZyJäFyðB}~uÃw>±—²µîãÝLÐø²U¸ä©°@çNŠM?)çŠ)ïV•B*Љ֣¥„‹T•+Ë–ÄàKè4pÙ®.žbdãÜÌ ¢äO@¤Â´Q†ÿæ3£ÃŒ‚3#%ü=,'7ÄÙpÎ/íqÎÚ”Ž}ñçŠYÈ«½X£yf A ú«ôw(½™§ŒI×W’œ ©P†­2§Þ†çÚ{庋4ql@AX­®Å S¬·<œ}¼.z{Âq]Veƒs?$B–Z³Øõ ”•é,×–0”Ü÷‰”„”}–ï&¹½³yç@mX³Jµ l¨ö\Ê£ÓŒ(UšJöáškíß®ÜË&%ñS¤½ÀoÜ„ëÞJ¶ó.-Dpe2Ânî57HúĈ᧧š-œŽ¸äVq^|=Êûç_ˬ/¯¬a5,™™»îÛï}+0öÆfÝ‘6G[P%q¸šKž÷Ú®Ô¢Ýá y5Þ0Ž2Ë‘–|¿ âZ—!c‡Hô-òʸ|hŽ8QÞð5]2×rçŸs‚ñ¢,T¥¨ …1ê%KõZ®Æ*.*„ä²PZ¿WO*–¨àÝD<”¼Ÿ«uÔFÚžs‡Ö Ì”TgÙôåL¢®l,)—æý‰Òq®¯ç©Þ“=1 Ïr¢4îjßìkÂ\éU€ë¡ ÷wý2” íÍÅœÔÖa4bŽn®GÞùÓN72º›¥qѪújb’ÿq–Š Ù‹Lxå¦l¾p}´ø¦x7ÑpU¶m ¯Õz(ä¾ÖÚð-åóžšòúPÓ–½°3+()pÛ¬ª—Ð5«ÝéààÐ º|bò>›ovÿÕ|GŽåØOz¦å޳‡ùyGùŠËqÀ~èŒ÷%F–E½Ë<MúÏ2’0ZŸÊE´†ÃÕ˜–5‡&xÛ “n§.+‘ü%*i+8äêb{‡àé¬gtDCRfj¼z Ôø:›~Í+€¼` `›ÙTdµL}ý3ÅÔl“—ÏãϲMXj°\Ë·ä•’ƒí(ÜäèÑŽº ÎV«ÀÄ8ÐÔ,ŒTpªÏMoxí[ Õ¬ÿ ‘3Ç{Êù¯CÖü;n÷Žï®Ó°ÆkÏŽFŽä=h»Íñ…ªTrÎ :ÂÄæMŠkÉ8¨³…ïˆÌ€áA% ç•' ŒÀlÛ§("õ šÅ)Lc" Åb¦¨­ôÐ’Ðm–ØÉ¸áÕÌk­xç‰åôõÃKZ4C¾Ô¿9«ÆLRª…Ãò½"þlpÍ ª7ÉÇck%‚S*±RY‰Œ59‚¿Y¬jÀªÌŸë¥ˆ€€4BØ$¥Ó¢l”±zfº™sÖmgÛBV’qŽ57ÔÖª˜ôãªZÍÎGH_xeLîªÊ3züоù"С®6a<”V™ñ|ÍV#n} ’ ‰1 b'“IVi*&ðqˆ ³i^ôR†q‹^~ò‘uÉŠÞçü;>V†YÊfÆ{» «³•ËiÚüJ'–ÌoS÷Üÿ¶x. Ú=C,ÐõA†<¯{†_)¦¾ÅÑðÆŽóÚª½Ú}ï©ÌÂí¸tÎúêþ~ä ª¬›ïV²^¿ 68!6f º¹‰³\!ÄO«Sªü&À0*î~í­Ë EÊÕ/ž^È»Võhÿ#µj$úΛ_ÕYÌð#”úõŽN™5Xfš oŸyÖUöó9•)rŸ.»¥ÈWñÏa5œÅñl!ùÏ —§6—¾mþBz/[ÎjóqÏZÙðîx¿ |Oý[¾Î ùãß?i\«Nçх»C‚ðáóøøèÁÕÓøíz9P¬ƒ!Ë7/ÿ¾>šõž¥ñåÛ óŠÇ¯ÜL·Ð0¬¸à†îÄMÁ’ctU‚lz.,\Øù¥ £Ï$†ø'*g0­»¶51Ÿ–£œï”Ž–«”ÂÅ0gIháÁ5±i‘¾Øæ¨¯dAèTE>A0§°+A78Qªšå[ë„p Z<ÂJ‰Y•íå¥Z3Öï¾xsZd ¹–xïæø_ÐTâUyL>Wõ™xÓO,Ø 8hXiêÕh!^ã.9oæ`Å#$0ë—¯˜ôŒbåÒ¹ÉàÚ9Î&£8A}Êï­óŒ;Ç«Ñó:y¤ 9éoˆ—p‰²n°»ÎøÑëu£Yiõa°‹^Ú%êß3¿<~×ÍÇù^œp„ußžÝñ>F—Á³XV$%+°mQÇJ•¥«¦3 â«Þ%v\Ïî×óž7·ó?·Gõùg­!!Š[d±—¢ ˜2‹‘©¥ÉCŒœ‘›ÛòNFÌ™}Œ;Ôä°Š”„´™ ^qN…+ß`ÆU˜^7ßkUO "$¨= ]•LW÷0xCBÒ§—Ú—-d¶rÝOësë`¾»ðÎŒTˆ<©v¡Új–Vv"¦} {š”öÜ´R¹Î$ØÏÁ¦Ž~ÙÌÁl¾v{^çK™´­òä‘Ç¡z·ÍRZ°ÓPô•j"LµjŒT´TcF““ÄiZeàXc`yoÛ5GIº^,ö¹Elà_ì<ÔœŒËYZ˜M«zª=;«ˆË…º¥€i+¥ï¦Óðxsù­cÿï”}è騔8œýb{*ëø¡›%0˜çCÖ¾VÖaËjs“TaD”l‰,ËvØ£k«½“ ukþ/ÛýaÆÝtƒ’øwFZ„«ª’ᆜ̒5F[ÆøôsfÌż\6Ÿ}Æ*l®Gµ{äîU¥Û.RtîCÖœ}Y·Ì=ø*#vÃGöëu ÑdÚxÉRƒñ!’¶")7¶ÈŽøù°øUÂpTÜ­Ea¿&>2»{ña[B:–˜ùìÐnÂ^8NÍ¢2:ˆÇöïmâ÷(€Ý¡Iò¹pjP©¾^x±}Ùô*cçô¡´E”´vËjbgIJKÔ,¼TÂ2‡Õ1wÉÇ%è[£JŠÓB“êØ^hvT5rv$¸ê«›ð?–Áêt¼ú™jÇs›ìÝé¾8ÒfY°®Íßí+…€¸­Ø‡šµst¾:W3ëýKÀ×N-3ø‚wEË+œ²áá°é~8âhbpe0!ÖBOÎ5riG»«!±þ8úº¶šÚ¼qÂû`fò8|y»¶'7Mx•v]»ùõwU”M“JÖÌÂy™ÖmÐÁ¦Ižt†ØhÜÙT-™µ“]«„¬ÚuÍ‚Älˆ°‘—2HJôÜ àÛkZXg©CÝÝfi9üßÍfßçÇ)¥ˆñaø}§E“Ĉ@B¨,ôhƒDÁ*׺ÐâtXð’½¹Á’û®ñ “z‡µæÀæÙ„{cÔqîúØÂQ¢ÁF¨ÈÅé¶â|o–jÚŒ5úXÉù`“£#ux‘@8½³Y“¢L¼ˆd·Î=¾&Ã]Iû{z©0vvÌõò3’ãZðÉ'×ô+§…ä>Ø·n¼vcû8rñ·Ÿåyoâöyœ»V]ô ä†Z¸T »zõCÂÎpá…ÖYÏ›†ª'LÎ2ù…4@Ç[EGûšMœÅÀƒw1úÚö?T}Í¿zëµ”®¾•vÍ©}Êòk#â–¥xèÅ“ÃQÊOÎAñƾ©WiMHžÃB©ÅË)SÂád·®ZÂÑ=pj•Žs¨@t…ýû·òêüóSg,Y‹«jo“âtF} Õå«DùJx´Êß}.±8^¥ã.â¹~Թ̰8­•É’P…Zf(Ù­Bçhøœn-ÿ[H8È!3Z (`<…#Æy>èkŒ^-Ò¦é%µÄ.á¹kaƒ»7ha¥æQ»6p"ä—Ñó~>ŸŽoù¥>xºNÛôíZÍg¢Çe{êâ¦=™¹ÔÊB¥$– <ÖFùžL„0žÃX©'5™ÅS‘$M|Ô”fNŒ°õR™ˆÏVŒ¢×ÝÀƒ®13ƒ6ßÅËþKßkþuàöBÎìMÖ¡Ze(ÉJ”bl^•kqDœ  ±Å}Bl¼hón)驎IjOÀ £i%>Ž[7øï}JJMꌪÉËU}6îž_IJw9Ž0Ÿ^ûŽ8sd`’#kÄVH±=bÓ‘5ˆ_˜¼f@ŽëdGè5¾æÛu!,$ÝçQ=­žks4`%PIàxµfB¢tÍ®~ºÑgY­$}q¤þ§{Ƕ±ÂÜW>.›åðŸá'¸.[âƒTFÆúy_:¤p~ºuJbÄþ¼bßÍ_N¸([Ò8ÄH!`Î^xÖº–)àêÑÆ[þ /MY\|Á¤¯J~ølmi¸tšsÚAt»g&§Ÿõõsiù†¹gﯿ¬`ñäad˜—å(}V| í®åW°²ka_$Aêb¨½a5Ë>%rdÃbNºpEFFÀŒë©¤¨!½¬KíÔ7îßÌJåü}êöiž-åz2ÂÌè…¢Èq9.QrÛßÕ×}gÆñ6Å% @˜_f\{.å7‚½ÙFdeÃJ)™ î4.U%ƒâ¿‰×Dó”aQÈÐ0«Ù ÷ÒåF%·ETütŸlé¹€§ÌF«ÊŠ~M£.7 ä'²ãpIBHì8Ü ;™ßIL–U,DÒŠ{$Þ5¡‰  Ç‚à(N›Ñ%ñoÚ°Å‹`½±0M ³¦p3²ìË\KG–´SRlx‡²k ™œ©íbålZ ó&µjªþ¯¹;ü>Ô§xJ«Ó3á=ú´Þ^ïØßo³Ñ¥ Tí·\~ZöÌ™«t®³Lbá£8Mªßn¥À¬À½©¯ôëÏý»½ÄÒgbÆ-£øº¥ÆõøöÉ¢Ù«žs—óÔ‹aô¸áà6s»4é8Zà b!9'7 aMŠæ§7ä1Qezt¬À½€êX3Rõ…VB³Ê-ÕòH  ”ýÞÍÞË’z”îÕU¸™7)ÃZ‹œ6*ÅÌ0[KÜÉZüÌñœó)Ç©µPê%ª àÂåÕc,Ý ÍïHCKÏ…à †>Øà:Ò ^ðþ «Ý7 ®¸¶Öô¯qÒÑVÅžÛ¾ôݸÐ\²#•¼ãëôÏÄÜ¢8%ð›­(øäKf“…ÑáÛªÕÏ&çÑØ?¢ùæ)ÅÇîG¿Ê‚7;ûêíà UI„ È[,IY ^„µ8˜K2(°Q¬]ôÑÕŸêÃÜ#y´ü¿Eï“Ô„.x¦¹^”´÷Ôw‰¼P^ÏOYÙ6IËvt/Ûçù, }/Mo/ßE‚ˆãÙúWîµ,ÉÛƒÁ+L*Àm F ­KЬ”Â5ÜâÅ{Á¬å(dŰ4öêÕ0\ðŒ%d-6QÂç4EQ&Ùšoe1Õ;ª»ŸijXˆ[•v)zXÐ ÇM¶-lÇMõÆ1„ Ñ-Ò§#v<ù.50QÜ+~µ¦= ºécû®8Bó{Дçk^ÝóÜÕÝÍŒW\¼_ÏröCÁßþyÌ0žÍìà°&s_‡¹‹sï?šÁ2Ìý4gÇw­6W+èaÓ‘´ŒsŠò¬©Ï n·Ÿ°nZó¢·y˜òf®×3›éq» 5ô¿~UŸ×·©á»*ÿE;̹ÛJ¼«Íf5šžˆŽSM›†#“m(¢…ÚW%Õ(‚øèÊ9$(òâ‰i@âÔ¶¼›|FfÙF1j$Ÿ…mõL/2pdë ÅäŸC—'ÄÌ…ÈäYk6ô·ˆ승áÇP›+úÐæjT0”¸…x­¿=TŒô+¯~9±µ÷ëvK^þ#Óž-Sì1+ ëó#^è5Ð’ëq›©É¯Ý$5“žO¹«í²ã7àÅå €ïÉ1¦+kÅÉåjŸ{^;+±HÊäU/K(·åÉ~zä7¹kykÜ[Áñ»÷å)v‹íÏ»¤ú¯”0ÌßSh½Všj¡:ÏjG”öd[Uþk¿/åübÿdçÕÎ7¸Wò‡yë€doÕæ l›'gµGqZšk¹w•¿‰Ú›!›J=PtPµ©l½òåE³`ð›$hسAFÍnºM ¤,`Áš6ë“Çt/ãè¼jS+xAÅJEÝà€©v2ô3×F›†Í ¢ÅmE´"Ó¼¦µýiÝ+¥%-L’  "äå,ÜÐë4A—†²Œ12*TõB Ov~ÀË&”3 òu& $¤—öJ35˜×³Nl/W5±vÚ–W€ÆmÈb’1ž8“, §’ºSÝœ¨Z)lgf¯JnÍ7yԣ™í³Ü•O÷Ϻ»AB]~WnX&¿ Îhÿþ)5O/;FW™À># u©}ʽ<ö?qí/ ˜?©§Æ¥¸&}¢QXK ·OÿÚy?MGµ{CO¾Æž;S_n0ÑÙ(ôñ’`–žÇèéÕÔN¡ÀìãëÓSß/k=Ó )B ÅZÎÚ3•Ór3-”)(V‹£X ae~ÕÓ´÷xPò)%µ*¸~ÿã< VžFB –ƒì\ª‚éec§£1+YõÁW-ˆ’ÖÚµøø™uòRlœ•ƒU­Ñy]kÐIÀ îÙk¢ÝÍ\N¿DQoLTc vÑ´VéÎÄã>Ì>*Vᨠ´P„4½jXŒ5*p½5YK?5bëjýg½|a9‰_™TcR…#-c9K»èñ•l/áÆUfµÔÌíëþG´kÎ@åFB|]c™× ÖxX@Ìì#dÙ/˜y‰aÕ½c‰~˜öLôê>Ö¼O»8ïˆ<õ—F¤åÓ2Ì—ñù[Ô×Q™– ïŽ…øÐ«çý æÙfÅTžTÁ|w³1P~Ç)@‹ šnÉŠ9Pvr)§òE¯ÉJp–±š÷ª+&´­÷:†ÜR¸Ó“Ï™,èr9¯< Á§É“æJ˜©ÃåÍÖÃÒÅ9wôÛ‹»—A{£´f—Ýx:ú„Íœzíl…{´’Þ|~,aÚPÑÒ T@*éHƲã)jɵP;>ƒÈpž©lF¤#<„Þ’ÓëÿjÖ‰~x„MOÌkËšÑDcJåˆËJû™dd™ñD«Å 4¢÷…nÚ'“E.06œ¶ÚžnZ&¸˜£viÙ­¡p›§!r  n¡égŒî؉§&¬aP‘q:† ÁoòŠ5$ï‹â¡‹ƒN¸À…g±,ã Z´ $6\n¶§ŸLï>'éB±ªþQÆfóßHPUýÁ®Ù`ó}žjˆ¦(øŽ4„¢håÖJ-y ÑÁ3pâüXlýqÕiQµ *¦)öVžíì€ÝCТƒFL™ên BÓ QîžÑÔí­mi‰œCÈË÷*Ý^ñ‘)—¨~‡¥ú>‚Cc*¿TúsÐî­ƒEY"Vy½µ'‰ Ùx¼.sêÅún’£Ù‡Êæø¾¯ä ò>í±§+vÖ À ±A²Ën*8 ÂW©B0dêÜÇÝÂ1Rá™>õHÌî%®¯«EÓXƬé}óåÇ:ÑYMçÔ¹íIK9^ƒ…ô w-o µi‘H#Ø,E¢ä·àÜ u%p±­Ç9ˆ“…ëgwÖõ°«Sâ‰ÀÌIÂÉt«­Ž@ævMÜxÜÐ|Ìn‚°R·É®íìí1{tÚ—¯ínn¿õ¶­ûueÉù÷Úè¥ýZWù{-ý³ˆs sœM—ií«.VQ´å¶Úä]sJÿ꺡§%=Êl.{+½c­Ïaéݤå[ð/XQÒ¡zÐûô£CüP,nQŸ³o&T„jØLåj€JOoB¾#G|~e¼Ïà¦Ï‹¨º;ë|…þþ³j·'¾9õ=/£×=_¥?†ùê±àMÞœ)'2Q™‘Ü9J+&*ŒÄn…ÆÈAFGo“³¹½í0Á Ѱ£žÓóv¿En—(ÝÙ3»¥ Ö {Þ–SÐ70)˜Ø®AÒ× 9ªÌ&o+k²,PšÈ #È’:zÍín5Lbd±Y¬ž,¹DnÃiMwÀ>«¥Gs—¸0BË AY°2R³ ¦kŽ€Ð†HJ$¾\™ûŸé©6/QÈsúªak•¡‰5\—Ÿ•a+Éh×ßj¬p[7ÙyœKbI>“ɬX"¦ï.âÍlÜJNç’¬:”K^ú2%-Vá/S忇T{Zî}„;T8eéãNQ¿üý÷:PS–óÒ¶jÁW©ŽTœ>ß_Ç:u>èso×üöÌV>‚[ÜŽ«]ÛÒŸÛýYi?Íè×ÀIõ€Ë vÐ|é÷i.Ñ’©HæŠî—г‡ü>~KºÇ¬Ð믆K÷ÞOôÝËÔÓò½¼çãu¤­`OÔÚ§Ù{š¿MiÛ·‚²íY—Q]æÓÏWŸ–ÚwiÌg›}}Ñ왞å9ôSg+i˜sr]d˜¥¸9ClkŽº_v9ãOéØ âÞCÍÐݬâ•4ê Dñ'yúW{X2kö‰I¨7Š%Q—žÆ$T=öôÏ¥Û³ÄpëͧʕÈ:]Ïo+i©æû$;aÌ¿ò²ØWó¼öTj1¼Úw„y B•%yÇ­÷ýšþ7õ'Íy¾OšjŽ9káešy™®•× ú5ŒvãNyÞF38i^VŸ7n²2È›ˆ>®)DsÖö9H°ûH„CxK°,mör¯%ˆpª ”Í÷:½å[§òÀCò4¦ï¯¸&4êµÃ'±ÀŸðQÊkJæÌ¢ 4YäQ¢c-•ƒ.j›ã=Zë^j>5M؃gÝàæ6a]3ËSà]a䈣lâ5Ô{ð»úcݳP`ž;@ö¥¨¹jÙ–oåÜðÎ|ó½ Æ6è.Jq*ïËkø~Ͷ×}Ù|»¾<™·á6ê7¶,LÉ»yr´óØ7õ4ø¾V^²É¬¢sïÀÛÙã ™/³Á˜cV=hW"w³Ëõ¼Â8žFÊú™Ý•l¼SžŸÎ´ml©v|®½Ôn‹¢b¼«]øãž*õmgd7þ™ÊXвkvÈžªÝå@0–ŽoŠÎŠ<”ψJ ÃLÀs O$Î4h#/ÜÔ â>ƒVÎHk޽=ÎPå *¿ ~æüîùÑü}`ž–RsðÎà ëùN+—»ml·¶/ÍD‚Du¯ÑŒ—NGÝÚ%‚ãY 'ÜØ»ÇŒyNjįÅÔäo|Õ«.NáSÙ âù¡ªÛ3‰_x¹òÁû%6.Ÿè5%‰ÊªHÇR—‘Ú4ÖËGˆuè›V1c6µg1Ï–ˆü¯{‘ù”ÙÑ=´ ª"ß¾H -SÂãêÞĵ?æë=_ÕáíÙg!úŒç£ŠŽÓ\Å›²Vn‚¸~8]0.’JUé8ªÛ5YÚIf8?ícð¥AÅä ˜¥ym»²ÓbïcÛD͘kË1mt_¤Ë¨KúŸlwT}0“…,äšõjK±¬'< ÅÉ£¥Ý­æÝ%X1-Él4då ªž#ÇeÙrðÔÛÓdYE Ùý "7ö—»w;Cõû}ÈGê4¹Å=‹;C®Ÿ‹†GÃËoÌÔG ”>ã,bÑ¿ídqÅýŽKæDÅ2jÛøó5= Q I~®mLYq\º<Z\Ä*»F@Iíæ&½ÕĬ>sÜôg¸9ZçD AEK4 C1[µÊ#ïljRÙwÚŸÉòù¨{úBþÜÛ¯ª°sMcDz´Ž¦ÂHö¤…ÌRLRe«À…ðÄ,·Žª¬^iå®=LœñVxŸÑMÍvÉÃ$ ¿ «Rß.ÛMÛO®·_ë–QHóÖÑãû4[sŒô¯á5룧‰±é¯ÔåwkÁ_,càŠ¦ï&ÈÝŒ÷Ãú³‰“ì Bª÷¬ÌWyE±ÏkŠæfœ~W0;´¥h3SûB»u!ñ6§°??sj}²¿Oš’ù›³æôq<>¯3q•ûØÎ½JÉ|Wé½í\è0²‘´å…Þ‡Š)=|øä¿ vOô¤ma¾ñšÔ–w­àõ¯ïG ÎYÆ@™.µ©!;mVñŠÙOì'N*¬¸‰¾y¤¸ ‘|qóåþ«íO™äOìŒ"º¾ìž»Ç~-~ý‡îç[²oEŸ:cÒ;ãè” ùMlúml› ý-\hׂ¹wìþ—ª7Ho¾Ü골Ɗ¾ø‚?½lÚ“%Þ¨rI£I®“Eµ8ð'&Ü;‘+†R€Jö›n^V¿ñ ìxxz躆KãR¯ ý(ÆÜ¿#_¯Uc.Ñ}U//¦UcÿUÄ&Ø=»¬÷¦]Ì¢Xq8›•%‹Sà ²Ô½ù!†v›ûwšnI¿ñ¡–bŽ æˆõ-~˜²GÉù*VfGs6þ%:í;¿äj²SÉUƒsMµ'Ž…óÛzmpTÎq§™e>~R_”A£[tÖGª“Èüx…dV¯†X nå_³v{Gw-©™.¥/ˆ•òc¯Ó´ ã¹ßjv§‘OÎ:Ž<¼PÅzÊßênxjû! ‘; FëW³½ßÙth‚ ¶e€/]X¸ õ•AäÍ+SN…Ëʵ\åÏ\Û‚ÅKi²*—”Ø =)D!YT»d¹õpŒXç> AM !Ì}ËŽd1Y©’$â›ƺ ·%±ØÉÊÝ4Øn}Š KÖ»^¨c°‚ ¤*°…I;›Ò_µÝÄÊ×vyÕ¶X…Óƒª–Ý‹Œöq«^µ´çú/­¶ã¿ëâáñ~©e¤`©{ÓÒÇ8·áÈH”«tñ'FÝ>GV7&î=¸¨ÝG||z>Pú ÿíU~ò¥z0ºP(¹~sšjûÝ.Œ÷‹ØU¥›v$¨:XFû^ý¿‘Õ#*éB4-R°11x11Fo©=>§RC|Áóc>R¸2¡„[1ZFVH–Ç»[+Mh†6Ï] Ù›6MN^ZMÏj2¶êï*Vk…Š@Ø—ò[oþO‰úêSçWø9\È5§¾õåØó‘(Þo,¶pá–—rûXB „2pk3¤¾z¨ÛSQ‰°nƒ’óWµ(Çý/kc¢/Â%²´Á~§û]*ÓÚo½šèé/¦© 40ƒòîOö7õ±ò“¥Bñ=#nR=}ŠÂ>(b6pÀþ¨ž…¯™wrëÃZ¡ *¹á–å Éú''û3WG;“¦çŒ_›ÚZzúõbÓzÛý99®RJ ˆ¬¹YpÒ,¦ñÈU‹;÷t9™dÅËÝBÃASŸso‹½À«ŽŸð×é\{Ò‹ƒp?©¸œô!òô«}{¼ŠPïæƒ9–Ðíg%Z®;¡ðuh“—Ìa'¢žøõL·ª*Œ0G>ùî„ÕLˆ½€í¡P™à¹ùÄŽ,Üâ[åh$Ç&"Î\£‘{ýåyþn} Èo§‹ú´¹[ò³°´]ô[õqÙý¶áqXG='×½÷ìžbã 6«±wþáy^ϳR”0g½‚xsÓ¾¸åVNÉPû?J.ÐçnFyÑS+Ô‰ 4IBÛªðR9|Š ­;û™ò[60Î$m¤Í@‡ˆ‡KѼ}ãÏôv+ÖÄoà|\œÒTÂE}ÁÁß(Êáì:SÙZâ?u‹×‡ï_ñu3¡k0[àìf1kÅÜ=:Pµ™ù Gg…gGRgªãÁC€ï©¬šFg«nv³Ãµi7Ë0ב‚/óZ‰à#ìq¶HËÅ#G•R.Égǧû°/+Å”Üñƒ€Éu²/‡V¿9F]̧K°2M~ ”¯z±”Š˜1Ù] š$‹€œ…pGíL»ž)SzcX~{¢ãCÓÕ— ­vdΛBw·ákÕ­•n_u¸|#1ƒm …w›SggkWMð!_ë_J†2Â_]ÊÎKävBÈw†°¹c÷²ÉÓ£ãG0ºsn£þÔ\ñ.æmݘõ]‚ ¡½·°&®\v Š.¡2ØM¯¤Í[÷œR9“g•¢¨…Vüyºìß>nks+øž.BvЇ7]§9µ&ô-ѤüÌŒÔ„Ž¢ÆâÞèõÍ7DE–Ú΃‚¿0¦Ï÷l!ÿ³Sü ÞZ¹Gü׊ERâÂJ@BIQÐ(’lމÕÀÃý]l;sü~îpZæj8÷7iëÚ#]GñMÈ;I&@èfU&bÂw‡Zý8¿º½ôÊ`À…(ÂטޤZÃôƒ „HüÐØÇãïõõL1Ñ Lk¢­¿Î[–êË=!ûKýƒ—]½LåM¡ÙýÛO“ò«ç©rºCBž€Ð;íDÕk]ÏÉ™£.õðïÈ¢,eÛ»—»þ#)F{u„¸˜vq'ñn§rýwOéÞÃøwéïÉïí¶šòñ­ýÉ®¿ÛzÌ–mìËmUÔÛLÁJ%‚)ØdÁ_.â‡PíkMy :Œ>DÎ[EÄèçáãPä½ZÏâ<š¯çv£ÖÄ¿ú5‘òœ g—5kâ:zŸ?5YêQVš«`óH}ÖOªúYW]ª©EÊÓ'm¡:l9õt§.fÑtÅ52!§>ÎÆb•¹É™¦Q=Q¥é˜¡¢rM;œBß=$ÀöÚúôþ_­)ü•gÔÝ7+*åÜœ÷|wKè‰rQ så©_8ýQE†–Ð1+·Ó„òd€4ˆã yÑ|ö¾6®®Óî„7±V5¼ÂùUµÃÔã³2„ÜŸÄzjd.ªÉb¶wˆRˆìƒ@VJŽP2KÅ‹6øŽÔeã(ê+cÔQ˜hîö:ªxX†Ž£¾¼B´ ¡M L¿Á*QÚ“Kgcø`Ó×ë2¤ X1r½µ6¦)l: x}Ì hs&j*ÎUS:öi‰/YÇdÎý–ˆ È"¾J™.Õ|déqåO#.öZd{‰òÉÃ~Ùyéùs§Þá)Ck~n¯×_¦ï™œªŸÞùê,âØæVØÁûÌG?™¾Ž=×ñâÕ úïÌãÆÖ<Ã^%°f—×ÏÃ¥¹úbÈ\ "$0!>ré¿×¢ éëldÉ4WͳíOnÇ/Q|°—±Œ»–¹Âc–Š8S{öËy{|µÚŠA®U|š;&êX5Ĥ}™vH±µÚ–[H"’ɶ†4?§0Òþî7ÌÎÛÉWøÙHÔm/-æéiJ_š.Æn—[ Z_.´TýCìc©ˆâ¦gI:J‰ÚŒD ÷ØæžVÖ<ÈŠzO2\S\œøX¿÷ˆ‚uE`›™Pþ£_φòá—4¿>L«•éáÜ^³uÿeoWÉ+cåùÀu\ÐøfÉVæVއ†jŠ—’ €ƒt€ ƒ¾v $ÖOw£ö‹¶£@²MI¦º¬k§íBç?‚ý±Ù²ÑWWÇSg>Κ¯T>¥5Ò–¼•‘m@£éEUs÷W³oE}ïÁäj,Ìký9훤¼¯M_33)f¹CN6ƒy ÊV·Ðø®t¡HÇÈO›Ð©Ü·Bg¶Ÿ;߼Æ?—Âf¡¥va7Úä:q ÛAIãXSG…eÎ|TsZLv2âáa*>f‹ˆ–ˆµ™˜’T ÜKó@«—œ}¯“æù‡ø¥vw­±¨eÝßNM»#®ò)éèÀæB½ôú6nÿ{ê9GYyåXIµéílް‘`\É ï¹nb·]A¹mX|ÛW–´ªh ¢-ºÕ·“g™š4áæ$]X®mÊ[‡çõÑ#¥¢eÛ;-Ì )kF 颯NfØŽÃX·y4kXÁvBHq~™¥K¦ìæOåŒ ÙÏÔì?‰ÐÖÃ¥äÎU-Â6‡)z?Zlòlp:QÇï6§o¥{ÅCOF.–O3>H÷³t§²V›a[iëèU"Y<°ìi׫’·¹u¹Y« RÅZ…ª‡Pf wêŸÁ÷Çùàºù„¥çÛ÷Ѵ¿Ÿ·HH6O~œÕa^=zI=0oΊ Ûçû÷„,™6É«a,Gp>£@?É»š]­ÀÁá¶^KU;Êø:öÓr¶ß}ï¸Ö<œ5v÷¾]Sòû{»‹éï¯~;±Üÿk´y-’CO+püø?PÏ’&#‰*Ìó ÓUÙ©£)Â5•èjQ˜`ÞSý}V“té'ÝëöW*ò:Œûß`¬ru&Wäæþïm/S­i{w7”›A 9ý­•a‘WÏâ‹I¡ŸÛ‹ 6·ë+Gã¢61†(¥d€†Æµa´ ®]yÞ¹»ð—ãVÃYe¡—3wW^ûÅœ„Íö–°ôöØ3œ‚qç1.±½\0n†uÅ ™'­5šÚ÷øÜŒ6*_·üØ©®PhÐÏîôS½Øo›Ú›ó.ú¶]e*ÇEÛ,›¹)AÆíÔSætS¬±ÒÙjЧM{§ˆáÊKIÓÓ9"ú1]‚Qpyß{‚™Ü  ¤›ö'›{c›Ã¼ÄÞxeͯÄgqÓþ}Nß®‡ï^û>Vt.*ÿKÞÙ!ÏV½j/)0úXvhP½¹€Éwκˤ¬Ü­,×V{ò ïëoÉ:ßòq^*Ž?3x }‹ 1Ñ8i¥IÙìö2ý6*¹Ùý…Ñ·î½`°>ªÝm·ý;ÖÝUûÞ³Õáí´õj¤½ýK™yö¹smêRW´ŠÆ•2Ý\ÙÐÙK,P¾b|^ßµ ÊrµØ*1äc!´b¹egÉu½¹­Ylɺe. Ä/ì)È%ªâoM34´X•×FU&¢õRÉI71‰¸sž#ªêP 4uÉ'K¸ýèé‰ÈÕÚ†$j*”–Šz¯L’_~¾ºRºe†t6ÁÿkØ´ÿDÚòÞßziôÐ:»^ýçNÃõf-hx ˜í:û[[IŠÚvaºIš@é/Ów°§p£÷1’Õê…C´{Ÿ¢[<°F½?!ßb`¹8ÈX ¹í!×Û©÷x…µ°sGKÑAã‹_U?ŽŸ»C’rspܾ?³ín^%Ÿ%P0]ŒAÔ³’q7¨åE[ë¤+ge ¼‘Ž+Ë;(5|ÙmçÃS‹ ¦"ë“ç½VK¥JñÛeíÄÉg?æëTæ·¾<Ê»Rïñ—è¡aê›ðpéîŸøwíûúŸô—†:“k®Ïî\Ýüep©)¦ÿ8TqÅ¥ààîþ]=™… M'b¼m6Ð<ã½§®øÐqlõî8ó…4¶"(“Q {d˜öe ý‰¾JÙóf-?ƒ.«ÅS9X·ã^DOOK‡UÎTË2\¨g°MÿWÍöst}MÃUÏÅ §»ŽŽ;7ÕWûgg·ëŠ17˜?k(¨îãÁÕ_•Kfmp·Ð3®·EWO£ÈiC•§÷ôýæNÝtúW’±3ÁáyÌo‡ãBÌÇ|pûX\ºÊ4s»ó…Â:Ÿ¦¯Tõ»<—E%}¾qúÚâÙÑM!° ÓwÝx2©æ‰h`Z¯Ž Ü¢®(E.EjhØ}>¿ª÷yþý<¯}ŠÛ„~êrl¦>6Kï;éòC\Âp )æ¬<œ‹Ø§7ZÑ œØí†' ŠÔ)fj¡:q—s L¤4éÃtDwƒ;ªýÇÃö`’_)ªØJEæ4z£/@¬ã= <Ô:û÷uù4´¸uÔdo·ÂÊ ¢ºTÆRkÔ²<ö¢=éË>-½ýñÛY§ y‰&‚îz|ÈÕ ŠŸ':”¾jų„dFeÜÍ uPÇÖÐ#2Øîe&tŸ/-‘Z¯·Z¯– ¥…µ¾ 0j…§qÅ“‰Þy¢=^KC‹Z©13°Í<ÖÆÓÏ7%=bÙµq‘ƒ4y+ÕÀ4E:6¬*{GŒšL²$±Q1Ô ¢BrD‚8¯øÅ2œÌ2W„1ßüXA s³sÔfB‡Œ¹bŠÕgS™Ë‘áf©5ª³_‡’.¦Ú¿êªÊXoÏ0×fÝ¥²r–ë2¡ÿ.W3™{Ãmˆš9EpC(kH6Õ¸‰s2_ò5ˆÃq’“Oic*Ú™‚t·ì÷×ôô;âƒÁŸ:¾3<:^aùñÙ‹]Z)2óìó_v7ý"FÔ/a»º?ÎH4h†^oKÕq¨[ò»î°o³èrÉÞ]0þ/‹ûxD¥vb&¼ä‘T¤²7y S#÷éhàð3‘ÅÝ¿?"¥‘JìÊÞs+ûØJ ØïÄøÏö— Ëp½Ä»­`qu‘vZnÊèÒ¶ãÿØR–$b¥‡­98Óª¹=v>Âh£fèÊg¸0Ž.BTMÊ+/ÓÄAÕ±O‹ WóaÀ\Æ÷SVkœ¿>îÇ£Ô-•?›ç¯6ž)ðê¶•]þýÓh|º©ÃŽ§Øˆa2ŒF€èëe„í3NXéß狤 +DæîWÀpBŽ®VŒõÐr®*ãæ¢—(Bø“3\'OgI¬y¼0èôzÍz48d==þ³gƒ™Ëú5¶¦¶c;¯ªó¿\ûFk úÒáȲ¶ò‰ÛËëéƒ*àÆìrí-ÀÑ0*Á些EJ²+I±dí”^`"†Z„ÓñjFk6¼jÆ:}%´çæ‡wGýÒ•¸°@¨ZMÏC•Ûõƒ’  0”Ö;53¾íb—)žùÉô— ¬o}È(ž³AåyjƚȨûhò¦<Ø™¼áÉã®Ï3)ÒÙ»œ¨}šï|±FØ‚|eq¨·æ†®_Fo–Oùëê¹Y¿˜ùL—¾IÎù”± ç›xH·CV¼^eœu§Õ$=’=çoušáÓ¸žõ•1-B3¥ƒ=nX‚_Ά¹6Ã9ž€úFÒŒ î§ Åž§†˜Û‰W u .¹Úè Ý4Vš9í^¬lбZæŠy$b-€ e ¢;q¯P鮃9r죕”†{Åg® ¼¼ÊŒG ôŒ\ÁúãYfª”2Ì ÓAÆœ¥†{NæI½Ö[Ñ’+obt&P» =ë6²|ÿÃù°¬C.… ]8 ¾÷¡7ú†ÐÚ\(Ž»šê“¥uX]¸›ó˜Æ÷RW>ìD99­êO݆ºßg’ŽërWÀ^Und|lÔ7n¨Gg,Äg!Z'(T¿"w¼¼ñO!°ˆ­*ш‚‹D^i[Wã«gã\la©IS—Î’¹’€ctiYO…CÃH2¦£¼Ÿ3E‡\_íÓ²=^{¥x%Byä>o•ËÊMlî½jiÏÊ#<ºý´;îé&¿—ëH.¦W"À>öT Ƭc?jC“«Q÷Y‚z“øÀUY³D8ÚðÉÝÁuæF³s˜gã©É^ÇË¢]‚’EžÕ“ÈY@bi<`üçAöVßpå£"SŒ"ÍôÚOMÜ-! ÒŽÔ *ë¢XctÑ2é …¤ñC–Ë1Ã&†ˆ‹¦ž4AbTcÈ´(˜ŽÃ›ÐzÀ¬ÇÖp¨t°ì F³®[Ý]\ßþàvÎ^§³?Å/Hí4óû< Þ ‚:½Ð¼ÙË®×ðs+k9ò™Ïº\%÷-ÄSÕ¾Ôt.ºïžaï _%4›x(³]n¶”…iÖ#nÝ®%¤”DÔV—ä²Wœ'ˆŽ%8Th Ž`¼f:cT¼Â0ª- ŽÝê²N›jR%Jd] ÛOAÒyÅ3Üp›–¬ôGÂÈŸôΘÉO´\]y›6yKêöáñÒ4”#¨+ëá ¥l‡x‡Œ×µàÖ°UÑ0ím¯µÅæÔ…(Zµr‡2_˜ïâ½e¼Ô½røÛ / oˆ`†lÍí¯-±?qóùÁPŽ¿ßGK“±õÛèŸ?æýì14Ûu`Ìr5—FgÇ«óüœ[¾¯‹†å;v;CF€ÕÚSÑ g¹??ª^‰7‹k¸roÍ-ü|”îôþmZ!·~ë;/’§–™ÞG1ÖÖ{‘ñfu÷yŒŸt³Æi>ØäCC¤rãy~ž ÏÒè¸íån¯£©Â«i©+ªÐ‰¯Þ&·©êó°´+³÷™fj°Ð½îUßMr]÷VèÆ² |É#.ýô¨ó]Y¡ŒxT2Pn~Ìí5Ð%x…挻ô˜i‚´×¬Â‘£éyYç, [¦M‹à i ›°¤¢JýŒãp@;žÎ9¹ê™*h¡ægS9"Vôì¡èè(€ÅkWú$Ǫ¨]—~’ãábPGPçÊNã ;ð—Uß©%Rdcí×ë×ò<ˆyú”(ÙÛS’änWº$³'£až[ëÇY Ý§í`•þnô•°êb%“Pëä¶yA+ß«C2ïÙÆL17í†/j>;^L>Ã7D0§ì“ï†sš§Ú5àíÕªAø‘TÍŽZ‚gòÄ;kTÖñÿ½†ôh=ËKXMuØK¯2LSúåºhÛ:6w¦»h‹¹y;šCжYQÎO·Ä‘ü9g`ÓјÏ[½¥9Š;Ü„™zþßÏ×$n#¹Ÿ©Õ£†s[p®y‹3®>õþd™åcéñäÈÎGIž†¨ª$Û_fÁF%+;„ÄïZä¶V©ÊèÙ1>гÐÁ¶Ñ ‘  4°ƒf¤kD9÷yÕx•LÕxyÆž`ë /˜ p éË Ñ é+ì¼!ì,½N£˜â¥–¬“å/ã÷g—Dµ‘dˆ…€ð2òº¾g.“~MÌÅ\ÉÏ“‹ÔÕEbÆŒ¼¼l•¦?-E]'ü³¹Âæ=ö%óý1}v{xr1©S ùß«¡ât¯çº4Ù"™Ò©m°óÁdÌoÚ©?ä.xµrº`·ƒ xÁ±#c-U*l¯/t~:ÑœdM¾[Ò¿÷óÕ&F»Ô0±Ô¦¹_¾sTó,aüˆ¦A\íA@ŽO=Õ€ƒ“[—…¼µ¾¶›éÙÊã ¨³³ûY»qY51ï:\ØŒ è¹<¼‰ÿZX kŽ-.vüBùMʯ„vuÖ2Í·/§ã#ƒ§9ô6¼[þ3¬îÌäîçi¢óÏåxí¦ã½b-Ï©%-0Ì](ª‡?Iö€“ü R£š‰æŠ÷ déÔ®[ìÖ÷ÛŸ‚3/îÝú;7÷ÿ6þAã×v^ãºáª~%|¸Bв wm% ?¯?ÀÅ–Æެt¦Qû#§gˆˆ0ˆÚ|  ^  ÍŒ‡˜?Çw”…Z7&ÌR9NŽ#á3-ï´#Nç ý—ë$¥b-‰6_›‰Å,ò!ÔÖ¦!à¯ÊÎPcæu3XúE@ô©pûrXÙ’Að‡bŒÕ'À|pœ$—xX¹Yš®óñÌZi \?ædÂÂQK>ãX]¸Åä5Ë•uR¿i‹Ç¼‰ü9¾Yk­9ïøQe/ŒgŽþ#˜‚KFrêd3ÍÑ׎î€?a•yðIûôbxí‹Ì:H«ßà–É(…R*lÊnm™(𢨝mDÃN­´Ôaˆ¥ 4‘9§g’Ó(_Y)S’hí†Ãñ²½„îW€ œ‰Op;–µ“èéÄê¯Àø…ÒÙ5.:ËÈ$I)úê'½æfF{©à?ùƒ4~ðsrÛøä+ŠQë,]X »)µFÈTtb¬+)¯xMƒŸ¤ÔEë놓ÕC\·´O%lšÙ³Zhéí>ÆH™zd£2̱\­??K‡_,ÅvŠè"ü²lf‹¾JjK}+ø“-ÄŠV30s‰†.²Ð´g gßZì;ˆ=\›ö)ÊõNø‡æìÄ\¨Ð,ŠhËvRל‘ ~¤Çí_(½ÐZ ‘ \õD—$Èç ”¹'óšs'Q†Ox$íqÐ&€ªeå´1\Œt޶Æ×“;Ûò+y”r—”So]«ë›™óXC¸à¦ñ£Ë>„‹ïk3ÑgÜæÍ®e›Í€2½* ³oE£Ë(J•êW\#6‡¼±³4Iˆ‡ˆ|EVcd€;^L’EÇÙ|f#.¦CÐ"ù$\ö3…@ $A\ùL6ˆd¾«Ål‡"zC ;/ª†CFRÕE=žTú™“”" 7þX¾ÔЫœéf„%ñOª¬7ª{‡Ìäzì½% ‰«ç2ðv ]6#%ŒŸÀMÜí8md9nK}êÚÚpfÜÎØWÞÒД» †žÒ·¯7ù*xÙŠÖ,`jo8÷kóý˜X `÷ËÀu6nËbÜh­JeÏ“žC•²´¬1>˜éʆÃêå0b€ €@h¯xïe=TFè±¹ïõ›·»|9ýe>µñð+‹1JÚµváØdõÑè-`§ž×¾Ãÿ]¼µJ%Hïµ=öšQ¨ÆVÚdt4~¼­D!*À1Ð;ì9%ls¼ëû…ìÑ-o˜ëGîWÁë'¥>{CÒKáW‡Ê‘D ™fvO}on´ä2àF£+ªc)ç‹N¿ëeU½ƒ*Ù–¨B‚ Ñ ä˜Á÷ã!S·œ ™O\ÉcG5”NÊÐip]SiZEB9ΛŒ¥ª‰ bIÌu°Ù€  `˜×>\¢‚,h°øØ=*I®ÆšxØJdôþca‰& δÎjxÀžF3ãlz´ð8~ítuããöùO°þ¬€M‹¡R¤æŠüªfþäæÎ's³”VŽVÕ-«âµÏ¾PÕÈ““¢þÉ=Lë•’jx× #9Š0Â_1Æ,b¾Œ¸]C4§’8w1z‰ü¢t%±7pËUë—ê#í«sQÇcI¶nº–AéÅ#(·´!u)ïh”­b1‰95C€ €cäS-|†9ËŒjh‡óÌ;[4¢ x¼ò[¯ZþûDOu›ã×£pcÖûƒ6S%eäÈ7Y§ÔïöøÆþ4ÍNÜë±míg=å#SN·[ 4mÉ7ýþŽ¡ô(ý4î’ú«UsÅÕ0I†Îò÷Ïž>j_^Ÿ[«è_¸?7QÉL.æ-£#E;&-<Ýt\ áãd´9 ˜*H«Ïq f†;w3JÒA‹cÛÉùʤéì9/—ôE´ý3d ±Ý-ˆ é Y… ©*T±ÐÐ.l|΂2¶Y[jÕùI—èo·w›ÒNÝ^¢ŒSjôÒŒ’ÍDbók~ÍÑúÊö9A*•½+ÿ⻓ˆ‰wªÄ²èV¼ÜÏÇûÝö\fD‹(F\ g-r5¨¯¹Â8…ætr[ßsë"!Ë.iflÉ+UÏAo5RÑœ¢—¡ñÈÔW/?l]Ov]s_4žÓºªÚMf[UY‡™ ñ@@x‚X(ºÉPÉ´W>¨™û‹xdŠËEÚÂ~Z¼eµÙÒlMTë2™Ï^‡…¹i5 à>ò=YÒŽ ¢u1_Æ_Õ0o,{ŠÎÍíLå`<:~0Ne;Óòc}²}ö™4|ŒyJs8­æÖ°>®®UÓ̤B‡É¤Í#P ÎÚžM¬.T8öt·OAzyŠótÙ-g,õ{ ô,aÆlÁ°a‚Ê¥V…¶Þ;Ð<$®Dc² 2†¾›!•QxòoVŽŒaUë!lèäN`j‡x¾V´ÂˆóÉ…ï-¿_{žLvM¶²>yϸi­>GšµÌéƒoxßÝéÆêˆç³CÈ*'*ç¿oÎõððzz8ç®òÿIí œZrU´zòÆ£JçÔ‡q•T«ÍH„²W£gÖI5"—|LmžÿLö4ºÍ­V詎ÇÕ_¡kÇjÁIƵ•µ2—®ùHRØßɋݟ=˜‚ºS~éAÉ’ž@uE¬wµ”¹(ñJVØ­«&Â;Y˜ôB´©N§Y—®+1f¦[7 êçSæ‘òÅX/k ¶¢÷ ŠTºë"Îr\³ˆ˜À 2 Œ¨°ß±sö“‰ ·­CÈb…‚¾aÏÔª¬Á $lüèÍÛ↙⚊.§žcKLn5…¨qh˜| ¶ÿ/EVd±dS‚´R>²Ö_wÑâÇèFå~¢Ö3LFûO³Öø=IžIŸË®Ä6”¢bsQSdzþ\Öq> oÎãV±£þª•ZÑélè.kÇ!à-˜ eZtfòöp-âÒYi‘K•ˆÌ^#uwµ&¬’äfê²x!S-¼¦‰}yî.ÇÏvß#lÌ(KP¹&¨l»RN”̚7Ë¡á5Sè&AS¢Ùx鯙* [òV­mªysÏ/Ì”rgÁزÂ꿺žbÍ‚¡ Æ1%å lL“(ùŒö|IìÒXVê¼#›5Û96«ÕÕ¹vEVík4+ójrÉɤ[¨fÎAeÓ1Xú›ŒªÊ'Ÿ"ƒ‰õª8¢ë„užnïQ§Ö'3‘ð] 5^Ge3˜|â}=ÏéžV)ÄUo¶ð'¯Î)þ4HM̬¯Û"Jå¬*=´AÄb—$¯íf%”E‹°ç’ÓÆé׆¯§ž›¹¢Þ`w¹o3 ç,±½‚ç!•^GîfÙÇ톄¾š“Ž€œù6^-6m,3¤¯(ÅŽ™Žš.â –Ûô”¶Z»ŒƒUyÏ•žÍa4c—åûÖ§¯*Òľ×<;_¡ðüJ¿·Ùúu¿ðQô¼ ‰å‰óÿÈÞ×K^¶OAÊ4Ç,<‚z­lš¬÷è'4‹1•Þ˜LKzjkÛ,–ˆd7á憒 a)´nbÀÓËPnŸ±P{¹³ò«l3!}mÈc¡¼›G0„qÜ鑼¥°14¼·%eÎÓ/~*É  ½´* ÷S5M%EƒÞ ½úÁ£1¦:‡íˆ;M|5|ÏSã"h©}Ûù¦*T²“—ÈáyU‰ƒã²ÈŸ3«ºª€æù³…ˆŸ­ëmgfçŸÅŽWøl°ÌþÆ;®2áF/À&xl¨TðϾ²?’Îüi¤Àl×ë°LŽmÒ/ÞK›~<?Ku¾½VLËA­ÊÑ®B¶ùef¢2ZWX"uÍt=˜Žºj¡jé{†·ùWHú:£í<ób’GÁ·²<¿Mê ;}oµ5õÒXŸkF¶ý;Î+©ã GjÔ`x#{C˜]ï êR_ÃvIJùmQOÌPí`ÊÖ¯iaéÎD]LHAÊèçP‡3“âæ×]Ìh,º.> J®<4rA™æš™ü,ú¬pÍåŠÜaCU±%†‹…,G:b¨°æ³ÒŽ¥–Z•¥Ð^®iŒzùdË¢‰ï@Ç[ÞNPɾ7öú9pëòÖ¦¢ª©Ï²¹e›æœoæÊÓÝÓÿÛéGêO:ß™µ-÷ü0¤Â4j[ý–éó®” ºjxÉê^1¦¬íйs”`ú×8t®†Z…bd„2m`,¶Rʆ81¬þ3\¦>‹Ge]1n( &]h† ¹Šá¾ÌWò~AÂ@D0„²´£a(v‚ertFM7Ú߯0˜P?@B¢èotºÁJMôƒ©Æ‰¿°ÿbcë‡wDKÿ›ðqñôñ27½ÁÝ™êgÕmí›Sµäª•IrÔó‘Zv„®®ë´¯ÜGh“$%#!µAPŸk`û{ï[@ª}Õ¬4>ÂV5ü1òºbFwŽ5Dw/éïaÎjÏ\LÒPè ¸SǺū5†ÒÙIÜýÌL¥f퇸Îdzz†üO9î€A•”ê,Yïì_zö™Z`)Þ_ùtio)yêDn«ŒÞjþìt]ËóM+ÊŒäyÅHøyƒÕX&v$“#¡±æŽc¤œÎÊÆ]Ó«.×ÈǶkQ¦jmÒ]0ë‰Føc[HzøçºI…[·2-ò]”¡ã®fb÷ð e„¬ cç|\ã–oêÌÝÛ[7×9ø¿˜€cÌ#]÷¹õh|4‹½E›ÆxÐu_û$}ò¶0lÛU?yTùÄ~àm·)ï[‘ïILl—ÏàéÍ6¤Q@à·ÈÃëM@nÏ{¬°ne»N3Qk¯@¯û^»dÁÖÛ3-¹òV:z›3¶¥\KôøZ½5úuÄ9ž°Ä¡EˆkÏRíPÍ.–&(‹¢×WÑ'j<ÜÜ™ŽcÓ;7„'KZê@W¤ ÆÛ&d¦Äv°œ§Ï“P½o¦¬M8±,|5î{ÝQÇ¢¯øCÊÈÛ»àù>$wcA= {|¾®…éæX™0:6õ¥ÖÁmÑ|ÄÒëw?Îk <"aÑû­,Ô#á(Ð/ŒÁŠi¿ÄÌ1î@2àbêåiNN¥ÎŸ%ÜÍ~rKFÔC™§áÜ´ÍœŒ¬ ÉlöýÂ~6& _CøÒz|DŸÓå>ùòþ·ï^ØçÖýë¿Qù/Àså²-)âTc Á±L6Ý  »Ëä³r³v³$8cUÙsC/Ì%ø Ré—1i2Qz!<¼ŒbjU¡ÉDë_’¸Ø¦1zý2H\¨ÈõªV ûÂwÛæ¬5Ø)|R{õêpò<îUøÑ]üÁ>-[¸ãø.95•ëõŒ(©úœ}b—Leeгö½&ºjMk8~KsÊÌäuV&kÈ{Õ{äë>Èî[ÏØä]Gîh‹—MDe…™«IM"¶ä·s D=5å‰1WeæEfmŽÿp?¤©.G‚ûTÙRCÏï8¾6Ë ŒªÖ«,툔›Ìf»È¥l]‹îLÍ(D»Bå/Jâ2ae"¬^¼ ±øSŽÅsBì9²sUbrs± lIaâÐo ¥YV®ó•ªaÎ?)‚Å þm÷´+[ÆUÕ·¼Vdä".2zNXg ¼& Üò~>Œ×ë¬ÁÞÈ!Ë݈£ÖkÉÛßI‚ÐO@“³’ºÖÏ‹¥Ôs_¾J#Ugʇ¬\„» %L€õbLêy놜=XÐ:Ò )NfÅÞ•¼k©{D§½éó÷5+ƒ¨{þ‡ÝØMKÐïôÿsÈu Þ§âí;œ¦™|㤠† >ƒºtëbüå–ÛÉ‚Ž.nÍދǨM:-0'Î=>ؽêtk§K¼¯\íÖ[J¢B¸Pæim@,^Üœ»1d îgc~¯]^¶”ÇìçŽ#¬ý½—E.õŠ‚`ÛP¼_^ó)X‰&ÐXb½SÞõŸvùª¾@CØb—Ì'Iæ4b«,ã×€3Ô”4{Y¥lE -AÛ!)e‘aG"ôsÐÏy¯)£Y?Í›Á•w#ýÅ þ{Z:>gÙúÃ-nfnøS Ù Õ¯ç>mòx¶s-R%à¦ÒÀ±ü¢ ;Ã{­W„ÌÚCbòæôÁŠøXSi®o^ñ+̼o—Ú…À©¾+Ûètãè¾êËzvÈ´’ë:SA¢ªz×A« °µ³}|óYCq‘ê"†Î —”ã1L3ôÌÞîi‘$‡IbíªQ-¨ªP¥ ¤ëæÑUòcžºw6"‡ƒ€ñ[Õ°`Ì]ñpOPmAåö•©Î†ž|dÏ[Fç;œ¾æ’póüìš«§Yð—Dª’àXùEæ'µöõÛ<ôä¾&†`›z½®Ñð èÍ6³óÊÁw”¦hoóÐÙ8d×jÛ2Þî[ðì½=*§·M)s­+ÍPgàÝFå6Aø~^jÍ¡j"Ê‚çZxêd½CqA/LçÅÒœå¢4aûU‡õù WëiŸ!3g<ÕLÄi?.o9ŠPm6Ðn»äãâðh_ú4õ ½ó%«Ìç9×sÅÇý…­Ì#ýd1¨‰³å±ìªà¶:ƒ¼~ïbn®ÅüÓ²®ïå«›Kή¢†Õq®‹i ~k\£›É¶ÔÝ¥-³ ÌÌVö‰¡¥ýúyÊçæ©f½Q@]%ªÖL·Z1dìX# €AjÅU†P†âÊyãg¶`m:l5Û¥¨»ÚŠsØ)|X©¤CÒÑ[°o•ø«tý4$×ÎR®úJƒ`®GÈòRîðjÿKEÝĬ…ÇX‡'CaljXR½È¨-zü-àV bÄÌý Ÿ-=.j,NÑÌzÖd¦`è˜á¯ÖªG Ĩ¿·òšùáè;%‚´ÜW~j×¥Í;l\?hÇ9YV¦*.ð CI•Òò×¹ôt6ÏWËkÏÎì´ô¯-H´Ø eœ”?» –£¢ñHáøÑS¯Ñú>G+û•ü­YÐ-‹<Êakš=~…ùu%DlÒý“4LE®£ôÎÃVW yµÁ#PøÆäÁÑ6O#§HÈù-eåQ£æÉYìt¤ö”õö¼i…ìŽ¡ÕÆmÿL¥áàw¥cÊÖU¬Õïk+G@x›Þ´t{:yÎøc2ý¶{ZPÓxèYòóˆLáùÀ 5椠fUñ|ÙÏZí`í#ÄêöϺ9Ǫnî²©Š‰çÌQèjF*oÊ‘U°w­l žÔ7*ùY??Ùãs2‡PX2G¯ÑYÓóm˜=ÞÒ¬íœRk…F>ÙÃ)¼Ó­„r8=—º9Q0gt&Œ ®Š•ß›®Q Äˆºyÿ‡Ì¡p›7îßó”³31BíWÕm_¿Ûnï3r «Ó*Ú[¨¦Ë|»ßí.XÕ3š0;ÿ‡Oé! ‰ë»T0ù­µ@@Ô ‰¬E×`ÆcDá×35Mø¾ž5ñö8Ï:ó RÇhÇݨÜ^á’=aÎór,âúÂmŸ2ÝyVRò6rë>(áþ¦r–xÍuð,TÔUk¼Œ×\b{rVõêFúè¡ù°Ib³«µi3š[ I|Ò­bY9ÔsÔÅÎÍ»¢0/ÔUz²{n²Ü£U7Øk€‰%‡dÍiíüòƒ8.+œ”wÆaô³1`BtUBAð)„Ë ‰”¦#l²ŽRw41¦‚ÛU…¦­4T+bÝìæñs3ÜÔ6D4Ôéê2`¸÷Vb×ã@7o*‹ M¡%æˆn„SQv:dRŸß\IÊ‘G…}ž+mGänÓ &äÙU¹pz[K@í_¹du<Ãlæy<—ÓଇD;Z¬ŸÚ•å™ €zbÑ©FªÍÓaóv#»©õs9¬Ü@½ú½8ƒ³B¦¯Íh‘ÀhÅžã§A@m:<áØÁe¿`ƒC~ßi å€”e Ønì­µÞŠ`67RÙÅ›‰ÇÊù~ž‡±Lr±Ëê­êhjù :_QßÐç†ü0‡ðõwý+\ŸG‚ÓÀ+áâN_*….öZ;2cÑ¥®ã¢g‡ݳ²2­÷-/!=¤qÔ¿ÑϤ妦¸âàÌR¼Ëº>Ķ¤ÍЋÙÛi™!÷¸Çp•H•àÌ :ú°aò+99ˆ>Õ@Â^¶Ã‘‘\£d^ìþVþûWf:Pôχ–à<»O™OõÌt÷÷ØÞS‹~>Å?«Ðä5/Ðç8æ<ö5ö4ö†ú߇sÆYåg5˜kgßæâôFúسes؉њôÙÕcåÝ™êGæÌ§wì!LÖ"btÍAº[ú7öv5Ob{æÉTØÍÂÃ]‘¥¥^”7[çvÇŸ űywæçE‡ °%HsŒ™ó}Z Îê. KÝYJâýE¿.RTÐOÏ+¤µJ+‡ªçìâ–¹–JÓÃmÖNR¾BÈÚÊãã'u®Á:­ê¯ÕÂñj†±1bÍ Å7^¹{›!p®Ë§Ž‹ŽÍ[ÈšuÅ ³Zù#î°X4iùš—¹Uzs³·[³y%«º¯iïÃ5jòÀ·ïºb­„˜’|¹ÒKê ¶Þ†&);4uy«k„é\sí2œ—ù¦ý²«ça³–ˆTg±®i-L2cÔVÇÕ슅¥µ`]©HXçîʧ…1Üärý*Š3ûÂ2ÛÐMÖŠÚÆÊË»Vj°íVòÆ?²ÿ.@\ýÒûL"ªAsJbÎ bqaÄ ª‡“<·îû~z‚OšüQ ~SŽ¼ÖŠM_ží’æ<(®2ž‰­“›×åYzЃàYx¢Sá¤vMi!Öñ·E[$ô-:pNÜ»úFH$HÝŠ²:pZU™Yç»Y‘üȆȋ¦ÓÆ¿i <Ód&ÖÑKnÚtZÏýÉH"ÊË?ó]V¶:¨égÚCó„ksÑÚgÇÓ3§—̯«’CD¹îwY.ô‰Ù¼·XuîP×$gA 1 I»÷F•2J_â˜î º Äâ»y•¸ 䥮ªÓŽ~M6(ÏHŽ~¥§• êÂå«VOG³1½_u9Î8"и>“J2Hmþ!³O«‡ôjgê#3ÕBv¯ÏkWIVO‘]{¯¼—±(´]Æ…‘ÿÇö[òUd:PëÍJ$ù˜‰,žî{"ýÏ%·»…®w)ŸçÑw-u:–uHŠä±wÝÇàÄ÷vŠ—e†¬¿ö ¯y0:+y:¨ß.ýFY˜<,Èš¶ÎÞqf\DÊ5 žÌ›yxÒ¸e£Ö¢dYTEÀ6w“¸9¬“Ê…M& ]¬Ž®Œœ)wß–ý|­<MX²yWÖî7°+þŒ’ÔéìÜoºÌ§…Øð:qÓ„Ãß{ÌwCsñDèêt÷>o_ä›åÊ×hdq^ßr³‹îJ¡ïP¾55Týâ<¤[±®#ø êi?=9Þl¬M e&ÝŸèIœÂ¯¤-Õº…N\DrD»r +<U‰¶«>^<‘_Pk¦&j9à?ÎãOÇ‹©z"‰ ¿u76ÒœÆz¡¸ej–ŸûŽî?0D7ÃîìÔT®"t“ã _ÒTáÕ·°÷ûD6<_r—!b¿œì,>£ƒÿ)âôižrL»‰~²¬áH¾Y/«Å^ ŽpËpåã]mζº–îò”²¸ËçY#ÕQ™o×Þ¯Ýç¥b9&/Ðî/¿§×”³E¬lŠ`&AÑÊš*½qC…&.žFgѤ',ïŠþw/“ºÇdÌó¬ÿq»É5¿Å_Õ¦Ú‰3=¯34ÿNÿcæ|¹”:Kö|õµÇ?Ù ±i'ù–ÁRê~m °´btåÿµìJó»«ó° ¤¼«·ÖG)/¥Y* ïß%®+ˆvL{GÊÂV©Í¼åÓ=Ëe­žÔ™ >#;TËÏQ!,ÆÒLT`3Ì)N‚Õêé#† op°bW}ò¾BÓ¶NftuÆJyf¢Rš‡U?—‘50Ô=aV’´­)µ?­û¸<Â+éø-·-°‡G8»0½.=x)¿ë8ŽÒ2äüØ©ù¢XÀÁ1‹{Kn4=M?S³¤ÙUµ³gÙ5Ë¿ÖYOãÂ0à4A嫯ú‡µ±€˜¡šÜ7†ÔšYÔŽõ+è»-šî‡°†mñ]ÑWíZXº„M¹'´‰Gî4ä»é›­#ÎÊ’'Cß5¤í7¹æÓs»ü0~­½´Û¸|¿Šö}mÞ*#^LL¢¾.ÚXissÇö•ÛËñ1ÖŤÉo¹Aḙú…1fš ƒ˜ /™—êØzU.¶_âZËhká§¼jäÆ ûñ[Ð6¿’È7İԀ»òÝ­V±i+×9PÖZ8‰šÝê_¬9ÕÃy (Ž_{›Ÿ'±éqþC| ß¿¶I|·ø%>Wʬûÿçê+Fü·õL‰ÃT|”Ÿ[~0¼ÊtÑ,Þ›gž³ÎŠ™mX‚æ¹ùVÕYaÏj1¿ì¿~E º9ìW¡Ò$ýÎhºËWn˜í@@°‹Ñn±ê»»(lé ;ßâs#|W:ëÍ +_#ªúS|·æ†S–!«çþh´Òè c2søôñoøJíêúuñsy<ºþÔ¤|‹yEÌ¿ÁìÁ­/OG^T°kAâX`ʲžuåä ÍÓÙÌóX(ÎhÏy6É«Yaë­èí¡ÏKæaÖI›Øûþš=(*pl0sËpÒ mÀ@C#eoü»¤5÷.³GP’ÎFIÙŲ´h£äHÅ«{ç9ÞòB[n„y»ãÏí õÞrÕVÚ˨&.·m– .«È—Qq¡R®2¸Ù}ÞyèZ½ÄÌñÙ1úª;ŸkrÖ2ÏÙàáÓ­1ï2ù´|«Ÿïj_]ÕZútWb¶¢]“~Mi«qsnÁ©×Ô•²æõ7Ì Ù¨H¿œj·³Ê¿£{`Öò­cl+…´D#,¸)<Al2 ¼{Ù1£âºé—]-®÷![%,¸Q%ˆüß—¡9T³s 4ú ¿ CÕ‚Ù~§Kµ¢s׉ëèÿ_WEOíØÜßâû›îè+Y–éÔw\Ÿ ×Ô`ÿA\±•ع?j³ÝEeç'WqÉ(ÓÝþ3œ–u­ô릶FxŸJºù°å„@tè¥j‰=Uµp°¡~ ¸`ñô¡½yê9mn[”ÉåÄ3YBµë‡XÂGnHáex<(¢K§^Lãþrw"m¾Q2¶vV½xÕî“0þ– cLóÏ%|äáëWè–Î:eyBÒ'] æžmË w.ë0ÉDQÈ[Ÿ™—!ñ›d¹ÇynItsÎHÿÞbŸ ¦¡ÙÉ›ŽKO΃宩$Ã0gË}{è\ç®Ó[çfKùÿß&ï„ʺ+lžZ—{ ã´ÛZ}˜Þ_¼ˆ–<È3öÕö¿Ÿ­ÂH#+wø¼Ò £]¨ÌTNá"ro}ÏÏ_˜Èwï@Ì ¯h0IÀýFëÒ&[æ¯Ü²WCÜ'§^ŠmW%LèX—[Ø=ѧ‚¿|â«^|½Kã\ÉMvçá©Ø½8.]˜µ5¤ôðY¾UŽaD©Ù¬‡ ŠRé¤R“I†bg ðœÉ¤:žÉWçn>­T¿Ì¥§É¦BŒNÓËP=¯–û?•ï}3¼>&ýw*r¨äÙïo9aÍ€@÷ j/’Gð±ò8ÎM¦ÃJq™ÓÔ£ÊYºÓ¦ÖúÓ¯·[¢ê§Ì´Éúìû–Ô|Þ¬mA;ß ß9z*1v«ë<òyï²l¸Š*úR ÓŽ’&ãîvu—ÿt¢äT›t¥þz½ù¼‰kXtÃWraàMÚk‹ç6f(õSÀi7I7ŠÐòüw²Z@üN§ì”â`ÚòÅw•–—c*Øf9m)(zP5b+ù¶’¢r·‡KШ3Ý–’W²lÍ+­ÏH«Îç~Lñ2ñ&ýמÖ=ü×Wľ·÷ÿgߨÊ8ÙM3h\/nÔ“ò› T½ O{ eÎF/X2 *#üÄÔËÍܾ'iRÙן/AÅ¢qѨ¤M;Ú)ïI^fkÄ |f«!1§´¸ÀsóÅVKF\b¶³1mÍ-Tþƾs‰Ûé”dÓº…rq;ª§áéý ¥\¿°ã7™Úú» õ…†ÛWYj­¨ò» Î áâ“‹WËnçéõj&ó=¸[ÖmçUddwíY¿uY“‡ô®:öÇn´LÓ‰ªs7¨0Cˆ£™ª¬Âþ¢ “á´û îá8%¥yƒ‰†žðÂ@@ƺø¹z½–Ûñ\%èÌŒ¦vŠ–3>yÒâÅP ÔX+³ pÕÞ'ÛŒñ¯,ù=lW /ö©pܘÎ~¸~¾/IÜ^ÑgfNÏÐæcŽÁ,³8Î?ò~÷ðE´ÆEÊ,ZhÔ4z²åÓæ±O1OVoóTì™<›]Ò0žùÞSDM­4ÅÍYºXÄÍ…ÿ¼&RzJ\î½÷¶Óæ›y~_º™RF þXÉ–C5Ñ4“>j¶cˆÒYxßG1W íŠþãw™i–Ó?*kœEªšr0Œ·ÀœmÒÂýC2¢+)Ø4à5Œ,"Âø;Ú<”䪖’ßS„iŸáÊé©Ä†m̰ªÉSšScÌ3½¦§gxÆ_ñäyíýUI€@n0œ¾—‘Òú=ˆP·Ç‰¦ÀsaÑlöe¸MšN‘Èi@X[oV¬„§A®)‘ yê2EÖøê7Q¤Õ µkwkZÍ<Ö.'ÏhRôLSërW¹Ð0éƒC˜îÖµU×Dyš¹’™Ð¼R>r6—‡$·C˜ùXúéØÿ#N§øWéÅèÿnŽŸ/évu†Ó{—–:,eš¶LŒ  —?hu¯£4¿<êPÖèò5%i.á,lŸÇÈ~v²CýcA«ðö™;|Qk_±þ =qÊ?+èãêÁã+þÎþd:ª½Uœ{ e@Êæ°³»W—ÊPç5K¾ÞjSõaÉ®ÎéÿÆ3I:YX:üG*²vÿÄ€ª^ã³Hub ¯ÖQoW¦Ço™ÍùQ›´TüX]·6΃—‚GÐ$s‚ {*:tcÇÔÿ—>\½¡3¬­ñªÕšL}‘4ŠŒ¢ƒb´7— 7³ÂÙüßAÛágÊÛ«¶þΟÁö×÷ùy¼oîìLo7ú£,lÜ¡+NÁu×qKö¯þüâȃ®›Ä/C[˜’J]Ÿ:³÷~£vVB bÇýlÍ*º  fZ¢ßßE=«óÓçãZ‚48îíë¦0"°\,|¡ÿ:möXÚ&äécùsÁƒFx1&aØŸÈdÙÛÑÜ·ye^°¿Ó«ùÏ(øþ—îÅáïÂý?icûº}ž¡^'ò˜Æ¿’Lf¹\Ÿæx²ÀïXËŸFÖyóýHÐKç$Lm-o.ü< •r§çò?/]ó38÷žÜýw;€ €½ÈŽâ'É•ûnŸÁ€û’<, oÀåSêÑAfaZ9¶ň«Ò/:O]_õ&‘fÛ: õ¸ýììW·8Ó1 ÅOµb\§¬÷äm$¹ÂuH|“ªømÜx_“sÈГiÅa"BÛbø4MQ¼Ù·jSOií,ϱØÌj¾Ð¼¬/löïr6ãXˆÉxêØ>T E¨øìrÅŸ.t¾™Â$Kœ´¡1ú’™¾RუY§ÌÖíZ߆£~T›Ê×÷¿œé)ªûs‰£]Ù!…hmM/6ŠN‘(óeOë3c£ jŽm€'.Ïç駆Q•q™M„å2?Ϩ>D ÕY»áö™vÆáy®DCˆ±HàóÈ#„ê¿ Ü²r•c…;Û–ÿ`yܘî81FkEÊ]è#°Ò8¬‚¿®ŽŽÙª†½ŸbñðK|-´í5²¶þ—0§Ä¥[›9}Xýí Ú ÒÏîÎFjÿTtBë¨ý¼î•Êb&4e99˾ôѯÎÐa®cuwóÿþÐÐñûþ—M±ù ¹Í…|¬zÇ'´É6fë[!éù‘!áA´Ã÷6ã;5"¼Z쥿8òÇ‹Á¯©›S¸AM·xP‡úȰ¯Ø5ìgulÂáNMcvø(qŸÛýOòû×µ™}Iߥ:´b3°Ãÿ—¿ú·õiÏò§ ¬ §ÂTfÄ ï¯¯Ï}ØQF,Œ˜-«eÊÁ¤C†7yÜ [0 ˜°Ñb½UÞÕ¼»­eæ4ºÝmÄ“{—± ñ³)6x‚W@$  å¢JóåDjVÞ]šÎgÌ7rÿšïÏT| ÛZ#O‡–OÔk]AÞ«CôÖŸd)þõá¿‹|¼ÛżâË‘åâŒ(Äm]åm\ëk˹×WÁËp3I­ ˆlôS—•ýßú­-½™XŸåf²ÿ¯}–DåªÁ†ÿºÔnJ‹qK}áZÕ&ÔÑ~xyŸ%hìHÜOª«3SÞ̦1'V6¬9l?ïs^,¯’|D¿§ÑaÙ³g Ô¿[ÚHÎÎ(¬FgS“nÆWebSæ²¼jòaašYŸfIÂqbâ­lH3 "Õ±ŒM*„52ªâ­¬VS[ÙŸ¤¥¦NyO ±f˜º` :ê²êpó/Ó¦ Ó!ʾ٣°?!òMT/µñV¼ù“?ÇÆ£ ú¾d‡‹½üÚÓÞ(ÍåGä|xnL«Yal…ÝN“8dïü}pâ$YÉ ÈÙÁ†c`q¨un*‘ç`À­ìö­ßkžHŒ~íÆn&=¦ôȈ`;0™RÆñ™·IÙqn+¹kíwü­ƒV8° ¿lÿ],*RAT¬ÃdJY¡@–йElã÷ó=žŸËü-ÿÍþIhx¾»çŸý&W&ÃZ Í¿’Szöt&ÉÞ§Ö‘™A,u!.ÙE´vL V¬ÍìÜ}Ÿ[Æ!q=Ü.ÍÊíº„ëUaÌœ3¦XaðÅÌö\ÂÖ;9píÖkew”Œò2äüvíÓ˜ñ+i®T)ˆt&ª,‡?å*éM ÞP»;lxkUžÌêQ2þ|Áæ®3+ªÈÀÆ(Êÿu°,F—¶7m“ÑäSP ZÚÌ9[?ò“è~äaÚ!wMì'ðÑŒ)‹Â@0>PmÃ5-VÏ‘À ßšƒ†!Z÷úÆè÷Wý?FÆçsñZ²plü•žyÿGòÔùüÞÜïú=†õ™˜£ðfTh¸h~a%†&i!ˈ$ɆMÿÎÇ.P¹2ÌQý%b-ˆž+v V°œ@Ïé4ž•¯þp-cW/ˆ6µÙ€ €˜7P=Ÿxvlj{î3ic‚RW zb^ÃbjÕªš§®ZƒHøW©:ªˆ¢EDµ#ž“õ9±yNøûÝlLŽ7¤µÙtv³ÝèìçáαwúM'nÇnÓíÜó®;GIÇòx ãQ¡~ï}hÎSºG0M®öÇ?gsÿ@Ø/H’ë››Ñçí ôÚúν5Á :²G|püõúçTM¢áûŸ“ñß9¢‡!Ç3· *ïo>/è¡fmj»9ì}õêY =ïŸæV7DzԸ[4`ºß™10æw½~|£Þòæ×ûzV[êñç®co/[ _ñ9é8T˜gº#WFlw£ß»LÎmØ»›˜`0=c‹Ïño³1c$†$˪ü!ðù˜”øö&}œëË„_/ôñv(‡©>æo…·£ 4–Ž&;S>ßâàÔSîÖÒÛå……$`ÛsÎŒk{òt3l¢çæÍe€ÆÊiö¹9ôŸuHXß=FêYkÖ²õ•\ùþãl?_™¦—÷ô9¿3S‚ pRUÐò ’¼õøÀ@pÌBt$ùEj«ÓLZl^2Ca4z’ê.> ù™GLŒÐ²œï¯ÙR®Åk6¢WÏM]õÏæÔ³ÿ$<6#ÏÆþs–ŽXu¾9úýú„s”Ugˤ¢º&Z'LZóó† cÚ? hÙü1,He1ÉðGRpå÷íQùÙÅ覨Óú;@p§ÐVµgÛö.ŠZbþcˆW¦Ö³÷šdب™GëÂÑF.R-«l¦Dâ§Vâà>ÈyiEŽW[re&ÎÆÿ=Œ5»…Ã¥Öçλõ½Çý,³J§o'b"L2iÌX<§]Õ#¿£Þ¦Qió‡ g~uá;†­?ìy5åFÓæÛtÒ)Б ÜbÕ&¬ù{K}1Ø~X¥÷0 xßõÃ:¬¯&älÉÓUW·²_jD­HãØ”2èR©Œ÷]áœD¶Ð:b¯ëZAÑJ|¡ßýýW)Üåàsp@aÓ~g#Ô›½B¾^^Q Y,úÌT|2‹ú硇@ Ußv,s—¹¬ç¤—)M+’Ÿ1‚\{Ýè¾6ªÜ¾n%#"{_³ðô_~«˜¿côzò™ôˆÞÊ}~VLæ•ÚSØv¯©J·ŽIa•ÙÓpuóÏ2ЬYÊÒÚ»ÛåÚ ôm–qü×.8È1êe¸¦èñ.¾m™Ê(zÍÏAÕ¦‚º‚‚Ýe›þžW\Ž×ájúíïmÀèÅváO3”Ë!~ƒ—f8ØÍü>®>SÏu½þcZê5{üºÿ™ø>¤ýÿ=Ü/ùY»÷:W(»:Í«4¹qÜ¿Ê3û)¨Ëêåå-9kÖz²b1¾¥ni­ÑSє֯ÿMVÈʇÚOi®ÆÓ@Íè‚‹Ñ'cAÌ#ÓÑ÷kq§G&cÜ2Š”2A‚=.g_ñÕk©Ðù¦l«ü¶ß»ò{ÕAä‡3º×9Åv[o½cç;{³úœPeAwË»³üûM%¼Ï„E§mzT§01f庶1+læÉÍ>ʧ<¸'÷!p(¶ –D¥‚EllÔt˜ÖKªó¦;ñi|ì¶ÿÇ`|1~q’Î6äxEøYѼÉŸ÷݃¨g«×=IP^+Eåu›Ò¤“˜43×¢õ ‹AQÛ¦÷&윹,pyGGÿ\µQ1WA*ÏqÍžëßAŽWÁä½ñv-Äû¯ È„ˆ‚”€@F) Ô<™Ú_E”oôÌ?-˜æ¯Ð;‹^TVPãó#y~»Ã_CÊÞ]ÄœW×Ìý¦i¸ÂQÕž/~s¹¼ž×>o%ö[_B:ä>÷󼄮¦Â¨÷4øßqÍãt4Wë.ú¥~Nªx8Ũ«‹¬2M¸•^•g ~y‘ç¤ý` >€¯SpÇmjëÿkJ¤%“çJõ¥‡U@ë ÅV#n¸£}H¦VrfuE5¶÷æ`wJÏÑÐXëI @<¦•w o˜AShJ5ï(Çeüc“åGú°?<«ùZ`óùÁùŸ—z»>¦Â ‡052“/Ä„(%ñԘ臫‘Œl{c”cãŒ"Ž ÁtÉ8É*oªá~Û»·Öf‰»#üwˆñ2ê4‘Wî™Gc§É¾8²P{˜LbÛ3HHC§°Ó ­ á¿!Žt&Ï4BÌ/ ?wå šÁôo%Ë?×ß³ÐÀ±Æó’ÿšè˜AßÄpe °‚ ÒÐßøÇG^ú¨³A¶ÒÈÒc<ô ®{ bªªn¬0‹i´i(T!´!$¯ãÝû|±a5¸g™¿þç~‹ ¹¶ï eGs?næ‘sÏ…ú%eõMFæm¤d'§ö9µ«‰2²Û¾Žuå{´û ¾U†b²52@ )óDÊaöMÐY›Yºfî¾’·Z»«1]-ÞÀ4¥ñè^HßQ=¸ôbzþ îwKø¦uÇ›OæÊsɽ €ãjЗ¢ï°Ï_¼óû¢kF¤SÅ _޼P_¦ñK½æÀ;ÿ®‹!Çmk(µH\!ÕG“D"ãÊ—±ûßàÂЀžIñœ™oêeš0op‘–ƒZƒ‡*´¡R¿Š¦|¢ax(†sÎd<ÞÝÇÉ×ã¿ôk·*½¥Ãß2yòØvÔ«1– Û35ÃÜ©ôæhm«JZ|Ö§ b2Ò,vñ‚|£ý££t‘…5¸K7ú;²Ÿ `ÝŽüØZ›Ù k'íªç€ |8¹œí}¨ðª,T6Ç+/¨Ð¥ t{l©èÚf ½¯ïOÖÕÛó>7ð½[Ê”þhÏ<•jì¨öûQäáïKŽ|Ñûê°ÄçE¨­ ‚òy]Ü@FyHy¥útçXQBÆ)¬©Ú”"ÁA—¶r"Mq®IÞìÁQ`]’VYr`ƒ­E¹g~t®YCˆÏ^§·ÓRa]]é/w´Àài)J$ª%6ˆ©q™ž|´| Zw¡q71QäJË™^r{¼>âîz<ëåW’‚€ÃÉu̦ÜTàê‚#ÏuÅ*ª4–e¸™¥ìßô-i~ÿ¸÷­ý–v9Úî’â¢SWº”h‡Ô­¥DJV^ÉdHjëo©¥Â—«ѯ.¢} ü"Œ® ¡L Â‡½ËöîHßòfJ·œeéÍ@@å)C¹ùF'Cêw3œ"û;öy5ý¿yoêÁ7Íãnª]é8 ÝÚÄ›­cƒ¡Évõ.âƒ\ÊÝBLoú͇e¨c ´cê“j'2ãׄsÜr?YÅUntú-9íµýª´ =5ã™7­µüK±šò '` ;ômäNê14ØG)2$æÜN,(4¨q])½eù‘ŠœŽÖ4÷0‡u òKw}pÿ/îøâòD^GFަŠR#’¤¡ÿÃQ³!¯›ÃVä¹u·iæÊµ´ã(•±ì£à”tŒ8à9¦ãâÖ¬˜Q42媎¿taóí+nG®M²› ©¿‡ÞOB½Ìò»ºê,EŸ×5щ~fîåo¥ÚÝ“rwÀ%8¥»ÛæÔÏA4 RtJrHÜÞšOÝtYä@Êén`ø†q'εfN7>±ñ¹*ŸV$Õ3v Oì*ÒæsÌSýíÓ°ÓùDÞf—X8¤º›šûd£‰I¼/ZdEl‰Zèhú^pïô³\¬˜Ž5¡\ƒ¾Wë"xsò”ôrx:Øzku)ý ÖNdž{dMËà?( $±OvI``«r,ØÀu§|ÊW˜½L»˜Ã5Þ&=Ý]ŸE^×ú€uØ8÷#9ÒN ÏLËqw(àÐloà˜\§ÞKªLl˜SqÙÛÈ?Ô`ÈùÕ`R¦-- á¦pªµè&¶ÈÂývÏ’à°@@m,Oœ!;о‚¡æqú¡©ªm³c$IÖMÅkæç>ÐV Ñ QõÎsCë‡[Yâ‘–aj°˜®?ÕO¯SŸšZ:?Ì–aË`í›—Š(1IU"îe_K0>Ó´†ã^ÙÊct›¡.fô>ÄàÉY4.8±À+»Šj{’í̹¹¬ÇWøìôxŽ È”çÀ8-fgæ8ÜF Lm>½ó”âè‰P¬—*$æÅ-~¤„ò{‘»WEùQ$4¸è©Iû’ðè FŒÝFuÆ7•ùì;ýžÝ âPÇÎÉ£¦?çYö|»ì¬.øÊ׸B”¡ø¹0m‰Ô*+#­ƒÊè}Öcø³ÖËû Qs:€ê„¾[ãË·)`g €{qd ¢«ßr+ô©¶bŸÑN«`ÈæTú48_ËD;"jÄõN«/ŠGÜf£b¢©J×.A¹r…uÎ¥çTê¸Ñ(Øa2‚`•ס*?îgVw u T‚|6©×žTÆ¢wÚÔÈ12c¥’x‘& =Sæi¥Yþ3•ÌöÀv–~L—üý¯‘9½,[«ƒ»´Ô!GO+|„?-Εà¶IG‚lÊ‘È%-ëã]©f•hNVˆ\‘©Ýå§_&XUEl‘µ­ qeoÓîEù= ^KEµó}l-¿Ñæ˜ØæjÂî©ó¯ÔxL8¿W©ouÊ+NÅÛW÷È'‘F+IÖ?ÎÁiµÕ¹9Œ°&[˜§ÔP=}¹áÌÏo2u'3uð$½ú%0¥Õ(!™¤µÃilÕ5ôÙ94ç™(4^jê«§%“¡ç(V[Ÿ¡L¸I±¤¨ˆš{C0g«Î…B& :Í&ñïçü|RVnŠNr, Œôí?—Wý¿5¾öÚÌç–eª1Y0ì4ü*Ò¼µã¼Z…^#`žZQxQô*ø™‹@•d{, 1lÇŠ9ؘ©V±Y0}ääÙÚaÒ 1=Åi•Sf,ùÒšÝÂŽâÉ ®-n5;ùÖèYÌIÁ[%EâE8‰Än=‰DH&^•àóüƒæÜG¯ìMJœ"8fŠˆ†kâ·Zçìüêœÿİȟƒ.ßQï"èE`¢hã ÄW¿¿dRßCÑÕG©Á¼?ÍTtwúúÝÅ®4šÄ[Ä›®Mk™ìC*‚¼ˆæ•ÏÊ"Ê<+%Ó :Ég&Òoù]©¸ Y )ƒ;ß”=¸µÑë¬Îö›cáäÞK÷u¿µ™|Í U¶ýç®Nµ¼_„Ÿ¹wUÇ  ÕjU§A/§ ¬4Ë –F)™;]Â#!p9d•}åâjÖâ³LEÛµgw{ÿ3+[¼HÆ¿Q«Šöô‘²\~)4Èé·Y´1–h“f<]Yì+Ûy´ BøÈZÛŠ98–/#.Ê^=9E–.§ï·Ì¢þ<–Þçñ8vÅGÛI)“"äÎpÒAÖy(K¨úž¡ROÐ8¸k‡›w‡¦÷MFI(0’ð¶ª15\ÇtOî)ŠŽ¶J”óH¯S/ªÛ¼5+nŒb7ž\Á ÖL©‰{ÖeÅyMÀç•£ù…ó™Jqn*ˆ+”·ÏÒe4¨‚ºT*q§ÌŠ+tÉåˆ.¹(–©º ©”)Ü‹ŠÒxZ»T=eÓžJù?”ê“÷Tã¯GZ~ ç58¤Öô~âj ðs–|MÖR&c’D 3%W˜™m*_a¨bžÙÒZ<ñ±´Gè^S¦f][Ä¥åPtýþÛj¡w&è’ñ¥Ó¦%IH¦lÜ"Px›iâvr0•ÔòÇ@MŠmÛ›æË(5Æ%.•Ýý6§ÒÁÃ+¿wï¶`©zæžœ1éYöÅgî wÅ|æî#–ãVÉÓneŒßJæ*¬aLjO¸Iy¿rþ‚õŸ&:‡ØÛþÈëÙãgR¹1x˜´CÏ͹1”´.É2:n¼‰âÊ«¡•œ¬{"ÄÔ$§áªt£/H%,êŠ(M^Â.}†lÇ_/ñ6ñW¬Eˆõ3Á¬”ý¸<´ìv£¼#¾*õ‘—Ð#ç—äHÊ‘±Ïά¢Gä­è”Jr‹ÈàÑNÝÑ>Šh¤:]Lš’0âдÊŒW¸IßÙi}åñºñÚVjG á8i&`¸ÕÏhY»ñä£)xuè†ú®>5†.AtTÚ®S:<^dûå+@áX·ÕÖ¶LÄ7¡^±FW€`G Z|Õ(Åör€¾'¯”]ß$ÿ\þý߬LÞ§[$bH«»ß¢Ê§¢'«efSö|^Rq«ŽÊÄ­ÿž{WzõjAöÒê©Êÿ×^>Š3•ŒnÇ¿ePòjreÎA<8€Å¬¹ÿ rÛv[±qÔï[Z׸+ _“˜Â­b>F)Z>(ÙòP-—ùŒV žöÆ/íK o-VÄ)1µIí7ôªæ—ž×—!b6Ü«| †Ä¦¦gQ’¶Kƒ5Qbkƒb¶|â{×?®# ­#ƒŸ—ÂØÌvl«˜€Dgdµ,}óÝ5C7C•gççœsyß>Öƒ—Á‰m Uƺ·è&:÷לmëýÖ†~^ZÚº¼œø 8ŠÔ5™äñíª&¡ÆòÜèò¼‘öíFbôlîLŸÀèÎÁ¸˜nx³†@y*8ØÉ•dH½–' $Ý€ZÍñCœS¨N¬t'2UcÑ-µÈ>ø±ãÍeˆ'¨„ ¡Æ ðöåØéO¤mbŠ–~ÞÖÛåµþœä«lßd¡NQ³Ý3ýÓ^ä”YqÕ2Ï«§ÑÙd*¬kÛŒ«ãóeêÉÈÕœ@F"PÂè±<)ªÚDz|ìíÔg<Õ|_Q†Ëž2ÒݼëO€Ì¤´‹ä'EXˆ½CrɤmþˆÞ´¢ e´ò¤ñ¾ª"Ë™¯{{¿[‰Dÿ)½&IVMãH‹#±XD—øßJʃ+XïY›]UBœE¡÷7ž¤*c3l‡—°_c0 ãØ8Ù«À³oJ¾Ru¯ámæÆ1"µÙL‰ÛBÕÂÿ“èà…vÁ]/}‰ÜLÑ䔄/0ÖpÝ^=“~óÚÔ´Å÷ïJ¿Ðµ˜Ø$1±Â×ÿæ¶s¢+ÍTÊÔUåº\“Û2¿’w[ïÖiHINZ]OK\Qç¸gQ’쯻Œ˜];~¸;E²1Ú=y¹½n>k~†[FíýEA .»Á¬³€ˆÅ_·ž¹ª‚£i@ØZÁ :aTß3 `cõOIP¡bàÅÆ«þ“O½,Cã ¥Ý890\]³aÆ5N˜–D›Á¿$\ë#È„•ªåï‚Ö¼»AWÖ×r¬³¾¼§ûí×Ê¡âm£õZ·&;Ÿ#aÍñȲëFx#éáT8rÕª¨ZÿõŸ_Új‰ïF«ŽA(BΊ•Êëð¥1Pàý|ׂRÂ7Š=ÃÝ(îy`”^;ãÊ©8Vulèu‡2oði’ ÅÀL#瘚štþêÈIFg™÷µ¾Æöö ’G$¼©gNÝßþî8#a륹Oof0Í+Pï¼C¨©_i³'’ÃïqiàF̃®šÒÒc‡Ju†Ù”ÏIOe[ª‰¿±z¾Þ5R4¶žž÷б;‰b~ ·*CшÓg‘ÁîófTÓxØ‹Õ, à)$Í\<ìÏVN‚ž-¼FB¾‰MY?KY›Ñzy…úµ”€#jÕºÁcj³#w뛜ãÞåg㾎­N9êð©ç9ûÊ–@½d  6­D3¦ãÎí¿ÜÛ媒N¥zêa°?ñ]Íqäª T¼Ì6™b ¢;š_ôôLB©ÆAÈ@¬x øždíT•·­rÛLÔ–Œä·én³Š„†€Aúˆ¢ÛÉñ«í™ÞYÝ»¥ö+©ÿºÄo¦N‹=y`‘5¥Ó¾`@Þb þ+ün+øã)±är]ç¾»‰4ƒu­ê›\oPò,Rñ9¼-ŒíÕl„FP€BúLDS÷ÏϬµghZÎi\ÜŠæø“þ¸ú|*e?ÆŸG–üÛ—:âÿn¬âûÑßzáÞ¨Ú>·ÅéW–Šré(ݺõÇñ݉ê¤e„&Q¾2^‚ ÔšÈ%ïßÝÆ1L¤éTãç R·Z¬•'·NÃÏâ^Q!Ö÷rJ"Vƒ¿¨ÂåO°§†,„9_Ç%èÖÖS-Ñ £.õw–îÝ™š€ý¯92¦jµçwB2ì\‚?ïO¬'FÖ2]Èæ§“Q¡¾ß«e¡ ÿni*ë4 mGÑ«O€½oXXëf¦Ò¯xÙµ’ÉÜð'¤É«Öþ—ùªw¼Ÿ ÿªÓAî6ov‡Ÿ ¯2}î:í¥š,GTîp³Neà?Q¨L€šýÐ\é|ïEj‡€Ôûö^êÚ>¼@É›(¡B^rÊ­¬ê£èîLOÒJÇWĪô,ëLVð€Z6r\«RÔ=殆UÂË£mGûÒiÙ¬¢yWßTi-Á­FãCµC†ÔzÞBáN¼ÆxŠ(aÒi._Õ€›VQ«²¼û°®"%b+Šœx%0ïž“SùÐõS”Ì´–"à{™"{…@OTõ÷Oð_}¨¬:WüíÜ}Y¬Ië¼}‡ò2 $•Q€ 2´ßúó TíÖ³ëš)pÝì¹À&sÞ=œ÷¹õõÒûïXÃ0 lòkÏÒWh›[0¡¿Nfx¸3ZV6õIÛX·s™M7ê'S½2dÔ‚£¼/‰³ q¹•×c“uªU±sîÒ30u›<•K=íÉÅ­‘•Q'±€ŒWØÐö5`í;0\ªš?s.²Dì Ô”’»h eÿ›8—õ»kúîJxmúÛg´Èƒ:nÙòÚ7ÕG¦¼Þ»haVe˜@²8gJL”]%â‘?Wä¹.#íéx Ö!ö'•¿Ý(rå·’íÅ(Ÿ²Â)èݬñ´Üß.*cø ɧs]ék¬Ý;­á¡i×ëS”ZDÝêͨp>̳gÔ™õK]νÇ.ERŒÙ¯úèÌò¢?1ôîí ªw@ÚÑòiæ&š¡ë+©áËWN|©F‰qYCåå¹o–,¿֭#‚#I%¶øZ,2}¹—›[BQÆ€Zò3üŒ5¯ëÛÆÃäDǰŸu¼‡eúßžBß±]M«zóT è8ïÍG °;m€µìÕ¦sZîžÿ=V½óµ"½R4îc11MŸ´vMŸ4ZáùÛYÓh3\|‚ëðºX4mwŠ´}c¹ó©ò€->•{Ñ®Ž«©ö-í8èôï¡PjZø2…ÕÀGs=Çnwks2LÉ·€&Ó×]̳TÃÜŸ¿ ‘&’=qeèJÒ•|wççŽ73î?ÙZ—~Î_Íá;4¹SPŒ¥Ü¾Òö^%ÿ\.:_øÌЀ47ˆ_K1îõÎéÄ;'ͪüobZuÑ6²t¤iÓÆ>à%ÚYGˆO·æI¶Z(¼Æ®3˜g [澯ô¢Ý­¶YÚ—í¿:(‹?íïmõúÄsÔ)Ó@U5]æÛ®#…ž®3`wÒšÂ]×ø›{ŠÝÌRðæFd>µÊÒ…S@_~W)ÎŽCYÿI™2ðu?Ë‹Ëþó8êœ.glʘ FÇtì_«ÓÝùj|‰Q n&Ž‚RÕn*޹£ž°0¸»TëÍîØrñ³¬~yšltö ¿ŒõOíÛ@÷NNÓU1Êã5'h€Œõ,-ß}ùîpu_ý¡ ¸0–T¾œŸô1ý‰L0df,fEíß3øõ\Rv<ŠM:ÿ¼Ò½°…tÆãgdYˆCCë¡®ïÝðɔ뀈ÍÑúÇv”å¬_©Èßö̱•©æàßa~Ë¥ÒÎ`(tœóÄ>€,Ž ßs™›_ÿn=€/“q®,Ó‡¾8¢È gò)Ót•}íãè8—yv«~i|Þ§ÜY”̦håª{öåO_áÍgÞ»,R¨¦›åA¨€A“°Ý–® ë•ýãµîb™¶+Žx‰³`"üÕò@.}ÙÔÅ:p´ŽSsÚýÎì¬ 9Ok¶utm'ÂGj¡ ÿXºøþŒ—)P²e{9vÐû¿ÎƒãÈ‘£² ˜ßÙ³#dètm®™Í’ý>Ç͹ù›Œ”|Aiý-ä|@/‚‡ƒ›ºÅɯîÚæÁÓ&ºº]Õ» Dßa7Oú˦+ýFSøúp.]¾¯¨/ V¸³fû\²<«w¿šFðäú­8|ü^ÙõBÜý²jPþõl‡¢Êî·t1ßé;ßãIäö!éö%‰Ö€ª·Yî]§È‚}@'5ü0ÆÑñ—³}úÏÓ€³¶ÿèˆwPÊcêÀOó}ý0ÂÖ_„ÁZKžÊ_$?ÔˆSÇæÂïJ¶•耭;Îd -_0ÏúEÂÞº|÷¾›š„f˜¶ §QÓ?¼Ö§õ D€¾Þûnˆ ŽGUãýÈN:V`“ Ðñ×øVÅ> 'Þ³yÑããÓêÑ’•rxaþp¢ŸGo-üÛ€«þ¯óV¤ž€^z·-ð°Bw@] æÙÖ¢V %Ž÷K£É˜÷h0X))õVúnÕÅqÀ_ tj×ɵ§Y®c-ÉwF倶üTbÆNy\´×õÉU‹œX)é¶#T-ÿ}ÐòGôIKÙwxÝ3ö¥Ó“-ÉÕ+ú móV­U„¤ ŒÿËIÔ}µ!ÿ™ÿú:{\¯~él úúV¯ó/©³øê×ÜŸäÃAÿ½ÿR+]ÃOàŽûö/õ\LêU¼.ãšU8cØÈ« ¬Þóô#:aŸÕzI­4x€ÿ›â¥4õŸE4×oøÿÁÿÛþÓ©°4hÚJ"áPVòµGr›—ùÿªÏý|_?ï§ÿjšÕ ŠÔ€ €(cúæáàE£7iŒM%ÿüé”[‚%l5ôÞ•ÉÁ呇:W‚‘Y  bÁ?¯œ!¸ØP {HH>‡t‡_VOò÷i€hк)èpòìÁÞ8eu •¥Ÿ¯ÿ»æðŠ¿ìS&‹ $T§,qD%p„)£•E/d`Ly¿&åËÿÈzÿ®«‡ó'ǵu‘0Ùqÿðüf¤­A@ÈgüÿÕ»<‘ÐùḠ]@ïA%$-©Kµë#úš™:5$¹Oy{õ”Ø:\ˆŒ#c¹ƒ–g–ƒ3“¦NC3„-£Re¤Š8<‹…Šÿw˯jOmƒUs˜/Êe  ÿñhØÖˆXA„Zµ›Iw ÒŠÍ*Ç¿•Œ?¼)pÿ#ÕíÜ ŒCSßvä À@þ Ø7n´fûÀn¡7ds?r3×@s!#ËþHê¿‹øÌ€±œ*ëRã“"?_àÞÐÑ 4޾x«×ÎQo{“€'¥ÑÎüèÊÇ‹¨I”»Áð{'‚‰ÔäuþÌ$x½±­Á©ù?á¢Ëb1›ú/Xy»Ö €(dÇú *Êö™´uº”D1ähåÀsÓa•ëjÞÃi¯ö>é>¥QCk?(»:,-È Ãr@ÍÛ'‡5jð„cí§%¹Cy<‡ŸoS0ü8ÿ%,»²ÝB~÷(:¤½hÀgíªeøïµ2a,Žx‚ðÞ¢SJ$‡C,äÄÏN›é’$ˆwàŒ(ÿíçWog(ùšÓžÉáGÁŸEcõ·©3h‚_›‘†„囨sÕ–¹Cƒ”CÁ „cŠõ:ÿïéÅoß-{óVr ÎIÛ½üI•®mþY†ú(ÛÇâèïõǟɈžždv]¿7­×0¾…Õšæ»Í¿9Û­@ÎÏVÞ¼yû÷ª)Ç%ìÖû?coq¯ô4èó9㸉µŸÉôaÀ·T<8ƒ“&ä‚#’L0‚+¼"Uqnׯ Qãœ9C‚I"Ç®PUÜyj)Vo)ã‘֎ἆ$ðª~š×R¾ÅÑ€ è¿ùz˜S³==ÀÆû6n› €yqÅ<Ç}‹5ˆÖJ ‚£%Eäà8ã€Åc¥ŠÔЗ°¸y—E‰úešvA3ÈÍ3×O–¿ÿOÐÿôœ+^7i§)uÿõºûW£AÀ-ˆý2%L¡†  NÀ°œöPèFOˆÂfíãQ)r£ËA@³ÔÒšl²ÕÒBMeŠ*ƒLa3òpvÿÄ?¥ïÆ|ôNìÇGŸi¡‹ØkŽ{'m›ê‘=»ä@(0hÃ,P!‡·èÿ¦›ù “#ÝÆ@²ˆ$Ó[”:X¬¸bFa˜A3F ú9~êÇéü›QóväîóÝôsü¸ñª½Ln¯CŒ¡Frk»Ç>AJ²Sd02“›.@¥(—3(—Q€Ä^}¤µ‰à3œ¬T„9© ±B>š½¯ê]âæ.úÄÿ¿ñõ}TMÓÚ†&To´í“Çz£Oí>”&­Ë¸  D ‰¯{L…Á¯—3‹r9ÄçÈB*//Ô@lR o.N¥ßþ2ðÎf±©ˆ_s·ñjB7)Àr,òY}5…ˆÝM‹é€lPb€ €,&ñBD++?“‘¨PH É'M!³HK!"®š¤ðÓ—UжÄ^‘SQÆ<Ä“9Xi`H²„—ußï~¨6 ¶èÓ³Îxt%Ì\½*¦/[î¼LRžÆ&‹³À@ÜK†ET* mE$DÏ—ËœŠt_ó`™»à†™>iÄa+[ÿõ­üÿVÿú ñqkñzþ|ìgY‰Á*Ž(Üd£#ƒ(\hã”Ó*–)Â5p2ºzGPª­J(.ä ³ÓŠ6Êæ°”@óæNžÞ‡…t ÿ÷~Ns ^—:gœãsÈö5Ý;¾û\ ÜM àÀ Œ7C_•dˆ§^(Ä­AO@•‹¢Xh“…Dzùƒ‚iŒš rdöé@þÚÿ :}úÄ×í@ïÙÈ3à—ß•Ótò:qâÁ|Šô’¿i`äEžTsLA‚›’`À ¹4JE¨¢ž…Ù«Ë8ÒÖ<–¢Î^ƒ’˜U®8èt×O67£0©ËYQ¸â >Á|2¡Üõ|ß-·òjô ÅÊôò˜B™,þ}vl»ù’ ¢Ò3€1†0p‘%µQqf§„¨«¹(p£¹H8R--˜ñŒ“C1a‰*vÈéÊSý=?oú†Â“x ‡ˆxhP^ã[}k¸g”ÜW@ c9¡ŸIqÎ${ÎÄ6¹Ð߀…ë”ÄV9£ yRgÂ%ÆlJ0/^rÌ×â¹£ßýy-zÞÞø:¾ Þ8óT<þ6:QWû ¨nb¡è.•ˆpa   <óÇ—ÀæFªl¢Ü.Qªå‘d,ŽA àZrzÉt0¹gÉ‘,ɧŠ$*i_Ëþ¥kæÐÆ/Fî^æämo#Œ¶ûlJ‡4 hŽDŽ8¡ˆ‚9b0‰. ž,O¤_& ’‹£OU– ©i¥Fì̈"qò•DmGˆA“  ö*’½ž.³¯åÒäçq¾^¢þ~Nº~bn0ª2»±¡Kñs5„™Ée<‡„æÓ9„ Pü æ1©åCÄ›påÇ£>éŠRC5Oy—¢ÔZŽJ9™Õ¢žƒ1«R²ÚªAý?«ô³þ®o>k\„fµ;Ù€‹†Ü'LÄÎQFL¹Ò­Ù™9ÏPS@@TI³¢´T€Ç'PG2¡%#è&à\h¨)fáÈP±RìÌÓ‡M`Š`š¸õÊÌæ÷7fˆ+\ø×z¹¼ârÍs~ó¶ ¼•)ÇŽaFÌ'з‡gýõb¼4 ¬Qrø¨Å÷I8VÆæiÊD52$$¡AG“34cfÄSËörgS˜ÈšSý7{ˆ%Ã’$dZò5N¦¨Eø2Ì—Ï)çå– ŠDD†žbP‘rHÔ) TTJLb§êS¯Šå!JäñÓ“L4éÅŽ¼Dòˆ0%ÈÈ Q6UOþ÷Y~ußß>õlïùØÞM{£ï?Žt\Š™Ú•(š8ã–ª«Ô Ç–‚õsJy&!$“anÓG¹S 9@ƒ!9Ît&&<åF›.r‚Ù€—( æÇ‹ù ò\ÛùM¿ÆÿøbÏ· Ž~¶lN$Ârpð<3æ<…‰£¯57Ò`Œ )8)N+Fló$¡H9&ª [ñÖPÝòE„¤g/˜¸êHm}2älÍ/[áéõúÜn¢®îÔ)u5hðq µSnRƒÎß-ÔÉàf €ª8üL¡dÌ?‡ßDÂèÆ@“@£ŒSZtÌ$_BšÖ•ˆ é•7ô|+8Ú,h›MªðÀxm×W¬M]8·``àæyQ@Û)±ëo¤x´)®îX¦Õ5>q'†Åë’H”¹·î¦ü%Þ-IœdÌìwØpnóø.êHÍúú»0ÈÛîó>´] ˜õ+}_CW§äÙãõ½ * ="ê4„ë:ªw0óŠ´Ê±è€ €ÚŒb†rOj‹j¢Ã ¬ »cˆ¹$#A*N,dg×~ ëTT™ âØD«ˆÅÄC/}‰£ zƒ+retkK¿¿;ûž}îT>?'Ûõ~+_GÊÜîÛ£—·ÔŸK†¶B¼£ºeÕ{fyâ ßè¸ ‚»±êLð,ÚÔQJÆÝ Iȭв%1é IM·fØê+Zµ*˜bV¨ÿ¯å4ƒW•Ë|a°.fog/LªrmóL¶îDx4½!§åÙX)H2Ó¢¤ZÇA¢2ÕËݲ¦:¨Àã‹:¢éë^|¶¬4YMZ( cN‰A xÌskŒÊ—¼R@²°0q3±¥wÇ…‹ê^Œi”ÔÁëNÏ—[µ÷–£“á†H§Fû½ç©¦7âdiTp0Æ 5DkÃQY¼j1–§¤I‰û'«¡xZ »a˜bø¨«±B`ù"«a@Žþª³ã$'œ])q­Ü¬œéob¶Óˆè^pèZ£­è^oë “š¦px!pïI‡Zë;meÌ"ôn©Ud â"-»oî,ÝhÕùT÷ÈâWAA §Ö(Å÷±d3A·Ù^Ç £í¸Þ™#=Ö-(äàÊ8DÙDôw¸ÜŽ×ø´é2ÿÎ\ê×õa`9]—÷K–Ymë†ôk™Q$‡æÂ*tsH$P,Ž%Bbž2‡ôÔZbg&!­!ŠíAa“/hUF½u%Î0 6†¤ñåÖ©“†÷<™–€W7šôuÁè÷Ý^¡ò ¡ – TÌ:n!J&èªø¥œ©&§‚tæ§è … Ô)<Ž3RÓ-mF¯RòгE¯PRIÝ=¯F³>Gc¨°9ž|½+UT`×¶=:2q\Âyª<©B¨¨I)Ûê %ÔS<ÔΫ¬’4G ónPôw(óîþÛîÏO¯ùd¸ïùy>'ÆëñÓÊ=Àý¥jÚ¡Çi7þÿ?[d¤%}!3žK¥Q×qøit+Î*¼Äl\û½ÔÉÇQSt]Í;V­î®¸[{g¼¼AVöÝìÿÂoÖä'N˜ú|i\òúéÅ÷éö~_S+†žæL|8½¾?¤Ö£9ò@óè-ž5Üíðè*-ɘÎLyyGhNq‘V&Fè.&2¹ ª&íÛßãS:Bµ>ø”êÔ'âyÞ·‘Gù½v.¡ #êlÆÕ·ÕéÓÛ“+¨“¼úÕ¶õ*ne…]¹·ª¦S 5åÒAZžÀ¤Ž@*“_¢Oe3 X¦J#VÔ–8šåR£¬=‘G àuÇ­rÐ× ¸Um¿‰fƳÍãE<§'ˆË‰{ŠøˆËš:[ NxppжIÒÏŠÊÖnôÓG"¼®&º: %¨bjŒ/H0ë>-Ëí&F›[ô+Ú(b’¤FNb›Ý¯ÿÏ ÆK¿O|÷Hãž2›G‡âk©›·÷‰æp¢YÐ×6Ï{ºQN\AÕ.˜¤›„ÉÕ³6 勿ãAŒøÁôKG ’à 9 ÆP¢‰ƒÄfg\CQT ë5µS|fDËPÉiôö}M_š½¹"po‡ü~&]ßsˆmT»4Išƒx˜@@4JÃf¼IüôÓ 8pŒô¨ !“˜RÉbÐ?£5tüÎîL™²£‡"‚ÌôF…áDªÐ“V£‡NÓCú£èa`O8~¡RH¤¼Ÿ´yaÍ›Ö.m ªõ”ìÊú=ëo¹]ÆšÂÿb¯:ùÉ@æa,{aÙ4ãr­ˆ²Ó*¢¤tçVŒÊFÉÊéáÂÍ5  xã‹„‚g4¶ ×¼;?ud]K Bózš¯yõR¶1‡ÝØšüËv¶à\N4‘¥pФí0$†  ‰ä°âÿxD™($OM¯ ’œ'&!lŸ0ñ”ª(Zz<ªMf†ç•Òƒ° "qK›åM+Å~C4ÛŒÝ3u ÓPŽà>CæHsIA 6¶%<8äiÇžg,|x“r4ӣĨÏ>RÕs@"ªªçЪX´H8Z ©­c&”€d˜ÔœŽ’'èjß"PC z%º.Ôz–å¦Y jîUR^öŒIi#Íd¤ß‘ ê>•Jl‹¹Û–õ@@mX²}zµÐ#jä2– ÔÂnÔDrm5ñÀ‰ƒõE3h`k—ͬÁ,¤³¯×Èž ãæQÍ›Þò³Ù>×O'nZ_Oev¿XØr]kµ“¯«ªBÉ H çä>2.³9EÌ«2\é´ÖÆ´DhMQó˜ÍXàÃKlÛ„Óµ0cÒJ«SêN·+ßÞíx~×*ÇG¯‰½hv~w¯å|<`|]ZFÆ4÷Vá–á©ßdÁ­¹`0C.N9Ÿ¿0<ÄšGª•‡×ëÒ¤’ # o·OÊ›¡•ß¿ieÍÏ=Æi'Û~M¯{pë8ßísÍôCï6ó›üP¹Ùÿ/©ôå 0Ü ú· xeéó׉Ç$h¦ÍŸJAt æ›vÆ…-ŸÞââñµþu±üþÏ{hmJ8»<Š–!4x§ô¨ñ÷,`uTZð¬4ákÓ™»²DÚÊ‚OaÑá&nmÓʹ¹IÊr]*êõrRB¾¬h“¯”£‹Ý!Ã%)—ÎùÛx/Scõ•‰¹/4(:E**•VåV£Íº *|nCZ‘вj`²ÜÑ™¥²dæˆ?>5‚¥EÚKºsZ÷ï.ìW8jÜ­¶}Ÿ/±m?ãèлÌÁÝ»·+Õçëò e$‘›ªÆ#É‚VHìLFÃm/É Mw{%5':îÝX/Šä}4¤³ƒêC(öUñìÔ >:ñ…´P tÑbùÜ™öþMN× ‹6x,z ºþ6·{‘_Þ '<`ø»Rþ8l[ù#£þÊ|ZuGäm×NO›™,;@ sqo ÛÝ ¸4%½ŠÈÎ¬Š®z™øíÙë yÃ_­¹zÂ+†ó׿(‰Ow¿Úîwv= 8ÌëÚ¯Ê:ìÖ/œévsÓ}¯Ç¯ÙåÖ¦¾fªT¡ä ®‚pD²æÅá®–µró-·Ó­SÊ7É# r~S±ù„¥D…ªRû1Ò7šü~¿õë©ÑÑVhèëškXð]æ¤fmäë®WZå2·¶.‰épšÝÜ0JKˆD†—¥8\dóˆÔd’z}>žAŒZ[åwmUì¥@N0—¯'ôöÄ-°´›È›•ã^í Êí€Î,Qï;Ú¨Æð¾SŠuìuéMoF”EFÚgÒÓßàq+¨Š`‚°z‹kEÙ‚0Ȥ,m®Fmi;[W«9ùR'X•özHùBÓ,­kÔ»CIÆ{7YþrUßtgÜ£å󡨵镵/x~nÕp(ê‰:1•m#EÖ±MîÉ 4Ú44FQÏh«Jhk€)w+Ô8šç 6µ¶'‚žå KH­R§ÚéüÍ_Z¶jR­îBÖÒÑÓK×Õ4>ÿ™c³î\z¦!gEASÚþ¤šC"”uRtrImdlµg&A>½…2Î×¢§‡‘mTQ«J¼=žBvu¬|²@9O,v»Hvâÿï6χ³^gíh]¿å=Ôüeytùø¸6¨(Å<ÏBðaÝÖ½AU[ÆñjϨò•u&ÂòG2nA0~Z#ÆV#ŽÕ³ª9#FK"ÉÒE¤Ó`Õ’ÎWô1=J"Õn[{I%÷dÊ9•GIŽëÍé¨Ê)HÄ@Õîê•`N~NhÔ'[bÈÚS NÐeiµ&š&ŸÉ?f•‹%éî_Á¤/"¶V"ûÂiÊãSj?2àœ®Ï¯?ZÞ&¼x»#5@Tq/ibL"DC~ϯ7æ¯6¯?¢h£W ŠÖ¤ÈŠÜºW@Ø}“H¼ª93+KäIŽ“àJk]ò¾”®4Û-#ÑÉ(¨ ̺=ÉÜÚÜ0Ư;¿˜¥”NX²ëÊ©^`¿‰ÿåedô}˜¸ðç¶ LrJHÚ­öÛÔÊqŒ_O:¿[ˆÚ6rü~ æ5tÇÖÕh Õ(©Â*Pbž¼SƒQ¶âÒµw ­[í*ð udgˆ¸A¡÷ý¼§…ýzwâ„,{¤zæ½ʳh vmÍ&jW7ÇQ– "e^Qç%>‡×csH•ÒˆLÈ]§ž‘(b‰'êúõ—“Z±œß½“<˜pxü®ÕrÇÏ™|è~ö׿IðÁŠŠB€ € rxˆÈñÚ:^Ž”©òSsx‘ŸJtó&¤L ßW>†2|ó°ÜEõëí6i舺IÄc¹Õ•ehÜÙzÐ(+F“‘ç*䯣70 ‚Ê4¢.H!6€“}mA ‘8ŸNzDPÑ×èÉ©á[_Ô'åy¼¾+zÚ†«·R4í)£ ÜÑ 9Ï–r&~¦ý‰Nn "tˆŠ‰&èèlëTH‚$)&“ E½5µÄÕa 1‹,n§v"®žžgfàrk‚ æã”×Ç‹^¶ÄºÆ ­E.Ç=N™b©hK(ŸÓ"Ç"Vª_,Èi餘0…¡ÈêdˆHzJ‘•ƒ{»éyuŠëXõÅ ^Îï_ð†!«NµÂ!\¤<ÿ·Açéh^ÌQw£nXÖõîoÿb¡ÞäËòoØB_Í3›j!"õc9Ýä;Û3)vôº¦]És_,«` ÖwŒã-WTDûž• ‰þg›àýuÉG_ÍÓv´øsûÚÔz1Lg±\=æÙPPb%¿T ÞáL~&üÏ<ŒÕ定䤖"ŒJÁ êVb‡!^Š^w1­‹|*ä ˜¼U]ñm¦¥î”gW6}rý˜¸0â·á>MMoŒÝè÷»¸v Ãr%¹àÜ÷»ÎúWŸ$måõƒ‚¡îº#^ Íã8-“=&¢(n]ÄàúG"oq9¢O²ç\Gþ œé¡$%1B–€"Ï¿«ƒ`$`c.’ÍNðt kjJÂVÿiÜZ:sC/_I?·ŒÌ½u¢9Þ1 ø9ÚÍþü'<˜É²©5BõZ· Ê×PŸPŠ»d8§4I2±ŠÂycµy§è##4U:ßÖKK@Úfk|ks­Z›i—1/µU=¢îWñ¬òÃØaý1Ÿ~/™J±¤ÐÍãòÿ^M.±}2»7cV™h@寛ÕÒ¯=üýw#”ïg›DºÁ¡ÇÒ[ª)KuXFNGxŽ®Øª>ÆÃÜiÞ ,‡J¦|~°“…ô»ëêö JûS“t)Ç֓ʧ¯¨])õ)ÓT@u:áýµõýŒ¨~ ÿ:³ÞÝbViýnYÜ9ÙÑy¦÷âxõµnó?lý¼A⋈žÉäý~—?_] ‚ld¥¼I>2°×ÂiCšg®7^°/ЋéÚ±Ó°`µ¬ö\GÝ÷¶¾Û·@ÛþB騣ù6“%G#yÞßÅ2H=½QJ¸¸¹T#BÒܸFÜÒRªÂ“e€<‚aºIÉÉÈÕ|´ˆØ8oí¶18SjÔ º4>{·z…ã×Ù“ø5¶uéM Å‘N±ö€@ϳŸ@èé¥2™aÃ}̹²†èïä•äŸ|@ësø±78Œ˜6ñNg[ØJiŸâ7.‘9$ѼxÝÈùåĘAvÞ€Û/ä ¢²8“ggÆÕI]™<ßÑàêô7«[ij©(ß «%W‘>ÇÞÐ7_Ѱ‡8yÑE½D‰Š8"N* )ùè¶ä-C5—Ÿtr),ršÀáÈJ §J¢¡«•‰ºÄ#‰Ï;Á×JC¸¹D^«Ð •êu†0‹œÜûsžã…$y ªJO>×íºüéWšzÝ#¢c¦@`du®•ž)ÂŽŽÅ6÷“N&¶UÈnšØîZOf‘÷ÙrU1 …¸(œi4ñå%xŽ®¥äj¤©AÁŸ1à:è-ñösÎSðÓÒ.Céu-6zä­:˜e\Ç––h¥íI èÏPuÊ6ü‹ˆÙämmŽ=Ž–ÕHGÊqjKäÈõ"_¤(--Û(mö…-"¥™äLu¡ÎŒPJr¤¹çOR¾¤PïIS¬:ß–ÑÊdIG%Rèó%PŽR\kd‡È»%žë…rI$ž¿GLu‚á¹™­“é¾È~v—Sn¬^ÔSÄþÄbúy÷‘BdÉ` z!÷µ°ÇÕÕo¥6IC«Å,]`‚HÉ4c® RG‘Bw“$ƒký:ÎÇEá@Q´hõð°É8¶œ?cÃÍ,xßTž ˜/h÷ûíª:ÍMÃñ,«a­, aðîî–ÎÕ e{ØŠÞôÉêu;Œ«¦V,¼©âxÓw‰0ájë(ÓC#*u´LWØ”ÝËfgG©yí«›Á1«´ÅÝ%¶çœ6PSg’Á&$‚Zhl¦iÔvã†H†y îìбHÕÊç–?µ5'+U°ìšÒM‘’ͬCÅ[¶m}†T„q†qN)˜?Źbp±-1°J Iš"è=2ž*ü±¤Øòøø;¾Fúv¡n@‰ûæ0âtó} :ˆS Lã™!ÝYåSœ1©¥-ÏÜ£IX¦¢¤¸RY ½+Ÿ~BE %²‡PFº]×ÈGÙ`ÓÄ@NEÖÊ»;6Ë•¶RT"¥«…´–¤A$Ô¶Dy–ùFòŽ ‘c|ÛíGùiÇ£5XÿÆ ^D÷%Èsí0‚ ûŽgY‘LØ9Ìš—•ÃÄßS5í $ Û" l0ZåîY5Ké‰[º´’ñõŠ’¹un¸ƒ!LD(¹M–ÊÓ ²ôó"óêè„$C„ +êž“H¿œ:‚ñÄ¡*¥wD¸É« {'<]”ª©á€B@¸0„®"•}Ía`R£$Ù ÊØKøaÎ1B¹¶ ‚UWt+- " †nÜÖ¡ªe]Ãh& ¦ÚA®Â±:$Qš ÖQ™u–ŠâjɪØYÿÒÛ”‹Mc“é[øˆn9è ëÎ7˜Úw(9BÆD–;.ˆs N‹sŒ"6¡ð÷q†µH# HG^¡Ë‡‰Ö9‹å¬Ð–ß‚z‰þ%=4âûCnž’|áNÝõíGsýP°wS{Wê«>Oé škoè‡ä'É­ÍUÐBùe‘-a¤D(H·œé_`¡,éŽ3åg¡¤Ù'[1:AŠo(Éç62¶X$2¦QóÈlu9¸ýcîSÛ§·Í}²²a˜Ûe$f¡®$nn]ñѦÜÑé”h ¸–<Êœ¦ ¤iæù;èÚü&¶f¹)^†µv³2W7$H|è”ä¥U.ÞÜLMç¥HõÅ£¶YÉa€Ô2tÑ#ݨÚH>;Nñ'žò›*Y_Žñ¹ÔYìÛV„Âfu±¬ë9žz‡°d@äÛ=d)¶§0ú…ÓŽàÉ<‹?g;´$†ÔslïÏÔ¦Ú3mDÐw¤¬Ýù‹¨NžÂJ‰o¡å1sÍh£#&–®Nâ_ÒZ âp =ÍLÅ•Cy¦Ù!ÃÍFÐ’"ÉJpÓv|™4BØ¿Úõ5vñèšpôg”¥Ù N{\AA‹„QÏ9“Že"XèVc‹~yyh€¶qÛ"A_UV1£zŠØƒ¼°ÙY,:óòH3ǵŸ ²ÐÖµ+=&”yT1Mέ±1šwZ\4©HW¤sy¦0m£´³ÚʺV|OùyÕSTʸˆ  ±Mg¦jM¡R™ = Mê¦LŒš“XÈóåTÌ$B5èÃ\®¦» S#XäcñØVÏ&p·©±‰®œ4 #æxé-ÞïÓ“K¡ÍªvIñ¼{qDø_ÿ‰+°šDé'À:žº¥0…Ëq·rH‘'V|¥ÏºTJŒš€Ó*u.c`ÀëzºÐ69¥†/»Ç3²v·­„‚;î–þϦ¢ Mü¬uy¨¬)’æ[ mùŒËkÜ‚9,‚P@&3š†þ.›†FJ±“þ†Ìë²Y»Ü#(ÌWÍÙ_ßt']°3S€[µ{Ud¡ý™¡Õ©$£¼ž»£é ÏïžïЦ•&¡4—ž•ЦþzÒ–úÄe<õ›Î.y” +Ùçgj DÉ¢t#“mÍktã=ÝPѪÒUÙ¼Âê9{D÷Êz;½6÷žº!r_\/¶WóV»):zÈ:‘ 4õ û›ôÊ—oúyV•!xR©†ÀW.ùªìSÃòi“(Þ8/T·'é¤3Ÿßò˜^Séò*v·+@!JFžÃqrü¯n=‹%ÛbL},¦õ •á÷ƒóÔ o%3­åGÅÊ8QÈn”ÞÒµõÎèâ9L'9ël¬Ž^-‹J9ÈMÛ£Œe‡§´e‡Žpô«MŠ7¾-¹Ï(Ê1¢®Ä£wV‹2e.÷deÕ 6Ã쑦Cr.º²¼tÆ¢E3`HÿoÓ42´+tÍìÇÖŠ0ž;’8Óƒ}å#_Îj·«Ï@jBT–ÁD¿»l¸BW€ÖÀî©Tšï·Ò@|m9äþvr´ó’Hª©y±±´òÎl•þ¯eƒ,9Ø|~ôfêÓrºýHçV„Ñ™vM¨ŠJF£­TO43¹:Å*Îrf­ â½oÙ×…*õ¼µU)Ø–.“\­U-‡eñ¬Å|)5žë€ÿh­žIñÿ•\¤£Ë4…7!uÝ–Zúkªõx2‹ƒ¥õ´„Àôƒ««Gߎ¾Öd÷¦žø}»gÀæÔšsû6ý£Êö kIÂ3¤í^g{ä*àÞy¼—“«­É•’Ê+Yì ¯eD,Á•nW)ˆ\SÒÆÎŸ8màÌîWŠ÷|z l|a´Ç€0öwæíšvÓM,q°Ž§í1¶cÚ7ô ª·‰E“ÕÚ:Ù%£„*Q”î žc›!'S= rY6¢²}-–:aÖósZ‰s”„„,ýÖHIF:öÛ¶Aîþ$胲xÌcǹ£ÙW~¿Ì·µÉ©\à€¼D9ç*N&~ô–¥i×ms§œ~˜Ø³Hð9ÎÄe¸7޹²7HÒJlº} 83-²]쉦™ºédV«–0%ç78¢mÕäÇÑ'šz(Á©ÂºGãµ ¤ Xn‚וQñÆßRÕ¯-)eçZ,ß4'„Û9²Ð£1«¤eÉv±z,HÆÉÜOsn빚ܲ™íš+ÞÀ~iú×ç¤éÅf*y7¡4>)­•±7o÷© Ò…®ð}ÎKü::ׂï1 .™‡«g«ýmċۤ[¾Á_wFÉ'ˆÌ=k»Íÿ‰ÓTŒ@¤ÇªXŒŽpaÀ!+‚ä‰Ôqîq>‰ [M7¦òÍÅf•ïºKŸÊ»A,2âòâVf¿sNîÖû) œxò|,½ú)™£ò<«0î[ít,Eªã–ÉÏDy©×¢ÒR5š|vÇB$ûk£;-K¦8 ZRÁyÚª> 9Í ÈÈÆŠ- Ýß_Àÿ?s?–î4åŸ!ÝÜBi…K÷Iºê’ˆQÐY;¦X‹dðÏúa9Çw~ë«üÙ'”ØÙißî×_ÖןWê^€Vn¤vœšÍÇ  yIµ*ÙGIi²ãDð8Û­ }„*¡Íî¶ì†£NýKתì>ˆ§®<7ÏrsÊÎ,.šöÚr¯#öÇ&·.ÚÜL˜ºRÿ­JZ‹GµÐLFØiü¯ÏêÑË„pzóâHb ^¨ªàÛ‚„QÁÀ…ØsçÑ»½QÞáRÕ&!F„ÜŠò·þ&Ùº—]?Öî¼7NÛL†ð$ˆ:ùE·‡K HasúdãÅåéŠ>‰¢£T HŒ0žT¨ñ‘«•lIEFáIteØÎA†~ž +ÝÍÞ·2÷{%U±0ø!poö/ wÚ7»}KWÿ‡ZäSüÎz|ˆ+Ë(¡íglÿÞ³J$D!¸ÏöPmüÍì>ù ^Dæ]þ³…{U¹Ó‡Är–Öm˜4ñ“|2¼Áú°Ü6ìqò驜ÙªD³F³^–våñM¦Ê¹1¬¾@šÉ\.“ž¸‚ UvJÂ&BaíÝivÆ_ nÿt…ÓtšC±Ö~àànf0/JKóž€c±I,m@ÕKMcé)Ï„+Ý{•[§K’ù?Y¶RKxZÕÙKýïÀñmqµnqøð Žo}Ë—BÏÉ#C2?R‚‹€nrol‹¨H“™ÆLIÈÔ¸‘¶;Öš φ‡å*_LÎõêg¢ÅTä<7§7¹á|סšÜêÀG—.s]Ãf¬—ʶhß%´“¾nO ¿¯œé–æåçm €…‹ª9ý2꛵ÚG¤çÏ ¢<³››žò?yò·æç7mTË4ß½8¨Úݬñ7`€G³ÏáKÊßÌ—ߨJwžÛŽz—(?>_—{{pŠë|OmûH!T\:UqPX»åM×¥Î_¶¬p,c?e¨ž(›csLPÂvËyÔW=…СÁã:í@IÕìHÞù5*#ËÒÉspK…ø,óPþտΖ§ËÚ%3OâΤîêôlÀB›r‡µ'‰¦^ß|1×|øÑ¸œ÷žÌZVã/S“ª¦Î¡Q—F=$e·&†vaÊW‹±”g5}©âhØûrÞ9®‘s€áhË'(5­ó.¿bÖººéiâ{--XþÞ¥,b–9F¾Ý9UI0œ£²´³ŽÏv8Æa‰íái÷U±¯EàóHî•:I(Ö¬¡ç:6Ý\µ%×Y¨U ®/kä»îU®ÕáàAaÙI¯zihÿ»ϯ6…0ÇÚèd¹Ë~mxО‡÷[>Ýþþ]Ú²Gœ8’,#j0¦Ýahl¸Ôb ¯Ž)¤·ººž¹+Ö¢›Œc–õIyЩ8¦!SnFÖ9RCðqÉèÔð÷µAÅϽ¦ØñI6Pñ*}x¼Ó“ó¼unjSà+e]õe†ÿ8<ÌV’L}½‰Yë>‹—'«KÍu²Î}ßàV£ºìVêñ­+*ÍI—MC"ú$í²cL&½jß~þ§…^ðêsyðè±³ÒŽNêl"‹ºªµÍ°p'ö´*}Š¡†Þ ºëAaÝc—ª$$A8¯£‡‹,àþî/÷7¿A´¡s¢Ö­XÁP˜—RÔæ6\^“b¼×h9ºA€®\ó"—ß‹½KY½a†ù’ÙÓGN.Å&S 1¶ï°=_‹½ÜkvBì@J úñY¶»ÝN¤œtçæl«EûDô¾¥ºðaùù¥g¤sŽXySÄ‹RÄu,q¬oZ?9E1‚³@{m}?EA€—ÈÛxõ›Ï :<Áñë­E+ÔÄâ7k°bý˜Y¬·ÅŠóçþŠ4n­N’½a“I¥"?pùW¶óó ¿Pw–UÏÚ$Z·C&1ÙÝLuMóúDáªÁ–—ñ8<ÍœæM¥åþÉÂæÞ@»;Û2m‘£{XN8—"ô·Üùt'-„ÇôïÇ&ðÁ7ù?Á>ë” @ Eiµ*hÓä"ƒLÀ’×)DsWÎ?Ï>ÕÓ1e+,ëä¾uŠÜ=eQÈÅÓÊ褉i:}º¬´~'{öèhÌù•v©­.\p–N³G|£Ï½ÀMv™±²@.ª¼ôsnJk.‹”H|äç_º<ø ¯^‰…xyñm^¢û[åÙµ n· @s´“ÞŸwîžãeãðÏZx‘ë×:ìn× )¨Žœ»Â`×A;èÆÛ¡"fñæòã¨e.áÿ­Ù€š|›aÜð¼…Ÿ2ðÃD~qæªÄ·sk*í#pE~s"…îߟi6>‡N«>#„IÑ Ó""x×ìTËú| ¤ó•.Æ–‹)=ì=Úò\ñí'è;'„(‹ŒØŒYëð‡²›WÕ±á~øÙXØ®]VßCz›ã÷£Q÷Ý´\MÀ-·ˆM‘Ïh$äH.3Tà§3hâçu!7|´ÍßcQv€sýã …ãàºR1­žu\Ö6ûý¾´Ýñ¿ŽÄ/·HîãËÛì|ÍÓ 2ú‰Çm@‹“+ÙadJÁÌÈÌzÔ¾¨'‹»3qÂÁÜT"ÿƒÁ×àŸ}ŸIÔ¿À•übqx¸ï.gLðü/?MSíîï†èaâ’h¢þºYè¨"QÍâ»»×l8æø¯•ržVcñ1ß]Ä/µmH÷åeãv¿¾ehâ°ÒS2øƒx¡+ ´Z¦ý¤y ¸^ž¦¶üž£.)ƒn=ý'øÀÖux«¬òû}C[®ôYJ„Ý銥÷¢Õ±öH úF9q!^:ÒŠÐ ‹_ã}ÂÀy§û¬¬*°ÁZ’xtOYƒ)®»ê25­•"upÀÖ·9Æ7W¶ÛP¡m¼O]þëzj/fÓHë´ æblz“ÝX¬£#âêwLÏxbA•aõ¸˜Y -O}²e‰M‹jÔÿw$¯$Á¤í̪6{ª¨Ÿáÿƒ½çÃù ’tƒçßÑKbµ÷•Ä*šº®ƒvMº¹z⮫î a«ùãTžèE"cí M¤8Ö.}8OÿeºíKÜ ä›2{®5}¬~™öË­WHú4j0Ÿfÿ½Ê=ˆG‚’$¹†æhÐ`^¤?úÐdMãi•@P«ÈAÓÔÑÐͺ¸€šP¨Üµ;æ·E.ßÑm¹~™éÁ.çÅÑÓ?ö‰º=Ü{so‘iú°µN^®D;°ÙÇ\Í_(ø\ìÜgö†ç€G{«×|“б‘ª£¤éÑ£QØDtӳ鸣ò^ŽNËÔ$E³·B!<9ß|³÷ù•Þ{W77 ÷oÂá›>‹ë£‰—ù¿ÙœRÇ‹yÄQ„™I R þfºù¢¨ºŽÛ‹tÏà[Ý>»oZÔËáP¦B`§(;Í<'»ÄåN=ûØCCbçábëþšdîÛ:¯.;ë Uô\ÿÇoу¨üVãµå'nCVšÖ”t[vbáÝ>8à‰¦¨õ×äÑ»kfó/Uáeïjiçöôï¸Ùõp<ê™uÓ[Ýã»}Î<ãÊôÙ^âWŠ;ï“)¦ÛÝùÔC¼iã¨T^ïá•¿Žª—=½%ÅØ€43ñ̉ÔèèØò×ío”7 ЙeˆÁ`‘â2¹ÃŠÇÎòÿUÉàzs|o/ !üÝ¿ÉV‘ä|rá´-sìûkÿžÆ¥‘ñ§€Ÿ6è%LÿðÖð@»«?´ËÎÔlÎî5ÃÑ;bt JS?¨£}zCÉÚ¦Oá:ú<B„#çh^ïÅíc-Ñ5æÜh³;j=nb4£¼ƒÎw:L´»­Ãr –幯q"0ŸfNeS7JNZ+qn¨Þö5jmÁÞí1Fd×M>ÕŸø3m㣴­Ä’mÖÙqöÝ7¯ãÒÁáp©|ªÿlKv/I- ’ºvF¤¨ªEº½dâ%¨1AOvÎO‚Ú…<G Ç7Û“7º>Ú˜ã†{¨®E¨Û•n, -Нl¬³¯XãP•DßÐ)*Ü=g8ŸK­¾³ñ$ßGº^Ü—åBÓe7׺¾öMýµ|¿÷¬/sµ‡£‰-¸ŠòWì½›í³³¢\×éõ>¸9ûïò¾—ß|ñÕëAQý¬l÷¹öœ²yzؘvT׈‡Ðë,’™ùý¸þ%¦ufçÐR'ý¢!‹+&±Ÿ¡$QMG‡·—ÛÚòñ·¶²~œ fIxñE4aliFÖt÷#Ç}y¾ûˆP}ƒé(Ë\»¾JD¨²€F³ãïÝ 8«Ô½[EA¯Š3ý(>¯¡únþÊzÜßÊ‚ÃÕìutØlßñ=³Yqëe)ø ÆŸ—ÚÍ5ÕºJ¯ë•Ù¤xC§WKð¬ÇŠÙßhí¦Ëéº-UûâÒ?…ÙÇÅ¿•Óz‰Ö¿à‰e…¼~Öù¼¤,SícYï=E}u>¦ÏþƒGÐ`e %Cª‹¹Žý¨½§Tœv+O¬› ìíZú³ºž{cö¥#o<ìØ€¢®!Ï7î]0‰LŸU™AÎQ‘ˆv†«ÁUyy3¤Ü¤˜¥ƒEü;,}K>væˆÂèî;|EÞ'lÛÛÆU÷S«â|ÈßIë¶:I³üáV{w›ÇÚ§J‡Ðiâ§-^ÁEà~is0ìÒãâFw}y°ecv¥u·UQ´…šèúYÔhi“Yô+LØÕÍâçmv>™'/æBç¥É¨¸ØéSl½)(Ðt6擨Ý@Tˆ[#¢Þ(µ¢èbjËü5®®ÌU6ÜKÅâ8+w=^Š’y-/&·<§v˜]L>M·»ff;:)í'»äJ~ŠŒ®£÷¿á¯Åt´KË£óÊBp6…¹68`  a=mY™±wÞ–ñ¢œsã¨C§¥Öã‘þZèdFÛÅû‡Õ{íîr w›vÁÕî#ºûÏ•y}–7ðI£&ýM¬«ŠwÈòoŸ[wë£Ë€ ýtÝ 5:!ùYô1¶ªU\åô:ЍC<2h¼>Žãã ˜`§‰ÛÛk»Ôv­{,ý3Ûyu»ÚOß¥ŽŽgÙ%S^29?,å,¾^qŸOØÞ4V«e"­r®Y’ä•ñ˜¥_WKØ„öªœóëÚÙÿ0Åk«SK@›aˆ±,¿T ¶ŒÿôÈ+s *%¡«k²š·5Ö=¯W*Ãæ•õÅþ–ÏÏìé·Üq/év½RyüZ‹Œm|Y….@Z²¼§@Æ•½ðõJæ_Ù/*×I…@|d;„K©¤—}W¤6™ý†íœ}·sé«ö Ù Sr›øGÈþöùiR°FÚ÷-·ÆOoÃο¢vU×§ožP¡~™¬‘ 8j?ÞŸyNýž:]ûÂ;ˆrŒˆÏ]º5¤äè(XMÝ#ÙkÃEŸÓÐñøé±ÿ6e ë¯déMǨݎeS°dæ8]—t4“Ý´÷ûÖ ?ŠlZ)çPLæ™Øê=šGNžç¥)Üã>ëéÜÕO>Ï'vqîiêúJ˜y¤ôº²|â¹ÒüÖ¡y>Rq1P_´bnÞBÝÆÅzŠ×ǽs>öÚ¹Ýë$•ÖÕ-¥%ÊÌ ZY•u¦ ÖñÖ^ìùæ…’ßÎKW òdßk<ï †•·PÚýÛÔju?$¶dÌÒäL. ÍP ÅÍ££†‚1П“²?7&; ¹æO>Ú€e‹:]´oÈß5áÊ­cpåù D5¯mþljçü³íøšÏLVã)jé«à9ÛedŠVÉ {ìààD@èŒe{¼¾U¤¼ý' 3ð}×x9g1.ÞÙêÃÀ­•ùàwœØÚ[jã÷1r=Ö„“gåàä_eÅ'™Ó¦c‡YÍt'¬[ÕþýI³Åšú;#in0ÙJ§½ßPlm§öäÇ,pt­ ¦ªHïVEsyÓ=eÆÿ2 |dœçy¨¡gæ=¿Û«D¶CbºËlÈ”  QÎß ±˜Îÿ›4Q® >o\ºäéxó»*ªAØmB—c^åoÖAæ’•^š §Æ”ÇãI!léì©·¤$òþ»oNÞ ½I(ÿ×¥<•Úœ@Žj‚–¸aNyç-Ì$szWî“&€¬ýèÈÄûoØY¬I¾XgµÊÓ¶Ó”cØkÜöšÍÛ©~û´9Õ½ÇQegî³~÷i¿Çû¤CJw/Ä"gÐùÍyçÅhZøÕ8 žêùS[ ¡}œËUI¤Þvb;$äöß)ÔDtÇ´„¯mR–D®ÝíãÀ¢·Ý~—ÑS†Óª²ì±þïõUV?†ØÍUR¢5!×óç%G?ðØ+öκÔ;â>UÆõß™¸ÛPíï9G;4¦4UR凯š\ÿ3õ/ЗҿWìt{§úÑt<&É+sí¢ÌØ+Í“gäcoä~~)JQÒ!"ˆì9©‰-Rußn묪ԷÏ_×êa_¼Ì,‚UÞô{¸¼ëHôÚë‚ µ(ΧOOjÿ´ýåg÷×ý*Úêœwà&ƒ{ÒŽéQMPjèutô9ÊÌXWwÁ¡i$ß$€ßÂnY&D§„ûE9•=ÒÞ7á¿fxçd/wÞ„­È_æ®è95@QÌ °7í–zûišÖï'òÅýRLhtR—¤ìÇxN»sÛö T‰Ü–ý¡ü$å¶×]]Ý õ'´}eÔ)¯Ø?÷Öë Ó`Qs)íƒòx_—3‚ÎARq°Uºªdñ”-ýy›Éµ­ÿ­oqäýä–+›Š5¿Ïù^z‹0}¬zoì{ꛣV¼eiñ xi:Öt,é4¡=dŒkŸHpq K'JtpH©’«/0é£ÝŸ÷ÊV>vGWœñ¥Éß.SlL7ÎÈ^àUÛ¤œ+™®‡j¸¿6uò¿#a¢î°Ùä±FJµy! «uÏÄ™†Ôp‰¼Š~¤‘äôÔ˜4 ¿¿Qì&ÿwÆxÎ6ÉqUÁ_±Çrá²Þ5÷› ZW|î4˜5°¾ñ©$ëÿ2ýUdß}º*YYùƒnB›ŒVkY3ÿó¼ÒXª;š>œ×77N˧y}ÃWòj³Å.v ¤\ÂÿÔèØ«í2i/Ôjè7óRÝÛŸÙ«e#KØ\™!þVZþfû¢ûké€@,‚†LœÚ%°Ú…2CfLÀÖ| ÍM«’y×ô_Ìiü“ |´÷ÚñÉ„zÅ~èÒx4s°Æ;ˆÙÔÔèÑqf§ÇcÔ.Ò›ÜþûÙ¨#@'Öº;£Æyõ†c¨fÎ'?Qh~çîåy©ÆÜ Àxy¼ñÞ [lÞÎè…=Í5®f¸£oøÂõNwжžÑ¿YèTp"ßÛÎ5[A¤w¦¬\ j EFÔ9Ú!UPÜJ‘5r«Lš˜ûl­*Þ¹˜à•úInç-Òâ'ßæòÜ^íj'â?:ü¤þu÷e6;¨ÿ[ú¬½µGæ v™´E4“á­9¶bG§ÝjMkôÎç „$ñ­î²n^ m‹„¡ß¼Ô½´N*?¾4’¾®:ˤ­}»²ƒ«j1šKÛ¢íÌj›ÿ0BÕ^1?{3r¸^½o/bâh×û§þ}Ü,(Xùû$Õíã6ã`?2-ƒÊ1 Àg”ª,¥ü®¥œ…”×Õ¬)YÐ[Î"ia=?ûg™9n)fÞÍUø~c_…BÀ© 7®žû‰½öcøp ™~ˆÌïÈù£öœ`ÔWÖE±n(ÀS?úxÆc;À¾ÕN¯“˜ÇRGjÞOhE²%’8’(בÕgGr¹·ß¼øé ž_õá×eËâë}Îz|õ¿í›'ú?»í×o;}÷¿fAñm­†çÂ`66ÌíQÍ9ê)I)•Ú·!€\`ÿ¨½_%E–•ŠVÕ& }Iþÿ'ÿg\/zpv=®ñ3ÅØt³[ ãèlvW@òȾÙ$Y2TI†º œŸŸç v¹ÁˆÁ»· Å.´%+ë"ùK½(¶?'±ÌKúÅÙKr*Ï,ÇýË|a—ý÷º‹F¼é¹WDÙ6>çF! ^1Qžöq÷goÓ‘é|Rl ÿËÖ8·ú§ŠMŒ™*'”mIn¸ Ы{ÏõáoËøŸöFóæÕÿœM÷pc³ÞwtüÝÂ*&¦R™F)w%#y .Lñçe++ÓÈûGÁ­ßˆO´äOuŸç ¿¼ºh"¦'>ÕV–›0ð§45Ó|Ì]t”­Ôü×xêÚÚ¹‰höâ#ŒB\°îo#û+iˆìVÓ¶&7s¥›v§NMO–ýÊbÍkՔϣgwöûcì·Â1üÔúËmÜ(Ÿ›Ç¶ûþZ‹®n寳º¯Ulµ-ª“¯È+sŠƒ›‚8Q`%õÉ$ýާXK=ƒÆ:u¾‡*[ Ÿ“ìÛ²y‘8!ì©Î÷>Þâgè:;ÑþÀ"‚·u=ýäÔzÕ4ó™fn!gÚ0ˆWƒf<Ø)ÆÏJµÍ11ïµiÍßIJVZr{—û]tO@$ä‹’'—ìw+³©—ŽZhrâ–÷ç¢ïjØ«§ïZ¢OÍæ¸ÚÊ—, ]ŒÌYMh½’Uøk'®r•tS§7 ð²RD» ô¤üò8dž^êJ@z°“«±X(Üa–?›J„iΚ¸­=ÿ‹û?È×üœçƺ×m)µü®ŸtÓv’Œ²z1ÐÙdMC¾­Å®ýžÛµçÌ7d œ‚Ô—4åŽdœ—q^•!?‡îÕ+ ÊÁõ£ a”ÿy¹µŸS=åiÞ¹~â/ï`‹AQ:e±½‚pZr9†cŠØE8«ÚNˆÛÜ£~RUÙ½—‰õý¡?“*P—Æcjöí’:Ûç™PÅO|sns¯Ý‘jÔc¢¬‘´-Ì.è¨ÔÈÕy¹Âäá] Úšã'ÙåŸ\äújU÷wŽ2ó)v Šhª°^ ™LúÆÝ¨NÈ2ZâÑ;iÃÁgòòÏOü*öQž2Íž7·ô€+>‹Ÿ)ãßà'ù³%ýLÍníÀZç>¢kÚj“ß):bœî(Ò¤ªZÔÛÐÝ—´u/D¤ü÷¼ÕSìëcôÜ+Y®=Š/)Ò®–<‚¬ÓI<>×P³Þúç_(êÀ N2Ÿ‡eZó3‡a)lÞàÐõªªC>7]óBÉ;ÄžóÑË&ê÷,/f÷‘o°ªà7ó~}jÕLþ—E}Ï*`¯Ñ:ƒ•TG íÅŠz ' CAáÛ¾…öiIWcõjIéw—÷„ÎG$òõ¾~Ö¬R’Íÿ_ü/1Ø­†W¤ÓãÿžçøãwÒÓvò8þÓ›õrºkR.}‰7 #ŽcW#-–ˆ£šrlH$º]rv—LHÙ©+à>ï[ýÁøäÿĺ:ýggQÆŸ!à—b”í¯+\ø/4oÊ 3H‡——Aƽ A:xÐ '%ª9±rÒÑ$‹ÍTÀü†òŽi:ù‰¤Ø ÝZGs§“¸%TÞɬô'ù<~ ‘Þ¼Å™×æÎ}޵¢º÷è-}8œÿw^;׳ù0å¡A`emÇ:… £ç6a½¥9ºõîuÎIo_G3Œ•+A—nãÿRíòÿcËÖë7žòê‰àoÞÏ£ÛìùÚÌêÊ]*w±ÐÓÙÁUOÞq¿ [~û¢¯`%éí¼¢hÓm©Å~u+£nè“¶rñƒ.¬–ˆ¸Ù:3oäèÐì1%Ï¿jþ 3+ªö©_þ®œ5IMÇç[e—ÜÎÝw<ŒþWÔ÷—HU;ó–>&ïT 3½¼ÿ* Ð¥œ¦kWóAjK« ˘dÝR/?IÏZ"®–U’úØã Ôp[¨Ý›_É?ƒÃTfË!>£}õÄ1oem–³k+éƒTÜB±Fï9–²’i%~—kÔC.ì \3µQ§)ïcLƒ{Du"›‘>Ç„ÚÁ®J-˜ôÌJwrßWYè]?•ßxgøüú>»HýõÄÇ5Ê!××wÇÿf-eþÿr«¢»6ìYëìuqft\Ɔs;·d§û2¢åBjÉ¡²‘7¬¸S{úhWãÛ':‹aÖ}Òäº ¬"ÖØ{>ÝoáÂwÀæi– Ð zJéǸrÕÜŸWJz}y"uGíB¹×SƒRÀR£©»æjË&—»Q¾üõyœü›Ûg%†X²ÝÜÛ·NH‚²–Ãb¶SÁ¯èÉ#Tc'Yg]ûízu޳ -MäÕ6âVš!Ë”öºZÂÐ'dÿJ—6×L42¹ùΑ²³ŸÒg qàâŽ&íbSKZñN[ÙÔþ9ın$äÌóg¹/q)ÙÊíw›qßyÕ$Z+Àóœ AŒÈÌ൵Ö_sR*ÿPèi"%¦ñOÔ©WåˆÎAÑ´)õf ¨Aê‡Þ¥â/Tý°ƒÐ'ƒI…8‘môÚ•šë „Ÿ7Æ£P&UHüu,™UüSó¿Ï!ò`üÇ0mµ'ÜQt½š×ža@“,.Öšf§y»›ÖQ:·ùÖØTËX4(]ú/X•Ö7›ªTÞ-”Œ¼2Z–rö§laÏéßý×#£‘¬ð¼ßöܧüáÚJG%ØMG„-¸ËƦc™ t¥ 8¯Ò^PçFÕ¦ÍrâTa‡–|ù^jª-¾A•.D&á…ÄÚZ¾‚êñOî‘€%@¢ëû69øiÏIVÓɶ%ÈdÎqЧ‘WaG8p‹ úÏy¾R½bŸ.Í·üA¯‡ú»~Ö:IÚ¤zÍ>‹yEHwÞKßFc1J.U®­…) Ñ‘÷ÜОã¤Á‡>ã˜Ó!ÇrÝEË<°¦Ù¼˜‘‹*œRÒá…å+qn½v/‰&ÊÖÀo¿=ЇR¢¶ß…ÄCÍù<ïïIÚÆÅa(õ_P×´OǪ (¹õ®Cщääj9­ÌßdÓ%ö†b¢ÌŠ=Ýີ¸Èû£uËéíc.ò±¸Nµð52^ç·‹oÚ|ЦVlãÒÅîh!ߣøþΑÏ,iÕ)7í·ûs“J‰ ¼úçæ#@¬É¾Û¿{Œå²g¡Ö[qe¡"ë =n{v:ëæßÎ “~§Ë=±H\ln§GmMc˜Ÿœj̽Ğ¥'<@‡\óqfçS’ÎtÔBzséÇíOͦõJæ&¦ß¸4ÕWC¢ßwÒË÷êç±’¦[÷2µ­k\ùý¤¯Ä?ôb%dÝðð²å¥âvITMjho6\:òc7ZË3é“êjЄˆgfòêÝÝʪ ×|”ÅÑÃ(M3]ÔøzöJ1¯ê.çï›ìÁâuka`q¸;m]žÉ` ÊcÄyÔÎÑkS=«FL.×ó\Dê¿þ}¡ÝÏÙ| Öª&Idchr«¼W‹0\É¿ùÕ‡øºþiü»xIVÆî@\ù"kÖãø[kmþõ™ÆŠé«T§9|N–”Ý&|ÚË5KUâ|tSzKq‚ê0£´Y?Ó¹òñû2ÓÜÑ÷¾}F¤Ç%búø ™æ3gáß§žØ°ÄÌûZý. ÏÜk^íßt½>=^ûcjx~£ÚïK¯~C3½À„kÁ8åÕ¼1ûèU¬YÀjLî½M¬‹ºÀ%uœHùW2RÕ´¯jÁÃA%¡2·«N‘÷åΨã§ÉÑ]ßÀ˜?Ðuƒå>”wèÀcçëºn6ý—¼~«Ö§í¼ùÃï}—ž…ú%‡}Òw÷Üsøf#( s†Âô`'’ZOCzu¨¸áú<Òê–Ÿ{¾nà›ªWö(×¹ËM~¶û$ɘäÕúø5=.\”m§oïêÑL¡o­)Ï^2ØÌ‰]¥L·®µ(>˜ÌÔŠ9'¨WHºJ+ö‡w¼jfÔ*"Ư ¯uç wôWÆfÚfës÷Üô g¸ºíÀï}ÀãRÑš‡Ea n3MÕy½^UI[s/9[)³2:-$­ùU ‘•7l©h¹ä±å²7ˆMÖƒ»t.̇çYÈ;¶kCÂ`ôÉÙC´ éÛùޤj²U…]d´¼Êyy„è¿FŽÉbNße«øô=›»úkœê:èõSL|î¸ Œ‹‡?äBï`ãR]*‰rç2¥ÓУB˜@q¥vœê@„6"svõO6Nßóx¬Kò³%³óËömMêªÜ¿« Nï}Ù¡#ŸÒÍÐtZµCfm¬¥ûŠ´µ4Áìbì8®²%Ò)¥TÖÀ²GÜÿI&³pZêV5“5wèõZì?¤¿‹-Gùèÿ\.޶™9Z»ÌÑm´7˜hZ$TúþgšêÚÕ§P}šU½¡àÎÍ=<™ÄݾY<)›E3ýg{•MÙ¦´Æ¿0Âö׋T¤ìûä¥Ù`c7*ïˆ@|§<>šæœŒ\Ý¥*@Óe\v²ìáÊ·‹žÄtyí¥Ò2ΤdÎ_t=ñq˜– %ë²k‚[+;6å¿ Zý;V§¦¿§ë0†{â1\Ñ®·*–à­Â´=íg£ôµ{i>¶¥7Õ|–U?d|©6^„kÜmïçb‡ï³æãÿŠk¤øûÄ’îå=¿ ÜÕûqj]´Ì’Þ[,Hvì&Îu£UǺ3­åÔH^¶QÔ­kL¡UuÉEë=¨>Á¥ÏÄê­ÀFá ?‚ùÕÓÀœ~è,JpEÁ¾)Óé.Ñ#逩… Tõ²{(DxçýF;ûãU«·Í©ïûshEªu‚TËMƾç"ÑK'E’Æ­¹2ýGºvÿ—µÒ„õ(`}øÍÀCX«·¶µ—g­(v €ªn EñïÕ‘kªŸo¨´Ì°ÑPu$I]Ïxpw_Úz-ÿ÷F.]ã«EØšÜ6nBÀÉÝ×¥QŽmÞŬ뒑f"Põ4²6*ÄÛ,“,?ôYêÓ› 6}eÞsx”y-­Q„-ã;¸S¯–‹+ò pŠYâÈÒš¦Ô•NE¨Í ¼O”:h¦‡Þ•º©uåúI5yY7¨ÏÔ6ÃÎC¸fyKQ:Õ½óíÊ’ÙIX6(l“X[¯›F9ýQ³y¢Ç1–ì”,F¦I*'Ȉzë4æùHÕjÒQ×ȸö n%yò^ò*N›O=Nñ×A#ä‡)QȳO1 yxVx ì "±† š2)㶬\ã²g¹än“ÓeÉ(Oků•M;°4n–¥þQ}Ž…:‚wvW³U-×a¢¼ä“ÝïôüËÞ‹x^qSQ‡ñtÉù¹r…8}‘øèö‹WŠ@/­¾ñy9Ùme­zÝ©‰´°,]t=êtøö½RG”×bm8žŒåƒï.²ß ÊFC½¼Â”•oó;< yÞîQÿÚ­¬ÝÕªºQ–]$€LŒml“ó†Ü“GL˜SPÁl¹(TdÄ2#¦ŸžüðçK­ÏèèÈw­6¿¸ßí™V8Ãíì´·ÆxØ==œíéšÔ$C ‹,õ8 ¾YÕj%×zkËb‚4&ÉÙQžåâQ¦l!ëö› o1ü‚¿c=­Ï³§Ü¼3þz–•Ÿwïøç£Á£ô’÷I2à’ê~Ì»dއ_ÞºV8l›ëDM3Ñ}çoãÜ\KYéY /éÑe ”ʀѯ/›®ÛÏ :ïyêò6>)8ÕŸvrú8‘“†ë@\mq^¬Ö¼Ò;¬g+É‘v$çSè"áh¨MlÚ¼²Ï¦c!/ŽFK ¨Ç'Ñî!Ž?~÷ßíì5« >>x¼9±¼bÚ5plnÀ%oéY«ÁnâÆÛ=d®zcÑ:*tÈUk,1P¦#šuM“»ÍE  Ã¨5wE[ÌÇò\àz;Õ:¼ºV ü쀮3Bowº–ô³!^ãgdUv¦˜¤’76F§YC?ªQâ¬ÿõç©WuE;szö®”{¿>´™ O{ÑOãöU¶pÞØ¦û’ÏüpŠ}^i‰—ˆ“r¬c/¢¸†ÀÜsoau@ڠܧ‰ w鲓2M§S۲ܨ!ñÇò)(°œ¯²sPâÿ;‹÷îR¸ç»âèÓ…]î)“ÉbÏtÄBn;6¦KmŒY;¢OuÙ›j«ù=ºÝ? ÓA7³£„08o1 žÜ·™‹£®½U·h¢I2O 7xòN€‹I;öç~èšáú£¦læKPŽBìÊ„óh¬aøX=+ŽÞöLœÊÕ?!mÿùîd¼Œ=Zª»ÈÇßÞ˜‘qƒ¡Fƒ‚Ô-3¨TÄÔÔÜŸ}¢DdHp§TSJÎÃŽ’3§MÈP “S©·]Uál³&~Pò~9ÿ’¿cš·Ð…ˆb>ŹTípc†ëADàFH3å ÝNæ° àkrå)²ŒþuEQŒU*×däÆrƳ¥ayÏîiÉB´uš+»Q}þ ˜$©:Z.²súí.Ä‘V4àCçM F´?$ÚÁɘÜK¥´Þ– =ÂIÛº¦ÁÒÏqšuÈÛ•×åÞ˜nzO9 $â•´”æ¶/žÔp4ÿYÒü|¦ºÄ™»7Ü>Iˆï‘3´çØg>3Äqô¥éÊ®’}Ï1¦Ç ü¼›"ïè…tÇÅ5ZÕï˜@ÎÊÕQ= ¯³9,'¿1UT3ÿY­óâ-]Úµhfþ²Áj3ϬЎΉâÒKb‹ÍÁÔÃp_kM¸×=c¶«fŠÁêU]ï8Ò¼H&ñ‘“6~™˜žW£¡c‚ÑïÿÎ*mRM "*"v˜w9ÔAI&þæÜÉ‹Ò {ŸlÖèúí»tz^â¶?»‡H?÷¯§™ì7Uèë~´Ë ËŠ:·Òçw{Ÿ&Ã+¹»M>¸ 4Œ×R­ “$i’õ1ªb-¢"ž™äãxöM¿J'î4‡ø³{_Ð}Þr®Pcûx°Wgø6¨€X3CCªã;NV[G=ª hɇ„Þ½Ö£ ¼‰ [òD0tÛ¤–h|׉bÊ´>3ȉS`³ÃXmÛ¦bP2ZXÃ?&´q3)®½þ*¡Ùð³¼Þ‚ûìÞaUd©›DD;Ÿ• †0[¤˜_k0ÚTõä4ûl>¼‡WcÖp_øµmõAŽ%à?…M¬‰‡€â+Öoz ÊV¤.³g-¸µíp¸;ß^D‹ö³ÿJÏ}¡·e_ûñíî7bÜoæiÑò  ó”‡ÏÕddŠ’ºNɺid®”£{ÆXÓ1N{¿ošÄïæ–òÄøÔ¿óŽ·‰k‘çüfHÌµÓÆÅ7ßZ(I1ò[ƒ™Î©ïΩ¼Qc`:'.²±±«1°)õ,>^üŸbþ—V¼ƒSpï³Û½Â9Ê&ì½%}lm W-òêMM`(¼“•ëqW=ôÎßfV5¸¢ùˆãkGT Ípü¨wî~M^ω}YÞö6þêXuI¨õ?UN¥#Y=qúoÂý&¹‘ÊÜ\Í Ö¯«ß£%(£ÌP)ûB1T6²PmÎ=ößæ9ÍêÇ5XFމ¶ñFä”òQAÃZ^8AX6>ã¸.T£`îG6ì¶ñ¤Ö¦":ª8Î_™èˆjw§e%tøûÆ&â¹¥ÍÎÂÁ‘]¿FÍÆÀË¡iØ™!3’[" “:ª‚MÅ„ås,Í,^ñ×k²^²^?“*Ÿ™bB‚iá¿{åHLuZ®6ðÝ·Skþ¾m®$ê=6QRÆvq.EVÝuí­edXñû©–Ýxã9$³‰2}v—Úó¿-eÇ1íë@sÿ¡ÿŸºM(Õ‡!‰]90cáÅ4LÎjJqºÚ:}¼ÿ|ƒÅcz–£ÐeØÌ¡Îˆòò_˜J5ÛaEp2=†kßuUÚåNá³Ëq4ÒªÕ¾šÆ%i6=;ªËÅìãoxÊåÇ_††[ÆŠ¤'*׸Šÿ=Lœýõ‹H®·­†½®§ó+€¶óeMô¿ì϶X±n𡸺v>xç$è¹F$Û¹âú}¼%…v°Ûg[ ]›ç¼{Í&ÛÜ–žö³¢Š¸ •e²T@Òquv‚òšªr¦¶³¦ãÒH;‰¤ŸÐ®ÎþÞØª+]Ò•œØ@½*ëø'ÖýÞ/ï[Qó š9N¸ üùÔ‘ ¹šÅCÞgEà:3nL¦Ü` )HK­Vàú¹mã ™Ë(ŸGç_­H¢I³¿é/¥?ì€l±Ø‹ÅíåYÊÎyÿ1¯ù•~¾ßcš»DõÏ‘MöE耗•kPïõS×ÍÞ4öæG¦æéôzze‚~IZe‡J^5ü$/ÿQeæåþíeÇkzΑð[ôØõõgÖÿ( ²×O¹'€ZY4†‹8¬`%ñt̘G<&Ú/¼¹¤ÜDi,ø]c(Ké-*œI+¬<¶ËnÓ|j¤°"þáw ‘ã¿FÐ¥»þ!Ú#á¶mJnm\ãªÃÅ"³-L²ëÀ•×GYË( z®Ü™Ú à<9¤†w9¶ÈáöÈ×»±£0v½= _ ]SdL¤$ñ¯¦¼âNvÖšs¯C¦Ù™Έ˜Z&yï Ï_äU½tm ð”QÂ)CetÙ있3ÿ¥qRÉs§N/Àm­Œšª^ƒ†_˜S?ÉÙ=k_£aX¯Áâ°e±XÏÑka`§œ@S¥ ¾10QÎmÅ8:™Cô /2ª–îê”»u9f‚l½þ—ûL')`¢·v×'»A2Õ³¦ªþ)—ÿ¤t•}¨€Ú‘ ³ÕÒ M·qP’‹„4Âññiü’?¹¯±TìXIÿæÙ«fA®å¦ÈjÚ- »™Nÿ0Y÷ kº[W²ø:ЄžtŠA¥YnìnyX™œYDç g÷k¾4,1f©Xœ‡5v H³Ý`Õ|®ÄlLTÚ§žiÈmÒ’>uÏÉCîrD¹af·{§ÒPãÿ›ï2¾Å§ïW/‘¡KŸMÂøo&nóøm5žÃjk³^{Øã”_çÓ™¯È°ÛȧJ[»Á«¼øÜ%ì¿éǨÅ3„Ù—§ ³êD%1•gTYò!þ> B·–0_ÆÃ“}ÐaÜ­ùríã0[í9ž?|¶ÉâþëªÍw–ræ''ñqg+»Ì²@+VO‘^Ã{IU÷—G%)æ%)Âd)\Ïéâ½mh|²j³ôßô£Éýqù;¾w“Åy]ã|GAdúã&¹cm{Cé)ŒRÅCb€G0ìlKä–ÏWeû¹¤VfŒÉžGÿó‰$‡«PZ2ã­± ¯ÏŽŽÔmÜ´à(Ó«l‡yÎnIîžÈI~Òvƒ•N¿ç¬ì×à{åbŸ¼™‘þþù8;Ť¯kï{ÍÞ¿ûý¤U²o0>úã ¼ôÒ¼*ÐÒP¨éB}4v¢^ÙˆSÕølÉUÒ9àºÂ[sþ>NÖ>¿«þàîܧš(w[f `óKO±Ô8?£F¦Ç°þ6&·Qç T µ?%Ñ`ÅÁ5¡Šöågì¸5Õ­€++µÿzW^¬NY`&V‰Xÿ”5¢gï}dR‰É7Ã䦖[Ͼb;‹J?¢¼%J™¶Ïïþñÿ£KÚþ)DÀC©µ,\Ѥ«~:mßF¥t预hÓ\µ”§)62¸‰Ç¬±S~¸­Ê[©Ã&¥¿‚ªúíq½² “âäf[´I¬|ÿ[ :¿YU¾Û°Î¸jpë‰Ú&'TíõS‹Ë-áÖDWÈ®óÏvt*Vdr(X"IuX4òN楙µÞ?C1yò®2ò°<'7˜‡úƒENÚW¯Œ¡Á’Žf^ðGF¦Zøò¬-)¸cp¤ÑC<Ü] wù®?D³{Þv4S•#ÄÌ.¼¯2‘7t§”B©[••šØ@]9ï‹™r&‹õ=•(î¤O>‘=® ÐëêqÉ­–x¿3ž­.$‘ÌmßÖ'Ñ#úfÿ#"o6ËUŒ|H@o¡¡¿ÕˆåòkT.5€ÿéú[·,½?„ˆÂ‰5“û;‚Ò 0$üøZíµ‹*ìA2ëÀPRëb”Ñ¥Q_•oc£g–O‚–IX…1ÒB§+œéYö^èð4­ä.ðe%}›Ç¾›·å Ýî"S'äÚFO]¤_ÄpÍì"¢G˜ùîÌf›Æ3èœyK¡­RR6i~ødi¶ªþö›£·Vš†Øó›&áK~¦ºd{©Ü[kZ7øëlE_ãöñöʯ±°1÷A±Y’$ÉÉ^¡Á'¬!V—ª¢þî2(—íº­^Þ’ëðåÞ*Get@QB–‚6_„WZ§ºãàK€Ð€Sc¶8äÂÙÄó”92*7+ö‰a¶_ð–¡EX†ùt© c£—½¹–G5Ò<­‰8¦eÓN†×üœÆ¯òCuéÇmí+B~äåTfÈ z¶‹G©FeaࢉÚÞhÿö%¼¡¾ñ*ü×Õ­Ïþñew 7ÞN|‹ÿqÞ§ÓŸZõ Õ­DÂÇ­W>1²Sû_¿ã÷óè&LÌbøW–á„98{sÆ5²z¨6¶ÏïívÛûz\>‡]Ÿµ·vœáõÚÔÜ|'D0›!iZÍÉ_GSO2OGã~)#ÞoOÈ(®kW¢äÅØ¦ŠÏpo²¬¦±ã=a3©¬Ìor"L¢TçD‘d•&C+±æQ]ëDHoìÏâf—¨ÂSl‚¡§O=ßÔë”×8¯Ô™Æ…&Å¥ÙI·°9ù2°Ž¾aï¦y¼ÈÌM·êçÉ×ý¿Þ²‰×¨dæ¸1±U€˜¤a™rb§Ùø]º©ÎÏ2s4sfÝ͞Ƽ}ª ç€à§Ç`_M•a〕6E72Kz‰Äb–ìwèêSôJïßn%:~‡£ »;6ã0tX•uô $Ïcœ`´P)Z©ŽJRˆó Oéì¦n½×¸­ä&5+k+·^rRå[Ü-ƒlÙ½)ÞR‘}Ž^Eà»_çádô}~º.—VKϱìºf¥Ö5Ý}i< Uó³æÝ3ƹ±3Zü½ä)ì,^-;&a¸ò›*¯˜­1Z¯O?k·ú’Žö™]¶´al÷2—U$SŸkMÿ+¬Ú©ô¤‡dk³*uÌ:æw½±Ø˜ˆmkg¶gå}üÌ¿r¹ûŽUõ Éí[ 4k¥©†fgçOv-s™Qо]ÆR6¼Ôæøž¡»…«í·Â‡YD»}§âú?èu ¤åGiH‚nÀ@„è³KM—¿ÆÍJìö«´€ÍZñ59Ó|&•ZxsäÀR†\JÛá{ ·—OF”RLj”_/íõ Þ¶nM?Á¥)Gµìpû„W×ö-¶›ðE‹:|¡¾e}’dg÷}=v­§'?ÒQÔžÙ°],¹üH²3ˆQƒ\"AdJ™›ÅF$.ÄF¦c˜Õ'‘‡G¨õŽæhxÖ>3÷L·Æ.@RH î»û^ç ðÝ Lë¶ÞÊ(Ë®?NÐÁøž_ú¯üò5”‰5Þ { ü±Ð‹RMÀ%Œ5”nžµÿKM'F?røØ×ÅJ–·±ÂÜý›kéôïêÇæÒÅ0-€“£f®]:8³“’T)þNÀùL–rò³a5‹g{´¯ÞèL½Çmt†f$NÕo”æSg>Õ~oWý¨¹,nÄwnzP-;‘–<&8ÔâxœÑ«Á³ 0êQÆ:]›\cù›(¹9×kŦ·Þ×ÖHÅÜÆ$Æ»'[ ½žDãÛ²|òâX·á¢•½ŠøwXYtöå‰n—ÙVéâ§ü›Ø—V„^¥C!“$‘Æ câ´ A…Í'7,ݨ.ýhDÉÙ¨¬¥ãºošËzñsŒ 1Œç Ë« —Z¾5Úérë - .Gfêä @ùüžëó*‘šÏ"¹²ð‹4q¬¬-5£»`(,¸2ø…6›þfß÷~KÞkSawþ,—ès¼° )䟳 ¶-J¸ÝL–˜•¶lØÙÆ^‹¿£œÆ©3øÐ-«pxzjÚ¯O€¡);r¥(,Ÿ§Zè q?½Vk¶å`šTú-¿bµžƒx¿ÓOLðÛä¬Ü #³2ëÕ-Ç^1¤Û‘åö±,íÛÏ/Ç™Ï9¼ƒQËþÕ¿¸0öD$Á2`á3¸Ôp£™èCP‡§P«ÚÕÆŽÑAâê~§Ô¢ª‹mxl€’ ¢‹\Ì…¯atq‘Ý×»-+=e^Ǽ¾+.c"eµý{è¹úD¾­ùájqÓÚJDò–U\Šà^Ó,`¶(ÙogJò3Q7«Ÿ€¬˜ÉU­T¥Ù»,`g5lç/äÏz7»hØx¿xßGóäð$tÁ4»æËde†pzS‚a ZvP6ª]nöPëÂR®vúN™WòŸo‰xð19²~³c1j2ïª70G>Ð:<œ/Çó·¡½c{↩÷yظ‚A¢«¸¿]ÆÇí+ýÛZmÍJy¼³f©K·'þKôâôÔö§1Ïüæc<*ü¬Šªœ†“À³ ÉÒ’çªô«ÊóGt˜ Ñq#}v¹›¦‹²ÌI7þ×bdLߦf6°œR3]Vêç>{fdh´Ü¨vê¶J…™ñùº½‡¾o‚³ëã>úªÿÅÑ¡ŽL@‡ %z¦;Õ§aŠ´f{:7<4ÊfÆs/YɆVá<: ¼ÖØÕ}™ð é5ø®ö GÆ?Ħó–Ð︒"Éúùå"õ­cf]ßv*z€-qLVàg)ÙK+¶õªlØäíbŸK¨gT«¯>™¢æ˜ìÒœ=Øäy §ÿÏŠÚÚ•À Qåôí¶ÑÂô£´"é÷ÕŸ\XãÆ›/½äŒua½•³4Ó%íø;¬±Ë§( æ¦J2ÇT r‰(ßeÛY£µmèÍ#Ñê×ïý,oîÊk¾õÔ#å‹,'j)ˆù”˜çuf•Ë[_ k~—žž‹[óÙ\¹#äê˜@]h$uœ¯¿üÀͧ svjÝJw"X©šÚÍaõ××òÎÐÙdyÚ UT¢|w7‚ %kü/?¡‹«Ë ‘WgZ6%SaÅÒÌme¦¶Æ‡Yíô°Ôí3·r{¿OSG¨™Û_b~Ú®w-eÌi̮ãÚÃT:fnŠ >½Ž¶Afíúe¯f³דMTb<íßµ8ÇgÑ-Í3)¦|by…žÿ¥d½Ä.$;£¦`èì€H`~¸0º5Ÿ…ÈBµq^ —o·œÖ±IB{¼\\ÃP/\ß 2ãH ZK.qÏæµwoöJ@ªµŠË.]Å_Cšœ¸»N©QÍY¯Gê·0jLç÷_HX kN”¤ãaô葳Õ醮ƒ™ó-ûW6BÀÊ~^¬ej²´¹EÕ–)¬Ñ;¯Ì2lЫIÒ]:ÇOõôýÕ9ÀWÅc-)•ê8=bßIâ.rùê|½ Ý®Zr‚ßHÁÛ?mŒ) OnœûÛ¿T!Âk—··±C)M`ǰò a‰&–&‘ ”g§ã/¬~»fœ¯M—€ÿ!Uv2öRPÒ»¨ ¿ O‚1×´”½šµµ^â9§o•Ázð!â)f,xgä±W¨&8±c3G΄W[$åmå_‹ƒñk=inÞÓ•Þ@[]ÃÜÊšÕUY³‹fìËJdIòà“|ÛjYF;ÚUz0vKÃ⛟™«­`n«g×÷ÒJ.]òA*W€¡Ôÿ¶Æ¿âû`ëþdx<Ï–Z•Tïߦ£Ô'’ð±â ,=ȬÑí’‚!½]tþhå°™8 ¿Ø€Pa§žÙlö·={Ï„^1p¾ï¦U»ñÿ^†NþÆ!Ÿ—ºÏBÀoù3ìº#p«õ:nKLí‰Oç+6ñQýäwqò<4™ŠÙF6.ÔÝU bÊߦ¤ÆëÅ¢µÈ®Ô)Ï¡½l¯ó8·|ŸâQñ‹;%CR¡O-I\k‘Û5ýõœ²š!&f¢œ{œe?H ë™×™F»ØPä¾Öô}»é}¨ wõK¾æ(÷òÞ2ûú¿– ž—2øå~1V0µF&Ö¶w•ãÕJgØs3ý 'Íå®öÆ’ÑÔFè¶üÈ HÕ=Ïþå*ס{Gãæ±Ç\oýß"ÿpï½ûæl¨åZë!Ý7k-¾o'·•Cd§÷ÿƒvÛ£êNÿ€Zm‡¬ºÓ_·ã9ÒåÀà™Œ TfD,1›·‡XÍðâ…¤™­¢2Á½’Õbô¥GàŒú¡œßöéO–Ããüx0nyo’\Ú|NÉ¿ » ÿQíyB…këÍlpô^LüàïtïËtT½+ž]‚OujÙ×ÒÇÆO—†kí§ëÀ^¬ ´)V÷jJU¬šïÃAÉ®Ëölû½z¯è@T¬ºg«²­'ªàuâËsåÉ÷@WÙ«ì#l©²¸™[Ý5Ùw0FÎK_ªuÀ5÷t9½ Î*¦õ> ¸µ2•y¾:š?B¹‚,̘ ŠIã³i]õd•Ç¢ûîÍmܹ2éó > V~©Ê»ºËcˆr~^“ݲ›Òb\?Á¦ç¬V¯ü Ôû犊R7ÈŒŸÃÆ5á,e´¡ÊÚf®&÷Sé:ã„s±éÔ8¥W }S¨~½É\kwÍoÜf¦Pœ/³”¬cZºmW÷‘$ôfbhd+ЙG} {–Xwd“t@[D¶¦Þ8ÚZöïôcÓÎóò0vÇÏ)ðÀTË>ø2r,ô¥†ã¨An,ü€#;[…H …ÃÜvÙªÒ綸ò;3Š\@ué¨+»^{½õ#ÉÊ—ÿNw:SqÌSU8gØ»¥Óðè´¥>›5«{EõC*@)côsÄ"ÁŸôŽ¢E7šžª.qqÛDãËÞRÊåËί}9ÍRâˆ]1¾íã¤\6‹ ó•½S}oCÞ¿qr¡ó+±Ï0«V3W3ÚoèL}.;5U:ßò²ŸÜWý€G«wwUʻnj¯_GáuIÃÐQšuˆ(Ô¨}Go65újqÅ—«ÿ£-¬J;î)_—ìñºÓÉùï;*ï›ÿý9zjæ7A¶¥rÄÓ²ù²+¶ùY7\™ÿx×x̽ìcîð÷Œ==éŸÀ 7VÕük|‘) {ʵ¥³{­™2<ˆœÙ'~'×þur—ËJ¡7ÞhOñéu~^ýV€ÉWAú¤\ DÎô2NY³z=VÃ>·1v©ì¯Jcƒs¦Vúj­(ÖàT)'€¯iš ü·0½Í-Θ¿(ã+"]9;p½ËÝ7Q†ª·«ÀµmФ VµM«|¯µj†¾*ª? oÙï¾Ö冫;ú`5%‚À³DuØØ°ØJû;Zx_áËQªjÍO8Ný¤Uó&s÷O)2ü}Ÿóíñg1WjÛvcܭؼ+Ûå>¼°ªDNâøÎâLßI§î˜Ç*Ÿ¤òM×ú pà1>˜]Å/¢g®œÚ€ðGsÙcƒ&fO´M0·ŒÑ¿$óäÆÇ÷áÝlš@ºyPU1“)åàqop|À.}ûeÿÜóEJíÀO3p6y9–´Qjéš@%Fâ ’p_åG¯MïÕ¨ê_<‰ü`Qv7wj‘úHÑýªQÑN°oËx¥ÓwJ*æ»ÏÊ™›—Uà—Á‚‘ƒ?mΤóv±ËóÀMÑ–†Lš°¬ïÐwø¢=û&œhçrô÷^Vצç9Ÿjç:›‚ŠFì‹:MÔ—±;yW;--ßBýäbOHˆ?#­›Ë·M¤ÆºL}lrVãsm´¡9Ú©^Hñƒà(ùľ-×RHË?`éŸÄ*vÌ@®[Hš°Ï×…Mò…dLÔŒzt` 6Ó_½$®Ô½¾8y[ôî,xg/ˆî»%X‡}GV*Ÿ—¤_ðîåF'ç´xý[&ÄË€ÝÅ”Ru$ž¿JÍÆ/tæ ØíëûÊÍßÂv£›-î0lÆá>/‡e»ÿ?£Þ:&x@K5Ò©?ê©knÒeø ZàáÓÉó@c¢ì耇•Ô÷)“j-SÒ„/ÖW±ð.½¾|Ct¬ìòýøÊ.ý2Nu^ç’¬œ‰:ôlÛ"Ko,›ÿ»¬ Ÿð çAÏkA—_í˜_¶ãzbõN]HßÕR­ðØuù¸IœÀóÊQùæ³Á´\âäU{T*^´õh¹‘(l©ÄJ~‡^ƒÑ¾lõ8ë4ä+K¸°´§´€œ›¶ºžzeö>uÏîôºò£Sú/¿ºb ¨Ðëݹc-âwã"˜!y6„ÿ )Í]·I¬«Ÿ”üÍœ^´/o1[ž²]÷°W濟ŸÕ¬\¨dû¹Å›@T{J\/®:F²LÙr '8î;#]Œí]‘_p ,B<Ó¹=ø |QÁayÞíh«—×€³àÝÿɺÛíf@“€˜ì[=œfß;3‰Ì`&Æ¥Ìâ  .ÛëßßÃápFÇ}7ßM²¹ 9ÿz9Ѽ <‹4—Æ oÇ]Ñöc‹øE%_ŠqêH|‹ñô‰ü\œN [uyÌÇÍdh af˯õùuŸä˜§Ó&€[&W» 3d~߃Î,x3ÄWÍât´­ÊôÀWT ú¨-Â¥'Yc±¹×¦¾¼^G±%SC7ÏØXp ¿Îý]ßÞQ]»‰oèDÏû<ï¦ÏvÂßä]ˆ}4qƒOúžg53àÒþ 7€¤³ð½o)¸mã°¾ –´–…×É'î¸Äw"ýd¿&¬Ï¯IcYÂ&Qä_ö:²µK‚ìk»#ßNå¬ïqÀ_ïèÜ!Ð ‘V¢ªÏ*¨¯òܬì5³›[þ€[¿NŒgK Hx SÒä¶ë >ÐE9à%iN“?:Ÿ í—ëwzï:=\B¸ .V «{øäx&Î6Ï['mA2æV¼=ïë·µ®wW‰”Õ€“à´\ë-´Š8µÝý¯B|¼yO¸ Í e5’|ÐÖTÛšÊNñ‘¸…~€$_F|Ã庮n…O¯k×9é.’‘ï¡>y/ ,*Ÿ§"Oœn±àK?›Rom,Ÿñ@MKÄY?l¾ÙIÂN#{¡4s¨SC~¯ß>éT¸ ·Ù@å,G˜«½•9¡Ö%¢šþöbè ƒƒø±¹§½z)û€¿ª=Ýä±?8 ôÊ™b`‹,G…ÖT”ˆ … ù/Ëÿxz¢%=`(V›kÔ%ð1d¸Û¯‘ú€Žð¸ö`.â*­Aò]n)YuÐUÚÇ.ƨpËN´^vÛ:)øBv+:ŒÉ ,;ìýÖ_ú€âq»„;*X,xø×—Òpà(ÿ ïÐkÝžGêrDŒa¾K¨T#Ù{™vßîIp ÃV?¶jñ}‡Å€©¬,ÿ÷´Ž@ YQf<¸ óßAê#ï\ÚÕ̓'l£ÙÔ)
÷¸ÍU¤¸ !·Ã½ (~0î¸ ±:·Â!„²p[÷š*§îÿ¦çêª~ƒ}åO¿ÿÖù_£óþ ŸCüïÈzýEOñ¿[ÎÔ Iž=ä¿«Á{%Ë.dw*©v“”« K¾«£¿àÚó¿±ä'­Jª¿ýeÏö*aÏâ~‡ÏÃo¡©ù¯­ý_¯úüOÆïçTûÍö•?S÷ÿ×úû Ÿîýwÿß_½©ðoµ¨\Ÿ'|€ —z;à‰¬û³ÿÏþ¸ ¶`wÝà{ ö…?¹ßÐÿŸ”ú¾êÞTýOûß²õùºŸç~ŸòÔþGø<ûXOÞAÈÀ[±—@[ܶŒr.ŸÄvÄr¡(ýø èÒ9SòÚsî*}gï¿‘ë÷U>‡[Sù[þŸƒß±©þæûÕOÏïÇÔú=üÑ}}Oð·òªSü}O‘¿t/çÔûÿûüþmO[å‹îê~#~ä_—©ö{õU>‹wSûMù*Ÿë}Ÿ?SþfþOé7ÚÔùÛï*~÷úïÎgçéj}Þû*Ÿ÷gÏŠ§×~»ŸŠ©ößÝsô•>—{Tý‡Åϸ©õ›ÊŸæïîª~+}OàïçÔúû:Ÿ­ú.}ò§Ôïûu>“õ|ýMOýŸkÏ’/´ø"û¡~Ü_WSîwùâûšŸáÿŸŸ©ÿãøÿWëñµ?¿ /ÍTý¦ûJŸìþGŸx/àTù;í*~G|±~Þ§è·ÑÔþÛ|Bû¡|/¯Ü‹þ ŸA¾ö/¸ð…÷ýH¾Î§å·õâÿf§ã÷ß…óEþ࿸ñªK¾ü/þ‚ýð¿P/¹êÅöÂýOçïËø¢ýнÅî/ùE÷u? øÎ~€_å ý1}¸¾°_iSõÆýP¾Ð_ä ý±}ˆ¿û‹îÅûq€/ø…ö"ÿ«SõÛ÷bþ`¿ /¼î…÷ûø¿äóöbýˆ¿¡Só»üa_“ÿ‘}À¿V/Ü‹øâþ€¿¾óþ€½ÅúA _üÅûÁº/ô…÷ûa}ÝOÊï¹ò…ûQ~€^õ?!¿ÑùbýØ¿Š/¼¸¿’/à‹ö¢ûq}È¿’/Û ùâþ½Å÷bþX½ÅöBÿX_´Ý ñb÷òEüQ{‹ø‚û±{‹î…þÕOéwñ…î/ß‹ý}È¿†/q}€¿\/q~À_t/ã‹Ü_ï‹íÅî/Ý‹Ü^âþX½Åî/æ Ü_b/ý¢ý˜½Åî/q{‹Ü^â÷úâ÷¸½Åö¢þH¿Ž/q{Ôý•6#þ{‹Ü^â÷½Oí·íEî/÷…þнÅ÷‚÷ÿ±{‹Ü^â÷í…üA{‹Ü^âÿ8^âý¨½ÅöB÷¸¿€/q{‹Ü^â÷¸½Åî/q{‹Ü^â÷¸½Åî/zŸ„MüºœMïS†÷Û üÑ{‹Ü^â÷¸½Åî/q{‹Ü^âûA{‹Ü^â÷¸½Åî/q{‹Ü_t/ý"÷¸½Åú±{‹Ü^â÷¸½Åî/q{‹Ü^â÷¸½Åî/q{‹Þ§áw¸½Åî/q{‹Þ§ø{Ü^â÷¸½Åî/q{‹Ü^â÷¸½Åî/q{‹Ü^â÷¸½ÅïSú%½Åî/q{‹Ü^â÷¸½Åî/q{‹Ü^â÷¸½Åî/q{‹Ü^â÷¸½Åî/q{‹Ü^â÷¸½Åî/q{‹Ü^â÷¸½Åî/q{‹Ü^â÷ðÅî/q /q{‹õŸÕÁJæé*¢Kõ«õ¼¥%(ÿK¨—ëò¤M‘oë?Û¢Ûm‘‘?ÿ~Ïùÿ?çüÿŸóþ|’I$’I$’IBI$’I$’I$’MUUUQ$’I$’I$’P’I$’I$’I$’I$’I$’I$’I$’P’I$’I$’I$’I$’I$’I$’I$’P’I$’I$’I$’IBI$’I$’I$’I$’I$’I$’I$’IBI$’I$’I$’I$’I$’I$’I$’IBI$’I$’I$’I% $’I$’I$’I5UUUD’I$’I$’IBI$’I$’I$’I$’I$’I$’I$’IBI$’I$’I$’I$’I$’I$’I$’IBI$’I$’I$’I% $’I$’I$’I$’I$’I$’I$’I% $’I$’I$’I$’I$’I$’I$’I% $’I$’I$’I$”$’I$’I$’I$ÕUUUI$’I$’I% $’I$’I$’I$’I$’I$’I$’I% $’I$’I$’I$’I$’I$’I$’I% $’I$’I$’I$”$’I$’I$’I$’I$’I$’I$’I$”$’I$’I$’I$’I$’I$’I$’I$”$’I$’I$’I$’P’I$’I$’I$’nîîîì’I$’I$’IBI$’I$’I$’I$’I$’I$’I$’IBI$’I$’I$’I$’I$’I$’I$’IBI$’I$’I$’I% $’I$’I$’I$’I$’I$’I$’I% $’I$’I$’I$’I$’I$’I$’I;ì›"&È›"lœªl‹E©ZÊ%ZJS*¢}©—àªTÊUøêTÂ*§Ê…äSì5›lKÚ/3ìÁÝR-ªÚ¥ †ÊˆÚªmI²Sd6€Ø£h™¶‘²ŠJü›å\ÎqÎnL<Œ«$u)T„½LÑ% ЭÈÕD´RÕt$U ܃ÑuÃÅ×@£@ª«ÍK%#%"Õ1 0µ ”%r÷K¼«Ñ=M$17CÏSI,LÓP*D1B\Ј4‘(ˆ¢¼ÁOKL”œ•4-3G@ÃÉB-Ít¨Œñ*<+U ©S] =#BJ±H”³Ì<©-H‘T(¨ $Ér Ì$¡\HµÈŒÃsÍr/ #Ô¯A(,‚¥Ð*T¢Ê+3tÔÑp”\CK $) ÈÍ)BóÔ\\ÑRB¢Ý#- SŽ6ÛŽnræÒ¡ëµVÔ…µ)[TFʦҭ Ž|½TV͵¶Tmš“a›6ͪ›mllÙ›4­¶cll†Ãkfmª­˜ÄØ[ ™¶ÚK5fÍl•¶¶Ä6–lÌÊ›UmY•[-›Eml36ƪ6#fÔØ†ÆÔ–ÛlÌ‘µ5‚´Ò¦ÔØøj*ÕSµ3e&Ä›-ŠØKkYU²«h¥¬ª¶[lŠÍ²Kc2l«dK0Ì…´¶R¶¦Ò ØÑl•m*¶ª› 0ÚˆÚÙRÚ(ÚÛiU[Sd6T6¨Ù#eµ%²•²›"£iMˆ”m°ƒj-ª-«iH|2JÈÕ)m)[EVÕSj†Ä‹`@m$Ì%µRØ j¤¶ l’6 ¶*&ÒRl‚¶ªÙD|‹Kº["VÕM¥+áÏ=s•.TžšÙ›m^dš/„]"]Ôd µTM‚-‘jŠmQCjFÒ mTØB؈ÚAðí_sœns„“ =HŠ@Ó2¤¤Ü„1\ÊÝ%4òÉTQÔ2´]M‹qQW0¨ò* —<´Ô1D· EO=S= Ï ËO"QTQ2(©)]U-%̯L“r ¢·R%É])CÑ*1<ó)70É*HòK74ñ%È´ %Ì¥(õ R Ë/Dô²(W4ÍÊ)t Ê­%JŠO) ×%CŒ22 3QWuR7+ÅÅ+Ý!/=O%2¤“ ‹0½5,\+ Ï=0È”·L¯O$BLJÒÊ*£Ü*¢ðˆò× Ã=wPÙI°Sh£el ¢‘mT=´œÄع³Û!êm*s,Û6ÛlmEÌö°¼ëiI²€6 ÝL©ßTõZ‘ë[6ns†kd]¥ÑvO2£h[B›B%µ"wjŽ*]Æff;œÓhœÕat\u<Ô¡U-¡Il…wg8æC˜ÛVÕSÖŒmm™K­dº.É5°!lI°tÚ6ÖÓ™­¶Ô®¶ÙšØNåhº.ÝgrÙVÄ6J¦Õ%MªÚU6ÌK˜9IÉmAÛ4M«±t]ó 6Ú¤Ú$.ä§5$æš' ºÛ5µ²¯ÞÂöµt”«bÙ({VÄl܃š+¹¬mc[a;­EÙ=eCÍT îI¨½z¡lÖm¶Ù˜—Rè½tîNh–Ò¥jZæI”™ŒÛ[6Ú©Ú]^j–ÄÚJmJ­…±6‹i%±N涡ĶJéšÍ±N‹¢õØ­¤¶²ØI²*-…²‰µQtNR7xƳjÚ'sl.‹µO2‚Ù ØR-¥6J¶¢Ó6UÍ6͡ݶÓVÍ•:—EÙë)Bó ¤¶ÙÚBmµ4£šl›¼ÔæKÖœÙEÍfÚÛ6¶TîÉt]æ)FÉ-‚Ø£hWm”sœá&Åݶ›[Ju.‹±çÐ×Òh³XŠ£Ê/rß#ËÏsô\" ¥³VÉôšÑ°}#5l¾“RlÍ–Á}$¯¤—ÒcG×ëSëö>¼ù&/“€Ù-¥ml¤ŸªNkh•ÆÛM¥|{Ò®d¬Ûe›˜/Œ¯Œ^vª››VËe[$­.äØا ªNílm³m¬ÕNÅÑv󼋘¥mm*M©M¤­¥ÍU´2ŽîoVÖÉ/XfÌÖeNÒè½;;ª—4I6–Ķ)²FÉU]Ò½jͪ=ss[ Ú²ÌÕ¶Í’v.‹¥æ‰-€lHÚIl¢mEuG6lÚ»„k6Ûi›6¨îY.‹×U²ŠlØ«iU²ÑI6„mTÚVÒ-‰ÝFÔÛ6ƒŠ§XÖËm¥9Þhº.Ç®ñMªª¶*-…M¢¦Ò6$l“ºXœÞ¸†ÛZ…é™™›fÖÛU:.‹£ÍUM ›)+`›R¶¢]Ós9Üô6*q›m¶kÝ…Ñu]ÕU܉°¶DؔڋbmImSiB­Š­—2Ùj¹« wÌÙ¬s(ì]Q6”mB[R–ÊC¹h9M„v¶ªv.‹©JyŠM‚› e¶„]kklÙ³mfÛ*î˜]jj+b«iM¢ŒÂJ­¤Š6$î¨æ¥Í¶ƒ˜ÑN¶Û4N‹¢õ×t9‘´.55A° ’©Ém;‘6Û6Ù­¶kd;“ ¢ýG´órÉwP¶…´6´¢lSj©mJmR{l£…O3ZlÖÖÓ™'²^Âéìd{b›Q´•[A=´VÅp‹ØÚmfl©í°½§ž{QˆzÑmFÉ6¤ØJöÔUÌ£j©/(puÀÈ”ÁB":¶fÛl‡°½…ìNÍSGr3 m¡VÒŠЛJ6«dM lªl³k𫽳[l±›Ú–à Ø^Ðö½ºô—¤õŠöÒMˆÙCjmH¹¤æ“—™/[l ™­¶ªæ^Ñ{ Ù.û'*2¹ªØl[´“d[Q6U6jU±í³fÄq)í–Ùž³‰=…ì/juć5 ª)´¨¶/m­Šç·9"æÍ¦Å=¢öÇñü=ñüÇñýÃðíð’I$•RI$’UI$’I$ˆˆ‚I%T’I$•RI$’UI$’I$Hˆˆ‚I%T’I$•RI$’UI$’I$ˆˆ‚I%T’I$•RI$’UI$’I$ˆˆˆ$’UI$’IU$’I%T’I$’Iˆˆˆ$’UI$’IU$’I%T’I$’Iˆˆ‚I%T’I$•RI$’UI$’I$ˆˆ‚I%T’I$•RI$’UI$’I$ˆˆ‚I%T’I$•RI$’UI$’I$ˆˆˆ$’UI$’IU$’I%T’I$’Iˆˆˆ$’UI$’IU$’I%T’I$’Iˆˆ‚I%T’I$•RI$’UI$’I$@C»Á$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$ª’I$’ªI$’J©$’I*¤’I$©$’M•$’I²¤’I6UI$’IRI$›*I$“eI$’l©$’M”SuS3333333íííííííò¤¾RÂ4Bú,D¶ƒ|¤9œ‚ù[3[[5¢|¡|¡{ öÉ-UÜSh›"M¨–Å-¡í†#•G¶6fmmfbžÑ{ ÚÚ5RÔ¯mR{fĶ©m&Ò š5Tæk`n{·¬ÚÚ£ÛàÂøë»§ê}JZªªªªªªª©jªªªªªªª¥ªªªªªªªª–ªªªªªª«çIèªú1{æ)=¤¥9È&Å'Dç›› 6)6)6oƒÛ½ }è7Z“è1Úˆl½srʸڤúkm¸÷})ôÀtüÀïŸÏçí¶«mªÛj¶ÛmªÛj¶ÛmªÛj¶Ú­¶Ûj¶Ú­¶Ûj¶Ú­¶«m¶Ú­¶«m¶Ú­¶«mªÛm¶«mªù‡q²ªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªª¡ôqǶlRl%)â^‚õ)Np.JSœ Š”çØ-‚Üó½„|ý|þnUK•WÏâ¥ëT­ªFЮi[ Ù+¸úE€S¾Žø¾>>*ªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªª¯­ó;»ÏUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUvÛm¶6Ÿ õòRŸØ-‚írT;߃µÐL„Ê£çõP¾"ú mb=l#Ö¬Ùœˆ!ÁôðwÌøøøªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªª¾ wpóî;Àó»¹;ÕXʱª±•cUcUc*ƪÆUUŠ>|G‚ ïÅœÇô?w¾ïô ýÑãîíãÚùj]3µŸ[îãÁKD¯ÝoUwÑt;+:«˜I^kÖŸ¯¥è^TE•µÆµ°Ø¾M«mšm´Ÿ7Z«éöl>nI›['ÌÈØÌÃhùš›lÅóióSï·ß~”Ÿ=ôAè'Í>gœ•9%-¥°|ÍO›–ƒÓhM¡é°Û¶#æ¥ó@ï—Ë⪪ªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªø*ôzªÆ+«V5V2¬j¬j¬eXÕXʱª°Ž!¹Âär«á™*bõ¤Ú¥´Ø ²«jM¥²XW1È9­°§Ã¶ÛY¦É>ï€;åòøªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªª¾W£ÕV1XÕXʱª±•cUcUc*ƪÆUUŠ9RÜà£á¾T— ¾­¤6Tˆz[*æj'í¶y”>8ï€;åòøªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªª¾;°nîä;j±•cUc*ƪƪÆUUŒ««äMÎ(>”,>F+d–Òªm ðó’M‡6ÚÛm¶¶µ´‡Ç|ß/—ÅUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUW9Îsá•èªÔ7®ùO†øw"˜R¹ŠôÍ6¤¾›JøE@ï€;åòøªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªªª¾8 ÝÏw:g¸øŽä¾F ƒiKh-¶R¶·Ã›6Ób¸%æ¶ÚØÖÛ6mËÇÁÇ|ß/—Å-UUUUUUUTµUUUUUUURÕUUUUUUUUUUUUUUU|qØî{‡\î÷\p Ó™!²¥|9V¶+›\×2‹á6šÙfh>6a|"ö—Íø~žõôGß¿~ä’I$’pˆ‰·DN[m¶DdÙdFTÛm•vQ|¬Kåf×õÆÑh§6¬h>VÂùBöWš{9Är©sTlŽesKLÅsm«bfÖm¶Ølh½…ì=»Š«Û 6›Å'¶ç8ÚfÔžÑ{ °y¶R»‰sPm(äÙVï¯YFÊ\Óm³1±³b¶;EÕ+ÍPl¶‘Ý.jõÏ[†ÅgÃjVÑzm›iC±t^tl‹i `»®e³Šä¦Å]ٶͶf+»Ö8ôyߥ"ÿÔ(ôÌ‚¥"œ©#JUÏ<0¼5ÒLôŠrMܲ¢*#-OA=W<ð°ô‚²ÝOED2/IRœˆ¤È#ÌÀ½<´GOH³ÊÊ\CÊMÔ/UÊ5,¢tÑ((Å¢O(­CËÑU*=ËÒÅ"T"žÏe¢¸•G…h‘’N&NUTBѵ)RmŒT£ÊI(¶ø…Ò»ª“j¶Û%Ùk³0;†‹¢ê¯4ÚªªÍrGuC66¨ì]a<Á§W ¹¨wLË3Tt];תªmËd½kB¶› ¥u.‹ÎÊ®êªÜãUnH»¶¶Í›ÌîÂè¾{¾vOl*Ú#êTõ†Ù¶ÏZVÑÏa{ Å#Ö*lîµ.kœŽj“¬Û» ¢íëIdN³RæÂžµ¶6‘غ.4»¨•6jYÁ°[Jê§EÑy!wV¶‹¥´‡uMY•sWbè»zËÍ*› ´ªOF°‹¢û÷þŽûU6'Á“™Hæ×1Y àY›Pøð ×kº*æJrÚ-¹Ì%Ýë9k!Ñt]yç¹¢¨vW*/Y6³[`t]]ÕG˜îmW­É\ÁzÛY Ñt^<Ô§š¥wb¸9ˆ½FÑ:.‹×ˆl«d'sk›m t.ÅÑzëÌsR–ï6Ñzôæå¶Ò:.‹§Œ”ÚŠîæs¹¤]Öl'EÑyU.çÑiK¹­Ñt_Põù®ü§Á†É5¹Ánd—¶Ö=b>|íÀ®ÊâS¦Ñ:.ÅõÏ^ÀöÉSonPÅ/VÛf¶‘í°½zì]ÕQë'©ÇHKÁÁ'dåáyEùo‘íø|Œ,Ü©mUð3mCà‹à¤—ìX¤—ÛGwp~wÜ>ä,~³÷"Ìí‹I¨wÜqŠ$*.ï¹W„žE¢fYëÍ•Ñr öÇ¥(èeç”EåR…î‘O¶<¡3ÊÒÂ<§m‹lÖ¾v¦Ç2æÆÍ«çke¶Ù|ì›LÕm­©ó²3S3kfªùºÙƒæãcScçiŸ;M†Ú>wÔ]ú™ôí9ªÌÉÙ,ªs¶3T|Á|Áy;¶w"\ÒÍh¹Iw6¶fÛlÙmUëEãÈ[PØ®VÉ.ìÆƒ¢è¾ûì’wRlžÖÍ’Ü‚öØmK¹ì. È󻦥ħs*v.‹×wxR²]õëUNAÑt_¡þ‹ë¾äjƒEO‘¥µzÖ‹ŠW£b>@¾@ºƒÌ î©Í#–ÉnU\Õ]fÛfÄt]®¼Â«ºÑzm*zÌÆÍlØEþÎûç"6*lC™›6q)í›iÂöæª.å6ÆÉp  í±°Eß2*î¶üã^¹K›nq#ÛU›ml³ÞÜÈyŸ¾{~oïü¢¯ƒ|ƒêõ’ž–Ê>¾|Ÿ^wÛ-¥ ¥Ë–¶Åq Ö°^ÂöÉ{;„¯m7¶ª ª¸¶Øa{ ¯ZwèK˜l¹Ç-*§¦Û6UÑt_CëÙ"öÒm½¹Ê\›*œÏ[36CØ^ÂïvªNâœÖÓm »¶Íª¶]EÔ+Îñ¬ÛEÉGvÖ¶•Ñt^½w¹zÅ Š9¶ÉqRôÙ¶ÉwZ.‹Ï;Ä.m°¹UmNékm èº.zì)Ü-«kEÊ&Êz­¨t]J—wpã5.l ºj®c¢è» y›¸Y+¹&¶QÑt]J^k¢á+¶ÆÕ]E×Ô÷ðàa»ô¹üÍÿ]ëñÛôúúmúÇÑÿkìy{ýÇ?%ø±oŸ×éõ¿_ù/SòÚùïï-êü¾;¯µßßã¾\_o¾‹~[_#¼¿)û=>_ÁÀü¦Ÿ™ÿ¢üï'ÙãßWà·ã€ðëþê€ÀáÕˆmÿdp·ã ^pFÔŽÖ Ù²ñødƒŸxäN»mŠL<ã*«L—+¶äZÊ{àÂÛÝÈî¸Ú¨aЃ1Œ† ý nlèt'#Iˆ\)=îIa -PÊ/š¡âiêAâeK@ü‘¥Æ qÄöŒ+ïäp–㇠(az˜`iáŽøf0‚Hg³ž7Ø=„`*0SSiˆ¬F†36ÆW×dÙ´Ù¶_]•µ±©ôÙ±¶[/®ÔÙ™­¥õØÚ¶4¯Íéd~y´}`úæm©øYçÐý·Õóë·à>¿äsì@’€÷áá׆ÎüŒ›Ž11â.…å6N£ŠPC,ÇLë‘ ÑýnÀàVR}•«_öZ?HíWîÿ.Í(ÌeOåþéA£_ù ˜DÈÚµd ¾ê÷žßíxÿ‘°èÿùÅýèâ ÿVÃþX\œZwÞ°þQ õûW¡/ø³o{QXl/Fwúfã1Cà‡“±£þÂIÿLû_~îŸÎßBà SLE€#<ý_Û`H)eˆ'ÅûeŠ ù6–§·oïõòtþÙëË1!„tΩ÷ÇLüAòçÊöÃ*L侜OÜa:\£“ó’à@cžÎ ’CýÂÿ±@†Äs °'ûãºDF4£ŠzV|;U"YzäÚÞtÁ%V ÇyF‚˜0à :èá~Äá„?]”üöü®¬ÏÒþø_+ÇÔ}?Áé}.Óûç/¨ßžÇÔå'ö|õôÞ}Þ_Qöü}V´Â?ÇÏÀü÷'ÎÊõÓó:Gâµ_/TŸ/%—“i?ÊÉýQ~{Kü «þ¾U_¤Àü¾©Õ«¥ö:§âò?c«åáöy~÷>·Û#'éqSû_Ðò(C~›Q>§/§Ð}fAøœ’ž²§‡Ûé/ñòÑjŸ˜×àõ÷ìŸe §ÉÊ—þ,_‰Á+õš)?ÅÃïXúÜO•£ç´¨þË*¯‡}^'ùú‡ù: }~IùQûm*—ÞõA~ûö\?!•ûŒ(¿Ìùœ¨Kôùû T'Óéïô‘/Ùi$_;ô?ÿ=QøíJ¢¯ÎiWúš©²¿M®e¶Ø\Ö‡ï4£6¶#Áª.zêî(³™Ì-¹.hMŒÀèº.w²¥³fªèlIëcke[t]D®êÛ6Ù.“š)Óci]EÏ ÍBl©Ô¹²£Ö‹mš‡EÑzõÖÈWrÛ%é6JzØÌ‡EÑyÞéµ´)Æ-a¹Jwë!Ñt^®ãg™ÄKmÝÍ™ ¸sõC¢èº»æçs‚Nbq¯\¥ÍOZZ)ëiEãºÍbѳe³m¬KÎ¥6›0:.‹ÇXÀ[FÔî7ÍVäu­Ñt^ºv”¹Ž)gyÏUÌ)ëV‘Ñt^wYHîæÛkjW-‘;­†ÌwJÂè¼]U-fª»Ü›m¶l«¢è¼»R]ÜT¹¶hÑvmUÑt]ïyr¨æ*s"ÚÚÛ!´mVÇ]^µ.‹¢éãYM©&Ò¶Û6*ÚRu¯[†Õ]EÞµ¢¶”6¶™+jl[¥Ní²Ø®‹¢ìõºÙ² fì'­cjEçãl¹ª6¤nsG2'qÎ\²®‹¢õçqBÛ)=DzÕm[ ÙGEÑy; fÒÔs@îÓdº.‹ÇrÒÑM¨¶ä‡¹­QÌÙ³*õ®‹¢ëÎã‰&ÖÔiWr¯H6¥Ëm½i]EßYâ©´£šN5®êzÔ®=aOYlõÊNì.‹¾Z…sœá;‰m,Ѷªèº/Ǫ’ëÍ8U%øª•>›õ_{çá_Øo[úo3×®sí5¶mt:›ë¥åÞyâ·`¿Åv,¦½»¯Ex»5±x›\l7›ÄìÛaij3cræ›UõdÍ•}^›5šm6|ì6ÒýÆ,ÓFÔ}Ž6¶Û6O¾ãjÚ™Wè³i‰ùjý6µµ}TüEóo›¨‘÷ÌllÑNI>ŸMµµWÍÍ•Uݱ¶ÁTxUA燑AEç—”T^EQåUUQQžs6ÓÛµ¶[ZÌÕ™šÙšk™™™¬lÚÖÓkj^NTUEQåDEEáA\Û4ÖÙ›Y­³m4ÖY›[-3M´Í­†4mŒÛfÓ6¶m¦ÚÓX͉¶i¶Í«[[FÃZÖÌÙ¶lÚÙi›lÍ«k6ÌÛm­«ZËmi›mfËFfÙ›[myÇ ;©-‘È.mRsSÓdº.‹½Ø”Û3U„îÓ`t]ŽÛ([6›a±Gas6lÊØe]Eë¾g:sJMT¹SÖªzÀãm•t]}k¨îªÕmª—=¢è»æ'u5%ÊvÙWEÑw¸Q±fÙ¡YºæÄ;l‹¢ó»mlƒj“w—yêRõ¹µ)ë[Z«¢è¿ §öQ6FÃ6ÖmR¹6*öËb¶ÛmUì/axé;ª†ÕÌÆÊ.f!Ý+cmEÑw¶¡´#3º)Ú®‹¢ë¶mÝmÊ£‰-»¶ôOYUëlnqWEÑuãe#cj7yÁR;£cfQÑt^u²¥ÜƒpSaÍMœŽb®äm3i]Eçšî¨›Si©.E;š‹¢çzÒ£i´Úb«mÕz¶&¶•Ñt]uPØ1UÚ£»AÑt^ºƒ`›K±²ÐáS²:.‹ÎÔîJ9†Ì¥mÂæPéÌ—EÑwÉq[UÍ…\\ÄÑ­§qX]]’˜lJì‡uµMl«¢Ý{¨mJwEÍmlj•Å›mEÑt‰ÕWšK»J9ªÚÔ:.‹°<È6ƒ·2UÜrTæ°;Eã±]ÔæŠæ+nfPpîˆÛck!Ñt_ñæËfÍ™“f6—Ç‹k3|xlÚ›m”|j¾=›Ç›K3ñ¶GÆ>3jøËãùOŠ’b‡ÄCâÃkU|Bø…çv®á[SeW0¶š”کʧuLËçE¼t¡²¶›ó‹mÆÊ®àfÕ]«Íв¥ä™Šð^W¬A´§›cÎ9­¤©¶‚õæÐ¼ÒW2ÒW*§’¹à¹åUy“yÆÕ‰—2‡¬e^ ×4y‘›@ØW”§­›Jð^!<Ë-ª«Í•^kce©x/O% õ‘^6«Ö*õ¬«ÁzòÒ4æ–¢¶MÊ'›YŠð\ñá¶ÅVÒssYµ*ä9„xÑ•x/Åö~ׇ—ƒ/ywÙÓ³*Šóö‘«‘H•ås3ök•Q ¶ «µª*!öPééålfɲ|Ìl63k´ØùŒ™›ce˜|Ím­”ù›e¶KæfÊÙ³KæfÕ³Kæ6—Ìf›+æO™µ}-Púlú~\Ò«ˆÙÓëJúqz„ó ‹m¡¶m³6¶µ›5¶ÆÆÆljÛm­i›bÜ/"¨ò¢#ȼ¢ò* ð‚Šòò<ð¯(«Ê€¼¢ ­µµfšÛZÌÚ6Y¶Í±³MmŒÍ›LÙ¬m¶Ù« fÃY–lÓkm¶­Zd³f9îyTUxQyUS„f¶f6¶Ö6lÍ3mm˜yEDES•yx{‘yA‘åUyEáDyD•¶m³fóÎzBô#ÖŒW‚ò¤xÑ‘[Š\Õy›M²¯¯=g­†ÙE´r¹“š¥ë,دÏY¥æ4+›R¾¼”¤}³î( (  ¯a­íÃ{:´D!ÁB+Û·¸ô+ÓT÷l%="I)šS¥=´š„?OTÚ§¦§£Q½Iµ1=@z€A Ðiâ¥JT`h =ª•TƒF€ÈM Ð)ú””¨’z€í%4<£Le3$=@…)Qˆ˜OF…?J£ÍPý©OOR{5LSõM‚O(õ Q õ4OPÚ™M$¤’”êjŸ¤ƒM=êOMwȈ¼±›“‘÷4¹îÙÒI¹6È9* ÒÞÛ4zº¹Ú ­gÛk±ÓG}ÜhtŠt O âËàò~/6ç3­t:ßxPÁ‚ª=üEÓÝêög“ÙÅ´¢€òÏ?Å¢" >YTÓ_ïòz;‡'Äìˆ*ð)/øQù”QqxȪ5óM¢ZV-2’…neÔÚ+­lTɨl+T°é­ÈY, —-–vL+JÆ2kŒæ;݆››V”*Ím‰‹²²ìŒ¶´5–ËR×VK+µËd¹gL™2ëjΉ»qGpXå™D™ÒÛ;rµG+Yl»ML”жkœÊ–\5k¹I.*6êÚÖÛHÖu¤ì´N#²Üª4Ëæ­LêUŒÕ¤ºKwVÖ¦ZVƒt5«Q¶´‹k²­C3­uÉ“iaf¡Û.c¬ŽËV®VÊnš®M7mƒ”B6ÆÖí4ÊÒ4³,‚’3mŒÒm´Ó’í¬i12\u™ZQRLÔÖÛA ¶´N6Òfq­º;H댓b¹VêÃZæ–¬e Õª Õ9V«­llìWN­Ú¤Øc1¥"”Ñg³nq¹ÚYºVÁÚì®-K]XšÔÛ1M9ÕPìà¤(§MfØÉ¶&“Š\t6ÌÙ¢7ZÛœ.ÚR„å©F–k ¤šk ¹2¬¹UÄN“b›€HÚÛ©Fk³\ÖÍ –5Bmµd„m±°m¡C`bÑ6Í·dX9,£­šÊÖ¬¬®u¢a¤;º8Û‡v‰FhœqÑ]m[„Ž¶ÌƒvÖ;–U–Õ¹uªYZ3VY²5uncuËf™œ—r]™4Õ†êÙ-†Õ¥¸µ·%»bÙkƒŽ,*ÔŒ²8îIÌŒ¶'1­³‚’ Û#·iÎç2ÄU¹K5T–Ó,‡UeÛnÂHeÙÖeÙœYiÆ´d£2鳚Ùq«6´Ð³•*Ý®³®ËZÛZN-'AdœÛ²3&µ´â6Õ¶#¢&Nc"ë X°Ô݉•ª]¥ZkXìÃZ.L«­dƦ·³ ݧjQÖv[d8skN’JGfnÚtIÇGaÖv…[#\YØU u®Ë-iÖ‹VêÉnŽÅ6®;f°3+6£Œ¬ƒ³Ç&Ünj0…rpê9Ùf®Ù­—[n–p¸›hîÎ̳qÐÛ¦·GXqfíLË;'+LN,vÒÊÎÓ€[b‚í0HŽ;NÎm–«K¥šÕ4Ú,8ªâYl¹‘ÕÉ“ªÕigjm¤MšÙÕƒ4E–œc–æíË-m[7Z\å5× I²Î°ªÎZ´V©[eÇNÓHV'[œ'NË4wi·D•²Ý«gjà)Ê´)˜˜—kË»V]\œw-³Ym4æîÂÜZìŠBÊÛ,ÙE‡E¥¦Fq(ŽBƒ¹·dvÈZËFª×,èÙªe“-µgem³³8ˆÊª‹aÕ1:µ®0¤Ö±™®¬“%iV²³¡w-­V¥‹3£0¹m\Òîk›YhîlJгŒ.––èÚZÆXa¹gL¬¦aÕke‡™j*ª£«hÛrRugmM-´få¤Ö Bܵµ£‹˜áC¶ã9³;N‰8N,¶·crs‘LÈfÓÝaa[[VPre™²ÐÜãBij&al¬´%ÒíKZ䨜«j®iml–ÄÍѶŽœƒ‚‰-i¢vÚ³³r£¨ë.Hh£¬âB5¨îÝYQh»2Û[' iœˆºm¸AÌÌÜwkj1¬³.Ûu3SnÒK-×EÑÝNÅÄt¢ÚË;[m¤Imfnç):¤ºVNâ(à9 5&Ys[„ŒëpâÚ°8í·q9–±fqRmÛ$LµYVQµmuFYY \‡6´5ªªÕØ,9c+¸Á5µ«"!:Ì­´sœä‹,éÙÚ1F”KŽ+:;³B[Y;i”U޳mÝ"Òmޏæë.ÆÌtâÉ»s0f\u‰E¶ÖäCnÚÖÛbÁrm3I·s5¶èäËm¥§( Øu§[l©;’Ñœ·7aÁ´¶›Zc1ŵ„ƒYÙÙ$Yem¶–Ú;#»–›sfuGqe™;kY¸™alÍ:H”(ŽANDQÙÑ-nèãkk•–qfÙ;´;’8¤Rƒ³ í­¶ë(.ä­iÙÜ2›;²ËeÁδ²µvå„;'5ƘÎ\7U®ºMZ®M6HtfÆÒ´‰šWX3eÚ»V2®kaÖ´™(Fµkn'mZ;Žâ³«:&[e•‘rNÆÛI·e$”edš³º:³¶5¤ìí+8ŽM,ì»)©'4¶ºÜ³¬É(IÉÖu³g;8T¶¸‹kt—¸›b9®ÛvmY×q±$’ÌÙÖqeœ6ì8£¢Ž¶YØæÛwbQÓ‘ÖíÛvI »mÍ4Ð3¸ÛN ØšÓ±´å;º6Íɶ’ä[YRfÁÇ$u”vhÛv“ls,µ·mÁ\Û¬I,ÔfÖÖkwa89µœwiØ„ÍEÜ[`‰Y­¦ue%Íw%·WcM13“Z“»n’…XëMDS¢Îí;i‹ZÎÅm¦ÔPî¶ådÔ—ešàîŒ-…,í¶LÛ»€:F³s»mIH›Kn·;mÒPÄØŠ“­–Ç$vkšÂY'R ™1«uËeš9ͲSDÝ­%Ö¶â"ŦiÛ\Ë+lÑ ;œÒXãEÙ¦Ç"¶è4KÓ5héir€6íKJÜʹ­š¦L°&iaWP+­™.g ]js•k®jJåŽeÚ»L2µk™«,ÄâµE«V®šYuÓYÖcI ÖTâ³#k¬×+- DZÓ“3Z™ÓFi™ÐÌãG £#"²fUWkŽ¥‘ qÓ4Õ¶[m¶I!¦ªXÍ­$[:p«Y[ ,æ³E#-n0iɨk¤eM’:L;4L®s)a‹Uеªåm`fNëNQR­TLÙYkBYI•«´SÅ¢hkM­&A«•f–+™4qέYJÊÔi4ÖiØY5f¬ê´¹c­d´Ù5ÖË-a­hb5Ze­MjÉXé‹#DsŠ¥Ó\Ö(Ye™Ö+k$Ò¬¥£5Mr–­c«IK†¬ÑERÕ…*Vִĺ‹Z\­(wÖ: Á­É¶Ò“‘-f†Ö"¹S˜¹“¥`K¥ —&jË1 âe6#”rkEq™­¬­”®§4V [YqmTêE™5¬bq±º]f,ã*—4‹™¦„è›JFu‘35YfÖl±%£4×a[sv”Îic[ YIUÖÍ.q“+T«Y3$TµVµj¹W ‘Tl4’lÙ–-gH›v²VÓµ¸Ô`µ†¬Î´8[1²Ëdu™\ÌÊ’vP­·fLG¶rMØ6fµN®Zs ,³ msUi¦U f²ÎÐřŒél+U®²+ JÖÇ*Y9Z冩]+Yd±¦5©gY¡;°ÜClZšºÖm9ªµa”Ñ ™ulÎ5œc£³cÒ5n­”M×0%f-55›jÖ%–˜2œ6­5kLh[ZÝgÆ™­ŪЌNM­Z]„9c4›¶ƒ6–Ì…“•,Öd–P1HŽKXÓuRÊšf¦U EÒufްå–]c.™K¬nbS)gk A±–Ì4¬²ÓYå“S8Õ,×8έRÒ2Ò£.u£™jÉÒÌÖi:‰‡§XÍ#5f’e³‘œ7i ­6œÛs6JjFP,jbµhë]%B5Öå‹i«hº3UIe‘ÈŠjÒÕ¡œçZp*mRÎfW i™–-–YjJÓ;J“MM‹RCuÕf«3NÄæZÕR™¹¥¦:Êq\KU–E,ÎMcYcUiM¦¥9‰ŒÓ¦Ns]ZË*ÁIZ¥ÙÚ³U§k¡¬nc¥•qQHiŒR¹ÚÑ•d‰ EUBеTÄ´®×$ëLcŽuT›N¥bÉRÉ)e.d»)])«j“ZÍYa!¨¸ÖJسc!d›[–ÑÑ+5mXLF´µ552hu³ÍcFÙfZ³DŽhI¦ZTÔºW[TëEÀ)5V™e²ºœÒš§N…«9·&ÜäìNY‹Y³DHÒ+ms0“¶kil̵h2¹*éinbÕCær¡¬LÉÔÁ•)ZX³4×Læ”Ö’¡3]ª¥¨å±›Y[Zœ¶ilÕfSÌTVéÕÀÕ­%uMgYuZ%Êf‹K%­ÚÅ4Ìeua-j¨ffYtÚu”êÔY•‹%’Í35kcM.e­›"vbè©\˜ÔÒ4mmdÉmÍJf°’Õ-e«)6.¥vl¦dÆÓMÑ)¹cX,6ãM¬Ö¥+i\ÖZec”ÌL²Ö¬Íˆµ£k`s·,¸–5¨Q'Y(ÊÒ—Ҵƺìb‰ŒsWU’2ZÎU§0i«uašœ¬©Õ¬®»Šk™ÅbÕ\Öˆ5Z2šTXHΫ]0ƒ-"«-kf”M ­4ËUR#N§ÔÅAJ­&ÌQ¤ZšnȬ;S·i6³S ¬¶µ¦ :V-1 ·PËdÓDrÒ9mŒµ³;†«ªµlË,«¨¥ƒE¶€-ºå›u£—G2µ)Æ´©¦i¬‚mQ:â#QÙ)`ek…JY:ÕÆU–”4ÕZ ÒÅmhšå:¤VÌ£\”™u–¤´)EµÃK Y*‰Z%«˜«ƒ:¶œÇj¢TÒë,2narÕŒfZ®³%Í*•¬ŽgS+‘iMvUÊ«JÊÊÙ™§+U‘Ž\•“ZÆ®¸,±ÅšÍ(R©hÎYf¤“:Z´í c²Ú.ƒ\ØÒe®œÌ×ug+I‡!¬P±kk§XXÙkZaIi+S"ØÌÃvˆL›ÛZKZsqØÖœ7s[¤XX¶ˆK[jëZË%Ö¥d,šÈª‰TÖu×djÓVÒØQ¸ÙšÖÁ‹4ëÉÖ8…SUŠÊÆZ¦k•—v¨ˆË5PZNXÖÅl¨Ëjjj±‘œ6¦LkiT5¬jgQÆh4mŃkKj­jÖÑŠÐë9ŠEk‹mkDjLdM¸ìí4ˆÂÅ´“— U†u.´©.’¬Í«­2ÑÕK2¬XeE«––.¹¤e)j´Í:ÕkX ;s@šgh)µnÎ$0­hCY¦š²«3¡¦SšªJBík1•ÊmSVsEŽ™ZÒÄ–­HµZÕÍ é3˜çZv¦bHº£ªKLiVCYY®åY˜k5˜MÎ×J».3YÖ[[1LNŠ9V–ŒÊî¶nef±–lÑ«fjÕÕhÁ„ÕÌ£B[4dÇTµuN «Y[:ÓDãM6±FYÙœªº›KQRêéš„¶Á³›hÎÛ¶$•«Vije\²F-Ú§Nª‹G)¬k3¦éW.aÍWIij¦`ºËvlãDͬZÎZY5ZíZÚç)]¹Z@sVæn†Ý¬i¦nZ3KL•kCa˜u(ºë4í,±µæTÌœšÔëhS31…W+šºM¬k]+Z«¤ÍXå# ÝZÛNÕ”â6,³gL±Ñ:i«j\ÚuÊi¨Ö7´ÔlëY“”pº2ÜÖ£PÇ\É‘%¨á4›9[[µkE“R¶cPtb]«©—NQ\…š:;N«ck´®i Vº¦ŠÙÌk*ˤê¬Ì¬Ì°DµÌ‹FçÑk6X2n¬³iUZ 6¸¨à“µMrqX³®­LN©`m6±sUkkb2£I–ÖÒÚ7#¬è¦·f 6a&;v£³´Î²ÆK5\eq£L[®­u¥ŠÔҪŃ,eªTjÚÖZÄÚã©iYXÂ…YS%¬gZÙ•²*®sl-ÚimšÃ¤±SLÊÒ•90ŒÅªMÒ§Y2±,ËZë,ÅÊ,ëFi«Æ…5®´Õ²êµ—IšÖ¢Æ›L²«tMM£šÑÎkšZY­M8ªTVNkK5sY͵¦ƒI³DÑ4Ó[[È,ÖUÕ5´eÒ¡vÜÚg¥2ÍÉg6V䉮ÖtÜØˆ²–MÍÎcZuªÕY:˜´¦™"ÇYZZIÓœK™%(u[&&•bWW&Õ•£R \ʶ•1i±K&–\;Xk 3VšH³­Y®Œí¤Õ¤&Ë[d´Zr³-¶ˆ‹kF™‰–3dnY6íˆÐ†µŒEÀ¦.e«­qÔ–j‡ZÕC¥4Ö¤-J.\¸¥YjV]Zu˜Zv£•²µœÓVL–¸m(æ†Aj-mMe`-¦e®¶jDÁ6µ'.94µmfÎÕ5¬Õ-–sV­` XY¦[À–›kC¹©iƒe(æV©§ j›rj5§.°ÌÖºifXÜÂEµ³v±¹…´ËLÁN®Î‚åÓ–í4‰©Õt%u5‹KeR5†Ží$ÖÔ›FÍÖ“Q“œµ°Õ®”¬’­qNÔ™ èé8Âi:°jk²iÅjfSg5ÍTÕeiª—h·2Í%£–”ÎÓ Ö¥]\µ5VVÓKf1T;6jº­ZÈ:(²#¦fn5Ò‘jÓ‘t5I´ë³šIVmÅmÊg4-ƒmk-­Y¢¶lp’a«k›.…¨m[ʬլ:¹[lÌ3ZÌçSLÙdtš.„Á¥Dv«Yβ&ãUpWJƒ,8˜Ø×KJŠ­%­´u;¥ Å%‹8–%Öƒ+­:g`,Ö™’¹+‘t9Dµ”ÚÑU,°ÉšÀY«Zm*Ík*4…×LNºrIÃiÖgF&XN¹¬˜ìÔÌhæ´×!fhæ“ZåœÎ&r’êËT]2b»&C­rØÚÑq]`Zë)]%—LU:éšêÒ˜V¦˜2Í¢ÀËvæÈ61S*k5®³CUY£YXWV‹%VFRJ2ÒËD¦h´kqcIUS!0Öí¶±lÍi&Z"ÖšÌfÆjlJ°êqÔNU1Ò]ްjŽh×gWB­ZM*”hanT¡ µœØt‘f» s£Y5–·WÕ£)V»ŠÓF—RŒ¬¥˜,šä*iM«2Ó:4t'Bµi3FÊÆ:Ñ4Ñc]&Ö™X ÍT·[%¬3²2mYhÍ£“´é²ÌC9£plvÝ­&vÃhLlM°[-ki,E1I fûªî«£ÀPOž7ª«ggE6Ö]v®–RŠËJ¬pX:“Jƒ!sk[F&…—MjX³´é¤¹hæ»Lpë1jª™¬©•Z¹;Ó™N) UW,ÂœÍ ZÂXnl·`¦“mÛ¥j–¶ª°]XJk*¥ŽU–ZÚ®f8­u²]Ö7S•‡1¡Ì’)vkËÍYªÖ[PÍ4¶Æ]©F3šbI›¦µ³[U«›V›5i6Öëp–n9[¶ÔVu5œIÚVf‚3-­s“Ží·HàT¶ŠSŽ£‘Î6ÜÃwgeq¶á˜—6È$›n:(*ˆä\qE'9ÉÒB’,ÒeÙpqQm7am®Ž+¸ä¶Ì–ÎØÐtwiÝØ\ue‚ͦ‰9,-ÜÍQeÑÛmEdÕ’Ý–ËR­QX.˹$RQÇ6»Mvum»tmd«kB#f’¶í&Ú%6îÜìîæÑvv°›Mƒ«7[Ñh6Š8vVé6 W(„Ëw:&Ö¹m]ÉeÙv[‰m¦eA-mͱÆÝmZ¬3KgW&»°«­,‚šÔ·fm´GšÆe³V‘ÖvP‰PÍ¥•®Ð­ÖMÊÎYÆK²Ë ›‹›¬·hX[–¶âÊ(βㇵµ¶“’tÆÄrÚv69,íMbvÕ ³¹:.D¦; ÛV9Õ]×5l,ꢈ.\⊓µ›µÇigge»Žàìá²ÌîÑ.Â%¶Øg0µ¤TÎ[²Ùm–¸å¬æ¸ZšÉk®Qukº¥Š§f†[ee±:Dæ™´gGkV·8eœG\rÛ,í.„Œ5Vª¡­Ae²]UW99Á¶‚+l@à·&Ç%'Vè-tvUXvл‹:Ë6VVqÀÍÌnȶíf3‡si¬b@ºÆì°ÕªØÜëª2æ*fBÎÄ8§6['9N""C‹m‹,sk6Öd£B”†5bƱ Å% ¶:š+­ke]ÓNXY3dB3£Y¸Ñ"-«Z—f²â) 9®Õ®èØ•¶—gYËk±Ìê]TFʵ'Z‘j(š·í²qAw6ÚáÍ'6ÖÚB+PWSZ³­FAtiÚÖlh¬(£Y%kJM%u™#0lÆ4&–6#ˆ™ ÕÅ1–ä’©e*]ÙYÉÊÛ;; ̤Í+ÒÍ8˜š¹ÒíÇ7.Õ­nºtvY­©«uF[1)•ZÙFÓ*ÙrFZaËI±N9µš%´š6àwXY[2Ü[ZKi­M£b•Ö¶ˆYŒ—c«®˜áµ3³¹m«µ-¦å++l¶Ö·Glíƒk ±iS“aUÖh¶Ûn2í¶ÖÇ@p¤DL¥›[bjÚÚ; weÕ®º™…vbjìfæÓ‹bç)-dq•Í®‹nЭµ¶ã¥­2Ž6èIkl²³¨¶šÖvldͳlvãl­±Cªµ—]†«µË$j´†Û¤NE¶ã­˜;‘]§)Ínµ‘Ô9Ý·Ö•šE2®©·8Ùt!µUI%±ÓΪN)²ÍrË“¨Ím—6µ[ ÀÇ+·e«m]Öº›‹E±¹ØìŽ:¶²ì8šÝ¤uŒ+vZZ’vÚâå;l™Àë\º–É™¦U¨ËRqmºç&´mlZ’@¶Èé:Ü-nÒë+ ªÝÖG6ÖZ4v›lìÍ"Û–VQÍÚvg'-UƒBc9Wl·j®J#±aÇm£œ"f [K°Å°°XÉÄ’wvciÚÓhÑŽÙiegÝvvPÍ4›m2hH†ƒh‚ZÐjÚÆUr[- “M%¡Úè µwWjܨ«ª;[k9™D“2‰Äœ·4»eÍcª×Vèu¨uKm!Že©µ[jë ¶5·'bDD³Z.m$qÁœqDvÆ×gIQ›9´QÜÑfÖ‘8îkmÜ ¶·FÜÚ Åɇ™ÆfIkIš­ ¢(ä(,ÔdÇ2ŽÝ¤“jÎèé"Q(×g gní.°íÙ MË*ªì;‹r²Öª°ÊÎŽ.SJ-+v(¶Ô–GJZvŽ¢³(’FìÆ¶›e§iÙ#ŽEµ”W¬³µµ³ ›ZHNγ‚Û!ÅmŠ;[NË´$rá&\šÍm¢ì’ ¡]¥rÙ]rkZ»]v–nwe©[²;:Ëa9Ifͦé61¬îˆK›3‹:Ùg]™ÛT‘Ô]d솛fZÚ4Aš‹™eµ’²ÎmØe†i›¶2W0:;œÎ°Ýdudemff„™ZŽK$“œQÝÁ[k&ÒÐqÕ”Û¹ÖÛL«:NŒgqÅâ¹Ë¬»MŽ» 8»m;-†Ý»».8"£«­Ô»ékW)ˆY‚âÛrËY°Âmgp¨íÀªëC’Ú¶­–€êš·$[°#¨ÒiG5´]¹ÜÚL×ê+8[¬á«´eÆsj3b³ ÖÒÌ«j].­j™£ ¡cZ´]®Zvu-޶tr;]5eš[0ÚÔíf$uW!ªåœ6rÛ% É®jSŠ·"µ&¦[–š3l™šXÒa5Ìئ¹ÖŽë‡« ‘Da–›lËld[DJYªšèYhÓNm \rÓ*Á–®dHÎ+‰Î,Êf’:ê°–©CY²dÑÆnÙݶ–f„-¸bÅ×])0æU«;L¬ãYÍSZë5«,£:²×2rÃijœ¬í-,k›²©il¸ÍQ¥–G\ºÅÐ3¸2#]WL†Ö¨ÑZšêÓ”k6q4ÙYU¨Æ´Ål¸éÉ`j×hKK'Zq•NÔì–Åœ)'5šêµ£”Ë[Q›¥-¬S¢áV±kKËMXM*±K-®ªÁŒêÓ44‘ÉlÙMi¥jÒYVµ–œì™UjY:«Šg9®ºF†²ëkYUlæI¹«¦mf9Y­†:K£ªÕM¬Ó¥­(LÖœµª¨™]⦲·ibÖŒÉ`ÍdË&(Ì\i™i2q¨CN‰š­ƒu †5UÃ\Ê„£ inšlÑh‰9™´kEJ³L¬Öª°f]X˜Ru©•¬š¦—i¬ÌTÌ&R‡ke¹10´¹c¶“h5 —)Â]k2±Brim ŽÛhØ’I˜¬YZÕ.³e2k]ZØÇC«6uYcœf«*µ©œY6X–ìVS-Î&¦b–¶‹bi4f]qºÙÊbµ’C’Ö²Åcæ«HŽ*ìn5Î’‡±YšgM”%js6Ú[‰¶Jni4(³V¹£U•f)AÚI¢t›-VÑÖŽËR44µ¢&ÚÛ¸vg#C†«‹KŒlÑ]mI¶µ–jLªÚ[-©Î\RÝ‹ªf–Z5f¬ fM,Y®BªVi’Î5V-dÍ®S6KYFÚl"n´›:ÖJjÉ•Î̤-“¦v†¹«–©Ö™K2QivY¬"©„js5fh¨éÖXÖº’ÎÄâÔ»VjeEÖšBÅE.‰c´´Õ¶æfì9ÊÑcb1Ö&ºâÅZÙt°qªÁ©”ÅfLÍc§0™e¬eXÓɬ`¥ VfèÐIV×4«d3§ËeÛcv²ÙË–¢0VÈ—5ÌXæ&-¥,YÓ¨;TÕ§ZÊÖ-˜ 2««2Jj]k¥2¶Ë[¹e°˜Å­h1‚V´ÍKf[]hµ‘4êq–³FÖIjĵwºdš²N rܘéš;kK¶GXµµ”Œ±k¥¬Yfk§,’Úk\Zu&bá:]ªk18‰™kFB¨ÆÔS-*éÕM+SŠGiÕZdén«5D:æVDãšiªÉ‰¬MLÉÕŒ¥ª4'2R±ikX¥i³Åk\Ö´VŒéÈ­]YѪL ºªî³(¦¬Qœc5´ÃJJÁ–'2´¥²l×fìÐMÖ’KbÚ–²V²Ó•*”ÆZÆ”U™i•šu`¶¬ÆŠÎÖkÖdÕ±«4ÌSªfZ±tGK+E5×*dÖQB’¦òo>wí~¯·°(¨ ¯%4Ó31Q{ìDP–ÊÓM™Y¡ýEçwyrAÛsUÕé­ØvŽ]ÔÚº†\aÚöóΣ÷ý»ƒ¥;nĉGFèÅE± ú¡:JØ$N{k;¹öºÎààH¢ÎÃZÓlé@¼¯^´·u•Ïnôh ‚¢ ¶6N—¤:”gE¥'"•üzÐHäº*(èÒÐq•é]v¶4|/Kç?|xìy{Ç­ÐeÒöÌóÆN˸ԚLta£±ˆ"ÄjЧI¥»b( ¡ëꞟó]Ø\xO=îë.ËÛD…ídN\myëøý‡¬D6­(¸îÈË1Êf›Z~+Òœï;+k$á:Û/nÜ;ÎÙˆŠ‚#žr4D¡Ðùi`(œ¯š´kr[}í{–u ¨¿Û»çÎã¢/¾Þm«ÚkpVežw½«ú޳€G9žï•åz4ï;~­ôá9ËÛY…UÉ‚²Â¢ˆ8¤¨n,çWHÜé\\þwH6—gdÞSÁ7^ÆH”É(¢„þã[¥É¶B îÒŽºÃ;­°ÙY—eÙÜYvÕDRM÷H«fš4ˆhª&¥KîSDQQ%*DˆPƒB (¤D¤)‰5¤º3E.G*Z¢K¨»;«)J)@â)j$”ši‰$CÁ*ä¢P¡ÝtŠ Zd‰“ùXt” $A@ÓJ- ´¬DC H%-w¥2 !ÁH”ÕL´”´*:)A*…5™>ì`”ˆtRon9¢_ºÇHÙîÒ ÐЦ€ 8A  ¨„¡iZ”6ÇÝn²”+B˜Ø N·Ýh¥âÄ )ä'u…Ц’ òƹìÝ@2Â.ÖIFJ‚¢7®ö`blÁ£Z»5ÂåF¢Æ;\/R5uKŠâ é×[m;`¡=q¸õT=!¨Ÿ!ø:è-Š39Χ.C]í´ÛZ9²š Z¥4ĵK¤-ª’vdϾ3Z z±7z?RªéR”ª{gÈsÐ%ßNì á>gmc½²G¸) «¶{»’±&,CÛ &´Û‹Öàƒß*¼Ù®º:VÍ[§çïû¼ Üåù›GãKñœ§_muÅY¥iÒY•m­µ~kµÀs,¥kv9ÈÚâìÌîB¼«/)8ŠÚP¡,”‰~”ª1¤á4– 9ràwãs”åãíè¶m4[¬cËÔú—׬&Ƽ²—a8à8#ƒS›‡Jauyw†¶^€ÑIï‡KIå£F»l‘ΉŸ›Ýòú‘gT'qÑÑÏ›½›9·dïàoÍnW”­k)i­-FǾvŒwmƒEë:¡»¶É"žÎž›Y –ª©=î’p”§D¤´Š½÷¾wùû½ÞÇcÑãë£×=_Û¿š«•,ëŽâÛYXp(èäè¬Ê¿k¯{wYkj8£ˆî"çHç·n¶Õ¶ëN˹#¸³´Y£»-.[RjŸ:ª!™»·E7lTg:44ëEV—öq¢’… '¶rQ•§~{ß§Ù<ß·o4Íå’Btœ•AÁÖZvì<@om ‚):ÓÒêš³€çaËÛEö÷^Û²;T¥vöÚZGÎËê<ɪË}¯.H½±^ÇiÔvHs§D䉥o{×^GB\–]—O›‹Ëôöëí`ä) ½ù«ÉÆœ~Ý’šˆZ=lš4nÂo2ùÚ-¨4kF"ŒÛ%µæÉ&ÛÏÃÜÓë×5øcÉ(€4”†ˆ¨õÝ×D6êÒ8èü|¯N¶ÑÑVtuµnOÚ/ÍbÚ#¹$mU´Iˬ syºÈí›6";1]—¨F¯LQ£´âU«NéJ]ʶ”E3 ZXŽøHÆ*¼‘ír>_±Ù ƒ“¹¿Õ¹"‡íêbïÀ¨¿¨"ë%Ô*‚€Â‚*‚€úú!{𔥠´a™ge”-ªÂç+;K-šÛ6 ͵…b]¤ÃD9Ma³ŽÑ¸ÓIaÚY¸XbvY‡Y¹­°°ÎK¦QV´l¶¦­€£dnL6v´·6e­,ÝKZ–ˆ)BuÕpYV ˉbr±FNQ¥šPšµaÌÎ¥„²–N›ZKS¥B¦´D)1fZqnskEµ2ÔÄdÍÖ¡²Ù-f3XË2Ѫ¨rº,Â,Öɱm­7%Íh´–8³¶éµl¬ˆ#-R‹–LI­v°ºÒº¦-eŒEbÕer¢ÔEª£6£,­Ec1d4!Õ-56bµÒN\&³fªÉÚèÙôÕUκªät³)3.§YÀmZΚ¹ÂÖÒ²lµµkv²L¦‰9›çQt•a´´êÖ5­ R«)ͦÕfNˆ³ZÍšNkrdMÃPÜX¤p«U2Z¹²¶’š:c®¥´léÕ‰[C¥“;]Z骔YidÎêæY±LªNj+Õ¤ZÍ-Y+¦r”UÔ¦·iɳ;&ÃIÈ™”Öº&­5‰®¥Ö§lÚºH㙊¬&„Än2M*âÉtÖÓMRFv¥l«-PÖ%9j3’ŽL†É3Qd4ΜÚ\5Ì´ÔsuŠÍ1U­©µŠ±¹J±¨2ÁštÅ«TÕ¬ÐÆY×F‹•¶Õ­2Ë S“"Ù`F¸±5f­eZZ’XSŒ¦±‰¥4ÎZµQhl‹LÍlAÉÔ«p¦[‘¬%•·ešnÒÓŒM¶7,"Ù­²ŠwYÁ+EV¥X,Ö–”írá3V2e†V²æ3§* R«EÕJHÖ «6«L¨µ¦+Lj5 Ó‘ÆÎÝ)iY´Ûfæ5–´ lµµ#)ƒb嬑‰¡†ÖUÕV:­r5ui8sFwZÕ™®®Ñ¥ÍZ슺”1Ú®'S£j««-XAÕÓ¦ÊæY2e³SMš Y3I¥k•£®µu‚ÔLµN'NÕÕ¥ŒæµŽ5hh¬¬Í ­$âk:ÌÁ³jš²3SZk³Kr¬Rœu”ºW+U¥­XmF,†´Ø3X²ä¬Õ‡"ë;FLÊqk6mK+‹+&¢®´Ì›LrÎQ«&V6œ•Ò65ÕӜ餮,´«f­*T’­\si„ÙD³ªRŒfmu’­*µLɘ…«™Î¤Žê+'2¬e²‚É:™Í ZêæÃQ¡¦©)¥lÖ©i“LµÑŒØI­¬Éšª§Kg,Å5r¶™ZáFM6í[˜9 ¶ÈÙ¸¶·lÃ-,æc)ìÎF™‹RjÄ·4­Z´³"É,ÍctÍ]Y±U]-M)RË\†S™Ú¡Ig4utغš:íf`¹iÑ…L°Ìå®Ödµ’ÆVZƮʫŒ“'#;fÛm‘"m¹²Ž*.µ]VvÎë”S™RÉW–m&³­dÂ2$ƒ*q·ør2ê×+áÖ‹f×#66¸ÞÛg!hëú1Å~F,,öô2ˆ2JM1tµÎvWÊï:¼C¢èââq9QTIT¡K@•JPޏX¢*PuȪ҉A®2PªZf‰v² B‰¤ÅõÏe×EwHUAҔ들®èŽº:î¬êé"¨à(fŠ((¡ã % BCDKHÒ k*ˆe„)”ý/ìcö2“TB” (Ò±úrš†Ši^±}Ẍù?\+¯Ø m¢“Ï9éiXØGH:Úʺ) ¯'téb KÖT$HÞetÀîèR¢—Mí™VãDç» Í¥¨c-Ü­ë.ªc®‰ÚÕѪÒîj9 $•™ —uŶ& ÅPÚÌF³³ˆ¿^LIÓÐݶÁÑŸlù'}yz‚*JëάbŠ&ºJlb$¢%wׇ®ž³¯—xA$‡tâs‡/ ré)÷¶öu˜›6'BS` ËF—B‚ˆ*SOÑÆ ’ìãFK¶j¯ªÕÉÆuÛ%NŸ .ƒI§AÝ…çYÇ—ÎÝ=·¶‘:q?‘Øuy‡|»Êò,¦÷£@uGI×EPgհĤ1}¦ó¼èà:VfÖ>¡õZÐDšË @+’[h‘Òö;€¤±ƒ™’‚«[e'»$_Zì/½®ò‚ùÞãØÊòï•zqÇ$ŠöŸÉ÷«äè}³¢..¿‘vj¶ÌºIµ–Âm-·r]4  û§3æÅíǺÒzÆi.–Z5‰ºLRkL¤Ô"c±”I¡ÑQMë>¨éúãÈÞÜr_©skæu”Û3ƒ»~¦'%#*­’D‹TÜŸ8²ß›“1 ³.¯Áãæäëóz§ÇDPhâåœ_Ã’"¨(nÂîÎTÕEÜAš8äÎ3ˆ³²’ÎŽŠ.Êìè.ýuûV>÷«{½ï!™:C¥­hª )¤Ñ¥¤¢ÙJ R™v¥ÊR]·b}¶[o<¯lF]dqÜvÛ¬²£›w•á^gIJfÚߦÏ"òÒpÌSú×½œâ¬ë ç‚P<³ õq–š‰:ì%ÆÖý»üÕ¹H²ß5¶ö´qtµ¸íµÅ–Û³y©Î(Ž“EœÒRÒ;šb˜¢ëÏh÷¯yǶt€[YÝæ†V’ŽU—gPTivجèêËÌyh)±¥Öݺsœ[köÛ\$trAÀpUV\ÓEû]ý’ïûF½ï^Þ÷y%öí×–yarW„wIÇVÚ¨çãp>ÝóÈý»}µfˆî.Éí×½¬»#Š:D ØmÇgg˾^\‰1/´|þf­×%\ÚµÜm†oµÞâ[B*\¬q BÈ"³n—½Îæò¶¶ÑP>_W“ÍXYQ6a”ÍÅ¡™¼OmˆëXåQåkf+X ȼæ1†5DLYqÎ1iÍ”„Æ%bÉ3™U^VÒD“„N14-t`,ca.f‰N„é5iRÔÉpçœ*t\pŽ—Z3Œ¸åçɧÀG3Œ-yÁžNG¨µ¦[ ®Õ…-(®2ÆN­šA×'sÀW ÈY1 =fKi”ØQÆ1‡‡s“›Nºu§-ÖIåSœ’Å`›\k™pv 1µœ4R˜Î˜Œ. 9âsi81†Òƒ´ çÇ€á÷ÍIÞg;ÏXõšcš¼ò,­N¢¦NÈŒVV´h2+’©Š"°Š×&Í2QcV5Äf.K‰­ Û fÔÕg8ÕÌ‘MZ& 4 H›‚‡ƒ"f+…°®’­d.Z« bÓe1Qh° ¦«‰ët£÷·‘sÎðÝ€UM@àÙD,¦::´ÐÃB/õ`Ñç3Î:t.RÁØäuÅA£L#³Ä'¤óÀŒzL§‘QfXCŒXè+H‰Zšm}zS{8ÉÉ„Àb 0¹–œpàW ç;Ê^ÍÞM$jŒÖ'¬Ø2c#Á¥)2$gCSo ä±ï>½s½êp-ïzöWHóiÀÛÈÔ&÷‡`íq“5inÍÃÌëf Ú³¤d=­`ðc½:My÷x5ãÞ¼B]ï>ö˜°TRpÐ=å/ Ös”â%_Nö™ef3Ak>#ÆSïoTN1¼`U6!4]aÏP¬PF \ç„d.1€×–O]ÛÞ§>€õ­y³Šñ—É´kÃm/{Þ&,©ìy÷¯z ÔÄÈÚ·€ÞMïk€÷¬®|­¤ÑÚÆ\˜/7¡ åœÙ÷­wÞÉçQ¸Òòc‹¨f|øÞñP5ŒÓçz÷keò8OGä=yòšH#É£9÷§ÌËÛÞö bôÛÁ#Is>½Þ=­ƒ8ŽrRMto²#›wµ^ºt¦½íx/o{ÌO®ò6ô/W3Y¾M»vøøø\÷À¦ß ’´9ÄgÍã‡Ç·µëÔڼ̗{H¸÷†Íö­ëÛÚsz÷*Æ|¾Á ãzÇW³â`ƯãxU-k„Éz˜xobTº/¾j¼Ñ½Kåõï{ ª÷ãW{ —0"1KÀ^Ö`2èqcÑçÓZ/bðщÞKëÝyXÔø|N}à.NÇæöòy*rU ym½â…Ñ©X6TõHr£gÉîlïW§¼‡­O¦OZŸ"Vža Κ÷žöoBvòá›x ܘ‹ÞÃíyu{ÑŒxåñxÎ|\¤š´³ÔšªB}&'Úáàlåp ÷K°Œ2ø0{¥^pooz/oFò¶÷^|õ39|ûÈ‘ö<{¢¦}ï)±1åFç>Eõìû£ú¼j‹Ü>±y¼rªã“ÄeµŒXŽqODbsY-y÷cO‰Õ¼úÀfóïÀl²‰•'\‰²ûÞ½ €Y÷¼{Ü¥hò}gϽÐØCYEQ¼;ÖõÔä2eÚ]¬Ì–.#0EÇ¥Ž™ =okص{Ë™_eö³¼öºypƒòyòøÈÕNlR´vAtBi™L8 hL€”ŸƒÞó˜I,Žg±´)Q­`Kyö<í^,so^ÑyÞó_;/šo[ÞiFUÞ5çz!÷lwŠC{š§9÷½ï5¼ð{Ú‘õ&´”ØÙ5¡¬Øq¤æ“¡ÂåYtXØ={Ø=¨m¥;*àÑ6Ãæl¹ã¤@^Jc£K#Œ ‰¤ÇF´í=é¤Ò×¶œrdmªóÚV±êb[ÛÝNË­wK 0›C¬èpð9hMjÒSšÌóµ™Ð™ó’ÀoZ²„Ui +yöñáÃÒQ•ɰÑd1ï>÷A¥æÖ¸¸p6qg6YQãTbMRq‹x÷·¼¯)–)¶öÄJ¬²Ú¶3˜Š¯éYרÞ&8º%Y°M@ZAB­éßøåG k¨ÙFh¡ ÌÂND¥™énȪ.ÌÅc+ª‹ƒ9@.3„#ºhÐFCBlFÈ3Ã|ï*‡²íZÑs`)0&óË<'Q8^h™ŽE¶XM“F«\CÚØ4Äm#Ñ›F•ÁŒNmvf×Xxxì$`Ñ Š!‘\u¸š÷±–LžÏM¯Æûap<ì(½Ž\¹Ò˜$ÔcƒF—ŽUŒñ¹r<DchÉÆ%:˦0KÊ£µ8ƒ&r €¬u…«3aÐV€6¦ÂtXÒáQæ¤rj3ب0l&š >4ØõB+Zã]Î1!ÝHá«Õ¥Ç/.hÁƒhÀ`2ò`r¹Ò¸ÑØ2t ughàª$›=8ÀÓPòRÖ¬EK¢ØX ‰²§hÉ“DC!9btèrju¥1¡*À•Ô‹«ƒº”—I‡ g%ˆPÍ Ó£i~yõ®–Ñ“W&²éÚ,Q\®5ñíîöÞ×FÆTÅ]’¢3V "hœÎêa1“QÆ5„1†û×¾_)½½¬nÈCݮСrÁ]^`7žÁ½Œ Xàr­s£X¶9ËŽ# JnmemL£‘1`àМŠâyMç= óŽTÔòhQ Q6,:9vMV#c™ÆW„Œs•D š\RÔÄ\ääNq€©¨¹xL½=šñ»ge¢!3\8p‰Á´F sÉÀãbh8xâ£6ªÃEYÖ %erKŽÄ .(:Îsjµ‡´`êÓà0VMga*°èµ[H†U95*šx4fÖ).Š¥Wà3š%œmÉÇyâöÅdØ‘L܈ëDÛozÓÖË<¬Œ`äÏ+9Eq0kK.¡± ¡Æ¸&1\ƒ@2X” )±œkBl 3HOÙk3® kFÐI³Vx1;\)6„J¹–k¨.CkèW.š#B¹¦Â& ÒãCÒcqæ<{] (- –¨6t ÐÈ£‰Ã­qVœñM“]1v\ª:a Ä883r:¥Ðã„te£1¨a­kV¥3£2ØU]ŽÀ8rr)–GZ1sÖ14ÅZCk˜SŒŒ¹vHç9z\RR€h† + ˆ ÙÎe—0º{&¹®NÁ¢¤ÖjNN&ÕVlçŒjÓ8NA9ç¹0›$s§1œœÙHè‘q9ke›6ºçë ƒi×0›‡¬¸† ›Ñ£²Z^åííîáMëÖÚæ—bZ1Ž˜ 3s"ôŒáÇDâs–ƤœãJô HʰU± „‚ÎI5 `UœÚ-dÖ¸‰ØÇ´FÐÈh@L/+ÖHŬ‚¬b âCkŒ™VmZT0§s]šé;W@æ½5­jN0˜s¨Q2GDZm9&ˆ6ªïz×Qí›! <Þ¶%¢`‹q;# „§[9‹´2gB×NažqÌè‹ÃfCÏjOuZÚ¬‹—F÷¥FÄìªd9a]NŠƒ×’sÐæxÐဃI!`î–ôõ¹ëY£Ìe¨ö ¶ƒÜ†‡÷¯FÁ…Ñ¢k1N49! i]É/q†YÉ6ÐÀšÆp€&§6j&$ÌDrŒ*DÔë˜\ ®u€“‡Tp`Æ«E@™,ðÀK£©ZÐÓ9åyư˜^çØÈŠÀb˜ ªÆ†œF̪c`§"ÚÛQ6([<(*gKeó›Æ8Ðcc@­¬YŘ‹22°8.«Y³ÂÖ„ë/AT©¡7ão"Ÿ)¢‡'ea1S£Œ„% 0šU&@[Vpa-ZyE2¼N’\­8l ±’±)A« ½ŒèˆÒÆfØz @K†À2v\‡hnÙq™tè¼›0Z5¦N,ç h±r&6KµŒQ$a[=p;7E\g™Æ2ôS‡—žâ´`0¡†Áe)ŽEºŒe±¨³rq8pÔâq©y#8Ó5•WFXg0uRs FG#‰“: ÏŽDèWëÜ`$0Æ•äu(K"Rº0CМ*Öá@]ôG8˨Ƞѡ{&3Ç«YÎ`L…žTq…È(æÕØ 3™1 óË 3ÌçJ„¥ ÑÑÉÖÐäàÌ‘¥51&F-`q*ÙÖ.eP³Q¬Ø8r¨¾= ŸYQU†A"‹cˆ|9ì †Ã…Ã°H!‡%XÇ2Î1¦^ÊËyxQõ¦Á)^ÕUɮ՗XŽ @Ø5s\ä#!¡ÀšÓ/2‘åçqlꛜœá¹f!œ‚€¬­ 3$aÖägDXÑN ãì96 ‰Æ œ‹ sFx9’çX2ȺgSÁYD]T㣜g§z.*Ö¢:·›¬ŠúÚvŒ²‘À™7Ça’¹¶M“ìb1/•cÂË9%1ÁĺÀSOZ{ eÈrHN…êɳ8L5 ­&ƒpè°#Á ¡Ó•C&VÎ`°BbiƒDµ®1ˆ{3ÐÊlÚÍNfJa\º*…,N¹z.Äöp£ÌìéÖ)èÝ%;æÉœK§“HÚç*¹…ÓBdÓTà1ã(äÈgGPQy†AÀô"Ø&ž´ó¤q-œÂratèÒng!‰L á Yصiõ2æ¶ñȇzzÇVWVc azG9–Â<µª¢ã¦tœ•ã^)o[=fë,Í:y8à‘6OLs¹*k&*¥À¬8&PÐÇd­(p.­• M`JUH “$F•Trå”ɤ‰Í¤2vÀ()Ø0”å¹®—VsWPög€; &4÷*-ˆÀ;=2h D>;ÞÀdL/x iqu‡/XjºƒcŒÔ‘‘éÏ”X.m]ÊôŒCѬ:ÍA GFÀeÊ(´–Àd  b0Ôdyç" 0(4Èð ´h䎚MM7 ìPY‡ œŽK.ŽÇÀÍ\”ì àDÁMªØÀ"Š.–m&Ä‘œ\ͬEL†F L@ƒM)Æp.pòa"aÙl9ÅœŠì©bj$…œÙ—hŠw8Ü„eȆ@Cfá\Ú:Y¡çOë‚D´¸u)¤•Øœ™1:@lk9É­gf)êÁ“ ëI€Tàè²Ûy½•éå<ÉnmoFfµØÊÑÌòd`n\JgÄÇ£]UZCJ¡¥® f\\ìc&ÌjΜ4 ÔçƸ(±ÇŒŒ4†t±9ž¬ue;D˜È. •™Ô ¨T9qg¡tTV1¦Ò&vx˘M¢‰éŒ›G&iC›%¦ÍJÚ²d×YDÎ\œ¸pé€2GNxÎtl%›%8h%ÄÓqV,ÔȯèLcRa0fCŒt^Æ‘¥ìkÍ=j5¨Ò^Z†Ï.C’¥{H­¤¶'¬Xy™Øz³)Š –K§• dZÄ Nãà‡58¤bP&˜Åfp¼$` cBXHÂÚér¸32˜d;£G5XèØsÈ.É‘1¢Å<ÙäÕ„ÇK=6Œ³}ä±È¦õ™.r`âyÀƒÌ+ÎÀá êH8 j®ÅZ2¹2` ³Ä*.2ð“œÁ££gf 0ä€0v˜lR%‘˜Í„®«§µ”0ÉFXƒ.ZlŠªéy+&!ʬÎWvŒHxpM`sˆÌáBÇg¨œ/zÀ™΢$»Jfš¶ÕÕ4 "‹ËÜà˜nQsFLÂ’bÆÈ‰ \Î,]"å„ãqDÏ\ä÷½Þ3ì#¬á³†&4’q/S3feÎl5ŒÀ»VL·3r¨dÙ4gUZñakf³ÌôÞ˜‡$½Ä8s££Mq¦MYÊ›ZÈkZÚ²W&“­%jL&‹(¤rKÁË€‘‰ÑBLX¡ cƒkXÜ‚–Z¶sžAÙîKN{i*ÚÆØ$˜Í$mcŒd8„1ƒ¥Öë9Õ¬0b—µ= “ƒPÈì‡C9P®Êí’lääH¬˜çB/í¦°X²öW£3‘²º ƒ““˜tŽäáx*»ÔK-&('0a=fô…ÁMJX1Á£°¼dºÃjº4¡Ú«‰äÉ­M6£9AxV"¤à¹Btr1 j«€C gŒ<¨r€ç$Yj°‡e­ò-«Fu ×NÁ©Dd)Ž‹ 4ñˆ‡\ò² ;:,BhØ6²*åÈ&L™¤‘ˆ‰ueàEYs™¹))Òåæºnn ‹ª 2ܦ0lã& Ñ…[¬O8Á³œ†MBŠàq`†Æ v$†3ÀØc:SclÉQªì-ÑT‚vˆ€Ìƒ€àËe –@Ô™¤À²XÒd‘T5&/,=ëEyÁ{Ï^å›cE—šõ–ÉÖ³¼¹¥‰´vq²1mCK Õ¦A6Œ5V®‹Ç4,t½– rÙ8rH+Ôevt)¨Õ‚…Çkozö9s½œÀ½X#V0ë9L`PÄäœÉ˜Ò ð2ÙÄ1ª¹Àj{GÉ£Írò¢mMÐ94Ž ¡v )SePÁÎ"ììMб°bMEŠcK:\åØpºÆÉS "¬CPÕ6&;SŠ$2‚DÒã°!N`©Ý.Å9t9äG·{{Òó[a˜¤¶ÄšpŽ”'뢱±†ÆJ_{{ñèttlñVœr¦›Nu«rLjtBÖŒÜÂe9'F&ËÐk:²"E%‰^÷£Çµ”r^]³q 3T`Æ ¢°$HÄ× ¤Ö8¶<#]d±WÐc'9Q ÆujŒÔ¸9 £d-:±SÑ”à£5ÙÆs«I›£d0bØW—m*KJÌc0[:µª¬ìË\Ó=–c^^{o3ËÛ¦¼3§m1" eÀ£Ç% KŠKÈŽUS„U ¸¬EA6H²)•´ç œ\ð˜Tl&ŠÔ†у Ð¥Uëö«xï[E–]P-ƒaztlf¡É Ý:{$éeÜÃp.¬^uŽj¥YR³ÑŒc“ˆÐá¹0„rKòvµ.mУƒÇ"½ Ï9Ñ8jÓ::j´‚®vx+;X.•Ì–¶R d&D[Fµ+M³µhŒáµ¥,ñŠ’,E,= ÈÙ'¸œK;&Ay:#š,ä«QfV-›FÆÄ'795“1U!&¢9¹ ‚5Øž°¸éêÎÁ£`Äò:š]C’Ä‘`Ó°.TZæDã*DF@-+#¬5:†6g¡]š²½Vƒ±uh&ÑHÙs ٸε—XÙ©dc^÷½‚®õ’Å,›VT¤°ìcŒ(n®XLV,¼óI¬rGX€Ah0U@ˆ9Raëk ²²<4a®&g:ÂõuÁc%Óe‘#žUÖ6Ñ‘‰Î¨Ô‘Ò`¨Ö9Èh¥b0+§.fкÅ;pàįHò˜;'!E¨ÎJAkŒK,(ÀÓ‘åEèÖÏ>mo'˜ÁÁ¼/”ƒ¢£‚bÑ¡{WjÁ¬¸ kYÌãÐ3 n¤š¯r‰ ¦šïK<·=/zgµVõá®Ñt·xT¶ìÖžuÍ‘˜Ù½ï{"pwŒð•IÎ0åØ&b6±9Ü=6uœ,Šš°»h ÓÊdÑ›91£‡HËœb8Öu88AáˆÐÊcYÎ0.Šsƒ&BÆIª…bÇ8DEÆ2Ã!Båì‘…£$K–Î 8èe tªRVLŒ±H…Ú£k –¥Ì<ÙÃV"=k:iÀÒñ çi€—/hÆ£²Ur@`uRãË€¸Ìd¦0`«JÈe1š£J¹NYÑhƒŽœ³Ú `Á=™V7N[ 84ä <¹ÆÆ{¥™ÖÖ\âæ 8YækK›!½{^6e^ö“Ô{Þ±«dÖ­• :ÇØÑÁ%ck¶D§ ´kH¨ã@°À®k¬EÁZÌl½:Þf›·v;lB"M« nÆÈÛ4°hÏ(ËÙ}®¥&•(’«YÖá XÐ0wh¬éÃÒeÍs›¥Mǽo³ìµIËŒ‡›" #k(HäàÓ“Fz0 ã;ÈeYm‡kI³ X¤'ŽÉ•l£ѦÄV'œtdØÓÅI´3£hÒw¡Üú‡/D»£§Á¶€Ñ´4—ØC%:yL£¬ ˜±Zé’L¨®ª€ ¢ΤÌG$Šå`³0¼á4F‹+ †ÕªµmB¬EŒ–%KO+Ò(ˆ 6šαmVŽ,gB2c¢5ÇhtöI£,&dë4éÍXkƒ2C¶<ư.Ŧì–7Ë–v”åïlŠ6ÕìkÖÙdn*ȸJ‰È˜8Æ0Ï3”pF§&§€3ŒâsÜH½`莎ã£Aa›Z±q£’Í„yª×3Âϧyð*]Žk1Wp˚至ÓÜžoW,Àð˜0™ÙºÒ±”µdç*³´A“q¨Žs˜{DjÏG&çcF… ,ö2å«§«££#X°Fé˧¥Ñ²ÌC™ÒäL˜Öx0º8:“X ÁÖ4dZ¥03Ú"5ŽMG´„ uš{FËœØyXÆVè9Z›3ºéìòvªÉÇTÈSjÍÃÎD›dÙ‘x¡@.ˆIŽìŒêts$¢jå2ºsƒd @#6Èʺ+„«&Œ›håcwQÊ á;œ' µ€Ž¡L­Fºi¶"Ýwt¼N‹ƒ¸¤ˆ³” `P]½{AãF¡DÒ¶CJc“CŒqȬXÐâ] <‚ì˜\„Žu=€ 8ì›då!ŒÄ#†GeÕbna æbrSM­‹€ÚN±“d|·¨É–'&Ž±ÉžÀ9s'ˆ kA§9ÆŽ³É­!)¥gX :ÃÆzœ”LChÈZmsc©êT檻¦°Öéi€´,èàÁ¶±1dW`Á´fuÆÆA–œèâhç…â&HéÕ¹páCŒÜÛN3ƒ´ŠÖ3k¨@1&^œN¡8œ BèØ0ºá+.œ g'.y)ÎÚ4–ìMÞ²`ÐäWŒ›q›8Ý.†UÆiÒrË'*5©q„q9 Q¬9ɈsVy ÔQLãS®œ. @ÜæL·( ëIaž'K%µÎ6DêLÌk\ô±”B–“tÊ`æ¬fܦy„ˆäŦgÉÑÖ;þŽ}„ñ&WV´ëŒòŽ0èZCK±¤œ`ÄuLU£FC€\u+6”(4`ÈÔ:žŒÔ.2«(Z4F].4¡£:1;žqÚj4(¹c& xÊÉg»GDQ/FŠÕMÕ­#À˜s€ÆT$­äÉiÀ’°<)ÌíïxM! ¡Tðs1ª˜Ø›”è=0ƒxY„—…¡ÑËNÁƒZµÂdœš :8Q§È:œjK ÙíIÄ"lNÌ=……x%×sœ%SG ’^u£6Æ LäÅÇUpdè,ª£‘J—²háS©€ì9M#Ó")2æõ´vïžãNÔ€`Ñ¥¡+‡(”C¨2E¦evqÖ^AyÁ¤Ú!tGR-’]i„¡¸(СC¸ ,g2&êv% ÍA±„–k µZ@r"¬PP!–'цӱÕreZ)Ï\«™Ó\¦l+’#0çÆ ëPC …v#“’7¥3ÖªéÆŠ;@Œ]T`6´wÇ áqÁÀØ¥šÐóXŠª³ˆÆ3ÆæxfÔ#pk>còpzzk§›Y¶zPÌä ]jˆMU܆'&;¥ÌÊÙÆ V­dq9ÈX­­TÎ9i^÷2øÖ´m (qÒcɸޮ8(Õ±;ϵ¼ãb{‚ªAÉÝŒ±ƒB¸”Yïø0êpŽŽ€™9Òêa:Íe¶J¬9‘Ç÷&ŽO g>Èí$iÚe‚•ÁÕc4ã,³hƒ Úa‚Ælë›ÍÃDÒœa4Ó‚T©£Öw•ée€{Ï'®m¶”ÂÆÂî…w4]£a(;¬j®!À›dÀF†”Œ,ªã<áÄgI—ÌìÄí€"™]\ë=‰§ ;ί$ñ: lQA†Çlm‘Ñy–ÚÛGqoÆ(‹~æñKõl¸îɵÒeÙÑU˜›Û&(“ÛéÇÓ{AÐÕ4ˆt)%?ÃŒDÇmzÊ›²y²tõå—IÜ…ÞÖ8:ó"öÜQ—NhšØÐÖ–&ÙJÊË#º“‹8è×óµ`œ$óuóÞïzÄòìŠIÜê¼£ùï.£¸½öÒô­";¶Ç=z·^Ö²Ùdjè¬f­zÊü/íîÕºó&އHb¥¤úq©¥è1—züõ¤ä½¬mÅæBOµdÌQ1y­†¿$õí"'ÖîM³ªtCçwK­MUÝÔM¯§9ÄNRD–ƃIwœyËt„qíªÌ8.uµœ8ãõët~1mi¿^ïzÕœAf—dGŸ‹Ï8¿‰×å·%ù¬+Ì’9šÛqÖf$™vÙ§yôçˆòqDECôŸÊ½$ž|òŠþ†×g™¿ ·ÍEM’"­¥®(kl.”-Ñà…ãirŸ»i¼i£½ÁÖêô~¾tzhˆ//kƒATêJR"r?snrl"î³®àŸ³uïh('gh‹0IT%4?ŠCÈV&"b-ŽŽënÏ„ÒPTÑ$y€×˜Ú\¾ýÍ_:úuÙfGi–Ú;"â:ï- ö¬ˆu ¡E7hêñÀ ÙÓô'w"yó`ÓǘN“ÕkÌâèÌÛ: è–Ý‚æ¾÷iíç)eo®ãis@Û%4—vMÝÔ½µæGvÔcfhh©1XѦ)¶5¡Á1FËŠ´hjN%4ÓCNÓC¡Ð•EhÊÆ¬…È::vjm[ÎÉaSprwª@À›'Žcå-°®ÖÚÛ´¥M£¹Ó ö=_~{º}d¼ÁÖƒœ²UdÚ]‹W3ܮɓ‡96DÉñ£3QÖˆ¿jìyØMÖÁ‰yß±ß2Ï•Ž'{ü͸Š:½šÊÍÌ7G~§ÙYû—©SM:YEdb—Q:ut1³3o¿ÏÓ£ÔûžŸO¤ùååñ›$yDTõ¢€©óÍ_öº¢£¸«:Σ¶ÁQ@„W9DShí3»,È5·sm9Ë´é¬Ê3™LÖ—.Y«QÐ5Œi¥GF³SX̲fÇ]iM]­I6³«)v%râŽK3hØ´šÊDÖ2²*]­2ÌYZ¥rÖZ·6Æfbu†Ëf×3.®V³ ±1 ,ÍEZº´0Ë¥:—]tÄÚÊ™£‚®ÕË[9kd´äc(jšU!kIЙ°fU3K\­mN¢ë0«ª9juµYÃWU¦ŠËQ†…bÖŽt)Ùq–…­XZŒÈu&•j uk)\-d YL¬lX‹mËiÚI&³±Ciɹ¬±hnV†ììì,L®ÌÁMjÉÚɪÅ5Lc*×J­»)ª†ÕkLä-Ói8jÃ3+2#lÖZIrRkZT7'vMÍ›e‚ØBÍCš]–&.Ö®MTÍj®c•«™[˜5©˜Mi&«E®á³Šê];&ÐÆNušâ3+I[«+]¬ÖTÕ8R$f­rÑ–²nj©qW"["M­® 8,kD¶,IÚÍ6WXL5Å®k¬Íˆ9Õ“ £\Ó5iR¸›$ºÍ,ÔÆ¨H¤MuœªÌVm2ZäcŒÓ\æ²Ç]%³µ¥Q¬Ì®uX] 2±’•Ó5Âã¦hÅ”ŒrZÖV[ÔªÒÍgFÂÔº%ÚÌŒU˜"YBë6Yc:Ž®t#Œr¦¦iªÒÖÕ˜Ç5ÎtMJ:¬$æj¸eumf%4TY̵¦‹¬ÐѲÓiÚh›\aC—mmZF´è6[7r׫)¬¹Í*›\±p©­e´bYj¸,,ÁUTæêMÌ+#Zèâi¤…beK[WLUI…LêÍuÍa,®«4uté–i «¥fºì¸ÊjJ+…¤-†¥²–ˆ¬Îb¥ ³UaÓCkk)É2³Zâæ˜Å²Y•Ðdšt*jØÚÆJ¥™l¸ÈÕ:j´Ô;VáÒÊ–Ä+G ««\ìÖËJfil.v1œé6¶ÒÖÍ)̈BEÔÉN³Ž•5Íkªeji[5¬¬†Ñ¬Ì›Y)e`ºh‚Ƶa»6í›p&œ¦s*YÓš¹šL³ZU46šib­$Ø*µ;Dµi–ÖãtéiJa2—\š«Y€Û0»­+NfÇ-·h×IʺZZÒJK3£™­H­N²Ø“4,‘e™¬Ve¹Zºk0RÌÖ’šbµÉ­Œ)¬µÊÄ®«:3Z9ÀÇXÚÙ¶Ai¦YXâ71bXÕRÊ"Õ˜µ«š™Ù4pÔ5Âd®4Ëc]4´¶Ó­ iËnrÝš fI¬Ò—KU¢å‘¬á5:s–edÌÕrÕ¡\(qªš:¥Xf\×&XÆr¹«L«–š´ØÔ ²Õ¬P´ÆµU¶-©±É¬Í¬DV霚u­vfv²ÌÛí½íæ±>„ÇoêVø©ð7¿Ú/aýîlø½^ÖÙ–®¨ÛI«>—ó=ÿ§cä–µŒòB0FŠÛ@l)ˆ ÈTCP¿l¡£@[a4 +Ji û}®J º9C”–u¤h&(1 ¤]ëJ¥*5]W UAÐQ.:Žî+’›å‚B€F°%PP3¶BŠôÚT]-*D‹³š jZ‰¦©)Sí”HNÀ•‚ rÄ(T¤ÐÙ `– ‰T ”Wޤ$’;Ò‚R @SH´#J *P¡Ç('›(©HEGÛ”4}’"n0 SMPìÖ!£í +±„ØF Êb´:F%"ûH K8˜û1 è™ŽE#óV_›ñå¦rüo&¨›¢ãÀep ñ-j¤ëί+Ó-ü»ïYGÚßn×óõó²ú€è{aÓ¦†|±?Æó¹ç³M”í«"›YÕ‘4´¤CøaèЖ`‚hn¼½Î"ŒÒT\Pqœ~;§ª' ¥‹¤ÓU§lšÐ”DQÒP=önÙóIÖŠb‘Ë;¢'·bGsÙ´tWâöÔ5«ÍJ³´¶È”žXäÍJí½ìŠþ]¨ƒ`îÁ@Ð^gì[ÖÊ\mTWg3 ÍÄKnËÌ,µ:p½»íg%µ¨+Hh1&‚šCEkQ-/l‘tï1™63æ0qÆÍ›n²ã::Ä»ù^wëi׳•£Ú¿—Ö¶úYMgòûPS¼y·dÒh ©qté:誚 „u"…̺-±ƶb ¥žwÞ„Y‡ßn¾}$^™Ý[Þ÷7ž³¨ëÊÂEÏñcÛ×ÊÉ}›q§N2¶Ýüªó³ù]’i-«HÔBR©^ݧµä}óÚïwN}þ—©ÃÇìxøûÙÎãóü?z¿_7äòwü'ûñô:>Ú=ÄìÍ#îÌPjJr)ËÍ™ÝÉÑEiÝ»dŽË¬ þ”ÑíXï,º<ìŠó­mvmº¶Ý‘• ][[£ú+±Ù(fìŽã“ ¢Ø3±¡øKŠ«¥ÎÊhÒŒšÝåwYc†Øì·¶ëkh{°AÙÖ#K@ÀhÒáMi;8£óÛ©ím¹µ/Õ­Q£zŒQ ÕðΤõ¶b½ÛÕí^hˆ4šöºy¶Œ´âö—¨u§TúÛ‰[×p¯&’ŠãÓíÛž×~zíÀLµ;2~l’C·ðð1- D­%ˆ­&“chج .ó&c¶"J"¾“fØÒ¢Ór£{¬Ù4š@Z4ÛBÝÉ'’iõz‡±«Õ¦–¢ˆgÕ,Œ»£°zž‡9Ù¬Õ™¸Tå$Cú^Ýx—+Ík4'ÃdøF*<ئóÖžÆÁ€%Ù^èÑ‹ííÆIÏ—Ž^™ÚßžMß,òùžV ðÜou–˜‘©Uá"ty|¿ îêü[·t÷þW›ìµö†-ù®dm¼rò PŽ8$̼wì*‰ëÕ¦ŸvDï. ÁºìB¡äÒ:”@Ò@é—BxHh<„ !U£JÕ "ßx¿|SBQ5÷”4‹£BTB•LÅDC÷Út&†‡H ¡ÒV³ ®GùºÜR(Nr€)6ÃE BЩPt­ÙWõvîé%Dw\Š”GB 4¥: Ð40D4д…øå•Tˆ&ª’ЉJŽ<iR#BѤt«¡¨‰r0(‚©Q{Ò¥* JLÓQ}»l(4…P3!Ö&V%JPûåT(¤X„‚Ä䀊{ð«Œ×E)Ð=½éˆØÕyU¹ÒÛÏ5QE6»¯P´C÷À{@êûã„éwNhØÉ¥*Äi £ÖÂ|§EçÅ’ø\‹]£†»K¹PG…çû䨽µÇÂYýi=¶ZöÄ¥¦“AºÐ[V«hœuïùwTöÑlé(iõ(8ƒËPá_­…' öµ~jÔ]ü?Ìõx%ë6É|¯oÁà÷¨4f­±NÖ.Ë#¸ìã¹3¸èäÍ nÝ»ˆèæí.:Ôm­ 4izëN¼íû"»Û_7K„"9fBF&.—lâC§®Š)ÒöÁÐD–Ôš=vzwFzt1&Æ¢ùmSØ7clÎ œš»jË.¾UåÓÛ¸¹:N¶·q½ïYÝ–shs¿u¾ÙígGvh\ûï|ùÓYî3×Ê×K³ÆÈlb""ouå¬Ç‰!òþ/·ŸGmïxî÷žØ—/pä½çxL*.rm§Zvgmö÷9I#¹"+O;#½,;^÷£Ë#ø¼ÕŸ{Ù§®ÆíH"gqYtwdI϶“Æ‚t?ˆÔ_ÄîîNãLu¨¢)h-Œhk‹ÛG6ÞX'bJ üW½ÄDTikÅWïBúN~~·?G{Ý?åæVw lÝÜTD"VaÅ·dEGfW%S N\ç™ÛÑÖé\]6¬ý¬¼‹Ž(»mAÙÖtIÔW·aõ˜¥ìmØ:‰zïŸÒÐÓTá‘¶~n0Õ%L—lÕE·mŒ.G+k|жë#›±Ù±¸¢ïÓõk¾SíÕfQthÙÑ•¶¤RýZÜ{i%S‘Vɉ&(í†õ×®Ý7aı^aÐóÛAjÐãÈÓ¸Ií¶ì’ˆáÜs•˜òÆ §K«wt®ÝÇFÙ4ü÷yN Ôùj;팰ŽìÇHSíybêúûfqÍ»³›”LÞ[Í$W6yÙ+lADEi¸3}—d Õµ™"`‘nì$|ö£Ìè£ÖÑͲé~¶¢¸ï•§Î¸A/Í¢Nw+jÎ;=½ë úl¤+Û}­ç“‚}¤³·›½±H¯Ùûw=ïRãŒikI!ZÞ‡…|þ>¶x8QUT ‡Êªˆ§ö…× RåÕe‡9Clí·@t¶Ë²åÙdÚËËf²dÚµ×V8Ó&̦Yfر­L¬ÂÈŒmekUÖ©™ÓZÈlÒIDºUR-i*¬Øš%²§™[L·jÜ9nÄvÙ‡ Vȳ.‰ªÕ®†YÜÕXXv¦ŒË"É£ZÒÇJ™4df­Z$*Ë&¥M*»*¹.`1f¥&j©e˜™eŒ¸ÎZYÈб5v´Ù«2p–ÊjvMl¶°&›FJnmÎ[n-ÖÈå•uD!”ÕÖf¤’³ZÄÅ\ÚŒ3³g,´M.×Mœê’uVZÖe¬§ue \µ$ÆŽÕ[F¶2'8‰›ŒÍ±©ŒëXÒè5uÕÒ²ºi8U4š©VlÕ•¦‰ÍÈm,Ú;Vµ\Ú›1@mZË6Mu4Õ®¦ÄÖ¬µ©hĬšh™"i*k*Šëb•Fk85EË74µ¬$¢È‘Ë‘™!9uU’¹®n vå·5 hÃF5¨ÌaÕY\Š’-51©KiÕj³Vq¦Ò¢UÓVL²¥ÌeœZŒ¦Í#ZU›0VK%jŒ²êÅʘÔp\ám…«]XÄZÇVŽ´ #SLšk 9•cŠÆ*.™¤Ãe˜§Vi«!4¥¥dë%¬—CLšÒÀëJÒ̶ ¸ÙÚÙKšS-nhiÙJÙ Õn±\R•M%2ÙŒ:ÊÙ:ºgVfd]]Y¬Úë*äæE[ #e•·0f:sœ$ÍÁ™Ó5[b²ÐhVA5¦–Xµ“1¶†šÑ£Z¦’ªP­¢œÕ\å¦©× £šÄæÕZÓL\Ò&§,´Õ¬ªÌ¢RÚíf]–c&¡ÛZ¢¹bÌ“œk™œåLš»Œ­®.ÖbSdVµ³jr(&' `Í“'Ls*;5vTmVš¨`Ž­l1Ú4"-u¢È*¶kU¡²çJM¡šÒÄ[VÆ*Ë.³aɤʙµ¤Zi¬Jµ©²®šê9ZY("MdÈkºÍ.¦³—!ksSZ«¬°d¦íÎm´“e–²RY™•k²%‹WL±V,ص1¤ÊÔɪ£VÚ´YIml­d®N,Ó•Ô¸e¶–‰Ötm*6ºÔäÖ¶R´-¡ k-bµÑÇdÉ…dic-»Fi1‹lµ´Ó0e-™¤b&HÕUr,¤²4ښ溙—Sµ‘‹ZµÖlMRÑu¢«UiÍR’s5«I¦u²FN¤ÍŒÕ¸ÚÍ4sNFiš ÌÀÕ©Øbº¶­d–Ó–QkêÒƒi®Öjæ]fiÄŽmm4WÌRl­cU£S4¦ª¹hµI9™#s!nM6ò“Z nÒmÍ#pÉe–±¥³™389ÙÆ˜cQs³$Š,ªÈ®EÔ®²bÚ•mvhÔækP²ÒÙ‰ZÄé6iØRu«)ʶ©Ø¹Z”å59,5¥™XäK#ñæ Ò%–ÏüÜøofæmVæµ&×y›kjy|×ö1­ýs]°h~²»}sÝ/0ê>·âóµáÖuyYQUç%G!ÖYÑwòmÊÙD‰¤ Ñ ÒV’ƒJš¤lÇ]'EÑÜç'Rˆ: „~¸ šÐÐTZZ»6QA+24­CH…'j!2˜) &ªJZQ )@ R¨ˆérS…RƒJP?žP JhQ¤ ($’ˆ•¤_nD×R ÒRÀ@4 €RA{²% P± %A'ÊDJ¤Í“¶D>»H4²D­Pk•ÆU¤CJ•@¡•¦ŽÀiPÖÆDßVAè@¦>«H Òf@¡k@ÒÒö2õì”äõryÇœ>vicÎSÈN—¢ø‘çžuê«“½êà윆=ñRà auùíwe³Séy°RÞ÷¹å¸ºÍµ–GyÛÉ» pÉöUeYÕ”Vw$QG_ºÒyØãœï¶ˆS5$æ2Q òtôþ¶ÇKlee|ï**}¬»æ‘Wµ¥íÎe"z 3£M³¢ZŠHÑ´jøözב» ¥Ûh)/;ÏÈ{Ìj‘éMAK¶ëEÍç4vŽ3r–Û^Õg‹˜Ñ£@И­¾=ÀGYvÔw½¯¡È丼­ ª†'lÄõñ“q—È¡)vÅPâ+g:Ñ¥óלϞné&¦%Û´Ngh’tïáVåï›ÀÚä¸(‰N髌Vz»¬ÄR´$UÓ¿6\=»¾QyK%¬6ímY$w9Ä^uàüÃçb’µ­|ívNÙf}[ÈÎõ]Ó“#Œnúȃ&M¼dؕyI¨¶51î‡zÚóh©}bÁ­“Òm$|qåÌWZ€ÎfýkÐï/ÇZyX‹kóÚOlŸoÇgx|æÑ ¿‡,è™dá¦9® °×Ãõº»{yÝxe¦¨@ƒ¤§8ä)$.JI á(9çN$ŽD ‘G(B#ˆ#¡H¸ ‚+‘ÒH uÇ:s¢tEI'\Tâ‡9qÄâ;’”TŽN.HåÂpâ;Ž£„Žè⋈⸧\G"C èéIp8ãŽ#€’qÇ')Åä@ttQuÅÈ(’EEÉuIAGTW‚ã¥qqDNè¹Î“ŽHá9ÂI¥ R©¢še†”iRWâ:)Ç)\uÇÇq)"ârâR]'8‹º8»¢N(ˆ(:¸âK®(¤JNEÎK»¢ê(Pœˆ¹)Êê.á…sè¸â—Rt\wAܤAÕ…)KLÃHDµGîpQÄuÑSƒ€!Ê]NàŽŽèàŠ:¢¢ã¨@îBrR:9:º:@’ç ¢Ž¤îŽ.\t\tJ!Š(*8QqE89(Ü]ÇŠ#’D]Ðr’]GîœIÝÀ#¸èî€áÈ¡Èâé8è¸S‰ "ê(IT.Žî(‡'uÅ'(ê)QT“‘BN!q\W‰r\]Ñ#¸âà\W(ê8ŠqÓˆ:œ§8’;Šäº'%).:â’w;’èB H’‚ Y¥¥¤™!ÊIÅÜPº:AÝs¢.DuÂ+ˆDœw:ã¢"QK‹Žáî".Ž…Ô(ê*’çrçH®QqÔt\tœ‚\¤QÔtQÜ\¨N$rWQ:å%GQÒêG$uèº;Žr)8º9:ç;…+ººNîrt]ÎPà»ÔqEr9ÒwEÉ$GÇQNâruQÊ£¨Ç\îè:8ÇB+§;©Ñ'D®£ wª.äAtAÁ rèèwqR@é!ÈÒÉKMÐQ R%"@´PS(BÒÅ¥Hœq×;¢DEQÐê•E.*‡@€£®(åÉ$»D98ªA¢Š]G8rHqÜrRt]ÉÐèqÕÀ+—  Šä.‹© ÜëŠDW*]‰QÀÝCp:Ž;¤îäHã‚E$àå RtÁ@8r8N!!tî'9#Ä‚D¥‚qR:SÈq$çTW9S‚ wÜ%8çINrSœAÅGrw¢r ÑÝtp'HtqRsIM-U0KJP´S 8è¸îââ98ãˆè⣣¸âK‚ŽŠ‹”î:#¹ÁH¸¡:‚î:8¸ “œã¢“Š#¤ ƒ¸Ž“¸C ¨¨éÎsˆs’㈤¡(êC¸(é.à:C àî8r:(# á êrîN;¢(“¢ ¸:ƒ¸äJ J8ˆ ä¨èŽˆä褣¸¸)(è¹88‚C¢ä“‚â9D¸JA(K£” ŽJŽ;£Ž¤¸@8*P„ç.’è¸".ŽëŠâä耨 ¸èŽâ’¹"¸£¥ H‚. ®©;Š:ƒ¸‰Êç(¢î ‚Ž8(”ާ8ޏ¡%3%- Å5 C5-$DÜq9ÁÔqÇ'9r]!9GEÄDãÑÉNÐWQAÐGwIÄQÝÉ%ÉÂ]‰q8QQpDqÑÑ'HqÀE%ÎEpE'$qÝD¢ÀqÝÉÑÐqqÑ'É—ÉœApDD]"EGRS‚ ÝG'”9qÉEÑÜåÇNQQRÝDu\w uÅÄwqtp]ÊwGQÇQ!×'IG'PRIÇQHDärœtw.!Ó‘Å]AÑI"QÔŽ‘r\$DIÄs„rIÂqÎ!G'PuÄtPQt\QDt”ŠQEÐqÔDtttEÑrtDqÐQIÝpIE(THIÉQÅÉÅ9ÅÜœ‰ÒEÈQps‡w ÄqEt…'qÅÐpBwGG'AÑtEÇwQÉR]G9ID'ED¥ s”R\AÝ9ÅÑ'Ã$‡]$t%QÁÔç!q‡$wDtWÜ‘ÜuEÇwE'ÅÜQGÄ’tpGIÑ@qt]QÀµBÐÒÑ5PÒÓ”´—EÂQÑEÒpDuÉuqÅ"‡w]qGGÝÉE@wQAÔqÁE#¥$wÜIEÇqÇQIIEHwÉQÇtQÄN„QÇPGqÇIÊT!ErÒGpÅpPGQÒQÔ#’HBwTTAÄut\¹qÜW$EqItI!I\Q×9 t‰ÉÁRBD•Ç'QIP‡KBÐÓ…%$ËCM.¹.(꓄订B‘¢¨‰+iáå|?0 ;£Xê(ºäâNÚëD‹÷u‰Â'9ÑQÔt^wEYÑ~ݸ) ¹ÉKZ;£¬ÙºÓ²Ë¬Ž¬èÎ*I|Æ®hèèè§CE;&d(RÞÕyÌ·W¶¼êÎË[.ç\»–ÀvÆd¶û)çVŽ.;Ž€¶í6Þb”#ŽvcŠöêÄè¼ ë+•CŒæŽ»²š¶í¦¨ö·‘–)Üsö´vÖŠÙ¥“`ÐPU-²é-uM.^V\W~Ÿjçµarsk(tmc±Š¦;fŠOŸ±Ñäè다§vOc¼ƒ§b;ÌŒD]öÇ›ËIçÞ»ÝóD&Ø4éÚÍZÉMóÚšzt>Oyy‡ê^{Vifkkóv Æö¼›wyç Û#§íÝéŸ4§ža§K¦>n8‹Õ1¨ªwlíf&µv¨4Ñ#BÅj;\­4|ùñ(Ç|“¨ï¿OÛêúRòtv0úšâ(££{¯8ǵ¤Ò2ïe!+µ ÷Y–rNOŒ]¾n:Qj¨Q‚ú6)t)¢‘(Õ hMPÌ‘0Ñoà´”¹† JR”¹#Ð H*QªÀ”* ¦  ¡@ýTéºF€¤B&´íœ_«˜¸ BiGB&•ZZ ŠT¤&h ‚($Q¤¡WIDß•!‰bˆJ+T¨P{°­ RURT4"HB‚IªìˆÒA7á]Z`ˆò%¡JRbB”S¥B„¥@¥‘)býxîHô‚bP t@'X’ƒUTÑ ˆVÎö¼‘È<%_';-(«C¤ JÑ@Ò­ôYCKÛD”7‰j¥b¥ǃm#K’cåÇf‰i<‡m ÑZzÝÆ3NÖ¿„ge'QÓÛ<çg5A±Åºpi¡ §§w¿ëcJPÈq¢‚ +;Ü-hõžnšMÓ¶ ²”šl¼ëÎò¶µÏ~›ÃÚĤqYq×ntAqíjòNˆë+Nà'.Oàv}l˜Úù{Ù/ïZâžµ6¾ÝŽê'³¢ì¾¯βy=• ÒRÍESäj¶ÃÆÍGNd‰¶›{Öç{ÚôÍ;ˆàÌàˆˆNíT”ë¹xØqŒ´¹:}¬öòôuíhç+íZn\Q4…!í:ÖtÄQD³£®--ΆÕw9PR maí%ÅÞ^föÀÞíµD—·¿ž†ƒÈÑé-ŒTüß¾Ï (%›ó_ží/ÔªêHÖ®ë&7QiP¥Dh(N¥6crm’f9´h‹lFõÇT>§KlºJt› ›œ¹=æ0»r¸;xt‡l:kZ Ggc{ìC›Ú÷†G–¥ýó.ópGyÙæöÃŽT—ï»øWÊŒŒŒÌƒooÅÝçííìjþ¨ŠsPùàõ†MP˜d4ÒÒeEáv“ZJÛ·qÂ4Ël6²t¡¶ÜÜ!–‹HÜm˜ÀÓi-­i9LæR&´ŠÛ›Ó¶[X7Xu’æ£ZÕ VØmu‡L¦KÖ™‰UwYj5´f£ÌZÚ¸eÉq5†„ÀšL6k!d[nÜfq"Ö(±g,–e†IŽLš¹¨ìíÓcåË]©\ë…ªÔ¦Ue«VG:tiΦ•›NÚDpËm I«išÄQm„KkVÁ±¹§U™;5]¥tÓ› ÙÌã¢Î.™u¦«–®Õ1”2k¸ÖYÒ±jêNͬ&©«±­3fLÕV*ضmLÙɹ­milk¤­]C1Ë–¥´N5ª#º¢r-SRƘ¦HË-1 ¹s¥²ÕmQKGZŠ.Fr4ËYXºÖsTÓŒ»UªŽX¸fµÉu«–µ£“”Lê„lŽ–°»­Wc…rfTIV2À®:ÖuQjÕ·Z–v [i¦†m:®ÌkV©-FZ9®['•š±ªÌ®–˜Š m(™S\´±Fµ™ˆ\c\ÌÕ£Kjj²sŽ™«,êœÄÓ«…’ºëVœÅØÖ¬ìÅ&ÛÇ·Mj,¶†]r\µY¨ÌåVZ´Tãkštºç.ªàZÓ ëQÌrëkÔÚºµ6´k3s–ݵ„¥5ªrt,i–¶·,ªç6if”ÍšYa†©™cafKTØÖ$95ÌÆÛ³láÜ“¦íbÉ6Ó¬FikdåŽÍÓ³£[c­m93”¢È—4¥YYm\5ªYŽUs)²V–™¢Z\[U3U©+šµÙ[4¬JÚÉÈÈÚq]]P¨#-Rl¥ikE«Q†«¬NkE˜!lM ™¦l·6d“’6qDæ¶k[-œ‚"]˜);L@°ffÓˆÖm.%–šÑ!™®b¶].KXeC«vÍjMbmÅNì&Æ”‹5Ú­6hmT,JÈS-JµŒ±,ÈIÉG*rµZeZ¨°‘¦V¥9£iTÁk:’§A²ˆÎ ¢Íe %¨•Ìæ®š´i•9ЫÚHkU4Q¬²±k5ª`s«WVYµ&MMiY¢ˆ£–e,£lå–¶BÁ‚˜v[VVa*©Åg*M5[IÕKYØdf¦M³e­6w`jN–±ËM*Y%&ÍYn56Õ™¢2Õ«0¨U«V3XIf©«®E£µhRtÖU34–æå6£M mËZÖ!m9²ÄfS&ŒÛ5-k”KXË4f¦ˆb×kS+ ±Ô,¨Ume-m%h©†–1¬rÙÉK‚³3"ËV´u˜jâV”ÄÓU]SZ‚[XnŽ2rÑ­&Í-k´%0æÝˆslÖ ¹bÙÖÓ+MVk®V´Q±6VRÕYÖ?.X¶ïý™>-7©>'i­læ—ZÂïÑ„‰²uR¯+©-¢-¢§ZE(´#@{ÊT£’’ún+Š’î#¢ºB‹…ÑßÓK‰ *kM"RŒÔ)ÕKTLÅ#J/l–1JªP))f”šŠ!ªH«ôJ:Qt”#§MÓ0 GéƒQ4P5GuÕœ…Ç îîPƒÛ¨(e˜ˆ{r'é€ æ´¤¢@¡J@4¢Ð öÊK0i t¨i¥B¨¢šT¤µ™~~x:B„N‘J4hBíÌQ©Z:DèøT‘4›ÉSÒ?¢éè]Ðè…h61´IÓ´,^^vNÆ;ƒÃ£<œç0ó®^B.Á^¤îÁUø÷N“TÄÛ° ÎU¾äíc«½oZóºÚ04?«>¸Á¶<Γ^qˆPçCø±YT]M7e«æ¤"ttPïÍi@è£OÂè”¶*¶sJLÒìa‰’#õ£QæÒÌLÌSäZó]ÖulÖuYoµYG mœM3¶¨;e즭ºæ"=ÞwS%â°´1dr|³øSìyéÙâARgË¢‚¨™ŽÚój%˜|Ïvˆ©,eµe¤<´Ó»-·Ëg¶›eÆYeiöÔ€‹‰O¶s¬±kv;ù«½{vEó§ñkÏŸ"s幘”RJnHÅ]¤ L<÷ºDðÉV¬§ðì¯/â~·wÀp¾U¾[£ŠÚÑI#- òß<÷›¤ÏŸÌûqmo"µ™:ýÕ…=«Kˆ¿[Þ· úµ¯lHîùœQeiåå~ã»÷Uø‹ŸkNWŸu;'"v £j›;H4ªâ¸‚Š3R¯ FjQ‚—lZU|LŠÓâIlº}o˜z\ýÞ†çCo›×î Ü‚¨ Š¢j"&(ŽK’t¤C¹ ’qÄ”AI: BC£ŽJŠ;”éHŽNà.+‘À„9IÂŽR\UPrt]qøDtwȸáq·N' ’"séÓ¹*NâpèNŽ;¨œQÑ'¨é$ÅÁ"N@èáÂ]ÂDq‹‘9Åp×” ÄT]ÅÜwDà“‰IÓ•ÒtRŽ¢£…NwIH£„èë©Ò¡!r.:¸.;£¨9Èp’’N亸ŠN„R9Nâ8»Ž.(¥wGt¤ˆ’:)t•twpî$DËJÐ…#@Ò­4Õ3‰¹;¨*âîs£¥q@N ä8¸ºä"”œRq@Ž@tQÎIÜ’ÅqTqQ×pw¡ „à\rGu¢ã¸”u@⎎î.8¢â㣩N‚–˜ˆ¢†‘¥ –)r:;¨+Ž”rG :”w !"rqçR\„‰'\r]ÇpドÈåtwQÓ»¤G(ë‹‹…Š;¸®':I"ˆ(£®(u¸;Š‘ÇG8rJ(9ÉÕ”NRCœ!S…GBRœÅEÈâ.9qÜu..#…ÅpIèâpDˆªr*ã¨QÑJ8î.áDÀ:9Ür9)"8Ž])Òq8Ž!w\HŽ“¥Üt]:Šå':㣮. å(.(º(à\®8èîâ)qS£º.¢(u9ÒèçœÉH¨…¤bBX¤¥¤JIib&’ã¢wtpŽ‘ÝÅr ŽrqÕ%ÊIœQÕN£…Åqî8®NâU"à© é.8’‘N¨£¨ãŽtèº+¸.¹ÜwQ¸ª]DHK—GqB]PI4DQ1 L‹ B‘$—QãŽ.‚GB):©$Ë‹ª9Ç\á8®8êˆwp#®$çtUC¸”\Wpà'uã ¡HuÑw ©Tr:¨åˆº;¹Ä븉Ä::î%Ü ¥ÝÒ‘ÎQT’*:¦)!(R¤ "–ˆ¦%f ¨)iˆŠ!h%•&ŽJ:))è¹pë…ÅÊ%$ŽG‰Üàƒˆœ':)΃‰ "Šp‘%p %΂Jr+ŠŽŠJ8R@ Hq98â:.$K£º#¸‰Îá#®N£¨âŽ“¹¸¸’Šˆç."¢@¤á*R8¹Â©È.N:(è :‹€îÎsŠCŽ£ŽNNâäœä¸¸"œ:Žâ飣£¢„œ(„â8¨ê8Hã¹!GJ $#¸ê8ŠŽŠ.8’8îè;£¸8¸䎎Kœ ãŽ8¤®:㊜Š!;ŠJ::’Žƒ¹(:8î(ê8Š)%;£¸¹(©89(è.#Š‚s¤¹Žàè;‰9ËŽè'N98¹'8è¢â:Žã¸ê(„‚â Ž ã®;®;Žˆ¢èî:;Ž8ƒˆ M4¤K-$M4$TçqÁ'rPQu\ÉÔqpq GÇA—!DG!Ç$AÂ'QN9AÁ%P’s”ÔáÁIAER!QIEÇ\tàt\¡IGQÉÑÑ$pGtä!GG'!D]åÇ$RDD\pœ ‘qQÅ]ÂuÇIQÉÜpEGrDU%I'Jw'ÄtÔ—AQtTIÝwAÄÑÄW9D\wÑtÎuÇqq\GÅÂAÇÄEPrqt§qÇE9ÅÇ'qDEGN'Qw‘Ó”rQÅ'SSŽq×quÊTèœ\Ñ’RDã‘QÉ$GQBWÔp\qÄát”UrwÇÈ]qAÜ$NuJB8”œqܧA\IÉÉE\]Ätå''DtqÑEGqw'tG\QGpr'TIÇIÄwI]8¡Iq$tAÐáGuIÝÉHqrEÜ)"IIÉDE\TT\IIE9ÝS§QÑqÇ%$rqqr tG)N‘GrTpNGt!p'$AÑÜR]N!×%:Ðt€WNT‚qÇEȇE$qQqHHqq@‡EPw$QpqwEÒsœqÁDw'ÉEDÑ@DNIÅE@ p‘DœtqD\\îqIEÈqÅAPq'GÝr%9Á܉ÐwÅÃ8$IQ܇Ü”q\Ü%AÅÑÀHŽ‚å%BAÇ!AuÐHprTE$t%$GE!p…QÜHq'$wuH”ÕA5AÌ÷·~ªÒ4 AÊÖîýVE'VÕ‰ÜvÖȸ³.Ì¿yØwI/k¬ê\'6«S³Nã‰mÆ/n¼°zÜGYÚ»(ó´ì´;:‹6ܦd–Ù6´FÚŸ2~>ÂÄhz{[I”•lÔASl»lvþ·Þ.õ¹õÝcˆÉŒ®1ÍX31ƒz,j=eÒyyh’mHzŠ]R”žE‰‡»=1ÛAL`ÒAØ7`4'I¤ü}—¶ Ö“v:4%s§AgoÞöïpŠ$ÛsíÖlޱ$Í=+Ä8؈i%MŒ`É81±Ó¼[ÍtJ ‰wurîÚÔ .;“I‰X„BáèÜãO´k¡Óå‰;nÚ¢v0§*OÚ³:Îì«{Þ¼.hÝtVDw8€¢ïÞËVV{ÖžÇKû£‹Û¯"³Ï9,¬ƒ’9Åm~÷¶kóÛ§±W™©"ÖöŠd•¦m1l0º¹W·³ìÜáâØÝAÕ2ê€Â(¸q–v‹k´©.Ȳ»”º‚!Ê-»ùëÒ-¬W@´©¥4¡¤ª‘¦¡wGuË…Ô]H_Îu¢‚Û1EÓ¥ ¤SH!J+’U#@ГS«{.š@! 4®”)Jƒ05%1%4 "(”:‘4)H,HT”-” @º¤F*e‚$¤¥(¤(w¤ 2r*JJ( TP‹jŒ˜(™_ _ËB>y¨ fZ¥›ëª@<‘6èTÒŽb4ý«Œ‰ i‚’+¾Ž¤iB“èƒHú½g k¥±íÇzÑpüƒ¶‹šv!Ù^UÅ®qŒluƒ'dêt` º©>Äq€ß^¸÷œÜäÚh,VvÁÇêרνvi=ó¯$3@[RCÛdõh|‡Ëuˆõo#%EÑ fSAKBitš]iÒQVÁ‹Ö31{ãx`ÒRô´çWeYerW½ïumŽWÍs‘Êq'' 9{ƒQÒfq¤ë.¿Úùß"{YÙÛhàâ“›jûd‘Ûa¶àât® ,ˆè‰ÓïØ:èzªj‡©t:‰~7ëñ|÷9Y‚%øµ·YWóWZ—EGryØYV™ÄUënè"އšž[‹:nÇ-¶ž{—jÌìÓ8" hH½òéÝši¢´ôFq;b‹zÛøž½yi÷KÉ0^·“ÓM%ÑEh£M!í“¥ÛtQ×Ù-­šÙ/׺ï-öã3ù6ˆ}ÑF©îÉl`‰îî¹$£…™˜Z­°#—)È1›d4jQfd$„½úöW—g>÷¼ë;LŒOØ”»¥Ç"xÌ€@Mõð8\Ôˆ°Â4*˜i½ýÓ——Û`ÞöŠº£õ*)¥¥ ((Šý‹²9IÖݳtJ8K“¨¢« ¬ë*r‚‚8\ïÓeåX/nÎíÐ:h{/v);8*(„íŠc£ˆòÛa meöÒ)G{ôÞ\¿fó½ÊÕ·ê]•GR_.O·žöÑÒN]¶’K{ÞOnì;Ùƒ2´PÛhb}JtPD#j[v;l„¶÷­ÉÊŤÂöC}#ÆÇÅvˆìÙݵ=¾^yfÚ°ì.kfyeÆYrÇJ]ú7ßo=ž÷yÒRµg»hhÒRíšöz¯]Ñù¬š’{YíyW™ïZk/lÓÒb)›/IyžØ-íßZÝó­)óWÍHä&yå]Zvu6‹:ËŽ2¶4µ/3(È ȡʅ ª[‡¸:=êÎ28‡pm™au{{ß§µ¸®;$Kû}ïÒߣ­,j$â•tNÏdöz{=½k«¿à>ÿÖŸü]ÉáB@ïbàks/src/0000755000176200001440000000000014673274557011502 5ustar liggesusersks/src/ks.c0000644000176200001440000006347613227237517012270 0ustar liggesusers#include #include #include #include #ifdef HAVE_CONFIG_H #include #endif #include #include #include #include /* Multivariate linear binning functions translated from the Fortran code of M. Wand & T.Duong in ks < 1.8.0 adapted from 1-d massdist.c in stats package */ /* Headers */ void massdist1d(double *x1, int *n, double *a1, double *b1, int *M1, double *weight, double *est); void massdist2d(double *x1, double *x2, int *n, double *a1, double *a2, double *b1, double *b2, int *M1, int *M2, double *weight, double *est); void massdist3d(double *x1, double *x2, double *x3, int *n, double *a1, double *a2, double *a3, double *b1, double *b2, double *b3, int *M1, int *M2, int *M3, double *weight, double *est); void massdist4d(double *x1, double *x2, double *x3, double *x4, int *n, double *a1, double *a2, double *a3, double *a4, double *b1, double *b2, double *b3, double *b4, int *M1, int *M2, int *M3, int *M4, double *weight, double *est); void interp1d(double *x1, int *n, double *a1, double *b1, int *M1, double *fun, double *est); void interp2d(double *x1, double *x2, int *n, double *a1, double *a2, double *b1, double *b2, int *M1, int *M2, double *fun, double *est); void interp3d(double *x1, double *x2, double *x3, int *n, double *a1, double *a2, double *a3, double *b1, double *b2, double *b3, int *M1, int *M2, int *M3, double *fun, double *est); /* Code */ void massdist1d(double *x1, int *n, double *a1, double *b1, int *M1, double *weight, double *est) { double fx1, wi, xdelta1, xpos1; int i, ix1, ixmax1, ixmin1, MM1; MM1 = M1[0]; ixmin1 = 0; ixmax1 = MM1 - 2; xdelta1 = (b1[0] - a1[0]) / (MM1 - 1); // set all est = 0 for (i=0; i < MM1; i++) est[i] = 0.0; // assign linear binning weights for (i=0; i < n[0]; i++) { if(R_FINITE(x1[i])) { xpos1 = (x1[i] - a1[0]) / xdelta1; ix1 = floor(xpos1); fx1 = xpos1 - ix1; wi = weight[i]; if(ixmin1 <= ix1 && ix1 <= ixmax1) { est[ix1] += wi*(1-fx1); est[ix1 + 1] += wi*fx1; } else if(ix1 == -1) { est[0] += wi*fx1; } else if(ix1 == ixmax1 + 1) { est[ix1] += wi*(1-fx1); } } } } void massdist2d(double *x1, double *x2, int *n, double *a1, double *a2, double *b1, double *b2, int *M1, int *M2, double *weight, double *est) { double fx1, fx2, wi, xdelta1, xdelta2, xpos1, xpos2; int i, ix1, ix2, ixmax1, ixmin1, ixmax2, ixmin2, MM1, MM2; MM1 = M1[0]; MM2 = M2[0]; ixmin1 = 0; ixmax1 = MM1 - 2; ixmin2 = 0; ixmax2 = MM2 - 2; xdelta1 = (b1[0] - a1[0]) / (MM1 - 1); xdelta2 = (b2[0] - a2[0]) / (MM2 - 1); // set all est = 0 for (i=0; i < MM1*MM2; i++) est[i] = 0.0; // assign linear binning weights for (i=0; i < n[0]; i++) { if(R_FINITE(x1[i]) && R_FINITE(x2[i])) { xpos1 = (x1[i] - a1[0]) / xdelta1; xpos2 = (x2[i] - a2[0]) / xdelta2; ix1 = floor(xpos1); ix2 = floor(xpos2); fx1 = xpos1 - ix1; fx2 = xpos2 - ix2; wi = weight[i]; if(ixmin1 <= ix1 && ixmin2 <= ix2 && ix2 <= ixmax2) { est[ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2); est[ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2); est[(ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2; est[(ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2) { est[ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2); est[(ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2; } else if (ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1) { est[ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2); est[ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2); } else if (ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1) { est[ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2); } } } } void massdist3d(double *x1, double *x2, double *x3, int *n, double *a1, double *a2, double *a3, double *b1, double *b2, double *b3, int *M1, int *M2, int *M3, double *weight, double *est) { double fx1, fx2, fx3, xdelta1, xdelta2, xdelta3, xpos1, xpos2, xpos3, wi; int i, ix1, ix2, ix3, ixmax1, ixmin1, ixmax2, ixmax3, ixmin2, ixmin3, MM1, MM2, MM3; MM1 = M1[0]; MM2 = M2[0]; MM3 = M3[0]; ixmin1 = 0; ixmax1 = MM1 - 2; ixmin2 = 0; ixmax2 = MM2 - 2; ixmin3 = 0; ixmax3 = MM3 - 2; xdelta1 = (b1[0] - a1[0]) / (MM1 - 1); xdelta2 = (b2[0] - a2[0]) / (MM2 - 1); xdelta3 = (b3[0] - a3[0]) / (MM3 - 1); // set all est = 0 for (i=0; i < MM1*MM2*MM3; i++) est[i] = 0.0; // assign linear binning weights for (i=0; i < n[0]; i++) { if(R_FINITE(x1[i]) && R_FINITE(x2[i]) && R_FINITE(x3[i])) { xpos1 = (x1[i] - a1[0]) / xdelta1; xpos2 = (x2[i] - a2[0]) / xdelta2; xpos3 = (x3[i] - a3[0]) / xdelta3; ix1 = floor(xpos1); ix2 = floor(xpos2); ix3 = floor(xpos3); fx1 = xpos1 - ix1; fx2 = xpos2 - ix2; fx3 = xpos3 - ix3; wi = weight[i]; if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ixmin3 <= ix3 && ix3 <= ixmax3) { est[ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3); est[ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3); est[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3); est[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*(1-fx3); est[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3; est[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*fx3; est[(ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*fx3; est[(ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*fx3; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ixmin3 <= ix3 && ix3 <= ixmax3) { est[ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3); est[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3); est[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3; est[(ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*fx3; } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1 && ixmin3 <= ix3 && ix3 <= ixmax3) { est[ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3); est[ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3); est[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3; est[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*fx3; } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ix3 == ixmax3 + 1) { est[ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3); est[ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3); est[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3); est[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*(1-fx3); } else if(ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1 && ixmin3 <= ix3 && ix3 <= ixmax3) { est[ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3); est[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ix3 == ixmax3 + 1) { est[ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3); est[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3); } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1 && ix3 == ixmax3 + 1) { est[ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3); est[ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3); } else if(ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1 && ix3 == ixmax3 + 1) { est[ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3); } } } } void massdist4d(double *x1, double *x2, double *x3, double *x4, int *n, double *a1, double *a2, double *a3, double *a4, double *b1, double *b2, double *b3, double *b4, int *M1, int *M2, int *M3, int *M4, double *weight, double *est) { double fx1, fx2, fx3, fx4, xdelta1, xdelta2, xdelta3, xdelta4, xpos1, xpos2, xpos3, xpos4, wi; int i, ix1, ix2, ix3, ix4, ixmax1, ixmin1, ixmax2, ixmax3, ixmax4, ixmin2, ixmin3, ixmin4, MM1, MM2, MM3, MM4; MM1 = M1[0]; MM2 = M2[0]; MM3 = M3[0]; MM4 = M4[0]; ixmin1 = 0; ixmax1 = MM1 - 2; ixmin2 = 0; ixmax2 = MM2 - 2; ixmin3 = 0; ixmax3 = MM3 - 2; ixmin4 = 0; ixmax4 = MM4 - 2; xdelta1 = (b1[0] - a1[0]) / (MM1 - 1); xdelta2 = (b2[0] - a2[0]) / (MM2 - 1); xdelta3 = (b3[0] - a3[0]) / (MM3 - 1); xdelta4 = (b4[0] - a4[0]) / (MM4 - 1); // set all est = 0 for (i=0; i < MM1*MM2*MM3*MM4; i++) est[i] = 0.0; // assign linear binning weights for (i=0; i < n[0]; i++) { if(R_FINITE(x1[i]) && R_FINITE(x2[i]) && R_FINITE(x3[i]) && R_FINITE(x4[i])) { xpos1 = (x1[i] - a1[0]) / xdelta1; xpos2 = (x2[i] - a2[0]) / xdelta2; xpos3 = (x3[i] - a3[0]) / xdelta3; xpos4 = (x4[i] - a4[0]) / xdelta4; ix1 = floor(xpos1); ix2 = floor(xpos2); ix3 = floor(xpos3); ix4 = floor(xpos4); fx1 = xpos1 - ix1; fx2 = xpos2 - ix2; fx3 = xpos3 - ix3; fx4 = xpos4 - ix4; wi = weight[i]; if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ixmin3 <= ix3 && ix3 <= ixmax3 && ixmin4 <= ix4 && ix4 <= ixmax4) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*fx3*(1-fx4); est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*fx3*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*fx3*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*fx3*fx4; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ixmin3 <= ix3 && ix3 <= ixmax3 && ixmin4 <= ix4 && ix4 <= ixmax4) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*fx3*(1-fx4); est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*fx3*fx4; } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1 && ixmin3 <= ix3 && ix3 <= ixmax3 && ixmin4 <= ix4 && ix4 <= ixmax4) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*fx3*(1-fx4); est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*fx3*fx4; } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ix3 == ixmax3 + 1 && ixmin4 <= ix4 && ix4 <= ixmax4) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*(1-fx3)*(1-fx4); est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*(1-fx3)*fx4; } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ixmin3 <= ix3 && ix3 <= ixmax3 && ix4 == ixmax4 + 1) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*fx3*(1-fx4); } else if(ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1 && ixmin3 <= ix3 && ix3 <= ixmax3 && ixmin4 <= ix4 && ix4 <= ixmax4) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*(1-fx4); est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*fx4; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ix3 == ixmax3 + 1 && ixmin4 <= ix4 && ix4 <= ixmax4) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*(1-fx4); est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*fx4; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ixmin3 <= ix3 && ix3 <= ixmax3 && ix4 == ixmax4 + 1) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*fx3*(1-fx4); } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1 && ix3 == ixmax3 + 1 && ixmin4 <= ix4 && ix4 <= ixmax4) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*(1-fx4); est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*fx4; est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*fx4; } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1 && ixmin3 <= ix3 && ix3 <= ixmax3 && ix4 == ixmax4 + 1) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*fx3*(1-fx4); } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ix3 == ixmax3 + 1 && ix4 == ixmax4 + 1) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1] += wi*fx1*fx2*(1-fx3)*(1-fx4); } else if(ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1 && ix3 == ixmax3 + 1 && ixmin4 <= ix4 && ix4 <= ixmax4) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[(ix4+1)*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*fx4; } else if(ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1 && ixmin3 <= ix3 && ix3 <= ixmax3 && ix4 == ixmax4 + 1) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + (ix3+1)*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*fx3*(1-fx4); } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ix3 == ixmax3 + 1 && ix4 == ixmax4 + 1) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + (ix2+1)*MM1 + ix1] += wi*(1-fx1)*fx2*(1-fx3)*(1-fx4); } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1 && ix3 == ixmax3 + 1 && ix4 == ixmax4 + 1) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1 + 1] += wi*fx1*(1-fx2)*(1-fx3)*(1-fx4); } else if(ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1 && ix3 == ixmax3 + 1 && ix4 == ixmax4 + 1) { est[ix4*MM1*MM2*MM3 + ix3*MM1*MM2 + ix2*MM1 + ix1] += wi*(1-fx1)*(1-fx2)*(1-fx3)*(1-fx4); } } } } void interp1d(double *x1, int *n, double *a1, double *b1, int *M1, double *fun, double *est) { double fx1, xdelta1, xpos1; int i, ix1, ixmax1, ixmin1, MM1; MM1 = M1[0]; ixmin1 = 0; ixmax1 = MM1 - 2; xdelta1 = (b1[0] - a1[0]) / (MM1 - 1); // set all est = 0 for (i=0; i < n[0]; i++) est[i] = 0.0; // assign linear binning weights for (i=0; i < n[0]; i++) { if(R_FINITE(x1[i])) { xpos1 = (x1[i] - a1[0]) / xdelta1; ix1 = floor(xpos1); fx1 = xpos1 - ix1; if(ixmin1 <= ix1 && ix1 <= ixmax1) { est[i] = fun[ix1]*(1-fx1) + fun[ix1 + 1]*fx1; } else if(ix1 <= -1) { est[i] = fun[0]; } else if(ix1 >= ixmax1 + 1) { est[i] = fun[ixmax1 + 1]; } } } } void interp2d(double *x1, double *x2, int *n, double *a1, double *a2, double *b1, double *b2, int *M1, int *M2, double *fun, double *est) { double fx1, fx2, xdelta1, xdelta2, xpos1, xpos2; int i, ix1, ix2, ixmax1, ixmin1, ixmax2, ixmin2, MM1, MM2; MM1 = M1[0]; MM2 = M2[0]; ixmin1 = 0; ixmax1 = MM1 - 2; ixmin2 = 0; ixmax2 = MM2 - 2; xdelta1 = (b1[0] - a1[0]) / (MM1 - 1); xdelta2 = (b2[0] - a2[0]) / (MM2 - 1); // set all est = 0 for (i=0; i < n[0]; i++) est[i] = 0.0; // assign linear binning weights for (i=0; i < n[0]; i++) { if(R_FINITE(x1[i]) && R_FINITE(x2[i])) { xpos1 = (x1[i] - a1[0]) / xdelta1; xpos2 = (x2[i] - a2[0]) / xdelta2; ix1 = floor(xpos1); ix2 = floor(xpos2); fx1 = xpos1 - ix1; fx2 = xpos2 - ix2; if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2) { est[i] = fun[ix2*MM1 + ix1]*(1-fx1)*(1-fx2) \ + fun[ix2*MM1 + ix1 + 1]*fx1*(1-fx2) \ + fun[(ix2+1)*MM1 + ix1]*(1-fx1)*fx2 \ + fun[(ix2+1)*MM1 + ix1 + 1]*fx1*fx2; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2) { est[i] = fun[ix2*MM1 + ix1]*(1-fx1)*(1-fx2) \ + fun[(ix2+1)*MM1 + ix1]*(1-fx1)*fx2; } else if (ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1) { est[i] = fun[ix2*MM1 + ix1]*(1-fx1)*(1-fx2) \ + fun[ix2*MM1 + ix1 + 1]*fx1*(1-fx2); } else if (ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1) { est[i] = fun[ix2*MM1 + ix1]*(1-fx1)*(1-fx2); } } } } void interp3d(double *x1, double *x2, double *x3, int *n, double *a1, double *a2, double *a3, double *b1, double *b2, double *b3, int *M1, int *M2, int *M3, double *fun, double *est) { double fx1, fx2, fx3, xdelta1, xdelta2, xdelta3, xpos1, xpos2, xpos3; int i, ix1, ix2, ix3, ixmax1, ixmin1, ixmax2, ixmax3, ixmin2, ixmin3, MM1, MM2, MM3; MM1 = M1[0]; MM2 = M2[0]; MM3 = M3[0]; ixmin1 = 0; ixmax1 = MM1 - 2; ixmin2 = 0; ixmax2 = MM2 - 2; ixmin3 = 0; ixmax3 = MM3 - 2; xdelta1 = (b1[0] - a1[0]) / (MM1 - 1); xdelta2 = (b2[0] - a2[0]) / (MM2 - 1); xdelta3 = (b3[0] - a3[0]) / (MM3 - 1); // set all est = 0 for (i=0; i < n[0]; i++) est[i] = 0.0; // assign linear binning weights for (i=0; i < n[0]; i++) { if(R_FINITE(x1[i]) && R_FINITE(x2[i]) && R_FINITE(x3[i])) { xpos1 = (x1[i] - a1[0]) / xdelta1; xpos2 = (x2[i] - a2[0]) / xdelta2; xpos3 = (x3[i] - a3[0]) / xdelta3; ix1 = floor(xpos1); ix2 = floor(xpos2); ix3 = floor(xpos3); fx1 = xpos1 - ix1; fx2 = xpos2 - ix2; fx3 = xpos3 - ix3; if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ixmin3 <= ix3 && ix3 <= ixmax3) { est[i] = fun[ix3*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*(1-fx3) \ + fun[ix3*MM1*MM2 + ix2*MM1 + ix1 + 1]*fx1*(1-fx2)*(1-fx3) \ + fun[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1]*(1-fx1)*fx2*(1-fx3) \ + fun[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1]*fx1*fx2*(1-fx3) \ + fun[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*fx3 \ + fun[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1]*fx1*(1-fx2)*fx3 \ + fun[(ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1]*(1-fx1)*fx2*fx3 \ + fun[(ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1]*fx1*fx2*fx3; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ixmin3 <= ix3 && ix3 <= ixmax3) { est[i] = fun[ix3*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*(1-fx3) \ + fun[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1]*(1-fx1)*fx2*(1-fx3) \ + fun[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*fx3 \ + fun[(ix3+1)*MM1*MM2 + (ix2+1)*MM1 + ix1]*(1-fx1)*fx2*fx3; } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1 && ixmin3 <= ix3 && ix3 <= ixmax3) { est[i] = fun[ix3*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*(1-fx3) \ + fun[ix3*MM1*MM2 + ix2*MM1 + ix1 + 1]*fx1*(1-fx2)*(1-fx3) \ + fun[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*fx3 \ + fun[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1 + 1]*fx1*(1-fx2)*fx3; } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ix3 == ixmax3 + 1) { est[i] = fun[ix3*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*(1-fx3) \ + fun[ix3*MM1*MM2 + ix2*MM1 + ix1 + 1]*fx1*(1-fx2)*(1-fx3) \ + fun[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1]*(1-fx1)*fx2*(1-fx3) \ + fun[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1 + 1]*fx1*fx2*(1-fx3); } else if(ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1 && ixmin3 <= ix3 && ix3 <= ixmax3) { est[i] = fun[ix3*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*(1-fx3) \ + fun[(ix3+1)*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*fx3; } else if(ix1 == ixmax1 + 1 && ixmin2 <= ix2 && ix2 <= ixmax2 && ix3 == ixmax3 + 1) { est[i] = fun[ix3*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*(1-fx3) \ + fun[ix3*MM1*MM2 + (ix2+1)*MM1 + ix1]*(1-fx1)*fx2*(1-fx3); } else if(ixmin1 <= ix1 && ix1 <= ixmax1 && ix2 == ixmax2 + 1 && ix3 == ixmax3 + 1) { est[i] = fun[ix3*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*(1-fx3) \ + fun[ix3*MM1*MM2 + ix2*MM1 + ix1 + 1]*fx1*(1-fx2)*(1-fx3); } else if(ix1 == ixmax1 + 1 && ix2 == ixmax2 + 1 && ix3 == ixmax3 + 1) { est[i] = fun[ix3*MM1*MM2 + ix2*MM1 + ix1]*(1-fx1)*(1-fx2)*(1-fx3); } } } } /* Registration of native routines added 17/03/2017 */ static R_NativePrimitiveArgType md1_t[] = { REALSXP, INTSXP, REALSXP, REALSXP, INTSXP, REALSXP, REALSXP }; static R_NativePrimitiveArgType md2_t[] = { REALSXP, REALSXP, INTSXP, REALSXP, REALSXP, REALSXP, REALSXP, INTSXP, INTSXP, REALSXP, REALSXP }; static R_NativePrimitiveArgType md3_t[] = { REALSXP, REALSXP, REALSXP, INTSXP, REALSXP, REALSXP, REALSXP, REALSXP, REALSXP, REALSXP, INTSXP, INTSXP, INTSXP, REALSXP, REALSXP }; static R_NativePrimitiveArgType md4_t[] = { REALSXP, REALSXP, REALSXP, REALSXP, INTSXP, REALSXP, REALSXP, REALSXP, REALSXP, REALSXP, REALSXP, REALSXP, REALSXP, INTSXP, INTSXP, INTSXP, INTSXP, REALSXP, REALSXP }; const static R_CMethodDef cMethods[] = { {"massdist1d", (DL_FUNC) &massdist1d, 7, md1_t}, {"massdist2d", (DL_FUNC) &massdist2d, 11, md2_t}, {"massdist3d", (DL_FUNC) &massdist3d, 15, md3_t}, {"massdist4d", (DL_FUNC) &massdist4d, 19, md4_t}, {"interp1d", (DL_FUNC) &interp1d, 7, md1_t}, {"interp2d", (DL_FUNC) &interp2d, 11, md2_t}, {"interp3d", (DL_FUNC) &interp3d, 15, md3_t}, {NULL, NULL, 0} }; void attribute_visible R_init_ks(DllInfo *info) { R_registerRoutines(info, cMethods, NULL, NULL, NULL); R_useDynamicSymbols(info, FALSE); R_forceSymbols(info, TRUE); } ks/NAMESPACE0000644000176200001440000000540214547755145012130 0ustar liggesusersimportFrom(grDevices, chull, col2rgb, grey, hcl.colors, rgb) importFrom(graphics, .filled.contour, box, contour, hist, image, legend, lines, pairs, par, persp, points, polygon, rect, rug) importFrom(KernSmooth, dpik) importFrom(Matrix, Diagonal, Matrix, norm) importFrom(mclust, mclustBIC) importFrom(mvtnorm, dmvnorm, pmvnorm, rmvnorm, dmvt, pmvt, rmvt) importFrom(pracma, meshgrid, quiver) importFrom(stats, IQR, cov, cutree, dbeta, dist, dnorm, ecdf, fft, hclust, loess, median, nlm, optim, optimise, optimize, pchisq, pnorm, predict, qnorm, quantile, rbeta, rnorm, sd, smooth.spline, var) importFrom(utils, head, setTxtProgressBar, tail, txtProgressBar) useDynLib(ks, .registration=TRUE, .fixes="C_") export(amise.mixt, ise.mixt, mise.mixt) export(binning) export(compare, compare.kda.cv, compare.kda.diag.cv) export(contourLevels, contourSizes, contourProbs) export(dkde, pkde, qkde, rkde) export(dnorm.mixt, rnorm.mixt) export(dmvnorm.mixt, rmvnorm.mixt) export(dmvt.mixt, rmvt.mixt) export(Hamise.mixt, Hamise.mixt.diag, hamise.mixt) export(Hmise.mixt, Hmise.mixt.diag, hmise.mixt) export(Hnm, Hnm.diag, hnm) export(Hns, Hns.diag, hns, Hns.kcde, hns.kcde) export(Hbcv, Hbcv.diag) export(Hlscv, Hlscv.diag, hlscv, Hucv, Hucv.diag, hucv) export(Hkda, Hkda.diag, hkda) export(Hpi, Hpi.diag, hpi) export(Hpi.kcde, Hpi.diag.kcde, hpi.kcde) export(Hpi.kfe, Hpi.diag.kfe, hpi.kfe) export(histde) export(Hscv, Hscv.diag, hscv) export(kcde) export(kcopula, kcopula.de) export(kcurv) export(kda) export(kdcde, dckde, reg.ucv) export(kde) export(kde.balloon) export(kde.boundary) export(kde.local.test) export(kde.sp) export(kde.test) export(kde.truncate, kdde.truncate) export(kdde) export(kdr, kdr.segment) export(kfe) export(kfs) export(kms, kms.part) export(kroc) export(ksupp, dwsupp) export(Lpdiff) export(matrix.sqrt) export(mvnorm.mixt.mode, mvnorm.mixt.part) export(plotmixt) export(pre.scale, pre.sphere) export(rowKpow, getRow) export(Sdr, Sdrv, mur, nur, nurs, Qr) export(symconv.1d, symconv.nd) export(vec, vech, invvec, invvech) S3method(contourLevels, histde) S3method(contourLevels, kcde) S3method(contourLevels, kcopula.de) S3method(contourLevels, kda) S3method(contourLevels, kde) S3method(contourLevels, kdde) S3method(plot, histde) S3method(plot, kcde) S3method(plot, kcopula) S3method(plot, kcopula.de) S3method(plot, kda) S3method(plot, kde) S3method(plot, kde.loctest) S3method(plot, kde.part) S3method(plot, kdde) S3method(plot, kdr) S3method(plot, kfs) S3method(plot, kms) S3method(plot, kroc) S3method(plot, ksupp) S3method(predict, histde) S3method(predict, kda) S3method(predict, kde) S3method(predict, kcde) S3method(predict, kcopula) S3method(predict, kcopula.de) S3method(predict, kdde) S3method(predict, kfs) S3method(predict, kroc) S3method(summary, kms) S3method(summary, kroc) ks/CHANGELOG0000644000176200001440000005214414673270214012115 0ustar liggesusersChange log file for ks 1.14.3 -Made small changes to kdde.binned.nd() and binning() to better report small bandwidth errors. -Removed repeated contour levels in .filled.contour() call in plotkde.2d(). -Fixed bug for empty contour regions in plot() method for kde.loctest objects. 1.14.2 -Moved plot3D from Imports to Suggests, so ks can run in environments which can't install its visual functionalities. 1.14.1 -Added weighted estimate for 1D kde(unit.interval=TRUE). -Fixed bug with decreasing cumulative probs in qkde(). -Changed name plot.histde.*d() to plothistde.*d(), gamma.r*() to gamma_r*() to avoid S3 method clashes. 1.14.0 -Systematised code indentation (4 spaces). -Further systematised default bandwidth calculation. -Added new flag grid.clip which excludes data outside of xmin, xmax in kde() etc. for binned estimation. -Fixed bug in kde(unit.interval=TRUE) for 1D data. 1.13.5 -Fixed inconsistency in code and help guide for kms(). -Added kde(density=TRUE) option to force non-negative KDE values. -Fixed bug in 1D kcde(positive=TRUE). 1.13.4 -Removed calls to zero.flag=FALSE. -Fixed in bug in kdr(y=y, pre=FALSE). 1.13.3 -Fixed matrix inverse calculation in Hns.diag(). -Added density ridge segmentation in kdr() and kdr.segment(). -Fixed bug in kdde.grid.3d() which computed the same values for all derivatives. -Added consistent alpha transparency and default colour scale behaviour for plots. -Added 3D kernel support estimate in ksupp() and in plot.ksupp(). -Added pre-sphered optimisation for "unconstr" and "dunconstr" pilots in Hpi() and Hscv() to resolve numerical optimisation difficulties. -Added REPORT=1 to print out progress reports for each iteration for calls to optim(). -Changed to use "BFGS" rather than "Nelder-Mead" in optim() in psifun2.unconstr(). -Fixed bug in Gunconstr.scv() so that Hscv(pilot="unconstr") and Hscv(pilot="dunconstr", deriv.order=0) give same bandwidth. -Added note on help page that zero.flag in predict.kde() is deprecated. 1.13.2 -Fixed bug in ploting group colours in plot.kms(, display="plot3D"). -Reinstated display="rgl" option for 3D KDEs in plot.kms(). -Fixed bug with transposed coordinates in pracma::quiver in plot.kdde(, display="quiver") -Fixed bugs in kde.truncate(), kdde.truncate() for normalisation and for contour levels calculation. -Fixed bugs in contourSizes(), contourProbs() which didn't allow for different grid sizes. -Added colour transparency parameter alpha for colour function in plot methods for histde, kcde, kda, kdde, kde, kde.part, kms objects. 1.13.1 -Added contourProbs to approximate probabilities of level set of KDE. -Changed from OceanView::quiver2D() to pracma::quiver() for quiver plot in plot(,display="quiver"). -Added pracma to Imports. Removed OceanView from Suggests. -Fixed bug for certain missing axis labels in plot.kdde(). -Reduced number of considered mixture models in Hmn(), Hnm.diag(). 1.13.0 -Changed default colour schemes to be based on hcl.colors(). -Fixed bug in hlscv() in calling optimise(). -Fixed bugs in plotkda.1d() and plotkda.2d(). -Fixed bug in calculation of Abramsom bandwidth in kde.sp(). -Fixed bug in kde.balloon.2d() in re-normalising density estimate. -Added pre-scaling option in kde.sp(), kdr(), Hpi.kcde(). -Fixed bug in contour labels in plotkd(d)e.2d(,display="filled.contour"). -Generated datasets quakesf, platesf as sf versions of quake, plate. -Added variable names in output to kde.local.test(). -Added S3 contourLevels method for kcde objects. -Added S3 predict method for kfs objects. -Fixed bug in names field in output in kdr(). -Changed in hlscv() the default minimiser function to stats::bw.ucv(), and default interval search limits. 1.12.0 -Fixed bug in rkde() to take into account weighted KDE. -Fixed bug in contour level calcuations in plot.kde(). -Added names field to kda object in kda(). -Fixed bug in plot.kdde() which always used which.deriv.ind=1, and never passed the col parameter to plot.kde(). -Fixed bug in plot.kcde(display="filled.contour") which didn't apply col or col.fun, and replaced filled.contour() by .filled.contour(). -Fixed bug to reduce default binning grid limits in kde.loc.test(). -Fixed bugs in kda() and plot.kda.*d() to handle correctly factor grouping variables. -Fixed bug in kfs() and plot.kfs() for 1D Wald statistic. -Fixed name labels in plate dataset. -Added plot3D functionality plot(, display="plot3D") since RGL 3D plots face an uncertain future but remain for the moment acessible via plot(, display="rgl"). 1.11.7 -Fixed small bug in help files to comply with R 3.6.3 CMD check. 1.11.6 -Changed dkde() to be an alias for predict.kde(). -Updated default.bgridsize() for d=4 to rep(15,4). -Fixed bug in kdde.binned.nd() to force when keval is a vector to be 1-row matrix. 1.11.5 -Fixed bugs in kde.1d() passing eval.points parameter and which didn't allow unequal class proportions. 1.11.4 -Fixed error report in kde() incorrectly asserting to set "binned=TRUE" for exact estimation. -Fixed bug in line colour in plot.kcde(). -Fixed bug in kde.grid.nd() not passing verbose argument. 1.11.3 -Fixed bug to kda() which set prior.prob values to default sample proportion even when set explicitly differently. -Fixed bug in partition plot for 2D plot.kda(). 1.11.2 -Fixed error in predict.kda() incorrectly assigning class labels. -Replaced dnorm.deriv() by version from J.E.C to compute arbitrary derivatives. -Added verbose option to kde.points(). 1.11.1 -Fixed missing passing of h, H parameters in kde.positive.1d() and kde.positive.2d(). -Fixed bug in kdecopula.de(). -Fixed bug in partition colours in 2D plot.kda(). -Changed default optim.fun="nlm" to optim.fun="optim" everywhere. -Moved rgl, misc3d, OceanView from Imports to Suggests, so ks can run in environments which can't install these visual functionalities. -Fixed inconsistencies in gridsize and bgridsize default values in ks.defaults(). -Added varying.grid.interp.*d() in predict method for non-uniform grids. -Fixed bug in 1D predict.kde() with non-uniform grid (i.e. output from kde(, positive=TRUE)). -Extended limits of grid plotting for in plot.histde(). -Fixed bug in creating factor levels for estimated labels in kda(). -Added air, cardio, hsct, plate, quake, tempb datasets. 1.11.0 -Added multivariate version of rkde(). -Added histogram estimators histde(). -Simplified calculation of default values in kde(), etc. -Changed default flag binned=FALSE to binned=default.bflag. -Fixed estimated group labels calculation in kda(). -Fixed interpolation for d>3 in grid.interp() and predict.kde(). -Added histogram density estimate histde(). -Added kernel density estimate for bounded data kde.boundary(). -Added truncated kernel density estimate kde.truncate(). -Added kernel support estimate ksupp(). -Added kernel partition plot plot.kde.part(). -Added variable kernel density estimates vkde(). -Added quiver plot to plot.kdde() for deriv.order=1. -Added kernel summary curvature kcurv() for deriv.order=2. -Added World Bank data data(worldbank). -Added any dim KDE in kde.grid.nd(). -Changed display="filled.contour" for 2D plots to not give adjacent colour scale bar. -Fixed bug in 1d kde.test() p-value to return scalar. -Fixed bug in interp1d in ks.C which had assigned values outside of estimation array. 1.10.7 -Changed mvtnorm from `Depends' to `Imports' in DESCRIPTION. -Implemented per-block calculation for large sample sizes in kdde(, binned=TRUE). -Fixed small bug in col.fun for plot.kcde(, display="filled.contour2") -Swapped order of computation of CDFs in kroc to prevent possible segmentation faults. -Fixed bug in default bandwidth for kcde(). -Fixed bug in default estimation grid limits for kda.nd(). -Corrected formulas for scalar pilots in gdscalar() to match those in book. -Added feature significance function kfs(). 1.10.6 -Registered native C routines in src/ks.c. -Added kernel mean shift kms(). -Fixed bug in predict.kdde() for vector x for d>2. 1.10.5 -Fixed bug in invisible return values for plot.kda(), plot.kcde(), plot.kdde(), plot.kde(). -Added more detail about eval.points in kde.Rd. -Changed default approx.cont=TRUE in plot methods. -Changed default to compute.cont=TRUE in estimation functions and corresponding plot methods. -Fixed bug in contour plot colours in plot.kdde() for 2D. -Fixed bug in calculation of scalar pilot in gdscalar(). -Fixed bug in calculation of unconstrained SCV pilot in Gunconstr.scv(). 1.10.4 -Fixed bug in plot.kde(,display="slice",abs.cont=) not plotting contours correctly. -Fixed bug in predict.kde(object, ..., x) by adding zero.flag which controls behaviour when x is outside interpolation grid object$eval.points: TRUE = 0, FALSE = object$estimate corr to nearest grid point. 1.10.3 -Fixed bugs in col specification in plot.kde(,disp="persp") and disp="image". -Approx computation in contourLevels(,approx=TRUE) is now default. -Boundary adjustments in binning() moved to C functions linbin*d.ks(). -Grid interpolation functions renamed from find.nearest.gridpts() to grid.interp() and coded in C to increase speed. -Reduced time complexity of loess smoother in kcopula(). 1.10.2 -Improved speed for kde.points(), kdde.points(). -Improved speed for compare(). -Fixed missing xlab, ylab in plot.kde() for 2D KDE. 1.10.1 -Fixed small bug in find.nearest.gridpts when treating edge points. -Modified pre.scale, pre.sphere to use sweep(). -Fixed lower edge interpolation in find.nearest.gridpts.1d(). -Fixed incorrect derivative order in kfe calculation in gdscalar(,binned=TRUE). 1.10.0 -Implemented binned estimation via symconv.1d(), symconv.nd() with unconstrained b/w for kde(), kdde(), kfe(), dmvnorm.deriv.sum(), Hlscv() Hscv(). -Added aliases Hucv(), Hcv.diag(), hucv() for Hlscv(), Hlscv.diag(), h.lscv(). -Added predict method for kda objects. -Fixed inconsistency in plot.kde1d(,col=). -Added 3d exact estimation and 3d plot for kdde(). -Adjusted calls to symconv.1d(), symconv.nd() in drvkde for feature library. -Included calls to RGL plots in help file examples in \donttest{}. -Moved dfltCounts(), drvkde() to feature library. 1.9.5 -Changed DESCRIPTION to comply with CRAN checks (e.g. imports etc.) -Fixed inconsistencies in graphical parameters in plot() functions (e.g. ptcol, cont.col, ...) -Added raster graphics if available for display="image" and "filled.contour2" plots. 1.9.4 -Removed explicit prior calculation of permutation derivative indices in dmvnorm.deriv.sum(). -Fixed small bug in contourLevels.kdde() to make explicit call to predict.kde(). 1.9.3 -Removed copula.grid() and hence dependence on copula package. -Fixed bug in displaying 2D contour level labels for plot.kde() and sorting in contourSizes(). -Added amise=TRUE option to hpi.kcde(). -Modified kroc() to be line with updated mathematical definition. 1.9.2 -Changed binning=FALSE to binned=binned for Hpi(,pilot="dscalar"), Hpi.diag(,pilot="dscalar"). -Fixed bug in binning behaviour in gdscalar(). 1.9.1 -Fixed typos in help files -Added new classes "kcopula" and "kcopula.de" for output from kcopula and kcopula.de to distinguish them from "kcde" and "kde" objects. -Exported matrix.sqrt(). -Added "exp" option for make.grid.ks(). 1.9.0 -Added efficient recursive versions for dmvnorm.deriv(), Sdr(), Sdrv(), nur(), nurs(), mur(), Qr() from Chacon & Duong (2014) Statist Comput. -Fixed bug in Hscv(,binned=TRUE), Hscv.diag(,binned=TRUE) which was still computing unbinned estimators. -Fixed bug in 1-d KDA plot. -Added sensitivity, specificity as output to compare(). -Made small changes to default selectors to be more consistent across selectors. -Fixed bug in point colour in rug plot for plotkda.1d(). -Added Hpi.diag.kcde(). -Added Lpdiff() (Lp distance for two functions) and copula.grid (true copula evaluated on a grid). -Fixed small bug in plotmixt(,draw=FALSE) to actually not draw plots. -Added predict method for kde objects to replace kde.approx(). -Added option to compute 1-d KDE supported on [0,1] kde(,unit.interval=TRUE) which calls kde.unit.interval(). -Changed default axes limits when plotkde.3d(, drawpoints=FALSE) from data range to mean of KDE evaluation range. -Fixed bug in default pilot selector for d>3 data in kda(). -Changed ad hoc argument matching to match.arg(). -Fixed bug in last line of lscv.mat(). -Added binned estimation to Hbcv(), Hbcv.diag(). -Added default binning flag function default.flag(). 1.8.13 -Added boundary density estimator kde.boundary() for compactly supported data. -Added kernel density of copula nd copula density, i.e. kcopula() and kcopula.de(). -Fixed small bug in plot.kcde(disp="slice", abs.cont=!missing), and Hpi.kcde(). -Changed smoothing spline in kroc() to be evaluated on equally spaced grid. -Added thinning option for persp plots plot.kde(thin=), plot.kcde(thin=). 1.8.12 -Added kernel estimators for CDF kcde() and ROC curves kroc(). -Added default plug-in bandwidths to kda(), kcde(), kde(), kdde(), kde.local.test(), kroc(), kde.test(). -Added warning when using non-diagonal bandwidths for binned estimation. -Added plot and contourLevel methods for kdde objects. -Modified plotmixt() to include derivatives. -Added 1-d plug-in selectors hpi(,deriv.order>0). -Merged kda() and kda.kde() into single kda() function. -Changed "kda.kde" object class name to "kda". 1.8.11 -Added progress bars to compare.kda.cv(), compare.kda.diag.cv(). -Corrected critical df from d to 1 in kde.local.test(). 1.8.10 -Fixed small bug in call to contourLevels(approx=) inside kde(). 1.8.9 -Added kde.local.test() for local 2-sample test. -Replaced foreign call to .C("massdist", package="stats") requested by B. Ripley by call to .C("massdist1d", package="ks"). -Changed rug plot in plot.kda.kde() to rug-like plot, similar that in plot.kde.loctest(). 1.8.8 -Changed function header of Hpi.kfe() to be more consistent with Hpi(). -Added option Hpi.kfe(, pilot="dscalar") to ensure scale invariance in p-values. This becomes the default over the previous pilot="unconstr". -Added 1-d option in kde.test() and its required bandwidth hpi.kfe(). -Modified binned=TRUE option in kde.test() so that it is applied only to bandwidth selection, and not the test statistic and its p-value. -Removed default truncation in Hlscv(), Hlscv.diag() for deriv.order=0. 1.8.7 -Further improved speed of kfe(,Sdr.flag=FALSE) by computing unique partial derivatives. -Removed unused function dkde.weights() to compute optimal deconvolution weights, and hence dependence on the kernlab library. -Changed output from kfe(binned=TRUE) to be vector not 1-row matrix. 1.8.6 -Implemented calculation of Lebesgue measure of level sets of contours, contourSizes(). -Implemented probability contour plot for 1-d KDE plot, i.e. analogue to existing 2-d, 3-d contour plot(,disp="slice"). -Added recursive computations kfe(,Sdr.flag=FALSE) which don't compute symmetriser matrices explicitly. These are then called in Hpi(,Sdr.flag=TRUE) and Hscv(,Sdr.flag=FALSE). -Changed pilot="dunconstr" to direct computation rather than indirect eta form. This means that Hpi(,pilot="dunconstr", deriv.order=0) and Hpi(,pilot="unconstr") now give the same result. -Remove pilot="dsamse" option as this was more computation than pilot="dscalar" but with little difference in the result. 1.8.5 -Fully unconstrained pilot selectors pilot="dunconstr" for Hscv(), Hpi() for density derivative estimation. -Unconstrained Hlscv() selector for density derivative estimation. 1.8.4 -Reinstated psi.ns code (more efficient than eta.kfe.y) and SAMSE pilot estimators Hpi(, pilot="samse"). -Edited help manual. 1.8.3 -Added computationally efficient density derivative b/w selectors Hpi(deriv.order=), Hscv(deriv.order=), and their diagonal counterparts Hpi.diag(), Hlscv.diag(). -Added computationally efficient kernel functional estimators in eta.kfe.y() used in kde.test(). -New pilot selectors for density derivatives. -Added abs.cont capability to plot(, disp="filled.contour"). -Removed explicit expressions in psins() for d>2, replaced by eta.kfe() evaluations. -Removed psins() and Theta6() evaluations in gsamse and gamse.scv. -Removed kfold arguments. 1.8.2 -Fixed bug in kde.points.sum() to avoid allocating large matrices for unbalanced sample sizes for x and eval.points. -Fixed bug in dmvnorm.deriv.sum() which had excluded last partition class for double.loop=FALSE. -Added binned options to kde.test(). -Fixed bug for exact estimation in kfe(). -Added plotting colours as function of z-value in plot.kde(, disp="persp"). -Added decoupled calculation for Hlscv(). -Added optim.fun option to select optimiser function in Hpi, Hpi.diag, Hlscv, Hlscv.diag, Hscv, Hscv.diag(). 1.8.1 -Modified p-value calculation for large -ve Z-statistics. -Fixed bug for binned estimation for unconstrained bandwidths for kde(). 1.8.0 -Added density derivative selectors Hpi(,deriv.order=r), Hlscv(,deriv.order=r) for r>0 from J.E. Chacon. -Changed vech(H) terms to vec(H) in AMISE estimators. -Changed default binning gridsize for 3-d data from rep(51,3) to rep(31,3). -Added verbose option to b/w selectors (in double sum) for tracking progress. -Changed LSCV, SCV selectors optimisation from Nelder-Mead to BFGS. -Changed Fortran linear binning code to C (and fixed bugs in Fortran code). -Added modification to linear binning for boundary points. -Removed explicit derivatives in BCV selector optimisation. 1.7.4 -Fixed small bug in partitioning method for kde.points.sum(). 1.7.3 -Changed partitioning method for dmvnorm.deriv.sum() and kde.points.sum(). 1.7.2 -Changed p-value calculation for kde.test(). 1.7.1 -Reinstated single partial derivative of mv normal for scalar variance matrix dmvnorm.deriv.scalar.sum() for use in AMSE pilot plug-in selectors. -More efficient form of kdde(). 1.7.0 -Added KDE-based 2-sample test kde.test(). -Modified output of plotmixt(). -Added "double.loop" option to kfe() for large samples - increases running time, reduces memory. -Modified dmvnorm.deriv.sum() to improve memory memory management for large samples. -Cleaned up code for plug-in bandwidth selectors and kernel functional estimators. -Cleaned up help files. -Disabled kfold b/w selectors. 1.6.13 -Added flag to automatically compute probability contour levels in kde(). 1.6.11 -Added own version of filled contours as option disp="filled.contour2" and different colours for disp="slice" contours. 1.6.10 -Added k-fold b/w selectors. 1.6.9 -Added approximate option in contourLevels(). -Added kdde() kernel density derivative estimators. 1.6.8 -Added 1-d LSCV selector hlscv(). 1.6.7 -Corrected ISE for normal mixtures, from J.E. Chacon. 1.6.6 -Added MISE, AMISE, ISE functions for normal mixtures derivatives. -Changed internal double sum calculations from J.E. Chacon. 1.6.x -1-d binned KDE fix from M.P. Wand. -Streamlined code sharing with feature package (all binning code now contained only in ks). -Reorganised and renamed internal bandwidth selection functions, mostly double sums of normal densities . 1.5.11 -Fixed small bugs in drvkde, vech, Hpi(, pilot="unconstr") 1.5.10 -Added drvkde (kernel density derivative estimator 1-d) from feature using M.P. Wand's code. 1.5.x -Added normal mixture (A)MISE-optimal selectors: hamise.mixt, hmise.mixt, Hamise.mixt, Hmise.mixt. -Added distribution functions for 1-d KDEs: dkde, pdke, qkde, rkde. -Added plug-in selectors for 1-d data (exactly the code for dpik from KernSmooth). For KDE, this is hpi, for KDA, this is hkda(, bw="plugin"). -Made changes to specifying line colour (col rather than lcol) in plot.kde, plot.kda.kde and partition class colour (partcol) in plot.kda.kde. -Added plot3d() capabilities from rgl to 3-d plot - removing own axes drawing functions. -New functions to compute pilot functional estimators hat{psi}_r(g). These are exact, and are more efficient than binned estimators for small samples (~100), and are available in d > 4. 1.4.x -Vignette illustrating 2-d KDE added -Binned estimation implemented for KDE with diagonal selectors and pilot functional estimation with diagonal selectors. -Filled contour plots added as disp=filled option in plot.kde(). -compare.kda.cv() and compare.cv() modified to improve speed. -Hscv.diag() and Hbcv.diag() added for completeness. 1.3.5 -Fixed small bug in compare.kda.cv() and compare.kda.diag.cv(). 1.3.4 -RGL-type plots added for 3-d data. Specification of 3-d contour levels now same order as 2-d contours. 1.3.x -Multivariate (for 3 to 6 dimensions inclusive) bandwidth selectors added for Hpi(), Hpi.diag(), Hlscv(), Hlscv.diag() and Hscv(). NB: because Hbcv() and Hbcv.diag() performed poorly for 2-d, these weren't implemented in higher dimensions. 1.2.x -Package checked by CRAN testers and accepted on the CRAN website. To pass all the necessary checks involved some internal programming changes but has not affected the user interface. -The child mortality data set unicef is used in the examples. 1.1.x -S3 type objects have been introduced. The output from kde() are `kde' objects. The output from kda.kde() and pda.pde() are `dade' objects. Corresponding plot functions are called automatically by invoking `plot'. -Kernel discriminant analysers are now available. Parametric (linear and quadratic) discriminant analysers are accessed using `pda'. -adapt library is no longer required. This was formerly used on the functions for integrated squared error computations ise.mixt() and iset.mixt(). ks/inst/0000755000176200001440000000000014673274557011670 5ustar liggesusersks/inst/doc/0000755000176200001440000000000014673274557012435 5ustar liggesusersks/inst/doc/kde.R0000644000176200001440000000531114673274557013323 0ustar liggesusers### R code from vignette source 'kde.Rnw' ################################################### ### code chunk number 1: kde.Rnw:99-107 ################################################### library(ks) set.seed(8192) samp <- 200 mus <- rbind(c(-2,2), c(0,0), c(2,-2)) Sigmas <- rbind(diag(2), matrix(c(0.8, -0.72, -0.72, 0.8), nrow=2), diag(2)) cwt <- 3/11 props <- c((1-cwt)/2, cwt, (1-cwt)/2) x <- rmvnorm.mixt(n=samp, mus=mus, Sigmas=Sigmas, props=props) ################################################### ### code chunk number 2: kde.Rnw:113-114 ################################################### plotmixt(mus=mus, Sigmas=Sigmas, props=props, xlim=c(-4,4), ylim=c(-4,4)) ################################################### ### code chunk number 3: kde.Rnw:116-117 ################################################### plot(x, xlim=c(-4,4), ylim=c(-4,4), xlab="x", ylab="y") ################################################### ### code chunk number 4: kde.Rnw:126-128 ################################################### Hpi1 <- Hpi(x=x) Hpi2 <- Hpi.diag(x=x) ################################################### ### code chunk number 5: kde.Rnw:132-134 ################################################### fhat.pi1 <- kde(x=x, H=Hpi1) fhat.pi2 <- kde(x=x, H=Hpi2) ################################################### ### code chunk number 6: kde.Rnw:142-144 (eval = FALSE) ################################################### ## plot(fhat.pi1) ## plot(fhat.pi2) ################################################### ### code chunk number 7: kde.Rnw:156-157 ################################################### plot(fhat.pi1, main="Plug-in", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) ################################################### ### code chunk number 8: kde.Rnw:159-160 ################################################### plot(fhat.pi2, main="Plug-in diagonal", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) ################################################### ### code chunk number 9: kde.Rnw:171-173 ################################################### Hscv1 <- Hscv(x=x) Hscv2 <- Hscv.diag(x=x) ################################################### ### code chunk number 10: kde.Rnw:177-179 ################################################### fhat.cv1 <- kde(x=x, H=Hscv1) fhat.cv2 <- kde(x=x, H=Hscv2) ################################################### ### code chunk number 11: kde.Rnw:181-182 ################################################### plot(fhat.cv1, main="SCV", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) ################################################### ### code chunk number 12: kde.Rnw:184-185 ################################################### plot(fhat.cv2, main="SCV diagonal", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) ks/inst/doc/kde.pdf0000644000176200001440000117526714673274557013716 0ustar liggesusers%PDF-1.5 %ÐÔÅØ 5 0 obj << /Length 3504 /Filter /FlateDecode >> stream xÚÝÛnãÆõ}¿Bo¥P‹;7Î 7uí"M€íH$Ei‰¶™I‡¤ÖÞ|}Ï™3¼z$Ù[£ú°¦æÂ3ç~î×WoÞ¾—銛XðD­®nV\¤±Ðfe¸ŽU"WW»ÕÑÇöÝz£¸¾Ë›*߯7’Ëh—WmÑ­…Ž>ÓDÞvE™uE]Ñø¦nèÇuñiˆ(kЬËýËY—­ºúöí{ÅVœÇi’w¼4±bfµ‘ PJéø«µÅ·¬0Ñuu‹ï®6€·2Éj#T,SM›Ãm:úßwy¹†®×ø›7ôº`bÍ#E‡K»â,NYÊñp€gc)å hY’À‘f6Ðlˆf¶ ™EE‹Oe4¼‡£mTßöYC3]MS ç>í!ÛmÈyn®»Ëý±EÛ5Åõ¡?‡Gõ_>Æ ™Á~÷‡¢¬«úG&¥ãºdÜSó#KOS O~Q!¢ü1+ï÷ùr:«èY¯ñOkÇËOEþ˜º£Óèû»Üïq ?ì;”¸%+éEb%Œ{VZÇJ<}ÊJ;VÂs[ÃÆ]Þä»)9Êë€]á´ždãþ³Ÿ¬èIŒdÌÁꚬ¨˜[ D¡Ë=œ‡¢»£_»"»­«lO£ë¬Ú=;Zd ÚÛ¼E^1 +¾u¬³#Š%éTÅÜ­*ö©¡SÑo¸Ï¶ˆßÇu’DÙmÞ†`‚¶Š8arΘ¶ Áæ*ÖpÔ…¯TJ'd ȤøõW0¨dUx"cÕGÚm õÃÕýh<ʺ¾°IÒØp;Eçûµ•z<Òiªc®ŒÜjNK ®!{Ú¶¹Ç‹–+ÐÏ zàÞ”>ÇÿQB ῎]ðÿcä?xU6àá®pÐA©”çÖWæDXKs£âÈIªA•AJ¹ŸEo°Ù5`|M''š‹Ã^siäŒ|t¯8ó$×JˆÎHš›,Ÿ™,Pd\OiÂ-[§XðHàQ[ÓàÚ -'_vh5Âôà ¸.¦xº1¸Æm‹²¨à(x­ó˜ö¹-Z﯎أîÒ^ a¡mú]SGºTŒDëL!'D)ì|Ÿo»ºñC"¶ó M`PÀ±Þì xµEz¼ëYòñ(J*Í\(Wwèú”bBn2tÈ+œè(‰ßRWèñWFkx¡)ò=±)òÆèâEÜrh оKN ËãvFý™ÔŸ”OÍ”ÏgІNÑÐ9„œ‚Jpe‰Ñsfg ÏÉž0åCeÂDôÙ„"ƒ¬„Ò ·ÐÄÉ%½¤0w¡qY7¹‡–[÷Nqa”m·õ˜çx®ªÇ&# × u†©&Ãx¿«ËPlk]8&>ê¥vÙ‘çã?|&%V&N #Nó˜[´ßXƒÆ¸mÜCb3ˆ«É®¯Bç©Ø*ûÂãıãäü8p5&z‡MÓ?‚¨@®'”X⒜ƥ )#D‚d†Ë¤Á°"»»i@,Î>3oˆµ ÐÆ™‚È=DŽ›Ð™:NxÒï¸ðÖç½ÇÌ·õç~ 'lö.·÷7Þ£ìrH÷e5Ýëˆ#7ue€TAÊlGºÿ¢f#b 2‡‡5vN‘9JZPX&2{ 1ñ¨Ï½Ü“£*øMDÛY2Á1ï·ÑePV"ö°Ù«Àµè‚1ø¦zˆŒ4öÙÁëñ°:Z Èq°³éi‡MÙÄ’ƒW‹…‘'T;fü¿P›qŸp™ÌeÐîÁÏh­ñÁ¬ïw~Ÿž@Âb0AS2œŸ`½9»N)‚\(‚æs)j®F)² úyÖ ð™(B˜ËXÚ™ÿsŠJDj¨8” Š}iúp…IÐCCÆÀÌI•µtð+l>eÜ\^fšV?QÈ„Ög>y==óÀa¯Î\^;î!JÆ ¢—Lœ …¤/¡Ú*ûP’§¦Î†}C±±ˆSž" Ï9 :ÖàC§âô4%)Õ «àG|¸v4–„X•.HIÀ“ªÆ©dÄM3 Éb“z ¶<úfô l‘»¥b8ò» XÌÔ³‚ª<Tm:‹ˆÁr׈a“«%úÖþ˜¶mp<-Y¹k»P6ž0à‹ÒÓ¦ªŸÏe™cÑIÃû¦¾Î® ,ÇVZÊÕ%LÜ@‘‹ùþE( 7Ë„`Ùì`6™’†`)ÝJç-baŒ8=Ò5‘_íßöÄàFO÷M À+˜eÍì;€”ÇÔ=Ûøt ¨…„²@…ÆàâIeB´ uùÒ>7F??Y£D KÄËP8FŽùáÏ!Q‚ÁË…—)Á$±û9«9ÿò¥©P.ÕèzžB˜m8QàLÝË,<9*‰o„§.yåúõ)·§)ÿrŸ4smàO Ô8WÎ%1‘ÿ©‹­×7!AXÈ"_ºÙ¥0c®é\€«êΟÓ¶Öç{Z&ÑÚšVÊ5UÈ3K'©¤»?˜y‚ãz _¥|²K“L™ Ÿðp`,’1ŸVo’‡K¼1ÉšdöT&ð|RO&3-Êï'|™;—YpäH(( Ódrì›?]½ùå ’ TäxÈT9KÙ–o~ø‰­v°øíŠÅ ò ·µ a¿úðæoo¾Æ´yÎ܃²sÂ` õ@¯e´˜'õS»$¹2,»Ì°’W2=ˆ&rØDi@b£¶ƒø”5;UuSº–4üîîšläp{W°ùe…ˆþ\ÑZ߀l²¶» UÄ×¥øê 9™ò<Ò§~‹“‘~£ÀâŒ@‹K 3—‹N”7ä= ‡[‚ ¼¼)Ýuv4Ýœ'Wù¨77°Œ´V„b2 8g¬€‚\ƒóÃ^Iz®§’RO%}n—yÊ™w€BY·§«.KןEŒ³&+‘¤¢uMRO:ÞâÓ_GÁÑ’”4»¦i&Ô_<ÒïÌ5JgÓýOCW›Å®o uwY­1‰âçAÂkò’ºïØ¥rÁÎßÐôm>Ê1CôSÄU9É/&ˆ¹•¢ æR Z 5,u y<äbÚO» µ¤ vF5vb‹Ä; ¡ÂU”I ÚµqïîÂ ÈÆ'NJô2ÕP‘ò1ÁÝó0¯TxD˜×fØcç:ÁŽ”Š<¨=†œ>Šœ˜»qñ´ù!PéF.%G><•ÎG  ©°G¡Nv;Ÿ+@ý,bÿC=G€gy.ÕaN€òœ‚Ì:X"œ_a°Ç)p3ÅY÷çåÚ{JA´kB„d)ºÑšì]0ÜIf\ûí<1Ó‹@©l´ÏºÎ}î¢Ì“¸@;²ýžîýýã̓Ÿp7’ =3Þ~@˜¾ÝCi® ðçÍg:ªn 홹b·?ÌõH³ð‚WkPtçMرïý}ajSÙ÷#ðÖÒ# NqFÐyº%y~ÞÑûõzqå²­7]Q —*ÙcÞ¿o/ö`k}h<À¬¹Í»þ2füHöú[žBÜý>´yïÎýÌ¿v‡1½ö7ãûýoæÀ°+bEìv¼âŸÜ÷ÌAöIN–ÎÉCß‘Z¸P"lVB­.úO¾T If Èr¤ï$N%ÉC~’%/ Lè˜ËÞù‡‡€ ‹I±²¹¿„Ì,ÅÈ–ÎNPónâöø"UÐLOýÞXD,»'ÊÎë_H:]Ñ3ÓÆLC!Üå2}±ø*sÐ÷‡ûï'‘*Cðš>€²eìtkNOÐQS"‘Ÿ°J‰¡> stream xÚ•W[³›6~?¿Â/™â£ ‚NÓ‡¦Í$™¾´ñ¤ifŠAǦÁÆœœüûîj%Œm’“ÚH«eµ—o/ðE¾È932]è4e¹ÈåþîãSF(å&KwÄýkžðôÕž/~mïþ€8‰a†‡þ—,q-+yΔJñDÁ_ÖwO_ÈlÁ–'9_¬ïœ –èl¡%g*‹uµx½¨·§Î.c)yÄ\ÆJ‰h½ÌdTt[; ]DÿT§ýRdÑf³Œáa›fÉ£謲‡¾ðô ¾ÅHÂ߉N~·÷<9ñ•홆öÔÑeǦ&ÜÖÛc8¿ò¦,†Á^°¿_¿sZ/€ÌRÎÉŒ¿ðrKïz‹ŒO_è|j¾4L p˜ãy¬‰çÒE"aüç™î[ùéê÷CWÔ[}NÛ¸>нml97z¢‡jNž–s3Ñ‚Uu±SEe}£ ²·‡¢¡kœpëµ"=#‘FMEÆ’s¦y.ðLIôÏËXsƒºpZýŒÌó“sRâO3(’önÅ߃.)¼ÿn%–1ÞSÔìì(o_€¸CEûºŸóœH%Ë\•s }¨ìœÚŠ3vE?ƒð]].…‰vtwÙYP$äÝÜu™`¹äÜv™úeSô=Éo¡êiý ‰5ŸTÔ#H&JªKlÝïŠÝ$hࡵò€{†ÙpvôKYâ{d‰3fo¡0Éý±lªÜ•M·À`ÎÃò38ëñ¬ëz,p{;ìi¬Šdc›—­„ylý!BAcôáDaαìB€ìfxVu‡jÃÇ¢+šÆw3:¹¹lã¶«êø‡ŃíWs¶8hŠk„¾’.gqhZ›ÆÒŠÊ È:.Ê_ü®,O68wûÉÅ®ñ’0ýèåÑ•CF©Âë|í¦©?ëɆ+Ú®¶”ÔÎ3À\ú}çÆ>E4Àˆ^bsvRœe† ÃªåÖ®@Á³ ‡éá°öq«àÀ‡ƒì@ ø™”Ÿå_MÆHÚPÿ'<ÑMA¸.ÛÃÛÌ–š)šûÉ@‚ø…šßœIÈ#8=HCòõÄ+BÝÌóHzóü--0Åß\/°<ïZœ pùÖ}!5uU`¨®"}üp»ùáG&0„§ãðó²/?Ív;(\Y`r…߉¼çÆa°ëAêP€–fü¦CÍ}“@[UfªÎW¿It⾃=+vi¢W˜·ü\o$ÔJ‰ÓÏ=̰o]·6˜ÖzPœò ö—aÖF:vŠÓO×î‰2Ê> /ExtGState << >>/ColorSpace << /sRGB 30 0 R >>>> /Length 7486 /Filter /FlateDecode >> stream xœ¥\KÏeGRÜ¿â.aá;õ®:ÛÒH -±@³@Èê ¼þ=‘‘Uõ±ÅÂÝ>窱|DFÝüúý+¿¾ýûÇßÛ¿ýîã7?üÃßüöõÇ>Ò;¥ôºÿüáúøÍ_;ÿ»}åòÖÿÃq)óÝ ¬÷,¯ï¾¾þâ¿þòõÝ÷õo×^|sÝÇþù »Ez—Ä‹ÿ;.¶w{õçÚkÖwk¯Ú PžçÝËë?þåõ¯?éÍ÷­½ùZÃnù¤bΖ_ßþîïìÿÎþúÏúÃ+½þù#¿~oÿ}ÿ‘ùä¿ýx:_÷œ¯¯O{¯np¾Ÿüú8ƒÏ{ÁlOÏøë@{µqN¶;VÅ­žùÎ<êzÖ»vÁ‡bÿ—І¯¦„kçcо–GÇ»LƒíÝ»Ãec‘®1hO(í¯Çaõë]x²ýmGí{3O¶vƒùýt‡Åne3àïü¼{P®ï…“³½Ïc÷Êý=qÆÛ×>x½û¸~_nxP­IßcØÆ¸9˜¼Zó{fâùNÀö¾‹xáµÆû„íýj{ç*\â⸼;îo&RˆëûiÄï‡Åõã]ù~þaµjj³½x«Ä« ›½?–ìÓ¼íÁæ# ·vÍ×ôe^iFÀÓíaÈ^ ã}¶=L{p%æâÍ+cbŽ}¦}T8P`[¸nŸÓÏo~ÿ°×ÕÝ>²ìÏü\¹í}-x¥Z¸à‡.89ÎpŒgý=…öc¸ðþæ1~{½šOÄxÎ~¼;Žõcž öU²ûÛl¾,ÕË_<+ü‰ûÃñ§ù¯9?*æ¿`æ¿èΟ~üY1ÿ•ÊñwÀ6qp‡é <·,æÏl"·3-ðÈ—¯-æÏ`Ïæ¦Ò¶ÛSÓsË‘;PëñóÀ³ž0Pp£~¢°M<‚H‹ã5ŸÜ®<Ò‰PÄý°À߈ë ÀýatLãÆ fs]_Ü¿›ÿ-+!</ć29MŽ»Eú9|ýó×¶0ËlfǶÞiÅ£t±eD{~p~F<Îxp™É×_ÉôÅ–ióó+üo±eܧp[ÄͱÇèÑÝ_3 |ßhóÿæÏʰi_Ä ˜ÑýçÇpªÄ\8ÅÜN^«8¦½™!c|ÍMÕ*lþ8ùñúîÀ¯ ÜxÃ>ÿÅÞÏÆÃÜâðëÇÏpâûXÆãfÌøìöh »ãù nÎp¥ãvû1Ç€ïoÂpÁD–¶4Þæhl"‹¹qÆËRù"¥ 3p‡!sû ÄÅÂB÷çM8öÒø™À þªXVçö@dï ÿÎëÍ1Ë_ãDdMðϼÞnD¼ÜÿáAf˜%â!^D8Ó^ñ¢ö=ð·~ýtÜõ}ö¡æðYůWÍ· ì׆…þ²ôì8˾l -P`X™_”îãê×78L ó“Ò™çaÚè1Ѱ›Öǯg öõ×910ƒÄëͰlâa&>#ÃßÀŒ ¯7Ä¿(ñ¾ƒñfWýúæç'§Xö˜m÷ëoaæÌ°pÌp° |½Åõ”§ü-DH,#÷§“ÁeÅçON4–¡Ï—a¬'`Ú‡-ü…û,;àNûÍEó7}|Ü / pñû1À¸³øH¼t¿ÅD˜ñ£X¼´‰†Ûb~džÞÜpçýã50ã#Öƒ¹ñÊç/æG3 [`öø‚øØ‰ýýÍÃþ-Œ¸=< ôÀîŸÍ±çDLÿ@û5ÌïC ¨Õñ$f<åù‰˜/N\ˆíqc(ØÛ¾•—ß?;î´'ÃÉÏô'öØÅçÛDçæ˜Ï3Cý$¦µ†aX“˜þöaeìÄö`˜ã‰Döe8ùõ“å}6ljˆÙÂ:ó!$.ˆ9_Hll`ýy¶pm¾™&ðý°Ð× iDÂͱ‡9›¤÷3G“€»Ç&†™˜þå§Í?Ò˜äçsà˜æð~4ˆë£šÿYx^FZ<೟OÿŒ4ë lö‘ݬ w&âL˦°96¤mÌjgàÍKöYÍßdTáÝóƒÚixH«_ßq"ÒDæ7À÷+¯kg~‘-~3ß®Ž}<;ýc¶e·üý<i+ý° DÆ2åõæŸp[ÆÌWë ¿dZ\„1^ž6ØÒjÿþAÃGÚÝ—°½X1¿†m঻}™?Ãõ£{¼Æxæñµ&”Ì'má¢L¨qÜ ;÷Çã+0ì Ÿíçާ¹i·O`»_W¾²q›ºÎÏoݯ=/=¯-÷çñ>mÆ÷Ñÿ Œêq<áxõ|‘ßkóƒú¨ÇxØxµäùBŒ—ÅKúg`s,ÙÌ>÷oW½‰ù0…2ÑíGó…eõh>±mÙ¤ù^¸_òü1ì¡*Ÿ€½`þ-Þ1Ÿ„=Á~ŠêŸ°/Ä;?>ßréz_Ù+ê…&{Æû”êùfØ{ÉO±°Þ »?ÂÀÛqsSc\ë¸ÇzfÚzÖpºÖ«ÓŽÎï¿b½çæù üÁt¬õíþ"ÇûÚÄ`¼Íí–Û¿ ¾ÍËYZ›–üÕÂóU‡?‹ú#ü]Šùµ‰±…ž‘fÏË_Fü Š´} ìë£Ð߃†™åøkà¶Ž?mãëÛý=iœqâ°û# }|=d&ŠÀëŠ7¤…Ú‰GÀn_¯€™?*ž]˜ñîœïñðÜ­]ÏóøÉ÷é'¾ò}뉿¢¥v|Æ÷{þíñ¸G<ÇúŽñR¼çx2Ÿó|€4ó%Ï8?Iþ;æKùæóÎO0ßžŸš£~Ö±å3°—êù–]xÙ—ò!`Ïÿ<‚=z½i˜÷—½*ÿ‚}{þo޹_ö®| ëÁó«ÉÂ|¯å{ Å¼þän÷úS¾ZÌë£ÁÄ|¯Wå› ÉŠ~”¯’&‹|ÖëöÈwéOä_”ÿ×ÈS=þHù3ü•óðgýø3fs§|œ4Z¦?”¿dþ>?.—¿U¾ìõ_§cßþõý·ü9°%"Û߳Л' ¾€½E¼P½üÄqø«ˆ7ÀõŠGÀ%xœû‰g+ÞÏyý?7V|ë#þ¢>b¬\õ±ø<øS^ïæÔ×Î'â³Ü?&ò»§>w¾Ãâë…V¹ê{ç?1¬9øÑ~ñâOŸà[ñ=Á/ˆ¿ÛYÿ?!>—Í?òËfh¨oƒß ³ÉÁƒ ~D|2ðrþš†´ùñÑÄäᅥ{¿Ã0êñàkÄc4çÇ ^|ó=â˹lx¾ÅGâèŸ8ߎewóóZ–ÀÌç7ÿ$>ËØû!ÌÏ÷z8ø,õ €ßö~Ü„¿¿÷ȇy¿ý `úçoÙa©)´¯Ñöõ °3îÞÁ!1Æ7ö07u€àØü ªÏh0xê û{‡ Dšwˆ¼Çé#T˜ño†P,8Þä-v¸€Ÿ&Ì÷ã¨ðã÷K\ÁÁXª£†@½#·NzX¤"dXÅ€*uW ôŸ¹ÎÍ ªˆ@¬Žž3žÁÀªc¨ÀœàJ7ƒ«Ž#ðò¥3ˆÁ«ƒ üøñÁñ ÆX©fb|€óÅ8£ƒŠ µ:ªH”|<“{¤`¸Õ¡Eâå+>±U·qttÉÈ‹AD¸Ë¹ãÞ|;šÕéÐóæÀN{ï}grpjRÛ+o®ÿa´[æ9ØP§&ŸF‚ºáÑgøÔ”§ÙZ_™Q ñѨøòz}û# Ã…Ÿ% Ëý›„‘âþ?°¶0V•ìt/³™KŠp+ ~¬‘ø¿•“$˜w¿¾ªÅp‹ñùŸ;xq˜IÛÜǯ¬ãÁë„j¶z¿Ãû„ûö+^'ü¿e"ƒÀÆ*üMû)™ÈO_‹é±y¹..¿üââ%Á¹8ýòkÕ¾;×þŠçV‹Pó~éýÁ.kY¬d¿J侉ÆúGÇ3"Ž£ëpî,Hã¸àu‚B}œ‘ÿ:ÁBôõÁë½Sœp¿â/öx¤›l£ÿù ÿÙ‹Qͤú33þ³W3E~Êþb[’$ýú'ƒüêõ§&] ì£%÷ët§ãðßlÄ­ê± e¹–EP„Íñ ¼TØÅ–¼%™µÆöÔÀhA„«FFÊ-cà~µ”É\-gìç´¤y¼–5¯/§¥Íû—ÓÏN1í90$?ÑBÉMÚ’ Ýr×ÀãjÉ#áhý´ðC¢-"pXå’’B-Ï–„d”Ÿ?R»µ%U’ì€bñãæãH¬$áñÎ30Có‘XIÒ“Õ O^J‘%áÉj‘uI&$éÉÆO’ž(!k]$åW|Ý¢Ã>Íä%‘’ŽH¼¡D4†k?I.EcYøY×KD“EÒåÉîÉ1‰j²š\y6Áq‰j²H$é|ߘ1‰lR!“xf\"XD |‰6PýCô±-H¢Ô^E¢›m(:Æe¡¶”,7;¬"åIFvàq ¨AÑ·W„a4ñöŠQôìç ¦¤{…-’jgFb¶W¨ŠªEÛbµvV4HËKäEÚR)}¤r`xezQôÍQ=$awbEãéŒ&c¤•YMÑÈ:£-žÏ\w'©QÄšù|DN"Áfv¡}8"‘äE뉨¨3D:©ë‰¨(ÑÔNÃE!QQÞ‹íl2‘ˆH‘CtÂ&Å%qR\¤ãê%LˆFDJô ‰l!Õ[4"RÜcˆV0~[4"R¤7‘™¢©#©Ò‹ý[¤’$ª(žQm‰HœM2ºáˆHDµ©ç¾è‘H¤Òz" 5¹ÇåxSa‹6À¦ÖK´¡&ù¸šèhòoÔÞóq4æ¯G´Àp5±€ÛÕäb“>_¢ —h£±yD¸ÿºDžA‘Fs‘Áièý[4ù ó~!Ú@Fž/ÑFeÅtD¯ ¥¥?¢ í!Šê´Ÿ-âÐü¡p˜1¿·ˆCó_³ìßþ‘®&hØýÅx{[åuÈþÊmŸ‰M×§^ö\DöDë¡ ŸÖKˆc=•.‘…ÖŸáv¯ÏÒc½ùú-!BÐz/¢xÔIÉ_”þ „(ÅׇüaW‹Dv“Yþ Mætù·²$OOÓYþ±<Š¿ò§Í0J±)¡0q¸šè^_A„ìõÛtQpä«Ñ¤—¨ …A»EÐѤŸ±©€DÀÉ—£Iÿ(Ÿ-.²>Mt¯'sŽMN<Ÿ&ú ¸6u Öw;ßW=›Ã^³DâQD“~j<2E§¾ˆ&}øßLQêÕD÷z¼Ä&ƒÄÝ\W}†ˆ~F½æ®Ä”°×äÛBw}Mú¡ú;ÑqŸú,šôKë;yýqšèÎoØ4÷½é)ÝMôá8šæâK€cÓU_W“\ü âù2Pþýj’‹¿©#øD~f7É£i¯M!Xhh1隸ø%¤qÎù¦‹ÝÕJðaÍ›âªçƒC™ì|ó“Ó’V¿ItH¾n\-g5ñÛÔ¦R5ñ¡í\Âån9«‰­µó‡L„OËYMýžcSÿqZÌjêwùÛhêƒ_ñM‡Ëù“h9«Éߣ¥"©^-gõ:û(ÂùWjÍü~$úN Ú79J«ö%6AÂØ4ÉMNÑ’öM”À>¿¾É’Ú8Ž_×&ª,¾"«yµ¬}Ó&Ëtß:ƒßò÷óMŸÜkâÇüïiiK²0b£$ C"™,`/JlBe ;Zà’0 ùÿ0Œ‹¿±[â’0Ø4¸ˆÅÆÍÿ…¤a¦Ø´ë|Jð‡!q˜+Ñ;ü£$³j¼´Ikó—’‹f–ê[ñÃ4©ª¯Äýn5w¥›ºn¾_jSÛÍõc› ×ý K S êü[.6î˜ä¾îÅ÷h'÷~¢ïŠfÇL~<ÞP‹}iG².w›<—³‡?åLòÝ>“³Ád9$ÇödÊY­x?Ãøm„=ùrn0ŽØ!üÜ亜㔢›Îs]Æ&çŠäoÊù¦›l—s¶Ïp=Í"ËuŒ]Î}f-ÎŬù,ôx¢yÀ‘±Ø\F4_'·÷bUp –È¢ú½øÜF÷J†ÜÎEÁr¨§Š`ÚïšÑ Åaý)v̲J>d¼‚;œë†Ån2^É9g%‹ã¹Éx%½G2"g½Éx‘ýMÉ{vguÈx‘ý%ÈzF½C¾ÙŸ"¹ñdÊ‚[Ý;TÓEÆ+k3’?²âïÉ\±c´°ù±Éø û[$›ýSðÞd9í¬ô!ãƒìÏJŽŠ7ïÉìVHWNÜEÆ{2\§’Íêdð&ãEö[ X¡erd|ý±c¶.OŽ‚Œ²¿h>-‹™é&ãEö'%‡­89¾Éx'û‹Ú±¨uïèÙ_"¹¶¦b2’Å û[þ>;Ù ²_;̨€®7/²?vDÃæ•ìŠìÏÑìèÌZO²,²?Çû™!÷{GgýQLúް‹œÙ/=Éþ{çà?.2^äÒŒQ]Å…ÈþÅÝ KsŠ‘ýØXïÅòd2µ‹‘ý)È€ÁŸn¸Èø;}|&Y…‹Œw2àÑ/>³Ø‹âljÇ_s"\÷ÅÉ‹vƒ"7ÌÍiÇëòâ3ŠK‘'OQ1ºR£Â"ÿ£ôäø·"s–~±;vq,2h³ãäSq-2iµØA;ýx;;^‡“Û3šk^ŽȬ•Dv<ÅïAˆüšÏ!ÇJ¹È‘g¿üyïHÚd„ȶÙã~ë“â?È:›ßQš|‡Ê&³Eî¡§ã䟓¡›Ì85¾5ùøl2[äáò3q«Å!·E6Ž©¢‰ðÛ"'‡æªŒç&»EfŽ"rž/2Rdga ïCFªžÄÆk¯GÉ22Rõh 1-<Ö+AFªžµ4ÞÉ?ù&#EvjÇ/ë§ #—ï?ÙdcŸOü"Wóú/ÈFÕç=ÇŽ''g7Ù¨únÅÏO!ær,~ ÇŽ-‰K6¹(~¡kÇ.o¿ÈÄ CŸó‹e·8-ø ˆ×a/gÇSv²Ïw0M~Ø©ÏůŒ.>c,碾?3æ—öuñ"Gq›*\ÓÅ/ˆ,µÇê~ùÓ/FYаÄ÷ïü©˜Ãoˆ> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/в¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGù͋љ˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñî§ŒOŸ~÷„óû endstream endobj 2 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpYz7VED/Rbuild110a3546435c8/ks/vignettes/kde-003.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 33 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 34 0 R>> /ExtGState << >>/ColorSpace << /sRGB 35 0 R >>>> /Length 12523 /Filter /FlateDecode >> stream xœ¥}K¯-ËmÞüüŠ=L^êzõc*!1 Àä# ãíe@ÑÀÉ¿Oñ{°j]X628gƒd,v¯îz,VùúýWùúÓןüáëÏ_ãyýëj¯Þ¿Z¯¯«~Õçyúõ¿ÿû×ûú—¿ùË?üío¿~÷óÇñ:Žãkÿÿçïþ~r¯ñõ¯?þñŸ¾Ž¯þQ¾~?ÿýéG‰ ¾þîÇ×W¹Æ«Þ_µ¢÷ÆiÇk´I¯qLª¼®>©ëõܦ¾qõóêWÊïãu.¬(êÆÕæ\¯q&¶ž¯¶t“¢n¶m¹,v·ûûÇÏy}=ûë©_å¹_㽈Sy¿ç¤ÎWé“*¯:›=¯×YM}ãêçUJʯãÕ–”tãjs®×ñ[žñºS·(êFÛ)—¥Âîvë^<ÂÖú«ß¼sÆë¹`äüe[»^e>†£½j3…öŽïJÊ'hÊ‚n^-N‹÷"±³‰M7(^­¶%—]Âîvë^®úªÇW™¥¼sî×=5Äë=ï>Ð25\ói¦ø¤ç ð,ù|¶Vtójqæ-•ÄÎG|-ݤx5Û¶\v »Û­{™ß`¼Úmªm¼s®×=_—Þð!´ù^¶Ixv¢Ð^?_WYò+ž{bEA7¯§¿î±°íunºAñjµ-¹ìv·›÷R®ùç÷0&ìÔ·/ÎT=ðµÆ«?ÎWŸ_Ü]^£›â×Üp–÷Wݰ¢ ›W‹SãÃMl‰&R7(÷h›rÛEì‡Ýº—çÄ×\ñÀÞ;çÁ›2»Ðø¶I}ë뺋9õ¨¯#±IA7¯gÞùHlå"ݤp/jÛrÙì§ÝþöïW™ÝB=âç~?Ç4+Úkø3;óû1ëJ}µ{ÉÛk,¬)èæÕä”çUkbËý:–nR¼s¶-¹ìöÃnÝË|Ïüg#Ç—äŒWŸÅìKfUÊ| s„›Ïõ‰M ºý½gÞë]ÛÐQ[7)~/l›ÛEì§Ýº—9I:Üfѽ˜ƒW»œºó1{ÂèÎkŒ¢ÐÞÙ0”X>0(+ªz4JÎ!^ØþÄaݤx5Û–Üvûa·îe~¦-ÆÖ9Ö©3§¼Îù#ÎzNZ«¯{ö%³K9†)´7»•Ù“¦üòH ¬(èæÕäÔ;¾.c'³/ݤx5Û¶œv»Û­{™£WÌ·fç~k|1§aF×ïøZëbF7Qô’¤ÐÞìüç`eù|LcaMA7¯g¾Wbç³,ݤpµÚ¶\vûa·îå¸bâÖz)ë{ç t³_œoE èÔ0ûÍxR¤ÐÞìVæ@–òSLcMA7¯§¾žgaç׵酫Ý6嶋ػݯÊþ «O6çÆtyðžßX<†ÙÍ7Xç²^fÉO¾\š‚n^-μבØÙO]K7)öclÛrÙEì‡Ý_Θh–{õÉæà½¬³¿˜ßX¹ÙçÎ^=zü;ûä6{ŸȉáNlRÙ''gö«×²ϵîÕ'»mÉe°Ÿvû{y°:šýP¹õ½ˆƒ~¼ÌncÎ^1Üxßc¤&Å/ aqaùü~7¬¨îBr*Œ4¶À,ë.^!dÛ”Û.b?ì^ã~Ì7gïP÷ÁA§Ïnö<¥AmŒÍ˜Ë‚â“n±IyÇ<ØXQÐíY8'.ÂÎl,ݤü›GÛä¤]g.®7Ýþöc„º±xxï..æ¼'¦?7Ç«ù5ÆXw¯ñlN0b^$yL0ÖT÷'‡ c¹±îê;϶%—]Ä~Øí9ÿñª'–œMï˜9#z»r¬Éçšçˆ™vJ×ä=>¾”s‚a¬(èþ–/œŠÉ‹±%^—Ô ŠW«mÉe—°»Ý¼—6ïïdgª~,9XP·ùlZ°ôn…;)¼|ÊçTfaM5÷cÉ)Ñ%öˆ~*uîDzmÊm±v{>†÷.ºÇÃó1qðÅÅ,):Ä“)¾Ó¢8Ãj1ÇK9¾‡ÄŠ:ÖìœÙýŽ’ØÁ©•t“òÚ18¶‹ØO»u/'¦´m¾ÈEã¾9se:?¼“í|ÒzãZ3Å÷ ÆT/å=VŠÆšª9’›sø­ l,S–nRœ©£§Üvûa7ï%º¶‚éæÁé˜%¦¨³Sĺ}>ôö}f1mÅúôIéÀ‚I8‡ÝbÌ1t ãæ0×R+ \Ê6É5Ä}Øên˜CTÃÈùÞ9›êía®OÃò‰‡0xR~b5n¬(èþÖÐ NgW*lÃKfÝ xµÚ–\v »Ûm7RáD¤ÆpúÞ9X¤…+§Ì^oNªæ®ž˜v²éx–üŠ›XQ5os¸Ì„K·ké&e7R©K^IÝ55n[rÙì§Ýê†çԿƼ³Æ â½s:W-–{e®Í±zèPD ]e¥ÿÈrvŽÆŠªv¤&§Ä@Ø#LN݇§ÝÙ¶ä²KØÝn/=|šá‹öÔØœ=f¼á¼S…“KLâÖÔx>Ͱ€œ˜h”Ä&•Sãä``ìÅ©¯t_kj¬¶-—]Âîv»Î‘\/É©ôy6¼¬s² ¸9†×nŠ]%gD–_x¹ŒuØõbNøôGbç×¥›”—ÜøP$§]Æîvë^f³×…7|.Êß;çÆ¨<|¸ð®æØ1‘ Ewâ‰Çdùí¯ XQÐͫű0OlÈnÝ ì}ž%—]Âîv{š?0ÜÎQ}rr:§Ó'_µÁnã†ÃÔ·1—|ÎÈËšjî““S0'6öÀàm݇ûäl[rÙ%ìn·¾ýιZüdœN&§sÍZépG´&V¡í6…/ #’rÎ*ݼZœÂõ.±ã‰aĺIñj¶m¹ì*^+Ÿ›nöcúæŸVÕ‰3Â1}M´>; £¢—)ö5O8L,Ÿ+—º°¦ W›GMbK8†R7(÷‘h[rÙ%ìn·îeÎæm†#M? ÑÎW¾…w¤ #ŒÇþÁÇš›K¹p´[>܉ Šªy1W¼ðBF˜+“à¥h×RÚ(Üfðê‹çw×WÔH ~fo8ûÎÁ˜P¬´OßZlÎAžŒèf›q&2^dC>ÄuFƒ¨µ¯`Û´”Ö·Ùº~‚ÖàLºFþà<ðGÏg1'm±°ž^ˆ‡ˆòo0ê’˜E&VÔå HrN8Ž„ítáI7)ÿѶå²KØÝn%=–—1#Ͻ8ƒ3zø£Û\IÅŒ}vþ±h'õ­ùŒ¸–?±N¬¨ž3zsªgåÀÎØ¥»¬½Ú–\vU÷BuÓýSÝB|\?lê²È‰€îTQ´K‡dF”í2Ånå„÷Àò‹ƒ¶°¤¨Û]8Wx-`ðÒMŠW³mËe©°»Ýú]4/ŸÃ¨§‘æ<1Oƒ‹®Á“ /'–w¬5Œuç“6çŒ@½±sr-ݤx5Û–Üvûa·VZˆeä÷’¼ó½`Nõù=ôRè~Ïï¥ÏbÄK,Go˜XQù½$ï¼±߃uü^ܶ䶋Ø»W4rŽ÷íAçÿÞ9:ÖˆÎ÷2º;®¨ëc*WÜ—91}›Ôpü"9üyŒÅÏ“º[Æ/Ü¶ä² ØO»=õ‚[9zÆCÓr¢ãìœð<7\£dºáhÅéÑ ÓCr,.Ë¥uójq.t`ÄF¿ÛR·¨oMŸ{Ée©°»Ýެ"ŠVç"³=Š¬Šƒ! L®¶ÆYÇcбÒŸ©ä³“© k º‡§âe7¶`ñlÝ ø+ªmÉe±vó^æW®z‡¿ü½1ž@ÌÉó`àg~ƒó)ÍD´5g¬£¦´ÇÈaœˆÛA3°´®b¬‘V¸”m’ak€û°Õ¯Õ€Óxö)Å3zqð'fÝqËqσ€(þôOŒ%–?|¸Âš‚n¾ôâ0jláK(Ý ¼Z@Û’Ë.aw»´Ãc°‡GsN qóÆã!Î."†¿˜ùUSh/2iŽ%¯ð@ kjd§mNý‹°uê®Ùi»mÊm±vó^f·Ð•ÛÃż-ü¯±Ê¸˜Ýs}Ålh˜ˆ¶b¾ÔSzÃÑ&œˆŒY™qÄ"\ÌS+ \Ê6%•5Äí¶:|F°½Væ4®.ö]ÑÝNSŽ×ÂaKù\-< kªäKhÎp<±˜C¦nQ\Š]î‘ —]Ä~Øm‡ãGs¦¡üœä4¯§cnõ“ÁËÇz”›6ËVš*¯“ƒu‰±s*v,ݤèðbÛ’Û.b?ì^£HD5Â#]s‡«´Ù›ÄüaNØÂ6{˜[ò(2W)?c€J¬(èÎ8¢UÂBÂŽËXé&åQ$Ú¶œv»Û½fögÉ™}pf‡Qø-F0¦Ãu‹oí1å¹z-g ÈXRÔͫŹ1'¶ñø­›¯fÛ–ËRaw»Cy0!iÈy|Ԁˆ6pÒÎilL²ŸÛ”£"ýZòÎ>IØîªújræ·|݉­È{´îš–n[rÙ%ì‡Ý~ÇN ½Ûy—œÁßòb.Âû1 Wã²ó®<s9(*yaMu;ï’ƒ0Vb•+Ý+OÖmK.»ˆý°Ûy†|ƒçÉŽ{stñ ´r9½$&ФGÇom9ß cMtÜ›ƒÀib+óÖ]Óqï¶%—]Ä~Øí>YÙ}Wæ0wdàu‰/.†¿¡<Ágev š%9½ÜÆŠºrx4qöÄ¢LÝÙGfÛ’Ë.aw»y/7Ç×ÚãÖß¾Ùè,Ü7óæoÓ$4•êGJ=9Ä »uôˆÍÀÂݸY†Ö —ªMJe q»­^È«w¿Â1óÞ9×¢LÍz8=þõ2õ­e?²ã$g.ƒ±¢ ÛKspN…½‘?eݤ<˜FÛ–Ë.aw»ÕuE@v0p¯®Ëœ;RP£³©':ŒX‘Ì7Aÿdz•6¿Ú–ü†óÙXQ§',É¡§ÒØ—˜uwÏY²mÉe—°»ÝúÜÃC¡ïÛ÷bN ©ßSt(Їå2Å¥†Ò™$¢ÇL,)êæÕâpA)ìy†­ÖMŠW³mËe©°»ÝËÁÃÌœ‘pV ³Ÿ«\kÆ“Ž)ûcÊ.“ˆŠX~cø3VtójqV~ÕÍÞxÓ}dºÛ–\vÕôUoºÊ?Y;½`LÎŹ†¨ÖϦb)LŠÁi q)Çð—XQ§ŒÉéœ× Û<›‚îæc¶M¹íêž—ÕM÷O}RãÒû½s8òž9QŽééüIÃ7HÊn¹ðxI~!‰ÆXS«§ GËoaci~¥nQŽ5sÉaWbw»õŽEÖ쉥›Ò%’óÄUñÀo¾±8š£n;M96±iË;³Ð‰5u;]"9'–KÂÎ~ª-ݤèÙœ$rÛEì‡Ýös$'FÚA•£[œ†yiì9ñ©Eà?|BݔؑJONü 5±IUO¤“s0ÅØƒ‰îÒMŠo¶m¹ì:˜Z·ÛíT©ÎL=¸DÞ;çÂc:Nô~þHP:‡)¼³!å†<ÒjÖT†’Ó™†%lC`Ý-Þܶä²KØÝnÿ. ©ˆ7²ûß;‡~X˜v„ÿs#J3ågY–Ão›XQ‡‡qs.îö¢OHºI9/#Ú¶œv»Ûíïi1ð;Ô'›óx¿ŽÂÈ=8¼ÐJ–r$%VÔásØ¡ÛùK7)^ÝžhcÙ%ìn·Þ±2ó¹#çê½sçs5~È×¢»“ߺD%¯‡³k ºé?‡~&bãÕ?S·(ê>ùH.»„ÝíöXùà—åˆðÞ9O\…`L‡/}ÜÛÜÍ—{ÍY*wnò¶çƤÖÓ& ˜P ;²t“úÖ¢6Ú–Üvûa÷zÇ Ãè{çtdHùÄ ïAm¦øœLù”<·>+ ºýF‚S8žN­›îÃñÿl[rÙ%ìn·Ó$›ÓǧI’s3Ì4GjeR éìD/JУ߅ÞrøàKŠºyµ87fÕÂ^ H7)ÏF_rY*ìn·—ø7~çùãë^’sÒWÏ”×ÙËžŒ³Å KŠsÙŠe¸åH4ÖT÷½$§áÙ[9ó•îšf·M¹í"öÃnÏa.,Tç÷uéw1ç„ûTFÃä™SÙw¦1_Hý±üªØ"¬©áŽÉéÈ1¶1 Aº[¦ÆºmÉe±vk|™ »`.Pµ•МkÖùzs”HÉh•ù–GnBkóUɱâN¬¨Û›Ð’ƒ¤±á[ºIñj¶-¹í"öÃnÝË|Ñs©y/æ0·4þ4øÞÂÖ©¶VSt>ž1Kù…š±¢jÞ‹9ô· {qŠ)Ý¤ìØ¼Ú’Ë.aw»×v¢ÂJy É9qã'’}c‹G4W­÷0Å÷;TRÞ‰VÔpHrZL¦[Ý”îê±¢†÷è$ñ¥Ä"¾”º«÷èdÛ’Ë.aw»}/ƒFÈ:|ß2ÒeéŠeÚR/Sßê)®3å×Áœ#bMå>šäØ:$ìTú,ݤ¨û´Oòá\(`w»stÐ7C}r´e«Äh„\l–'hÕÃ×=F”ŸÌ¦VTõÉÑv1b gÒ]r¨wÛ–Ë®ÜjvnººŽùÍáhqr*ÆÙÊ4劽ÀøEaúªMc–cI’XQ‡£Å攇þ|b 6Í[7)^Ͷ-§]Æîv¯>¹w&ÏÖì“;³¯‡V©+3ûa‹”ûdÌ9rܽaEuOr’CCŒ-v^@wñ$'Û–\v »Ûíi~a‡÷ä;&ΉÚ1M*Ü;ïЄ†«‡”݉¥-ù…®ÙXRÔý­%„ºî«'6òK7)OïÔuC^ìÞºú/ìv$;ıÍÀ‘6¸öHq麲 S-¬©‘59ÌéˆË!ܺ[n!sÛ’Ë.b?ì^}rüν9§59Üó°XN?½C<:Rî“åHÕó¾aEµÜ„f—ÆÆZ¤»ØI›mK.»„ÝíÖ½Ð?›EåÞKVU‘œ.Žƒ9ôâ‰â“ff§åÛ%VT³{/9Ì10¶8ẋÝ{ÙvÉ„Ë »Û½2xæÓxP/é½10‡+t®7b‡~‡—tþ!a'êT,ÅÞaãH´Ü,FE’™p%ž¬µ¯×ܦ¤´F¸ÍV{[Ø=Txÿß;‡Y¥7]ƒu‰ngEñ§gg$ùS˜†C¬©-Ö + [‘uoݤì›AšŽä²‹Ø»í5F„èÏ•’œÊlBD§ã“»OÆÖ†)v/­,$ù…5• ïæÌ±â¼[†nÝ%Þݶå´KØ»=T %ÔHzoNÔ"m16ÀG)¦‚¼Å¨˜AÊ^jv’¼³4”°¤¨Ûq“S)@5±5¬›¯fÛ’ÛRb?ìö.µÎ=G8ÈÞ;‡ëN…kf¡*¶"‘Âpqùºä + ºyµBI…eY4ë&Å«*’œv»Û­¤æTìÑxo ”+š/el©‰œ™ò &ÍD´¹-ÅŒH{Iœ‰Ãy}b*® Ž'­$ •mJJk„ÛmµC²†â:›­ÎŸÉKá/?¹JyXé£4Sx\C5Æ$¿âq&Vtójr òìŒÆ¥›¯fÛ–Ó.cw»=³çòâ„?æ½sº‹íEöVlê©p²GÜ™”Ó Âe$ùU¹ä&ÖTÍÌNs Ü(ÆtÌJ÷±·j[rÙEì‡ÝˣΎŽ+øàÐsŸ·ê_,‘ù¦ìÀoÏ’wŒ52 Ç–k3kÍÔ]=Óζ)·]*þ¶Ûín˜ VGq}£ä G1:”HÎ:ïþgì$ï¹Ñº1¹Ër&~+*ӕĉyßóóÛ+u‹âÕÌ÷œ'Ýí¶GïqÒ‚“©Ä)…EDˆ‰B+,7r¦¾ågRÒDg­Î¾auŸOþŠæÜ˜{¥W–nR¼ZöRnK‰ý°{%¹ .xŸ3“ܳbbR°?#^ÏóÄ»òb8îBpÄrem +*{ŠäT÷„ŧîì)²mÉeWu`plº]«© ôîÚÄm΀w˜„YéŽ4¹‘³¼ÆÚ¢â4V!Öuójq. ¦Âƞᥛ”“5俾Ÿeéåô‹¶éþé~O$jYé`ÔòÀŠSÛÌÚm*{ÙkÉŸ¬Ît,j…}ÌÁ†cc/÷ÒMŠW³mËe—°»ÝžÙ3êVrG¤9•%pëæÖÂ`á…‚…µ¬`!ëæJÞ.ÓˆUrGdrèt¶Ðé"Ý%wDºmËe©°»Ý^Õ+³uCß;G5*22;ƒ)rýDqå=0·\9ˆÂŠ*k9Mù–Ä6åbR7)^Ͷ-§]-s5ïM7·qÙÇ]ïýîõrõ¥¨+UþEo(‹™JÞL4ÖTú´“SXZR؃ é>²î”Û–\vûa÷š‚1ýMesÍ@,9¦C7ß0Yª&<;Ìx0® gâvq3ÎpN cMj%á)XK©¬!n·Õù sdÌ›6GålfÔ6o‰D½é¶v”ÑcyÇÛe¬¨#³!Å™_}+‰7|uÒMŠW³mÉe—°vÛ‰WèÒ¼ì(NK:•Ê® ™QKîROª óþ,¿à?0VTVEH΀ÈXmK.»„Ýíö´%+e^›8u0³&ÐÅ­jª ”ymYHÕ´Û†Í AÇÊ‚§°§°#Ò}¬¼6µ}¸~ç³aw»×t2ºíûð©ähvÇJ7ÔÆÌîqoE¬`10gEwˆÖ*’sÓ“±¢²Œin¨bÙFa;*IZwÏ2¦¹Y‹rÙ%ì‡Ýž¶ Ë/>ÊÃÓq¸¯LiÊ7âh˜h¦<yî%oè1‰MªûkNNuþ.°%fû©»äc· NÚU7|oº9¤¨[¿s‹”9ÜîV9\œªl²’[¤*“-gnš±¢îÜ"e è;‡£ké&Å«5I.»„Ýív®ñà»»@Hr0‰‹|ßÎUwaiħ˜B{ci’GRóšêYÈ×ì¿I,6ߦîâüòl[rÙ%ìn·½-7’Žü]Ì9™3l›¿ñÜg·³yRöˆ à”ägôú‰5òw1§±Ðƒ°•bé®ù»¸mÉo‰U‰gÓýS¿#r»*’Ã4æZ™ ‰I\¼µ˜â{0\@³q›Ì³aEu'T$§1ˆ$,âp©»8¡"Û–\v »Û½ l`;ÐH׋9¬äÑPG0Þ!t9kC^ézéÕï äp–kj¤ëÅV1¶re"Ý5]/n›rÛEì‡ÝËÓê\Ì'=­à4.¬ïÌå|à@«‡){Z#™ÕrUW"ÖTÍÅš9Ò¯…íð1[7){Z¯’rÛEì‡ÝþöOnµnYòÛnÜŽ/rp‹˦önÊ_óSS)h kj•Ì0‡uAŒe%cë>²§pÛ’Ë.b?ì^[=c$ž}…Š5%§aª×™\PY)VuQÔ•Æ€¨t§| Ë1ÖTq8.9‡«˜wæßÜK7).ÖØ¶å²‹Ø»WÞt, ‚¢ïƒð~d°ø½l(^‡ê©ë¨„Hj®K~ÛǬ¨ Á&!,c‡)¨ûȬ۶\v »Û­ïåF¾_áËúÞ9x™ÛÍD_¾èíFÞ (¼ÓíK~Áme¬¨\P™ÃÁX¾úÖÝ2/Óm[N»ŒÝív\5OkYlNÅ„òø%·m¹,v·Ûn¤ÆÓ®<¶Äœåo¦u¨jØ'Ê!ßãXr)"¬¨+-1ËxcÕÑ¥nR^í`Ÿ›ä²KØÝn¯Åº8sÎoJóE,%¶ÀEq•'» är{ÎÀ˜‚[® ‹°¢Î,+d=üÆÒÃ_3œã9¿Û–\v »Û-×Ë é\ÓUNÇÌ@Dùæf¨†ýo±ò6 CÐ’¶+„3Q½m× TŸ.BÊ©•„:¶¸[ìF92œ£®Ëœ‡Y¹ ƒël9^ZR\"u Þ–sk©±¢r–”ç7¶–r¡îš6·-¹í"öÃî­–O"ñ)EæœðœÅî~¦ ŸÜ-Ñ‹)ºz½Ê’Ó=m¬¨•iir ËrøÖ½2-ݶä²KØÝn/‹ 0Âæƒ÷ÎÌòãy;Á“¦ü¤Ï¶ä'³š„ݼZœÊÁÂÏÛ »¬2¼j»xÞÖ7ìn·»®ÓÇö€‹S] r®Y9Ί²O[‡+!s®0O¬©uÖ”8Ú¯*ìÁäPé>ÖYSjÛrÚ%ì‡Ý+9äaw{ÀÍC:R6 ‡ò‹ïìy˜ò;}Ü)‡¶°¦jN Ì9è G`¥nQ¸šm§\vûa÷ªæT¹ÉJ¹`ÉÈ܈lz¾.(¥ysKÖéŽÛež”ÇÁ kª9,9YƼ\Ö]œ –mK.»ˆý°{móÆ1imT2|pÚÉ/àä¶qœTÂÊß÷ÊT›íæ„+èNlRmT"Nái#Âìõ¤ûXyfjûp¯Ø7ìn·Ý{§svµÿ#9ܼÙ@±ÎÂ͢ݔ]pgKù(Î5¬©Üÿ‘ÄÞ‹ÕzêÎýÙ¶ä§sTØí^ßËɪH® (Îü0oÖQùöXô°²(>;Tþ±œUŒEÝþ^À¹\‡%ð™'ÝmåÌ©mËe)±v{)Égž2ùÞ9Œ˜u&Dó¤É˜÷òµŸiÙy>¢å'¿ÖËÕå× —ßZ¦‚Ã] Æ6ÖÎ–îæ l[rÙ%ìn÷r#EßvUOÁ’ÓX¹¢_Œ=g܉Ño’²©'϶%ï>ïõn¿°{-¿nþ€ ë%‡pqobø£0OJåÕq-9öE&VTÉ@€98»ÎØÂû¤»dé·m¹ìv·{åæßLõXiNÅ>þÊD‰ÑœõFÒ ³U,zÈuºa»SQ=VŠÓfòÇQ,ݤx5Û–\v ûa·ÝH *v¦•=+*‹6˜'?‰ —×ÒMŠW³mÉe—°v¯wìdaqç;zºb•@«"p‡)¿c(DØÄ?ÖTº]“sä±,Cp§nQ~Çî’rÛEì‡Ý[nþ`ÆÏµróÏkc¦Œ¶ŒâÒs|A~}[rme–TÏñ%9·x`ûÅNÝ=Ç·m¹,v·ÛIÔ³±,ŠYfE}Ë;€’y”sø7ÖÔ*‡d‹‰e¹#ë^åÜvó©4ua?ì¶Kìò‰¯Þk _בQÆU-)§œ×’W–X&ÖTËý¸æÔ,ŠÍï÷ÙtçiÙ6å¶‹Ø»=îs[2|”œŽ°¼*wÞÁ˜š(ûáð±ütfÜ‹ÊðQrŠ#ÌÀŒ Kw†²íÃæcÃîv˽טMV„lÆqF0 ¾"å ÞûîÜÕù&žÒ{K‰3Q¸jÖêÂÀ¥•.e›’6'»n·ÕÃ#ÆJ‹xïœÁêo,ÙyLÃf9OR\´£œfÊo,ÊÕÖæMqPŒ"±¾zë>2bä¶%—]Âîv¯i~¬äã ¶šÓüàðž˜jßìx±ëRžæ‡'Úò %EݼZÆü…íŒùK7)Oóu†üS—¥Âîvë‰â‘}w#%‡yWqÂwlŒÃûîFjQuº-ù@ ÆŠJ7RrXÇŠX¹‰¤{s#©í”Ë.aw»×ç^Ù«i nrN§+=¨8õ(>Éç0åϽÝK~bg¬¨Ë»«’sò\Cbc^¶t“òçm[.»„Ýí^Õi.ïì snŸ¡¤¬Æ³´z5õ­¯ }¤äÉbÆŠ:3»Â&XËÜën™]á¶)·]J×Úí^›S ¼Ä7‡Ëô7åc ¯°A_K|E$x †°¦Öß.ÓëàÞaµÄWÛ–ŸN\8_Øí©1w\]èëß;€xõ wþzWc…JŠqÆ“v’_,¡#¬¨Óõl’Ó¹·QXŽ%ÖÝ\Ò&Û–\v »Û½ ÏæÌy ®9‹ÅýÜAŠn+l{H97tkêJǦ9'ÂÄê”FéÞOqdÛëDZ°v;òÍœä IïS]>3æ³9¹“¨SŽNgÉYÆÉXQ‡{Ysb±Ö{òŒ^é&Å«UæIrÚeìn÷*½qñwô÷Ò²X~ –½ÊcDf )¼Ó­ó·–œ©õÆž~Gü½˜S˜>.,O°î#¿·-¹ìv·ÛolD(OºÃÅQª_aV©’+k¯‘┙–3UXQÔÍ«ÅaÖº°…yêÒ]Ö¶kµ-¹-%öÃn»];£Ë™)’œâ#M*‹<œ!+óÂÓ×ÎÀr¾ÆŠÊLq°ªNÚ‘ºEñjn󶜵6Ýíö÷ra¼êyyrt–3vÁ`S&½$ÇmŠßgaºšä•©lÄšÊCÈ“ƒƒÄËc‹­;!϶)·]Ä~ØmòÍ"øñß;‡um*¶òFõâ–öpý¿V9{“[Œ5Õ³þŸ9•ï|õ̯mºKí¶)·]Ä~ؽ6è5¦-ycÎàÐÅ#Ý.Tcˆ }üΤ0Ïh<Îrg¬¨–ss*7â [¸/YºËÚ·¬¶%—]Âîv¯þÍ=äÊàI} Oñ>ñ˜Ä>ÚÅ}gÊXÎ7ÜØáýéç 4בÕDŒå„õÉýâ¯VÛ”Û.b?ìö|L»~óD€ä ¦¸Dè –‘.Yš)†;Wv’#C'±¢ò\€ä`ñ‘XÔþOÝy4@¶-¹ìv·ÛãKánlâ|ïœá³ˆ]‚H7S²Øé€P”nźPò“ûZ„•[F“éƲ؀uç–Ñl[òâÝ7:a¤mºù½°—ŒŠQÎvûåÓ° Ý…M©cËÓHœñ,ygu\aEõ,íd¢^‰­(ÝdÝ5K;¹mÊm±v»OæõÇ•ZÍy0°œy}ç&¤p[‘b¿Ië,Ǻ1±¢Ö¶dsNä“ { nh¥îcmKVÛ–Ë.aw»?ªk¢Om{uÍØAT«+\Æî¢Þ\S”×/­/9ªg&VÔé#É’Ã.ÀX‚ÖݽÑ,ÛV1lÙEì‡Ýþ]p¤wÛ–Ü–ÞÞ]7Ý?݇3òY¼ÞçÁ 0jæƒ Œ­™Ê£.ù…Ÿ±¢î=©l ©[¯f0ÙrÙ%ìn·Ç}øÆ óOÞ;ù)ØzÁðîÍô r˜r:Äx–±ÄŠºó­1¹-Âf³H·(^¶S.»„Ýí¶ß’í^¢NŸñ ã‰yF– c!å½DÝ…4ËQL9±¤öuâ\9‡’‹lõÔyñÙ¶ä²KØÝî59*’ŸoÖæ'õ­ [$k[δ=c{NÈ|u÷¤¬ÆÆ¤êIÝ¢x5Ú¶Ü–ûawÜ˾þüã7ù‡¿ýí×ï~þ˜³†ù‰íÿÿüÝßÿ@ÿô¯?þñŸ¾Ž¯þQ¾~?ÿýéGì^<¾þîGOÛ¾.Ì+Þó¥ÁùÂ$ÿ×××Ï_ÈI¨ÛqLpðþ¹]€\G^ r» qÀM Hîì6¥‰Û¿ý£Á÷_þGð—ïùñ›ÿ:?„¯?þÙá¼$à¤Ox:Žþãûë?ýMÿÏ_üÓÿòG¨þu,ªè·\ÿýàÈʘ¯èÿ~lìkŸOraÿí¶ŽÍ…ÍHºqBÒû‹Tã‘xÖ¿#%Ô&\d^K.r» b£#Ð"÷ žº· r»@6ù‚ÝÄýö·{þ›íQtD×x@ó_ÿÁáµÙsýõ_üWÑ8™ó­¿ú“ÿ:˜%}þêoþ«àãGû·~t}`?úÁ§Ã>À$òÎIëù›üŸ_ì„þ¿?Bd(ET9Ö~aÞÿùÜ[5øÒƒù¿ÿáÇÿâa¿6 endstream endobj 37 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/в¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGù͋љ˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñî§ŒOŸ~÷„óû endstream endobj 42 0 obj << /Length 729 /Filter /FlateDecode >> stream xÚ­TÉnÛ0½û+t¤‹á.±h{èt95zIs`$Æ&*K‰$;Éßw(R¶’¸-Ô:˜3|o6ò‘&>šhŠs®©ÖL'åvq»À"gBŒ€ÙrÜ¢‘gŸ·<ùÐ.¾Á7m1’ãœN„ÅOceLk,„J²YïV‹³s^$”`M4MV× USmŽÍ“U•\ s·Þu6Í8§ˆ½J3!új»ÆÖÞÇPe›Þ¥ )ËÑCÀÙ~p[3Ø>@îܰ 7õn¹&½­m9´]Ÿ^®¾œçb^K¦æIÆ)V0ر’·i&iŽ>õ垆åëìèúI$¹s4Ä“zO1Lü$ 3Â],é%¢€±diFÑ¥'&Y„RŽ¥ÐK—€¥„ D°Ì9Qœ‡ÔÁ‚(–ƒíÉ$ð¦ÌÓžÀa¨-N5O:ë›=ïWάÿѼT˜ç,¥óö™,„o?¶>Ÿ÷îÓÌt!„(&“Œ¿yã;’H„M1$ó¤ÅÇÕáŽFQEÆ_~ëåTú³‚fžR(‡€‚(qý\Aü´‚@?½û»‚² FÙµ}`ûTJdjW™Áµ§dG*1S $&±`ñF­6¾"Ю)Û¦:ã[וiª;WŠ3FƒìÏI$ÔR× ÷'Göhȃí‚cؘfZYþ²¶‰ÊvפÀÎé†>xï60±v꧘?¾YÃX» ¨2þÕ1ÁØgŸÆz¸FµéÖ¹5a^µƒØÆv½çªv·;[?¨s—S¢²ÍÆþ+×À‰Ÿ¹·ýÒ/)ªÝ¯YÚë¶›ÓÁSí¶~ÿ*NkÞXþ`ð‰7c<¡éÓ³ñ®Ãóé 8§°&Æ÷÷?ÂâxF³4‘Ï({gïüЦà&¢×¶±©ýD¼ÙÙ²ÝnmSî„ЩB/Œ €× .Š"@ø¤õßc¾ endstream endobj 23 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpYz7VED/Rbuild110a3546435c8/ks/vignettes/kde-007.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 43 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 44 0 R/F3 45 0 R>> /ExtGState << >>/ColorSpace << /sRGB 46 0 R >>>> /Length 4597 /Filter /FlateDecode >> stream xœ•[Ë®%¹qÜ߯8KpñY$·,,Øê´´’Ÿƒ¾ØAþ|EdF²xzfZ3‹¹·óV‹•ÌGD“¿}äÇ7ÿ}û7ü÷«Oo¿üî÷ÿü«ÇŸ¿{KÏ”ÒãüùÝŸ¿}ûåo*îÿôŸ<žºà?ÜÎk=Ûõ¨ëzÖùøôþøÅ¿~þË}øŸoÿáñé›·údoøMÑå‹!Ì.©®GžÏQl„ÿÿã?>òŸ~h€ã©Ç`øçº0NzæµG({|裯gjQŸ­=j+|UÁÔ{yüß<þðøVnøõG¸aÎ ã®Tðs´üøøëßᯣ?þúöÇ?=Òãßßòã·øï›·l¯ÿ—·ÕŸ#?jjœÀ;Í2a^ÏœŸßV{¶žš4볯GŤñ˜ø`^ÏÒÜ,æz^ÅMθT¾È†*ðv1oŽ÷–ñìæ¢#kÁ/š9¥g+fËlŒ‚yÕô¼ªÙ/ÔZèÚæ¤Zë3›ÏknÏl6¾ãÕþœÃl|g…}Ñ´o¬u<g—K~6̧âýaÛõçšÝŸßÚ2§E{>g7{Øø˜XofûýxQ)fûû| Ú½ëÚÒ1~ƒ§§Ùݾ§ ómÏdã÷ôìé˜?ìÕÍ®n7:ªÖö¬æ¯Ž¯Ã?}>k6ÿÙ:çË¿oûû²¥}ÙýW^Ó–+ùý—åN¬f¾°p´›ÆÉ—W‹Ÿ¾\¸Ñç?0¿ëŽ•<°ž˜/—ÁïÇüÆZyb=1–Ñ¿ob¾óŽÄ<Û312‡¥ÝÍ߸y.$ãÚÇŸÓâ+â<¯lߓоùú'K7Ú1ïªÍá‹-kÒ­´¿¿àÏËmÄóÔÒêó[A*@m@¶õ×| Í·["d„µÕkÚ×2ÛêA½#_VÜÆ¹/¯W…@$÷áñTñ â‘mÎú í‹6ÂÆÞþÄñ1m_ïËmÔê9mΗmÕÞ?ÁmÓÞ?¬Pe¸Ùò‹6‘Ï–?fØUñÂû1_à“êßc‰™‘o«ïGÕóƒQ&ûù=È'ͯp!rIÞ?ÂDšö|·B“ÖóôoV=¦ÿO2Ê|«Çú0bý.>ŸØF´¾°“¥É½þ°=ß/‹v}–uÄOôLj?´9‹ï‘Q¼7¾GÚIÙÎ=c32¬ž‘þ¶îqÌÐ+BFÆZÅTÉÈè««Â0Bã Uqh{Å÷ Eâ*H5Î7<¦ŠFÛ2BV…ô‹PÅ$4«Šš³+.Bc¯¨*²€¦Wl´šªà´g¿+>#Ê;,l{^¦‘QˆzUaÄEDªÃ0b‹ÛÉüÌVŒ…Ø®E{ùõ‹¥cg;3<2F/7Uh¶æ~d;$ý©šq›w DÜêw³u|g<[·UUudUë´Uaa`"ã¨@ìð¬0Q±Ôñ3׿­NˆÑ× v*w…,ÕbWT!<7"©–áQ…`2iöñáì .”g $Æ{t!¨ æÏZ—»ce°ÙbþA©À@»ãѱ#¹¥¯»c"Lx„(‘Žh`c=w DŠ_ËÆc⎻£ÂÅcÖáøâttè@Ȱ-Qc¶¾üвn „N„à¶#z~–3&ægº†!¿žØ6‚£ =œ±¡[Q £Ž!‰Á”¨àbgp #g¬Ý¨îFb|vÆðœ!q6g”BŒB¤b” c«b  {näûñб2MŠ3Øò‚˜Åpi;ÃtÌ4íΈ-q6c¶47Z’Q‚W!´²á Ý @ {Œ@ŒžeËgÞŒB ˆÏ¿nFÂÆa¶:¦…‚e´üg#âxÁx¤@h{ü%?6cÊHœÒo†ÅVˆDÛ Ì6+‡N$7ófµé{U»ÙàòPò¸|æÁ5×’&ÌÔoæú¢ïxc.L$ÎŽy¼6êÏÇÇ/¤µ/5%—ÖBËL‰‹l?üÕWwJ¦¬õ@R¢Ö©Y})Ýý}Í ko¹"†€›@êÜ´ù¿^wër%—sÏ÷ï™Ç –¤iß ó¸à䜛ç çœö~ŠzùUé‘Hz=˜m®<~h?]¶´&]χËÏÐç¡K³.±Ç ºT1\RïS–f Ûý#:¯P(†Íwk¾Ò¬³4¸Ð¨‘ß>î ›ñ…f[hÖ®Q#¿]a˜Ž¸ƒQ†fM oãI£FX9ƒž¶Y{3TiÖÃäŠ#üÍp¥a#ßÝ_Ҭɘ‡lbø­ùJZS" Å©ùJ±Z—4½dù{k¾R¼06ÉÖ€¥­äõ Zê‡,…Ÿmïg¿î‡, 7¹&‰…¥F¹5a×¼­LÙüŠQû[vÍ›eο¯š£oMXš÷°°¤mrkÄÒ¼/ÕmVÞ±4n—Bi/.ü­3±æ¡KsDo¡yS#nÝ5âЄ¥Yöš¸Iw·FŒ|©íЄy½š°žÇk»?o›ã·F¬ùpÚéøžØ¢zskÈò~õ%ÿñ}[S–?éF›O±oYë{<´©¨mÍø€þس֗DÛ÷8ŒjoEòãŽàbòÝ#|š¤4TØW½3„@äòëÅ4Æ­IJCe¡ •ÞÖ$¥©Öä‡BC¥t–dóxÅÖ(¹½9RkÚ£Xwj”ÒXY¡»* 3nk–Ô\ë¡QJsm#vŲi¢[³&&ß%Cy¥4Zt¨»téˆXVDÛåˆŠŠ¹Ž  TÓ Ò­Y¢5r=¶F)MZ»ŽË4ä­QJáwÏè­YJ³ŠÀŽp»F¡ÑrÙ¢ƒX…ÍR,–y„OÍRëªÑ‘Š}ßÖ,¥±RÁp u°ÞšeìŠ-ÅS!š¥:,€F‰ÜëQÁÕ‘“v¢ƒ§X/ì»#Λ¿³kr»ÃAä¢]f!@´¼@Íjw0!fíúïÑ…€bW'7×ÝÎv}k–B`%ïr°5J!8äc:N5”}jƒ2ıK²,!V!N¸Uv±ë[£‚­5®k¦,d¡¡šÆ¥4Ó:„ÀÈuš¤4ÔTši“&ÏD)Ç.|h¨”ò!LÛuÝš¤4ÔÛv µ…&Jqñ@|¡¡¶%F$ 5¢ÙùÐ(¥¡öšj1%nk’ÒT»ê‘mž§BSE=0„L•Û¡IJS½Bó%´»R eÅ5G.ä©Qúqœ[“|a ^á‹e8+¼‰/ÿ”©/iï÷Ä¿ |Ý`Šý°?ð%¶ÕfüÁÔ?¸+ˆ;ký »/T [F¥Ãµˆdô»Ùé”Wéê“<7C7y!}?ɳÆCD;™½N2Mpp¸‰¥_Yd¬½j®*«ëÀR÷$›¼)ÔPºü@˜Bv>È(É_‰Pf¨ïçú«Å«µÉßqÀª,—H- Z²YÚö÷(5ynÃBMdt¿RþZ‘Úóô߼ʼn!;ŸþžÃÉb¬‡JÉØ¥'ù£X?•¢‘´³|þ;tUº®˜¯;æ}•>¸Í·wh“,æù?J)RÑ[{B>È¡J/:ž¶³†‘MUº›à‘UØ  ‹*ý-¹ëã¶“ÍÔÕÚa¯“ Š|& $4eéßPDä3å›lÚ"LýxÁÔž€6‚ÆëÒü—©474´^Uñƒòq ( øÒÔ(äSЕ5˜Ðw“C‘ÏÙ|ýhÏrC8ŽÐy“A‘Ñ)(Bò™ÓAEFá&'‹X¸Q2(2 ]ýúðíö ƒÙÈóMþòp(,hüqáeòÉûöFð¥kÖŽåÑQœN™.–°¬E6Æ9ÁN"ÜàYäb\ÌsÃ_ä‚;z1•¥òÏ"3"\ä`n°?_üã3ÌðÈÍ^q}W)ì“nîˆØ_:âÌŒâŠqû”Ì·~D°2tËÊ`îveø<¨¤+`j»3L$-‘=U˜,úIÛŽF«BeÉ•D¥×q *åâtTDŽÈ)ª(QTQQ¡|ýUK ¹Ð;Ì®hªè¥DYQ!}½ða«S„˜ë¼O—‹i¿€íï±¾ºË{®Õk~=ðPEžxdy!:ní:0à}èz€÷ÛÞ—äññÚ¡1´üzà{…Ž ò+Ø"A˜t—×/“Ï6¸0^ÀÉ@\Ø7¸bêM¾û|6¸ÂêCóéŸB»DÖs³ÿ—à×BxW“\ωr47EÛ‰H^iT$ʬH¶/üR¡<ß~‰™C>:©hÃ|8P¤˜tÙ}ÃÌåØ1ûmýîÝ endstream endobj 48 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/в¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGù͋љ˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñî§ŒOŸ~÷„óû endstream endobj 24 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpYz7VED/Rbuild110a3546435c8/ks/vignettes/kde-008.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 49 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 50 0 R/F3 51 0 R>> /ExtGState << >>/ColorSpace << /sRGB 52 0 R >>>> /Length 4850 /Filter /FlateDecode >> stream xœ•ZK&Çq¼Ï¯è# ˆŸºžÝu•` `Ãò. Áƒ Yb‡†DöÏwDfdU}³Ëõê°³“ÓÝõÈÊGDf¥ãÛ#?ùþýêýË/úßþêøãO/çã<ÏcÿùÓ|ùåo Þÿç#]=ð.§ž¹eôG¹÷¯Çw_ýû‡ÿþË7ûñøÓßþðõ‘Îã«¿ü×øðõ÷Çûo_þå½Mù›¬1ó›1MκŒ#Ý+sȯþ÷»_éû¯÷?¼`ûê›m0ü::Æ9iÌò;?Úxœõ¸Ê£Ö£ÔÌ©ò–üçñûãGéå×ï —ûîwœ?¯šŽw¿þ7üõjÇÿ¼|÷ýqzIÇ·ø÷ÃK²éÿõe´Ç ­ä†o×—Q¹¡’û#µãÃËHë‚89S¬ ±äG9)ú·{§x? Ÿ–GNÓ‰“=¾ªÉ9dû:+]ƒÇû˜K¡|>jßÖâÃc¥ýZ“ûÂ?¼¼‹ïs¿gÇV(—rä j95~¥\÷µæËW{”®õAqùêÑ´¾ó:ò}òÏ”ë£áû»<’‡4Èõqú~aY”±^ÿ¦.ó=–\›ÉÙÆÇ\x> ?a=”G³õ$(¶V›/5[ϸ}H¶÷1¬3Äz ¶Õšd˜@Á¶|½ùä÷Ûè6_N´Œr^T³Ë÷0ùò÷ '.\¦ŸëcÜ¡?ª?ñ~ªür4¾ýùø7õ]ÒxÜIrãšåA.ÉÆÏ¦f—ÍØ ”r{¤ä^L†b벇TLŸ%_ü3䚨ςi’í§âûbòíÏoN¼ì­šâL¶ùZâB)»þZãD”»­¿áýarµýö“¾»Æëø¾˜ìëéÏù›ÃlÉÖëú½ÉýTûþÊnà±?fâ|•.Oùrý…¾®›öXræ1C¾²ô †NýëyõóŠó¹µÀŒº¿ßmó|°ó¿ßÈ6ÿ8-VL{²˜‰ Ûì1ì †Ýwû„açs³_ʲg`·ÿaó/fO”3Ï#3^\Ë¿2&ðïûb  œ©ÿéŸù4ÅNÎðçLÿÌtcÊ—û·ü?ÃüÇ0ªÉƒç5ãE†ÿöºâ åáñ¦ºlf<ÊP<¢)ãÕU$ÿr¿ü¼s²@–{¥Ú\îŒoÆžã ßOTåÄ…dæ nŠÊ0ÛfóÁ`ˆf]ýy£âr«îÿ9›aäû‡áÂñr;¥:Þ¯ƒË† í¯v£ËûÛݶŸRiȹB¿6ü¹`ý5i¿ðç‹ßŸT3å›å>_Ü(_6>˜¿4î”-Pnï‘_\ÿðod .ËÇÇŸŸÒ6ŠÀÍmY|¡b¿³…Ê—íŸþêòmóeÅË @qòy÷ô—áï´¨u¸|Ó1([~Éðo$rƒÅCÊt^ð†ýõ[ú‡ ?§Þ=Þøçg|òù홲ùka "W<)äØ>ö>^ÄÁ' ëçÍü‰ùZñøJÀˆ²žw—O÷÷‚À‰øœ¨ÿæA¦ªüiòu&ô¶½_nŸçü˜×|c¾?bà™³©y[?â‡á[îoœ&—]ˆÒÿ’¯MŸÙ`žË8“Ïí|òç•8å”·óM—ôw—϶٠üÉí»× ¬×f Ï„ý¦äñ®0>c=é”Êþ!ûù»ÿ¤SøIþf0w,ÿL§ò廈 ÉékD!¯DÙ„"–en®0W¤Ç·-¦Ï#$wh¦JÙ<,Ãuμ4¢ˆ›qÏÐ0}Z\œ°2¹E° ÓÏ÷²e0sß?\¡ŸË˜yâa‘ʈ©ps™M™ë  \7™ß̨Ôox”Í_—ÇÅúh˜SÞ<”ûimy0¡ÈhËãcÿ­z„È­S13BP÷Š(Ôí+"ŽÉmE({?¯ˆÆñ¯{E¼˜ïRàú,‚*bÆ~Ä«ö;¶ˆúb”¾;SýŒÐB(ŒàÒ¯!Fø'•ñ…ˆÔo³GP̆艰ږ1„ÀÒ= !8Ë8§ä¶e !@uG,„}e,!JsG80̶eBûh»÷º‡ýfDRÂê»Ð>æˆDMŸÓ*¯V² ›”ÖE%NŠy/òÛûc£ð½oÿÞ›Fér ÿn¶i@ò+¯D›²ÝÔ¦¡?¹|Y›}Π¶ ýkoÛÀ¿<ð¨M“Þ´iØÝ3GS›&µi°#kS0‚ö½h¡¶ ‹6ŸÚ6,rZ SÛæjC «^,©msÊl£m3ÛÉÞÆA|ô24áµµ=X¯Í[ÛÃÛ<­ŒŒ‰‘¡VÛ#YÄZm™Ålsxˆr™EµÙæðÀ¸ÚTìÞÖp’½ÚÞFJWï›Ùf[M´;Û íØèjcxۉѧ{›Èª—«ám'FoK¨ŒÝ«Î(sC6{Š2x/šÌ"ál{¨lÞ³Ú *³Ãm ­–ªù£-¢²<ÜüÞËøìFúûÆ#W›Dm€&àUŸÊÞ6i†PV›Dm…ª¢;e–ùgÛ–­ˆKù©mÂïëÖ6Ñx%Úl˜ˆm¤ÙFÑúŠŠ~lS°H?Û(Úi¶ïÏòËj£HYm„Rì~Æj£HŸ,’µMßÀË)Úz,϶ŠÎ ÇìûÑy ê~Þ¬uúy[DZmÙG° {â5‹ºÙgжˆì?Å~äOImç(ÁðÚÆÌ¢ìœOM’hÃF“$pQ“„×@šâ‡=&š$YöM’m5I<›PV“!šjš@v ¡& Ôfëe<ik2¨i¶z]ñ3M À¶ál:¨‰‚c;£mΦÉlBtˇ« Ñí~Øj:àyé[Óß×­É ñëíצ2A¿·¦׳7´Þm|ÈíÞšWqÿ¦€}Û›ÙÔ™MÈyo:°»°74‡‰õô-žÙzóÖdÐ~®¡kÔϹ5ÔtÂ1z~éÅ›ÑTP iȯÍtÃ[«‰ ¦Ô}GÇ›"ãô¢§á—½‰ ¦ÈˆkÕú¡«‰ ¦‰¡7E¼ ‚4fþB|Æó›M5EÆ­ó%þ+[SAŸ·ùƒXM†'Äaˆà #¼â Ìð¶3«š•`ˆ¡S÷Nî&»^òç XàôÈu*~u[‰ ;ÕV(·tƒôû]—Eúõþ•<÷³ÓfWƒA„¼“4 ‚œÄW‹Í‹t s铤 Â’ô©¾¶+Qtš²“~Abp»l”WhÒFY<ɤæ[[œêÚI¿ 9¹œ_¹³+‹ÔŠ„ŸŠAÂÏ •"áýÊ}ÿÚ¾)?SP§˜ß)Ìå¤l®G”e´%Ss?—±óµ_Qž!lF'l»¾è„;IebÏôƒØ¯œåò>´Ëìóün·—yÞ¢l—b³ª´“f'á—°Eò…/û^˜ö'RÞƒ²áàìJY©à”Fªƒbˆ”#dhþaúY¤ù´à$É"å-ûH^Ácy’æÓrÕ"ÉéË“$3÷í$YãÕ¸R<«OʨõÁmÆ0ß«ÆX^ç‰×$ÊB0ƒ—¡}ËF"ERkõh9I©Z¢É£ë"•AR/i4Hê- @ÙwÑℨª½…$j9"€{$?ë’Ïk#‘òpLs»ÇV³èI"1°L¿´Ì–óv‰Ð"NÛ,à)‚™FŸtüêÄ O+š~?Ý9ê,õò΂}œVñ†b¿=ÆæÕ9ŠÌ5,|¿ZæªÛ-nòÚ˜L0?0Ÿ”$—­Ýünr³³z $¸¹‘å:vnú6¢¸o–;W+Ï oa±þb.¹m˜6Wn–ELÆðï#·z \¹yX'wår}OˆvI&6žX`´§+a±¾ëV6¼â‚аáuV´›-ë{aÅKñ4°!oÛú}â5¿°"ov<÷ м¢›„•ÚŽE„p #°‘aа¶qO,X6ý X©ø…ˆ8'‹2‹x²`….¶Üvë~²ãËj"§µIõ~eWSôkžìXZ,Vq|5ÄMÆ<®ßð[÷¶˜°îm…£…p5^‹«:Å6­ÒO¹ÅµYABä²¢ª‹×V“/#©ÝµŒë1„L™å ÉÅpjyf@5‹‰!U΂AAö‹lbXx_ ÝY-Ž@í}Úæƒ«-.j:ëª#¬Í/¶ÝeûЧÕL†Ë‹oûÅ@}…Ëâ7ŒÃÕ|¥©‚¢õ°pZ·õã5¿H—Ý 'Ö~KöVè£$ßï»yBÅüüÕâʸ7¥-YW›uâ\AÚ,‚+¼%Ÿ}ãXO9gsÆãyû÷îN†û¡nñ°»íäîWÿÈ~÷ò5!ã endstream endobj 54 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/в¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGù͋љ˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñî§ŒOŸ~÷„óû endstream endobj 38 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpYz7VED/Rbuild110a3546435c8/ks/vignettes/kde-011.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 55 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 56 0 R/F3 57 0 R>> /ExtGState << >>/ColorSpace << /sRGB 58 0 R >>>> /Length 4525 /Filter /FlateDecode >> stream xœ•Zϯ&·q¼¿¿â;&@4þ&¯ì ÙEr0|r‚žG‡äÏOUw5‡o%­¥Ã~‹~3Ã!›ÝÕÕ5L¯?¼Òë»×_ßþÿ~ûùí7?üë?þöõçÞîë¾ï×ùûßÿòö›ßÜÿù¿^i\ºà?n甯•_eõ«Ì×ç÷×ß}úößþþõù»·ß}¶ÑŸõxþâñ쫬Wš×Èöôÿýñ^éO?5ÀñÔ7Ç`|Ç8÷•Ö!ï°ÈW[×]_£\µ¾JÍ|U^ëjùõ?ÿùú÷×_ä‚o?ÁsvŒ»îŒßQÓëÓ·ÿŒ¿Žöúß·?þéu¿þã-½þ€ß½%{ý?½aå YO½Þi–ô*7š¯ïßV»Æ ³^9»y˜ãêf½*Ì”®”h®¥¤zµæ&Ÿ…‹Öò›ñÂ’ÓU‹Ï•\1gš6’ûÕl¨É‡JžWá{æÓ0t^>„ qšo®f7>G{t³çuãþïrô„)Nls©×jfç+cüÒ®|›ç±’ÒéÚxÑ„=¯Þd—âö Ëuãz…‹Ù˜~5{Úû æ‹ñ+æg÷ãÅ¥Ùõåö¼Òmv±ùãÆÉùÚkhÃwåxmWOfß6¿Š}¾ÅÖ×ÒÕ籞V®äëmnc~ÃüQ–ÙÃíðW¿¯Î÷Ýtm<ßÿ÷Ê•<®jÏw<ì^ê“A{ØûFòë)lø;=›ŸFã Çç7º]XIc]•±Óèf·ÏÐJóëyµËn÷‰ivNl*í•ܶ÷¯ÛüqM»3°K\/W®O$Fº§]µ÷#\ó°¬ñýB¼¦ óµùãFäsÆ0“v¾Óµ l„E5Ûá¶—1‘ñ&@)™Ý¸Ÿy¶kúõÁ|Ï\f7{ òLîâ6>O[íÌxÊpk÷ëÅæC7Ûûï¼^ÂîLŒ<äÏŒA~ä¾®aóÏ772÷Î4¢¿´‹Í`È=ÓÍnc Úu˜=ø¢ (“=¹ß¹M.6¯Óî×mã#0‘(¹U¢mlÆkùj¶:aaŸm~HŒŒù#-’‡Áu2L`6;¯7ÇŒüÊÅìló©@a·Ùûªûv¶õ5Â)lƒiÚÙü…×N[/&ŠÄÈe8~paœ§Ydãíbïë¾ÿ¥èzoL [vØHÌùɾeàŸáWF>5Ü7Vó7r˜í냎̀%÷/’sc)q{Юzÿ¬>^QüLKDns·õLæ¯Íe”‹ËÆcÀu„åC^h´“Í—õ þ@˜™¿˜h…åSxÅDd¼ ¿‡_"†­íwA|ãÁLofëýÉ÷Ï€aXµü!°¸˜66?šdöäüJÎæ¦Y6»™††§ŽÂLSó·!Ÿ¿56øÎ4÷ùXÿ x`xV°±vƒ {7žvu|-äOZöíÉzc0dóg!iàCóA}ÀÂh[=)ˆßŽñ¦â§ÔÊJ3{<Äs¦}{~$Béà<Â+ÚƒÈÊl&Æ'ɰù#¾í¬÷¡žt¬· ÓnôW"ØøHÜTÌnæOÄ;€ŽeÄׇć£SWüÔ>Ï?Û|x#æÇa–l,<µæõ¯ ˜ ÓnwNÄÊ ÍõˆïDz–7‰o,›V &>y=ÆaÛÖ·? =iÇï ä|°m­œ×‹ãmäOâ¶/ú‘òŒù‹ñ­ù!ߺùõÀŸ²ò;Ö› fÜ?ÜŸ¤úþ#Ý´øh?Iõ7öƒùÒ´_È_Ò«·±Ÿ(ÚÏä¶êmÄòÁòµÿ°Ò,Oñ…|hëˆÏ[üëSd\òŒyŒäñ¬@åÑ3šo,GƧ»;Ã"BŒs†Ž$v'¢pE}=Ä;Â. í!=d;FDÃ…íA!mGXGÄ”Œ8Ñ6F³wDJÛ2BˆËµg©¥‡c‡…Ð$ŽÝ<bD„áI$-BYZ¹žˆ(U„àqDîÆ€vª¢X„zŸyaG°*#|y³ ·3 #‚¸þÈU4Qg ˆ(8vgK/3(2P’DÔ+ì4ŸŒe)ùÉpVXÆK €*0‰¨W¬jÔg#HF!aTÁI<ÝÕ*âF(–j"P š‰çmó)¾ß€VªÛƒ˜b´}¿áèy ®JB˜Ã̾1¡ÅpÒœÞаÁáÅÒJÞd¯H»Bˆq%„±—¬tí #ÆfÄÓ¯;ˆ ŒjC0F2\·­TóîΠy㽞Š öžÎ`ƒñ2!^Ø9s*ê(°Ð^Ÿ /Æm à`èÖP#õ­éabü9E‡ä‰!rý›±¨Ã £q†ÏRV¤…vòŽÇ¨·É;4¯8›a©¢mñ¢ŽÉˆ«ÙÈ/2à`pê¸6íèÈHd-?™(©=ŒP]FGçþðŽÏÂÔ;BghÁ8Õ12Ì¡2‘ÉP‚±ªÃ¤]Ž”idNàh#VÇjihë!cï£V‡KÛ;8ÚÙâ+º:fŠ1z%ý _·Á’ÛÕâ):uìŒÁâ/%gàÑa¨Ã'LzŽ@%£Ž„…%ϧƒ‘b`Ž+ÞDÄR†ÀÙ“‡Œ0kaÓÑa¡ü0¢!C60¢ƒ‰ÿw{‡Ü »np™Ê°›Ç ŒÕÒL Çj)–üþúX[¿½>}!|}©ú¸ðÊU""~só×~ü| 3Å)<½‡ìtªJ_ k[U‚›“®YasqóÝ´ù¼îV÷æ”—™û÷—ÌãæÒ?nyÜP€wç Üœà<àÙÏïøWŸfuüì–ýadÜõ³{þÕ‡­á+?µéJ°·z»wÂ,#;_“ÿÃüoxü×jß hJR‘ð[0ÊßF©(jxë¤øMhÿ²ƒz6’ª(,‚ãå¶­2`÷ϧp¼û(|‹ä–5‚rW^‡Oö?9>dfÔ:/š(ꔕ·ì)Ù9‰Ô²ª<2©dhrÅ)Ra÷‡Ì*Y:I† :d›¡£i2R“ç%C³¶§‡Q½C$ª¶c>’¡aß§ìŒZ/™Ùeâ½>ÉÎÀ,—µ‡É$$Cß±? …ùô§dhØ5‹TÞGe’¹ºG¥é+çþ‰”®âM Ik_Ç~‹Ô"³$KO#1;>D‚uNR–5UO<‰D³) º$„¤»œñ)R>$›° Ü2¯Hü¸ÕôQ̃$E°eÝÙÓòc˼j*“ncãHš¶Ì«&½¦ËªÒ¤-óJvF˜x“Ê&ç”y%;×å¢Ùý™W23H«š¦Ê¦ñ‘y%3SÖ²Ç)ëJf.Ëã•l}¥CÖ•Ì RÞ½I´îë‘q%3—M§Mü‘q%3Ãö¦M2s^júÑS&Ý2®dfÖo{›ðS¶•ÌLûhêYè|ýØîÿ–m%3çäù²2¹¹‹ Óš’-ÓJfNCŸ$+“«[| có»‰ ™Ûìþ•¬œRÈÊjúԆ̼E›9\v&S23¿vØx’™ïæùý“âÓ´Jv¾cý( Ö´F,Ùù¾%3JD"=7ŠêÔ:d[‰Pk‹R“ yd[‰V˜Vu™¹1Ù²­D¯)ÿò{›ú-ÛŠÈüýÙºçGÆ•0ÿüžVQB²4{]—]³‹[Æ•,=„Çü¾GQj˺’¥©^º,]M”زnqÑi˸’©a'ÞÔžGÖ• ØoßOÊÔ嶬‹ø¯‡è2b‹ùTSßY—õå”u½Þ<2®žÇ°w<ÏùlY·º¬¼e]ͯÌXŸ¯Ët²˜ÈðȼòÝ2Ãåyåß"‘ŠŸr?dßl}d_íW–ˆcû»ÙWûM›ÏÚ²&6òþÆRDYjË„’9AÁ$k橉ËRï‘ %k–ä;˜§él(™+öe³¸Œ²¡dN8nº¬¹˜êl(Ù“ˆêžì‹À##S¼ÙP2h^ÑH½(ón›²æ– »õml(™¶>Œt>2"…öuȆ’MYÁ²ì~ʈݠû‘ %£bšŽ°@T®ˈxÑÙP²*2RˆlíÓ##JFEFûz›Q­GF”lJD°ë,uí%‹NÉÙùȈ’AWÖ|ªÉ&Œ(Ùa³v«‡Œ(™s QXÊX16‚ªBÞIˆ­ŠJ™ÓdÆ@dUd0²åŠ]öÚ¯ŠŽ àˆ„´ŠBŒ€äd)Ö“\ÆÛH ¹¿oKŒ§¢‰Á¤¨°b@9ü‡…r=[6ƒBbÿ°ï ŒÄ"ɦ²³eD1¸ÜÅP§¯gˆb€°]Vc‚øATÈÚÙPŒ“Ûದ˘¥‰ñ ‚G6”Œ‰muFŽ@fEÞ2¡dͲ>ä`˜8Cí9>|»l(YÓ¿/~oEè-JÖdúAßß2?  -dÍvëà â™ `Ë„’91ŒdÏdŒtË„’9›*fÈšHC—…±p;²¡dN,«/ٜߖ K ÙÐeBÉž}wlÓÖ ?tˆ&Î3axâ!Szú€Ø?¡}Ù–þHœËè§íë›ýØÊüûß":˜:‡tÆšÂÝ`÷ ÈEø?殦ˆ¾›=NE“Ź̓P#Ú»ù¨ª»YA±]gs£­[¡À#z»9ÒÖ#cü –BÍT?CiEê(ô– >BuEsÔûÇñÚ€¾ÍÉæžŸRcEó¤Tšñ…6׳קԛå9ƒÄ3ÛJÝg¶¦}j{ü‰TïÇ™›Hý¡/zÄðy4c­âÐò± èÁ2ý 1ü8sP… æÍ4ÏÌå³9s¨ÛÍØaû¢jãífLPÙ5vxâHÕ€Ú¦3nüôcÐ͘ îÔ#¹~š1A;{¤ø¢e_¢£ }6_*°½ùà~4_*5%JÈp9¾ÀD©bÆG³—hŽÒ–‡JnG³¥æ/δE³‡RáÔ§Þþ…(š-5{ÈY5{Íös—"5Igè¢ÙCsåÔ…gŠŽ/dÑìÁ­ë ¬€N•ºIxJ§š=~(9¨·Í©_·OWO)V³ÇfË›9}AlÑÜÙ™Ž§´«ÙCåñ¥faåÔ2Û™©MD-~†kêŒKPQÓm‹Ê®ªñ–SÅMeD}—1€ïÕ‰R. ž* ÄЕæKÍߌfâ6Ç?Í—š¿)*ËæÏ¨W4_jþ†ö³øw”§ùRóGª¸›Áƒ:–äÔp7[ „ƒŠRUc³ÔõÓ^1Š2ø}{¨/É)+ù)› ¯ò(€O;ÐY•òª5v Y;ºÉ¬È>6:çgÇé'·¸‘d}“[‘yxÌ3nةڇ܊¬Ãã1ÝŽc=äVä|Níx·ÓÏŽŠŒ¯ˆ0’ï3"D¾1‘1öM<"Jä{ ‘[eå¿°[="VÊ£']vJG(ÃyFÅÉw"‹}2Jqë”2d*DI"_@©Ä)ÇÛO½mDsC`ŒSîBàø)Ð\>ž²b&}£ ÄÍIדQƒ±„Ø »~&D¿ïv¹r#¢*DVû¼Étœú»½oò« Uâõ2ý]U8ªÖÇ)\"ú–K‰H›ìª¢–×E¦uÊ“dšþÞäW§¬Ý6ù]1‚ HŒ¡ÙF%ò+ÆQ•<#óAvƒ\/߯`4MͰÉÝ÷A~Å€Z¬¯ùþlò+ÆÔ¦N%ócû!¯Ãêñ¿:Âo²[û‡SºÁغò oŒ£:‚o†òQ:"u?…3H}Þ_*Èïùé—œáþÑûñ?p,`?Ødœ7_¤ßþ‚Ø»Ç endstream endobj 60 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/в¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGù͋љ˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñî§ŒOŸ~÷„óû endstream endobj 39 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpYz7VED/Rbuild110a3546435c8/ks/vignettes/kde-012.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 61 0 R /BBox [0 0 432 432] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 62 0 R/F3 63 0 R>> /ExtGState << >>/ColorSpace << /sRGB 64 0 R >>>> /Length 4608 /Filter /FlateDecode >> stream xœ•ZI¯&G¼¿_ÑG[µWõÕF²Ì–ˆÝò3à瑵|ãñ0æ½É×ݵdåYñúêŠ×·×?_~ƒŸ¿}ùù¿ýåç×x Âuþüáß¿üüËŒ÷ßþåŠý¡þÃåØë£ÝW¾Û#ëíëõõ'o¾øÝõ§¿ÿáÓ+†ë“¿þãû?|÷é7×Û¯^~ñÖ¦û2i¼ôÎx&' ›ï+ŽGOî“ÿ|ý³+~óéõöÛw8¾úì ÿ½Æ x¯Ò»¾êýåêùQÊ•KâTé¾5]ÿúóõûë{éä‹7ÐÉ ãÞ!ág/ñzóůñ×^¯¿|ýÍ®?½Äë+üûö%Úô¿ÂïüèãJ­>ðí«Ér{@#ßAN‘!ß4$cêÔÓ£TÉ\5›MÎÜ!Lj’9þÀ|þ~áFÒˆû6¹Úø£>b7¹?R…Üw”\øýxdÿ ¹¡ø¨66Ò#äLERŽ(ÝåÑmý1?5Ö¡“ Ÿç`ûwëË!>‚ßx¤ËLÅäþhP|hZäÁçXV2ëÉWÆ2š—‚½mÛ”#™1ÍðçÉæÇ4·í'•GÆ÷ø,fɵ›\l¾Ô©ïœ V—Ç#d“›Ÿ¡ŒŸê£›þ2ö‹ý¤ö(¦ß\¨˜Œa’ÍŸÇ£ðûñH¶þm?o›¿T*:§ûQìý‚ƒL&7ÓO1C0Ùæ«iÊÝôU±Ÿr>¯´?“m|˜sŠ&g“[âF8¿Û ¯˜l}0¼m?®ÿõÜ#tÛ±ù{£aP?¥KîíÐß°ƒ§~]ÿƒþrèàC¼cvû…öyƃÜç ùÎ&7[ï¸ý¼§=ÜÁíiè¼`¸ñ´.÷»ì 21ÃmÜ_îJ÷ßöy›"¶=ã?Ùí;š>n(²nûO0ô.¹%Óߦÿ$øïÝ·%.„þרÊæ˜Ë?^·ÿR¦ÿC¶ý'ø'rù¢?Âq ÃÞGÄ‹G¼HpŒÇ¶l=ð?øSjQóÃà?©Ê^S2ý¦Úü|(s}«OÇÁø5s™”ý-U–òmñ©÷¯„½¥Ò9ËÔoiš/ÛA§’yL”+ý1!°ß6–JpÿIðWØ'ÙyóçÃkfÿü0ãýÜÜ^R1ç4fï þ ÿM ~ÅãgNÜ6dlÏešý'øs§d/PEšñ€Šc~€²­ŠåxP“ùGj°çjrµùp0ðß5Iðß0Löùà¿ú€Ú«­§[¦¶c±ù`H0Ĕо‡¿Â±)—3ý²¿ÄQø}æ1A†ÿ ~ofNÙâ-ålûƒãÐ?`&ÝÖÿ¡½@6ÿ¥ã¥f²'•öžtÞtlÚ7ÓíÍßÃ,‹Ë¨šºÉÕì—fíò "þlþO¹r¼âþ•“ÝÆìÙ™Ëfÿ |5‚È.w; {±ÀŽõ0­Ú÷8Xênmñ!Ø™¶m08ÃBu¹2E ‹”a 3®Ÿlù™aÉâCÎ7 ƒaÌõEéNÚò ò%äÙ#Í Ü©Ê6?êŇ ûE"°°lóÑ!üÏåN{b·øiߘfdñ€2ü‘iÀòA¦ãRÎÒ'Љ]çOù.ÓŒùk†ýg¬ñÆâe@HÊÁßt”؊LJÜíŸi-ÉHÌ‘nbëi†—bÒO³Ä9­!0Z|Èf˜Ëv{i–Ϙ–ÍÿsóýC-æo”åêþO™ë‡šc>ß7·Øã1ÌÌù0QÄ1÷¡õáà"ÌÄÏ[ëÍÓþ°1îñ!ûŸþŸ _±ø{*‡þ`ÆsµøDÙÇÓy¤àx&W[8aÖÇyÆ©?œÿ ÜdϲX§ý¸}1?Ɉ_”ë²O¬?*?Lû†?Y<¢lã…iÿæ/„¡nOî_‘ùýÞþH9òý73bØ[Å€lØ(y#ÔÝöŒŠX”KßÍ€nÙÏvPvDŒˆæ!Š˜Ü±gÐ~ûޤE\aC\ŠÈ”GÞ›õ ‹áާÆá)´3O(ÞÊ#îT¡lˆ]†ÀØ@u N‹PF2 ŠS>,ŠÌ*)Y SqHÛb™ñZÛM™5-žïS¿ÓC8Àò Îîíqs=øÌBõˆ5=4USäò`Ê´ØéñÜ2ÆŠKVáûôÀq8¾E,E¤9?’ô;fDS†gÄ3§Ìó›Qˆ  ÁA¦}Ì*DGR†e¨¸w"!0öóÏVT¯ˆ-DC¹ÛøŽ€â˜Æ3ÐÊBTŒíûd¥èÊBhÌ(Í›•Ï+ãáQvDFDXwÆšˆ1¨eªÅBV†›ˆ3¨‚L¬ï!'‚%Ðõñƒ?ŸûÆkî?(‰g†žˆzF  НŒ?z¬ª˜„è£ý¢\8ñBªl›^X…AâµW$TK´ç€î†À„hTáPiV@Á“Ïç”!,«ˆÕYñO¦ŠË›ïš!<[Ó÷B„ªèxÌÎ(xÅ'à¼*BšIð qp# ‘ª‚LÓŸYqæ°®*P“½¢Žx«ôA{¯1«Â¥Ü½âµÐŸ¨°BöõL®Š9á5C̪¨é&ÍæÏæˆ Ñ«"'wý{ÅN7óŠRƬTñSN6¿3æ¶ÎPxE7+ 1”og ŒÑ°0ä Fö÷U±L†j4LYŸŸ—:+vÈ´÷Y!‘¡ázfEõÌy†«æÁÌpõz½ž3Þw×õæêî]šÊ©»'ªí³ÀC¶þNB Œ_ UvÁ*¢ìäÁÞåÿ7Fü—ŲaN°qmÏÏ]jžø˜Ñ}ðzI<^°¨Ö 2òÿ¹ÏÎ5­%/| =úA:ŽPŒp¢‚?+ïc2ßÿm„¹×§ÿ˜±±r>þ[2sÐäþöÿ˜7#4ösÑkÃNº"Å!’½Š‚%Ô²çwžôréxá‚ø`>—x¼À˜ Ÿž/H<_¸Ó9ƒÄã­i¾p.ñ#Øçb\€A»?pàü8zûéÿà×LZˆ?yäþx‚þÉ3ÿàÇV·å÷ºì¥׎ǀ)æž&HÿSü4þÿ²õ7`'“LdœÃ(ÁÖ+W4 °¯/É'ûEÍ\¶ØzÊñ`ë“çþÅÖk<’,Q2b³÷¤YÇÁÖ+—ÁÔ­Z¥Löf±÷ÙʶÍÖ+⤲çN£57{¯\z—ù}5vc±÷HÒdSÛ©\L’ÓÞG’66²£b³ÉžÚ|%;{:ÙU±Û¡)×+b6+¶›Ø®mlAR§N¶;·ƒÝ»M’'n,“'ö›ìvŒÊ½b·!;ViƒI{'¶›çÝ7ÖÉD¹»¼Ö+¶;D£Ü™û÷~Åvë:›> kn} Z7eêWØr›Ø‘Øo‡°åÄf£ö$–[牃&–]ç/lÊÕ±j4vvÙ‹°ì0HçX—ØrÙÛmémÛ£°1¶™\Æö-6XúìNMlÝ'ÛIÐ?6YؼM¶r­›Ìä’6YX¿ÍÚOìßb“YÄçƒ=Ví€È¿j Ú÷b“a(ˆ½›=&v+{¬Ú„XrH&^ì1ŠbáÅSÎ;¬ïáVa²ëö|²ÃZÜÔk±`lÍÂÞoÖ‰äbůËBHýÊÞëÃÑúDçª&èèé¨ÈWÏ;ºmæz­«zÈC¦ê ý6 lFƒl´®j¢ˆœÕ4æƒêšüñBëªH›÷£z(ª^)³?±ÐºªüÙ=ÂË€ÎUMÔàÕêŒXuzL6ÓÝè\lÁP5AYè\´®ê`x¿a¢sEdXdʪ.Úa±Ïƒ'ú|Æ€¢ÞÚ¤ÌzðÇHòÝ4õŒ$#Ïh» ÎëðÆx› GëØÅL´³©¬)÷jMÂ*¹Ý:v‰½‰±Jêlµå.¡5^¬^ÒͦIvmZm Öþ7gT[²“ŒçˆjS­wÙÛô's,µ)X{™¬¶‹|Om‰8I µ% ûøÃI‰µ#µ)bôcžmЍ¶$#r=5¢¶E¢­Õ¦j³1Æ“æWÛ‚ÝwÞ¬È^4¿Y˜4u0ûÚ´¿·1 LiÑþѦùÉƃæ÷¶I =oF /š?šÿoš?'‘&­ïm#­üýÛiúIó£h¾LmÊÝÛ"Öß´¾hZÈNÓ‹ÆmÝ9ûÓ¡´¿hßV§\Œt]´¿hã–çs§•ò-™´ùjˆ– .g#ÑV[@46¢—%NÊÔçj HG[@´8ÂŽ%Š\­M¸ÛÕØ Ý¨æý»-@´m7I’Ìx—¶Öcï϶€Ö›ÓlËX¿m·´_¶ýã¡àËá4»%òÝ&>ÓܯôŸƳ÷/wÛ@ç—âl›‰DŽjÓÈRP›EöD3ª²?‹Ýf½BöýÊþùÜÛtÁÚ"ë}÷Žîío”å_Nú“Û‰Û)/ÒŸ$tš¤õm×l6 .Ò?MÒ^¤?£û½ã £¿Ù;U«I®&@Ò5Iú' œ”›‘Ü‹DW Û²){X_¤ºšÌ^CñÓÞŸ$;Ò8z“ìp¤ûhCñ9IóE²7»CµIvŽ×R]ó!m–Ù”¨'ÉŽõ±M¹Hv"¾xìSŠåÒ5Ÿ°Zﺂ6¬A·×'¬6‚®`òÅ|ìGØ ²°Ja`ÞXAØWª&–1,1±‚°ÍteSXféWX†Ûv 0·ÏçÉ¢¼º>mäÕngÚ%3¯6ß—ëÛõ™™‡ÿ°?ØÎÛõ¢õŸìXZ# p«Ëf AÎçÕ©aj}œˆÒï#lD©SªÉ©Sžjì¢Ô)³‡Ô&„-«©íáתŠî„:+Þ/s«õŠóÏŠÂ*¸YÑ©âàý̳"aaç´W0¼ÍrV8ETû¬€ÊPÃþ|>*BULE×f…Uª*RÈ€î»bäÅÕó"Y¶ƒ»bÔ÷[ÃÇóµ*FÍÇŠ!ë#‘詊íoU”ÚOu=÷›£ïïÍ:l'ø:9’]£èÄr+¤Íó*󸟮ºM‹ð”«?k˜Q¼Ó9k˜'‹ôùeÀl ÷õz^á{ýà=ÜGÉÕj3¬ó]?øÍËü&ôÞ endstream endobj 66 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§ˆ&\S†“‹áÏÏMç‹ÅÌ07#â1Ø™YárfÏüYym²";Ø8980m-m¾(Ô]ü›’÷v–^„îDøÃöW~™ °¦eµÙú‡mi]ëP»ý‡Í`/в¾u}qº|^RÄâ,g+«ÜÜ\KŸk)/èïúŸC_|ÏR¾Ýïåaxó“8’t1C^7nfz¦DÄÈÎâpù 柇øþuü$¾ˆ/”ED˦L L–µ[Ȉ™B†@øŸšøÃþ¤Ù¹–‰ÚøЖX¥!@~(* {d+Ðï} ÆGù͋љ˜ûÏ‚þ}W¸LþÈ$ŽcGD2¸QÎìšüZ4 E@ê@èÀ¶À¸àA(ˆq`1à‚D €µ ”‚­`'¨u 4ƒ6ptcà48.Ë`ÜR0ž€)ð Ì@„…ÈR‡t CȲ…XäCP”%CBH@ë R¨ª†ê¡fè[è(tº C· Qhúz#0 ¦ÁZ°l³`O8Ž„ÁÉð28.‚·À•p|î„O×àX ?§€:¢‹0ÂFB‘x$ !«¤i@Ú¤¹ŠH‘§È[EE1PL” Ê…⢖¡V¡6£ªQP¨>ÔUÔ(j õMFk¢ÍÑÎèt,:‹.FW ›Ðè³èô8úƒ¡cŒ1ŽL&³³³ÓŽ9…ÆŒa¦±X¬:ÖëŠ År°bl1¶ {{{;Ž}ƒ#âtp¶8_\¡8áú"ãEy‹.,ÖXœ¾øøÅ%œ%Gщ1‰-‰ï9¡œÎôÒ€¥µK§¸lî.îžoo’ïÊ/çO$¹&•'=JvMÞž<™âžR‘òTÀT ž§ú§Ö¥¾N MÛŸö)=&½=—‘˜qTH¦ û2µ3ó2‡³Ì³Š³¤Ëœ—í\6% 5eCÙ‹²»Å4ÙÏÔ€ÄD²^2šã–S“ó&7:÷Hžrž0o`¹ÙòMË'ò}ó¿^ZÁ]Ñ[ [°¶`t¥çÊúUЪ¥«zWë¯.Z=¾Æo͵„µik(´.,/|¹.f]O‘VÑš¢±õ~ë[‹ŠEÅ76¸l¨ÛˆÚ(Ø8¸iMKx%K­K+Jßoæn¾ø•ÍW•_}Ú’´e°Ì¡lÏVÌVáÖëÛÜ·(W.Ï/Û²½scGÉŽ—;—ì¼PaWQ·‹°K²KZ\Ù]ePµµê}uJõHWM{­fí¦Ú×»y»¯ìñØÓV§UWZ÷n¯`ïÍz¿úΣ†Š}˜}9û6F7öÍúº¹I£©´éÃ~á~éˆ}ÍŽÍÍ-š-e­p«¤uò`ÂÁËßxÓÝÆl«o§·—‡$‡›øíõÃA‡{°Ž´}gø]mµ£¤ê\Þ9Õ•Ò%íŽë>x´·Ç¥§ã{Ëï÷Ó=Vs\åx٠‰¢ŸN柜>•uêééäÓc½Kz=s­/¼oðlÐÙóç|Ïé÷ì?yÞõü± ÎŽ^d]ìºäp©sÀ~ ãû:;‡‡º/;]îž7|âŠû•ÓW½¯ž»píÒÈü‘áëQ×oÞH¸!½É»ùèVú­ç·snÏÜYs}·äžÒ½Šûš÷~4ý±]ê =>ê=:ð`Áƒ;cܱ'?eÿô~¼è!ùaÅ„ÎDó#ÛGÇ&}'/?^øxüIÖ“™§Å?+ÿ\ûÌäÙw¿xü20;5þ\ôüÓ¯›_¨¿ØÿÒîeïtØôýW¯f^—¼Qsà-ëmÿ»˜w3¹ï±ï+?˜~èùôñî§ŒOŸ~÷„óû endstream endobj 69 0 obj << /Length 778 /Filter /FlateDecode >> stream xÚ}TMo1½çWì©òJÅøÛëܵ9ÐU@UM[0ÄvÑî&(ùõYéÉöxÆóÞxÞ\O/†7Öd\Q©ŒÈ¦‹Ì2Z“YÃ(3:›Î³_dì¾ñÕÌ·ùïéhx#‹Œ3ê˜ãÁ²€øä{]ç¢ »MY½ÏR2rEãz›Ö²šÃF8rõü\®CNüî˜f‚1 +§1Ÿµ¯óqWÐBkÈÛ'l7—ù@iCV¾É9©üÞÑ ìu®9©»ûP-Ñ$ÉÆw÷`ddÞö1Ð"ÒxAä&ϼ°Ô)Iµr1)‚äÎõ ÏÕDÂÖº=D`¦'c$(ÈP–³\X²ÊB¹„³WX;ß´¡®¢APA# ÅiQPjA57ñÝuµ„ò ãÈ’\¾äЬz¶L’O¾é˃~s_µ¡Ã|OÑàÛ.lÊ.ÌŽ¦¾"¸YE\/Á¡5a*¨Üt{÷rýÔ†O’,ê&š7HõaÝÁÃá±gÜ„²ó=­X²ãÚÎË®ÄHFB w¶”£z6ÕxT?4 °À•‘I„€× ;a9©Ý®lÐäÏýWŒ*U¼|^Ì}䣩Uvï!8>\ô¥»#8ÕâÒš%ßÇŸ£¯v§¾Jí}ï»n{9îv;øONþ¶@ª´n–ÃGÁ‡Ù³< F¥G< 4 .5Rí$lꪾcRöbädDcÿMh<§~7o‹öVÊ}²É©ø4ùÅÇA|Ѫ¸>éìtáÐ0y"¥É¶‡}3ø‘:s].ò¯~¡ÿÄ›ºY%ÞÂ*¥9Ð-ú@—c,ä ²e–|CsÚ'=ý‡9æîÐ.¨±ö(<¥ÉMn y¨fÙË$ĹJºBãéÌzíw‹oÄæwÑ8ª+ŸL_¤ßl@i©´êu!ΤØÂRaä±·8¥Šý”:?¢8Ž(1à.5ÊaÆ" ú]´þIs¯®Aý)t¶kÿ‹0d‡ÿκ0CF ï.>N/þt5§ endstream endobj 92 0 obj << /Length1 1369 /Length2 6104 /Length3 0 /Length 7037 /Filter /FlateDecode >> stream xÚtTTíÚ6 HK Èi†îni˜†˜™¡[DN•%¥CBJ@@éFJ@éô}}Ï9ïùÿµ¾oÍZ{ösß×]Ï}]›Y×€W‚°…ª àh^> ¨­`*@ !>H͆vþm'd3†"Q0\â?ŠH(±)Ñ 6hx¸B€€ˆ„€¨‚@âH @ ì ƒÚ|€E²)"Ü|0G4¦Î߯‡' ..Êó;w…"av`8  F;B]1íÀ.€ÂEûü#‡”#í&ÁÏïååÅvEñ!2œ<€ íèCQP¤'üлBÿŒÆGÈ:ÂP9 öh/0 ` .0;(… ñ€C HS0P׸Aáµþð.àøWº?Ñ¿Á࿃ÁvvW70Üwìa.PàŠÚÍ€á_@° ‰{‚a.`[ àwë`@E^c&ü3Ê sC£øP0—_3òÿJƒ¹fe8Dáê …£Q„¿úS‚!¡v˜{÷áÿ³\g8 î÷÷ɇØÿâáÆo‡¹{@Õ•þ`0&ÂÛ hà> îÔÛΑÿWC7èo§À/3f†?7„`³‡bþýP`O(€Fz@üþÓñÏ¡€Ù¡[¨ Nøïì3Ôþ¯3fÿH˜7`ÂÐOýúýëÍÃ0îâóoøïóëª)>PÓåþ3ò¿œ oÀ÷¾À+xˆ¢˜—€æÑÃþôñ±êp{ þW»˜{ú»eÏ?àø#NàŸ¹tæBŽÝtd‡yüŸéþ;äÿÇò_YþW¢ÿwG*..¿ýþ?Øæâóa®£mF ðÿ†š@ÿ’®Âòß>u4£y¸ƒË¿®†RyC!º0´ã_tùËnôKh.08T‚ýú´¼ Ðù0ê²sÆ|>PNþvA1âùgIe¸òKe‚˜ ƒ‘H°!fɘÓ}ÀO#GÔû7‹~>8 0Ãö$ᯊ üj¿L„ÿHkçDbÄõ{ñ˜šŸ+ õ†ÚNŽ#ì$;U>n:)—§÷â]e[5Iãäõ›D¾÷8»‰—ÂYöòÑ<òH>¥§lfY™ãPnŠéÒo«¾/¼!I¯ñÜÿÂ:Axµ‘ðËЭ®Á¼-ùªNF^C¹5ÿKwãçkõØ-lÙîb7us)O¼:T½«:‹¦ûžŒ¯ê­•‰h]ðÆE[„äf˱Í£e¹æeÄç¢Øó&ý|x4J‘5ø“I#›0`;F赟ق`ìé˜ïl‰¡ ª•î.-ãµCоav?…õg4~o¢ç¥'2EŸÆuO=îHÓ4¦c Ô¥E>.ä•~¬ÕBy·5ËËîÒÅØnP`¿ÇT…ïÆV jŸTÊÚS`Á :½de4H4ņp¿o»Óº½cÔÂåŸG+!óuÀõΑ4åRsF ÈÅé þbð°ªnŠ=Öë”΃[#ý9Êú=¨˜êÝpaVó\91 ¢Uy“zïÖ;.ÔÚ:ÉÈPˆ!Dü³x!/ŠìQØ”yl8áî3‡JÆ {R<‹õ<Ðè·;`ýA+rŽçZ4N䬟˜]½æY់wý£~R|ÓÂKN²ò+6Í·ùy&ÜɧšâEY[²2í¯{"¨°é0SíµoëܪRžID¢S’oï$ ‘ewaáÙ.flwÀY‹ó;=$ç:Ûù–J7LDÉc ·rSðMcîy û½ø[: ›JK½¨Ç¡ë7—fëÛE\èà [ñ'®–M{ßy¶ø§ ^[<¼+rÛœ;ÄH9’µë]´v΃m’ÁJ>¥‹E­²¥"Z‹C®E¹_¡{ùo؇RDÞ<ÞL¹Ù†ËÍ‘rðÜ9ئ’ViþŠ¿²'Å™ƒ¢'16Î9Ì­ì&ÚæÈjCáÃY8íÒX……ZHàæCJë£ÂO‰ÕWÔøç”vl­¤TÌS„ž¨›¤iÅÝǽֆü–³EžP” ’ȵ§œ°ëñ~p{ôõÞSöÈtmöJÑjòzògh³ùóëHÆ ÿû)Úùïî³T=âj.#maÜžÚ”*HQŠ®âiÔÚ·wâNNØîm~ßò ûh+«'ƒÅÔ~ìÈz­xÞÃ\n*`q}’e<”°¢ž ÌEÀ8µ¬Y®uzòµ¡½ëñ© v•ۺ䖽 €„óñB5ˆ¾V}¹M†EEé]½e[ïÑ.KÿæÉpEb7À“¾k\ÊðmM¹X•·# j$Q½3JœÛ%eçRص;X{Ç×@Q­r­B+¦@äzƳ>tg+“ÎŒåiv·‹Ãçk½ï-rŽ_GNâŠÝÛm™_UËÊn´ÁîˆÅk º×QËIŽ$lXª‰­ÊRí/Ãy{Í,ð…â×TÂ(#¾ë$ÊQŠ‹BY¸ãŽPø³Žü%eÊ„æÓÅ sï¾~ïRx»¢;k «!¬ìvb2Áÿ¶Ê*«i3Õ“™ÓA6yIÿ3`ÃAr %¡’´$I.ï}´Ì¸ZYpIÛÖùéná´+\¼O ýŒ7üx5ûœºÐœ,»ä['Éâ¯h'Q"|¡NíYŸ%ß7×xeÍöêäÖÎô©t`2û+C[¯w·§=#ŒUÆ,ä+iµ¥¸ d/Å®3ÝÀI¿¾»Ëu£¥‹ñ©Žúm=Ï|qa߯ŸO…'Ž]CH¸'£RÆæ¼­åÇÉL¼ŸR»ýF«‡_Nç8òl¦FÕ_æÑNûÚ á‚–‰>òÒ—û5•œX«íÆGY…›7C)]>.-‘¤Æ2^Gë|VZÏ"nÕËÈ´HC³ò…ô’†xŸßù Ï+¤ÍŠe=ÔÀ7@øó³ô¢çaè½û¥3tB1ÉÒL±šè¾4\§) m÷7×¾X!s?D=†òqQ²î؛̱ÔчyÙÙø“Aw©h¸:щ¡c7Ø0éj±Õrü8–£”Ø‚2R?‘Þµ<M'ÓöÅ}žÑ8±xX¤³XÕ'˜´NÂA6¯·ÌðÎå–Ë©E$`éÜ]*fL˜~‚½»'¿+-=x‚<ŒW¬‚ޏV]½9 Ç% ?¤-ºzË™O¯²?(ÔfÚŠÕ›’Œ§k–¿¯Ž~Kñ%‰Fè»Åù¹[Hœ¸2+°ÿœ¨}"î9ž”d„§ÌjTäõª}n«Û@ÿ^©s“¾XƒhÍÁQÄˆÚ (`Õò¸êe†ÎH.ƒ\c¹#Åz†M™¹‚JxóÛgW#†—Më”Õ Ðà&®»åLp„÷¢&@vؾo|·Ù<×¼-ú2,ÍÞH’Nø**¦2“Nï†Sý~âÙ㪄ÈáL©óˆ¹f6F£ÛT­MŽO_6Îp|Î×ë'²ªZ*Bsóy€Åê…ô÷h͇"¬-dcMП¤ºÃø%ëcˆkXÖô޺ζ—f½à,=*¤z¡kì6 nND$¿¢ÄòY:ɩйgÑ,7å²=¨!ë (»»O&?>¶î]"mVʪe_ˆ‹/ç½kò{Û“ÿ‚¢ØŒ¹ÚÚ†9å2ǵÛèŽýèmæ Ãd×È4ñu-El‡B ¸­AÄ™VC‡{ÜíäȯN÷­:įéèWi~}+ ¥^ø¯dÛ^,§|Ø•âÑzvgúzp×»¨e)w´‚…dŽ|Á ®åÅ«W¸w*éíi{¼||üÕ¨rü6ŸÞ’rié_igÄM#Æj&ßAMOÔY µnê8“QˆêHôLÔ ïq"ͺ"¿“xVš!BÍ–i|5ïc`MáÍ$.¹Q”ì4!ì$bÐ «Ú¶sÉI#~S¾øÃ‚™Ô©'wQï2-ŕƔîÆi1œ8ªg §>Q|ÞŠ#“]YìZŒ R&-ÔWoG ™2zÏzi4è‰3Z%Ö¦•dioö‰x/~Ü&oã{¶Íx?ÆTŸØ†úì}ž¸M©Ø5i\ÔRÇ” …»Ö£>ë±+¾ž’ìDOðæ‚Ájª¤¨[Ÿ³äµe-b›´s^rû*iM°E„š>IW?‹Júæ7S²òpÿŸÁ]ÀU=rÇbçOac¼Ñ¼llê M“† ±ïÉAíB”±Øçäc/¬¾$ð~Ó™äHc KâÛO0SƒóFÓKc{…æË2ßì{ivM¸ØÙ©ÙÕ!ë—ñ–œ»ÙC˜ñÒŽ?[m}°˜ûNl›‚ëF-¯èñ­Lƒç‚»—O¸¾w*'D¸ß‘l‰kä_#mÙ4/ Ö"ß!C¼k£¯3$þ1ƒ_˜£ò4¹RFº‰ìû±½ÚÈ»ñ™YÕ(vVÓ0â$ía^=ª×ßw®lô‘™¶ó“Ó\<ŠÒKùçþIIÙä¶HC±ÉD ÑJ=1’oURÄlÞ.ì!ôù„ãoÉv—¤TªsçY-ê-yV‚F+äPŽ‚›.*_¹:#Š^LÆL}±É#ªp3e{óÿuŠöÇ2Ç?¾e+uJ/ª-}úrD”Õ3¯¹ýhˆ¬v¸ÈKPxŽl"¬Âo»£hÎYvŠç œŸ#´~9«ÊŽÅï¦ÚÎFšÂ8)ù¶ ´F‚*Më¶¾I¼IuÚà¥ýœšœËàõ`Ið+Ÿmaê²³Ùr¬œTTÞÉ££+_Ÿq—Ïœ?/Y*é½CÍÊÁåÐÖY#WÄHúþ¤`–NjgpÆòWߤ :)òŠãÝÃéÇ*§LÙˆö½ÎP²ÂÌ•uU-RŸ^Ã)ÑRÊÓ»KW­îOM«sO]ÒíK….óAØÎˆ×@2?ßHÎS¶ÄÕÕÝ]ùn5€“-Ãóð%£ÊùLröó,6ìdÕ4 TNÍ<ìX;·ŒØ JŠ@¤›éôÑÅioûš-–ßµŸG¦oÍ_çvã §bîɯ–Mø)€ËÝÙ×s“¶ QìÏ‘<Úº£ÚÑ\þ*8¯«ÓŒU¾:‘6§°jw„ ê®\/CŠÓ¢¯UDðè*Xó‹Ü]ÿI%“·éQ£üº:U7¬†R„¸@š‡+!—3]uœ£×J56´XFîÞ?ûnÝìP¬ˆT¬åÑÌÈ"Ë”ÅhäÌ^K‰ž#>»9#ÀŠÝŽh(~V_GC­Îdj—¡{ü%ðiðjüãÑŠÕëNáKêÆÑdN'Æo¦ÀÔ@˜Ä ásc•]<Â.N9ŸúôÿëÇb¶Ü))¯¥igÙäkÓWNr%ºþ ÎG·‚„_©\m?Ap™^~ø¢ùÆ)?}ßÎÍ2(0]LªUn9¾-M¨ö•Sýª`Fø#¶lqQ½»¾ñVím‘82®Z´t–Å}žnUµÑ=ÙÆÇÀœ7™Œ.8•ødûDZÆÅ›2Êì…õ‹vc‡MëëûHð½XïZëûÒã›õypº/.øg”G7ñ…— ¶ç_|Õ¥^ WêíLô¾Á)†?ºÝŸÄ6žÚ¾¯g)õÑÑg?8nÿBÚýåû:*Žâ%f³éæ1M%úž»¶ãMŽU©\SÚñ´|ëLCjõ$Bj·">b…jĶ_Ó=ILaýt'NŸûÑb1£¡á"6G&'Sí…²¦dš­º•åÇ•AŸfgQ»¿¨{¸×˜ê%Ý—g™wŒŸëä4¹rqq´P´Ý0ƒÕß²Ç!âì Ì&ÂÖ]×­bºþyQ@´pu\³Ú—KY’vþ´·Ÿ–ÜÁ†ºCÙ£/ËÎv‹Æ§&¢ÈÔ$UÛx²<@±°ÿ”©ÿS†CËz¡=;O·Å—ñÝý) þŽíàéjÓ\ê·²"‰ˆÁ‹ªÖWsÁÙã®{IŒGÛ:.Ì£“ÏvõÏ?+5eZò†c¯èîÛsÞÛö¼Ž,nSÖ00oq}-‰±Œ¿‰ÄaŽn÷oKV{[ì„Î脦õ¥o”$8‰š¼Ùd}Q5’xc/y¢·ªçcƒ‚±*GÐl…Æ|E=ßD^JîÝúò ér‘#Ù2Ê ,òè»$†½ª…[9! ”ÖýÃI¬-¦}HÞXl‘qkÏ3B­5Äkå²w]¹êœÕšÌB¦ß¼ÞSaié¯Rø·¥¡ É|ôGv]¼øùòYòim÷°„´I{(ýg+—g¡y…n(jJ+1M´s.þäÝŽ‚)Ñ’M€X FeîÔl×qiâ©ùš€®{ú­·©i+Ž:"L4Š(¶õh†|l‘¨@ÃnI¢‡ÛàNKhã  uÝ œ?e»eÉžZ6½dùÒyÿ[»AµÖ®Ïöœõ•Óɧ4ý‰Z©ÝeàG_ý>’¿®¸æ…^ÛŒãòœæ+êøvÕ}5VÍ,Rsvâ‰Úµ»Nâl:Ü’R/ õLd×?TÎ/uÉOÞÛàf“óñÑ3Øòè%ßM4õ½Êtݘòž8x ÓdŽy‹òˆJ„DUMJ#i+Le÷]šBSúŒƒL >³Ó÷5H㫆×g¶@[Ã'D ޳šõ+1œï¨qƒïOœÝ~šRãòÎvµG_~函%¡îWmø]}&®ÄÎMK(}8çMF¼ðx¨ ›AöÛ3‡J¼š öZ€ßdš’$­O6ääÏ4 žË:Ķ_×Ip²™z]«RŒŒ³#Îðu?A§V+ÅlAš¾Aý fo¿Ä–³ð«}ªüg“’BRϦ7{k¤6S#+œ¨8„ÛYõO{–ûròQ¹kßmÝ©6±-ŽšMƹŒÄÝÑ~‹žç1Øs°Æ·¶¬É…E–nòªöšŽøV­s«ú'âçØ['"-•U¦çÁîÓ/câó] Æ,k´á¯~¦â{r”+UòWLiÙ“mDM@–v'õßJN«¾•k|06$]¼ò±ZŸ~û³Îá•kàÂs›J3ÁèõÅ¡ÿ(Ü!¿y »T=ý“dz;#ºÚ£¶Vlƒ„ThÁíÙ…¹’;ORW3ûÉ,©¹6ñÁ"êFzœŸr°¨í!±ä×ÄóZREIêdK%¥ Ð,Þ+â±<½ sißà«Þ׺9\šŠFî‰ÎÙXÉqý^Á_~‚BãàXQOú=0:^íôZ}ú¹ÝÌtÒ¶üëö%–ÛqGmøtå·ã>‚¸t"Œf~ÊÞã20 ‡ˆË$Üùò„†'/•Ô7ÔåÊßwÏû¼ïhpûqÅYÓXuÕƒ÷ÕÆMDP¡ü`X´©¦:ô©Úù@ì ág^£-®Ã+¼ òÏc²þ$j±;"-\÷ö+d%/Uéý¬.É£cù‹XXÜtÛÑOoœˆƒIÆÒCC$\ÞƒgçŸlt«€ÝUébæ!)û£ç³Ÿˆc™mJØ•ð­Æ{V<^¾¼.ì¯ÜÕvòKêY㯳Ó ±»Uoü>~{Ìè½³÷qÿéH¡ƒéÃí‡îl o³ŽÞÅp[ <°òar°z¤¤r5g¥yX;ú0ŒÊ.J﮲[„ ®od¢à›»3‰üÏD 9ÄôìðÏÀW/ Žžá¤‹Ô¼z¼æÿ‰‘= endstream endobj 94 0 obj << /Length1 1361 /Length2 6104 /Length3 0 /Length 7044 /Filter /FlateDecode >> stream xÚw4œßö¶ Êè½3Zm-zÞ‰n˜Á(3ŒÁhAôމѢQ‚DKDïD‹%"D¯ÿI¹÷þ÷ûÖú¾õ®5ï9{?{Ÿ½Ï~žw­áf70T‚  ê8JPDt¨¢«|_ ‰ @¢nnÊú× à6ƒ"½aøÝÿPABÁ(ŒMŒÂàtp –;PD ("qWDò.¤ÿD ïUÁ¾0PW¨…€C½Ü*O$ÌÙ…9æ_K ¯#PDZZRàw8PÉŠ„9‚á@]0Êê9Ñì4F8 (ÿ¤à•uA¡<ï ûùù =¼…Hgy>  å4‚zC‘¾PðWÃ@=°ôOgBn ‰ ÌûÝá„ò#¡@ŒÁæ…{c"|à(ˆ9h¬©Ô÷„Âÿ€uþ€ï("$òït£%‚Áƒž`¸? î t‚¹Cúê:B(4J†C~ÁîÞL<Ø s;`¿+Õ• `LƒÛóvDÂ*@Aè¿|a9ºa>ÞBþvA1ºùç‘jpGä—ÀDÅ%€`$ìÀŒ³Š`”¢S(,G 0!@LsÁ@'ðkž’¢@á{¿L€¤uôA"1ºú=vÌ™ÿÚÿ1І:f&Ž2®um'/•˜ü×寏×Ìñ Î ßøœ‘âgòÕä=\D)e~|O>·¢Æ{¨8Ëv¸ýú~tKºaëyÐ…]ªÑÈZ+`z˜¶g¨h[©¾›…YÐDq=èÒ+È,Ì çõ-î/)RƒBª¿. t}wùçþ¨É5Ãõ m¢‹òQÁDÓë°ÒqîgO&è9ðP‚,·) ÉÆÆ(ó‡®Ù´RùÁßÅŠ-—D“N'æ+MD½ß2p1XÒ³àRöÜ TÞÈÒ¢› ,+‰}õl|¢ôžÿnëêþZïp-¤…AžÉ£ l…t ™I´38˜~7fwUA‡ËÖMÇj‘cCê(ó #ì¶pizÞ2BˆëÕ†îRO0÷;ýÎ<ìkƒ/ȹæ<&­9î¨sné1Nút7ƒ9Îç?ܘ !Å»{#+ùUÂ@k&³È`’,V8J°¤Ç0Ãöꇕ'C/e÷Ä:J1¼Kåá}NN¢'Ì¢¹ÐÝK“¶¥+£hn߬Â(¿4Iùƒ#kñªq§Ô¦»ìùÁ }¢ÍéÏ£œ¬¼O†Ó®†tÓ…h»îᤚ0éE²½ý‘­®D¸\£0¨¯^÷,‡Ý}0ʤöÖ)öNPEÈ+kîÝ;qq'ÙÃìƸëXõ>ÕšèªéWã쉈ÆÍ(ºÏº†W\=w§r_5)eŸiSkΛqµæU2Æ©åHªÚ¿_¹Ãw©v–ÁŸoüX{-ñaYÎþüóDÈ›ŽúÉŒ„Tæ×ÅÅbúzá¤ô®—-;®6¹C§¯÷®Ä!JññÂNavÑêÙ7#éNꆸ²–ÒÍ s‡ø–,¿!ž %Q½ú î7íô+;ƆßPK×¶ «ÓY«³g0×e7K!ƒB„$—7µ´)‰² VtUR®@“ºBtï˵<#eºsÞÙ»Ü$ªvécP|Ÿ&ÕýíÕ”©~ž%–,ò&öùÅ “äš+£B7Õ†]Jü~ºÛö¼w8KrnŒì¤YÝŒaÈúú‰Ï¨ðcQæû*$¼P&]#“oß[¢¼ÖÂ…‡š—¸æÛêê¬Ù œ-»[²°æz¡ù‰ÙúÑ©SQ|™˜øoù×·_JÇõAyJ|sÁKHßįv<îLŒ ÁŽ;V1L¯UüêP²Ê·Ø¥ò–&ûü$ñ#èý<$šOÃq}‡H 5ã2Ûƒ—›ND¶¯Æ×G;Þ»QxúRs6ñ½˜Ò¥ö°Òô-½±/Þ¥([õ£%ô¨Âó#ˆ8¢Ûòð¶<žƒì½zQ™"™*€Ôé¸NÉ2ß^žÁAmRT¸ndždŠÜgžqÍ™ùXF×(]a鄃¨ðZ—µ}¡éeþLm§®­ÙããdiƒæÆï2njß!GèÁ2öAH7¾ƒuÑ­nXÚÖÖ Ò½×@n\ó8 KM"ªÜ›ž4§°·»ÌuÍ‚-!à÷øóD£<áûÛÇ¡ñŸ)ýR>´8‡{ä¦Ûj·ù>Oïzjô8-S¸7{K p+ìe¼;'¡n*èOUÀÁµÝ?Ý6Ïœ†TévÙ‰$É“úÉ“N¯¡]ž³\îßþt¯éó¬ÊðP —¸ç÷ ]ÂðíZ›nÂŒŽû –ÑrôùÖùòúýÇ+?¹D¦DÊ”ýI‚UÙúJÚ 14}ÍÉ;Ë/òƒJ*a#qªE3ă3îù¬Di7_&­$oM­ôŠÛvëNAfŸ§; (“‰5õP)”Ùú@dK =ºÓÌa¹ù£MÒ«nbUŽ,…ô'Á¬Úq¿tÅmn篚hn†Ù€Ã¸Adj¦»Ðxœn‰”ü­„új>Ô•Z\½}½êºGnø»eoKvÔ¶«âpå‚*­u"2iÔ‹™¢³â{9ó¶Tñ¬~ÔÓð9¶õÛÀ—'2Á€ìk£=žþØ€%%]wpîài¹œBLJí:É !¶6Âx3"eÄ}SüvŒß Vº£¬ÜSCÝP‹²ÒŽºš¥ÔÝ£”ÑÔ(ÏMB]#dÎ'í+Ž™ðït€ ûoÃö¿N*=búLìš4@õ5S.q;å`yBdÚœ¥­?e 7Úʲßwt'{€á;㇊¹™ïã:¼ 9½3BñýLb‰°ºª3%ÉéçÁÂèYµÔM“†/X™æ7;9L~@¨F÷//¿þšåøÉÚL³àžñ^ÉeñÎzÞÇ‚iO)¶m?Ø©õGi7“óŸ%=o¾Qô>‰ º³< "9×âú(ãB=é]LÊ@s¢ZTDÎk¦±RÆD¾ òPu£!®7ìèÄ‹ù/Fß/÷¥"œfØÌÜ5n1mÚ íVUeã=g¿•V3÷m'Ì|ÊQGL©ç£]öNê€ý\¯Hv{¬Éé(H%dA®Â–„DH1ÖìG†x/gË8 yö¶7)ê;²Ðœ&ÞÇ_ Ç(YO˜Uì‰Ò©­eãÐÙñ©BeJ¡ò“N#™GH½ç|úù£Tº{þb®þ 77föi$ƒžUWÁƒg¿1sx~ž°¼h« ?QóäøfÎPk©Ø‡8¡ãUs"ýŒßX£“LÏ€ÓNïÒ›[?£qÛgêBT¡€MjÝ»  àTnd¨ãí}qãTÍ"”´K‘ùNß“7É«ŽÐw¼C Øšo3Õ'K¸ºKV†²2;åX¿DkHžqMÚá)Ê4Þs.‰Î®.ÉR“ú^*<&p2¶-Ç>¹Î¼îó6…qêLnnž`ݱËÜ y¤c\ÿ9©èþ­þ¹V$å“‚³¹sÑ’6ü=¯DéÅs~^ØJ3ªÏ/TLJt"ËE}*©ÚM寙dÅAsÐû²DD…ƒmó§á¢ /‹>ɸ†=¦CÞöUÜÞç<¨f ÎÔÀîg½¯Î»;̦°:r4*åèT£HŽrhؾŒUk ÕdÏŸŠ¡!ˆ6Kú¡k» “m¸d(óƒÊb­g%cZOüh§M‰µØÒú™Ûþ+½9ü¶oo"K È›v¢;bˆ÷¥‰¾šOð÷½G2ë‰^Í3ÂoaÅ'b©Hï-Y‘ž4^á1Ç·SŠ©‹æÃ§ô±­õêLJ>²§‰ñ:1J¾êØ<öúœ¥Ïz¾Ÿ)6X¦m{ÍyÌMÍ^ÎÏt™û³ BKÔ%ëgE[>JîÁi1XÑæàTQõ»ùÞ¼ö)aw’;éá”¶êá˺WG ®~€ÕY{Óœ‰=oEu?=ϵHÿ 6’½­?’ËóÁêI ‰ûKšVZ†ÀÇ|½®Ý$ö<ô`’øh‹ÂT§0µÞ“ÇO ¾mj'Š—ÀQo¿¿ Ë:HöDR]gïGTv.½Žæpôºq†³8™ù¦½X?Ž¢"ךј8: •.b6 1³jhåç;Æ®NäHÚ¤´ šÐz<|¶ÑÊÎÿÖùÔòn2¶ÿ= úÁt׸ÔT–åÚJo÷i²ï\•4ò~Þh´ÒÇCO¢b÷+îï]óѸIŒìÅ ›ø[Ò**Ñ·|’êvìø¤«Êë÷—æÕÂJ‘úÀPNÊô fçÒÿ¨ÝëÛUŠG«µ1µéV¸ùÏÜôEÞý(Ì*ʰ׼™‘¾:¶nPw^BæAi&Wê]¹2#ÈÛå=yð`¬?­Y'fïÞå IáâbeÕ‰o­¸ Þ WJÒµ£—ÂËü}ðú’¶Þg~š0/l‰+- .gízX­Œ”1åë©öc{<²èc²ÍÁ”äCÊ€¯yQXîåuwÔ:ÄêV-^œ÷¸4¹Y;F°f]sxYÑ&åz²Ã%¾±‘\¤ŸØ¥§›.ˆK”í`äæB†ôÃÔ×~ï¥8õÑòChÆÅævB«Ó§»)÷ööÞ% ®8ÓÂ5W 5êÖò'Ç‘Ä <é›!p²íOTfx,«D Ø‡Õ¤âÚâòð K6ïdŒgÑݘî\{²)Ùl+o.*X#3‰W$âãüí;Óµ6™!«Ï;[ÝLÉ›g½züÓü-v±#BÝÑ‹ëîÙ2›£JÃRäÀè "ƒ¨fcc“T2ÖápkK»AEÂâ‹Y3wQ~RöÙžgA{Ö´¸g† :6²qÁø9&-žØ7Û$ÍÓ/„âá'Šú,UT—Õšôk— i6;”YM{-ÛØô™T†_ŠKí4º[Ù¨TJÖóÎLe„xÂߠדiôØÎ‹ã».Y¦Ê&·¦pj ÌÜën¹n™ú<Ð_Ž^ΣMiÈ6 *R¿¯¬PTþeqÄ¡§íÛ¶„ 5¹x’ü§5æ>¼¦|ü˜ÜÁвŒWëÝÄ–›u™½³8óVåhó¸¥JÔN—Ì#ˆL)Ö- ™&”À zßL=7>oš•:ÛŽTEZO Öä¼äçÇê~§GÇŸ….°ûIæÁD©èœö^^K27¯Ÿ*cUOòtçÅøÀ÷œ’ 4lÐtêXSoT}Ñ‘ÒiŸ¥_>¹”Ñci‚s0þOf.·ú#z#øY½ëìîU䍨Î5–P øR©ZXÄÂͼT¸aUôuÁ‡=Z'ÄOÑÝìÈ„OÍ*=ŒÅêÜm¨¨I ·Ø—y8bàqXا3¡%¦VÿÖ×cYÍtý2Ì$8JÊEŸ¥,®ð0í¨ðåv?’S²Eµœ*žÙÏ__2ÉÈà­ö”›yvPb™îús1º ¤äÊ<{š£ý$VœE&gÁ–•¸|ÌZï‘},Ë7ÇX¬’ÏNÿŽÝõÌɤåX¢$éºOœpéy!6áìîó+ßÜ£Ïú“|ïóG%³§sÅBˆã±Ö¬îÓXvZX¾Ž™†ÏëP*}¸Ð`t+ê'â]\8ûI¥6:Ϋ²EãBÐЕµ¿áž+DŽt‘žÈ˜â5·›Uˆ·}ÅËaÓ éÔfÍ#ùj,é*L„,U¨.ª$éÇÂR¬ Vk™ðVT?l£ˆx°ik°‰sxèLÌ †ô$î2­Q’f•8»¬?Ô¬c½~æò‚—¤§tIü6†fÝïN×êQÿ†'?ÿò©û ÉX¥Dô–c¯–ÂJ´KøòŽ ü³EµÊ{)€X¥ïø“ƒjrÄÚ‘}ó¸b°$±‰Øàû.MÙÝÍr•^M‰ !GˆÃÕqvyÆhï­d:”*ÀÖá¾ËsDu©+ydwh׋O3ýÆv,vtþ‹ÑÖoøÜÏì9‘À)ºW<¼Ñ!œo‘‰+î³øŠ£_$O¶´'&VX£¯d>ÁÕXSá3»G¹ÂÇÅ”8¶Y¾šY¡)b?±<ŠüÀ WãX <ÞcÖÌàÑKJ±¬Pr®Ö€Åm4É?ÒqÀ'P|ºÉœìu<ÚZâ«rº³ÌB²ŽJÆÊá›~úÔzvÓ,³Ñýø„‚i¿!ù×9c ÉýÖš®:ï ]¿ ¢Þ2ÖØ@°ÑÜO16CÖøÝÄÓä¬ñó8ŸðÀúŽˆq¼[5ÕÛ§i¹ðÄ2 ³ã,nƒY®OqämÍU>Ó‘od6j©¤ðšBÇ5íaž?<6enbY•9k}Äáa÷v:ñö 3¶Jñ0¹Å%„ÜJ÷xÇæñú2y”‡Ï^2 m¸§²a%àÔo3P)Y*ô-0üö£©×wÞèØLz®Éä…],82\Ån¥Z˜” 1çùöÉjjxÝR×Tj—0[ñx°È9í—ð-Cˆ<î‡ ›¾Š½„"¢îýßÕ?9ZºÐhñUfrà5ÂÒÑ´,}PgQ0\º9eÙ¬ŸJu6OV©õ²nø~Dï–”dCA½?“Ÿ…‰UžàÐî^KAË>¬v9sN/ÿ6­‹ŠÉ¦óþ3† Oígt ÃIÆù·VËÛ’™Ýž;¿nÇŠºÐVIV¿Ã¾kfùW*1ýMLÁ¤TN¸„w»–ÏÐJèw{ãX/c­ϵñÍÞÂ÷`ÒRS<­)/uE” ÑÔÕ Æòx0¾åUðÉÂG£û44·Í<ç™/]ô¶`N³»1\ø6eº:ÓÀ¾é2¤ý§«ͧïÉÕÆ½Ã—tuƒIuÏ5=lç.Ly¥ÞÈË~en!h÷DtµÚÜu& ?/3+óìkžÉhvÿ>Zˆ5Lã 52&ŒÜ¿'Yót~ðç|ò5¥4ä ÀîõÑÕ+ ßÞQƒînþ%<é)_9ÒÇñd_T‚îP4ÿ\´w1  çÇ–ÇöÑ¢áÆ¦õ¼.âV¥ê7“ ]š?~²˜_JñÖïå{i"hgÚµ ©>Ö(CÝ[a´7Åý!rø'®¹ñ½ÞÃ'&íOîù‡H§RèÉm‚õŠ 'ð}™¯"œ@;v ®F} g»±y×Ç+Èh³ìJ{Ûn½ŠUZ©br$N-RÝlº9gU/H"/$Âó&w@‘7&ø.Ó®ð.D¨Ì‡ u=ïõµLÅÚsTÈÀ¡YòûЀÿ¡™‘p endstream endobj 96 0 obj << /Length1 1497 /Length2 6463 /Length3 0 /Length 7459 /Filter /FlateDecode >> stream xÚtTì6ÝݨÀR’Kw§t·°, ,± ËÒ%Ý!RJ—tI(Ò-(HKJHK R?úùÝ{¿{ÿsþÿì9»ûÎó̼3ï<3,Œ:úÜr¶pˆ2†äòð‰4•L€|>>>>~<(Òò·Å‚p‡ÂaâÿÁP@@@È;›"yGÔ„Ã=œ@PX("ÎÇàçãû›GˆAžP[€&à1qÇcQ€»ú  öÈ»{þþ `s€bb"\¿Ýr. ‚4AHˆËÝ`3@†B>ÿÁ.é€DºŠóòzyyñ€\Üyà{i.€éЃ¸Cž[À¯’Z ÈŸÒxðXP÷¿}¸Ò „€î ÎP0æ~çâ³… w·ôÕ4Ú®Ø_d¿\€?òÿî÷¯@PØog wqÁ| 0{€ÔÐVÖàAz#¹ ˜í/"ÈÙ~çòAA6w„ß©ƒÊrºÐ]…ês# ®Hww¨ó¯y…¹{f%˜­ÜÅCºãýÊOŠ€€ïÞ݇÷Os`p/˜ßß';(ÌÖîW¶®¼†0¨›DMñç΄÷o›= âã@Üo°ï¯ |\!¿Aà/ó] ~®pW€Ý]¨äîÏÏä  ¿ÿþy¶P0`±‡ÂðþýÎ ±ûë|×Ô`Îw'? €ï×ç_ÿ,ïf ‡9ûü›þ»Å¼ºú †¦ÆœJþ(/÷øqó ¸ùÅ„ù@A!A€ˆ? àŸt@Ð?‰ü‡³Ìû+ß»‡ú;gÏ?"`ÿ3!€ÆÒ‚ßI`ÿ·Ò-ø„øÀw_Àÿo½ÿvù¿ÉüW”ÿ§Òÿ;#egçß8û_„ÿÁA.PgŸ?Œ;éz ïÆ@~7 °ÿ¦Cþš]Mˆ-ÔÃå¿Q5$ènä`öÎÿzH¨»2Ôb«E‚þRÌ_vÃ_³æ …AtàîÐ_ÛÀ äãû/ìnÀÀNwÄýN–¿!ÈÝüüóJ%nûkÐø…„ äƒw׿»“Àx7‘¶ïßBðòÀàÈ;À]y;8ïWOùù¼6Ø r·í6P{û_„¿0à¿°ß}üðºÞ-Øÿ¸Åþ ÿë$zçäîáâò{ÚBÝ]A>¿ÁÔö@Üy#ëî®à¿Ï¿7 â ãÍNÁÁaޝÃZÏkåî{q¯H}fY7Îààö›E´yü$ÂNå¨É YBœÉ¥v“̯*±ŸÊÎ1\ûí47`G¾{®ÛréeõLol½of”ªÿSÑŽ\}înÙ ÿk7£`'ôfÔŽÇ,yn¢D:äç^½*Þõ}å_>FL­ënÔ«ã_•sÇÆY—L°äÛdOÒ0a!¹ép‘xOœž}&ËýtËðø'^Àn¼@±ŸÙ2ÂŤïB¥¿{'íCZ3:ôS²c¬~ò›i©§ýJ‹ 'ì\oUdÌÏ ‚‘NYlBTÚ:àѰ9Oµü |ê®<œ/e˜ð£ç@E"¸ðÈ+wçþi¥‚¦­:öôV S?fÏJB—A@+–’È;•À¥|Îó¾âð—"q}ÙTâ¬Zù–òk¸!¤øE#7ž—Ýìd_£ÐD£¨)kób:™q( ½r ~î[ñâ‘H±‰‡§×Ë^¸’Ùˆs{Ö©ØJë} NÅ*c¹â¾å¥‰zú=ºµIlo‰BKD6©lÑ2*ömB.i¾ vjõòÓ½õ@»ŸÃvDg.ü! ¥bN|Ô"M¿ùc¸«óêö½ï”ã^!+’Æ×Ú&ó$¸¥T®OM_°¾PyåL¯È>G§—±©QŽí¥üì +ðy\×DZhí¡VH*—:†mßeÁ>…¥­'2³\ÑphèÔTóSE½µ)±Ú¥±L‡qMüüAþ ´ÍÑ|‚RõõD½Q¸f²£hñ°vÐL òÔL~ý%ûù&t¥Ø.‹»5¸Ñ\W¥ˆ£ë鹆Þ÷øe_½±ö ¶!G9i" e¬©“j¾?TÃËj}¦ý²¿"ê+}o£Xì‡ÖX–'÷Ü‘gá2Ç•A¡â#º5€Ü&·µ_K¿ØxOKÿ8£ó hKªfÝÙ‘r¶œ¤ß}1ó¦!r»HéylôKÉžy¢Ü“´O×g„Qæé¼ÄëJzÃïõ¶â%Rˆ˜ú…ÙûÊêÕïË(0KKØ;p´O¨€š¿ãŠÁtË:%Cd©7"²ÙÉ$çP£â3ïûsFpLêôˆ¤àg-«³¥Oss_-1¯ª9¼nÚìYW>ƒ¨]rGé×cÐåÌ„1ÀÖ¾}ô÷+3DÙùxÆ òU_vïXzjGÊ.Äš‘}ŠÒbVÛ¯fCµ›÷¼ì שFwM<¾Á|g½Ý­ÚÅ`_ãÿ …I‹êkW\pT'×:ôÌ\æYÐÅÐm@zÐg­/}ÜÈSŽ¢'JjìOæ“5͉—ãmow ìJÍÛw:îð]ØòË›ÇG(LñWw˜ƒ¿*ô `÷UÒW¡]Ø­_S_x~€UìÛÈÛÏ>±{¦MQd\‰NÒ. ¾`ÿ­êq°9ó"}ÿìB«9χ©ÎÄV¾÷ÑA’+?J,›>zæTO¹`›ýØ/Î$Ȭa¸”j36r0%fÕ»ñ«bäú2ߺɶ£…¦¬K´‰Ïs*² ¥™+[ÅÙª½¡Ê¥î,»7¼²ÃÖ)E ì§\hô#:ÐT_Hzjuvi5Û«Z0èC·AŠæ3+•÷[+Ê'´ZoÃâ|–V½&} ÑûŒ=¢^)·‚‘1˜„36CÝe@"ñ¾vs·ŠÓ\0IþøʃÔ7Á@ ‘¼zE€z¸± šþÖiÔÒ±ÅCsõ½‹Š³ƒ'l~qW:Fóå¹2|f–t—ñ“ ü )ÙÏ"‹°Ÿ5Âxþ´¹§“ôó\J¤Áy?C8‡ZXjŽ­ˆä GÞ‰°sàG\†{Sª ¿H¹rëîѽ> w^§®Æ}5hí¶7áóš{l¥]ØŽ!²EñUS±C¨–Ý}›ÆNrH‰§EÕkÑlGÚ¾¡¤wX«˜JÈ +Ƹ.!Š/;·ôBüé§JÖmîDvÅP>¾7¥vN tØæÁ¹ÕióÞ´q lRÞ]ÆAªGeøÐ@Ö-,yûí¼6û€Iq«Ã~2H±LŸÀàå=®^‘ž7Ín A»g’ÅYýit»¶êl<¼=º½KÜŽºÂ„T³k~—(?„¬$‰è×»y.å,·V?[ó¤`ÊâEgܰ_M@§s&¡‘`qÚi&d‰{é¹Ë„ò–å‹fY7†\²ðT„–_Pø{–w©–ñ4Š—¯­%6¤ÕvZë7 û×훳òô¡›3=p»Ñ{êgq¸@žOáÁBÁô(“ªKs—íùr´]…ìJæ Ï™¯•17f×5ŽøZò¯-C‹)x|ÞBOz.q…ü¢âÞ?dB!—¸¿5mzeY‘`Çú]½®3 «…û£T†ésyÖÏ÷}å¹<ô_e¡˜•ÿºÇa¤Ðñ-m‰©ð™Å*üµì#=Ãi®„)á¨5Éû݂̂8ÖqÞŠý½¨pq¥)äz?Í0afñÈôˆê—–^ßsN3äÊsRÙ-ó¥./Õ£0Oèö•äz¤ŠêFÄ?hjO€úÇ9LÍW_Ùp-й€f5ÇÐ$ ‹¥›ÅÔZµ´F…ú×XÓ}’¡ ŠEz*oxÍŒ ø9úGé¢é§x׸Wѵ=ë„ÐûH»d}‚ÊbCÄ–ï˜GoQ§ÇÞ_£zFø¡|gÄýPè;(i†r OàAÚ ÈÛ½»w Û´˜3Ô]PË2^©åoA U¤År¸CGTág5Îj3›c‡|°†@ yxd§—ô| ÷çë9¼¬üœ1>3CÂÛæĨ³ñÎSªÖwÆ|´˜×Ok/sêeTãµ¨ßø-ìm ;ÜÏíWKŸgv†6|ßô©N=‹óðA¾$Ù€¤ÍŒš ‚zòö>Q`ñçJ^×§%C¾¼ y:¨¶>FX¶¯òî´Yhã@ö)? “Lß6ž¥µUU™€[T‡ÕbkäÄ+#‹#øA¨‹X)õq¦m˜¾pЭI¼L@DçjŽÆ iéãOO|dC›D €v?Ã×÷®F!¯i]*ö¸”ym×2èb&è˜TEîÉt;d„œ¥ñãm8ãi78Û°+›Š«þ\ªÛ¬åé#„èá ¼¶,•¬{La#ë÷q°HŽ•G?ŠpÐN:Þy¦QØŸ/œÎ»ð­#¥§\[µ"·R‰1p2ÂL"Ž× ‚_Ü)ÜHüÀt­sô‰ž'¢aK³Zë|Êh\mlÖØøÊûI¶ŽÜÖÈäíR`åÆyŽz6¦2Yÿö™µY貊ž5½Aª½À®j—àJö4ÛD]к|>Ø)›óMÅ"íÊ" ©£ÿJ³È<|¹EUKË,:‰ô±ý(–àxÐÞÖr¡ßSÜ74-œCì…Çßçfx¶1ÑO–GTàæg>äžuÏxÔ²õ[9“ÄH(G 4l9£ŸÛy¤´~(î tÑͯ³£ïqW9Ô±åØÖ3å_é•®öá·–Âx:BÛbHœzˆ‘ÙÛ™¹:jF}$ùH¦qmƒñ41‰g×'×RÇ!Zkj´b HþÑíiÏ÷ „)¹ýþëZÍ„½¯¼±DdÛ]x'V&ƒ3#k-í;6¹%W9ç‡{Ö¦ØÒnã#•VaÛïQ?ÚÍH’ ;©*ˆ „Yëä‘·8!)§‡?³ šÏ ÷¾[E†w;>‰åÞc3‡Á|(µÍ ŵ×n&d RʃJ’Ú|%” •rŒ’€$^+ÂLîåQÅ\6økÆÚkºl±Cu#g²e3<Œ†Þ£GÄ=G)C©‘bE·:õWKɬ9b[—Ãn™#Š’âK€ÆšXÒ ›6ÖÃÛ 0ðÚ &Ëœ·eÅt´FuUW/:ÁuÛ:h¯)CÏ'"—ð‹Ùœød|ó³„ö³'Á°i_"Ƽˆ+øWh“¯ö9ú´ß$kbkÚö“4o$ÝLìOxM.£ß¸ƒjiø˜ö1ä*v*N!ˆ+MÉÎ èòZÄ1yˆ\Ñì±…ZÓo±X‡Ø>™m°ÃlNãÏ00…)Æ·¸pí8’¶/?ÁEd§ónÄíB˜¹éiM?ÓØßùL2R%^¿zòuv]£¤ ($0ëèé{â3þ`rKlëMJuÌFú}‘)r³SÓ3,Ôõøú§Íµ»–ËŹß$_R„+<|4ç`®ÜÍ‚*¿©qÒÎÌYäóÆjÓôtßÌ)¦סÊKNzÃ/š&GÖ¹ª^Z;­°Äf¨yØ€®`7zs‘4¶´‹˜„ÃâÈy™7µ»¦¯ô~^æó›8Sio%U[vù•:&áuÖÙ¥§FYE»ú§¬oK™ig°ç,¿¶ÞQÅÅ{'Dø2kŒ>RÄÖç=ZÇT…¥Ñ`ÃäDØné¨"ûˆ±y/RÏÂÞH¸ë›—«~ŽQ3¿ÙEZ.`ÆK'}’ Q»d< "R.‘¨ñ*à[\wÔoÒɥզ̇ìÏk‚ƒ[”@«µ’ˆý'D/0Êðæ§@\ÖõIhѽ¤bE»MÝåä™÷&Ðòšc¥<ƒF¸Ø¶•Iìm¤œiß_j?ØÀõò•H”Tß{eòo4è+k¢©êïÀú Æ›˜›~åO»å«Ý¨ÇįƒÌ†û8ðÕâ?É'r£É±Ú$øÉ»wòlaÎ/Götû©f/úð’´Nî—T}êšç™ºpMjoºE=ûÄû¼·;ƒŸjÅ T8TèÚíQVŠõiì=‹Û#ÜoÊà1Øñ©&ï~[6&[_y±ìèyû¸o>ÊÃA»‹õM—3R#"°e8Ox{TL烬£dì%`ÔÀH_hbêu º ¡ˆk³œ¼ñú¬öSLÿ5ƒƒ$& âJÃà]ÿ,,¢Ë Hi[gè ,d¶º—TËÞ^È÷ÎØEºláí4&¯fªD ¾WW¯¼$ÿœNë µÜêr—)/[7Ht7þƒ‡‰¤t}c Èe©Áå<$Ã|²$÷âšµê4A¹*‘Mùå'/Rl4«Ì‚cæ¸àèdÐK;Pº¦=æƒ#§ž¨šGòu?=[&J}™ÎÅ`ÑyP¢iÁ*áá5ÿ%˜UŽ©„ï|÷Z1Û ëÚpA²õÐ8—°ü!xUÃëÓViw >N–ð†i‡D‡‡>nàORÅÔW!îJ©÷H„¹«’k›ª±ÂMßÐŽ^ºpuñ2$nÙ¤ø»ÿ´¸{O endstream endobj 98 0 obj << /Length1 1548 /Length2 8423 /Length3 0 /Length 9462 /Filter /FlateDecode >> stream xÚ·T“Û6LUŠH©ªô$té½7鈔„–DzGz‘^é HïEšRéE¤ƒTAzøðœsÏ=çþÿZß·²V²Ÿ™gfÏìýÌ›„IGWÖaSB8cxÁ| 1€¼¦¦*ø@ ~"66}8Æö;›! …†#œÅþÁGÁ ˜›sCÔD8Ôž;À°°XD ðƒ@ÿCD Ä W¸5@“ †p†¡‰ØäHÜÖs³Ï–('üø±ÏáY' …84!;˜ÓÍŽPˆ#@…Ã0ÿJÁ!a‡Á Å€@777>ˆš²•âä¸Á1v]†r…Y~· Ђ8Áþjˆ  oGÿéÐCØ`Ü (àÆà‡ÂœÑ7!Ï­a(ÀÍî=U €6æü'YãOà¯Ã€ùÀ§û+úw"¸óÁ(á„„8{Àm6pG@[IƒãŽá@œ­!ŽhÄM<Äw„XÝþ(P’}€ÜtøWh( ŽÄ ùÐpÇß=§¹9fEgky„“̃&ú]ŸƒÞœ»ð¯ËupF¸9{ýÙÀ­m~·aý 4p†»<‡©*üŹ1ý×f Ä@¢"¢B˜ æµþÞ@ß ûà þm¾éÁÇ ‰@lnÚ€ùÀm`7D^hˆ+ €A=‡ùxýÓñoD¬áP À f w&úoö3ÌæO|sÿ(¸;Àt#?0ôûõ÷ÊìFaÖgGÿÒÿ¸b ºŠ¢’ÚSî¿ZþÛ)'‡pxñ ðxù…@0HP r³ðùwü¯:þ«êlƒüNö»Þ›ƒúOÍ®‰€ã¯ áü;™âFº0Ç•þ $‚Þ¼ÿŸõþGÈÿŸÌgù¿*ý+Rzîèø‡ŸãOÂÿÇq‚;züŸ‘îsÌÍh"n†Áù©F°?gWf îô¿^U äfdmo$Í ä þi‡£•àî0k8j÷§lþ´ü8G¸3L†ÿ~ÄÜD@ÿã»™2¨ÃÍc}£Í?]ôÍÈaþ¸Èßv3Tÿ®CÑаþ=}üB ñ º¹ü$ðߌ©5Ìýu€|ÎÌMà¦g€ Eôû¢E„nõÛô'µþF¢7Èøo$ô¼v'Èß0ˆ´ùíþoRÃÿ oätþo¶›ÜÈU ¬ÿ6ñßð‘ÿà¿ pÿþ«cèsêæHþêÍqüÿñðÁÜaP¢é Tü…}õ‹–ÓJYz7ÞÕA üÝÔSc~ÞÁ|sL·âˆÅb¼ÞëŒõ·JÓ]`%sûv-9—Ó̹ɯ•š‡µ‚Ǽ•ÖmZÅN\c%x00O6bå%Ë={ƒêÆÒaŒ$ë ”¶¶íÖc#õ~TÝ9•àÆNU¤$¢Ü¢ÕÖQY”¡AÍðØp{IÓdÕ§¿*tHó(kÉéeÐrÔ³ð­WÌêmgDñnd ½ŸúfÉ»ßG¨¹®î±†`ÈÅûoWÎæÞý•õÆuë…—•2¥Ì–uwÉ6A¹=¬Ž¯g9‚ÇŒøMESíZô¥ g¿hö7& q…§–ÜÖ}ï½]`rìå)x•{ÐÖ&Ç>ãRéMå¹P3ëý*?ç| üX M«Š’÷tsZ<ö®p_4ÓèuBø”y¶8 í,L’ÇD!Ä3‹7Û#^DÓð5‡3—}]î ;zPŠkU¤•]r~@ý¹Ñ‚%@äZu¾åy¹…ó  E„¬õÄ«sOÜp¤ÝV¸7·›Gjšçú§Tsж¿š “,hàRW ~¢‰‹ªÏ«;—ºy…â1ùb0üùêf⊶ ©/åWŽ’@ª'}Añ/£Zd›²"9G«èÇvî´Ið=.þž±u(`KOw!y÷¡i›ÅÃk«k„Ž=9ÃüùóžÀþÛõûFÌwhqVê_§X€ìo ž*¡µã8_$æ³íZóï7¹Ei½+U‘:Rð9Ú#øú:®Û$énËNmØþiymÑÝ[í`«¶¦ µ·oË6»è7V~œïÝL‚v¡¦ì ûĽjc,S["ÊT;Z½†à\v|? äïȦ̈́9r1aAÔMï(èi¸q°î¬{°ÐÎh “©¨áÂñôNŽK2‚ó5<àŒéMK— Ñ‹™G ½|éÍceÿjϳ¿‰BÜ,h!wky“Ÿbö¢gð¼Ë=ŠMºmTWÍÿ ‰ÈP¸È( Çms—úd3©RsnúÐL_ñø½ý+ݾ_ ÷ñÊ õ&àU*Ù7GÑOãÙxdõaVúk&ÓÙëz¦zqkÁÜ<ýÒi!1žåsW*¨úK®Ù qnÏ¡{·£*(bX2Ÿ‹²ñcÿ¹G¾ÿ4 ’òkšÁ ¤Ð¿ÚÿµÐþ§b4¯èû$L”‡½À§×PCUÚJ\a ”&ói8&ØÚß?èÛ[­­~3(²x,xp-¹Ëø•öÎÕãk…9×{õé*ïg‡S&án”utš[y<\u]®B>|o Þ¯Ìb†´Ô’ß…™l¼ò‰Å.æ÷Qý°0¬¤|uûëðp‡o8hßA.>¬çÊñ‡”[üœñ¦Ä’ŠItež»+E+¥ÿŠ/k:sºG§ƒÂ]+ŒxMïÝ ûüÆ®×Ä+Z´Ç‚e£·ÏÒçkf¯íe¹Ò-mGBŸ¼­ª&ä(ü.ÞPù„Äàá y¤ÕúgñÓNy¬Ç° v%SŽH¢Ü_Ág±g…Òƒ4Ñ(¬ˆÔ–f‹ÃòàÛä#6ÚFs_p–‹²ŠžŸ¾ÄE;ãËÙ+„Žî|):«Ÿ" ¨â×01|úÖj[Ð8Ótô|ÀžZJq̡ڿšÉ¸»C#¶­þ4§7Î ¥ò8} êx®¯ÂÐçÆïVÅÞRòLd-ló7˜ªãà:NÚ»»,d3zU»ùTª¨k¸ØÎ¿Z%9$Tü2©ò@¸©·/ÿ‘âžW­×šŠûÌ£ŽÓù¾ts¦áz Z=탳[Üûè ÁÙ¦Ôž³6©Uþ)“6ª\ßLmh%q4)¨F9ÙZ~—“%¸¼-¯×æÀoÆû”)ŠöAæ–Z'‘ 'êÉqlã¡ãŽ)j']—uÊ“<5Sšx_› 2ÜÇ<É%Â5K÷-Àçy£¾nâw[Å z<øÒà—ÒAIJcÝóŽËØMC˜wÙqw b)~l¯vЖ8ÉŒpQ_ÄêS=ŽýªØ„S`¤mË;“<ñ>Š´Re¹Fs¢rë3Á3ÏüËdÀ‘Ñ/í ñ½^ƒÃÀAúUI†—#BFêÞžþ#Ë){Þ3õ˜ÛùörÖÁ†’åžÓ3új¾-k¦ÕmuWã–—®ÙWHgTÁMC£·éÇ|ß)Mx1ëi©•eÕüž¯Çƽ(À %’ÐÆ€ë»¾Y¾³zµ•ÕŒ¥ºE=ÌhÁùúiÚÁ˜ŒLñ®†¤À*´ýY~Q&2Ñ-DÖ뮡ð¹]65¢Þï²[«ñ‹.@#r'æÇ}_#9pÀ9»zf"ã+|6úíÐpqš¤sº%:héœ5GP‰ûà@Ñ´˜·ô²aûèîzPQ€gkòêKT9>éæÂkÜ[b `Dìw±F9ê#ÀÄÜ$»¿2}æÈ ø¹0È m@˜äýª9£ŠZ0kþ ÇÆž\õ%/¹CË»¦Ìžd‡‚ÅËwêÊ"¿ÉÌYäË>Z›ìxÚ<ÚAAÀê #tS0Å-‹îˆFúöï•ï…ß2žM½u’õœ>úžgݦ%μ’æKsk8¾¨ 9ßrœr¯‘V^¹ð®‚Ö—¬Ê˜®yoq€úíÂd²ó'c½––+möb@bœûXOk^ú__²q Ó?›:¨—[[L˜‰“ø´¹ªŊĵ/óïèÙpëÄsËæàåùjoÇ2~Z ºû÷Ðë&ͯíƒ}r‰ž}À¨ÈoÞÞýtö H¤°¸ x÷‰Ð ëZ¾“†[*‘SAß¶9téýÏ+Zêx ŠÉ ãxC=ã'OÊýã3C#ÃôÓÆ+$+€ý'ä i-«âÛx%•h',Îùûsùø¥˜yÝUxº#+ˆÍÝéR>MÌÿñÆw‹AÞw>¾Ã2ßwJ?ô+Kv™—ØMmr:T¤7Á E­FãNeÐæ÷OÝ•¯ùâBH·h´OB^=üa_±¬D¥?Ûò…„‡²D§ûµòm“X›—?õÌÉêFTq¯ù„¬Íãù¿WÜ µÚä‘/w9èd¹4¦ÜRÍKïKÅCµ>ûZÿ!ðãõÒ…. aé¿ØŒ²¾”Þ+¯È°—¼w«ªx¢Éÿ2vàp+7h°O6û!Ç´™æ í$9)!Õñl¬y@Ùð¸vR•ã·¡/’ ‡EæžBŠQ´ž¬WýÐ9P¿ƒ^òe«ìŽÝš8¸5Ì_ís•CdKÓ€²¢Ýûºô½ `@åB9½eÒëe™Ç‘L©ˆg2 ý3] ñ6©g{Å#:®÷'â!¼ŠÃt«Á ˆwúVKßÚ…V*Úµò'¶I•l«»»t,|´À¼a,K߈ýâ=l4ÉÊ\SŒ“©hNäÛå7_@Õˆeq0ªH•ÂÖgt¾ÕŒ‰WœLõNæ¼ðq2ÊØàýY§¹(;艶D[ùÄfFM_v¢e¨,Ð-@œ”Ìe¥°ö¤ÚñŠJRDÎÙ=S¶´Â>†<ÍÕÀÊ1·ßgn @Yáòªvìhwù@Ÿ¹äÚµ=²ñÈþŠ+Z:5ë|¶_–q´aMˆ¯-ÖfÖK/®Â_ÜZ-íØ^[ ¥¢X¹ÂÒ)89d‚ä}ÝäÅ€c¦€tÀÞ8+‰¤„ÙK‡ àª×ü­áDDú`6å‹—Ÿ{p—· D•ÙÈу-·z4Š&;_~`…êùoÑ 1½Ïl%E,ÍÈ3%×û®(޲ˆŠßñËÙ¯½>@ã1f™oVJÚÊïåÆÕ|“>[ßšô¼-ma–Ђò½ùª’PÄü"î;éSáCX©øzÿ]ÿ£n¾"kßå+bêâ !Vþœ x¥¾ Ú^aøÖåNÜ@cEIbÙò»m¾¶›‡ÙüijŒY4¿‹Ó+éL¾©õúócq¯£#ý&Á@²Ž¬/ßhå-ÜjsØZè²^p¯j§(Ùô&¬À,Þ Ðïð!iþ¡x±“Îÿ¶œ…¤ /íj³Û”°2…h<úüÒ£É@7Àa7öæIݧpú’Ïí!/¼[ŒuƲ [»96<ç†7¼\[c:µ0\Y?_AɰÒªèâ‰-Ä1cs¼é¯7ÙÞ¼©ÇW*:ÛM£½súm§šWV;‹£<ýѦŠcö9¨ðU…™Ÿ?•A¬¾ë[nµ|(À$ù\ñàÉFb Ç–‚„x}e×§†óùiÕÕ3~4cnýü…l­mˆrÂg-½ÞhÿدæI_È·e‘qùsy¹ìÅÈõJ*Éo¨ØaËä}ª G—2~³oíiZ0ú×ê뉠Æmé7#HpòÉObóèÔý»T÷È-u|õ®Jª»–V<[S§OOg¯á>ÏC=cå'd5ŒºO Ú°¶–L rJ]‚ðë¸p,³Q€âYâ²ÑžÞnD,}‡áŒD±àØÑᕌóîƒqݸ(çÌøp~¡ êÓõ>Þ­«oL©ê±”ˆÒ-Fe&!ÅÚÜ>½Î¦V„R'du–vJªm„¿QëµÜû˜hÌ©ë¾;¿‹ÛåN(4Ì7cWÄBЬ B»™5 êÃÖù¾o¢V+¸³Ò¸P æ‚GÝÓgç¡þ­öJI\Âä:ëKð/ßNïý’†yÂ1rÊC} ªý²ðaÅ®l—ÕfKo^íÅŠc)›£ç!èÄý=AAG^KPýYý·ý5iE–ŸºÎ#'IÌ›‹,ß²ãeô²Ï½Õˆ9¯UžáÑߢòáIq5VàôïOÞ5ËO¿Ìÿ‘ôn/¾tinCùK/ùû|¬†Wú¬ëªšÇ‹‘X¯D©šŒÞ• í,Y’–Ât=ø8øqÊAQ‡Oà¹ë·$¬Î’2²š°‰–³h¸ˆK’¤ã)õóó%ÑC!ÙCªYÃŽV>Ë,Ô¦FB4ýÃ>$ß§bùT ó¾H=´ªXzF|£*:\Jùæ=4ºÙ¦ý€sÞåË*™R†0Y; 'OÒ½º ¹?r†dªÐb­[ò:ì’;¥bDûñjˆ¬ÏMÎ3âEsRIw‹¨÷%øï^vÌçk„ÚËÍ,‹JK¿ø¾ª€ÿ.+ŽÁ4æÐC!qºý¶“Q¡K¼U$ä:ÙÛ“\Õr´¬xCzЭ¼|j¯õ‘Ì}²2ìk2­;ÔK£À[·‚q:’Àã—˜a:bw⤃#îj.‘ú]•!‹Ó'‡4`W´¶S>vÍW‡q¡kcôy²JLUµƒRdÕÅCwD¢÷Á¢;±ä!o†kÞ/å!*ŸJ“¦žÐ´±ÃŠ!^N8ÎO'ÉyŽRWÉËF¶‘ÅØW¨AI`i;;XIßGÑìg‹@Îü=6vÍFè<ùüDB«³åÆÂhnBY·{êyD'­†²¸ ±wRNÝêy“ê§ÑÄ' ªýÆ;¾<£g)+PĦÿ*Ÿú„U1‰©n˜©Äèi[†0Œì òi»³ºÇ»'ÿsjÜH›]Úo+ª®LeñTÃ08ô,4 Ú臋õÞ2jáÞÇKQ¯iiÿ5¡çö½íª_Úf(ŒšÙÜ×H`32‰èjI‹,۾ͯӰ–Š ÅQ¿N"s!0þ쇞–˜hNõ#[¼Âyé9Ε1]+¶á‹DTƒêºT„0ÉrW¯/g*¡vL¢Iª14÷"Js¾ÆH ¤Ìe)ýYWI³ˆd¡Óÿj|Àv9fÙa33ãŒ{…Õ&Ô&µ.ÍoŠÛQïÚ¢0Õ òht!¬•ë¨gí”yDöIxƒÉhkh0ö\ñ^}jÓ®ð)Ï…ªþ{ÖÐè«AôéÐ Ø*ÉHxÓÚµ!Ô.ӌƑýë|ŽRç7±¦dÚù¬,}L bEX‘äXÁlœ8·© 4Ëéšî —}½Ì)çÒ ¬)¸ÎÅB'@»üf"îÛÍ‚HârÑ.>–ô*´CEËC(¯Ø‹¥WQ£Ì¿‡%»€w»â¢ût’eª6+¾Ÿy.° Á¯ôèÙ½8ÑÍQõíåîL;?šª¿ROÒ¹ÌýŽYãmpÈÛnÞ¬•?±/ íi•`Hï, —>v7$¼ pfuéðy8LtzÔ-Úpi=Ǧu€Ê c %°¡7Ní>¸×œw^×X‚œƒ/›’ò¹Bi Kߟ!šÒJ†³}7ÿÃ:ƒjêÉGÄ]*7êÓéåGƒµþŽUãðŒŠ36zAÜðxv›f÷¬Áh¨ƒ‡¢AèÝý”W«m™OM:gÍõ]šËĉø&~Ä]¡_fîœh.Ç•{Ñ»Ó{á[(´Gïµ"Ûï½u¿w% ’±9øÖeú>q{àJ*b²nÆ–Ýj wø™ï–ºösüÍŒé}y—³ò²ï¬fc…À°{WJw™ÃëÉ~i†ÏAiNR”‚¨%ofš>ùû¯Pãiå”91VÁ?ÞÊï.ETª ¾óÜ×þX|·ÈDv¨É;Çç=¸QÍ>Ð1rÇ=• áªNñÞÊ1)¯h[Á‘ùÂ'h<§~§fj{ƇâV¤B•œ ×½€»vÅ\U$?ÁþT<´øwSU=”ÔäVŸ„Ó‰?ÕÌ+½s1[ÐÂpHùµKÿkÑ{ißrXqHDÍWœû%6]©w„•ÐïXÖÌŸ«ãÞYT°1 5nÇ èã«bQ÷áû$'EfŒ‘r÷ÁAKJè/'Ôήë{Kß°ýý?nCßxdÚR˜é¦xü"Ô¡"ï£a)]0{6nÃ@äü#]s{Œd–Lm —“ÿéGû£ä)¸SîÉfÆöª½,¡=œ¦or^ý´‘«Ç”­’Š:I‘Ç­¦@x¼„³Üñ ëìÜÄwYGc‡/ÍþÁ™U¤ç«4×_¾n@l÷L3kì-;ã—×PwDv»Í8¦—û„Ìt€R$¥¼U¼êøLËû_Þ|Ï¿uËõ %ó£Ïb›–@–úoÝ=EÚå±ä\.½¥…®î¶xíó<Þì€Msú•Ìw$„¾u’åòÊ.¼ã¦?ƒŽ\ð[¦×Ñø 膃éLkMå¼UÑ­«òq<â9¯¡ÄgܙӨӣ§ÞÑ [àŒÙVÍNÀ5a>ñsÍÄ¥SÅ¡//¦ŠŸ~ŒÈçTõ<9ÒÅn¨a¶h¹¥d:þÀ<Ý ìK1¥«&^ªÿÜÈD¿Ò™c?z”§K%(vGܵ‹ÿ­)ϯXxY?wZ„5|êÝs™Å’¢§§ÆµX ï§`(%+4BÖ¢’²g)†“Æ!¤®^´áƒ'Êîh\Ý;S^~öx­Âþ¦ôëIXÍæ€¬`‡#e_ôl¯j³5â4 uWö?‚ µót§–i”_g7̲½°ÇVïÊM{/³]zã|iY›Ô$/áå©=JÀ‘êS˜;È»4Mô‘ •-‰;v™üèÐ RÓÂSÄÛÕĹ|Ë".(º€ø³ÑM:QŠd¥i(ÆÛ™cë¬çdtLí:mú¶³\.R+˜´r‘喙ºCÍ*½x¸?ÈœGgI!6E(O±IÇ-±ãTW…&’¨kÒGÁêºïQUà”–ÄȽFš’ü šâ:0x•ÐT-K·ùqë)ƒ8Ë9߬ëm1£¾O3vETÞÑÀ™X˜ðmŒPgiÚ2å$KDtPãüãÇÏJƒÙiJ:â׎q¼úL^®ÔúÙàÿ39OÌ`ìUý¶÷ùwªÛ,¸âkšBÇ« ;.8q ¯f¿+K"à…Ͱ*ï˜T‹õÊ`:Tm( Ñ|ø“­CxÈÔ)‹-•®ëq”AYÓ¢{ΣwË÷ mÍ^çz >}G"1šÇŸKd"÷ªúÃs_ q'Gýön+hßPëNš­N]Ã;Ì‹éLi¶–Óõ\ øÒòûœø‰Ñý".FÊ8°@®Ö¢[­ÐiQQý¼: ˶e³܉¡Rk·4 îkÉa-cµf%ˆîÿHOé œàTëay$c4$µ bëÁJÞê9xã?bO: ­A¯Í¢®f™RÞgK¬?ÊcŽ«ËÙct%y"ë¢PÄq/=ò´•d\ögóåD½;ýÔ`¥lÈ(ïÄt™M¥—‰~{‡O¡Ú€jJüyo÷øî֌Ζ=;mûx‚(ßC/´8CŸR›š‰Ãòäçz¹`~í«)\SgÃýÑ8ÓóûC£ò—gËwÏ4{‡Fñ¶æ­¿®Z‘$«Pg— ©Ÿ¾yÚÌn")ÛZ_Öï-yÚý“ ëþ†ÓjUÙF;ܧh*í =\›ÎÜÑhÖýÊÎrÎtßÒ]˜àÃy¨…ˆÆ=&?•#q’aðøyßPŽç1;¸pOŠCij9•Î3HÕŒ(¿{ÀVv÷b1ü2뺨8o±Þ+ ­mËFŸ q”S)Óg ÛQÜ~+–=*¤5)ÊÊ3ÐóìwèнœÀêëIjR™OЈþxÖ¢7³‰-},ƒ#½®KA3.®p37"ã-^՚ܵbNVŽ&áÀZ¾µØ‘t¼Œ×úûÃ|¦ _¯Ý] Yl?-ö£Àw?ØÝÊjðró“~¯èéDguk" ”ÊõÅ1_.[摎¿Ç yËÝS>Û·$ŠÄ™¿¤Ê=,^D¬îZhH‰ïìT6¨~ðúhj‰-®qK@–•V>#²ëá žl¼cCa–¼¥'T~¢½Ÿ£HêÝhó›K‰lfHEHbÕ²så"7}Güb'ÍÂ:GØ‹ð‡º…áçÆ:x‚ˆï.mÖ+e¾^÷ËÕ§K A´¿µ}&nB/êÓS×µ3G·Kõ2m>ÂGêSTÚ#ÏQÍgãëè\5©"{2[n膃bÊþÇ@÷À¯ô£ülÛ%=¢ ÎÛ -Íô ”ÿŽɧbs%.zlÍÑ Ÿò¸œØ¯\¬h7çÅC>Ä|tÄá,aÝ¿sÛJØ`EÇé—/ײûàv‡%ue–‹Þtùiùî҇󢑦26m3mÜrA:qBE…zWWl,‹yë©ZBãž{Þ. ²‚Ó|ûKu`-Ħµ­Ÿ•q¶DHˆò`¬ÓÉ­ñ‰z}˜Ð»rõ!1åpÙ—h·ßÕDJgØÛ6 ×·†í²ŽxØY"±ž©©{å“] 42–xU¥p­ÐòMz(­†?Rü)÷Ú•îu3ÿ>q¸år@ÌQkƒ£œÕ2\(㻫€6ä*—’-ÙArl&<©Ûj*P ¨~ÆÎM€ÔTä@,ß·ØÃ•ÿe¦þ˜ P‚ endstream endobj 100 0 obj << /Length1 1433 /Length2 6673 /Length3 0 /Length 7656 /Filter /FlateDecode >> stream xÚT4œ]·"„ˆ’‚(F¯3=z‰Þk´Á`3˜Ñ£÷ÑEïD¢AB´ ¢D/Do‰Þû$ßÿ÷ûï]ëÞ5k½söÞÏÞgïsžç03hêpK[£,a ($†ÌȪ©) @ ~ˆˆ™YŽAÀþr1ëÃ\ÐpRô¿d]`P Ö'Å`qj($@ÙóÀ‚¢`!QÀ‰ü ˆrÈAÝàÖ5€2 C1Ë¢œ<]à¶vì6ÿZجØ`!®ßéiG˜ Ü Š¨A1v0GìŽVP@e‡a<ÿQ‚M܃qååuwwç:¢yP.¶ì\w8Æ  CÃ\Ü`Ö€_Ô¡Ž°?“ñ1tíàè?~” Æê`¸ ‰Æf¸"­a.ìæ%U€† ù¬úÀøël`ð¿Ëý•ý«ù;je…rt‚"=áH[€ h(¨ò`<0\(ÒúŠ@£°ùP7(µÄ~w(Hk Øÿmåw yÐpįy•Áž²<ÒZåèCbÐD¿ú“ƒ»À¬°ÇîÉûçf(w¤÷_† imókkW'^=$ÜÙ¦$÷ë"úÛg à a!~a~Ìó°²ãýU^×Ó ö;þåÆNàãí„rØ`‡€ùÀm`Ø?"o4Ô À¸¸Â|¼ÿ{àŸ °†[a–0[8’èïêX7Ìæ½|¸à Ë=0ôë÷ï•)–^Ö($ÂóoøïûåÕ”SÒ–—åü3ñ¿c22(€77Ÿ€›€Á`@»ðùgM(ü¯6@ç*!mP,þw»ØsúWËn1€í/u°þYL…¥- Àö7ËM@öþsýwÊÿFñ_Uþ/–ÿgC ®Äï0ÛïøÿCáÏ¿XÖºb° PCau€üO¨ìjÕ`ÖpWÇÿŒ*a X%H#m±læ ð€þøáh¸ÌZޱ²ûÙ?~½_ZCÀ‘0MþëqÁf@ÿà ÌÊû€ ±Äü‚¢±jÃü¾Æ_6 «§ö!´BYÿDuqza¯kAÞ`¬B­a¿© àåA¢0Øvf€ Ê…è×5 xu¹~[`À ÿÛÄ‚ùoS àE# h»ßž´aåêâ‚íó7{°=þËþýÀ`0+¢¯c(+±`ûªàÆãJiwî¥/âø[©Ç†|Ü_ Ì1òCæsñ:™“*¥ _ÛÁ föÍê2ÎÇY3ã{Þ‹Õk<¹*¬Ú>´|>vuxm$Áûˆ–qŒ´'× Y†^´Ø¥G“.Ь妤µm‡3éSÖª î,wK„”CÕ›Z*K2TïÓŠèÿœ×Ƽ³ìÒ]‚ìS±fÏ;Æ-D›D¤¨§1ªˆ7Å»WÍ~êîš"o§k Pv \Úf ÅË‹uÑÊ TèOåÝ>Ï.vû‘Fám©xGö¾ŽZ.ƒ6Q—l^â»-*“s ¡}F„]Ñ*Òyi+…v(ÿ”~H[pJöc˜m¤·µø´‚¶Ó»ñ°î!Ê£¯ø³Ý‹â–¶ àS»×%»~¾=Ò:Qö©åVÆÜ<9°±;¹íîV â£wŠàq¬Ž‚ÝÔ¦Q=;À‘ð#­3AÕ”]O zwÚ’±¤Ü4®JËäÒnÙ­1®1:Ñs}ðÐ\AÍZB}»(Ú{ aû¥ôc Ô¢,òã¬cÒlÅa“%Þê‘q¾·FÊè¹gÝ `Á7Þk Ã!§Ú?ÿ†[YuÈÿ>75¥ž·¹Ùâ,ùs~¡}}c·ÿäÚóÃKüÒQJu˜&”J¾U–v°°œ°H鳃¼'¯…œ4Õ­t…L?ÓÑÅÔÆFÇ'4F燹«'£5¦{‡‘æ»vúãJà åÃÅ‹¾¯¥EâíNÙÞn’G·Lh›SL×Ö2$ä7©1}ÄzEÀË-á>Ê ®òÉ3؇¼UEnûXöJ¸*2?Z+ã›Rš¸*¥dIxypÜXëÙ#±©(×€ú’Z†ýsós7}Îs¹]nq—üà]:ÄP|¦³ô5"”¦JÜ~³çüýÌVüÝy|qhü»3MÑôš3K©,\G ë1åÂÞ”MMT#lñÂÀ¾ §º“ü@Y4c›ß‰l†î´ð6:,iyy}å6óÓÿý¬Á·/FK™¯HÒ5È+~ºë»W”_\fššOVÊ7ÎÑÔnÚìw’j’ß\vkaœ[ê\%¬@œLÝVÙ¢£À^~mN;ßkž×ø!oÊ/æøT‚_ÖÕ:.­OxñÐäÙ.1ØÇH_ût¸ßí™H ŒüºóaÏt§§Â綯¤m)¸û¢µÃ¥bœ„ŒeÚXvû¡Eh< Dq%MI2_U©íÊ`*¾“îr³¡¯÷“ËM°ºAÖÅ:šS³q̹bºÏLßvÅæÚ£¹¸.BÊÖkeŒ´ ü©7|Z´„šÐÚ€†ž§tÐöwQgôá„Ç_‚“rŒ¶k¬»bˆ1~ “Jăt¼ÁÖ¨;±>:ïVÍNɱ)‡“·d*®íz̸“÷Jü Ræý2ÐTŽ-üo½øÆØô îOõÓ¸Šy¡¯À­ª4½ÐÚÙP°Äºš6å­kã¤`vkç¬#(—øxàCí䤡6™b#¿¯æ¹UÉ9uïFŽÁbyú/?Ü;å05|)yF"yDȈ£b\Šc/¥%Ì›iÀÌ6¼ X­Ÿº{âäý$‡e2ïÃÏhncÜì\Ggç—v% Çég ¯4;TýnÔ [+ã­û# a6åtrO%ëªÏNZeìlËNC*Ó‹},*¶2*í˜(Ó–™ÒbuwpS°Š´5W0ôu\<ê=:kÆ>ɤ­ó%zdR"¸x0üÊò>)Á¯¥sîÔÓ5Ï<÷r/˜^0xÛ”ú°¹?ÐóQë <'¼#óðw2oãGÇL%KªG­R$ãŽCùÚãñ3s*f^ɺ±QVëŸ=v•Oñû”t/Ä|ÅXÈMNãØnªÝðšrÐa ¿vR©œ?6Ó³_´\… òÒqVˆJÖï‘Ttñ³\-roË0Üå}Òx‘ïÖšîU'¥²Óm\o¸ð#ØŸâD+<ëõ•¨heW˜P5Oå“eÐ0ôÊú® 0ùû~EF;«@vu¦X6H¦~ηþ@/2 ^$PéùˆÌŠ?Dp¸pözh<aad =œ¬ð!iµ}ýdeÅ÷vÀ\uÃüc­¼ÞL ž!¨ÊkoR“3ÞgÒr“Öÿ@;Ûi”-ճц»âCËÉ,r2|¡GÚ¥îü>s¥fuóã|{Cð°ÙEÅ­qòëW´FTOYƒN‚^‰Ÿï‘gß—'Ó(—æfº^õœõótt+™†Éý’“&%Ùë‹~•OÝ£…çªõÒ‡[rö ´¥ÞHñÈdû”\/ŒÌ äëÎ× Ö(鸤 ¤®Û!1]Ž×Wÿ$½È¢4Ò˧ Þ;› Ž™œ–*i!è‘ɦfÿ)o˜Ai›3í*ôÉ ÞÐîqGäkUžyf]+_Zt×Ù4%í^ªºÈšnÖ1à ë/‚å¬Éf»Q¸q‡qÚ³-®úãÀÝ OǨ÷cé,³7<ÞßÈÞŠJÄžA¿Oè™êpRÛt5âàñxHˆß<êÇDÄè[3P®qàs~æéäÉWRìox@ Œä2òV ÂêÅÙN`cvc"òßÚÕåÔŽËÔžIÜ£ËYæÓXÿ¬}aJå.GåTý³‹ÔúDÌÇn³‚‰Û;Œb šn¢8Åê3¢¦+JD…°²éó§ñ“‰Þ¿‚¨ˆDÇÒmRLíì:f ¿üÔ‹¼ Pè:’à–d¯eì*¾Q'¢Ÿøéí0GQŒ»=ÉÈÛãžÉ$êähT†êR+³‘/®±ZÏE+@”¤0«æy}צּΕ@HÿöчÜè»Y« š Mýˆ0È[¡³Îj¸^×°¢‚Û.¯†7.<4·„9ë g;ÇÉÖâè`\B|35|&}=W *{n.r¿ŸžÅ3?ÝéNáçŸÇ•òpà’Â}êѲ8Å›üup¸;yœ]cL\:¤šÜp_T?úºhH(ÂÔ¾M¡í@qJêàü<«Àdí‚ Á9¤AÓx*©hÔ‚ 5èÛ|씹ž[e‚È»)$aõ’ãL&_Hÿô¦ÍyeÎ5E·Û‚wLðƒÃn·ê ÿ`üО9@ï>ðƒ®‚.ì¾Áb¥ëld»F9q3¶òžCfö|‚ä Ö>ª7%»Sô™E9d4ŠÞù®Hr&®£|ìžë&ø‰¨çAÕû8Œ0€8z«1F‘x6 ì üÁPzØfÞw9bY!ØŠOÕAa0ŽK’Õd¢òñ^ƢëÓý }`XÈ[ ¢³k˜“¹Þ­åY‹ê‰‹#úÓ;ègµÅ6<¯Ú?í¬^÷\•ä!÷ä!1Ʊ@&‚s~Ö–¢º²vµ/E¾¡»ù¸¶#‡V™©Z«%5kOxæž‹ò0R¿n¨.ô ¢t0nÚ3;ÁÌŒsÇ\ÁI¹XóQ)Þ».Ä+Ò°©€Ù¬“ÔÖqäœßŸT‡_”‰ŠíátôHj¼÷‚,ß="Ó`G?¦ê¼¥5àt鯥ÙQ{ò‹j5Na®5U«­x"ÿögVBAÛM+Ì·‰dužDì|}…¤–ûqs7êŒOh‚…ï5£—®ÈSû7ª=µ+N÷MOµ'ßôSå*ªÎ¼á‰p4NØ6¿“®óáöšÚ%‰´ÞóÀJþ…Sʧ]1€ïXªøPÿOææùPž‰&‘"ÎiÝ€WW%?‹&½¥;q¨¼bس­’ …~¡ª{fÑ÷x\aÖ’%8(îñT|%ä燞ŒFÎÈ1Æ+ÂïÇÇu4rcIÇ jZ£ö·î:©ìyž%ŸfJe$Åùhm_~âtÌVx(2*7šÛór»|ñ&I ?¯ù>ÃËÉ%#ê‹P¤:^áb`ôò|2ìÎÄ#X£ä‘NŽ×kùпW´GîóÄ`ðËGë¬WZ÷7~†ZŽqèš;øñ}ǬWNÂ*™ŸúÙ?.`ôŒƒ}qÅA ŠIŽxzöNôz3Oc3ÂÄcLn©¬v|VÚ½{Mß$^/ß×÷•a$åÕŽ0Cë!L{AËQY~”ˆ¬+s›¯ºÿÇ®Z6­UF/†S›¹ÃùÂwäh?¹¥&ê¡LðCaÒ/yé×îaã<ô¨,‘õ½à»áì3j M ‹ÜŠ«ÿ·ºy‚$\jH£Y²YR;ïñ¬~ææ]8Ó ·8ñD˜ž¬³TG4ËÚ‚,Ï+Ò  èÁ®¾„ å Ä¡É" oޤÞô1õ8tÓ¶©cÇä“å£U¿¬œMË¢3cÌ ¤ü€í«os…àÉMÍ݆Å_g1,Þþ–𡼆#¤~«óõúÀÞœYà êØ'¨¾ëùîçæ›ÆWR³‹ƒwLIs”rÂÖ‡\täßôÝâÄ)çzKJîy]”,gp©PÅ&{Õ'ŸZ~2’rS oÃW8 ñÁ¡ÆíyV)Oä½j&b3voÇ5œj‹Ðg‘§«Æß–¸úv.´Î.µ«rÊú'k¯w’Dø‡“–2™èƒuº¬6@s ŠyyË<“ÃËbßm&®W ¢=4ײî›’I{Tw»­ÛÓ(¾ÍÒPê Š,âê…,¡–_~o~Ó&>ä UÄŒV2±°Vò#|£ÀôøîÂZK‹×s?þƒ§Â ЦÂæoZïbÌœÒû;Û]Ѧ‰Õ:é¼Õ4”_c]›e Öª÷8£nDßÕ1öwÞv²,§Î߈DÈ•Ï&½ pk»O1’D/(&ÖÛî ]û½¶[#¹¨UûT Ý Î‡rÓTÌÆÔž]•Zî$±Ï6Ç®j¥¢’º!Uû z~™BË Wd=áýÒŠ%m`—çŽÏÖ®H6ú2Y | Rß#ãÐ/{ÄÖ#œw ä<æ\òZ“DJ0ËIû£ ËfRÅø³îÍ]BðÌ8ì»2ÿ8Ñ‹à é;xSÉ*Ýq«-tjÓãmK¥h ¯YÑC´Ý=nÊ/lœ4cOzïx}(vx/c»„é4¥ºƒŽrÍØ-WAæ½êmÅa7¾WÍ@(Äh¥ÿ“Àdp־ȱµúg)þtLVlZŘ®ê¹ß6³èÆâý¡tª‰üC<‘rnÓXoó“eðK!/–ÐW@S›w??€°U"Ä_ÞºÑGÔ"ÇÂmz§òZΰ؛8ŽTmÁݦ„þý”uT¨X~|* (ù3ôšoC+÷¯]'së”êyO‹|Kß 4Ÿ¡Ø½ >¡~“+Á©Š[¾r©æÈJWÊTZ©ÉÑ|¨6áõÍ­öMÚŠ‚Rå‚Ê@ÔÏH³ëŸãëúCéQQMQâN[ÊRïÊ{q˜;R¦§³pPw]uC§†cüið‚áS·^ÉóvÕDÖ/” Ø#—µš1í&4Î%€‚äšzÒ× ¿®§×÷ž…Oñʘ´²áÅF'IÝF2#‡H)··ÎL¯$‚jlZÇ\¸Æ­qèçbv{§ßÿØrþpÌ™Š>—$F”P|ðî:‘‹'OU®Ø½é/ûP–u6ÿ,Ú³DV¾çrGù{ÞĀǫ¡Š4eôQ:Cñîê9Û]ÒHªÄü¤-Úôè‹0…X:Å‘²ûú8ê§cíªB2bu ” Õ(8’„{;zý°‚½ˆ¡¨ ‘g/ÕNºÕÌ"x½Ã½v'7d~ç!Mú2ùVVAFôç&ǵ :%‹G +fÛÓ ›mÊ8>6[ðÙ%þ² å= ùešS¸Ftíž›‘TÎcê«Ë8¿³ÑpÅ'G[Lƒi&ªî¼ä»_N'ÏFq‰l qyÛHÉ5=ã#‚tzERŸÆœ™jŒá²H´/[ûZ &ã K3¸Äj]¤ž¶ðõ#[t1â{¸´\æó¤•™é\~£á.g«ª@%8%4Bë……3Q4}éÁ]Å´*úÆ+’©}iÖÜMƒâsçaò¤[önKÇ­AìÏîWSlVîܾ=ÕHM {:J ïh¦AX¸’ñF-'â£\%Ì'ªuqÙÍtð¹éCî>\ÄáÆêõò÷d¯íWÇhãÃÞzµnó7£kéœüN;˜nm^ð…ôi?–¤ää:…e[9ÕÉÚ(SN§aš»ù.Tßö”º˜=©~œH]Çy‰ä~UYXÞ>¯#ðæk[íˆwÝ—±˜B£í'T”56í‚ å_~‡Gg¯™SsPÈxDàÊŽ ¼¾ìeó>êl}ÐGrð±ÓФ¸kÌsVëK-díg©ù…‹6¤-BGyÌ öeq¤Îe?ɰyÙuœ:2ïÙ IFüBf{Eg"ø5—L±Ð&©bðQ¹O¤ò€¥/M_zLª{5BºŠw¨£¥Ží©I¢;ÅË5‘RZ’iIAo5å©äüuöL*¢N_®¦'r³ +ƒ+d«©N÷BöÔì´´²‡<ÔúS^ȹM˜®E7䌞–½šS.åx§;ŸœbܲûF2¾70‡Üé~‘ô€ ªÒ§[Rú`i†öŽ}Й²Ml„Åw\ñÿø÷y¿Í®Á<°ýþØ5L§=‹"DiÖPŠ“Ðñ½~¶6ÑïA@%ÇéÖæ\V —6^Czº4²·úlxhT)^ˆbp›v!NÈN á(Œµ¦mþ4§€++ÕYðL”ø>1{v9¨¹ù¡­÷6K’PÄ ï½ò ꪴßDò FéG[%LKmÓE-s1ÃyE9Y{›~¥¡[àDÈUU†‰Ä½¢áC»Jà2xP¡7Æ$ËBºÌà“Ž¤}ýãùä QN[ru¢¦œ u; ”wÛ¢lîÅ™èÇŠštÛ–€º¶Ê_Ε†ª0>DkPrHwÑWœ#ŸïÓ·ËM\OB¾û$!æ¢7u¸¤¶òLxZ sf¦»ÖE×úÅ(ÂÓjûˆg|€7½RĬ%­®¹ʦçujº¦óµÏ>°>}Øñ2gi¾¼=?O÷-a› þœUEÕ^Ɇâ‚&r—;ušÜå^ìù¾_ÂÅ\á%ϾßdƒšðBDuI7-æùæóŒÑ6M­g‡‘ à‡ååðÜøDÊE|ŽLë·Ä#ä7}D|K€#ŠOkÌãªû›îZ¥ëñ¹Þ˧¡nf=Ȩ„"ЯIÈ%2ãü•Òâd`ªŽâD)×Zb¶zÅIÈ+öÊ[»w”I¯±áŸÌð$øâaùöΗ%jánÑŸ^!='ù.?oD=YÙ#ÂÔH‡G¸†Ë[–Ñlµ°pær¶ ù¦_¸òZÏaÀ¿òg÷~çVT_—¨qã¸ÿìHš·­êCŠkÑË3ª7²^ù¸Ê~K>†ø0ªìntX’ ‰ ·ª™ynln¼Šúñ–hr/ÇÂZLUXi·•;Âïë³÷ æÁûéB{=¯I›nOÔ)‹×w$Œã.TcÍt_®3ë'‰T#ƒÓL=Y@²|•*Ë`Åq÷Ùé1§¡ Ô›©YIq±Èdýv-ž•Ô˜43Q]gú¢¥WH6“1F§ÂX‹Uù-ém4¬´¶î%˜ÐFv?6ÅÚ?=” xï ÆtaSE÷ôÑb¨×|>j;:hzOÐØnŠöþGN̘ ;°A$’"¾âçqŽÞºBá( õqºÉÊH –å.…ûi'ë¯ê&WÓ&Ó?¾w[¾G°ÒRP^Ìv³Ð±¥0°LÕoÀ°0ÀúánK8¿…LJúÁ˜‚")y@Ú°€\æéS mñ´Ö~±wÜÓ †¬¤ÅR6Sþîžä¯þ]!U#„>¡ÜÅÑá%#ͬï˜u ¶ðûWïî¤^ óÃUPn„¹”MØ®´FŸ”i÷°OpáM¶%·þÁz¿ìóÁÐRXîONèC»=Ï„)ó±‘9‰rB­þ}ÜM¥'h=˜º=b å^ĉÿ´:'X¿DF]mv¾NäÁ¹0-Ì {v©Ío§\Xµ§þälضAøƒÙ†»…WKö%^ !ÔÚù§f^½Ði¯€×éö’ûu7?7œ¸’¯6¬ô½ªŒ|Bè÷Ñ0jkÁ¬«õ¸¾½S¼ì{8þÄ£z ”#üåÌHÛ9”!OùÖ>¯ï'_EC&æÍÊ–×2›£Jäþ9 ŽTHÎo1S_â,dKÒ”®áT„+§êAÎÏ'AýO.<Š…£$wÇ&®sS,²„.ê­iQO{+µVößäµ8#ÒØwis_X¿!÷QMÌHHUƒÛ;M7Ð `‘Í–gRfIS{Éç”é±ç Óâ ô3„|µ*æAªÕ³mL¬÷¤U®ð­aK0 õ±®uóžØ‹fyïÆ"aK½KãÛùcÓ&£y‹ 3² iÀ`> stream xÚwuTTm×>]JHIshPb†éî‘t˜€a†–’AJJAi ¥ i¤¥;$TêCŸzŸ÷÷[ëûÖYëœ{ï}íº÷µÿ8<F¦BÊ0Œ=\ƒö ƒdU}}m0Ä„A Q23¤ þ·ŒÇŽuGbвÿQÅÂ!—:5ˆÇ%RƒtÄÃîr™ A¦(îáû¯ürŽ®²""ÞÞÞÂwa ÖAA@ðFz8&pw8Ö ~5 @\àõ&LƘ9"Ýÿ°˜bÞ,¸T P8ÚýÒÇ ƒcËô€©¶`è GÿÖû üy;Xüw¸?½B¢;C PŒ‹+í‹D;$ jè {øx4ì‚rÇ\úC¼ HÄþð»v ¡l @.[ü³Aw(éêá.ìŽDýjRäW˜Ë{VGÃT1..p´‡;Ù¯úÔX8ôòâ}Eþš¯3ãöû[D Ñ0įF`ž®"·ÑH7O¸¶ÚŸ KÙ?:¸ ’–“ànÜê(ò+…™¯+ü·üK}ÙE€Ÿ+Æ@\6@"à—2?wˆðÀzÂüþÓðo‰ `H¨`w@¢Éþ‰~©†#þ/)€EúV K‚Яçï“Í%É`4Ê÷øï)‹˜ª¨j«ëÜü«ç¿­**ÀO,‰J€°(Xº<ü;ùg!ÿá«F`é?ê½¼¨¿jöú“ü.‰ðïX˜KöÂþÈn ’A/_àÿ3å»üÿ˜þ+ÊÿNöÿ.IÃ…ú àÿñÿ .H”ïŸKúzz\®‚>ær!Ðÿ ½ÿcU0(ØÛ´= — ¡Œv¸$µX\$þ‡é®ôÃŒPÇ?hó‡â~¹=¿òK†_îÇ¿ª£¡دE• X,Ä—ìrˆ—’à¾Ü8Üç7Ma4ÆãÒ¸,=@`°d¿&&- ˆXüRý–À¢ @Äç·ø¯,PO,ö²Œßs¾,á/ù÷îÂá>p(ÙÄg ôV˜SeXÓ÷WÊÌÞBË¢âäû1$w–Ô½VCTå~|{ñp\¹udåá9êMÑ¢‹:g z³í¬úCYú‰ê&B„ý¢ ÿµüÏMw=OÑÝ–Ë{bh˜ý×z“nï¥O×Ýö–G³Ö5åxëËË¿óÕn _ň{HÌ --v/¬Y9«¹\åµò§‘·ÚÜîûJÛI¬ë?Ò;zE%:'Kq‘óéÌ䮺Y OÖÚ$™tï Ÿx–ÚÔNøCšÂ‚ùíÝûŒRÊt×2ÆOº|é÷ÇÓÕèVYbÕ¯«‚œ:H|!ž*µ§¤æ“6ر;*]¹ºµû~RÑgA%Úw¿®ø÷öm¿”– uõ&Š®¹X”áÇ `B\kä±™§TÇ!ö#oat]] =òhEô¤%žøñBm}þK†gxw6»âÙN•Y$ä—…¾;J‘wZ¼Í«0"8½ú²E Ì®éÏ7>Á Ö63s[w¼‘ç=önl¨®‰Ç±õ3CúÒwçÔUN"Û·}†”hUvÃe'w”w©ÜÐí’m IGß;(¹/JŽé¥`2gùžõGõÉt÷ôóobÉnïÉΖø’ÞhÑÎd¾Ù:\ÿ ƒ°ë{ÿ¼sù‘¸òd˜6ðIÆZ§ÜÌ {æ{3$߀ꨥގޞiË¢¹¡ç„Á=,]ÃÆÆÏ†Ê–ד—UÆYV›)Wj†ûvr§F꞉Ëe½ªgÉlm˜êÝg+† άî3.§r$`Y¤ƒ°bùg´GVü å™;*n3¬ ª^‚Š5 sŸã5>Òý|»—·*ãÊ”P¨ÃgÔè›V§5ܳõN… ˜IíÀ'݈gqtíRh³(”]\8r0êæÌl•Vƒo>ÍãO€ÍÏ:¯ p…wd¬ß¤Ù¨çºÕã.DºôõDàÔ®´Ûi7sUŽ“ƒ<·>R¾T/€=øäTÀ5ÝÌ;¢TT µtÕ®óôõ*)ALH?A$šfî›!vqØV¿å:‘ÔÛ[ Ú|o"6;ЖŸWk2]ÁçBÕxa“({sn—eFB±§I&ÜËÜÖX˜M0E !Ca E\\iÈÜ\†ÀÃäñÑwÐý<3ŠÚcùˆwûeÅÓàö%-­/ÂS\Py'ãê_¦ôÛ µ³Û$£pÙ0Ic«m«ÄæÀéN ‚*Ý?¬H|°WÝœ8Y;ç¸ÝØqÝ:Ÿ%oþ¥¤c€˜o“ËP"´stŒ¹Ƴ™ðu>u$ÜÌñÓB¢ÔÄ»,KpÄ9#”§2ûÈÆû@‹¸šçuø£{†í\ÌüSa¥o¸}ùÃi£ï tñ¾s¼©BûZD÷ƒž® F¹—ÅîømcsPšj„44)1° Ñ®¿Öµ²o®b°TBŒ6¥6“žX9Ž?ËŽ:/'$T˜#kHëÜ–ƒµ!é 9Ð[åƒ>ÜmC†öšMÊ9Ûl?§Ñ»à„ºœ;]¾²N^Ç^·©f5¾"‰?!Ñm]jþ>g*}o³üw© DVH½ýªoò.l<èüÖ<Úæ'±÷]b" F²Ã©û·jãkskwê¾ïŽOÂ,âè·(ÉyM”²}ŸÔÇñÑÜ{4屪AePmH#¶ƒðëNl÷¹yÚÕHûí£'ˆøÍ¦IäÆÊ¦—øgì`ç(|„L¸yìPˆ~2=8÷~<ö[‰Ê·5MTÁG.Y»÷ÝÐÛn÷ÓŠÀ\3·pÎ×>É۳ώm…NX(*¦xàW”Xf[䡫¤çÙ¶Ì9ˆD¢”í(íJÊ«šl¤ƒÁÉ~Owtíâµ#!k¹çü-©Ü¢úkÉ/gºt ù†3;Y6øÓ³C%©È"'X³Y©œ«5yO}ùwØ Î©1-ÍK½Q3=WC¡Ì— å¿ LÒuHh7²6™FÚÚ¾¿Oµ%³…¸.¶©;¾»Ì,º@À!é›I4ü³eŠæ8/¦ö1¥6€_”/L¦êNüAƒä™¯´®Ç€¸[(Á l§V'¡+1ŠS ¾VÌޓN²%§¬\ã܃Addo9¿ÉßeX_ÂÜ ÞÃÖ~bÉ tfÏ,¸D‹¯s™4ì )¯3oSŽNÍôÁùëÌ=•:ÿ餲ç}lv‡N£È–¸Û{Ç­*ú!ÎMÃî8Í¢¥À‰sÏ•]Û$ô;·‡ÊÊN…òÏoqÚ+!Íhs3ó¤BLL&¸fý®#žZXâèæ~o³ ¥“Þ«_O .pªWð~}Ëœÿ{¤ÌµÐö‰ºÛsÎômœñ°rqD™ã™vö{_µïÏèÎDZÂÏØ: òg[¶¬hr®¼æSêU>Ml¾Rô­c°O£àuùÜ3û ^‘ŒnÎŒìsߎ£Ú)Ab·/£@4F°eýL5=XÁ±›W/aêhÏŒPõ®Šò½‘,øH¯EiD±Ê“r«úUƒ=>…mo0¸ù ƒˆr·Oƒa q«ü£õÝp®èWY§ÄÙ–µöІ"á¥Gœí]rqÂKGcyÚŽ« t!L$y5ˆÈîu£€æûЇ¯Ýmâ2Gë ã_PƆ6-v¼ŠVžéKsœ³_¨ÝÌ€²¬)VðÊï©ÒÖð] `$åû^Ä-ƒ´NÈ?‚Ç ïl£á?ÛÌg‚èc«¶û:²)-QÕÑlšxÒÊ90‹ÊD%Ý4-|ºÑΫoܬoyÔô9M-­ŠÒqâŽ`àÍÈF®0yºîæ0¦MƼ,b¸ÙÑ€ç¼yB´{Å-ó&§ÚG¸«tSâý½0«Š"¹ÁôÛ–1¡ ÄZŠ22×DoëŠ6m±ý󽂎G¦wk×bäâï_5+à’RøœlÅÈ e7Hl?m˜Õ^y ‘mR0ÄwF+Tʯ˜Íî®´µ9ˆÙæÞ¸û)œÏç R2¿bÊ='§£trÑráç³at˜§ä”@rc(_Ñ¢›õ•ú«Âýî×–+YH‚ïŒw|ZQýFÿ<^g,9í™Êž6”­ ÞWxz=øb7ÈÌÖ” ð¼t­)ëODL´F—§l VzY¸½ÈIG˜ ‰”³‡E©Ò×|@ GÌŽƒRŽ{²ŽL")B¸±RÁE#ÏäS4DzÈÉ‹x¾›—¡d¸=Ž'o×®ÀP#òÂá˜w~†¨ÁÏeR]1ç&zñÌOâì$‡Ó»SEE“•|\.æô~H[b‹4W‹Xsžû…-¿®çøžµU£Mxˆ.‡† Üòˆudˆ}󒜩2_ nùN2ƒÚÉ# {NÝÞͨçËu+'Ã0ÖµÄEÜ5üêSëŽÜG&ßUwá# œ i€üŒ˜edFî¬Pß^ ¥e¤Ê—@¾ÄR¶¥£´·ÛGB|6Ÿ:Êí)¶Jùí¨Èué–8?™—‰Ù" ùJÙE¸JÏ <°“,LRúûŒ„Û‰KßST}0/E]÷ù…lwO ¶yÂMþRŒ:R3¢ïZ#Î.cªÊ¬±'GÕ{ þ€yA v„³×ú{B1Ïsñ» »Ÿ'IW«ëî}œúØ0ÈùC%†P›øÀƆòÔÏŸn‹B:«‘4ÐØ‹ÉD¯Õy5ª×d·¸‡s"ã¦nŒŒ~Qè<¾Ôlóœø;xy1›† Á·a1…½ïúŽJÛ¤ÃÓ,Tö§©Ê­’> Swd¶s´a^˜­|'çªW?y(4;B-Œ•_Ò·˜ÿk¸=š¹G# ‰z>²FŒcFFjŒpr9+J÷Ó’%Fµ»SížqXž/6TÑ/>\…84Æõ@}-m ·ñù±Ò–¦{«wâ‚´q™[‰tÆ¿Zš‚ùu¤b˜~¤ ˆª¬ŒRÒpû.«c;iáÚÛç•{Ïœ¤gäísw°ó¥K­h·Ó â©kEk™dÂyœ7ìí×6½!Ѱ°.0Uô‰l ΜÙ+k=ó1xŽ»À÷øX"€jxvÊ *Óo7ÅNÂvË^/gýÙ -VMÝx˜%yŽ!-Å1–×_ÏDÿ@%Û¨íoVLb®5ÃÈ6\‹–7“´ â´dƒv*WŽÞVpâ¿=>_N~ì`¤¡ Äyñï>Ï2^Úñ6¥ãÞ yz>[MÄ.`|Ý6û‡*º¨L.±[«z–¢Ž¿Ï6máA¢.ø,R3ÿ9‰HŠìúBœ˜»¥œ=ºŽÓnÂ1Pº›B1±.|«ÄÑŠ-uSO”ïXì]§öÙêW¶ãÒ'ɤ~~±PNжZ¯åç4\ÐüÏ»Vr*4«):äÉN½õ†Â¤2Ô4ÏÃÙ?hÿžLÊ¿!Md5þMTn½á´‘›4ØÐ©»q,.9hþ"¬£É}¤;rE×&ÚÜ8âd8‚:ÀH…W=¥\Ǻ@áWò^„U7½wMõI§™“ð`OdÜ•)Õ,æijz'4þÓ=½ãt…è$Qr‰Ì±#™!ÉÆ,ß{þBÎl»>B˜DõÈWÁ¡WΣÅ!Ò…Sè]Y½ýͦxMC ܤ wr)K1¡: ²X4~¹"£Ò†Ù tº5а©0>~Gk)Ý´T7ÂaŽ5¶‘~Œåp£< lejß`3ŽS[Èó ÔLzµLÙ®-mÕñ«B@ ª†Tp;ËSîéþh“ÕmÉ‘ ¢/QVí ÜAw‘6Íüx«~ã³úËÅ2õö•hã„"OkD†ÍâVç|}ƒÅjØz;‰æ-¤ q ®øÜœXK˜#h'õ¶ŒŸŠúÈöåjêô Ê´,ΗD<¾SÍÇڬ̳Eà¶xp9hÔÉîT¹Bo­›v]›ìè ¡œÂ)ÝË w^.Nº¯öx.Ôy¿ýý÷V¾·‘òáTˆ ÿ—G®/-EßÁ¬+7g=—tYøùCT·ÉÎÏ_©Ÿò¨AEé|¯Å*øDK^¤KñÀ®ü¬€<sÕwUò®TàÐxpxèH'¸"ð|WÅj…»|ûúáGÉS” Sõ—!¾/ùƒMnùø+Ð]¾îL”«üâ_ßë뜔z±¿é?Í¢¼Ñâ5:CWEÎÐnmÕÈs:¤WNƒeúV â­ØÉ§>ÝPÔu{$È~$˜=;À‘ý51o¡f®)tÌÐCò‚w®9,Ç'U¬ö®¢N†SsÝÕÙL«Óƾ”žw‡WäíjÎOŒó(Q¦w)<¢“¨Û–n|±¦Œ)Pí©zÆW¶³çmØ1˜_B1pMùcŒ,AñÑÉcà¥]{A_ŒÓñûS¨ËIyl²GÆð‹„ý8?Sh—9llà1ÁÛ Þ¯S£RÛϾóSŠœ˜–|vu•O/Ü'Ž.Åp÷¶UØ_ûÂ2·Š_-:¨e\‚—†dê\¹Ï‡D„ï± wŠÞƒ:%<® áËÙ¸ƒÔØÎ¢³ÚHv»C9· dDç¹-£º{ãKÑ;ÎŽ̂ûŒM…¶à …‚ó>œW,5‹UuN‚ y¥Ÿwú¥+‡Ã¯„ÜåH©?iø8Úu$´ôÒÝ›§ñrª5¯}IÉÀ(¶ž- 8ÉFÇ> è>ÔÖ[¶í´†Ó~dý ñb&¸ãšÚ`÷¦{Zk”_?$+¬ |.“‹‡ZŒt´k D¢W>¾LN 8¦S{‰þªí°A-¿„ÿ’b$>ðyµ ¢7‡õá³öÑ@M«Û\°nv–7ÈiŠCýõG+¤å³ã ã«U ,tûGÉM9TqÁÖûyJá\ù[C8fR·ºÓ4Œ5_»ëkôëÌ.¤ÈTR0‰NbÅA†L+ÀúÕ¢ßÖzùŒ|ÙÀ­Ôõ8½Ý|:”xéa4.ðdíKT&è¤Z´²„¡<5¡ÑÊ” ýaŸ[#>Í, ò~°kWCo4m¾ØðTLЪ[ âN‹ûˆÉë¦ëþ‹¹×VêM¸mhkÉ€²¥.Lv'ˆãÉ¥u!)yë •bïU¬XíôJwr½„¤¥’ÒgŽTÇð‹¾P‹3)æ–½ÊL’† 6:a)k6y[}œÆ¡ÕüY°;§î[ÄíaÚÄ£§S¹ÐÖ+(íWSâë~Ÿ­DÙeIô_¿ÿV%ÝER84†ì+Ki–ˆs×RLHl¦, ¯$ÚµÃ3® ígÕrnŒ;© ï(zÊNñ²‰(»:%ÀÊpÑh]¹ÙèöJêÙ³^ª=vâU!„$å»^·ké´•¦ù.‹Ë:²yüËô ÒEÛ1‚×.^qòb˜n²ŽÁÃ÷âÄa¹?] R¥õ,þ`â›Ô×:hH“P¸F}«áTfÊÓ]Þ©ÌægyPpk|–­¾E?ᕜµye…ÔâvEÆÝ¡×I¤%&ŠSÄö—g{Ü-úàÍÛá`ò²ˆû±v!»ˆix¾¸ë£'ßwŸµrQ"F³]QûúÁ§’"Lò(ÄÎ5§boŽjîŸ%(:B tºá»Ñì·Å‚ƒÈcË-™2æÐ­EûLÍÖP(1„4Y_­·ÖñÛ¢3 2_(Ú¿q+ª’Ø-ÿÇC|`€70¾ GFÿ±³NÐä›UíKa’d[M#}GmsÝßêLq„ï%©?©ziAa(í>¿7í…{/U3ó‰’ïöG/'ò#ñ¨P7¦=Ãv0áˆÛ™î¶ä†Nó®ù¡R÷Äž«¸_•ßÉôO÷0hÕú0žìeëŸÄO`Æ<ÖA íoÐ «…– sÔmÖ5^ø×Dl|r¡¹),MâOõ’ù˜-±Tü»j¬èFvwÇ;€8ïëâYžåËOoú¼2Hú£×m†23r8ÜEݨÍlCiº¿§ŸÖ©ŸÍY6Vœ?\ èEtϾ‰79û²éÿmD—Kóü2†)1 endstream endobj 104 0 obj << /Length1 1669 /Length2 10661 /Length3 0 /Length 11760 /Filter /FlateDecode >> stream xÚ´TÔý.L‰¤t×€t]ÒÝ=t90ÄÐ]Š„¤H‡4H7’ÒÝ´tHJ\Þ8ç=ç|ßZ÷.ÖþûÙ½÷³tÔêZlVP °,Ô ÆdçH©h9œœÜ윜\httÚv0øoNìêfuú)W°9ì“6‡=Û©@Šîäò qr¸89ÿeuH›{ØYTØŠP'°ÔÙÛÕÎÆöœæ_ŸFK&PPŸõOw€„#ØÕÎÒÜ  b³;>g´4‡´ –v`˜÷…`±…Áœ…88<==ÙÍÝØ¡®6¢L¬O;˜-@ìvõ[þh jîþ«3v4:€¶­Û_¸Ôæiî <;K°“Û³‡»“Øðœ ¥  Ps;ýe¬ü—+àïÙ€ìÀ‡ûÛû@vN:›[ZBͼíœlÖv0@MV™æc˜;Yýahqƒ>û›{˜ÛAÌ-ž þ¬Ü +¡0nðïöÜ,]íœanìnv?Zäø#Ìó”eœ¬¤ ŽŽ`'˜ÚõIÛ¹‚-ŸÇîÍñ×fœ žN¾ ÖvNVÖ4aåîÌ¡ãdçâVþÛäBû³üœœœü‚ܰ ìeiËñGxmogðŸJàðsþ¾ÎPg€õs`;kðó?4_7s0æêö÷ýOÅKh@ ÀÊÎ°ÛØ9¡ýý[ÿ%?/ßÕÎ `ÈùÌ= €ó¿?ÓË êñþÇüÏýrèéèJK(²üÕñ¿u’’P/€/'€‹—üƒdüÏþÿFÝÜîï2þÃWÁÉ ü«Úç1ý«b¿ Àø÷q0þ;–*ô™µ`ã?$7âäå´|þþ?SýO—ÿ?†ÿåÿFòÿ-HÖùSÍø§þÿ£6w´ƒxÿmðLZwØó¨@ŸÏÀéMA࿎Vleçîø¿Z˜ùó!H8Ù@þ=F;7Y;/°•ºÌÒö/¶ü…ëüqe;'°:ÔÍîgÀö¼šÿÑ=Ÿ–¥ÃóÓáöLÉ?UàçËùï”2N–P«?NŒ‹—`îêjîö¼äg‰à |¾E+°×Ÿ$p°;AaÏ.€çöüÖPW´?6*Èà°´sµtw´†<[>ëþ„Ï\ãÿ[|ÍñWãÿBøþл¸?ïëß6\k¨»ë?1žyÎý·È#ø,9ÿCýìàüšóYz~Šœ `ëÒðÿF]ÿ+;÷3 qwû`Ï}ü“š÷9—ØÑÎ yì¿QÞ?Pð Ïáa¶®àÿèô¹ ˜ç?U¹žæÂ#àð»þ¥ÿ¯EXº»>— ûóTž·ô/ùχ ö[¢-ÌB-…Ãì«ÃZo¾Jy²m¾™¢Û¥0±ù.¸¶¹ß½zù‘©2#dÍõJâã@öò¦ ã¥ø"ÕƒïASíËw͉-¿ýîMã5'¶[ÐæÇ ¿åHÔôR ’³i‹ïø=¸øé; 6ÁS¤Ëqqx¥þïÆ³GΫ¦·tiøíì¶ÆN%Ÿú}é$[ŒN´QpÑ4]®Eæ 1 2Œ…÷Ô kúòj 7{ì‰J1žÍÿ0†»À×`ëÃíŒÏÊm.·Zb ÄKÜá z_ɽdE¢9ßâ‚åïó^M"TÙ˜¬Ÿ–Ù°Ù÷¸Ò+í4£œº«<†Ú[9a dßñ¶«JðM]iÔ*[L£ð`UÜ`нvkÉ•Ö+ÝNkûOä‹‚“5O€iL}Ûƒ ÆNßßµ.+ƒßÙ®ß~j¾hÜìô‰õŠ‘Ûx™el #——„Ò)>½(… t€À øØ¬‹î_˜‡‹ø†œIÞ‘7ò99 ^tÓóæø}ð/¯;Um_ŸÆ™ƒxâ‡>…ò©>Y|ûtzšObË-c–ñ–17]¦ó}îj,ƒ†p5¸n:´”ãt½œ]ºÄ›J9aƒwþ¬¶€GA®{{®b Ò(ÄÛþV0ýX·\§%Ž7Oƒzë0Šmd¯.¸_24¼GðÊjo㪬LÒ„à'T÷ö´Ü3BràÃmþ°Ü÷íšAÙnj©rQ…×›3ݨ¯´%=¨b±@Úoî6G: ï=YK¤é Tz•~мó[0ž—.è&ù$Nøþ²!ˆ«ºMí†#{å]ÆÎâkÓ¶ݳGRI2-í›Ï¯}Öáºç&G6©‡&6i}!åFE<=aÞèTDH<=õÜfPÚÅÚÆèÈGëÈÆUO‰ÚEòªôéïàµÃþa$É×/Û¸9ñEGåù% «Ê,øO³‰Yö>™âwT±o*ÜŒªWñ/EÂËÞuèÃÓ®I‹Eè죎ïÜó}Ìžô’ËÕpê+#K‹$Ù²«¤›MÊ¢ž›@Sæ‚#ª2ÌJÃùµYÈDm®‚b¹‡…·)2Àß_›|¦‰§Î-“KÆ&T…À³þ8.ñJ«’ýÅBOµ”â(œ"¬Ž¾Û—9ޝ íu°îx÷xÇÁ̬ü YâEGÉà+Ù9OYÞ¤XÏ3­Z¿Ê§`¨†Èèîè'tô ³€z‘LJ g‰Jê“HËOǓچÛùŒÕÛÓÕœ°µòŸA(uu_¿¦‹ ˜‰]ð'¢›ç¼0üØ~!à¥.œ¡¸x»‘‡[ÄÂæKmjôwUrþäþ¦#÷Â.°VV^=1ååML9û`IIg´ærdÔܼãÆãAm˜nz­ÐüùëIMå=Á÷Ë(RÎ5вsÅ[•l}Î/ñ¾|kÄI³F7Ý×l—4n’7׈I™Ú“úJìËJÄûL¬DrƒJ¾æR’rYmh”ÞKŒ¯"•Þvs©qNé~ªàSP§¥…üþy÷¥‘-‘hݰƒø`<2uÓGŠB¢4·´G·ã×}At=åß&<¬þà2†–v¯*LÄl‡°Ü÷r%e‘‰„ꀯPæ“­PJú½ù®C:W²˜ÅaÄÊN¼Zçžô%lS_+³ä|nŠ÷‰ßô3$l1?´*ÜÛªÞÕÎún©á^¥•=,g ^ØÕvÐ÷Ó¨î;öˆ7€fÞ–ç—;.[Ðq.¤¼`¾3K;xÙ“›¸–Èa}Ù i£S—áÊ¢Ô•`Ò’TæØ)%<æÅ+•Ê’È[»Î„Wøu 3¾nž×{á$‹œ@ƒ—&à«9MÀéìÑ&bÜ2"XH<*ùAUÐMñf§>pÖÉãÛ0½·Töžø‚ÊrªLÚ®ÍaYT(âjÀƒ\ŽL8Ù3ª¤znoÊש›–X**sãr½+†ˆŽ_žš}Æ¢e‘Qq!¥·e4]Š¢`¤y]ÉX$¹ÿ®.SÒ²Vÿ’ kîIðjщ&.a#aA[‹lLbi›z&û^›2ÒD× R!±©á¨žnQßÅ/œJLmõ§KМ[4ð k‚óã"i¨k mâ]c¢;áY\X´ Ä—å¬q=g*nMÄÇM*¿¦nî1áC¢3B`ú5{ÿ»A’¹t«/·2V…¥ÏÂ9è­°ïJé0jp¹7áž®®ð|nëHÄÔ+LÜ 5vµ¬u;ÖV²YÊÐÃî·á6ÙoŒÞ¬‹ÓÈ®ß)ÉÞhçZüN‰o× þa„ØTtf.ZL–ǬP7î•Î$Ï ýx¸5çÑ}” n¸è4 ÜLŽ>»Y}Hwzå߉ò²oøwЍ±²juç"J¨Ûê#ùj$$ˆ=Þ&®NB>ÓïÖÝèoÏÕ2,ŒÂÇ·Cé;>s+IÒ8ìPí&ÝQ-²H»¤Ðúz|$ä f²¹áѼÙÏ s¼š6V`@ ÷òÒL g‰1vCž<­y. Ï$K¬$ÿXLJlK.=\½‚ÞK“JM°U&ÌÂïl”ïéE‚«¿!‡f²Â+S1²¼—H€KžmºÜ”,Wwýãôg#¥¤oâh¶,àÎ~áÎr¼¹@sw!*|Êè3P;A‡m©\ËñBìçLq Á™í'Áe½ÆŽ§÷>­lãcê¼t½/ëq&2iŒ_û*rö8ïÄ̤xŸz5ß[Ìî*)øž–À©Q_ª\ x×+µ}],7î ún:Jé+|Δp^Çþt"$XöŽ}>Zõ"²^xýGÕã—®ßGð åØ÷kÌ­‘"ß s?nöâ…á•"ºu ŠÔÉÅbJɘ…ëþül»Þ6ÐCG ‘8*[UoëÙ˜À ¤-k£™àæn½Vœ<4ˆq|Ï`¶YN‰y躺RM*8ìw›ºìKj G¶ÅÈ0¡­„Ósñxâ–@|Ðwh§¹@'xÕž+áßÕ-žò¶iÔ ³ñG@Ý̲¸Æ+…fÞ¥ÑU¨IF»Èd™K—í´ ÏböLTWV¦y*f¡ùoµè´kT›•øc+QX4Sg ] Òïûõ/{ñŽô]Àþ­UóäQ×óúïÔï~¡ÊÂëÈ|>¼xŒ>R í˜Á’f¨®…7ÐÅRXX‡Ñåm=%gt3ñ÷зßçØŽ:ebT‘üÔg˜^2ïª×î>cÐs œ^Kåp2uúˆ¨¨œKˆŽÂ˜,ŽV½›34ž†V^§Já«$[·¶tjân±Úå´@xd¦\™gŒ²dž¯“ ßÔë{ë±q¢¼¶Ü¼ó™'ñ0…6ÏÕQâéØí#~×÷‹s'ƒ5DæÔ–7%׃n—§Ðt| qßxïYtj))ª(_"KvP Aýá å'‘úº…s¸›zŠ-Û„ÑUåèt:½²ˆó7$ÜF,†_S@z­ÀÞû@Or=‹õhŸw±G(ÕÓjIŸŠÁwÙoq‰yL·¢ #ì_´^νëÖƒ7Ï£ ™ÉwDK^zGlã¹@ý å†[U_Ü[Ë8à,<!µü•œýù%•*9Óƒšóv¯!R6)ceíî¸DfÍÜ0"ÄúFÝ‹'H><‡Àoÿ%è¾VxIŽr[†[òïÂÅýÎÍy#š©2Cä`òhê‘%¯ê`©³(¿b‹äÃEâqd½ ã¯/ ÖW0QÊ ‘D9·ÞÆ`!*MZAc>|úˆ;Gï &»:’HŸö]W â(åMI MIsÈ·¯Óœ9ýL¡j^•¢J¤µã%ü“È—xZʱêÈ‹sØç‡¯÷.äRû¯š%HӃȃúK·f5ôšh‘å׊v±†a ùå­¨ý3ì úägâóNe®Éئ¿–ƒ»P0ÛÕî7aª/Ø”1’£>‰½0]%41 _²ÁO¨ È"ü+cFùuü¨XIòÝê)¸§¾s™ª2¾M4iòMÃ9þ-iÙçÀ)¦Ž·õ™·uç1Â/µH0Lš.…¶`HŠUöâûø^e5¡2ŒœrƒÂ£ÍK.Jú’´šN³°…³Dâ\Èœæ·Üƒae)`¢^€w„ôšKü\ËdA°ØVÔ¸PJ‹¥ƒ§Ï§™tŽ»(ƒ¡Õ):G× ãü¢‡Lßó/RåⲚ‹Ë%øi~úJ¿Õáà‚=¨ôöýºuÛ. úí ¨ ÷+Á:¯$cqãiìá7 ÞB¡žÌaQ2v?ûþêŽñnfÅ%-þW‹ú”•(ä{/Nì—0)è'$ö1XÊv"¾5µ à.â¾¹LVÍݦåó¨Ö"ÕÖÇyÿz‰¸BùÓÏhÑñ®X:ÜÃÕáu÷m ,‘•®œò,+Ë.¡Á¹ÇüuŽÆ—Ún¤+xÛè^¬/··”_|¤K`âÜ×`Ü´ôì=•yqQ{&­Ù–Jk„èÓÜç§€àõ#]iÅ¿OÍÍzXþER®‘ýDßµû%ú÷Э‰Rò’Ôý…_DÒLµ*]M½×Îix%{}g´Kúpm2ˆHAŠˆ.œŠÚá·¯ 6ê Á”wŸæ{Òè}OaÈÔàî·özÝøŠû‚JkX¨ÔÞçh‰®¦ÿ’Ct0Wíé8ƒÂUw5¸‹ÄËøkr«(ÓÆIvgq$ó‘ÚŽm¸›;.+ŒÍªJ_Ì€gWiT¥ªÑ©p.?=òkœªCL[èH|^ö„hWõ ý¥–^ç g•ȦÛuÐ,rpu“¬tƒmìû|6“‰Ö‹ 6Êp'OÿVôÕF‚q)ø_•×èúè1š/Àu>]ÓW§Ü­“Fhõó°£ÇGý¼f±D²¦é^1¡ˆ:Y~Q z u}ïÂäıtf±ùШRÃv—¯Dï…tM# ˆÎ 2+púNÕÑøÎ œpë,ú|?pߒͰF^È¿T_+d’‰‹À ¹×)ȸódª8µÁ‹"p™Z·²]ÐT¹LaûÒ¥½ £Õ ŠÉÖ$ž€¦„(Ç»ISî§1þa§ôn`pîãæâŸ×£ñ^1Ãh¹†²œüœÐòp‘8èþËyÖnœøž.±ëúIPÑ1QñH2¬èÑŪDåz^>öQw5..-å>]#¬ÖpyUD§SOˆ)dqIÌ’åŠn½ õHµeYR‹WðíMhg |LiM÷lOFÍžöÏÖ•-‹P/ òc53- ÚùËÍAdtâf¸ ¾*µky·¾OHÁ[z"×OÂV RdfÁþhlöx>£!È9ÖðCd†)óH°ÉÜ ¨@k[=j¤B¢TŽ!´‡°gùލf|rXæ„k~\^¿“¨’òH 4ÀV-Û™jôCJé.Í›ÿv nMËV9Èĺx©¿êôm9Eê {^Ýìéô}ßãû4y‹Çˆ§RÀúri½îŒ½Ê} ÞË,Ø4T¯o9òÝaÒ’R OäÙÂ#Ÿ2LÜû´ _Ê"‹è㺆³™#w‡[êÆXu=ì ‘öVuLàî…%iÃ[bgàíÏ|8ûµ¨ñC«„+Ñ>mɱ}oºtݵ{¸èA ä’JÏ9fLöIvÎt¿¨OE>|ðêKèaÅœ·'&íÒîÎ{à %ÞK‹D%±Â|O}„Ð0«Ÿcö员W|Ò‚QÄÙ.2,Ù+¢o(¾­Œ×^ur#!»Å<õÓÑa¥ýþúÚ˦TH8ûò¨X“]Õ+P5¬ÈVIu'é‡3k8»¼eT7.±QìȇDs™­naZ‘9UÓÉÈ8êâØŸ;ú~0Â|:ãhT³€¶—%]$Éøƒ•JŸ$œAÞ°ÀõX€Û}©VÎê@ÑØ+øpšÆzÕ·iÄôµø¨hsse Js!8þ¹]s _’˜Øj˜™â›Ô€‰F–,ŸîUL­ä˜@ÒúîÊ6HyÉ›—óÄðÎÚ%¥?äØcηû~­E‡t£+ô}'7q³Ã()û‚ÓÝøðžWåGõÙk;ä¨ä_ƒtÊ©\qÑÈYD@Bd˜"c/%2FA¬wÂÛG³šZƒ ¯eâA´ùzˆÑˆº4?C:°J'5ç‰É}Ýk²CN7ȯä.§}pV¿Ýv«Xsž,XËêºs6Šzzýƒ`¸Â24zTSB¿6Ž­ü7&ìí2§Ü1>—2üDø×ºŒÊ¤S˜ŸF¥l7QVR&†Ç¦FòjtÞɲ\–ì©NÙ6Nœh#+ûûS®„Jž#Ÿ“%9 7³[€ÈLÏ”òãR’z¾*ÏäFKxEtÜ – ü6Ã,¿®Í…(¿¸9<”˜¶ü!Ï!@26oxSûÎä•_ÿr¡äØv~IŒ WÈ}N4QyWò½­3¹äžÖ*#TŒ0”­§w=ý•¹˜QuζJÇ¡‚JøgC3ˆF’2‘-ì'Ís1â«‚ºœ›í0.áJà}' j›EFªœ‰Œø>ĆŒ]MíÏË#5 Düqcç<é¿ÜZ‰1fL³ƒzš5f¨ì{1˜NƘ¯ü[ûÔÑ>õCS¤Â(Õ´˜N ‰¶æÕD®mKó:Lþ4bUX:]ñ…&þšÝ8u¸‚4:¹)‰œ…·ÊD«Cˆúè ¹¾ñ—¨Çs…Óª‰Ä[jëûNîtÛÏøÙ8sA‡³´9Ç×ÎmÇåâ»zë˜í:‘æÎêÆÈi&£`±fãÇ)&Jh±­,‰ éTÆš0 ³öqzRgÆh‘fŽû5R)[³7Nø#¡d ü.êp¿³ ð8Ò È~–œS§¾‘w—¦Þ‘V*Ñ!xË“Ðð6ì  Õ`jýÄcµˆ°âøò?¾ŒqzMåú¤YœßO2A¢CÅ4ü˜æþ•™Ñà¤b? ü /Ùäu‚mýIìÎä2ñ2Â÷o£Ý»”€…  ²LÙ“5‹# ÷ËUuìÔt½}ÚW„ÔÊØkpm_@6A•àÆû!0è˜ékYs1ÉA¬$ØŸ@›ÕÌöò“ ,ƒRâã$hG'Fšÿ>Tä öh*H³ýåg.™_¯êi±mÊìÇõÆ-u,jô¸yB?wcQBÙ^„ÊEN˜Jí‘ãÌg†@ê"2•Òð8@ ‡‘+¯ϵOwóÙvO€ù3îð/ëO°ä0O”îÃÅååyãUøö•LÅ͉_iºÝ|•9³ ¡•z]õŠs$KoˆZøÊ–™_v{?¨¾OÛp¸ànˆ,é“ vwqJ¬ÏÝë8; u0IN -Sr“¥Þ*ÈÀ¾eÊÈ}ó58Æú2¡µÝÔ—$¤žú¥;Ѷ^†I`¼CôUeßÔ—¸pV7cT# ª©”ò~Ñþ®ž ;åU÷kĶS¢«&Øb,¹$˜B{LäM5Óš3ILþXqîX‰~®°iky“ˆ&IUfñáû¤o• ïûUêz,:f;OB¤¿_~N¹}˜A˜Kë®U}xcóQ’Sa…ºøj$¼\Îó³ýk1æ7~v_²ÃÔ ö+î"Æá[ÛbêðµÂäz‰—Zn‰nf_Y¿ry¼˜å–Ç_ å`kp]oµj(…œ`þLQ‘“¸y( ì|kq4œÓíílØ€@L²ézeû:s<0g¿@,yôRõÐ]cýØbîF μ8š?h@Šåh@yÊ{“â ®á¯×¡Úùš)‹»[GytwÈAnÔ´²ž/,©›ËoÛúŠÌ•ì;èVÒ… ¯…ý²M\öq¶þZ«öàWëñÅà@ÁÙøWD*â;"ï7¢iü£ëTB“Ñ¿ªÏêrˆ7½{ϺQº ]'5ôh'ÚÕ »]ŠÐ¿wä¦Ó`ƒö>•¯ã+Ým I­b¹'§k(@™ › ¥,!(SiöZ'¶.“/àðÃÐz²ýÜí·©Rérc€às 9,{![lPvÞÇŽä3;sÜFD&ýt0Y%ÃÃÕ&@…†³t¬ÿeåZ\òÎ2}#xÉ/ܱ\½ñð©þÆOŽÎÀOlÈ„ÅMïžÒåGÊTg-¢ž~:K+ÀP%a—4îý’2î¨(á̳ÁkÉ%5"Ó¸o”)Z|  оÀ7Wþά©4„ðã{u9ÊâW`w'M¯Gí~oÔ NçÇ/Oµ§®¬0ØfèÓ/œ2B+¬r[å% ¿÷ˆÊ,/…éꔫ‹6üìcˆ>qÝâÍ% ô§”Ý84ä~Ë tÆXiZÿʕ'Ò'Ó`¤íø9¨ð¤üÝ* ^.À¯Àv ÓÁ¹ÚéËì™Ë¹Zw*À·â¨Ë¼\1„(@ µÃÈ·|mSyŒ†àÏ()=¿(Ä!^uÚ#‚.l·G~‡¥ZÙx2&Á.\Ar,ïs—“=V]8q_ëeòG·ºÖåÒ´ßùäLP“P\míVê*•§ÿ°ú8ÂL¼ôxÕ¯m,¥ì•+$;A^°ÕsJ´Pù¤k*ØG\ž·þ¹kå@mwÅ-˜yÕÿ+i=BÙ§²$Å_Ñk¯zn8:WË|Âú¶çªÑ)e{gÑŒ§¤¬î$‡Q-eÖ¨VMÖ&&•Aôz°!²[¥›Gî7C^¤!5Ýd©úìVœÄØÓÕ'"“:zoä¸/¯Qpxç"}d>Rd‘hT…®ovxnHhbª{C ÃuHý“püypË|+\0¥8Ý9pN‚b¶yè8.`x½qe¸«„N¨ÝW“ƒK™º÷vø·ß£Tá¤Ëý¼±}P ‡¶6†ƒx9¹¬ÒS+Ïõ˜KÑk§¾"›îƧ*Ù¼‹5Ôcz™?cëîì(;‚ʺßx'…5c4¾úöè²\¯RhºâÏ{ã[f¾¼ .2±«<ݧ›¸’N–âõªdœýštÉ©~³OÅŠ€Ìn ê :úW·‡ÿà ¦öÃôÕùh;Ò ½ûî£3ù¢DšF™7wÂ)Mlï=x‰LÃÑ•<›a9FVÆOû=1†4„&|ôÈ׌üµ;oÙWÚ®³É…>ü ÷Ë à™•ëWK#gLÔw÷_8OBŠÃɱÅùJ©й®Œ&ÈxZý/ôÊ]¹§ûßeB;ª©Í¥ ׿@eæ7@ÞŒºo×ùQ9‡UÁR1J8 ~—d¹H˜<Ÿ-{Ð ¨ë6M‰;‰½Ögi>?Ö (L=L÷UãûÖÖ%[òd|PÊ-b.0œ¯ÜQÞ_Ó­øŽ‘ùw¿áº5^!i–ØÏ‡;Øcž&i4wª±,Î(¸Ë.,½ ëîÖ/_<‡˜e¿Êjs8@©øeD‚;€x|±ø k£qéÎy*F PÖØ¹fa·}¡š—±^‰4öÛÕí§;ÔÚm´ÆÄ ‡%x”¡l/9],n¬9Â×弊®©Žæþöºf@^ò§¨ž‹'NÁ_ÅBekÐ 3…»Z¡‘©l«ÊŽýq„Ë:Ϋ;Ï(Êþ<úçª endstream endobj 106 0 obj << /Length1 1391 /Length2 6189 /Length3 0 /Length 7143 /Filter /FlateDecode >> stream xÚtTÔk×/]J‡¤ H÷Ð!Ý݈€ 1ÀÀ0ÌCƒH7ˆHJ HHŠ4"-%-Ý) J÷7êyÏùÎ{ïZ÷®Yë?Ïýû=¬Lz†¼òv0 Šàò Hµ ÄB|‚¬¬F`ôGKÀjr‡ƒaPÉÿeWtY#P:%kÊMhx@@!PT(&) ø#Ì] dí ¶hó4`Pœ€UæêívpD ªüçà°å%$Äx~‡ä]@î`[k(@ÛárAU´µ† a¶`Âû_)8:"®’üü^^^|Ö.p>˜»ƒ 'À Œp€à wOà׸kÐïÁøXFŽ`øµ!Ìáeí °- Gx@í@îTm€¡º@×ýã¬õÇð×Õ€|À¿Óýý+ú;ØÚÖæâj õCö` «¢Å‡@"xÖP»_ŽÖ8 oíi †XÛ ~7n P‘×X£æûk:¸­;Ø烃!¿&äÿ•uÉÊP;E˜‹ Š€üêO ì²Eݺ7ÿïµ:Ca^Pß?g{0ÔÎþ×v®üÆP°›H]é/”Šà“@nÒÖ‘ÿWr#oWÐoão5ª_W˜+À5ÈlBýøÂ­=A„»Èß÷þ-;°-`rC þÉŽRƒìÿȨͻƒ‘sð€_¿¿OOPزƒA!Þÿ¸ÿ^.¿¶±š¦¦*÷ïÿ6)(À_^!Q¯ ˆ( &CüÿEÏüWÿĪCía‰?Í¢né? {þµ}Ž¿ˆÁ øw. ± Ç?·°E}€ÿß0ÿòC÷¯,ÿ€ÿw?*Èo+Ç/óÿaµvC¼ÿ²£ðê@a_†bô¿]þÐUdöpùo«:ÂÅy¨äïKÃUÀHaëø*ôÆ¿CAz08ø×ƒà ü— Å*[gÔ£Gáñ· „"Í¿K*Cmav¿Ø%(" °vw·ö&@­%‰|(Ú¿ àçƒÂ¨j<€=Ìà×>E~›êÖQêßa ?ê¡ú[FUæGxÁ~Ëÿ*mëáîŽ"Þoh úúü›å dK0= ³•zîTý¼ù¬RžÎ‹w}HzŒuýÑ+N^ßi÷ "ÜΊÌg‹î'ò)}ŸHæV•9Žåf¯}¿5Ôâ†7&ë7]ú]Y%Œ¬7|ýBÕ=\ðM¾¦‹Ÿž×HnÃïÚÍÏ$س½Mƒ5ÇÍCœH/üÌ«SYÓU2;6¹®¿Q!ªyçªd”7Ö8Æ"¸hœ5×&k‚šÁË€ÇEv€$?>#{=|˨‘ÈMà¿+ôÆ×lI0î|Âg¾ÌHþ‘†…ÆŒšó˜l`„ÍWaë¥Æ½)ßâ7‹î©T-ƒË./€-Ÿuƒ3÷ퟖl¬Ã’4Ôü˜êÑ1‹jI.ELI8ÂYs‹HõŸb 6pÝ$ð­.D¬šÈªóÐa¤æU \Tj#å“^jPVjm}¹?‚ YËóQÝS#J¨ó1ZöîçÒ'8? .¥çöJêÓ(7³ÔVÜ!Rjø’ÈD=~½¾Ë·Zit?{c ÏÔÙá6 tMùS©t8~zm˜®)T:¶º”õïÊ勎«*MºW;˜9 ‡-I‰‘/©9 V½Sx>Â9¹Õ£¥V]Ð>qݓݔQËla`Õ‰m’g·9œÁSÅ!5ëHù˜lÓc·Dh Þe)¯?ht.,‡¨pB$ýïßV­™$b2h{ža³ªê´NuÝã-láxú%ò”9cS›$AWPH|&Œˆá}¢®úÌv­Å^lm<¨pà§àÔ mâL¾¿€èf³›ðm‡Ü!>’’‚j±†LÜ‘à9å´¯ƒu­ý¢VPÏ„{–´nѵ­Ã™ÞJÕŽg¢D•J†Ï%+ɳ’3Òç ¿ëÝTþ1›ý šðÖ¯—;;&ø¥Áš„+I– õaâãáM9µË?†Æ„­>ÄÃpYÄŠ-–Ø|™HŸD\î¯k|bf8D±yßE/Ä~P¸r¼‘“á•¿vd†^aß–ƒà—jÐ!v`Eåxh;èûBž+aª ÚÿŸ1¾uEÍI)éEÀºKd'D0ÉL£ÈÌZéƒXvO€"IßÅЗ¨q·Ì²Ñ0èßΤ ïð‡*µ¸Èb Õî¾}ÎŽ·{NÑšß°RLnB2Ìm0M¨zíší­t^mØéëDf«E…•w ôjÞúÔ:?JFÛI‰_¯²­­ãÀaˆµ—çÄÚm”•j–Ó&ÈßSÐY,Sm5=ŽWïÀ1¡w õÞM624UÂ6wþvïÐc¾Èlï8$¯2¶¤‘M<øÜ7b/UŸÝÿCìô$òArÏm«£Å¥1®åø©ÓÛ>çZÉzƒŠéÐ(°ò¦q#IÓDîÉ뇻Í+u|–Ùª^͆Ñ'Àè$à2„çýM­‰ëâN­>4êc³ÛŠ©¬°¦±«Ò¸t]¨8ÝO‡/¥”š_öÓe!LäÐàé?ô›iÇSÍL.ÚšØËüF±ä,1S=BqòΣ%½.æï•àæáµòjx”p&†áŸ5µ3 "Ýûæ3±œß@_ÿæfpì-­±÷µ‘ëÅGÄÔe}\ä««AáâMkÑQxqRø&θnàÁx*†e¡ÆbIþÐÕ%ÁKäE±q¿+$"'â1èQàjg¾ß¿qnÇf‚ì®yþ ŠޱB†’5oYÝg¡Å11ÿ~‰“¬fv}ªYqénáuœÛœèÁºoU®Bhy§S: -q‹"e<«7L1̾å+ïÔ쇶™æ•(}LŒ!b„ÜQOçÆpE÷Ú+ê¥}Ÿüµd¥‘ÐWä' ÿŠIj<åa4Yå ÐÖcª»Ò¤‡í3[ 6'Ì¡_*ÝŒtÑŸháXµPÈdœ”áŒvj&kÇ¥„ÂJ_#µ[£#G<úî? i²™Ù$D§øÂ×W"7öá~:]t?ì ½;[…Ζķ,eçË9±ìЧàx[¸l+ÞÃÿWuéƲä…Kò©‡ ¬ðgUL¼oVúÄM3¹]Ù„OœÒ(×hêNt-›UoŒ\=c•B-S(:;qcÒg ,¯ÞïqÛCŽÖI‰4¹bíÒ+4@CQºwÊ mºp€ðÒ¢À“[ÂÔüAÓ´PÅÂ…þ9Fòv|Aš@ìîF+ø­ˆ‡R™å»ìê"þµÈêl ·ýâ¸ûne3ù=¼–ó±äi}e{Vä§-fciÐx?ÏRϬ“áJ¶ÏÀ]·/\½;\³}¦é\qñàHøOC{䩾nfôøì˜r=KvXgPC½ï5³f§ÒQúmW:þ¤0bÖèq<ñ=‰ô[SaGþ]H»•˜Ü R3zãη­üoäm¹¾£½•o·fÒó­Åa¾Ý˽Â9.[û¨UÑ>-Íd"Âw§5Ëß]sù’êÛ³5Ã4L.vƒ él¢ ÉÖ>„ º•°® /õ>MÐ4ÉuºJ¢ô•8).³×M±ÏÚs¢HTa1 Så¸)%y–]Àðá«ñó=bÔj:ºhVˆ?u¤…à©IÌI2æT¨ G~ åg}ÊãBUZÎÞ%Ù0\ZÑRµe ‡ƒó ùf+¯©pv­Y¥”n=ií¸GÙ¹/œ=z¼«p—T¾ªíjY³ù¬x±™–þ#äx½ûÎ#ÆÊ'œú±¤SNöPÜø~WÇ`’uDyšÔEÂ\̉¸äoDuµô5¦±rˆ ù£wJ¾J]bsM í'tìA Ù§_Ö/†N©’#PîclÕ<ž©®ažÊuò@üF¨yþ™Šú­‘ÚR¼}èÓ$´“tÏÌÖ”oo×¥YÇø¬7¼Ñ5Ãâ#˽µΡL-†ÎÞžÌ"U³J>“Estì´}§‹ChÏõýÙ&‹{?Ž}h4 ¶rfkžÖÈɹ+ì2Dxhþõ£yr¯£ßІLr‘µÅ ºÞ؆û½Ç“cåD«¯~®Ñ´[Ð{A—Û|·›õå?`C1‹Þ6¨:$««½Œy‚«Lx^It˜ ì2ɺý°Ô,,ÉŽ×o±jTð˜þ°­&§~ée‡NƒìöëÞ»{„Û앟­êxÓØ°\-<]—ÎJDê*Ò¬&iÔ Ioù<äuúï}Š~7‚»Úz㑞·ñôñâ8UoÓgùÛå Ch~§ÛrÙÇøwªß ĆyÊ\ª1V•‹vêE%dRÆ®ŠÅÔÂhåS'¨[@#G徫_ÞåÄÊšŒõëäó¸ê%/@†^ß±wÈ–R¿iYlO‘ÍcN›U›SòþR}WkW[†@yž§Kضp²Ýo›€Ö?‘(‚{]úb³’¢nIi£¹/xµÑui¹¾oùèËXÞDìR?Dù`_YcÀž_Ý@°è®SÆBQí8TC…x<]Òί1…õ ÙX†õ5êsdfœIÓ7¾R<œ¢‰zBXÔ˜(sáÇ&ù@•´ $;ßІ‹=IN^@O¨ÌêŒ7 ô7ÉãNº$º8 öÆ_ Ÿw^ž{ûƒ©·úõç=$@ZÌì§YOfívöÛ—…ߘ&JbeýÙ(HlL ^Õ»M•£…îa’#]®‡âÏ/ÍD' •Äòj×¹%5Çw"“«–ÏçTì°tJ€xªC¬'˜IhƒSÁÅ/ÞïÎ/¬u͆½Wjî-¬ã”RD {µ(B°÷6Jʪ²Ü€*íúö9@¬¥®¦è ‹”ï;rŸš361fÖ9º"èë,cÒé•°¾Ô\³ÖWqܵ×ð‰Xµ¼'£–=¥¦ãà¡ Vp0>Ìúy3UOHÞþc2¢É0³úÜ£k3ú±‡P»|c‰1þU¿Ü£1“á#+2ªy6Z«Œ2ªPi`Ú¥ùê”ê2²gë.?F$su]a…Êé1 »¸æ!Wblޝ'ÄÓ¦ÿ€MI.Z³%8)ÎS÷Vt½ñØ£>ÖlyZRÙïñ){f€É’éqáe­ŒˆÍn<€0ù?;L(½³|Ý€sêô5<,3‚ªˆ dà¸ÎKËwß‘¢X*ÖGû ¬ûk™mE<ñðIò‚ƒñü‘:ÊëN=å¹Î@½ÇJÊW“ûFûÇx ÒÅ袖“´W7,Õ81æôIö kC{“»//äÌá“Îh§ÂÎMrüï uK†¡5,˜]zi×÷}|¹5Øt±/Þß`g®sabGh\iîO'ùaq´æA[½É'ë½Â}lNV­-ªe¤ |ãʯlÌ46Ä‹R¨ĸ˜Ù˜Ð‚††²ÊÈÎbÙX~øªžïÆË>…´¼î(á ‰‘bóó Ó&¢nÄ=ûØ‘mq’Mdhˆ§œO¤;pÞ⻎èÑɦ‹e“{ŠN2j˜#q Êï?®}ù£2¬wd8š*µ.Å63#pÜ’}§z÷( ½½µeŸ™ëø¦Ó‡‚C‡#{¨ã²3íé~:)NSɨÓKŨT —b_zñ¢:YV… âÒ=Ü Ã6õ‘[ë)"HÖ\ÌøÐÍny,ÿ±×þ Izmí%°;„[®d-ä©á(”Û¹³wL3ƒð ÷¼ÎÍq‰®\žÒ äÓNÞ ¨d]g‡•Æðùv$r²*•Óð Õ´﬑Mëu·‡Ó:á¼P‘¨™O°§ÅÂ7wþîû4äQÿÕTã:!¨5Ûpô=ªm{Ä'³ •žøw?Š$nLŸàÖYni…ÀHm*ÒS/û#ÜS‘Ëå›ðQ†ÅÒ@ù™tuhöx_^yH5‹£Ôw—åúõ÷Eíìh=„¸K'mƒKŽÍƈF¢Çý€m O¤6ˆë–s"îW6Žø·FÏ7ïúr%¼®cÂÕµìÙ&”*ü(ºØ=Þ©ÞŸ“æÂÛ”;Ä›ý5oDkŽÄÑ¡Wœ¶yùŠ;¼Ø”‘ÇÝsV:%õð_Êzóî fß ÒLñâ…Ò{é>+Ìl<ÿÈé†9.`”ìŠÄÛ>w‘уŠä‡†Þï-lŒûÌ1wÁz­ÝfõD}â>挪mo£[ ŽïüŒN:iÆ õÎÆL릯µxh®*þ,¬u9xúS1Š¿ ·ˆ§cªk«¥YÇ=UÛM}üt‡ÔÐ/å˜Ýš’IÌüdÒŒ-ˆOÁSS"DØ7`°‚]ùÅ=ÈÚIáâü¹‚Ä]Q‡OQPCb6ÿ½­ w ¿éVV“¬Ô°ZOÅ"ƒž.—²0«ÕØ V°¾w$9 ˜Ø‚rðÙ!øXjpÈÂðDSY®ì3ë=œ®}æqܵ…Ã=V¥Ã·‘eÑq”a£ Þ{LRTENOùÞ2~HÙá8—¨ž[¹²Ãg*z! ÿ’ß½3úPælÅ.îI–®ÌGi^ÎÒÛÒ:;†³B‡G?hžÖD5þU•xeµÑ ‹;Ý-Ñèó¤~Іå߇Ö<.WEß‚ »®÷Uß)´B?ÄÒԈ̺Òe¸ŒI]Ícg'~ë0‘u„ˆx?—OóÌé¬7»‘c¯ÝçþèâýJdMkö(e×Á6ž[“ñSïAÁ%¼Ü›6¹ÁZ„THLÎÍ™4æPåƒø™cØ·ö¶ÇÜ%šÝ³ÁñQ*]ˆPŸNþ>ÎAçO|x²!BŸØjïXÆÝ-¥ˆ½õ¾ìd«/³ È ¯õäÎ÷4Ý!oüæ¡(x©þ.‡h'àõ­CHZ\úŒ™Ãu]¿¹è$‡Ù@ÿãv‘¡9Pȳ¹ ´ï‹0W}ÖúÉÆ8˜BèüúU£ˆòU1ûÝ+Ÿ8—Ô‹´'å‘s.F¾ÃÏÓ€fQ‰¯8ºïW]{f"t?ÒÞ—Ìuþ’GM“!£9h×» Ï»[ºUõ"µ}:_3Óì}Zsä­.—ÇdðA+ì¨HY4ŠôaÈ÷üªWfN:¿UðnSÃ0“!­íFê²{72i<œE»Ý—Š­_„ãS!ܤÏI6µ¥+nd/<6…ŬToßz¡O¾ +èÏŸH«¼¶ÿê'¶Ý¸µcEžBÖ‰.ÿ³ç½ \FnÏý«s t}|(œZ>Ýöô’"z¸ÙNSIÇ>4‡±b-PT ;HøA›hòq°Äâ»±›ãÁ%²USÕwºÆ·†&!o³ÙÓÙTb/˙ĵ´»cƒt8EéŸMKH޲ÛÐ|l–c7 5 YcÝK/>Ì2úÑ,+²ëÚ\[k’»²”|¸^‘ŒŒ&õ‡Á©ò)Ý·Ãú@×wÖ©Þt íG4f±})ÀªIóÜ­é¤(Œ ( ®hX./æiSyƒÛxŠxå5š´ ¨ª«xÔ’*Ë|0@T ‚õÅð_md€–¨À¢i€®WÓ»K†ÈŸÝûä[–;›%« é.q}‘{c>O(¨£ ½W¾ªù§}ø²w“é¯\w5Þ¿\Ã=¢PªÓ´Qµßaž«¦W~ìSÎb}ùóµåÑV\%†YÌùèáyõòÍkcDOó3-¹[Èdöަ=æRDö3ôš%ÉõYÓ%?–‚l§ÖÆ]wºâu¶ð;Ž4þ|•_¯ms_ïuq<~K§á§ èµl“PÌ5lHý°Š×E–f¯Z’ÜiWµ‘( û Jÿn¹þ~†D ^ûþX·‘ï¡iŠ%7mtÿ‡ae´Ö7:=¬È †º"Ž„ï»{¯m/ñÔê9õ1Ó9 ¿å©¨¹\l/<æ§!²CŸo0Ìñ†>>dóÄ–*~¡Š«I¬º•ßQƒëw¡ÁHi[V³duºTE÷nYâ`md°—ZM0,ñ<dòÜù;¼›œæ›*Ͷ¼¿'"ô9Á÷y¢ì— x¬D)·ÂW+¿lÑÑÁ8…Èðdýð‰vµï¯JSwجæÑHÅUkjÐÚ}Rq96ò–k㎛³Å¥&–<Ånü$ uæmØä<¢'9!à"uP´K~ú²îŽ(ÚÎÏ+»&ròp4Dâ„Æ·ïçxï·ÖìÅ äÚÔ„Vç€q^‡ôþ<ÏÜá³¼dì-Ͻžúo#qÑY# ãg82\zôH!,4> stream xÚx4\íÚ¶è½F¢×¢÷Þ{'ʃäè5z‹.D'A”è5ˆBÑ¢E ‚É7É›÷œóžÿ_ëûÖ¬µg?÷}Ýíy®kϬÍÉfh"¨ä„t„ª#hAP ¢gb€@Q! P„ˆ“Ó††Cÿ¶qšC½P0$Bú?*^P0cS£1@=$ í €D qi„4¥þ"½¤ª`˜@O D@QDœ*HO/˜‹+Sçï[„’’’øPò€zÁ `@Œv…z`*BÀp€ ƒ¢ýÿ‘‚GÖö”öõõ{ „^.ò¼_Ú` EA½| N€_#ôÁÐ?£ qL]a¨¿&Hg´/Ø Àà0„x#œ ^Lu€‰–.ÀÀŠø ¬û@ðgs !пÒý‰þ•†ø †@ž`„? áp†Á¡u]!´ZF8ý‚á($&ì†ÁÁŽÀïÖÁu%#3áŸùP/˜'%„‚ÁÍ(ü+ f›ÕN*H("úÕŸ*Ì Á컿ðŸÃuG }¯œa'ç_c8y{ ›!`÷½¡Zª0Ñ¿m.P4@ (%..@ï ~Wá_Lý=¡¿¿Í˜‚=‘žgÌÐ`˜3óEˆû@h/ohpà:þ¹"N0àu!ˆþc†:ÿµÆœ¿Ì`ÄÐþúüëÎÃ0'$îÿoøï#Ö75Ó7Sáÿ3ò¿œÊÊH?@  ˆ@PJ@â 1@ð?ó‚aúøX-„3 õW»˜}ú»eŸ?àù#^À?sé#1Ì…xþMô{@1 sýŸéþ;äÿÇò_YþW¢ÿwGêÞpøo?Ï_€ÿÇö€Áýÿ 0ÌõFcT ‡ÄhñßP è_ÒÕƒ:Á¼=þÛ«…cÔ „pÁ0ZtWx÷/; ¥óƒ:ÂÐ׿Xó—Ýì—Þà0Ô‰‚ýzÂ`¢€ÀÿòaDqÇ\Z~c÷âO½dE幕.öû–‚¢vŸ&o_8$概Dl3åà’CŸ(‹ÂYJJtý©^(—T .q¬yˆ*ÉWIJðds¸ÈBËÕ¹…h® óž¶dφãakýΕ±ž gUÿI{/D—ÛØôŒE½ÐñKùÓ‰`þ8‰Ê…- ¼l±—‡0’ÉÇyÁïž±Jæ|n¾‘ª+=錢½@áíh>Y—;%m¢£3|KIëÂлº1Çw!—ªÜIkX3Àr.á÷=…%~8ì³Éˆ™¡V£s ÎxÕŠ!Šø`̪®³x_ø³Ï«ä ›â¶ïó,­ºœÅŠ{9šÑ锲ˆÜIÚOFC|Rg9šØ]Í*ôN÷VÓ=ùæ"%|HßÑÉ-‚PÁëd銥úibvG¢™q¦›’ÎíÑ£ò˜ˆ6ÞÈ&’Z Ä̧iNÒ¨›‘Ä¥—iÆO¨NNœwc&í(T]?œÊn”æä¬)gŽîÎ_¾\(D‰8s«œø/¯ÕÛ=›y¦h a‘Lo@0HÜ玠ÿtC'öê®»ÝþÅ93@Zå¸À’!Ríá2Ý'wì·9|¤„÷qlÞ³á>¿|1>¿=sÙn÷ŽãWÂæÉ±-…$<3yãª+2ò•pÊÚšo.›éf(˜ä{µö`a²q•6¬Ë„½X&y;óÊÛË$ŽHIì8G\YŽb±Ä±˜cºNø4'¹a¢“=j꼊ÊVL0$80vjpX±h&‹ÿ!S.ŒôÆu<î‰^\ošù8XM¢_ŽuèʉúV½sTÀj{·ª_J¾±Ykþ~c|(¢\•k{º­0+@Œ½ìˆ.†§g»’‹§þ4ˆõ鵯À7ÞyÓ(•b°Ã1Üg3ò<¿ù*…ñ¼ÍKü¼ó•LY£uÖvÏà†Ï<‹”;±?Em³fš¿§ ƒêð»u@ºý>(RúÈ8Ò¶v)Ï1@CE3<ºTæË¶Ý5&ž8Zfl¸ÍñéE›ôÏ®·¼Ã]Ijwæ·Ëœ¸kû6{ÛZˆ_‹` 38¿$Ç,ô0&Ì #öª³¾bè/Ñ%•œxyº=iž@R/`ׯçéj–½m×ùpzÒRÈ›ý•à~!ÿ»Eˆ•æÀXGYQÞ "*{½7 ׂŠ@JsükFîÓü‰†É³*´Iåp)™“ÂRƒ% .O„UDTÿ¹»¾E³ÕË}õyÙmåàÆ¬÷ÆÅ>‚ã}ž÷³v"óœä@l¥«ëÏ‘X ãd;ð äÌëÏUwé6ÛižFûP¥'¿÷a¥<äRZÄÒŸ¡Ö Izû|'MopïðÓ3fÚ=ô|Ε&47zÁP¥iÇGÿ`‹]C<˜¸ü3ÉAQ¡m3¨X‰.åeBÐV=ÏF›ùì—•xÊZ¸D¸Ž3£uÀ²W‘ý¾Üw ö@î껣іz€ržžÛ}™ чôÂ[¦uû¬#ƒÚâTo° Þ$û¶ TÛ÷l‰g°6¹l=QZ©%Œm|½J?ߪ´‘Ñöª ïn×¼.¶Á2 óCë•0GÆñ5=†B«’ËË»±|ÝßÇÜf¹ÐñJV&ÃØ'ûÃ9€Ìó?šjúÌ`=””q'[’ƒA=eòN?ÝüMsÃúšÍ=†>=Z¥¿_ñÆEe„>õ$õpþe{î ©¼ŽMÓX•åá~ôƒöÔìh¼ $3ßþ#©„A¯‰Y­‡ÅH·:³}3*?·âv¡›øcüŽ ú¹ç_ˆû´ò઺‡FÖ8¢aç-†‚äiçq·ì¤U× >Tð3Hø‚4¡H",OMìÎ#§˜/¥È¤¬@[•jyŽ8ê7\Ø&cùž Û±fðÜâß êÛ+{Š<–LÐñ5rYJôÌçÙQ‘† ÍÅÇŒ÷î5ëšÇ=8}ýóꃄ'±å¤þ,Ä^îÒQeÿÌèŒ)„aÃóåÙ@Ë* [Ì_È2"‡izìï—OL%¦ÅûXågN ¢÷šš7ñnK!5%„Œ`]µÆ3æUÖûq/’÷ÀUF;œ‘3·µŸ·Í[–2”ôŠ¢µüýÝO„pgîÄ Ø¤ß)eŽP½£ö6<ÍÓ³N¯¦òkÃr¶Ï¡—ÁŒÓ8Bq«”bÇY*‚¯Žñ³¥ô×î-ù¡Î\i¾±e]fgì±Dš>©4.pš› ?s M0æ&ïä{¥4—¾z¾Ý0K^ìEóî€ë“7JÌïyo"Î&n5ΚvfPôå¤,ꎯjïïØ©Â*¯mʤQ²ZEzÙ¾P/›Å\åaeÝÁÿ&f¹ç»$æ¡Tî+ˆ"ÙÚí+®ÃA²+ã¥Î»Y¦öë)î.Tóg@HŸ†¶È?ðçöøŒ’Í1Û “òd³ïÆ®0ê|ÿuƒ»\?O£ép/[ž{ r쥛m%âDÔR^»8@aAu?TNÿ"êF—#‚„91ð>£<û‚$ßöÃîá1÷JÑ äZüÖ@èIœæz/I²ªïZÛ…mi(qc ÍžÓã@µ4𑲾‡¾a ô«9B4æ«[® èü³wwwÛ»{u4¸GçÖâË:ò\‹,ÅãÑ“ï¢gåÄá:®VÒ{žÜh>yÌdš~¿[}J×þǬEZbÊè-ÁϤ@®ÙÙžBü0ãu˜Ä› Ú¦½ü'_¸ùh¯ï[†šÞ@£egÜ… ªD<èy G'œOÛêö3W×J.üïð¬×9p¾¬f$¯`ë4bÉr¡VüÚN·”køkg½ª­Ú §Þ»è0%=á¸äx.GÍð1ü]¾Â®“¢(„—ÜV´š!‹tw–i‹“O/X“©!g§¬PFÅéÉj6Þ°—g}=ˆÚíºH‚¼'…bFgîkpŸ¯©ÜËj÷äÅ}Ê{†ÿ#ßž5—¢„·a–§í2¯¬ù–ž^;Qsc$ÉýA®–º‘s••MQ}¬]kT£¯©¡w`*-ª³}¢þ°ð!¿`ÁÓªb³‡Ô¿š]²wèŒGGó×tN[uC%¨€5<™g„0'FÆ\Ý&w©W”)MjJ7={¼j¥ÉìJo^ëÐ lõ2¶³šÊ¥' zÌ¡^ù6KSuáØœAÁó˜ÀêQµá2§º\GxïРr þÙÂY±†Î¼¡kãý&›V\­ûÝØPÂ#SÚ€åiç›[7eg²2®Ân8ÜÅü*«æ´¹;¯W—8‚oâ“…¼¹WJ±¦/âî\>AzkKˆ„ï6û¬îöœ†è,ZæëÍ,‹*¼KN¬Ò;¶ý&mA/CMáÚEžg¹&ƒ©Ìšv3®Â’*YÈpUKF¡ˆÍ´4 tëŠö3Ι™íƳˆ4¹)Ëø·®ÓÄš^Ƈz…ê¡åýåê¢ùÖÃUpºØ‘t¡•¼ŠáÜ-å™W'7îžÏ 3é†aïE¦sI ÑdØgÞ2:.wߟL%²È¼ðæeßi”¹ü`’™ÕëÓ5ztõUåA£BvD‡pš:»¸™P=P~ û”ÿƒŽÑL$ž}d¦fJn/ì jgy?©m»ú®šíÁ’PðÃÂPôÞhˆÅQ‰ùÚhU#-EÙ‹Ú§/¯˜%âÍPá¹Ïæªý·6GúÓ L‹TÄ®4¼\ÖÌðûÉÅÞp$Çþàßêï=¥š'})ÌJvuO/¤2Ç#=MnP8‹Š^§²Î¹Ë¢ëƒ[5°]Ú[µEF¥˜oZ›˜åŸ±sê9é,y½~ì­ŸG|ÂzÝ/)%t=WѱåÈ࣠áQëáW +å‚KZ¶ÊO"\Noj©oI+ô¢3æSðMÌ Æ!ƒ¤n"*6ek $äÃÜ-.ÝËJªº°G-ë¸#ž0—Óæ“ÑôÊšU<§¿(©‹tr[D½#%‘'ιÃÈð%k Ím™éD#ů&Ñp¤¿O˜ã” 8`{œàÑV Ckð yaÇ»$š2'ú‡VØÑøVYg?X}‘žT§;9l9jý†ê ÈÛýn†dRkÇOÙó^¹¤+Ç$Œ,$Ê(Û%T¥‹~ëáS¾%ÚáèÀáx±¢5Hºû³ÈÞúÃá-k6ǡð`A)#áþ²M½Fhh¡Âü«¡áÝéúæ,äŸ]7plãËzn^ïtìÖØïŠÈN]Œe òã’F3õV×ÛLä7ÄÒkÿ¨2P,åÔKÕ _Þ!A—7ÏûàGDÏÞchv\Ÿê;Ö3W§Šýçw×Ú±’wÎ ®z Ÿ¼èo9EÇ´Ôül/v«Ž.ï b¯hútz2GT44lC3£ïZ„dEIÓÈ;“p§Yù‘~‡3„@3í–@­ßxÐ)ªå·_¦Y(«®C±¾37¶ïZq«ÍZ|òv–‡Àò”«‰ …5e2)θömîõç#r•|õ7 Áðjé8Ž ±ÃHß>ûÉ85`<›Î²hÜä(»ÝæÔ7—û?¤€ Îjƒ<2žìr?Fœ!ï„òÜ5Pû³*£{aëñ7 ¾}Ùs !¬ðLËl]©Ô§SÉÒ¹VÁö­/ð &¨ù0bò–ã™Ñw–)ÝÓ’“ÔVZ¿lß›»Ë™£Ô[—óð¸H°Gn¨¥ËEµÈ8à’ û|ç¼`}_gj³™Ã·= ‡bdzk<2cpK\©Áô?BûøÍJ{;¥Òz[ûÛS?7ubO¯~âÈ¿õÊ Úv^ç÷‘[Ü÷x]Üò¹ yª‹ý«ç<E_[çróœÑ^›Ý{ õo;,vdJ¢>µ7‡Ö ò|¦ë˜ZDnø\nš¶._Ž–¨ïœåÝ°É ðŸH›åÀ…‚jÝgÅ9+}vàˆR L‘Ãó·ï9ŠÝ¿L­ÛòWü4ð³Ílõ’ð÷+…>M™!H`Ö‘¬¿²£DœË/¹ýO±*æía&¡ S1îb—÷Æî=¬Þ1ìOìÙ!çÜÞ#A4lò‹¢ßM‰t}OìfŸeÜæÍ |aÜVo0w)Bn\ÙÛDæ÷z{5Öh€ðÔWÙe2u­óÚ亣0ý]5<Ïñè,@¦­![ËöžÚ;¼W£»qž^•îzÄ/ñ{¡\/‹ffÃûFâÐ&çŸú9™Ãöy·…=‡.–^V³ßiö&ÊVƒ¦‰]ð÷/Ô£*¬¹©Ú*¾fÜrŒ5zÎpzTo§,r8¤Ð3øÐc‘/‡®=à¢Ò ¸§O]¹uµ~N«¶K °5™¦Ó£CRñR †iC†Ì?–JO3˜-ˆLµÊƒbß½Ng'ë1²hapýu*õ.ŸAÀ÷v!Hë‰$£ÆŠëjåÃüüUòç]¹= Ê «Û9Mg¯v”Ä#Z+êñ™Âù WoâsVIlÁ(¡&}¾­ç¦í€)©”Ÿc- r¦,Ó Ù·=V,µ.7{Uû8-éû¡û7x\T— ò¨{í+{·}ˆÆü¿0Q­f;ðï,X( qdBkž1µpb½®Q*Ðêsp§h¸8EøûúåÑD tŠõô'–Ͼ‚SáÔûüȾ$¾0¡Ä£·þ¶“¶˜)Vu›õCy¶ZÖnkÞé ÄP°q¢ïDÅ þªxÛÕMÜa+»ðUÇóÓ{¯öŠˆEʗ̼S\xã!‡ÉýïÀÖíÐ×6„Ϲ´ˆç$©†ï–÷ißÒj÷î<¨&T”…Š}|Éöœý‚,Mp¤˜M´¹åFŒïÀ‡¥Þ¨6ó ê×*cÖ8M âó/€x%KˆO¯ñ í> stream xÚuT“ݶ-ÒÄÒAŠzBèUz/ÒA 1 JIè½W‘¢‘* "M:H/RD)"½Wo,çœûŸ÷ÆxodŒäÛs͵öZ{Ïù…›ÝÐDHŽ~€Ð@£pB" °,PUßÄR‹À`Q7·)çŒø ¸Í,’ý_U ŠÃcjPž§FuÜ"b@IY)Y0( Ëü‹ˆÆÈÕ H8PÔA£X·*ÚÕƒ´wÀá·ù×#Æ‘‘‘üTvA`0( ¨Å9 \ð; Î@4 ‰Àyÿ£Ÿ¼ç*+,ìéé ‚º`AhŒ½¿ Љs#°Œü50ðÔñg2€hê€ÄþÁMÐv8O(ÄÎH…Åg¸£à ¿9ÐD[hàŠ@ý!ëý!ÿž P$òïr³B¢~'Ca0´‹+åDÙíΠ†ç…BQð_D¨3χz@‘ÎÐxÂïΡ@ e# ?àßñ°0 Ò‡a‘οFþUÊê(¸*ÚÅÂa¿úSCb0ü±{ ÿ¹Y'Úåûwa‡DÁí~ ww6C!ÝÜÚj)xðÌJ€e$%Åe€7  æ ü«¼©·+âwP䌟Àß×í ´ÃðGÚ!ð?_,ÔÄaÜþ¾ÿ;ðÏ@DGÂpÀ{$ ðŸêxa÷g¿| Ò ãµ'ÿúüûÉ//8åìýúïû6ÕWUSÑ¿õgâÇTTÐ^@_!10PHFB(""-”’’úÿ³Œ!ù· ðrµQvh ÌŸnñÇô¯Ž=þ €ï¯9øÿ¬uW-È÷‘[ƒ%À0ü—Èÿ·Ô§üßþ«ÊÿKäÿ݆»³óï0ßïøÿ†º ½ÿð¢uÇá  ÆÛõßT ÄÓê#àHw—ÿŽjã x#(£ìñbÅÿàH¬Ò 7Dâ`$ó7ûe5g$ aˆÆ"½[ðY`ðÅðþ‚9áßX¼.‡xûüs_u ÿå3Q I ƒzÀx9‰âïÛWoH8Âë·’Â ‡OâgôÚ¡1€_×  » QîØ_(à•aî Þa¿€ßö_ëßvF ¼0Àô&æø*¬ñ¨R™ÅSèÛéç/Íщ–Q8žÉg¾z—35GÝTîÃË®÷¥N½ˆ™ðcØ8ô²nÍžKÕzŠ#Pï\PÙÊí=ÞmµGÜ„£™í‡¤Ò¢¨7Ý ¼!†ð>¶˜8 —j[._*8oÆ.[ɬ]€…Tì©,¸‡Ê9Ík„±–ì\_pï-­©âi0twSm³¡sÂXdü¡JcÓ±÷$>ãýìlOvWgšØNžn{…dY-{N_R7ý£l:,ŠÄê‹”D¦ç.ö×{5´:7)Û㔄8Kd_ö–}ŠÏŠfõQn?ôðªÚ'¥¬¤ˆU4¾‹0¨È£æaiøL+nq¬êg"½æ-·oº³çUû’:Õµ¬uóì\£´ HŒM¹©`jj¡ÎUóÉjõd5™¬›±ÙÔ„ ¨ÍyéÌ¡½¹àÐm‡¤‚N1¹‡jÏý‹6øJÓÃc—L‰hŸ¢('ÃA*æqÙ/ÉŒÔRC’û ²MÉ&z£ñ(¦Á³iâ`ùS·/¬­T1¼ñ‡´_\_9Kyƒí¹ª¿Ëi™õ€Ž{©¸SÁ⣷²D ¼*ž–„Òöbä6¶÷ÇÉÓ¨ƒÞ3“¹j)ry‡Vïuˆ±ÙŸÁ19a? oÊJfêÿÌl¹,þAµÃœ-Põ³U`<䀹î¨-A‹vV zšÚCOâ ¤<yÜá2ž¹è—ÇX%µÚãõõ†#Ɔmõ%QšXX §iÑkÐc¦B¿]Sªáìnèrƒ¹æB×%š<ÔUòžËÁÛ(™lGý ,1DþÚõë66¸Ul>Q®‘û±ïÿ’Çd¹4¥ÓOØ‹—]_.‡é—J˜ ¸ (D#Û/AìƒÍ”ú% ëúyÀ#Ýc£"cIÌU«O1œgÝ}k>]'¦í{ìò‘Þ 6*S¹¯½NûÀÅwo6¹qY5±åÀöÈfŠŠk ¾¼ð¸¢”)"@Øšˆùv9Æò…ȶIrbÈ¢o°%…fÂÚÒeÛÒXñ¹wC&»&PΛDýCwfÿÁÄÏ¢ýÚW"-BÙAMæš ÍYaPK A™SæÕ½Xαڸe «àŸarÈ¡œøb°«íJíM×Ó75 %ïTôÑ =ïÃȹ¬@ê»IÇjü –4 ˆRº k‚V?žWv¨³ˆGµØ&P6­Ô¤ÇÎßþéÇ=k5Æ?FÖ]à—®³ÒÄÏÙ3’ Âöm3×ü‘)uÅ‘ä(;ZÀ›ÎÛŸž>!>#ZXHøúéZ¦§Žt+†,º%{àš†·&_Ř\;Ì4Ýå m˜¤)=n/¦ô3ŸÛÔèÛ|ãö½aOÛ\œnù”éCD Tì+Q­Eö›¾îÏ;jÏ^L ò¼r`ïš}75̼îöÓXE`©Õ=¥P^xN eÞ7>Þ™ª’âuH2ÚK¬¯&*9ÂzºXÀì]‰Îx#›>ÈGÍ¡¶”£uáÜÙÍlã%hȰP0Z{cyn ;õû[•*kî.Ä2Ö;~:^ü±8‘f|ÌVQ¾õÆe‹'áÕ«.›µÈB”EhÈõnXþâs žÆk’ç?'­¨M–œ‚ §Fª,·ŸÒpuMõΔçì¨1(xNŸwYEBßæÐ}“îXh5ÄÏÈòÅq¥~:|°xà†ðL g]‚flÆ©€¦in.pó·s‰¼ËYÿý´<üyyšlÛ]÷óÑÅ—•KïÞøÔjp°Î4p0<ç1úîŠ49l̾ܮ¶ aIà[ Vñ0¼É#ê\À³ãÊ ªÍpC÷Ž7Þã€È=ùº¼¥5 9{f_vVhzRpÓqd=Ò»?Éyžû<)Qþkl]ö|þÌí #ôà`ä ?x±úIÛÞ¸¤Rûû}þJÏ Á$ A%¦¡—ÉSÊŠ¥íÏ eûñ˵¯„Ôf¥òŽG«¦uy¯±Ÿ¸¿Ôr¬Ÿi½\—/^ {È,ÜzÑh§¹ï0¹Ð•X•c”OA“@Lñɧ@á¬ïõ`˜VKRfÁBú¸È¢º›x¦,?« ÍPª4;i¬Òw¢íÿpë™ýîçÈÇê%{#U‚ZW”ÕL¤õçbŸJ[4áPĈ¼YóWÆ‘²ù›ƶyfø?¼ó>än…®æqmò{ß´(ŽUȰú~V²­¡àú™6"?އ†ÅÈ|¾k®`ã¹Õ|Ù–£l§=OÍ'/fʹdâœ+ãêàEt} .…jz[*ßv‹zq—,GJ×4Jñ ¿íôK?ƒC¾EqQ¶IÔ]'×mõW· óµÞ·¤—³%u-l£$=J¶”±_—˜'†S…_¶×÷¥×°{õ†:)m¨É|Ð%a½”ÌÑ CLG'PQäªøôréE½pó×4‹I³_@õœ¾Ô®ìV7]—ßìó8½­ò5µ‡)òçF,Ê“½Ëør˜í®6Û»ËoH`Ó·Rø ëMyêKÔ7ÆÏãG÷/±|žìì,Ùc0Uqû Fc½¥$˜Ô¨Göu ¤Oʹ×õ­tÍ­4~ìl̬;B¬\ew÷x‰Mõ-¡Å²¥Æ–O[Þíj}qKs)ˆ ßžÚA-m»¶ZYƒ é±(LJ[àµtÙW›Ã2 &ÈŒ—å)=GŸúqxtŸ›®.~¦Ñº\4\fÔ¸œö´~ ¾B‹ålâ0«ú!šÚ§êNÅ7éE5Ï<¸Øˆ:÷!ES‹ŸµF-‡™­} F€u¡I|u)§!OÆÇ¡™RtUw;|ý,öóT¬*÷¡2éôZ…­wš¬ÚÛì–ç{¹Ê„¥Â«­˜ÓÈIC†8ý!-ë=FNäɹ‡¥Ê»r?î8æxt:‘ˆ›.Ìô^ûÉ#kdÝã7C±õÑ,T¬¢?ÏyÉ"b^'*oA]»»J8g_O>åo¡¦á^Þ]ú¦ví¹¢ÿÉ•š…ÎT÷Ü{!’oì¯,Ô×뛹vö1 U>…»äK»ä¿º•‘=…nøŽU•îymZjåV·×$…Ì¥æ&ŠqJ4è…¡éœÌ8-â‡`´ÉôM!J¸1OI?U=©héâtÝìêÿQž¦SB¤YVg'«>½5õSÜ2ý㯸µ­â¥Y»Ô¸UŠCG2ÎGÉ Îž¾)/ ‡h-е>Ù¹¡DO<ûJ]‚ùmçÑäT4C˜u3ü-F£Ûð)¹k[LøtðQ1„g¡IL”NÄ£Þ•ZÙ`ud ~ü£¾j]  "9î%|»–ã,7)i*¨ßOFZÜümûŽÞˇÑ‰n”láîǧuû÷D/Îúî#fä€ó§`¥6û#>°@ÇpsájKÞ󓘰ók¬…¸m¨Ù{ÙÉk Btʰ˜ˆöÊÔµ/9ÊÌ”ºÄIwHLƒ«0º'!¥ôÕÐRþyÿ[ßÌëÄ€üζ–œ'u÷¡#^Ë” „eøZv…o+·¹úáBDâŠë¾ š%Õ§Ð|m‹ÊŸ„ø%½í®[3µ«ùës3V:M÷•øîõá¾K˜wdßMÐåDä™:{ü—RIu¤bd´…ºjœÏ:¸òL_Ä]@«‡qNœÛ{TÜ».*úpºÁ锓™RIošùªœÐϲñî”c!$ƒ¼mÛò*­îÊæÄü7î!“¥b«¯QkÌJ(§xþsÅŽ3q‚æ”ÉYûm²£D…hêÛ•™?\æóØòûû’6ÊÂ{…$_J1ÁçªNyêt¡ÙNn‰•‡Áˆ–‰~ãõ[:² Š›Ã„X¸zÉ|THOM’µp¶ªý/s‚¢î¨ôÑlRì&é7\ëÞãÆ\’o>Zmƪ4ÉX"Ç'wE®ê ^ÃòLŒ½«;o^”‹ôgÃÊôÎ’-‚÷ŠÔ~Í;WƒËF‡k¨÷˃ëÝ6+UÊúÓä4¢Ï™T*›ßÞ³ø“'n˵6wýGÕzV™J£Š1ûŒ‚1 ²<­éÖ—{è‘ÊxÙf< ºã±–@¿ µÅ¦›T5ºs"O(íÓÍñƃŽׂèKÙ“…9HVÇ“Ž.‰D¹²ÇÅïE?Ùê5ç¾´Cýp<Ù|Ï@Œtq¬% öˆìkÑÞ{$¥')ÖQ`x†¯gÔ¿,.¨‘S«ñÔ#>d×Ç?¾‚„f,¿RëIŠùÀÞEÐ2Á1 ®óÆûAû-ç©l㑞O@ZØT:(m5Òèç)BÁÂnýø~ú‚´“©jaRÐۀ砎Æê܃‡EP%,’k%%zA;—»çìÉD±ã¹r{\¦ÙVJW>»’ ¿1G0e¹oË,½¯ SW¸^»hCͦι¾…( Å%Eå;¿²…Ú4*ôý×§H4Û±$¹Lj£³‚DlÖ `™IµgZ¦TØk®w¬$·Žwy&€žýd÷@zîVï*.j 措u3”ÐgÿhŒ “Ÿ„NfÕ¶‘äô#Gïæ/EV´ ÛÏ÷ùMa×ëøÑÉþ‹¹·>yÚPÕ1ÒN_u¯(}V\#œW &´*í™]WVÚÔ¨éÇâØ –•ýÉz¾7'ËÈÍQü„Ï•TœàMÓvn¾3¹©ä) k¾6×5añ¢¾;hVU{e ×ðÓ³cšþ³ÿ[(Ýzæ{ô‡MÝô÷ïOêÇõL×È%/Û¨¾íÝã[F„wg<àÄ¥½ßZ}¡Èë’Ì‹À9“¨‹yv¶è¥ÇJ‰…ÂõZtxÝw'¿Ò{l×K’ïèÙ\ª!Jv’ Päðoäf©”\ª8Q†Š0åO…ø{ß>µ¸]:ïÎôóÛ éͼIêã~NûåìFž÷'ØbKmlTB[àC ŸÉ§0Ìô&©êBÙÌk¶¡ÞÛûÒ­—* îE¥Yz…Žy0J$s‘ï¥Ü•£„ÏVú‹;Ë¡$ÔVoÎ*ïΞ6U­B‘mWiÐjú­(ª ²ˆ*ò}A¿‚— üdK­ƒÌÛÞÖ|×Öëp×ü!Û–ä ÊW¸˜¸ íüàŸ•LàS»Ñ¨·uû‚,fÃ3™<ÈpvsõòåÆÝÚ"R‚Œ«J¡w›8оU5¯Þv‰™ZÌOí˘;¶ÍRAæÕ÷XÓ«)Ë¡Òï,OÚ?¹Zx]šŽ¸.PG~Û·íäz—XîÍyÓs‰a0ª©-åï´v×,j;.c ½ó’ŠZÌRç©ØóS²&é‚Bdµ—GlŒâÊåxC¿aG9ò9+˶¼^Ê¿~V;3ipïãY×ðšõ¥÷Ý«Á_»7¢™¸DÀRDwüÕï«(qINXŒ½ {ýui£!gõMÖXºì¦\äež¿öÃ2/¿ž<ûcœË«¶ÕúÎnc³Ofïjßy€ôÈ +éÇs%¬T_·#Üß¡hêœB¯¥àâÔ…_Q!t‹/¬®}Œ£mž½†½¸–ãv4 £@R\fšKÏSq‘c˜DIÕQÝ=ª÷cŽNEwÌ.d8ìÁ ÝÕ¨èœ/ÝØoâ3Ý ª äýdø˜<ø8LcDKÁÿ´ßZ™™ÒýåV"™¼œuø–PÖ™?¼,Êg´’¬L“†ßËëcΣ83H» ½wlðpX9 /¤étÑ¥³¯~¦’¨Êïó]CõHßEh#/±ÐwIݳ–ª7Žò–wÈwnKšº|ëw}0–¦¯’ë^é·£SæXÿpleçRKNù2 Å ¤â¸”Ü}÷%k€íhû2eÂgYŠ;འG+ª-á=WÁúé“Ö®X>"…U*:Èü°a¿I¹FãA¡&D9É ¥P|÷§—E,PQ#âÈþk‰"Uûäå]þ˜úÔ8Ï¡C6:Þ²q g½ž-Yu1A÷Ìí 1OÙÒqÞ–#宇ö¦Í[Ÿ$†+XéË=4o ~j5»šêÓ©mØn¹ňYéDz¿æÎ'o"DPç¾mDóE|Y ±Êê£ëê¦ #‰ 7g¹$_H…ÄòÁ2ÌË|+س—õc$”Âühìž{ži\Þ»|]YMYÉÜ^­¾j:pm.ÿ »÷á…÷S«;îèܶs5Òdö÷j`—5·Ó©Zög¼pä¹ 5YÜJïÉ=â_Ì™ùUC¬T¢fÔLª¼Ëû~;øB¼Cö endstream endobj 112 0 obj << /Length1 727 /Length2 14186 /Length3 0 /Length 14751 /Filter /FlateDecode >> stream xÚmºcfݲ-\¶ízÊv»l[]¶mÛ¶mÛ¶mW—mû¾{ï³ÏûÅëO2rä˜1WÆZdD¢¶6NÊîvÆLtLôŒ\%QAu&FNV#= ™ƒ±¾“¹­°¾“1@ÍØ dl`b0322Á„líÜÌMÍœ”†Tÿ6T­ô̭ͪ¶V¶.æ†fn~Ggzg^ú’”ŒNfÆs+c€œü Y1¥˜¬ @ÌØÆØAß ïl`en674¶q4¦˜Ø:¬þ£ mmŒÌÿ…É‘þßl\Œœþfâ`k Q•“Uˆ1( ômŒÒâÿ¤Û89rýmlhàö¯îhNÿ+Yéÿ¯dð¿ÒÿzÜþ#ÑÃ01ŒÌ Ʀæ60 ÿâMÂÆÄÀö³‘³Ý]ÿrü€ò©FÆ&ÿD;[YÉê[(…l­íœŒ2¶FÆ6A[+#€ˆ›“±‘±Ñ£õ­Í­Üÿãÿ'BÂIÿ>lLÿañ?&sGQs7c#ys§(7Ñ·ú‡ªÛÕŒÿs>ÿ*ôŸìJþ9W#[+÷ÿ)øO3ÿ.Ç ­(©¦þ‡æÇàßnC[#sS€’Ó?Œê;ý¯áßny}óÿÌСüw¦LÿW—Ñwr0wh2ÒÿkHÿõüWÒþ¿Q‚‚¶nžtÌÌì:&6;3 €“…Ñûÿkèìà`lãôoÚÿéå¿ú¿ÇÉØØÍØf}ÅÖðwEjKH¹Há\8ÍÜŒž Í ðêtÏ&²ù™‘îáν2Nž+þ¬¿ÏßË]‰ûfbýXc¸ãÉ©(¢0íœ×b’ SŸšíw…ü·ú0k0UlkCY°›ð;§ôõyê©_™9ç5’ö¥<ÈbLQzeýàH;qk‹6bTÚ(ßpÆû4ÕHsfÁ×|¯{ë ¡yánȰuÍ ÊýíC5òöõ;0Î7™ƒh¬Fm#Xþt´gM—ðVŠÿ„íJÿ]ßÊ…ÄU %g%J™_ ¿DG9¿+¹jÆ.´Ã—«ëjÀ.ýÿ\¯ZW Ì/$§Ìü,k7Ó<õ1ë4ñºeQî˜åfxÍÞ¦›ó[SùüBrà«bòÅ@ Öíú ˜–±_]®‰·=´îvBy!ËÅ;äÑD!·þ3: zCí¯WŒ~žØ]v¿X‡×F¨Óf‹Ÿªùxµ³Öº0†k¢@:‰nЊŒö.,œe%³•ß°+i“ôÏ¿¥¯Îñh]ÝÆæ7Ér ߆yÆ–«©'ÆöúÜKŸçÉö¢À÷M‡È+cGòh Zµ\L)D¼/“%ëÓŠÏⶸm©ÆÝ`„èf7œ_–ŸT`Y[ûuŽ£ö8²5õ&Ö¬§Ò)ÊùàÚÀq¹(tD36-;‡ ´R¡)›¬ù×l1 V¼Å•±ý¯ƒ ÃÙ#5ÈçPþyÉ$ÅI7h1ôr´XžÐRϰ-µE³¯0‚_gó4 81Ì72D?I8ÔR=CpdÙä9ŒQŠC½‘e"¤aBê|æËá¯)H,ë€á!Åú¥xz ÍYKñ,­±ÌÔ·Ò ½Ýë› (Uº YüÒÑ,ƒAÃõ¼_. ÊOð8ßâF&àÄ’r˜b|ù­6ÆpG­–‘”Ο°§ßòY6æÙYa~Ûî„?dJñCô‡Ë‹Y ¬4!Ï «OêÁÓ¬zÇÆÞfvéê.Ým xEZ:ýÚÐvö@½)€}šX;r̤ÈP|Kdî߃ÑL”„R«>ÒÑŒJrKyF°(ß"ÒWp5#JàP%S [†éßlY,l”%ž&¹œ…¢k©Þi @¨» ž´Ar$=ª´°½òœŸMrXÄÁá}NKkšz,ž>s€mýI $2Âu,àñí}ü‰=â.z»ý„=ÎërœRé|«_!OIeJN¯n?¹‰WÒ`%s¶+ŠÎ‰¸pÅÞª9‘t\S ݲ:»Ô&Ÿ«}6ÙÂ×pjÝÛî×lj\5Gí³‰„woºL¯ÝÄçÓÿi͘ŸGÉ¿á½?i€_DlûVuy@Û¾(»ÈÚTY!lÓþb•Ô9¾Q^Ò>îÔ–~w/]MêÂÔ,Eƒ3+>@¤…!…š›¸ói ­uOÇCX(µ¹3Alߊ“çqãÕ²–FÞV".À-aÍYNVo¬Û_ÕìdWŒ`Ñ`ËýÁým®å•ëKÀ"’ƒús”"ø¦füÔ¡Èí1ö×SÇŸ æðô$6©Õ@ÝÁ ÿ”©Eå¿¢l5Øiˆ11ë&ç[æ(J€TkÜ“¼xŽÔfbÉõs’¬”71ÿíb¥«(A¾2EÈÞé™Q¥Vê¨Ä0Ë’#“¾ÛÌ}LËöý™b†~w¸¤€J¿x‚¨dEË¢'ýcZ}Ó&Ó¤=ýµ,°¥¢Ï=ÛQ-ª¢Ë^å-!0ÎnÔ\Ùt)k+°²]e£·YZô¨¯`@¶Üö{;Évk !ˆëÏ™û6ó“`ƒ]+™…U×§ôcáϾÅú]_o™d>ËPuûL 6G^žVvŸoQÓ‚F~«4Ðz™, íç/£A %Üž DÏš*’_zÙ¡v£¤>¹µÛ;½íÄe4¯ß"$û›9Ž‹ÆOPcY„bÈ žá©5¼l{+|V= Q B˜ô…·Ý÷’:FrÚu¤– RŠ‹KO"K×Xž{ì<õ¤CŠá§è„ÍòÙŒDºxQ.2îÏ%y†|™¥âÑ ÆX--nDz°€t¸UxŒ³•‡VŠ¡ýæÎä1ÿ: €š; ç¡Bõ)hfÆ%ß^!žÇ‡úé¾z„`>û_]C±KÒ^Yä¢1‡/jërm_±µË!ùî0 –ûò´[(dØ!’åÇ6ÕìšC‹õq' –d˽Í™˜{s(¡ÉŽE>§b^>»¨ª a…ø;“èJHþšï‰Ì‹}}fæÃKÄ€»?準Ÿž8-8ò– Ø}8Cw¢:ØO=ûàhX4ٮƂØêP·NÆ"÷<©8]‡ BùÜË .l–š”UKÏEVí¦ÊÓžï“©@N¯ ÇÔD=cÅåÀ\å~Æ$¯ãHÈK¹&¿o4 6éT³N¤×|Î$)½’|L7Ká¯0 dñh>N,&º¥h~ ®Ø££mµ´‚÷!:ðŒ|ë 4«’ÿò¦J‘³PfGщ71ž½ò³ 9Ù(D}K½B‡uˇó“‡`›>¿¬Y…ÍîÏV¡©ïViKÎ ¡þ­qH8÷|,*«èäñ,¬¿z´c¿Í{2®ÙMæ¶w¢\F¢J[Œ3Y2&½AW% ®sÿ®Æª:Ãêš…N¥‰Ýïî­LbŽz8å+f?èMÚä>DÞzîr¸—’óÚ=›[Ö7÷ÎmJ™ÇMÈ’2— ß›ÔÔi¥ˆê)ÕÏu§n–¿cŠq—]ËãÏ‹T$ æ§h  ƒ!­œxOL¡aÃkÇ5Ÿ–DôÖ©|¶$(/À°«IÇŠff˜‹¶Àk2s]Ñ£îÛ¦ ¡/F VÉ5o4^¶f*WÌNt%ìi6Xáý Ö×ñ$_ ‰Zƒ»Ib"Å¢*ü;¬»ÄÌ”@•½ì]Õr½nµä{®÷ƒd5‰'7ÝÀÐÑÚ½p Ï­’ºÜÃÇ@žá¨Þ˜ïFLá£G†»‚5-„Ÿ[ür828µDÖRÈ`5-þ›ÂB;¦ç¸ÿ“R§evýýžá4¥ú ƒ÷t„¿†€ƒœ!VYÆžg'|cq½„}›OIáx,s7W!Ü·”Þ1§…ú¶.±^z#pSÅVµ$Ñi Ò¦FŸu-Î O¿ýYü½€†"¨ž –ÉÄ]˜€ih®_à…7&:îÒðá-Âyjð•¡š¦3¢ùˆ¡|‘lhçøI~Ã"1âCv?>å˜m0¦â#;e“ÈKt{¯<G”dò™¢Ê<<"˜_ìÃ~Ò×½Ôèáb'Þe=NIÀx,KŽš?RÜ{–Óf«¯LK%s¢oŒ)GeÔÀ lhõÖ7ðz˜‘©cr;Ököžrïw<¬IuÚ*dÞH̸PUí||rsþì'Z‚µölmÜó䘳æ.TËöªð¶a>sô²fÇ1Xä¦êÁtþTV³¢Uì㌀¥Ï¦Z8Yܯ¸nı|Ò°vE¦˜«ÁÈÞ&Š…| X°>f s.íejZ z!•}cT¤öˆÓ‰>sw"è²b‚e+ômUNÁNÙv…q¼Äð3Øš\7)íûÍýƒ»è\BŠ¿h2˜ýÂG¹¢¸z]–¤‰À_*‘•m6GO¦ž€H¤QGy^ðGEâäeÉŸèéª@ñ¶Î=¦öÐL/ŒÏÙcSÍ 16Ø¢^_æÄõ˜óB£M] Û̆"ë[’'Ô“’ÉÓë!«æ㌛¤û{\t‰¨‰è‹2Y!c / “ìÊ*}¼L¼$rÆkð%¸ç‘m´nØ5PÄ®B’¶HÒ2þ‚ª [ã•s™‡E‚^Z”bœ`˜=«Cà ’JBا÷»Šõ„Èý-TzÃÃÉÌ隊ÀŸbgÑŒ_|&Ì´’r-°Èc¿ª ®}&{ ÞLà‡ÜnÛuvr†4¼+Ö\ù¶§Nöä)f¡ìÔy™ eFq$DÕ¢ƒvk6D[vVUT7êa"œ7VOµr-úä“þ`ø¿ß‚¥œç­ÍmŽr'ç”öÒOþa;yÝÇ>b­y`ïÓáxat Õéksµ[!pk\[ù™$¿ŠCÚáxÃÀY¾Õå]*ÆVÌ¿LòízWÿdF˜û÷J%5“Ÿ&ÆÄÇ+S¢þJM88E$_¨}¡?l‡ñm"ë6ÔYA:èֶ“·¶Uí£oÊÎ5¿„F¬OØ*A4NÓ—¸F…Byß*qžõÁýxÖ°é‹XR®Y-§rÇN øæùíî5Z,º_ü·êÎC¶5„&^0‘f0ìÜ#; W£39zU‰}YNðâIæ–…þ­© ÞÆSÑŽækÇÚ"1Á^ÕyýRòs¾ijjgÆ obWGô ÿmýW¢¥þ7UÃV;°8\™-¦pþšå6 Y¡=Qð€Ÿ®è6s‰dàÃâ3‰çv1†7üd ç#Ù3Ê)a2©:7«ô­µ,Mp±ã'¡dAJcâiZ'â…“Ï®9§ŠFñ#èìàwìË–j§&¿G|ÛóÌã>yÞTÈÑ×P‘ÀÈ ®`]©Gþ±¼v•O9)Ðp6¸Qè.a oUsVþãÆåÞΖ;‹öo¨Óy—·` Ô¢&:g¥Gú¨?Ȉ6¾È7{ÅðÓlÀó²žã¥=ðoc(<;Ã6‘û U ¥¶,Wu;•”\>ólmhñ!ÚJ,À·ö²jËç™»æTõ£ñb²êAw¹ºÙÄŒ¨¯_¼µB|$¥ýèé,¹†,)Û´3ê jïä:þrÞ<¤%9¦3×)p 7ˆû<º •QàÇAŽq€/àxk²ACë*)&(ÿ÷^îZ°vV÷þn]s¸Kúàê}Åor¾Má;ÀUŠÑCmÇÈ„özBÖø’óà4O»cE–N‡¢„}æÜûƒáÄ8¤[åµI¡' jí8µw9“@6 OüñæäÖgr-D|»’B%Åtœ`ŸêñrÊÒOwLÃN)©â¥ŸãVéï‰YÞM¤š­f І!~zÑü×€þ;ÖBfâ"ÿn«þÙìŸw¿dêd(©ùÎ9*å„ÇHf¬>©9²©Ÿ±©2¼o¤ØiÛÏB¥À]6~ ›å¨-ð!éºÅ+ÍÌü«)yÉ’B,£Í-Š7¢ò$¢ëK¤¿ ,¿/ž¨Õ{Ìu¤»¥¯—}II"Z¸(0m€M°©gÐÙ«ų¦®ååaübòŒº¯îµWtÑ`_Œ®3L A‰’( Ðz¯ºÀHú†\‘i9ªÔbj:„<‰—ûR¿SIw¨¤üX(+ùwm¥ëuuKÏçÓe›rªâeD Y]cÇ;³È®ÜĈd˜ !Ç#ÃäÔÎø:¸8mÈÉs4VÔD·¼¸ ̨í³,öçN«íi%• ™…—w6Æ 4ì©åJÈå6¦Ã.GÚ·¬}cíûA¬6;  K´æv°¡¼¡ñPprn»B»®í2=U.y{»,Ÿa«`Sþ®§›6á}ºÐqŠtÇòwçVGŠw¯WLq€.Ðj‰éͦæúãÑxç7[xÏÑÁ¹mæ%ˆhäˆ;6M,{’l߀Ý E@˜À®JܘâµíÚ$.œ€ýaK3§ûD4„åó¡Æ.´VXòpóŽÍ ÐÖRu<¦øzècBÚS>A‘ÇÂéœ}â1suüé3DÞ±ÃÇÔŸ›’¤‘*I.mÆÔH|8s×é Á_UÐ `¾dÈd•·ÿ-î3„xÞÿ½’O)G‡ Êb{ó«—ÇymóÇëX^÷"NêCQ}ˆ÷âyGõôLÔ9®æ NÒé´˜Håõ¢’ÇCrr3§¸X\Fê#i õ=Ý‹@,"î:sE”jTªÜuò=Œõ6k|ßa–Ê3˜Md «²f…I¨¦RïïÎ=tFÆÌä°àGï6)0¤¿òð vø»OŠÅÀWO6­öqÉ&RÁ±ÇYS 뛘9vÕJßX— (øk’õʺiN—挜UÀÉàÓ̬<ÚøÞw•Ögfĸf/¼þ8<ôôÁ°èþ­!ض &[¾2øõWß(åF_¨v%øê$5ÒX~ÂøÅâÅrâU^Y‰è ±Æ]^Wð€‘¿±Äή¾d¢9SÝÖä•t‡Šò""¨è N7dO¸’Þgë: JÈx·Á'bú;Öá=NKæÑM*ƒ”R¡f¦¦Œ P!puØ~¾# éc„)ayUœE«®ßÚâ¬üÊ8Vš]%ÚÑ~¿› fuŒ‡ƒö#ú^À¥öDˆ)\ÕØvÏIØRy’×éÜë6Á5PÆõ spò ­k\fÑ (ˆI˜vÕJ Mä@béTc×8ÒF Ð$ކš0m”<ÌWR‘XN}kå+|lH¢Hûø‹EÁZµþ$‡• <$×9P²eSH(áÆÆ£_º6+³ÌåÔ þ<„5ñ±9gX°Õ–½heå~åI¹6?Â%ýpîàïÏ^vSŒªÖx'¼3„Ùìコ:R§1—T‹’ºë©É‘s¤Ã©#’‰¿7i,ŒÒM¿´ü Y‡¨iWˆ#jõŒÀ7Mñ±Ûa‡ ¾î/mª[¨éÓöEº‰ k}Åýš9[U÷ÈBܼ©^ ……;ͤ€>”3³¦tҔ߼¯J„Ë?e‹0özí¢vªƒé!‹iœÿ"¹Îm´lç,„±gë9~·5øØ6‘«Hy„´Ë$à0’À"OÀóÃdÑ« ÛÂ+"¿ç¥ÙÑ|~ FM)_]be Âø»äì1lXÃÜCe xÙÛeÒ¥AøRÚñ¶œw‡ˆ:Œ*2¹cP‚CxtiXîêæÙúûᎲl¤Vl|o]¸˜’°MBdNÈ:¦³Þߢ¤¶/ª2+äìë7««0¤icøÝÛj6$0û[óÚ*AÖNàm*²¼[ƒ+Rè4Anr0Þú2Ÿº:wõ7Ìv‹>®_ùpÏp«ÐÎF^žÊ7”-dvöÄ‘7 c©DI˜'9 q É‹É9øk¿„CuöÏÎQï->&•¨ ¿ô§ £«rW¨Óc]'»ð­Ä KðD~k÷VTè³4áiì}¼.>í‰vÊíø¥…(H6u~‰é=_ùØv;§z¹kÃxÀhQ*¼ØÅ˜`»Â`Ãî–ÙkºÚ²Ðcãªß®ñþ²CDvÀ`ÉéÌU°rL.\ @cØû²¿þr1ˆv`j éðVÚ&d…mJ 2"ã‹f½å–;‘‘˜#7EÞýPèT)óÐÀު؉˜÷øHþ–Gâ†PÔZ[ÏÏ¢Þ©ø^æ&À…^#ë•å]4ûˆÂ$’ay'çž“å¢ìðCõ×7‰ëi3V¥3T¨ùdÆk€F„>'5¯íF®æ«UtÃv³N#Dµæ4´uÕœš¾Mß²ò\CƒG}ÈÐq)aǵŒÓ“¡?~™H?ækÚx–!â}PÆ(`lu’z³‘">ÿÈ1iö–Òj«¯»‹ÿÒ#ÜÝ(sCþ²ª/Jb8 «]Râ4”²ÿË\Tt^ßa²½v0q¤ïg|Ä€|ñtºÜÙ ­r«ÄI½–ö «ÉÜÒéYw#´ëöíÆ7fß ¥Z{éeŸ q®&,ÈVÅ®1—Û¬°Ô’Œ¶œÀh!‡ò¼÷·{óé¢TZ˜ÆhsNÖ„Þe43€ŸCw^á¹pׇÎú®nC×ü=ùJ´&&n„U»e·žpp·¤¥7ýÛ<ëÏn#U^ª Í"^ÝÜz›0ùí Þ„’ÓüÀjì±L¬9¶Äá #Çp'9Éh/`L©¢ÑØ’[ïen|òæk×[Lh¯ ¥Ç𶯇ùBŒ„{¿V7•ë ψÂð[½$ql­AÛùÚÖ‰÷ÄNN–x;›Þ®÷nøFa `†TE·˜º£Ôľ±wå´æ3 óÇ"ð' à¥ç¬¨~¤C8p‹Y\8­7¯Z7 E 7ÇH \‹(ó#nr!¶5/µ#£X¦L‘ôîÿ}Šé•ádöñi(ŽžÔàÅxfG6~ŸsØ•ëäl-Ÿ=yƒ3ëÀKkÏs˜wvƒYÛ»ÕLiÐÚ¹I=âåkøy¬ûˆãÚ?_ó‹-ÆÝžZ ÇVmð–€NQÛ6‹ß°PŽ#)‚rà_ vtšrñ½K1/žM*fÔ™}ÓK ×ÒæÛp¿®P7åK)Æ·À…Ô.ù `ºX9¿:䳯~1ÙÜÈ•„¬1¶Á=XSÿÒOÚ¬´Î¼ÇÆ!rÿåe¬¶¶¿,¹pPÃÍ×Î̼ :DµŽ±rÃÄä_/ð!÷@4ÓOóQ± `©ÿ\[,•£WÎÃ>OÕ& ùW]]¬wâaßjf¿áOžüŒž,^É<>T¤‰¬øK¨£û¬£ÅÜÆ›Ì+ô-©gg?þ²ß®Ëã0¢M‡‡í¯˜¶ÆpY?ÌxÇÙ‹Z¬M~.G`3õúÞŸ±úùOàꉹµÅCÉ<ãºØ@|%¸§³Æ ¹Žª±\¢¤÷²ciîô!š×í,â#Oݺ“GËRµ¾WÚM+ žR1Úÿ ˜ÀÒ¢Mæ¹ žpnÞ]À|ý)6bH"®GK°¬öäcÓUÃ!ÁG¨¦“HŸZùb‹>¸½r~ÜÇdþ‘¿tÅzÕH𯘠’ }Ûc“`]ù›ì0~Vz¦´åBG>PK‰áòt?Ú6M[ˆ`à›¤ãîÐSI3RrDS³õ·15”ߣ¤ÁߣEæ‹™i‘ò¸Ü!KÉZt‘çÏêËyBpÿ2§f\~E;ºÈ¯rƒÃ¾wD‘]Îú_~dvóÔ” jÝ,Œíß±éa„k»è2$UèJèt÷+í¡‰–ðèš?ü²‚®ßÿ Šý‚Vï§f´Cî65£-ë>^ß²¼‘i›ݽ×uë÷2:Íg‚Ëxü‘ðV2=Ü´$sµÉf‡•6´Ø… O¸Ä>”[!BWA1N¢ˆ dÇ&A?ø¬ÝËÒ2L²KØöjKÅJì20~4?ÈǺB羟8¾ÅêFÉ’Ú½­JûPÔ›ïƒQ×ù¬TŸbÈ\ö_·lù2,_þùM:ß‹îÒ…Gb–l$Ù+óô33 àÞ÷ÄOâ? Ð.ƒ¯X/¡·Ó´^‚ÞYŽY&F΢£P›U‡:2¨i•<¶¸-<ñë1™þ!Ù¼—€Ðß]%]Ýe勤—, ¶òr6C»ƒmHƒþNöÔ4û;¢¥±2 ?ù£Z¾Ç…Ñë¥ð 6KÍ•ïÃ>23ã$U{@ÂGê·¯ ÍVf#ÌK¹Hå8»cüÄkÓ‘*Î#Bépß{z…×,ÉeVÌæjúÛãφo À=ϰJù»¦tÝ8ƒ%L~ˆ3²“zšFÅG´A¾GÁKè–ÞÿŽnGIäf¹[¹­`±Ívð÷^~’ CÆ&s”Í–±*yeÒ[<´›FÕ«I­ù¦é u{0” C .àÄωbÛÄ;´F­«ë[$ô^‘”|šíeG%QËV€Éq ñÓ·Ù†gðl.¬ÔÌ)ë4°ÿШ¶7g"QÄó‘g ÷¡è'*q“P•YåXòºRŒ: *¾‹&Ý8à3SNdd„èùIarWÔsãUI«¨7šh¸?²z4CÖÍ.*n ®³ý:D‹ÕA’–â3=îÉEÓ³¹jÖQV¢>¼.6üâÖwH8¡ Åd!‚! õp›b„‘lZžˆó}^Åáä÷4!tm î;I_›ªéª@Ú¾œÖ/Ä¡îúÈ#nBi%]\Ud/²Üæ&9´ý0锸‹Às&® ‡½¯ö›IXÈÂyvlĺˆd¿4é|]ßÚ›}A’3cI•°¸úZ–ØÀLw$lõãzî ©“u ‚ÇåŠÌ*Žç×öJ,cÓÌnÏ—.›tµ'-?b~óÃãR¬•¶5fÊÜn†˜ØJù”Ùý¾ÅÀ’W90…4@î/”±Ä½'fË"ƒWbý¯=Ç¥’-^­ÁH k„M³Õ$g”,`xh|×õ¨Ôèù8%(Þ\(c0NEìÚk½þ¸öûuª©öÕ"Ô(H€¸f©TøþzxÞȇÉH× ™šŽ›³ÆfÏPØ÷†N¦'2TðeuÑœl£äqA§?‡ˆÉÍPmñ{1+rRS5§ÀöÉ´»‡®O©´vXݪbßn’+(0û÷ î뇃Že-MºÚ² B®œm7ÙktNÒ$ãœêúµ*.¿ èiXªxʱ»Ó,ƒq—x>¡-ñ»|ÈÂAÚà¤RTCéf•1…"F­»óoÚüØ«·ßÙâ–”ë$´Ñ剥J1~w>V©Øˆ ›˜¢ÅŽñº¤8‰Ô?9ÂÓ‚†ª—øfÊ~¼D<|”iÌòÏ·äÛ:ê8 LNk{ê×/í­4­rãGÑÎríp¹$fÁ£»X°ûÃ) }ä¹Y¨œž^Æé&t¬ºedbøñ†Š *ã’®ôJðE=þñ-O-º +™oÄO O773¡½DcèÒ…V7d¶bjŒ.í#qýšE*ìß!fvÝÏÕ5Jjü‹õ7àFΕD¬  šó=lš~š¿ßÈò®ºz/+–êýV•‹<g ’êú ÞeèåþìÅè‰*Nê=ø@‹OkáE nvO“äEÜÍÝ]}þwwБ™Ü¥¿Çl”\䎹·§‘ŸÖúœ+” hÐ_h© ¥$»QC-aƒ"ú`D>û+wY9·¬ï߃¹# ž“‹øtùÀ*o}Tø°ØåB2&Bq^žmÔÍüQfóÀ‚ô :Ì_Gñô ÞÊœEÖ©®ðÍÈ»Y3õ.áL<¿˜Ó„õë™þŒ¹;>¶MÜŠp°@Ì,̳€L\NA/!OýtÄaõCU v&¤«k©`G˜¸%»*žñgÌçG ŠQJ…ûˆ‚¤ cY$ÄÜÑi%œüµ¦6½ŒOÃªŠ«¯ ØHI.JJ&)˜|Ø8p01ý1’f¤“"4ú{ó¨6vŽ€Ñ·¢“aôzv\HÙ«öøN‚ÂÑc&DÛÔ^š÷EWÂmçµJ„欚¾è'¸fͣ‚jÑ®ÅqŸDt¢]ýóÚZ-Ö›Çòú¸Á¹ZÊí©,K¨´¼^8œ¡`ó³Õ²à.€ 6B\õǘ¶óŠnÕc#‚E@@S‹;Á©gÔkíʼ ¢5bFÀ$KÝâí g³HÞáß#÷¹ˆm÷:±({avn]¯ йQ8»éc’#é!—Â/d\`Ý«r#×±ÚÀuºfG#¼É+%ù»ÿ[Kqò×…Ó)Óî(Ûª§¦©\g?WØð²vìpÈ\T•Yd^®fŸá’ïˆ?e0Ä…àß9D;è¢Û¤Ñ¸ž¾»³™ ^7“Ü·š£zï×VߣŽÇ²‚[ø¼~{?µÄ¢«7&LIÆ Ñl^»{Û±:êÔ7ò´ ª¼fŽáýj™’¸B:ýF„‹š9R„SØ^ŠlžE–ÿ¾„0§ qù¡žH3•‡Î!·…ƒ•]äB±LdXÈØÐ^ä˜UŽqûx©<0jWy\VCºùtÂL‘Ž¿A[É̈ƒq•ÃáŽmزhfÓþ%ðtpÛ_5ï­iîl`q—з0ëE‰î­1ärãÓfûóÁ³ƒ8øž<¡Ygû#û¥R7žJul‡;84ÐËë·|h&öÍ×Aek¦í… ;˜"qËucOqXe¦Øâ%C gf§)|Ð{fm)K2®1}ò®µñ[‡%¤S¸ ì~­vdMžˆ±2>Y}Ý…ê“W"{]Y™7²ê®t2U+RzuW©µÏ1Ç\ø±^æ8΋ttãÊîÖL‡}´VºÄTÔôXvCÉ‚P°¾äMŒ¤+‡ømaßÓð­¿Ù¯¿4ìÌý Ûá-½ìŽÄ˜‹+"wÂv*õb_4C¿};(eqWªóÕgt~µCjîàO窜>pŠÒóÉ<ò¶0†CÔG{øªBa(gÀ5zʶhìÔi¹w ¯C…Ó‘L8¥ƒ—ÚÔ£ôg½r_2€îÇÄ«ÑË®!”h&¥c6Íúöañ5žqÔ ŒÝTGFÎì™'¬ȆRIµD“sÔ§ˆy<…u©Ù[÷šõk›–-†ýìíK™ë\@÷T#N =ÐëÑ çô{‹lÒ„è€siYhê ¹ÓõTÜMâ©«ÙßÄîïéà†é 3«?ÚvеÚi –ð¢×[¥}.ÄhHÐÆÍñ/ ŸÐ@~%sïE®&.ZÏQ`¦Ñ‘<(×êú`ÉÐrÈ ì›Á—“Ù~ºÇ‚Â5©îÁ™QP.w[†Nóº„Nl XÛJ€$un ¥fº@°‰ƒû6dÝ<ŒÛ²VG‰`AäOrãÏŒc”¦ÝT•?hnÇŠ ¯b©8s%*ûŽ`&¡üÄÓžD7Å›ÈáÃäÖ§BÙ‡¨1Ÿ‚:H‰Ëå¶z¼_耷ä»sê<Ø\íÓ[Îåâh…$º[kù)4ßÕìz%ÈÓhødŒ¥õã6-vT0[øîÄžÁjQBvEÔsà`h75áì»þ @«¤^‚9' “àÁ/ô¿Mª»lwÁo_–“ á ìG×¾<›éÍÆZW qA-ôM=&Í¡…æö¥bþêÂf>#¯s4Ã꾫tB›Ãš"uöþ┫-p}èá Kþ°cêØlÚÒPZê”(Ÿ+þögŽá¤{e3ЇqHßçò˜{(½x ¦z¼Ùêzøp>YõDSÃEþÃò§62r¨yÝœfè œI.1#BXf¯™E[&X1ÿð?û½ÿô¼4"&{ñlTAfƒaŒRâxm–|ì!\eâC®ˆ/8%äéR—³à àÆ%rî°a´Ûh«¤Ä5§!L—ÿrpq>®?RL#û ýxå\Ár;<øâ„JñîðÏÚ}|¨VOBÿš¹Ÿq‰žb>+¶Dj—‰á¨}£çn­ ¨aš]>…¢]Ï¢¦±åJ¿¡Qû:Rô5ÝĹ€¡ b; )hÒæ¿‡ISŒ7¶^Ú†&¿œX™S„˜¤ ¯§VÔ/¤É>b šò$ +H|úæá8„Ì&\!h ZžfùªÜ`\¦íÉΛPˆl lRÓ4¬Y‡ázùíÓë„Ûï•IŽa™N\=Ì¡€àƒÁÍ`3ÞÏÎÈ]”¡Å¹ݬãÄW¡?'hMºÓSÍîVµþ°C7h­)#Ýè^JÝE{‹¢Çuw4wô[HÈ_²3]yiX³«‰Bõêué6ñ^ˆš&.§*p«XKPÃ_סk$8–Ð Ië²ÁFG8˜)Aÿý’ý‡YËŸ7ì_f2ÕÛ½ÜuOùP­¿íx8Õ96ƒûE³=>›É]¼Ïâ&>D2Ô0²¢ñ¤ë€Y%©­Wzû·Rv¬jpr¼þŽñ^5âîù+̺ÃIZ>µÑpþè­ ¾¨þ¶ß“H†ôm¶š®kË—q­J$ŽæUyò :ôLµêªãÌy‘;/0ª|àÀØ8I/!±EG‡³ƒyÈØÞôu(“C'„ »ÿî¢]ü—E^Á—Køé Íe¿‰iÂί)‡"mú ðÖ¨üÖ¯»¾^y roèP·1Cß óX(‰£DþàÜb¸.ãÁuoÄS?qlÙ¹ÍÀ‘-"Ôþ.6†äÕxG»,G¯H¼;Ï•šý•õ|]âUñ zÎ;-X‡ý ÂÐOžÙKÀ†¬½æ@Š}•q‚“ ~¼PÿNÞ‹W+&ƒ|FÓg6KééÓB3þÜŸ<ª¦¦ÅoÍN¤4ïdK(ѪvõáŽ.ìû(œÿèý<Çnî7»¨ *Žé ªØ#M•ãÖ××bmlS0ªú¯#>Ê>Ù˵| ³=@'dâ"Å81 뀌6üõÞ<Ñé¯iÄ ¼!&úÏ&¬fx›<Þm­–ú{w‡}Pz Ù¨©v8,m]y¤ÃJo/*Ñ"Þ;`AÏàÀUV甼;UÓz§¥h—H«°A׿¯¼-t/­Q§3²arè‘h.¢Õ"oªÕn-œÿÜ`µö4žÛY½¢A=}¢d^UØT=¹ç®çlÈ,“ ²Ì qoÝu/2ìE¤º¨{ÎÁ¡ß{ f(¨PNî n– Ù\âÚD~ký9är®): ¥è¯?-I#†ñ½h<¼£:O×þ2†óõ4æó%sÕð=S ç±Îö ˆ4 ‘ÉûÊY¶$Ç|Éj—ºT"tü~ÂÊ…Úá°âUÛ»¿ÞC› \Å‚=C/ÛÑx=½¶24†j¸œ¦éýéIY.¹ZíÙ«E¶©-ã¹´£Pbq½N¡q›Srˆ ·Ýàõo믊蠰Vÿ÷9²¤êús£2ö‰ÚTK»àŽñ÷-~ʉ”!/L­·ÞìÚ¯Câßf §O¶“UíÅu´m!—«j´±IìÈì»ÒŒñîMHx%Å”ž‡÷DÅéÉºÞørëÞJظJT•H882¤@ÀòæKÀà1fìbF8tÁ ¼÷mXÜT‹òˆ?ôŸk;Iøº¹àsŒ¡Ú÷êín¢ÿw'tÁ;ÿ›á*nŸžDd\rz0RÕ·ˆ%j×½‡vÎÇŒ±Œš'þ ‚Âb›´ñ„$¾Ã‡Ó@Ä·`u/½b¥à('¾ÄG@ͱcªâ•hDÑ (Á#Ðj±$bÞHE¡19Ë>ä¨öhi E(^è'k—UÏ 1Qž'<˜IŒº¯‚®c”ÎsíÃE5ßµá]_" ;ùç4ä©·2V”Ëž8@Ó"­”Èw?Äßa„ó &zžq$ñW]GTÚ¢·„0B»öȦÜX¦t7ï">bÄö¦ñ`Œc~rû*¶.ë"O ™—f»›vX`¥•*›Xo¾hRÊéuYꔕ‡Û!ŽÅk@ý‡© ¬}ùNsýƒZz‡÷WÖ]¨CÍý.µäÞ̽¨‚¼7¡U•-"¸`ç› W>)ö yÒñ-â#QóÕö•CoØ»oì&‡bi€†dá¸íÀ <Þ'IÑ…x8º?÷!™×îQ<Ïzú’'f¼€ÿþ(k˜ ¢ø³èÚ…ÔI@y©P©»%=š˜/½8#lÓ¬õÔF\Âg!’ÜãͦîKA™Cw¸Ä·Cçµ—‹böÌ;þU™³³lvý–~¾B^‹“ˆ|Ÿîªï߬n1öžù ©,I ÀÀ³²–ë³QëºyÑ–8œ¡¢Õ¢ · Z˜¯š›$•&Eþ›²$](T÷°2¡0(’€[k¿2»€@Ÿk ání¦`ô. h%£uÊ© 3Àw¼ í*Õܺ‚hÖ'»:D06éŸâ|z4GXîÄìAƒ«]B­Ewx,…Vª Ö›)ôr+y§¨6Y¼LdmŠ2p3)õo ú;Iþ³™}©v¦k›¤cwRNH‚$Ý—{z¿”ÚP”íÏâÙõ‹DÞµ®¢ žd[)"‚WO‹ ý ³¬¯c›®ñæh=Àzó€íOŒÀ‹AI_:ŽÖV›žr½èÌdÔ/XÔíCïYÉ€ƒ‹¯ (H­®B+¥ö¾™9!u¼ÒÎgd#í×vhøº}Mk)3 ÿÈrÐwÚE‚9\ý±/P Lìë¹ þhmÓ!¶úz2‰æøN(—¯„õúd´ŠaYNnÿö}ù;F•”;•}¢£V°·NÀþ(š„ ¥#Øñ¦èjCÛôNÿ R€dÒ”C2dXCT[#Z͙ۄ® o6úZ‘è™Ë JÍ"÷|CæÈfKéòÊïÏàuEj ˆ¯ :{gà,bXÙm“ž[tWB®è>𢮄GbYE œeU$Ô°PBÝQ ü e™šôò¯ˆ?ÌL¦@Zñ»FQU©RwM?Qõ½PŒ L2¢ bœá…n±i²›ü^z›‚FÝA”ÝÑb˵ðñÓ–™I`™™Š ÂÇ ŸôN)I¨\S`^ÐÿM?Ò:h(iÞ;˜ ‰’Mç%­År—õZ_MÛGt]i¼m¥8:›öV|Z ½OÏK÷¨Ïâ¡4Õ²W¯Æ»ê OAœ×üâD;1«¼×VÛ Zo~îŸK‰x #VÏÓØ‹'ÞdÖ6=÷pD:7@‘¥Užißo}à 3óBQ”Šb½—±îøP E™}pÀØ9ÿñîsÿ¤G!Óü1â,µ¬S¡³¹//Z² ³7ËvæŸÐŽlœ6ÎKtŒÍ ×ÊàñÃÁiŸãÎôY7z¿ endstream endobj 114 0 obj << /Length1 727 /Length2 9282 /Length3 0 /Length 9863 /Filter /FlateDecode >> stream xÚmxeP\í²5nÁ=èàtp ÜÝÝgpww×àÜIpÁww—/ç=÷œ[·ê«ý§euõêõô–Ú4’`Wu/GÈÂÆP“Õrr²ØX8PhhÄœA&®Ö`qW?@ dP9€ìv66 @ÌÁÑËÙÚÒÊ@oÆðO igbnmoí Ðt°sp·6³|vwwqwqcqvbù[¤\­@ k;@LIYGFQ @/¥¨AÎ&ve7S;k3€¼µìbX88ìþíÌÀæÖÿâäÂò°;ÈÙõ/1 g{€‚„úI%Eu€„«ºÀl—þ[vuáÿ‹™™zþk:&€ë-;“ÿZ¦ÿµþ›uöü·Å‚Ì­Í\¦ Kk0 ë¿t“[8¸ÿ6wsüOê/!—¿üô5d˜ƒ,þ¢ÝììMìAz1{G7W3@ÁÁä ˆ:Ø™$<]A`sùÐ&öÖv^ÿ_üÿ d\MþêñlùWA¶‡¬]$­=AæÊÖ®%·0±û+Õ?q-пÏç_þ]ý·à﹚;€í¼þ§áßaþiǪ"%¡®&Åøß5ø'-6s0·[Ô\ÿ*jâlþßÀ?ieëïШüg§ÀÿõL\­=zl,ÿZ¶]ÿ± þ%*êàéÃÌä0s¹@n.vËïÿÐ5ssv]ÿ‘ýï,ÿñÿY'Èd†²¼à`&j“ñ#¼Â_¢xºžqzÒXƒq½8Ñ}®‡eÝoen´wJt¥NTàJ:d*¸u²)sõÒ$„ú{l<–"Ò`è¡”ªÒÒ¿výI¥ð±!Òð Nó£½™"ÜyÔ¥ëוå`åOãœÙyGµ²Ne‚XRÀXãòö^xÌÄ¥Ù°ƒv–î*h‡±sÚ*¬èÌTøa{&¢ d õœæ£‘UQEã^„nÁÎI;þÑ*{(£=ÔøÅDxߎ'Ú‰°^;[h_T]«ˆòÌËaÂ[𝽠pÌÂPš©QÐk߆-Ó°HMZþlú•8I1žqüßøq&nWw#+®½¥ÊмçG‘¿ž( ûäžf:õ2[4{3íË“þq$§8ßçÅÒ„œyÊv—m×äõª–­¬¶{Ú†-¯«ï—ß÷%U#œüð1õ®“sOLIÎÒNdበ°R è‹^¥zÆ/PSÎÔ“0êÉ«ž—o!ヅ¡>zEâ|Þ.®I~+;P¨Ò¼±¶Ð*ä—õ \…`öí¶œ¡ñͬÊ'IgÍ,ÇQeUŸ `\»x‘$2Å`]™]\.y•CyD( |Sô¢ë÷ƒ$Û+“·ÝBd Œ.ÊDΠ{s§ÙðMŸK"JØÅ^ž°‡ßJv%¸âà!^™©þR ÙÜS¬9¶k+Èê‚0Vþ¯ë"¥÷T!²è.ɳ—‰¼°¶£v<1»‘¿¬I¿ð‘£HØ£z=pmn/SßÓáU~¹¸É®¤¤¼ŽH c<hÛ|*Kë®O¡ÁÕ"•ë·²,Šð“L{=£ÛãŽú²Ñ L#ÝúÎá' R|'B©×cŠ9H$QÑtâü7û=FN£EÛÀÿ¤é³Í— ¥yÓ„1 Þ±ÐîÄùµ4¶Ôp\WÎ{ûqï,¯¼VÏ»úÁ½¥»ó{CöÑ%òÁSç£{g˜–_ìÛ‰Jqú9‚S+òyít)ÀdE ›7.\¾Gï²àÿëvlèÉm ú&…kÙঌE?ìúfùþãÛÁÁ^£ŽBÒ¹í×¶g‹a§3­k¼¬þü¤€µsÕED·ëœKº6ì¢/9õð/àúâ°§„9[hޤvˆåü†×åôDŸ8Èç—UKãIéÈïÎL‰=nÅI–ËÄ’L]m2Öè<*V ºÉS| öƒÊ£½)bžÇ#SÄãòj «’?п{ÛÚâê’BPüm,ýèö¨­Ë,@TKÓå;†UdÛ)º»i‹Q ù³ ©¸É!V6€Ñ²Uàɳ¡½CŒ…µOÁ³O=8ˆn\}ÒöP¥¡î)Ûr©á ©»(¶ôÆí:.…yMØX1/å ?Jh[yöù·š é‘…'eø·=á\‹Ö‘)ùX‰!…y˜§ºÄ­öbDÙ) êíÍ“•º²ÊfD×ö3ü‚ÑG‚B˜î›Cc̱ܭPÆÉ´ù¨î¥=½¼´gºB·)À&+ê’x5”¯à3o…Ú…¤Ù­Ì–цԟr:&дòÎ~úwccþÕÌD7¦u ›Ü¢î̈]Ñ@¯ÚÆÝúÂzo7âÍžyÔ ø·=Ÿô²Vmà5sŠoÐAÆv…;§¦ŠÛ¹ûmgxŽ©ÉçÀã±æ›$¿ÚðÛè¢Ô¯æïÙç¬ ¶ C•[ÕÈæMõ2‚­0ÊœpŸ ºòR)Ý'ŸG±à–K(¨€º½ïMNÙ_Ë2ÈUÃðÌkµLÜ×UX`›O‡”ÆçÇeùlg©üèŠÈMWÊ!é_Ívá@®&µåWUðÚªWéÿ"ECub‰ÑO§y%Ué ª… óèôËíŠ.VFpÔèƒX­(‰¾ ÅüT¿ ãBóÀAÍ6Ä+þ—ê+&ˆ\Ø7è•ðå¥PyÑ“²ú5ü¨€4 °ú²â{¸MÂG¼Âf¡ß‡ÉÑv<"“pbÞ¼´&šƒæ«÷°_C-È™n¡·Qr-ÆÂyhè©é‹>G¿„ÑH‰°ìh(ÃQ;b(â7È-²¥tBãwÇã®ÔçOÔ5÷áA íO D·Ùo¦¸€Åå‹Jk)ø¹¬·è­°}r@¡ÂˆbüSÈ8¦f(×_ þx9øaïü}±!~ŒÏãðZ ƒ{¯Þí¼M 'Éd€VJh"tNCjŠ£lîYq‚–Ö=  [Pú s‹àpŽ*›¦´vYbù&¥Q>…þYŠd|”3¢ÑH̤P|;EþìÌŒ«‰ø c!¨·áöÈ^œ«ûǾÊoõaΕa|Kˆg§Í1íd«f`Öä‚°T²’ #vI=NÌDßb¯eXGµq>| s¨ˆŽUÆõã ÿ7·ýèee¯š’=÷Þ,«îS'‹ôâ$eÕžÕ™lìZ¾ÓWŸ®üÈmLã„A` j©=“iñ][mÇæ ÿ¢3?^€wÑ”V–ä羺J…Á™¤îa0‹ZvÖ|Ü~s¸*÷¨*´‹ð¯ÖpÔ1#'ÎäÖUß‘ Tf#(¸E{/mÊí7‘£“JÏg¨á9-f›‘:çºb*e°;õèñ-¼I‚É“y4)›Ò}×=š¥Å3y%/D²°–É@žåHŸ¾Ò´Ô |ü1“t¤`ûëÁ¬C—/=Àx §ÑˆGc7U>Î@?èÞ%SλŽß&ï 4×Ùl~”Џ¯Äá6Œ’EQ®ô†Lé¯<ø ‹NÆ{«Õì°þÀ…e÷OÖ'úd×z,Kh§o[ÞÈ~¦°$vâ:¸ÂCAYô|¬u¶÷^ñÉLrîX¿’è…ù¿‰­¶žàÇÊ'ܩԢӥü?Ðë‘9¼Çè’-ç)`D‚°pk­:Ë÷U$Ø §Zùwê,Få“îäÔ°Cˆ›§«˜}.lSº0 ˜µR?uP;vç©=0/!¢Q3B²Ñ ©å5òØœ€iÛ‚Àym)ü\`fAmþ÷&½|}NÍßXXÁbž§!³‘}sZt@ƒZCÛuøþ¢‚tàtË ©úIºQð1…Šfs½0Î:aýÏônØM‡}Î/fœ‘ë‚aBï¦ Þ¹öX¨œT i†ùž¡Ä·V¨dƒûk'ܦv9¹YWJ¬ç¹˜¥ŸÿÈwA¿ïõÙŲà?¿W²‡äN¿ÕÁS°¹Ò!§Le Ü[* ó•§W?÷‡„^Û<¾Ut0ç||þZùAŸ+Í÷J™¡&ÁO‰¹{É÷½ôÛ-­•AfÕOI‡ •žÆà$é±ÍsùDÿ‡:;ê/<‡ Ñ wã¡F{›©&¿Â^ã¦E¤} LŽYQmŠ•ÊÖ×ÀÊfŽÔÆÝô_äœh;Ãe|Ô®<ý.gpå Ðĸ+êí-Ùt”`ýÔê^ÄŒ÷²Û› ~´•c F(!H6¾/«O£ÝÇ(“ð®-fU‰ŸÓ›h­â 8Úo°ÑãXŲ S ÂUZ3¡7Ãë½E2h“Ó*#ËEáoU3™1o‹8¦/¶n62UµÉ™;13zÚ²l}ÁéPQgœ+c}ʵ0üh>y“$æ[‚J¸MÿògÆ]Ò–ës‘¡¹ÍÆMo‡&øÓèÐÙ~em›ÐCï\÷¾xg³ƒo;nkžˆáâþ¦¤x€&AvdôñôÄá›Sýëc›ò.Šöð*àkNóÙ¶ã©f°íÛK.Ïékd>¤v½´6m¦ù[Óz®O‚ЬÇG*Ø‹¹‰‘ÅmS(øŽ¤rPçwãúMUL¡+?[¼±[ó•,ŠäOÞL]FÜ$’áOU^Š­”–T¸âF[&z~£A¬SñË¢TŠ_ÜMˆþ [p4h¶#H_õóµ ¬•{ôÙXΩ©%@…Ë? ›Œ|z "ýÜd¾‹BíW2C.ú"LØr$²Sÿ„ëönÃßwò`âk–õ+«1U Ê+5“«‚8Œ_ÿeÎÂq͘ŸKØœøå¡bdÙ¼U܇ Ì•`W‚tÕWN{V—]ӹɻ€Jô²ˆ’Tö¾zèÚˆˆ °”²vøûؼ _oaP^üFú°Y­õwUÙOu"üçÍÞZ'ƒñj¼ç eeÀ·r°†,Sˆ§3î3^÷š3OÁÜ Ï@GWáãïy~¦~(5uª… ßwÞÞ)‘Ц.sy4…j` º?GòÁ›Ób™AEIÑ0,^{­˜›çâ"uÖ7¿7O9ª“XÔÊwÝ®¦aªo’E[½òpʨޮÎ6¬¶h^{u×ÃǺ$óDá%¸‰Y[ìÆK²ŠÚÓÔ8ñ ]:Þ0I@4zÙ‚07¡ª`‡›(á÷Ià3û 0l¹O >4ÉZÖq+ Ú¡Ìž¦AÑ%WÝÏô°a \Ç£`AB|mQšÄ52p°Ñkfòµó\l^Ÿ*Òƒ·qdœŸ¸ðøf&zÈPW»‹}ƒcýBú†ÐÖ‡‹>$eÎÂÕpyçPÄSÖ{Ž€Ò.éCyTHñÒ$SyÅ$`Ì—+*7§¼ôO“™û,»2ˆIwH,~mUx´¬+¹R—Ö¹/sŒq¤{2ëÚõûñ§…‘•1øANÅâN„ši?Ý;•wS¬.³ÌW¼ÆÖàÁ+ÛV¡8OVþÅsý¯ÏŠ&˜ÖúFöœÇ¤=ºVé¨ýޏ#;—‘¹ØÖ²¤ºÀ-sM?9+ëTѧh”ÆÀÊ㪨µ‘Kuy±<ݰ7Éá&³? [ØÉ̪}±ÂX˜y™ÜJ=ºÛ¬d¯Ñ][)ŒÛ‚¾Ñ â5;]¹½çÔuÅ®ú'¯Nú朊ŠÑ‘¹J„YÐ&VŸ§ýlŽý§»AÊ–¦¼ ã:Ïv˽„kÇ̓/ëŸ&Õò:7|p«ÏÜ–§žß$FKQ­“pWôtR¹ð&{ជ‰È&=¦Ð-#Ð0‡ÖÜÔU–„jƒs½NS}9#=Z<#|êôˆÅ!Z°mqÖWÏPB}ZuŒc‹+Ôóqq·>qêŽXó¯“ã´-G9>YÆMoGµoï‰ röíäpí¥¡²¤àtÄcGž¶­ùûÛñ‡Î.ºÒ@ˆRJ5UØ€•YïþÇ‘©,i‰òšÛñ':RFi*!@·¥E«Ç•Ô0Ý C\UëâcåÓ »xKß*gš›ìR)ތϳ†Ww¹‹×—îqX$¶‹•¶ÇˆP´Ìʽ­¥µ[þÞÉȈ1Šö·WŠÃó…Pt@Æ÷4I“ï×/´8œÛ·ÀL²þ?~Ì?æž_öl$ J„ ͱ®)9Y•Ü•»îŠ%WÓÓ7™ˆ™:ÍùI9+Tmfò»™Œ‘M&ïåú<#Tû!U,È7˜ªGÚW›™;‹jÔRܑВÝKcȉ¼¾v ¯ì Ä?QÆðL<¦gЯ¢x-Ýð˜àmkϦ÷òò4‹&|†'ÞºÙ<Äc=‚0ÿÑ=0óÌ~W¬™(>G¼Hš³r&¦Üé=S66î{ó“õ%¼:wÒæÁ˜Z¼äÅæ ÐößbÑO<”¡¡Ó–Û̯kêÁ¾Ylï å=K­Á"Xo$[MCå¹(´àqWæ'k•pÛe€œˆ¢bõ„Ù;(:V{HNäÁ6ùºÄ£Œ¡ÿè…Õ­.´¶üçðklÔç=fO ú…ØÆE6†Ï¢XAj“E÷ÃäæD™ŒgB£{aœlà–ß-3ZÙ FÔŠ–rÄuÇÙ}‹ :¥7Ós·e TŸ ÀëŠ õMzâc9\Ša7fÈŽ PÆcÑÙz>qü9¾ÑÚ"³”Ø÷ã+"Â)|tðäÞ_} ½?²bKŒ`±ö¸¨&X÷Ý鉎ô Õdkwect6 Þ k±Õ ÎI8Ëcƒ®æ·‡Å.ÐÄdœnÖºñb·<þãõù 5 ä&xеÿÖ‘Jsç§¢É!ÉTÙ"&3ÄíŽÂÀsRúÌÝgÎî͆uï˜ !ÿ“œ3†—{¸-ã gŽÌ›ttËÛ•Ì_Ajª1?øçüêÃ]Þ-ˆÇGÝdF>` ¨­~C\+Ä=YXàÁ¨V”™`/ºÏ*‘Àl(¦ŠºÕSot¨³@”ÊUÚ.qÛ ]…!aÈ_¯!y·VˆTG™×“ÌJP\Ã×NNý8g ­”Ú8|æëB ñ•æm¦”$k—ýîÄíÈß¼(Z“YòÒ¨S£¢ 7…ÇNfuhÖèwskQº·WëÔ«hó»ñ­0_~̹>¼$œ·§™©‘˜&"ž¶ Ä5TÙÇÙCƒ«?yÕû4ÕEšý-º¤Ú‘iáX…Ï–“êîï’kÝlžôÇÄäRþ¿9€¥RˆWT>¬OŸbŠou ¿S쟄!’šø,¡Ë›®—ŠR^p=¾.^V¥å&Z=ÔJDÆóÜÐ|×q¸ m"£Ç„ãà^û åœ!Bð3Q[·-¥?Ó,³X*Ýk¨½¸3úK7Žˆ"”lã7í´;''GµÃ'§€Ú›Ç‹ÖŒo˜ŸeRIç’ÂXº5S*Ч²å“γH¾J“Š ¯¤ <÷VnKˆýö¼—í7Åð›†â«Ô°zWmßF²`€>þ¢ðŠþ3èúÀ!!aU*Å@&ñçd=Ó"ýRv¯í\”Ì驲!)"°µàÁ‰ÌB’a ëæÎ¯M¿¤åI Ã*W~Ѷ3Õ"´¶wX¥éÑ[¦«ˆ¸—çlÛÞϼ¥ãödX]07ÞBК¥ëi¿z7hC± “»”÷V£Ø:ó²QJw¬x•7}™fA©mI.ß’˜õÕöÛNI@çî-Kᇜ&_<Ú´$ž“ýè’i–Eï·Þ§öjŽž¥v)©È4cýžžÃZ  L‚ôï/5W–Åî- †!Ͼ¶ñqJmµ®ÜƒA£ }ª(ZŸµtA‹5˜/}Xn– Àù|æd¨°:¿£¯ÂJ…|D¬u[¼6j¡èor²Kˆ£âÆ!-ÛìLg5M#`¡‹Å˜û•Uý°åþàZCˆV«nBõ8kG¦:¤þIÞCœSzî“—îQ™ÁÞË‚ó±\k…ds¨Ï*̺Z4v“GEÝÓý’À>8.¸ 61Dy·ü†+ª|sÄÑ`éÎø2ããúYu0í\¸¡ôÉŽÙ“ä®Z2w %Ü«/¸¼&ƒ¯YoÅìÒÂGY˜*ázR ™wÃÅËü˜Ä½ñQ·¦ >Ò.][+¥Ùî :æQ–`y§r[ŒDE•Ì®Øf¯YÑÓ±Jpƒm˜ÄMD&ý:j§rŠüêqìz$jóqÀ?òæ Cg}àî|؇BÉ"†eI µ&ä;xÿ«8¶ïÜ„RyZa•Ûy%øŒÍ@¦)úÍûëËhj×þCï‘ _žÛÕ3^{ÞÞ´ÏP4/ï3ŒEKjmø^![ÈÙÎöξé^|e¨Ø+“NÕ?s{ï ‡ÇÑkk5eãœyåâWú IPü#ænò«2’1Ôq;'Ö~ĺG¨vMoŠ—I+9ßYªƒýžh÷i PK†©s@“_²$¥Îj/!È ¢ü„užî4p?Jw 0„ž!¹XÈ´ô‹£Û±~’7ó@!"b¢þ§Vtáˆì&èy"R šxd›7ÇC¸~õ†óÔÀšã‰hi7/|Vuû3ýÅ-¥–ú¤Ò¨¹I,.çÚoÏaW][çò]¥s¿-¹‚ÑiTrŸí¡ïšÁ”ˆGç!ÿ÷?áÖOœ>.¨jI¥»á¾ÀŠ<ç)üdÌ!>‚ˆRÁs…bÜ»âÛž¨=ŸÊ ô?œ._&ì t‹§Ô·ÌE–u—žaМzÒ`&;TÕ}Ùü ßÉP ¦Æ]NÙo/6v²OØ¢f©ÁCßÓ:»iš¶ÊÖ•s©:áø@GÕwÙpߎ^Öå „µÇ*`|_ïb±•J@(ñ X/îÓƒ(ŸE œ(Ä÷e¢?'0–‰*^^?3’Ú°*±/Ç‹?wœu1£:¬_“0—ïìLªD}v¸÷…Õ3Ñ'¶k’cR9EQ¡àé÷/ÜšNz0¥«ÿHàÏÞµ ·az™´^-Öˆê4*.|!Ä] ç§Ûâú(c±ìõ}ÉŸcøJ”ÂÝ&[l3BHy9?~.•ß °¼R§ ™0ÉW<šdô¹¯‹ÝˆÑ$ÎÙ}[êu!q¢ »ÄbÊJʶÐ>]„þ› áÑp-ö¶3÷FËq’”‰í^ˆ¦ÚÝ™Ô|ÁBèña%€vF`Ù ZÐruìÝéš3cÜ®Z…f¬Êo ’$)¿ŽÜgm?‘ìÇ6A¦¦g¨|„!ççÞP;4)÷§5ÅeëÎÅMó^Ó “ÉPfS¹Ë.¶æ¿² Ã0…¦H1¸y~Tü)¶b×!ëá h3`«?0pý&|÷h©`C“¥¹·²F¦ÛÛì9nyóÉë:ÏUåH¸D«¨Ùí„x«_”‘]3{^ý³4wn¾ Üœ‡Â"ž…")’ I³I­AÝR#˜ÖÞÒ¢!ý­>D§}™z»B·Qá§ö:1¦³­^œD½V¤QßòG2Òñ*mƒ2ÍE!…¬h>Ýœ‘PŒÃá»”ˆ'?ž<äLZ¶ØIÔE}àü„ ð·t±/ À¸jcÀ%R¶ÍÄ’÷pæöF¤©i²|ÿ¯ _2NôùF׳ C‹éøÂœ{„nJêªàû¶= ’+œÒ¶F©M77®,´_`ûLðþF³©T3´]Ôçôö‰²ÖovÃZö—Ÿ¡g€ãÔ¾\tŒ[¬€ñ\ËHaþÅaò|.^] Ñ­êÇ4Ú¨ÖYäú¦³¨/>¦è€NDë]„Ê£©âXuôèGèà³ç™Õø€1®VPÅW8’ßr­ÂшuÓQÂ$°ññC™œ!¡{;±Y¿gK l\°1´ý&âQ{5lÊj”%”ÀlsÇQEù >Ç(Ÿ—8F±FÕRó7‡S yÒ»'ÑQÏc¼W ¤ðÐËGcºßèY³­Ó½ØõÕß ñͬ~û­³ÞÖäòãO^BBG¬Æ]ѧ¯Þ¢o8Oz­^JŲ^Ä8›ž¬’‹bsxω"Š’²±ÁáÔÝaÃy»W÷Ò®ïüª%vKnÕå|¨Þjïcã²êV†Å2y0‰s!ôì]™É:«ùö3M·cÇÊFÐuÚ‘|ÖÅ‚jDß9“Ÿ±oAÑ9…Y[ž­ï,eÏH3 +vFÌ_yY<åtÁ.ÁpÉß×? ÄYÿÀkNÀœÈéÈ%ÅGp•¥ïë4©UIdûò!ÈK(—=: Q¬#‘ùÔc{5{@°Y¤z¸Þ³Žñî=GLOèƒÄš3BYÙ½Üߦô˜?‹‘uØ„… *ã÷Û²½kûü8ÁfÖ(2BÚA1m¸6ÃÍ0ò,HCûn-œ/ißù“(ÅÝ¿.·Ûs“¦Žj].‡ÎãŽX4Ÿ2¦æ«Ú&LåéŽG‘ŒjŠà?Ü*­Gíñu¼ß®°štõÏ6æolaNÌ e¾ ðÿÖ œÊœ¹/ùì â«ö `Ñ/ä– ò–æ8 ùò€þm7/Ò–SÜ>Á\9AÞ†jF)ÔæM9d—=¤B÷Ó.óˆ ܪ‡ì|ù…ª_ph¶yŪ/ΔÕÇ©¡dßh,ïÝüî~ӤŪüUs=êY Ž-åS›ˆÓ·‚¨,ïlPËèÐÏ8Õ\ІVŠ*"~D¥x[ål‡"ÞÄñKñ⛬ØÇÃXø)ë]çÞmxù‚ªÈ°’KÑœ“ºF¿rgá[¡7j»’…Æ7³\û⇬ÀT†Nø•ýHº%ÈIÍ `­¬XT„ÛÈâÀcÄ•r7úMŸâxðCÞ—ð¼è¹êÎË‹ý/ÄND7\ªpã7¥FíóieÀA!É€»,a"äÞ»qab/jÛÇ:K,ÄeÄ€Öž,V÷ÁלtÖT}´ñ­õ|š Ú©Þ¹§9.±¨g#,Ö¦º"oRJPö@’…J]=áÄÙ vq®¹>&·ä:ÄWb-zÏ^©ãæKÁƒŸf»õg‘£ùdp—üÞüÕ3t"WÑU)Ó \d¤åþá"Ü»mEêyß*N Owrò”ǰªŠj]$†—8\A²ÔÕ_£ƒÓµ³0¡÷¥¾ˆ·ýÜ@b.{Uå´Ð/6!+©êË£t–œÞûtv,Q£YNØ …ÂÉ?+øåH9øÝ:@sQóù žÀPs².Ó6𦧛¨;Fmê‘cK®=Wgºž`Uj‘Sé±ü¼Ï‰ÊK—¸‰bÁxX„ç %fj‰¯3ô¢(Ñã4ùùm»R:'Õwc [LÉóÅOÊtÆûÔ%ó"ÉôhšU;Û*­/>lvÚhØcðÜÛ7|sýÛ¸b¿wû5ù¸»åSõ8¿{Ì÷£¥y¬ëû÷såfNN½NXo²Ôð‡ª3›åµö5wEûÑ™ÎH‡Ë™>Ss9R¹­•3ÒÝZÓµmqþj$Òý\PþB endstream endobj 116 0 obj << /Length1 721 /Length2 25307 /Length3 0 /Length 25873 /Filter /FlateDecode >> stream xÚlzSnͲmÛæjmÛ¶mÛ¶mÛÆjÛ6VÛ¶»WÛÖý÷ÞwŸ'âÆ|ÉÊ5rT>äŒ*RB1{;OSFZF:.€²˜’,#'+€Ž†”TØÉÔÐÅÒÞNÄÐÅ”  njP6u02˜aHÂöžN–æ. cÊ;j6†&–¶–N5{{7Kc ›››€›³+“+Ý?Iʦ¦ S€™¥)@X^ASRN@!.§ 7µ3u2´(¸ÙXd,MíœM)föN›ÿ,Æöv&–ÿâäL÷ovn¦N.ÿ3s²·ÈŠªŠÉË©D…éU„†v&‰Òí\œ¹þA›;Ùþ«:€ËÿX6†ÿcýõ?Q'ÿXt0ŒŒKc€‘©¹¥ ý¿t“´3³°ýÇmâêðßÐ?„œÿá øGCJ€‰©Ù?hW9C[S…°½­ƒ«‹©@ÖÞÄÔÉ dokh÷_”¡­¥çÿ÷’.†ÿè hgþr ÿqY:‹Yz˜š(Xºü#µ™¡Í?ýÛ¯núŸs‘55±tµýO¾é?züs¢&öv6žÿwËÊø÷†ôJʲš¢ÔÿÓÿ‹ÚÛ›XÚ™”]þÑÒÐÉäÿ+Zþ§{þKæ¿Ý`ükYC'K€6ݿڃá_ß-Ýÿ‡²÷ð¦eäàÐ231YXXœ,L¾ÿ‹®±«““©Ë¿ÿ§–ÿ®ÿÝH¦¦¦Æ0ë+öÆÜ!Vé­a~¢¿ç*Á©çf T©WW§{o´‘-‡-Lô®°ïU° Üáðfx÷.w%ï[ˆ ãMáN&§b#tW°_Kˆ+Íýj·ß ß"ô®ÁÔ°låÀn"ï\27Öƒ¨¦X²óÎk¥Ëx‘Åc Ê;À‘vÖíÄ)uÉQ²´¾áL¨kæ,B‹®ø_÷×A ‚÷ànH±ô-Š*ŽÂ5X)Ú9w—:ho%H’CiQÀ¹<ÁúìФS"Ò‰à(ó"ŒÍÌhPÜPœÄ©_‚:®G)_€ü¦GÊÆ"ÎÍs«þb¢É™®sldõÜç¤d)Ý€\•‹¢ë*RØ%2ÿŒ9˜ï‚2]J .½·ÇÉÙt¼7ã&ó…´¤w† u3”£aÌOüVч:P•¦`£ÎlËû¢À"ê«ì|z©h?X/fç¸ç {S‘Í…Ž½•vÙéC؆1‡Dñn¢÷ -¥9ƒP[onüá’ô0 8ÔÝAê¦-xøõCgÈX½<’N¿*Oùó%ÀàÌcG‚Wf¶bZ0ÕÎëyî$„£܎DŽ©uÙt;¼Ÿ¶¬Õ×Râ9º*²£qø»(h×¼@Vë|ìïa¤±O -+MšÙãj¬?^)3çÊùz)ËTiE.¬gë4þ_êžF,®Aøs_0œl8Êe‰*è˜5è“…Ù±qÿ%úÌï2KßK/‰&Ü®¨}¦5™qéênE@&QøÒÁÊ(’vÖé˜ølÎWh‘ ÷ ¹{vt?m-u ïÀ)Cn‘éôê ¡÷(ò6]â5ž¢¤P”duiì>Í]NÒ¢éÆáUïËDÚE²d?Ù­~‡d3¼rè#¾¢ÐŒg¾è3Э°gD¡ìT-°ûBGŽZ»YtÈä‹å}lf‚áÈrT,$ЋLÛáfPk‰3ͲÈ(ù#"¥0÷†–¾ë7˜0]âhMSPÍ¢¾XK¶ÑAxœ¹Hæ@h [ C°z- @Ñ5D²ï´|š–2Ý),há¹/8Ÿø’›ok5Ñõµ‡Zü’äa”>ñ[9S·¶¬lc•Ñ kKÇfóaÚŽ,y@Vö†zÝ_.9:…k î¤_Í£õîJÜŽ75ìHï6bvLém¾·’æ`&jVÈ‚q wàÙ㬘ڢÇçÁc‹D7÷’E¬)ß­°yV. 4LêÆŽo*Âyã’Ë.ÞcÒÇ4P=:À*Û¸›ýçsû’ìB› ofÑ8VˆQlUOMѰƒwþ“Hý6.^tfÂ6ÑÙ/çwÉ•îqÃP„ËU N¢Îg°˜œ‰GÈòÆYk/lW“R£÷°/‘žàÊÝüÄÕËf¨ä°5]&ž©Ã¥ÇqAc%ú†g–sóüa‘ z*ÊqåHúÐ2¨Om6¦r³up׈U97_Í-}:¾üžÄ™l.v˜¡šíåøsä w𽂛ýxÆ(6Ô9~$Úh©Ý¸ Hô¦ûëIt_/Æh æ'¿RÛ²>šÂr°÷¦|Æ·ÅšŒ4’⯫'9øà]ciôdÊo¼у“m¸‡½ãÝ´'¿54±–WA Qf%/BXe ™øc$P—VK·j!!æ}®å&3x.æ¯ð?¿ïMã䄨 à/½‹Òî*²ŽŸ^Ñ×]ÃG'>•Ré]»çÛÆ »ñ²5˜­VhËÜúŠoI Ü3 A{ƒ÷÷õYS>ü1Ye½®\Ö=pí´yrÜ Ã9b9§0à|©Á[_z@_"µýmSªW÷üó¶% µ[~ìü< 2p˜©õþ2eÄ%ÝÉn;÷ÿÁš@‘%ˆ÷£fïù1/ Šn9Ô­3ýÛmåE«‰‰à¢Jl2Åãîï·qaºˆÇº½Rêû šªqXPÀñ™a¿“ŽSŽh§¬<Ÿ¿€gý)ÎüSZ“lËÏs‘µbRâïN#LêX*tÉhœäSu4|ºD„²e  t6?´‚×2ÅÎÿKeîTÖò¼Ço†Ø5úMø=þ±×Ïo#c„lð÷ÒtmD}Œfâ,SÁ‘1X(²šÅ6‘»z‰þ¢ƒï{bÆN? <Õíii»çD/d“äåü¯ÇJNwâKƒ›  ®£¨ú›‰]¼JB‚»¬Ð|á½ö¼Ó{d&p¦ZL‰ŽöqÚ-2Qxå`×þ’B®­!ÕÊÊ`/çŸ rrœ#Z§BB€r,±ä)îOqŒu„Ê$ÈÝvÉ‹ ÍHÓÑ´ŠPXþ¸ÜF×yo[°ncx®^Õ©“᷌來]†Æà†‚"ˆ•±“U'ýÎzcbâÞ³ÿ°¯FŽ`CŽy®.^ ?´šË—º¾€ÅyÄ‘}B* „ÌÇ,¯:4ŒËkòݺbš•÷sbÌdL¤wÊ~ÚWq|Dƒ®ˆaw´!ú‚ˆ.É|ø/ ±…ëÕ*ùút‡9QàKX[l­ëªãhCY³8l[X|4 ÔáÏjf¬>ñ½¿Ê¨ y•ÛÚ¾•‚6Èð¿ØÂˆ¬¸­¬/áɽ§Ïöê¡XœýJJǺáãV\Ù⇾wöAz@¢ÜAÖ;(žÁ‹Ìó¢ñh»âpäQÜ¥ëÊÕ‰ä Ýáù~[Ž;.\ÎðÄS (¢Â¨F6õE1w‡6¼^ý5{Y-”X‰ST1Ì÷ Ðý€£mß¶µ,Öaû‹ª-±D– ÿ5êR9êõ¡ÖùÌ“—ŒOæ #ɤ¯u”RüéE¢Nÿñ˜Ì1¢Tæ7µ,£@ y…‰VŸø£ÁbR’3jÍŸq÷ͧ~Ȧûë*Œ.VÒ#n|+j¾ŽÍÛ¤éûyBåŠú³dJ¨QêB"eâú û¥Û ?Ï”Òù5óß 0¶'5±ÂÜã^hqL Û†š ÍZÖâ)…Ž’N°GýD|h ®”eÙô¿ñ/>Ю޸Ąª9Y˜šiÞ 8·:—€æVUê)I-ð.`*ùu:Ù€ |Üœ¤ºà)¿Bæ‹J6?p™GFÌìÑ޹æêûÿê¤4Ìḑ)›Yz‹ç#ø¢Žynçl9©—3®g_å g›\8%7¿oZ|OD—I”òØEcó*øŸ‡b5ýŽ7 YeÃŒ #\"E5ù½¼à<øàYïayŽ{Ës¡ûh·§‡ ­»ä„Á0t¨d«c²!³&ÜÃIÓ8L¾BˆÏ™`}ÎÎÿX„ËÕÉ¡²ÿì¡ÒRÞÀä–5ÊKóá ¸h²ýh3œrŠØ áâ_ŽZ(áʺ H1(ß’b£·–)ˆÔ¿ýò¿Œ ¿=*öóÙ½xä°„µ³mÿXŒc9'ð\ †qùé³.·2bœ¡†5þ}V^;.GRI¯Ã?QœíòüBÖ>ŽMì÷8U†‹ÂŠ@‡ ’ôPn]±Â®|@tÂú`á b©&¢žY©›BvÛZ©_ð]¸›mÚ£y‘(̦R¯Rr•p_Ù{p޻ɧ8¡ÚÆë¸a’Âògwo›Ãc¨ë˜²6ë:¸¡7>Èõ;kÖ[i/ŒSªçÜ!8v_9"a<ª:R«8VD<§ÏßGððiî´1ê®°»W£E'(JÏ{2Úa«awÖQ‚Š4ÕÚÁod‘kŒÇÙö´Q[%CI•rˆPûJÀ¾ãä&ØG?&¢¡ò 3ô:Ž»ã¶ãI“É }š•êáY_l[pú¢êú™KçUÓgâÕ‹ŒÓùLØÎ?®Ù~ŠåØÃ²iÞª5ùC†§`6üÚûÍkìÿkF¾» €œ¹šéBZØ1ìIwý¸µF„%bs©5Ù•¡üÅ£D¿·_¥·=R¦lÀ”0DÀë¤\DwÞÚb´e0óVhÎ y“s1?]ë·sǤ2(¤õ¶_T÷ÙŠÚyù¾ý8=‡›EìcK¦ 6žýòdu|Íÿ3Ô'f¿a³d¸÷³ óIÂlÕ7lJ‘(n9cËJ²­q`²<þA™$K“ö©gT6àÈÍvŒ~û¡Š/ôdç˜$¹Ö9=íÇ!M‰>Ó·ÄE]Y»“ÔüEj+vÉ|1 f¡žËÍ’>À ‘ðd÷<«ˆµ£²½¿nYĆ=—:M( Ixtðy˜JÒ`õ‹ØÄ후ržn ç¢ýs;íWâÏ/“BæloIaÌ* èD±Îê>áïr£vqI;ä§²s«ä{’ÛùFqV}í@ÍZ;ɘuÁŒø8öQ¦^r`؆ӓ߽,CY²Eàˆ2S<.œ#¦…4h}*ÀþËJó°à˜Öɶ+Ö”ÇǼMœۇȹ#ý•‰1q»%ù“¿Ã«‰sfþša}Lî*‡¬ÅŠ©ECòen’k/8{Ï³Û æOówŸ²·LIi¢¶pCê˜xv)¸Çˎ̳bh÷v…îN,›Ý”siA2Ä}Œ„îz­Ø‚å\·)^G’@ÉEø¾Zº”úú Ï×@ýÄâASþ n^­Y¹ -!Gµè¶k]ö ‘7!Å'e|IœÿO?Ý»gX¤q¸1±ýc«Y"eË];èÞ§†Nœ!³ø¤N³5Ãé+ ‘c¬Ü•Ü";ŒˆUX7•ó1±âðõ>«èÁUƒ7‰|O3iæÙ9ñÍ.£¾Þ(T×ôyíΡà}p³QÏBà Vé}3ÖY*É]:­Ü²XÍeçÜ^þÂ'9-¯™.¬v!¸U~¯“ü¯°t©øÍ$ûn]ãFA·ñ_ò±u_­i^ëõX‚^áv;ÆT$C·¥š'÷]Íç• 0WBÜcg%)tw`ÂÙÍ;u5‹!ö×tv¯i㆗Ú`ûcDïKnCЂ¹ä¥ö+].œÞ±aŸ½`ýÏš*/5+é¸ì¬ycU`Z¦qŽf¡ŽÕFš ˜|kl݃sÑ‚ë+¸yÆ4Ì G5ñƒ“`l!6ÝÐà J2 ˜f¥â®!¡Ã±¥MIã€-á½ßå–î6rÛìü¿Åe³íÍÈÆùt“†}¾ñ(9ö/ùJ>Ñs­ç[ìsÎëx¶Š|BŒ4¡„VÏÀ´`¯ÕFÏ’ú®0§ C!óÈí'7Âü.þ×4 N†J­ûГC(Wù$¸ôg~Tt ®‡ †gå'¨ÈËwÏUvî mΩSõæ|1ÎÑ#µÚWº2Šñ»IsÖÅÁ.ÖhQîÍxün84ók²‚¡Œá ?‡’k(3˜dýÎ Ѫüà —Ô_òŠQváÖY•OMÌZ£½K<é"~ä´ý…'™`=<¢!Ñ$м ïqU˜ßÁQ­K½¼²—äØÂØåÂ9Î(+•çöoÙÛݯôŽÂcn­L^3Üð„a X)Øáµ~ CEÈ®°<ü1ŽÌ¢¾J…=èQ š¤KÁ{qÌ»å%DÉÖ½RXú¶pÆ{,4‹Iî.ö¤_2ÓÌ22ú5xjM1B>DÝïmÇ|¤ž’æ&•ü)¥¶TJ£Ám¨À×\ y\¶ªÃ¶õyã ”ÉÎgWw!?w˜ëx‚Ü”Xï¦.J°d](s·Sb%îûûTÂPH:¥ îéïn^ë”|NUÿgU&Zwv1ìZß2ÜŸúyG®k·¶Õ­B™~Ǫ„«Í-µXÎÙýðÍ)¦)æw:»ŠÝ¶¥Ý/7î˜ýøòÑË©zÍ=wF›Ì‡¥uXµ@…æ÷þaßKpÝ~ïŸ<'¡1ù߯‘ú¥Tc,Ôkuº›ÏÚäýsß™`I½ë—RΛ`¶5v1"Ø#ñXR±FwÈ=BCÒÔ¢µüY»¬J_eÛ¢/ÐkD·cør'ºÓ½tœ×ÊÀMãõ,‹zóßP9(g]wÑ­Ý®ÐEs¼òL ™¸+Ðàv½„— ÷¾­p€;ûiÕA¬¥ÚmþZï—,˜·Ëû[^5iS.N¹®Lk®H±fƒ¸ð¹ûƒRÔ}Ò)œÕxõƒ¥%<Á^6¾’æo.˵½@N¥:É3á­›‚¨”àíspÿ¾M;V!üä[²‡šùN Û}îH $Òê’. vúØö¾¯ÓWe¹‘™©Ç?°øëv™€ `ÕÇö_s«;@ºùêT»ÈÛàM„¹Óao»“J0ysó?ã´ï!1\wÙ²Xd$bÎâÅL1G ™¹JržU9lÃÈä…Å1ZYc‹£ˆ‰èoIÖBŒ*æ¼K>º[æåÁÕï!_Æ ƒü­i6v¨0·aÖò~ÂÝüÕ]zÄŠIâð©­"èÛÜÔKÚéÈdÕÙy™r·¡Ø[¬œ¶Ø¾„ñÌðSçΈ’+×jBŒ)D$Þ\Taeç¤@ÿ p‡lí›ôiñ»žË–˜Ÿ&¥;ð_7ok(H=Æ¢{šPZ2n,Ííê,¡ÓÍLQÁËr»%ÎÞwЄá¹‚štÑÞ¤Rª~m¬l»ª³w*§ü†Á0±WG®ÆéX”²ðA 6(%È‘Üím”IçÚJ•Îæ({€î@ä½MY“d} j é†}œeuHå$ºE†„v,'s¾Ó2‘¦©Úß)Þ7§ã? ¸ƒ^ÐÜíœR3¦²À¦g±º"½nèwæ†OÊËrAQÍT“E¿ÌCžà€hj¾}$¶¾Öž×y×nÓ,|ºùV¸Ònæ0¡ú z‘RA™§â¸]³±ó# Íë›Ôû!ðCxJ8$¬Å>"¸×bÇ`$ TÿM—à¡à2lýk¾>Nl*†ÑÜS‚°óKxH]¥Úõu‚Æ#¶ ;íî7)1fèŸÛæf2 Òé/¡ïvèŸG»aí#¶Ðz4©¯!?ANMtßwØd#þ…m }’lBò•Šœäµ®Pøô#;‡c† Á.%æ‹ô­¥ GÀ #ÎNìÙ?±¹~@Oå!jYæ-X„îÅ[–]ŽÚÂ_.SO’@7¤3²R”%ï”ÍîÀ±Ê3Ûhiì&vB­ Qk!ŸQäxóvM>ì@_q”O ..ß× ö¸æ2@ðVR§6.-ƒòd^Ò&ô.žl]ÈØ- \²ÿÕ ©è+,2_áJNžoËlÇv ¼O­× ²!Þÿ½=Lªg‹à+½@âGa3j¬´>‘.¾Lq P˜å¿¼ÊA¶j«vßÏVÖ÷±†„à tøòF?t%OÏ•ø=Ç‹£cèÛ¼·’Íûq´›uOK,•6€ØàWd*©èç¼h2üKYvLašÝ›l*N¥ÝæËu™Yx‘çE9l©êÚRgôÉ)©ªÐð…&\ÑÏlÙk'7STc [>º@û'ž 8VEBÖh'ª6! /HœþõŒ·ÕEüÝÌo^èøM ¸±n¢KZdz^skoý} ­áÉŠ?DŸ^©) ÏW½ÇÇqàTïµ ~æÂÔQè¶K6‹±\¾$<!~„ç7dÕ<’:O(“AЪ|¡-ßL½ ¬´Fvt¯£oçäÝíé™äb6:Üî•âÏþ¡þMÌž)Å¢:»«§ìMµ¥î#ìdkõVþ·HéX %Í2bhvâIµ°ä/†„ Ç0.ÆX}Q¬‘#̺¿©+g7paÁe:`}J‡V…»'TXü¼é™þhZi¸ËÔÍ?ßHø›Ûõ—/aGU©—Ï|χÔÊÖ9ÙÁPªæþ3Nñ‘o=ØŽ#@‰®S=R¾µ*]Ñc *åi.Ò™z;\;PJª™kúŽhDeâË@ñžz¸0ß\>0ž iÀoލ»0ZeC»ô˜ÂÓÿJn´ìã_ ØþU™Jañ(Û¢ïx¬6Öý)ýma .”¢lr2% L›ßæÞNUÆÒxC8&jt´ŸŠ!^dÅÓîmøGÚÒ3í4Xr¯ú›_L Úiªýr<Ʀ­h…%Y²êÇGÑ^ãÕDÚçËñ‡¶$‚3©œ—>MØoQg8×#žØÞŠuù®é1žª‡!jÕ3·¿QtšwG G–Jx3"øáœ3=Ïþœ8—€x¶ÝÍ ›ðÜä8È.’L“Wø†¾ÉdqâAûú«°#\¾ò5^pï|K0Ío¼6v¶˜‚`)Ç2ÒŸðiùhø㟟¾KöŒÖD¶PêIB ΂Øá[Ksæó\NÇdF}hˆ]ÔíØ÷ÅWQ„°‡…3&RD¦¤0™›ï|ßSbŠ8uŸCV«;ÌÂaÊÞcû¥T5Hw¾õFë• :í£k!0:x©Wt¦Ô‰hb±_züˆ~]¡ëÉÄG?eg*Óè=Äq®¿ Â¹n­êâ*€díQ=‹ +î5¾,„£·£ó ±Œ!øôÄÓ.RÍ62SÄ"š9æ¶¶jr†ÄUœ¸H ª€™¬ü×éV*ý5Ýñ 7Šoé?,ž«'1*,)Y¦3>.Š¥‰‡? 9êÑÜÚyr™µ3 i¹D :ÙùʯÙ;Ê;×7øîF­ÄpQ/7RúHggÕ\¨¢òÕ6¬'DìÊ võSêBpÒMÀ(Kæöü&¶ËÝž¿V:ÄÕÏgžuaïWgˆšyä*¢(ÔKXÀÝÔØ‹ö°í·­ äKõ†áâ‹DwØd9š}çÐ9 *ö”m'`qÛEÿf¡mÆ€¥‰b ¬¨3Þ«RLˆ¤6•°˜¹¢Eoî‘àDÔüÝ~yúè3©½õ¾g%M'ˆ1F‰^?°)Iý©y(c{ª2fÀËfl¯M 171Y†¦f‘Ègʎ؇´C€‡9ñÉ[ãÂ+ï…/Gà@6ÔskqÔ^£gåý7œerÇ2ðYZ="sʵd”¥f4ÏJÿê—È@dÙÝ|Ý)™P¨<÷ÆÊňf"xPuŸk¦]ÉÓdçÒ«¸L_•FF7¥BÏ^íyå ËÚHv™VLUëÛj’Ï=8ú2~G¬¥òÏ6C%h>K¶F|­°v:”¸vF®î \ÝÅÔñÒ'ãÓx~”âºC½e ~Ÿ”…"«%Vhš°«rõžVùÊžKÈDHX'©D»›´—•AƒÂ»K”ô‰—tÕ_€ŒÜ¸˜9±ž}òû¸‡¯ñ7ÍqÁ Å ½zñ£Zr×vÐö(ÒØ€P'I1}™-†ïÕÌPªÍ¶t3Ê£ P5Ï6ÿø*³÷´8`!@Ñ bIú«' ¾§ó Ý—²ÑaJGé¢Ëð+cø÷½©zkpНrPF”®U­f|¯j›DêX´ÃIwLOst»p¶MWȯê´xºã!S©Co{z7ì“ 7Üâ»if• •ÎÅ.^VÙó'å™>I··'­‰7)Ä<l>é´/ ’Þç—CÙ$Ÿx‡ÐïžDâì¼SEêN¥'ìÍo çgÅÎÁRÒ6^qƒîw÷_ëR # ~K¥>-EÝk3ÓÕ…7bîww¹Ä‰ËÙèë‡q¹4@œm*G®Å–¤Ì;–µûȧŠÖ>Y8^¼\2Œ#z-s›+d.̧ξs/•¼¸À0E‹«QV)ù†xû)|ùî>™ƒö—Ù«†–¸åDvJX¤=òÛÀ†ÛBE­0À@v! ø»9(LÕ…-¡túKŽB/FûÀ›øÒb=eÛp ÔnÐA¨dªc`g«&˜CþO!È„4²R¾øêöŸ ®I5çœY¤fQmˆr{•àüÒ3uSêš¶`Zöd(SŽDÛK¸zùS¸øç †:ow.t’©iUª@ÚˆY{²<`0¹»Þ½Œâe¥Ic× O'‹ 1¡ñïªQÎÜ‘5‹:Óºàë7{Øœ—ZµÑ† À:M1^ÆÅf1p§ï“<æp¸®Õëk‰ú5ê~º+n8ûUˆƒ”l4õö룼þUåœࡨŒi8ÔáZ@Õ^>®/³E< Ô«eh»ɈYõüª.ô¢J¤Í[ÒI»«Ò˜ à@7ìY'ž¹Ö#Ì ˆ­„Ìym¥6øÙ´dËçS²ˆ3ð ÂQŒ{¹@×9ój|~̯ÂÏw}³Œ:u1Ü\˜oOò¯Yýtf5—!†Õ(_íî’žæäâ?A‹Eø_AÙØóÈîª1  ´?V½Ô#’Ÿ“J¶é„ñ;°ôæ’‚” õ~x•„vÃ_þ†Wˆ¥iI¶´‚-“ʼnßC–IªÃ²tªÆ“Ò‘ilš.\þpNѾ½LqSèÚÜz|ëég-Zxk¶Ã™øtÓ ÑýRŒeFWTã´°sû,£ÒX첈<_ùÙ÷¸¦ûøI¢U¬êiM|ãs;Æ žbMyuw lec•P7TÊp$2¹jwÁåh‰É(Ã^?úèÎ[{ÂBM¯¨qì'<Á“;‰Û˜•#í;ÑÉàì”Bÿu­Bƒ¨ 4“Ç6µpIàp¿†X«×LjÃFÂ04;:tÛ ÿF‡ì§@îàë Ÿ øÂòÃø®† br>ËâËH#EÎPôÖ Dð­QòC^¨7’FsÇ%'ÅV:Ç„¬¤$¯ð¹§d·rB4ÉÐQbÖU¹ØÅï•IJ‘Z•øe.n€Mõ^kÑ'3hU8Ýó‰« yò÷¸ps¨èkuËy ï1°Æ™ƆÈ)Á-G¥† ý "ÿ;YÑ£Û™0 2Œç9êàI¦ý‰Îh Þ™þÆ jpqÂ= hô–{KÀ ÃkC„ì}Ó½˜<镎“Ðý9žî[e²Šô0J]òÕÅF¬|L-þP:EįÆkˆ™Ñ™?R¦Dq GôÛsŽl|÷èHÉ…¢üD6>Fˆìõr˜äf¶`“nStŸì4'KcûxALŠQŸÔ”d¯äÇOr„‚mÚÕû¸/ïªÌCÿÊ”Œx‡ÀÇ샟ö`Ú‰gDr7ý©Ë2½*(³{n6¬?׊b¸Ôý;áî(¯4’”pe'Ñ&;ºŸYbmYÿÏÍpæ^yORÀÁUç Y}åË<ÄVÒ:€5Møb‘áüÝí"ZxIj°^+¤‰kÍh œ†p6De×zÂׂº`ÄH*<þ¦{oB ¤†RhÃ2’Ç|âÒTR쥟`:pô‹¸®c…áãw¾*µ ·†Ñmkù0Ù Œs"™X¡H§5®Wzèû ¿ŒïDÕhŸd f3¨Œç58XäðÕë÷wľ”߃ºÛÿ~I#N²ó;aµõ}•½U#>Tã–‚pE¡Wɳݺt¡ÊÞÕìÅ4@{BÍN!¥êc˜çQè3tqë¹í–·˜þk{D6r_^dès ß»‚pѼþª 34ŸÓ<²ZÅ?Ççù1gä#ÃbߪœÁ-/ÈÛÚ¼d¿,ë£Ãää)¿™¨B¶b‰–”[nŠ桾`ã$ýå!g&0Ée¤×o½;Á‰iûlDxlNîä ¡‘)|ýÄCf{l•¡EÒl£{8RÜÍy€²×Š”×ÓÛ—d18]7³ø«BQÂ6<©ù±>ó"1 /©ñú§H•y²X{§97P0І *´Üâ\ð"ÃÛ+aJÆ+§Öþ2¼À«s è7d¢v$¿»æÄ$i7lëìÃçY&Ï€•š/ÛìQñÔKÄÛ…S[hׯ¾€Y·—ȇ#ÔŒdßÁyÌg4å ݸ ÄèÌ”ÿmfÍäv–*r† 5ÃO'³ÿwý{lAæ—?Öùõ¥ —NjzõÍÊu÷d‘é€pòàËv]Xà]ˆ¸$lÐ5Ý?TYÊÙäq¤ ÌüÌ<+qn)ä~†& ç¼Á«¹Iÿ j1Þá%{/leæÄ鄨x"•“>¡`¸(ß­Š4Eh¦~î«ñƒˆ¾§¤=ð‹»RZhH¨ð‰¼MOݦünß/UhæÈžJƒUþ[Ë .ïPag^„÷uþP“Høí\(ƒö\$dÙs‚¥Ó)Wøh9ê_Êh¯u0Ž?wÿ{ÏÇEÉZª€`9è@¨+ì1@W}ûé·6# ý“>ê¸Ôƒ2˜¦ÅE,}¦éRø‹«¿u‰ œ?¸Bø_ÓÝ@e68¦oOzí¨k‰µ$£Â­aa†ø\Ý13·uištÐFâÙ|:TïY£È ¹2Óø‰$ñ—dÍ¢Ãm^ŒÑú­¦Ì»¾|¯mh,µbïëϬîwú-+t¯£@ÂVÐí ȳ@›ó¨‘Ç ²ð˜!}Srµgf¯ %fsðC•ˆl¶pºøIÐK7†C±m©{¥l&äa¶EEoWqúÍÆã]=#âÕèÕÄŠ.'±€Í×¥{ýš>±"ub/c„Õf³b‹":¤­i41úÛW‹Ã!*tø‡ÿSÄw¹JÐgŸëWô©ròÿ¾8[K´^d§‰,r”Q‘ŸèÞC[ üRsâ<×úíãü~ÇÿºÁaÆ~?»c¤ùýjhüÝR×îG¿'pë@¸ú†ÈÚ§¶5¯7ê/å ñ½ü,»ˆIâ{#6$鋱o*(ÃO8f¢M™0Æd×Á²É€y¾$ºœœ ¯rÏɃ±ÛD¥hǬŠU0HLŦpÚì ?+“Ñ’Â9;˜ß‚óYIÁïÞw0oæ ÷ÿ(âß®/"âH+¸i‚ž-)7íS>(Nñ’>\Ÿª; iʹAmaبnÄ/šÛ&_ ×,¢‹°âlÛ Ïí€'ëî„XÔíóÕ„Ã`·òŽ ü8E؈*M‰Ôû9àÛàV;ÈkɈŠß{qá•…]«‹{iq"–<²ÜÊ?«qĉÚ|Êb6Õ‹Þûª7ç[ƒˆsìÕÍý’ò¹TÁ›e#±ó$H'&Ï|N¢Â¢® =óæË4f‰B…ÎìqaJóO¢Ó}Ó°³Œ¹_Ö±I3ˆdSóxÌîéUb¬)_~«aWr)lîwL±ùÀ"®ÅšûÛšÐ{b¶N9‘ ËDƒ×à˜‡«óâÏO¹šk"áGèï”—av>.e‰ÓG¤³ßÈô2¥ïW"Àà¶ —@ ½[»àDþÙʈû]æàof˜Z²ƒ'y» ËÇÈÎ$æ¡gŸ¥˜ÒV~]‘‘yùRÞqï¢ 56lùýˆµg^€RæÌÆ8ûuøG-”qæ8£íK||•Š··¢‚z½ÆñÌ1l^¨]•E™¹cgžÇsć¹F4†1æ„Í}AtO¾_‰;X¤±ÕÀ–îÈ<_™ÔrÌ‚ÜW)7 ]Ïú8u•'U±Ë¡OB‰êºP»¦ò¦.]øÀMW©«l–Øþ ¡='kƒÛdü¾8Rñ -ùP5¶ •­(*¸×ö6U>õé¹j5“Ð,3÷3Î#ó\n fr?|¿hgŠõt¾«|³!‹lç}q³Ò‘Ñl+ô/Ã0öõ¼CU[È5®E–¯5Û/ŸÁhP[ÕCp2mX‘HîÒ"ºì[¬*ÛL§¨–ýˆÅ°šœ»p1¯þ¤­a„k$îå(¹Aq—KDr·H[FÈü²}ŒtŽC%£“•ê0Jl4VÜH݈Ý'eLq ¯:*Ôæ÷ØÅ®Á¬x éšøÓÙ=_ÐÆ€,@º^î/¡%á`Š˜o+ø3cÍÈÆ½m×¼`F û^L†œ“3Ëào……Í"ŠT˜¼Ë®,ÔîJÝç9óæœnNJ¢(e¨>Û+|»ý„=)èvšRízk\çùÛè´ÜOjSÉ×Ý%8Ðnÿ²JFž2zBm·åO:ÁÛ—œ¾W×öz¯té äë³òžæZMÈNeÝþu¹œGá8£;—9{Ru_§ÜÊù#œ¯ž_èe¨ÃId±Y=ƒ×gµÒ™?¤Çà× ?lıÍykûÎ`HÓÆ'7v"U~|7†ÌߣmU´ÝÌuñçtò=ïÊ*{G‚Ùj¯¨^ÚÕEeCdž’°‚Ž6'H¶PÁe籈¦€¦Ös°­œe ¶½©Hf¶µ¢)Îþî|~ªöѹM‡ö5eÈPꔨb+ÁàH¶Ó¶7.S—|eCÄtFœIø­1‹ÚBÔ{z€×§e‘7ì€6~p‘‹J /gH«Ëh8Wå7’Eÿéù¿¯`)Àã ¶øHî]mË%‘àr홬éI–}ðì©ôyÇμ»Åñ-o̼‘%àl¤¸2È£*Ų¤DÍ NÒŽ½øÙ^VËÈL+­Mf<'àãd…[¶hÔ(€†%BÓ¶–hŽ–O%Ö_ì>k@ec½†èoö´M“ÀpPÅ=ž„¢›gÐ_œ2§ô×RY\á¢ZzŽöã`„`a0›A‰ôò#¦µª¬%祹öÃR"]0ÜÕɽ’í‘0UeCžd$ÊdÃæ¬g·uæß‘`½a ÌîKÕG9¾ëWB GjÚÔ.sÞ›•zˆƒÖœ&[פ´ëE0q³š#7YrßBéž¶îsp¡3˜chî ùEaYîUo|Ö‰r®“J&ßÔ@jû-;6©¨l‰×|0«CÛ»KT‘”´9ZŽÑGAŹˆš:º”³’ãüKË¡z‡Û”v+7 öTü¯|0މ´…A¨%¬GÝ)¡ÅCžlÈ%ödnÛ½ÀƒŒÙ?#,ï•׿ã„*Båróä¿V‹PAª?¯ÊS f1÷¾Oè-?k—fdë ±†·ëÚyÝrMN—æLiAÿÞAµ>w³Þ•LÎb¡ f=©…©d§öÇcÀ¨ëÖ3^ZÉ¿W–J¦ö’«¿/ÂÐ&Þ˜d'x"ñ0îÒEyó²BÇhÆ2yн]±øAÈ©·. ¡æömÒMßHB‹·Û¢õ¦þVìÌۨǸ3˵¤„´¦‘î>—HLj·Ô>Í8óY™¦ ’ÜóÛì«~tá~Oê²\Ùîþ‹)aþùgO”Û^ië›Îªpüèž{vø2åG3_Œ÷Ñ™›ËíФ‹iG>Eâ l~éaÍ–âÒ³ÃýÔL¾½ Þ|fíŒÊ@…«…vÆ™¯õ±†Š—„Epn³}À­DƒØÜZà}…X;\i<óVcøéksP«c®%ûeÿk2H/àGàà·_tÛò5Þ^Txý®“nüÖŸ¬@ñ°±ÈÈe0P =·ˆtäHÅ_y†àuvâPýÙü‰"~äx›€=’[ åMOO‡o›JOÆG^¹TV>1¨ü® vÅÛMöœšÙSùXYŽðøPPÉ?iKã•DZÅËÉ®þùmÚ¡å"Kj-î1ß ð8óò‘Æt÷È\ɨ‘ý‡¢ðXa NŽ,ó nÓOx;»g*üÕ’ [(±¨…(ëù-æH Œ¯ŽMz#/á/CÃH5r"…ëŒ=ýø ôÞÅ¿]yîGþúò«°ÒF¬;âœJŽ^neSöF%ëãŸÑ58gÅaÌ^‰ïºžWéZŒº§ðãƒëÜ›£ yr}Œç³ÖÔ-r¢˜Už¹]DíÏB&‹HaÕy(ÆôIÃŒêgI: —F"ùHPRFuGš\Ïò½|ÿ4¬O¸{jú¥'@qÞÓh NÈôzŽ’:îZø– n‰]E®4íÍH¼1`‘z*Þð¹QYäÇQd¾³Lx“]a,‰÷ʼn+S"h.£q˜.<ÏŠPu±ÉS›Oö¤u:•œU |10º2ȼ†mP Šûcöø©x#Š@¶mô}k¤Eÿ$›7øh¬’f0Xnا!á¾[sÌ]uÎJý¦ÓMØ¥Îv«:ÈoÀ¸{eçJ±:‰SN熉œŠk±hÕ3­¨g[°—ævêç,C×hëZmÉ;{ G\Í\Jø©¼7Ó6åìl¢ÌÜÒùgÖk!G6|j’Ûòó§Ê@À’- Zž•ÙJ{]{íÏxX߫Õ7¥zÚaì8夯¬G×Á‹3©¼8™bÂf*ââL¡>üRøt›çPŠ`×_+?+íó¨#ô²ö©ÂJ©WÛO*X˜¾÷„‹'C„lï†í!¶m=AÃ{„QŒü0ÃMAë™?J<õ9k©¦×óFÎ1k³k ­‚âõp˜Å¿†6þåãÉ{üÕbmÔË[Ü›ÍFB¾ Ö9Š„pˆEy(Ç”ä‰/o¤E˜¡Ð¿…¢úë*fðOüP‡Èàh‚@ïýžŠn£¥e:Á«‘lÐ|ccÑ0gz÷à°/Vÿ±"$ž¹nQ^ñ7ç°ˆªWáëÊAHÊ£˜ÂYÑyÐ]ý´¨ëj4[`k&È ›Í¸åL³áüÛÍÓ€hdõÔ» ƒáýs.œ±òÒC;;¶×LŠk8~ÔrÉùЪÓB?ìój³'ÈßÞ‹X%Ê ›- ‹ÚŠE¥ÈÔ[ßnô ×Õ*{AˆÙ˜¡ÿ¤ÏpkŸ‡u~uõ`dµâ¤mZ~Âæ27úæ¼NbÚÀ~h¼[¸“®4Š'õVÐ+„IÞ.ò.q¸dѽf9—  ‰T`ض¦®hv88gt\-"ŒÁûÝ ‘NHýÖã÷OGk¦ÿOd$›Ûb2†}›+Éíå8UdÃó¹8^‰2& áJß ÎúªÄ"Ãðú°‚´5».¿b ]þiæéO7·‡ !:Ϧ[”z{#|]¡ch>g‹û1¹®÷êpOø^—¦åØkéÊÏ`šiL{õ¹èz|ÉóG¡¿¾ä°Rüä vÅdk2ËÜ›l¼R{Y£aŽ_ä•‹c6BׂËN^nß’^âLów²[½¿¤l¹W»³e¦ÀèQ_©Rw8B3>ù'ΰ©(®ø"#ÕÇ? Ì3Øü®4_ rúóœõÄJ Áp¢ T¥„ 5™ FßzÞ»£?|¾6jHwŠI†ªBBƬèÒ2—)7Ýä{³5DoeE—VI¢óÀ‡ ·9;Zà@´G ¢Öî±Ch$Ó|w#4–uáE‰‰N4+ÝÕi¿ð÷Â^!ʶ…Þn¼Ü‚vô1¢ÍQv'þ›Ò{Ïí* ]âèR¤)y±›ä5‚Øçù+çÓvíÛŽWî{E3?—¶•4æcÃ×·G*–ÆêEuÁ,¹|<¯"Ð\ü! ö–9çón&úÔS3ŽWß"ÙnG"©îPì\Çñ^ä:ñÁd´ËökuÅóè ð–ùõž5þtì¡è p³J¾–ÊÐÍÓSy•!Sü.{„ãô7[ÁóK3íxFœ\Újép~%Õ05ÂV„t‘ûÉxfWu,ZtáÃÊT@ÕÕ“î*±ÙQ΂¦ó>i`ó¤@X¨1Ͳ³HÎÝ1‹(qFÛÐχúƒ> G”˜ Q>¿¤šIg?ÁIàÎ!sz‘Ø]…¯UÔÑÜ8ìÒi'Ig¦“a…•¼ê-ÂnBB9–WE•A•@°ôð¢«ÒY@iÚ?hB³I5­­¯AŵmJzª‹r¯/ «º+t•miüu;»7žA6уz²{SW†Mñ*æÇ­¹‚&ÃÐŒ®#ëü2MõÕŠ‘®w E±¾Wƒ#hÀ¢‹jóýÄæù-Æ$¥™qéáIÁb]®EJ£¼×€Geædg¶ÒõÚžü‹Ö€èìnž!xhTƒ‘Èå¾?çr¼ûºÔÏ+ DÒAÉ6nù![õ%I×d•ù&ö“É*&š®ä£Ù•F©¸¾#ÏÀØøE|¿bzÁ.ÕvÜ d Dé<:[ѰäeQcô£P ÞÐN ^´|¤‚Ù˜Ö×ÉÍž.ïïwjýÕ±šÒê¿•”“ÌѹXY+ßôè(ébSÉ'Ùã XL1Î1Ûeg…kºÉ8… Ææ ‡xm‚T&©_™|ª ‰‰:©—Ö%¸§Þ¿V°^iõK9àˆÍ-,xÄÜ¿gH [‡ìgYÇ=šÞÕç+拌PæÍvySý׿ùg“¬…uøQmÏj»˜£¹Ò©ä)°çßâõñ§®žŸŽÝÁîMP–É™%“&&íîh‹÷òÒòÁ˜'>¿/!¯MEõHÓ)Öi‡¾Þ™` Å-pWq×eª¶a§OWÊ*ôì_ö„G° œãÞ-ßag è•&„S#‘±eó2¯ò›0Ê0õdmAòV“¬ˆ—«žÖËfG„‚…s\ ¢’@œD§ÑÈÌ ¸hóƒ í2*gzä׉ž+}TÒ÷rœtD2o?Âr´wŠq•–ެ޼ôÜf8m–£,â2ŠÌ¡¾lsúãœJëæjÂRÕÖ±QN1ÝÞbñ%¯ªóss¦8ºNÔGiϾk(Ÿ1ñêéhæ^z¾{ôðO cÃs%~"Ùt/ž¦ßA\‰\kîò kf…›@ŠÔIÙprÖ'´òjsç0V'öµ¥± ŠÞvsÉ7p;À¡Æ£ÑÒá>…•Ýÿˆ˜ur)zŠÃÐZÇÏ/ïÑ&„¨ŒO¯ñ¼©_èÑ5Ç ƒY­9çP!<Û¨j©[ dy§ÎMÄkD’‘;lmÖm7N†sS±ŸâOn/ƒ¯kR/´ßŒïp]ÆE(Dp/ìÇÓ±9ÓpÖ Í@Öc Í2ü¥ÀqÜŒÝ ˜ü'éd%DÄäëk/Æ'ëÅ×:v²8ŸvàÓæXèø23î]X 7°b]*ì9¡=¼ƒŽä /ðÊ¿†Uu‹I¶Î÷ðZF1wîÀ~A.}Ë.ƒ$ L ‘´Éö¶g—4‰_utÝ <ÍwëeÛä’‰ .$x®0ÇÆæ-U¯P%¿†íïfÅ‚ºÏ´Œ¯t!þ%g[ƒòú"8pOÕ¶Ùä0§™:I<û˜Ѧ!`yp„ ½Qm²O•°™…¾¥¢ zÛaØqCXm>7v?ÛÙ}ÉZrlþHc«Ü.ã0¿|Ù&B$ÃQöBM&³±åéí]"W¤r ð.Øs?Y*úLj}^˜f»`ߨY[ÒßÚ»¥–å°å@l 6ø‚M2 wœ”RNœ{= ÝêD©¬›†âãNû¼©):yùV&Z¬F–/,›3hÞÏ ÔÑd´'(‘Ö(åølÍuš×÷u6ø›’YL…€:’¹ sXÊúê%“‘Âuµ§¸)àENš¨ÏW\ôü›\Žm0H Ó\8À?HñédNæÐ¢öÙß]‰‚O{Âúq|TÁe—'V<$¨ï)¼ÛÞ(-ÜY°ó…xUè* ,5ö¬üõrS–bl–õÚ`GÛ5ì"e63A5—Ž ™Õw cØ5¤÷‹‚³ýNïÐmš9' ‡ìŒèôºèÓ‰@:ò™ð¦9[K꧇jÑùÄžJ|Tâ{äœòÃè.^aÔeé<šu¥0¾+$ØOŠbzŠz¹¹«$éï4â@HŽ…ÙŽÔø“C!7. îâLRUmnÌÀ§±ù%5¯ t‡yƒÁ­È¿ô1uP¢ýìç¶0•¯šÙQä>û7æè:é!bŽÐnÍèÏŠù¤éSéûÞ^ yøóÈvý}µÔ%oEÑ$.ÓÊ+[>,éÞ=ÐÏžs稚U÷ÿ— ¦X¬Œ±Ÿ—&Bú­9Š[:ÿÕÌ;²’ bG *xmè=Yn ygÙ"YÍÚò‹ =ᤴ†eE2{#lvMC *”9·‰I¼ÇÓ¼Û"êð‡›?À'an¡í[§ƒÎ;õ–ˆq÷Øõà¦.,'Y7&sziá^óuvYn_Þ+‡ Óž†-¦ÌéK#eÏš6®°L\¢>;—<ûnƒeküºR"ØžÏe<[íŒåÜŒíoY­,«yЄ岒¯+ªA ¯wâ‹)AÕm<@¹h%ŠH£³LÙ¨ãVRÃ&¼÷yNáºCdNbŽN§à“\ÎvƤ.L¦D–Ot^Wz”oË„Î;R°¯å4Äb¿s]PíÐ2WÎ×ÛÜŒ¼§=?¾ä´5àôŸFŒ†z%%@j’?º ·IŽe7‡@®þ¤.›KbIÒH~‰m4NçËì×dY £Ð~æÔlÂ7¥S Ô¹f(É[ûéOFÆww¦Ž(~­É•å š¾ÑÒD±á}ˆ²¨9- ÔÁý¿¸“†ÙŒ/^YAv@Ÿ{Úýä9ƒ|2m¾Š-?¤–g£—_ù…O¬ù2ȶ£ÝcŽ.K°(Cà}itͦôŒê¨Ë¾¦Èû¥¹( P%ÈÊõÔÀ»Lâ5‚º¸AšÈ‹¡àoUˆP?y«òwï¨d%—áQ¥¹¼·c‡®|€–ƨ¡ó¬ C^–wD…ýZa§á* èi®ø*òJj*²²&Ÿ(•¸ÉgPü²}?ݤÉ=ëãóóµ=j[‘ßÄt5’êÔÜÿ0·ExÔ{›ôTRe –óu)ôæPô€Â;TFðFDѦg)¬_ÝxL;Pá­“ô!ŸY–>)Øì²<÷iÅÉ<‚1rp7<‚Ã0`ÌäQ™}‰*ÝZDeÎr“1*îÁtE%|ÁZ-5¾©¿†0Äo‰ô“E9€ÊjÈŸ=†\}¨7í ~¸»¡[L¯›¶çYê#gÖ µþùžN]»\k¹vÜD†”WéýóÆ@ò¸ i.NúÝ|)×#UË_«ÕžU=œy-•:O‚^_ïΠ\ûÇIA29F¦ÛšÅÐImØ\ñ‡òIzÈœæÆïpµ ~¬6F;·Ó5‡¿DJ™o²~ÓšÍ2yÅ+¸sh;‘Sâ¾Mô­¯¬°•¦ýµÆuð²F^;ÍÝ‹ ÏA{­ý–ÈÅ¿ǯ櫫~êG°ŠN´’¼ÆÇ™¢#÷Uö¦xÛ9³ÓÞëÆKž“´•¢©ô'»¡Âá$£©ÕˆÏw¬@š÷Öò_ûä4Ý0³áó˜kXeU@yÙñ¿Öã¶Ï‰;·K1•(V^†ØÕŽ%1E*¨ë'd†wàmæDi˜Ÿhy;c¢õjØàÑðM{oKBØvh×7[@<{7£rPkÙСázIÅágÿÕ„møÕÕ/F§ÄZyÒƒ·|£>H¯ŸÒ)3Œ=m‹5f‹ÝϪþâÔJ)_j58¿¿Ó‘¸ì LüxÄ"©#9¼YVTSnHzÂ/[ö#=«D§z<)6·GN¶"÷Ïù¢™Äió.jyï³c¨—¶Õu ûY!Þ!â«ÿ§D„QÅξln ØÑQ.z»ÓÖpâÀÃÐóyvŠ_;5|Ü*J£—gÙ/tJ8u3OZDáúþAÖt%Û‡¬‹ ¹ ²ÿÃþÛV±@øô¡ÏóI[Ùøü(ÉœåΰV£õSŒd`΢|òÜ·»ñGÅçeyhÁ~4²iЦÍTÄlÇ ÷ûœqí¿ÌÙ ?8¶óÕ¨tèÁ®k¼¾Î÷~¹éÚ¢I>O%ê!‚Òt*E¯*9\®Ñ¼å¶ƒ9Àæ]?ÕM¥‘%šã…Íuo:¡à,Ô„Æù; ˆ÷š¯äω€ŸÂ¡ Ö È“ÈÑšaLˆ©êì²<ÁöÝ#®cÄdˆv âuÖL½M°RD²Å_v+©$,=0#ŽšèÜ-¬2Í™à³\öÕz9wàªé†M–ãs~’F€/ÜS€c/¦™e¨½vÇÕòAhÚÖC¤\gC÷H_øbñ¤®‚-¢{wX¢w¤_ºˆŠ®Y2˨P,òvtõŸ¹¤Lë“íãÕ$Þ`d‚KëØzˆÏÙDÅ1WE¡ÍÉ·„˜W{¢j×\TÔ²ê2V®“¼TõUÜ.Æÿ¹ËÕQõÍ—•‘ð3[ã·ØðwQpòÉôp ÝdqGƒÏÙŸÈ~×hòhNÕaÌJ*ÔÎSÁ7oÔCú: E¨Ûcµ³MúKœÿ$ba„2:‚X/ÁšmTgë6Y}òè~™òœŒkYyøü,uA o/”]™U0¨â0\Ì#ayàåÉH$øOÅ àü7—+U²Ó,š!Ö×L(˜/Û}MTø¤?§Z TÉ3N5…a‚ÅŽ!0úg{l¬{½e—c6¥y'5C ®»izyè™VÜÓ'®i»sÏ=*¡Šžñðܦeøèq®\›6‚ɨã>¾‰Ù³}«Dz£`1|¥â¨Îäír[šUI­f~Ák)¤¦èÚÄæ\r¤*€› øw/š`ÂY×O¸¶P/ʃ¨]TMQ ÙÌ¢V°¤Ù£Imr ºÿ«r(j~å4fF7ÂÞÒ‡(ªŒ1†/®ÝUƒO+†[D6-Px¬ø¥¾ß%Ö[)6¬ ´ü¦ 4»ºì}æ{6¾c3áñ³÷?™é­U^¥£Aê’&Jy[ÎsTu<|’«‹ÁxœÌr³îNU7±¿ˆ¢˜M¡;E®;ìž‘û؇ñÂ?œ°ŒQø¨GÞ{Çd†õ>€ël3ó§qÊZè²}zE$Û–6!S‡6Q|´/‰YïšÚ6˜rcÚžO7ô<ÇXûš–9IÿøÄ7Ú©m-o1­ œÚZŠ=y/ä–£ .ÞJOp‹„;UCäehV@p]©WéÚ4hJ‰x¡LÏwæÌlîâû`§oþ•²´W¤<ýYÔ}-õƒÝùnáåÔsÀqø'ÊO¼øô‰sSŒX8é †z!væéî^½L†9 ÅÓÒmS¼EsW¨q൥ˆ(=¯›è¬ò±Zp´ ;ÆúIòb¼tÊ•0Døh5,tðyj0ŠbÊõ×ÙC·xaÌhÙ…ôt;S7¾ìÓÍ¥ë¢ kÚ±çHl -{¿¢Ø•è·í!ÈÀ„ótÌsª3Î_W—Æd¯ÓV#ª!ƒŠ ºš:Pí ½3xº&xð ´'òM£qâçœBñ·ˆmÔ€­ƒú9̵ê¢GC£¬µy6²éIcUXØ6˜´Í'šÅˆ1YøñåEê·¡‰Ø²•ÿýg,÷Gö¹¹cPm"­Øšƒ[‹Ô]TI ZÐtVl±=Yçæ ¸Á+Ÿ|æ4’à2¤e8îöÃjˆ!R“°±óp2ÒSPá¼—ôNp8‘ê^*MÆ7)% F¡Ú×SÎB¥ 4ÍÛ㔾¿|ubUáùê•@Ì—ÌLCÆË^ľ 9Þ!´ÑvvÌ `•Øý˜} †îõ±?@ ËÈŒ†áû«ïår aÝcN×EÖ›¾ÞYQþƒ”•Êv›Yì^8ÝJ\”;°ÇwCŠS>‘{k YŒRj]©ƒùœÃ£j¬çF=$ß[«±§>Ú« ¾ a.*¿Éç¨kçy õQ÷ \ÿ›³ßÇù Õ8 ›î—º`àu³#‹¡c«å7´Õ¬[äÙeÚì¡õD5þ¢ëœì:àç@Dí™ÜùÒe1 £v’qÀ$s1c®HÜÈ«âO¾ÿ&.°jg)·:PGnŸ …EÓÉ1ÖöÐSV4çud¦¸‹ ÷ð+‡“LM 5ktŶX{à™Ê°už ß_gJI¸»øŸ£ý9µ8S6;òýÞY ät‚g^aø„£žÇ6­øîºÈ×!ƒ ¿©(ÿð!²rñ²pMë.† ˆOû-pq´\b¦ÂÿÑyµÿÑÁÄÍä™Äs³M\*b40†;ý’VúQ-“J"ò.JB3öè@(-’x’T1*Ë&wyÎrr^uHW¢oà ýàÑÚzšò«G”ctÁ±¦„xqbåÒš´¨7Âܨ“!·¡mëâ¶gx=Û*:Þ!Ðéºr´1ÿ÷<.ÛO‹Y¶ÍàãñÕvnÒ$§íßt_$sv¥•aÇþY¸ Ç(U±²“H±Ñ6¿ŽËO† ˆ¬0aùÍ@..>Nq¼³ K©.²ˆ¶Ûa‘l·úÇØ:þ¦ôé{Ë „ã9uºÑ4J9‹yw©zCH2u7#Êõ!£»,h-áRàfi+¥(§W¨<:w£—û¦fdrÛ$ò䥛Hf2j3ì u<UتÚm8¾œä&»ÝÙ w ]êjpësi«ÆÿƒyàTB߀0/óIøn•ÙhˆzÕ’VÍÇG¿‚-‰• ¹™&.y„2âEÙb\œšéÜÎú¹GâEé”%Må A5 Jëµ3‡:“5û½…h÷`ZËS^!B×døw¦+hÜÕ®¼òü/D£R"¿ž#!š)qœÊø~OX¯hF™TŒRElúŒmZáµð\›Õ¿´w>HŒ•7U¹wÄÐS:–X™åIýZžÞüÄ4°­£É~Ì ýGÓóÇšè&¶&UøÆK†…Í8y»«6úìÔC2ÆC¢êf¿«¤T7Iˆ ؇æ\SóEþ/å[0ÖY‡ia€gŠ3àH´üTD0×7¦»ûIÛ¢Yvpâ]vÙ^ìm ofsNé–ì—ÿ›0fØŠuÈy`.÷2¢%³‘F’ÕBÜøÔÎÖçJ ËHu™TEã4z0Þ„þŸφüìÚŽüwwkL°>6¾:Ðv«©gk9 —Îùœ«¥³žFw¦£í™d=î8”G5ý>é‡u5,tE*Ø´«¦ÍÍ븻ޙõ ï ]Ò/“EÖ‚ÙѺ‰,ÿ0.FyW©Œq‰¨wØf9SæÈ®•Æ ÈË3Ocd€`­¾˜QÏ‚šÜB›1•`³I– Y\¦8–Pºf|Hq¯Žñ¥ï7*œ3ÆŠŒÐ¦9ó” ý9º/¥bÿæ#Ò)†¸ ó¨­žJÇåÉ”ŠMgAð;ôÃšë¯ šãJbIAþ°%¸òS¯Eq;pîæ Õÿ)kFæSžzÄE€üõþ4¨ãЩ½@5šUÅâI$ü„ëíéÞDƒèuºÑø&ä†Nß.r%l žç,Z简 }Yáç‹mv­~β¾MÅ'”Âkþy-x5V¼«€Ð¡zeJˆ¥@Çy“"ðOxíC‘|ÙîŽË¨Å™sÿnd¬/ÉÓn""XnüÝ 7à–“oø` €õ’ öatvX}é`xе ùCOù]á÷jþ3k¬¡’NLjj‘%¶}»OÉhÃñ&]jm8 +¸0-;Dl¡ñ!:-¦P¸šÅ(ä_”ð§Z‰?öÐà[9F@¸,‘.%\å=³¸lkMY6…Ã7tä€ÔµÂ 5Ù£aÅîc„µø„nô|ø´ÒY šLIæÇð¼]˜ºÎ¤cyŸð4p&–Ië—΄(ÚQ÷9y˜m™CkìÆ jÑ­Âí}üVJfà«dH„†pÇ` yÐg¦ZÔ}B–Š{Ç’5v¸À'W7ï w qŸÊülC£”>ù»…ŪdÑ—©êzŠ¡ÒÚY¥NáQÃoVºgmº%c/qz1/ÓéEXïi<>óbÉø‘C¬˜%yûY -¾‘Ú%x6—aÊÛçn¹¢%X­·›J.VÙÊu[rÅd!T?Ý›ën'ç}*Åã®>å½É›¾G»ì"7u5EÝýÚïèÛ¶?DéÈâ+. Šˆ©ÐÌ'ùPGiV<cXC§Š×kânE© g)쫾åIöLë³½ze¨jËß)Vj³Ð¿ŸZö#Ø,í™å—®œÆÊ.o"Ë•5v*–MmÇÓu£.µwNÂùŒÄÑ'ÂGÆ<ÕñîºÃÚ¸-;:[’C¦Šéqu@~Çú E´Cý/§‹ÒF1=Ôñ`»ö'ºK„Ðæ ¥IšMFì°Á(w»Ãð„lö¿Xhî7ö©aRX$0e ¨#Æ-žˆã}DBOX†Ì2¹Ê¾?c±¨B ©|-s» ¹¾"?ÎCËŠ¸°®)l¤üV HˆNd Š!ò=iü ÕÙ_¸š¼—­`!Yx4ªzøE¶DÖ§øò·ÃÖEIùAidJäÓú¾qŸ+í­ãøN_‚ñ‚»î6e»9U:éÄd›,È›¢xô4•–EZ\ c e;×(Y®Žq!çNwjÙ+’˜#Þ7®Û³à2Ÿ²[Åüj¿43²g™|ÚïÈ¥ŸâŸþ˜X­ Žò{1Áwž n ¹øÃ3•IÞâ7úÀÏÝùº çLRëV†šÆ¨^{«¤é.NªÓë'ÄÍ^Íà!GVXÏ×Êœ¿<áÆ#ÇÕ®a2ÐV¯Ãå0SóQ­BDðOO5×9¼¼£ÈXa!^›htªn™äd]_ìeôŽ˜.f»øb©üb‰úäÇ(Ô­Ó9!ãl(º˜Šå¨Õ&ãÃPæé0z2í}¹7|®Ø ÞÈ?jÊMšÝƒ!»OyãÑUX+7•7ÏüÛò³`ìùJTn§òö|âêc¿õˆŠçÖ¹®–… Þ„|EOãŠDoù­Ë÷®æHžùâ™ò’Q/ƒ¹ˆcSô!Àt™ªâØÐ Kˆ&JßÅ}»x\ <#~–ÿÄLEzƒâ‡ôˆŽ7¡~öö‘„~ÿæÚ!Ÿ:@ _œ Xîƒr¤éñƒþ‰ ú²v9ÿv ÇZXß§£Bq Â.×2Ø58·¤üâÓ Ýc/-!÷€úPµbhÆ“áì€ ®ˆÅ‚½Ëúß4jðä?¡|V0!Fút®ÙªiM?ĶDg—„#ªá¡j8ЬH ? \pßÚOP˜º¯š¦s^NcTYŒ—ËyŒíUHä –”jñ>ö®ºšÀ_O°}¼ç8k¯ôlåõš&½éòÍÉýf>”Ýcb îš6–¿98â±È#6]ˆ¦ýÄh€S©IÆ”[”•-oâ)IGãIègsˆ*J%³æ6GcFèbË+ 'Vã`rÜ=M(/@´rv%3æÅHE–ÛsòW-¶||¦;ß&”¾ñÅ,¾ ùûÎ%GWœ…Cîøbu%= ŸkÅÈ ‘—J7·ˆ7\Ë †ì6Àœ‚›ñœóŒ6,UÃ]HåHçZ&ir½vg‡g‘0õå0†Ën÷Yq küédrdð& ÖQdt©ÏEH‡©)HÌC¼fH÷8®îczÙºÓ©£ˆ@"­§2ømCr·ãfmþãZoVóQØQ «ÔµÔ8à°B O¥QŽŽ/ìúÔ+Æ¥R>Ó ]sÅnÐðf¼T]ŒYÏE|c{‰†Á~¨ý»±!2UÄF] \B² ©ô+¾0äö³ŸùÐ×¢\c¢•]vª•u“[+=Òê7AîC;CRˆ^©“潂A8­$3ÌŒný9ÅX¸×XTí°Í'ÿ™X"ͨZGžÎÎnÒ£n8 IñÊ1Š`›ª 7ùõãu'âà—i_¼,Ýæ»uÙHŽé_ñÝ>Î>å@J¿S"ð5–©š¬ÕL6%I0âŠ÷/9R}¢ò\%Ø*2h˜5µƒj}W rÅîò’؈Tp!^ÆÎ7:`àî{±•ùŽÝeÞd½ö¨³{QjŽ\*ïûª »4Û=6C:S}.uf‹dEûpK_Ä×8¿•ú¿,ïáb64ûèɯg™B’¿ï¦–{‡!ô›'Û¸š¯¹’……“ˆØÚaNñ†7cíÈPû¾+›šâë&»bódÁŽT2]Ÿén[Ìf%”l¢ ‘ã—vw|¢iŒ„«OQK½Êò`ñOËÓŸLü„hH½˜`(29È,y+<”Dõ?«ïËÍ‘¸ßoAó3,ºÏ£:(“3C‰K“Aïà endstream endobj 118 0 obj << /Length1 721 /Length2 12485 /Length3 0 /Length 13075 /Filter /FlateDecode >> stream xÚmvPœÍ¶-î.Áap× ÁÝ‚; ΃ 6ƒ»»C€w·ÜÝ]ƒkp—@€—óŸ{έ÷êUW}µeíêµWïþªé©åÁmO' 7;7×€–œ¦ 7€‹ã==½´ Ð r˘A€z@K€Ð ÀÍø áF£H;:yº€¬m & æ‚]{3KÈ ëhïè²°ˆ¸¹¹I¸¹B9\ b‹´€@İÙÒjꊪò&yU€< t1³¨CÍíA€ ØÈ °rtØÿÛX8‚-AÿâäÊñ°Ðò—˜•‹£@EV[RNMU +Í©- 0[>*ü-C\?üE-\þÕò_ËÞì¿–ù­ÿf]<þmq qs,A€9ÐFãü—nŠ`+Gÿ¿Ã–P§ÿ¤þrýËÀôWCf€%Ðê/jo¯jæ0I;:8A!@€Š£%Ð Ðtt0ÿeæ²÷üÿâþ¡1û«ƒ$Øú¯r\ÿ\å@@Kuä¯ÔVfö%ú'®ü÷¹¨-AP‡×ÿêñ÷D-Áöžÿ³åß6þÙSFQAUMõ¿ðOZláh [´ µ4s±üoàŸ´ºèßÓó2ÿ™&÷ÿú*fàǿƃë_ë?–Ñÿ¢¤¤=¼Ù¹ùìïy¸Ü¼?B¼<¾ÿ] ¨‹  ùGð¿½üÇÿg€@ ÚÊ¢£…pˆí—ae~²Ó刬ӓ¦:¬‹Ü°KŸpAý6–&{g¤×Ú¤¹îSæ¢[§›Š×4fñ@Œƒ±ñê£EÒÇ"Úrk¿êõ'¼ßuÆçº$ª‘WŒÕ• u–qÞ¯ÙÇÕJÎ%¢¸òÜ1¦¥­=ˆ8 Ës·`yf#F¼LÃW àkδMhþ¹¹øãö \xnðÆ=‰‰M~Y Ó^¸Á S³Ðæ| û¥âµC«ûà‡†(Ó§ýˆ„á*[²klj1}|wÕ «z*SgvÓö§`½Qvrn|¶íìã­ŠñR.ò Z4ƒÑþë—2AFªüÑh÷%Ô/ˆ¦œ'’(&V¦r̈‘ŠñŽDBSª‰ç ¯8‘?Ñ”ÎqzæÚw—&ôkô"f§ü$éê·È¸‘¯?‰LkG¡“ÕâoWNØ–w¡IÖ•&„&Ü4®ŒIÛN4×ÂTr¯ÏHiû½M±[ò9¨?4|-ý¬æéšJ»Õö ÊIø&䪼àcÊãåøÊol‡7Çp§|êD†áÐd½ÐŸk1 ÂG.öÈz &„B·Ý=ýŽ> ?ÞÇ.Ö‹I%¶}æ2?åôŸ”9“jiX¯J½[%æÊÂÿîãZý1-óu eóĦ4Á/”ÿÑ‘YÈSÔŽ9”¬Km›ä“ÍêÔ™q3µ¡ÚC{i·‚T µ+ÌðžðÅ}†™ìeâ$¬ÐZ)]ArI%v'",Mh´°”Î]7ÖšÄ×ꣽŸc5¡¬«xÐÛE\sãù{DÉ­¥©åø‹Aá„“£º¦]Êë‰z†^¥mÈå38˪Z½ ]¯‚J¯J‚þÅ©¶5ZÏî˜G© 'ì«Ø‘Û·[þĆè,ÆØ¬¶äq‘òè>z…›#Í$f…-BÖuèŠÔÀín蚟Å(Šùþ@PȨíÚ«-Ÿ2ÒýCøûµ¼o,_Þó'ºšû0PHÓÓ½}ýMê†ï¦å¦8µâKéÙÞ2(ñ®²Ï± ± /: ˜p±M6%—ÙG[­KÊ…U èÊÉ9aì]Ö¾2ê'‡F<ƒ¶œ,ó­YºpeüaH‚…ëö¢Pl¢çÞ3°ÙqlOƘw¹føÕJ€ð°=1sÚ 6}¢Í£Hè¬ P¿Ò“AU–5°msL™™ÏyœSã^‚ü⟙ï6ÅÅæ‹ŽÉÞ][Ì|…ÁÍi{sÙm+R»m2[+ÓÞpÙ«<'U³"îž5]‰X±sQ¡¬“Ve½v .Y`‹õS½ì¦¬l½©t†e§“¾2=&hˆ_Rf¦Å Z,k{ðA•@£æå*"UA3ŽÏ·tâ±ûq_6ÌPŒÑpTG‹ä:|yÜÃ÷˜ëëK;o-§·G2½ÍSn§\]¼±¬¾x7â<Ù6ð"õ¶šr|ˆR¦~ñ7Ð_w•ÀHXÓ¿€8)Å '% Zá{*QxäŸÅ#ˆidÎKšŽje¶*©ÖÜ5YÆ$ .N¢WÆãýû†ÉWíD!ÈõoY´uÉ´Úž—ËÒ>ÉqÇþòM([ع†T=¿Hù͈d‹+0¯[w-Ýö‹Uã £~Vt–ùà ÿ²žît&˜OC`.ŸCÿ]zõ=bÉÑcç% ½ð¬”Aºø‹dôVÑpƒéµÏüƒx ³ßpï”OÊÓf„ͤžãghÈb¡]8ºþþîC¯ÍæØï¥å²¿Ùä-iÒY“á•¥eïW-(üFø=åàyWéTÏ»êw–å u_@±`¬/tC¬Fà±oU¢šìfLVbå&詟ÂôèZñÜùX6§ÈTÜ¥òºÖ’fajô )?Y|#°ÚUÈÙ6Dz(x®à´‚~š;èO°¢™ŠÚ.„üò° ­Õ/{ñ2—jÎuµ¤`¢éø’b™Fÿ¡Õëñ¦ý54¾€{B]:Kê>,ù롊šÀÊ~*Ì7Š47¡0Ä»f<þ·$žÏaõ¦üAÇ /×´—åa̎ձؘç/¬t×3Øš+ì5‘QD„pfB‡Ù’™Ÿ¼0`}Þ¸…ÏÒ7ËP4vdÈ”>ã5Ó³oþ.›’ˆ‰Æ? R~³Y0…¾Ù) á÷´Í¦Zñµƒ9=áüµ._,kcC·0Í‚£žß‰)Ç··“,í²èÑôÛ´”3PVœ±ƒ…òtç†Ð/ŒðÁ¥Þ“%'N|¾í’šgu_¼€ˆ»;Õu=Ÿìs¢G†rlI1† Øts}ìt.ïèÉ'ðâÇIGt¾6ÏÂq1Î'’xºýÌ@_×nü"³¦(MöŒßßE…š·¶…WEUìÈÌ0í ¬`m=¸|¹bwbàÿãnÂļŒl0óa…«FÑ YZ=ÔèE¨WÇÞ×÷{6LY†^H¿Ä‹ÍN3Çf£øe$\IŸkx:†1Z ÁT}ÿé2ô}‹‘TøvýÏæ 2Ž+ZÎBʶ'~ÂäI[RFÐVØm®YvªlˆAD¼?ÁúhKÄ$µ)?×8"G@Ø„r±×Qs‘Ö™ÆýF]¶ß葈9s=Òô_aS¼ýÆÿìHpˆÙΗ~ ùåÕ o¿“Ø ²"6Ó˜›k ]UÙ`ç`è¼d–FòºØ)èžëaqkcÕu"JL–Õol¦'çäÙ¼˜½ú£*Œ"\ œBrÊ~~›H¿×õpCº_ÑH—Û“xÞ^Ü?cs`uÞÿ%3'”Ôä¿Ù!·BýKJd ‡3}•Ÿ ]íù¡Ÿ’aHAUõBоU+Í¥6¯óq˜=—aRM¨Y)™Ó'qÀ…ª8“:§ÒÒñUǼe!è2Œ‘6’\—­3sO‘BCû°ü²®¾p†W` ¢§N¦„ÜMß1:o—£Bÿª gD@¬x@ÃXÖIÍsá{]. §é›?*ve‚éÁpxCß}gôøWý³¼f_xËŠ+~´N€Û…Ñ|ÀÊGª¾Ð‹úÕG€PÁÿ½|‘D³øÂ(ò [?Dг#õ)Ó—qB°&°`€—ø¹9a`ˆ†¥Ñ¸-¿ú ´Úû²S¨:®$ñr\Nêè}ÙÄvÔãmAÕiÍ3Þ4Á+Ò&·…êÖv¯Í1ÖOžÍ *ÊØúªuš6²3…ß( _ÐIÒB3ck­§`„ s½»Ô78FâÜ7„¹>©/†úM_Î6Ìñןöïê ÜézñÜ6ö³Åzu°Ü¸¤‹1aWêVtÒâÃ.ñŸÛMñúa¿ý€Å}áßBì2 óUÊTÞºl{•¹$¥ {©õ¨¹EÒÊÔÛÅŒ†Œö ó©¹iÅGÿ±I§Õ+ê…ª?fEÌNe×~Ά‰ fÝAfƒÉò.÷ÔsÃÅ‹©Òžu Œêkc„élú·S’‡&f¥œšçjøI]ë`—`qóé= "Ãí6h.}Ä> ò;L¦~AêäŽÜñÌ׿ x³‹Ç‚¸4/ŸmìFB™k’³ºµÞe‹üÈÁVéoãçj­sw¢þ3cÒô)\,e?Їï- sê#b€Õ4lÇBÙÐëgÂÖì½iï¡Ø’°`£6ô‚t“£žÁ€Ë~®y7-÷¢)å’9­hãY´Œ¨ 1Ð’ßæ€Ù1¼N9··>ò±ì¼»ç…kÁ ,™ÇWÇ>D-ν” îê««/iYÌ£”V#aãì‹Ê›Ïåqä!Y#°þÓȽÏJ)É’Z0w#TôùЫX£/G—äÏ^^ %h'úÒÃʧœö‰1îé E8_DŒ„ì$ <)æ?,_T•öd-x0A«þ¶ƒ¶3´úÚû*¾Ùöwl~s„Mù%ç%bÓ‹½·c«øí| %iiUH–þ<À$¢—ãÈ«óδìz~Rú€2f‰‰&õÍð]|‹ qŽñ %¦yC?0Š\¸Ö+Mq>2´:áfkw†Äýx{þæÀ%˜ú)jçÞÿåZc:DU÷Ê7‚²á¦^‹$µ¯£)~KÍC¹®æ<Ø?d¹6ò~2…9Íò,/"¥‹WzÆë‘SƒNei‹jS*Ž~²{E^>UjY½Ç‚ ðÜ )¾xeKƆ£¯dt°ÂÎñ’‡ä¶ †¬„Z¥‘‰ÂßË;UŒ-énò w[Œ¥tvï™enåh™Œy«•°è.vTÚÊÚ:±Qù¨ºeÁ©¢£¶^~íÔ­‘T*ŽEÊ­°*4d%Å„Ú}öW-x´iÖ®7àÖ–ŽØ’0 ²Ÿvœ øðLæ0y ¢’›ºs¸8$[.#º®EF쉕ú&²¥óøÇ)I9§ y³ó¾ðGÊÈû6‰GFÎ\¡öÕ5´ó|¯´¾°´IˆL"GØ–4q=ÏL(d5—ˆo[œ†‘Í;ÚéJœI©d¥È:©­èvÒ¾ÎÔЀ9çÁ6qG {c9äÒbÔßNâh7û\t¿é—gD?tZ?ì*z µàpî69±ê;½A‡¡}$ŒtϹe¾‚GD>‘Ђä¨©u´ù4ÑAk2{° t^ƒæG£P2Ã(/ÛŽß ËÈOãñiªK%À7Q´­Ãœª?ÉØûJqÖíDt“ÞHr¸KRs¤\ð`¯‚n´ D„¥/ðïáó«²!u°Ç‡²Ø {_ ¶åz¹$GqE# þ°ßúÉß`|zã{Ì{ƤC'M›~ûö5FHIÕç³éÜë¡Ù£/w¬äöoñU•Ÿ`{jöê& [çŪAŽ3w§ ÀcCç„Pdå5Õz‡ë&r15e¥y7ÄyLdk n·‡ø¬ÔÙ`3¤QÊÀÌ®Øwj¾ HLB-×ìàI®-…6§Ò"µã¢@»ÈãwÕ$Æ™ó÷£¤½R® T*ˆ(ƒ_¦á÷êKÚ:Í#"wJ¸6ª*ΔŠ+ƒÆfí ±Ö‹±qãßd ؘtĹ7r¿[¿¾Ñ­Í0*›f%' ìÆ±Ä7Ÿßcú"Ìñ¬Ésy¸¹w´-“&NazFI>2Ç!ºÁÆ@,²‚»ºÌ¡„þ©AÑDç×yø¿zNŒ[œ®W.o“ò¬åº<¤—r#6âiù.í^¨ßZOœIæÆ%¹ì ®ÈAçóƒãEì$º‰#Ð_~ÝzvzD#í» ~Ì„žgöNû†`ùMTëü­å”^«èzF‰F¾Z•4μJê]“éÃø•q]S\"ò$Ð5A®ö‰  ÷,yÀ¿I{BÝQ:úIiýæ€ß‘´GX©Ì㵸xÛ©ƒxê!¥®¢ ¢—T‹ˆ:°=dâcŸ<’‡+˜jYªÕ*c–yN×rÝj—nÛ¬µ'=.ä›8¸a˜åß©ÖáfÌÒüÖëo·É¥ÑFpSô¯k¼HF ‡4Žw’*-FVžß³˜µ}â<ÿ¡sùm‘ñܽ¡ê.ššêpC-áû]‚z.ì(?^ÉD$½ÖvöRsäVAÓ Ç· Ð¥Nñ÷ÁÆ:/Q;{¦wöB6ƒØ2¥¤åA?W[ï5­À[t~?§*l!ÂO!"G1íãõ/Ì'ÜGöÁÏZ€5f(Z>A PZ’by\n$e˜Û„kv·>dJ¢D¬Ñô9×€[?Í.Ž¡·•¬fÑÆ_´×¤â\Äc¿0$¥-ų>š_&’ùãjy™n¶Z\Ž)­³6J.†x ¡¡f$7:˜B?Ò¤*‚²ŒÚ±­.‹Ì$„…åsr$ŒÀ ÎÖô‡dΛ˜ M;›ýSÕ¨ïiÓÉw)Y×A™T!¶+„º}]ûÛBé#zhÑŠÒ‘©ÂWù=S×kTgÆÔ!æ1Àèš®ÁèÛ:ÝÒeBãõˆÀ 3P¥\°DÕTX^¬¢k"~t•ùO"¬Ò.“¥Âð;î 3ŽF­§•xÅsÔÏwvC¹¹/2þÈ‘j4þãq¦K%Cˆ6Cšam2¥½‰¾×(ŸsÚgˆTªã«p¬»hr¾)›ùº™ŒZ„Úã)ìÆ!¥«xO“Ý~=!p˜/¦&Æø„AêÃ’x_Z'LÊÇô©cž`¿$ìT GÙ 0ÇëŠ ®R}¦îF‡ðSÐÆ7§- Ï–6—4m Jî*FÜÑ2UPeÊǪD‰åzOGZ$xäÖ>—O6¡WLµçf¹Æfî&§æÛ~kÍÕ«ÚîÚ)˜ÇŸS}°W°-RòÕ*†@,, ?tt=ݰ­Ó Nüðâš Ä#ÓóÎ$5Nvb}XdŒeëgi·×03П=|W‡ß/„QÂààÆ $ùjx„uS`‚uHí»®&|žK_Îø^ŸÆ‹H¬;ëÝ)²#m´»BõÒUÇ_G‹V_¾àFŸØÈXøùö²PÁšý€õ'ÒO¯ýi,±WÿÝéA<ØÓ{j®ˆoìŸ Ø3oÐÔ »a šÖ"WhŒ™1»F¶·ß˜î$1@NNÞOßXd¾(?–tSè©\²›I;´eÆð©˜üJ±ÛéȲ @œé­ŸV~üŒu7ÍÿDx|Þ±1»M꘤»C@*'d1µ|¢éEE–«y6£cP* 0Ã9´ëž—§Ñ6[«(`9`=SöÒ¡¹ßnìë/€ L¢%RQ4#¿y˜gMòxëÞV—GhWTwÁS0ø~Ò¹oÔ{_,ÎÑä÷rŸ^+ÊR5äè­øs«§Ý-sµPÐ?U"~$Ø,ÿÓ|d½›>ª‘Îy©× Ž5‘ª|]OÈû/NIüi•êÄÔ%0 ±®ï ß“X5ÌÉpt©Ï†•}D ùPÖU‡$Œ„Zѵß×ÿ[î£ðÐM¼aÎS/ê'C¢:³Zž¸é V%5ߢcÆê Z/î#9¬,Ú†Ó·»RþÙ¥0-(PL⢫ðþ‰…G6E4Ô¸ŽŽô17Žù¾ ±Å† ¾%ªûe–àãäœVÝ®¼b‡~2MúºÌDúdM© šÒ5ÔiBð«c´f^bƒæeHÈb KH@¢o6Oé&I¬¨\¬M³s5Ÿ¦ÉŒÊ«ƒLc9ÔiËËXÔÍ£“ŠTÌœµÝíxý:ÅB±ž0ô¾¿Ì‘I<Ìp¡’¸ª¶Ì»÷OMøl€zò·+€–šGHx‹Û0´•y.Í>' #¨µ«HZ.GÏlãf ìäÆè§®½¦™ÿLCÁ³…u~ÿªî?-, '…Ojý€/ËZéæÔÔ$#7±ýæ•—;ÛÚ X½oØ>²×õ8U ¥iÍgp— ršm6ìôzuÅÛøQ²ö ®.ÚA¸KhöØ—@tÜïøØïð'dÃ6snå"\Î R5»>éÑ%+]`øµH(\ÊÂ6>ÒÄ{&í5Ò”¿ØÄí‰ÁåÃm÷á´à9Zø;#¾Ïñ—…ÝnUý{šÁ<þv—Ê’~ÙuʪrÚ—´Í¾_&¿tMEK†ê1áŸ#G3X6ÂýŒ|pûÞ'ž "? øn'Y÷ñ¾¡( R—†žM(ÀBTÞÏ»r›yL^ÇýX¿î¥™\Áåå±ÉkºéÍdL ŸÐ3ªÇcfõ³À‰|4…C&÷t7Í÷ÃWˆ;†2™ï¼‹b{–Qž rPpáÄi&XÍzì¡üÊK…‰ª>Ë\ éŒ%!õýÍ}qãd%ßxéöY?ÈÄ7wFÖåžêYISÞ½¶•Ú!j#/ãß$bH±4Ôhn‡ÔÛÆS-Ùõ!éDiê¿|ìoÊüHnì'6ËvŒÞ¾þM+Y¦Pú¸ÎG´zŒw×Àk\(Ñ‘¶ìd¤nxñþt³C­©~O–÷v•Ï<¤”Η©r ºU/†œó­q™(9õBú&9ö6¼ ¿!õÚÆ©igÈ=?†S«z2q×˰æÏ|½@O-£G#µQçh¾.ê!—ÝY{ðÎI¹±ç÷Ö|.„ÍMúÞ!¥´€ã)ZIG¿\ÜÛ2¨·ÍñÃäÜž}®ÃQÕƒ[™ìVi“›Ù,‘,¼xwR_àz¾žëM/ÛsW_æ,(3¦ÛþþtC…Q|Μ©¶„-s"’ð¢8‰—.JMçÄäZ¬|^øB6:¿ËÓÝÇoÎ`S’ô÷òíÞ ÆØâO8;$sËÝ6dãÉÙƒ#Ñ1ØàxV]Æý„\~ƒ½1ÆÝ‰‚oøÙ©PM U4Þ68€Œ‚½3Ôä$†;õ¿:(é2ò’y±+Bì&e¾Á“‹ l¥œ§FJ³O«¥…êgáí<à|Oq¸XÃ,ûèúG¿ Q4$@‰©³KZéDËÀv™Ü¼*ýhn¢fW/½åŧ+@ºÞ>Óq7þ©Æ¨óŠ/z Ϝۄ;àJÌw·¯} P}a‰ëÔñâ™”ô¼6*â _4JÐt³ê0Já«|jÐ Ù$¢Wd‘E%¢|é4PñZ-í§Jó¥ãáÍz«Ã0sIAÊvîâ’„ò³øya`øo†ãq_ŒûÓhÊlpa ›(û›ôeמ¼ê Ÿ/Ú¿Óx[ õwaÄ&ù>¾Þ.Þ ŒåY(µ ²èÒ¶Ù^ËŸOy‹Ÿ*ˆí£Ð5ö  D !ó¦d^{”^ܽ€¤‘¼ÚÈ `hžîrÜ•£©ùn(•3ò¯ç( q¿ÐúWdh@îÂÔÏ/‰0§Ø%¨»üÅ’möö’æAÜÐw Ïòw¢&H@A¿ÄLZ•S¥ôLì·Åg³ÕÜ|¡ƒ`Š-ÞÙÛΡ[|N5¸q©9§æ(nÖåsI f”Ö®T¥:è^CµÈÏ-¯Jµº14lË€´Œ™¶(|³~À¯9ŽóI"è€t'ŠÙ{íWj„{¬ÕN BÝP8êšÕÔL›â LK£Û1~Π˜fºº*Ûnüå–€­ÿçâÌÍKB¤÷„ÊË‘ܱ"µm_t£³Ïñg¢¯ÄÖþ(Ζ:òb^c®`,˜ü’«+㺰ŷßòKxçŒa5¦¸¦s¤Ó´=×IÙÒ}†)æX•otçU„-L6Ï{ö«%žÊ¡ ¿ã‰ßbÀÍ‚û&Î)â×gׯZ:ZÒõWƒgU #¸£ …Àl”õFr0†«ÍÀ!CX³ïjê9›ˆ!êž?£{¯&Ì›Sœ|²Øô…Aæh@¸Uá³ʪ)Å6µ»kI½QJÕ,«í˜í6ÞÛ‡ñ˽¤qÆ·Ž+‹…¢=h”mwž×ëý:©USC·ˆù”o1óëF~…5ðO­:á¬Q!ZŽïÐ&ØŸH×ó%A{“ÜT=é›üý‘HïÒî¶«2'Z}›&V)Oÿ¥9öÝì‹£Ü.“Ù\%‡E‹[q`/6b!˜%šÊ';9|LÌh4’ô3'PZÿï[lo' Žß‰v'(íÜÖÞWêᡃʂHЙÅïXŽèïYÎÌÓ›õžÓ²f‘Ýäd:øø§\y)ÄømŠñTqÚ›iÉå<EJwÕ.«Pá dK‹P!زßMOOÎÑi®~ßÅ$T´ô28b¢3˜T:ÅÈõ"âTt–º¦P‘·Rœ4›ïÊn%\ÕÜcszSØwE$*.¸¢ež¨h?'ÌÔÉ6„$OYÐ9±DÚû0Òašþq¢Â%Sè ö ÿMlçän5?òˆ$噌?²ÜŘýf55Q³eA?HÚy‰H$ò®ºo¼j—’Æn–ì$Qöå@mr<àaíÐ~kÞ›pR,\¢çÌ Gg›·Â7'>•b牼ó²–(ç _NÛª[ ú&…Ïô¹È¼<_*ýH³ø˜U dÃ`âþ5E¹ÁŸ|ü,ä‹D*n;bl€Ý·! ǵ;^ñÈu2ÿ£ØòO¤b‹7‰ùûô°»Ôg£éIš½TÏW1ÛÖÖÓ”«UÎÚ}­´¹Î+ÒÕàY—Zvuã©k<ïB‡ªl9Ý$ÕŒ|Æ…-X|·0”%{bã¡'¿EÞDÖ?º…¨Ÿ ¦>4õ¶=­fG(KØídè"(¨p[éÃÑÔ» xíéhJF¿!;Ñ5ˆ¦ÿJm¯L«¤Ù¦úHþžLµ6ÿóŒ òE2&ARŸ«ËÄ~ ÉdÐ]Õ‚BP…™·û¢‘”Øøî9ƒ•6$ì¨òž~íÊàûJŸXœr7LÑ™'âECät¸ú8s\ߘw,çÚj6!€·ŸŠíKk¢8÷¡ÙˆÕÇËK¸‘sÈÛ+7óíP¯W‹ü§3~\v8’ >úijàSSNÀgMÞ­QË$ÍÝ•à+¯€)¦ÁH'XiÚ¢3Üáåoßj!;G+%ˆ¦ª/˜š¥xèÕªÒ“ùØ__¦GärShvO'ðöý¿L¢|Hެ&š=û¸6@`4íPò…z[#O §oI+!îE+W˜m /êHÆaâ~í–f‰–Ší~Ò=º¸d‡Â9hÐúš9õQcYT;N+:Ù[½²æñã È+Bn¹Ù¸íÌ¥“)­†ì˜æ»þÛä0ý¡NñÊp4¦I„Šˆ„ [~(PRBíûµÔOÐ¥†Éyp|[¢ûÂ¥|@ºŸvTùžµ’VC.¸ŠK’$󜠢¡ÉyŒnÃÔ‘Çÿ-‰·Hb=ÔŠò¼U†„YÖ\ÓsȽ_¨ß¯ÀQýxATZÛ‡¬qM q6‰@Jz“9íIÌ©L”?îŸÍ&¢Š•ÍI#x|]XSoÔ·M(ˆƒ>~Ö–  ÐA'ùúí5–þfYÅ“éD•³ï§[ŠÐ}íønLný¾FaÔOsó[…ë ¶¨ï°Ž¼¥ÊYmêÃózÚ:w$ÅÞÉþ kf´œsB!©G1¿Û¤‘ÂÙïŽ;ƘO£yó Ø T€–J9jŽìiNbúûp+Òƒçï>9µ'w_…1¡9"îΪùHDþŒhÊ•íô›Õ2 N‡xQCPKøëwÅVq63D…i{ÕŸ¿D"K×6?ú-´Ø…ö'üi¶Öʽš=j¡M}ä$»Ÿ± ­<×¹z2ŸÊìH%N·¯ׯß?7w:© †Vû²ÍkN£tq‘OJst×)<—Ѿ´uû±;Žóø¾J<îOZû½ÌŽmª1H'(s¾RÀ(ë;Äè\WŽ4hÓ·{î‰yAmîáÊÚI³]}2\Â>t±sœï˜•Y’P¾[Z/{¯Fû¿õ¨î¨3ÌÏ|¶Ýði\NeDFrýeg?IܪÆ5Þ\Î@g&æÿríôsõôÙÿÊä¸÷ëû*//£Q½¾ƒŸîfa,C‰‡‰¸À±¼åxçyh¯–¡ýy„Œî«Y Û]áE],ä꟔ïœÇtPN¶Î²À˜´ÚFïUf·1§„mqÌ1¡lèêØØ‡·‘\ç&1aÍú Ìty«b ¬†•|Œo–ZîmÚ|ãe+¾øö¯p%He,Ã3áŸD€ECêÁÌ\*Àyꉓd¿v`ŠFF1u²5>AfH>oF·öõs“Rò†8Öû>P¯Åƒ¬Ï:Árưe'®‰tHØ,Òœ(˜Ÿ ¡Î´»ª³ñÆ<ŸÓR:0'ý.3€ÇÀ|  MT÷Ký“8œq*hŸ½D Z¿¥ˆ×óË8æ~…ž…ÑKGã¥<øjܼ{³÷‹"5„,­’IÀSG<ÉÆ©!}o6õÞOpL.>”ƒ¡¨/ôwÇàœkîBúê%£Ī#gYå\ì÷Ö“WCD-²CÁ‹8 §”ü­š;3úsëu2áÓëëíž6Ü"Ì/ 7ßת ž\Mº¶g¾—j}j:ø½“nǬútm]€ÍwÔ[€þÅÆ1Â4̾òði-é‹&n)X›oGk(Öë‡A´C/&´)¸Î+tÐø€ Æ¢W_¤&‰1ßþG=Çž·ûë{š¡rÇ—1S§äK1âÍOÁäXKs}œo}÷Š_4ÏðE#ܼ½_}¿|ÅŸp#÷³N‘CŽA>Åê7CSKÑtu²uÌ#zZ@«‚åÐ ùHU1‰-í"ì9Žö ˆÈ¼gZÊ^FÏ$©¦ç-¶K²ïD§Œ`õÁ¸G}[Y{ Îêƒ÷Å`<,'3ƒ õÔ>Ç©µ9òŠEƒÍ!OS+>AHl”Øù j,EïÛà/Vìÿ`u6¹[̪ŒÛÎf³¯J¼V'*:žÙíð‹Ÿc!Cµ_™·VxܾÉëg ŸR'}zOEùkVÅ úºˆpàØ àŸSYø¹ OɉVõeÌ&à¢^6 ÷V³R[—އ•ªQñ=UÄž¹&ª>µÐmqˆ~éO®ŒÄà¡ß“¢Ä.±–ÑO¡Ò6ÝT|û—AÏ6´MecUÚ =í´ü‘0XºÚ‘cXs½åID˜óù-bõ2ßP©zäó¢n÷1ȽF‰ßSÌ Ûây½E˜ÈfLù¤èyT3é u(ÊÐæ•Óo`m“,`ö]Ç`Yì‰Èøcr/¦´ýÐaé'[äêì\–¯™³µwr\EÝ(g^{U5|SæqãñËþ›ÂÂK^òôV ˆmÔT¡zÇÉè7­ÀëªcÎ eM…+Ãr“®[Mƒ|âÜ‹ñ„>ñhnÉ… žÙÛó-/½'„'Çe©MigÛtï]š²¡*u¨m)@=AŒK­ó"v>ZœÒÌn¤ñ³ø‘æ´Ù“(­ã„éÚ”ñwôZö[¥áJñ[ùXÒŸïçnÒ.iÕæ5žn ì]”Ä.VX㈋ÔÔ"öæ’gÜbÃê±PÛJBkºæÕôë„¶ÞÞ¯þÎ…Òð¿}†çz.Rl{»`]œÓÙBlùùÛq#DI0ߟ³e”=›VÚnžrlÒØœ‹zöfL³:²v7öÓÜtµ"¿Û¾ávÄciêå& ÌQåïü4È¡ { 9îCºurʬdfÀ&çøø'¼j~„Ù€…TµUêÒAH5tÙ† 9ô1yÚÂx¶çÃoŒ}Ö£sÃeoɧ°_õk‡e'P{á1"¢Œ5^¨½‡jÚÆ¬nÒŒPžîf‹muÃ0Ÿï)Â`)°¸–3‚½°|R4ƒ'¸ìB€ù«œî,cøÍºó?Ù¯-ÊÅî˨…EHÊÑ“¥¿ XFdÊv[‡±&ßEfë$-„«ÇY rškÈïy˶©ˆ¶H;‹m©^•¬?[™ÕìÀ¯¶dÉ„³âX:t¯}yΘ´SÞëɨ½ÔÊ·é†B`±à*¿çÿHšy xÞè­#vÂ-&»»=’(,¾ì%qªžEçÊOãb´nøzNúirÿÀ]—î)ÙðI®„®¥MªWLÐ*ã|H›Ú{Ѭ¡¯sÀM¤ÔnÒ$æ|(š_N@K;«í’)Ð`Ç•ˆ’Äßo6ç;?ÔSuÂÉæ­å+ ýTžx^gé\_»|£ì’ñ•Gv¡.ÿìðêé;‚® Q“æù¬²øC»¸¯Á‡†i:÷²H áÎhoÞo€—Tþ¾5æ)¿ç+Ч¶aØ1 &:'v…‘´Û"ŸÅÛñ!ŸNtÂu‰ë˜ìÔ^Üu˜o~ÒÍÇjÚùÝûÑM2Óé—£ªÒÝÂ&Â2H qw´’?êEçg&ÐñÙĦ@Ó×±$ÇQú.ÿÛź 3‰ìäÞw´F×ÅZßýÏÆA¬†A­Mš·öšÝn£ÒƒàMšØ³Ç&õ ãµ€$\ E¾¯äµŠ QrÝ+Í1äwía׃áµmiWK3-ø ,©ú\ûúñÎq6[Î‚á øó¡ên‰K†ä¨ïöÍ¥ U$ŠE >ÖO)„ƒ‚_¢nñ ³¾øØÌÐËùïµtº¯(ê*ÉæAÚ1äë_¸Ç×fDf306_ûÚKMÕ¹p‰‡;olóÀbÙ¡nuÖÛ8@ .ÚŸ{N7 ¿Î0;rƒ½R§¶3 ã:Bù7Ïß¡U¦¹½ä2 y¥®\>àºt„“`î­æ× ¶*²T<ý&9äìêûÒÿæV"m endstream endobj 120 0 obj << /Length1 721 /Length2 9088 /Length3 0 /Length 9670 /Filter /FlateDecode >> stream xÚmuePܲ-6Xpw ƒ»»»;̃÷@p÷œàî ¸$ÁÝ=¸»x9ß¹çܺU¯öŸîÕ«k¯^»«6•¬³“»¶ „ƒ•ƒ $Ô’ÕTáàãä‚ظPéè¤\!îPg'i wˆ PjA\€œ@Nˆ•(åìâã µ±u2X1þu,ÀPG¨+P×ÙÁÙje öôô÷tó`sõeûÛ¤Ým!@k¨(¥¦n  *dSÕÊAœ ®@uK¨Pjqrƒ0­]ÿN€VÎN`è¿4¹±ý#ÀÉâêþW˜µ«³#PEF[BVMU(#Å®-´p•åÿ¶;¹» þeC¬\ÿ5 Ðý¿‘ƒÅ#ËÿFÿ­ºzÿ;bCåà‚¡Vî@Kˆ Ô •ý_¾)8Y;yÿ ƒ=\þSú+Èí¯> Ã_`ˆõ_¶‡ƒƒª…#È åìèâáqª8ƒ!®N@MgG §ÿ°,¡>ÿ_Þÿ0Ü-þú ádó×9п!¨›,ÔV‡ºÿµÚÚÂá¯Eÿàz¿‹  õpüw?ä¯_ììäàó?Wþ㟠ÙuÕTÔu˜ÿ»ÿ”eœ¬œÁP' –û_/-\Áÿþ)«[@ÿ½=ÿóŸmrüo®báî õØþµ ÿD&ÿË’”tööcåàã²rqð9¸¸ù€ÜÿG®•‡«+ÄÉýÃÿÎòŸüŸE‚@¼!V¨KóÎVB¡vé-áå2…“æÉqsæyØ…±®s#lè€-Øì÷)É•6I¾ùD°¥ÈæÉ†ÂU3µEmod4–*Òdžä¡˜¦Â&°fíIãëc}¤é‚.±£•*ÂyÔ¥{ÆòÒ'u¦Qî¬Ü£Å¥"Ør±æeí½¬õÄÅ™'9FzœLÃW4È6s5Ö¤mXÁ™¥ØÃÖ\D~È&Ú9±™mAy°ËïâUh> šÍáƒ4Ƀ N%ÃìÅàkÛ»4µ.ÒñØvcS]#Ã’?éÃ--¡<~EÊŒ|îb ^îI¬…Û¦6#¸ö”q6YlbΜƒö3„Öä€~Âà8âHISk!#ýÕÐCbÉÇ«wõ¼&ˆœZ×™Ñ[9nܶ‡·rØUc`¼fDµdzµiÖÛ%g«ü'©#S¶¶Ò`ó®}î•"¶\sAÉÂÝ~RȈlm½Y !õHýW\ñ󟽻¯A2‰µ³ƒ]ôÈkÈæ¶úÇm!ÃÚÀŒ—‰[Ž×Ð$˜_ÁTðÑ^–°Õ:þ»ìÍ8ǶCµÙj¢¢ªoûÅ÷xááUÄ|f )-·¸Ï¬y W“u’ű’3ëjy– Ç8ÊUuCÞ®òaµdöæ^´L$hpAf£rbEîÞuµ/¬*1Cç N˜9H>Äè È@v¨¬a'/8Dš0”Õb8=?6ž¤r!¹ëW“¥m׈¾ÃOO+dÙlTòü\@%‰ÜQ«l‹I3Aòp4N‰µæ/ü äÛNK;~]¨Z/ÿ±ú9ÃIôG4’/ÕÁ¨°Õ¬j'{Z³nFÎkêV¢9ë½j%¯¿]åÆN w»íKt²¹ìClЦš`çþ¢´ËŒ"CÏç«-Ðüîpàcù¯/] Cùý„¾Ëï*é‚ⳋx`é46C}´{ÏŸÁ¡ƒajêÃK ’Z_/ó76´iµZ´»Bbf¡ÜðÅÖÞ•g1! p*œVŸ°|UhÀ&„ÀÞþôÅú€Vûþý‚OhywõµIôƒ[ü—‚e}ã-~ÄØÇ@ÒÔÖè¡¡J¯_N³R óÚñ'qýüˆ¬ª¥x…Z`†XÏÏ´¼G•<_±-õËY®— ¢\ÐÐ$;0½iä'9ætœjwÒ«zÛ¾¬ú–ß¾@¥÷|;Èð`¨o! #‚øyÆÓïÓ?ðØÊ.ªn¾NYj°O™+]Xž˜ÉSI”Vhä•>!*¹é®Ùy ÓÅx0»ç˜ýÔXÔ-ë%0ýÔ å¿3FîKë¥É?¦“‹¾'PºÖý•LÄäþ¹+h¬ñ£vÀ„¡ DÌDys"8Ü΢ö©K¬O—ž¡ r%…E )Z¤ýÙà„‘:±¡E-U©®`šº½-ƒôî óññ"ÏÜèÏr}AÂ¥3yˆ‚»Âý)a­ ÷ʾÞœ¶ð8Õ#ZxºÛ-þÞÔj0ޱ±/ÿ^fd@ )I\wc|4¨¿`œ¢wò)ýNO$‹Üó{rŸ%·yy︨ëtyó´Ç-+s ÿ»-¢Fyiì¤8É ¾ý¡¨R8M9¸¼` Ü·ùþ×Ißi6¶]k•‹ÜV––™¿³b)!%ñÛ7UBìÕÂCME§õ­à\•RI©2U+1…Y¾»ºà‰QšO £¦¼Ìîè¬hk/5O­_|xO‡2›ÙÊîdK‡­ÉµÅ0Õû‚ñþz"Mn=æÑówK'ÊG6ü¬`Ć.Z}PŒ£Ž›ÃÅ÷7ÙËJ¡%b„k—SL,À3A4LHøš«@yÙ"Y\cM¨ìԓʥˇ˜ tâ“R±&š‚22)bZøË[œ”ªú¨ß ®¾sÆêø6W³œOT± y)ŽºZèq¹KÛ§Lµ|•”k¶ïæÅ¹¯)´æH„ŠøªëÁƒ¶ñà4S¬k9ûÁ-àtȾØÆ%üÁ6Lĸ£¦bÍKá§¥YȬ|ŽÅö¹ÁÖ nOïók¯ ~eÎðØí/Ùž÷«VòM¨ÉOñâqt{W–íaô !²ÈAìÝiä7Óx£dï“㈃yùÂ*©ÒúS†ëËÅé³RPºE|±rO‡d=tVÒÆÔlä>/Îþý&Ó‘jŠ z Á‡@*æoÓ°ò"lùé §d2úæÀi"ªÃæ.OÞÂߣgÚWž ¿òTUþ\½Kú3„uqþ4œ·ºFJ HߓҡKíÍÍ@f—Þk~ÊÕÂÜê°/ŠÂ+{Åïn¨n·]ëÊ´ "0Ey*½ªfûÆAÒ'òD¾¦æl̓Žð¤ðõ©žö0e˜Û6(fL“6X´;ãa4fA1Ï•œ±ÍÏmé¿9ì§ÈÛãT•IJ7²a•ÙImQšM Íà‹ŒâòÐPÐh8¯É ¤óz=Ç$Œ©ž›É>M0ÛGÁÆõ."?ä!cÉQ1†WM¡K×½ì·L°UtŠ]@®éõôªU|]²­E7¨a„öeùÊG߆´mšæ\»4ØüyûSö—''o~êw *l¡V6ÛnœÌocH%|–à ¾kI¤—ˆ\ݧâ§KØ>ÚP÷W–8"Ô. bÓ&`;M—˜‰‘ÀãõC}½…í|[4å/‰ÐA7N¯š]ùÕzO:1 s¥|`¢4£°")óèÿ º[ y}ÏÑÚ› ÿöŽ®®¡T¿œ®O8;nJ-øVfÃAPÉxa zGM=h%©Pk‚4³œ-±Ð’Pš½u€Ò45^$ó©ÝEÖ'¡ù×oáG_ñŸ6zk6Ó¹¿mu–=%¡'Jc?€v0k}v€MÀézøÁšO{׫þ‰@¬d¿rà" tÆ[wW¸­ÎêþÕ_*¯¥?¹vQ;ˆ[;¡ä\ÐÂÖ&¡Uÿ¸»!û6QÝï4ñXÜ[ô2À§\µËჳ‰â‰Gôâ´"îdº|íOËðSãLi.ÐáÒhw0ïíh5Ù±bgOÈ åA ö+¢ÒÊΘ*Ròê…çÇg¹:öÎHó@ƒb`_í‘Çë㬥z-±®fæÒ—ðyÿÈé†~ùËŽØŠ— kM޳%°PPŽÝfî¬xЛòűD–ã¦H§bê§ù—æ×²Ñž& úËŸ´÷Ï vÍûmÝb)ãz˜ XÇçoÔÆ±#þ<°ÜÀËJÁ{‹‚ÞøµgH¤É,Yè×·wW¨„)‡ßUÍcînR~ÄáWs4an&ÅNžÅ¯—õ12ÿP]ÏãÛGè[—­$/È4&ÊyìÑÚö­’•aTd³àŠËô^%šmm~P6ñH_•Œ~¡$þ5Jù\ 1GRÙk1–/þÏOüÖ¢­yÒ´_W¨ŽäÎ;†SÉæfTþ%¤G5Έ”Èy.MÎj<Ȇú¹z^MáÆž…-ƒÓY”?’Ü(H3†[£#D?¶]´®5RPÀRkl$ÛñCžOæåÒ¤‚ÞÙ–Í4¼©ñÅ7]ÖÒ7âößë°škáe”Cð⺼Cœª´‡Œ>t0±×?‰m[‚_çç «wKÝ|_¶9,ìÅy7Ÿ¹Á”½6ã›ÛüaVäeš¥k2w‡ÓqH/À—aâÔCì ”µK×yòÞv4K;æ›ã¯ Mc(Òp|ÊLÌÝ<¶mÓ~ºFÖ×aªeõgêrs…þ=/§^q,ÊÂtw7Øœ0¦ûªˆ\É •@[¦dÒ|ãІ…¦Þ¥1g_;ýéJ8ÞµáÝY¼hßöá÷f[®K Vº›gœš –dÞŒ÷&&Q’>ƒeïË?ÑøPñi:¹r ®Ê ™2¢«owÍTyÊ7XŸïÎn7!ØöèÔàIŽÐhÅÅð%P¹•‡ŠAH 9C%UtkuÐy|/W:dW ¶]Ès¤•´;|w ë j½B]’£ÕëÑzˆë»k$ˆ‚¡uЗ`Ä® §o×kÃÞÅ Ž<þøósŽ…ÓÁP‹!RoµÑÔåÊBâ<¾E1"5d§&Œ0 ÜÿüͰÍm]ˆêö„±ÛSÕÖ2H*á™ÌA™½å %æ„ÂHNS¢ÝmBïKÁ'oÎ"Ø|oõÁ‡?›SÐèP ?ná8!‰Í¯4ÏßJ Û ŠF9|,ë¼Z¼åWúòçMµbïJ˜ X“þdÁúŒ"Q¯_pú•—0¡=IZõ=Ë|ë>,Á_fÙÇç’-%ÄCù91ÌlK®„ÀBÕÖ«HËo8žMÍgG dY3(vc§µ•çtåÜkaÎÇM£7Ÿ!1$ÉÜõî~­ª³E3ˆ6ƒ¶“`æ Aa{™aVßóeÀ:#¡@|°i˜üûQâbÖ§9ÕHÄûý¤^j˜ÞEéñEþ—u‹Ghˆä=çTû#çû<1m½ÚPzðl+ßäJ¦éŃÍÖº r0(ߟ¼ðX†=é@tú¥¤WÄïÅu*OÁÞm§ü¤6wÊ'¥…|˜Ô ²ÍÞˆ‡'<<Éx¾KÖ¤ÀË?€=+‚€ÏkÛÑѲI¤J°7BˆÈ=öÐàÍâˆ.Kˆ'¯Íß4Û×G?![3ÂK¨¼`ôDaÌpŠäŒ|‰<%Â7¹F( ®^ô&+. jèÞ.7ñŽ¥øâ‘xDq—ÒQe$—šÆ"{z?hÝ—ˆ_¿K²O3¢ÀŽòÛ$ìrÙ;6æp†Ä-k8Ê5ˆTÛ£]ƒk”ƒvÍýÏûŠz@ü½5ã‚è®}UUÈSòEà¶ØŸìµ9»ô¼j×Ó1c@&$JÕëIÉ.Á2ÍÍüä þwýŸXW%6i¦j¿¡™EHº;òÅñÝ6bŠaEêª&æA2øu±iuL³ðƒJiĵ.¼(œ¿»‰¥ ‰”."î“Ò.lJ[ßTMDÛ¬›¦wºƒù •¦%tF¢ÑtE&\›@·Î»FS°ˆ¯uò+¬˜/NÙ“®ýRÝBÉëdè3Í,îG{DÙ½h™æ‡¥/Àèõ©0k¼=€¬ _Ku^b:¯–ê>á,žÓ7áf;Ü„ÁÈàç÷ïy‡)?‘êY˜5ñÛ3zŽõmü°¸½ŠRR_ê´Lt±·Ê—añטr4Óy;ó"ì¼@K¡ÏµY:àll­ ÐMv¼zùr]pÍyZ‚žŸ0‹?ºžÉ­,Ý-0æåE`·sœÉ÷Wœh{‰ÑZÆzîßu¥Q(™Z“¸å##§|‘·T‰.å’|ít–‰®9Zp¥™õÇ ð()@g{+ôA5ÚÅÉvpèxluPÖD'+Ó:6X¸˜æÈO9Ls@¶-E{Õ0ßÚnÊI¬ƒ—DóžÆ—}‚–Pïä µžˆÖ{µ ¢U.xO¢Ïƒz¸ ŽÔ'©)ßÌ+äÛ`¡wy¿#õyb<ñ+iÎâ´lá[Iu"u$ÁÔg°Š𪜇guBôðYQ]—y§¶àþd;õ…A cÈ€rÀs_Šœò:7KϘ»òB¶›SóXÿžïø}aGE9‰³§ZÖÀä«2Zš_LAŸD3âá]°j5Qh(×O7Î8üˆ+»;IvÏ…ñù+mu8—‘¤—·Æ ›)/¡êh}\…ô<1)bõaìW¡£J”ûßòµ ékÓήç…~¾µ˜†Z`/t[s¶¯ h¡ =KBäû-€1VÆ0JíŽp¡$kEðÜÆww ±‹{1KZ7Á}Ú“êi7¤ÂP9ßî™DTLߢ-…7XHàô{Nªq“(Ï^<Ä=Ð3Pù˸®°þ.ΕýŽqGŒÔÊ£l÷ŽRȼ隵âš4ý{¼d¿Ã˜¼«ë“›íDNÛáÒÛÉV—³ªðœú±†%C«„Ï : –ÊcyoÌh¿W lÖnÒxŽ{ ö²¦£Ò´>¬HñŸá.k±P[ñŸÔBoˆ#£`SJŒÚ«”‡¼WßïuZT¦+ \ËU‰§¤ïÜ9`ĤymÜž œ™ù÷ `;¤¿Žþ]SDµOXKH{Ü"Û6Ùt\Ê*Å vò×(t™Nê гy¾!twDßÿƶãL-¶›—´oàÞ†–t ZA»„ÙžÞ®ùsõXtð£_$x–,ã•‚a‚Ý·†çƒeDÿ«ùˆàÆÑ;ƒCÌd~‘öý0¥kkÙäÓ™n9[{a–Ê^<11BA2Ì¢#N‡¿‰{´ýJ-)ž$ȇ8®!¬¬sGjQõÅÝž‰BûbçÙЩ‰|]y§Xº÷Òj~<ÔW£õÓº`ñ(bU$ù¥ó“Ï9FÚê‘ œŒafRŽWfÌLÆJ.Ùíï„"åî§¥Å^/Ì70JQÁƒ×Óéøñ0ªGaý˜¾gùï$eسI"Zñ `*;Ë»AKoíƒ)ÎXâ4Wïñr§pÇ­‡ñ}9ªg7´%^DÂaS^/3|æçO:“±Ã¿%::ïu§wê—ÕT?Ï6$&Eâ•ýYÂý8€o)¹¿Oz®ùø·ýñd—n/ñú‰ôfí ùË¿ÔÂE–®ýat¡W¶D¶ògÛ¿Ç£~-‘i¨‚ˆi¾!ª•aðØîã‰JjX´”ZÁxA‡_]0¿Éãtˈ+—|šÊc¦z¹›ê»ªÑ5¦{ûð¿H? AżƒcãüåMÛD(ÏÓßyµmg‹‡£ã ‡ l,Õй€5ƒílJoö{ÞÈta,¥K÷\Ï“ŸØ|Ñ*hû'È,4ØÔM…ÓFËC_ 4µøñ¦*–IôáöFº¾½Õ¦æãi¯QOI?‡öúÔQ|‚!Y7Ç_±I.kM¡MvBËã9qzê§{Ž—Šm†ýñ>¼°!ª~åpÔgñM]„³8®õ+›®nÊ>g¾´\c˜Ð q”Ÿh±Õja'ù]+•UtˆøÖŸÞøÒ6S…CyùbH’¦gÖ!‡ò+Ç>³±§å¨O{‚Îß9}T!oIªF÷0óÞ‹\ìé+  뇋»Ù;ôǸ±j9½PNÓãba_) [‰ë4çVåÂWêR×5i™ÄH¿uj—ŸµJ©_žsûRMekÐ…j )?LÁ’iç© ta˜÷ó;‰ä4î&Y«e¥OU±ûø$`†ò«Åé§V€åev÷û}˜~*Bá¦ÞG8Ɔ±p9ØÑܶw'<õ# PÎa åȱJÖS\2ƒœZàüL$tÎ m0ëtø6¡%nq•µûv`¿æb¯Æëí¸n¯©U-í7*8e˜”FÏ€Úèùöý4Ýù‘Æüw=&ή6ŽFbþÄ­n(ÏÊ-ãänåÙ\¬ùSv³âíàQ#›Ÿõ…cŸœè¥¯ç_ BÇꇅP3d<¼'j¦å¶U̵²ÞU[I^º“ðŠK_O e ±×‚næõª1N{iDœÜKSOç¢ +!(3ÀÜÚñáakDÆeu³»êcwú€&vEGœzRÈDíÆtè€xȄҩƒu÷‰³>š%ê‘òÆÜ‘¢)¥Ò)q<3>N”3<Õ Æ7.ä$Eú©ºÂÈÚJ:ÖA¿ÊTå®<õ6äPˆœ"mt± î.Sêƒx±Œȼ¥”P­ÎS¼E=!’ÂNk<â§þ°yG§ÝÒ^Ë4°ÞgÆæ„~¦ÇóÜjFÞ¶¦<øEØÏíYÅGsæaðÇ]G‹ìôn¦Î[Yö® V7†1Õ¹ã“]šú€#‹ŽQLüC~´Võ’ª–ò\¿cuÝwôg/9º>ýqP¯ ¿H‘Ôm± ‘Ê¥Nׯ§ƒ¼c‘Úu“A9‘“9Úlyþr±ã­my+öA bráo.÷Ôg³ÍÛd”DV$îäR¨A¤Üh'DWk‰äËBe:,‰fYü™Ø>D„:W…å2÷"tþ}e+ÜÑ!QôðBÝTí<ŸãÂ*›{z®¨±Fs‡EsU›•Cô¾Õ¨ÔÎAìãáƒÌ¿Òy¯9¿pBgßãÚ 9³±;Äæ@…æŸ2 ㈒`÷Ri=凗Kso¥ŽBÜ8Æ~Èq˜ Bt¼²Ý,¥‘ ÅSS8 e×?'îö#¸c”ÆsLfþVKÚ¸·Ž'ÑH›×96 dßÒ:ÛS¾ø0JüÇU|ÓÀŠmEÔŸ8¬üjEÄÍ>-m†°¬òñ‘ày™xφÆþL³ Ô"bãax†„vF­Œw¥BÃ8JW¢†_˜e4ðRnýF®äÚýÍœ¿p±p.ÈžÈ ‚F°[ÌʘË÷‡=}§Œà´yË“>ŸZ< ‘·W  :#ñlÝ®Tði}*—–Dåë]BÒ·3½RºCÓ$œ1®÷kÈŽ‰µGk´'#Æu(œà{2ÊB,bÀÁîPÅ-±N½täå îbÞ&½›»‚~-nM/Ûˆ¢oèzê…%Eì2¢¤[ã3$ìI™ELóçR,~=,Ù )†àë‰t|p0Œí¹pºX”¯ÈÖEYÚa®j·è;˜[º/ è=IÃã^b̉’?9i6³ƒª§ê}]è°xôTo ' ™Ó¼i÷ž·dRÈ™¼!.5­¤¿/æ3èS¢šk@eUáÙBˆbjXÜ,„F“+.·['ö­SsJ‘ßúbÈÀÁˆ*øg Ê0Õ|üÎR' ø1dÂC¼ˆ\z'ѱ’Ì{aŽ2/B 7A+‚wÞ&½“+B—>ÅXêd%ïԽན5l.›^ÓÂF•ÔÏ3âŽNJäoý¶q_å®8aL¾—Ñd–þ ±w±ñVØÉ=K;q´7ŽÀœÙÛˆ7Ó›D¨¶áãû8!‘­àê!@ýœÜßéX×ø|0UçóúcÜà<Ù;s/!°„‹¾íçÎr"ÉaêDÿ‹j¨Ùù¬`€èýùj¢åð’#Rwºu×vtú¦fž=¡_KõpŽu)YÝ@º+þItsžÉÔPçøg°¼å~s'ÞR‡LH;m¤ií äF7Qô¯¿Ûñ™Geç£ËÞ{ÅÁ™¾# ÞÖføH,¥‘‘çÈyI¹pcÕâôOÚ6çî}ÿ›´ù¹+O”ãp^³(+njúa,·µ"S[Õð1«mn¦ ãÃ_Y8Çpg{¤uÆuâ|*]ÕpX¥ðaÜùtì„âÕccv Ò¦Ÿˆä¿5ì·tˆ¾ÖZ¯ —°ñ­ Ó’{ʈÞvŒsãvš¬Ø4„ ÒùçE·`ó¥Þlê>!¹ž_œTsûìÕ&Ïõ±DM?"—4éöÆlbû–ÒÜ3Ñ'$¨œH Ý,·L`£qøUsM1âêö•zY?t]Ĥ:dë¼-—-çuÐ~ž|€Ñ_f=oÉþ0à\ºœéþk‰TˆUžìP²óúâªÒè£Ìâí£C„ÖÆñµ“e¡ÈMw“c»IÞwâëŒ#|ÀôrÍÀÒ»ý¹‡2ÛU5®^8Ê«O"’&µqH5þ p¸ìJ|žéدâѰ~JµòPÄ?S»&m‘¸’bÓt< ¨Õ=…ìF¯%UâÇÚ)[CÔ?CÛ.<ÑžoT0Ø1ÙšòLT•Škˆ¢& E{i×Ì;~Üøgx¤·O7§zI¹í“ ¬›-Pà¾*PÅâòšÅMÐ)ô2Ÿ#Š«oû„-h)ù^bà®vSu:–êù™ƒ]º9$ªà6âŽJו=‹vDƒW€%´i¶êá…Tkx;LiÆUßi±}o2Ø\ÄñðÔó†ç0’¸£ýnVVö%ÕPÕˆ‡G2ø`ÅÞ¤ç1[“'ü4év+kWPʈüÜÁ„@ endstream endobj 122 0 obj << /Length1 737 /Length2 25030 /Length3 0 /Length 25543 /Filter /FlateDecode >> stream xÚlºct¦Ï–>ul³óÄ6:¶m;OlÛ¶ítœŽtlÛ¶m¿¿sÎfÖÌz×ýe³êÚWí{UÝdDbv¶Î*ö&LtLôŒ\e1e&FN6#= ,™°£‰¡³…­ˆ¡³ @ÝP6±01˜™`ÉÂvöŽfæÎJcªjÖ†@  G€šµ«…±9€ÇÕÕUÀÕÉ…ÞÑ…þŸ$e€³¹ ÀÔÂÚ ,¯ ))' —Sˆ›Øš8Z\Œ¬-Œ2Æ&¶N&TS;G€õ€±-Ðâ_˜œèÿ ÀÖÕÄÑù`¦Žv6YQA1y9€¨0ƒŠ0ÀБø'ÝÖÙ‰ëŸhc'çUG pþoÉÚð¿%£ÿ–þÛëèþ‰–‰ ´0v™˜YØÂ2ü‹7I[S;ûÌ@ûÿrýÈé|Ê8¤MLÿ‰v±¶–3´1P ÛÙØ»8›8dí€&޶€qïæhñ/‹²µ¡í?•üWŠ¡…µÇÿoÒÿ‹t6ü‡A[³h¤ã¤geÿÙÂIÌÂݨ`áü÷ÎŽ.&ÿ1«›üçœdM€.6ÿYÂä~þ9a ­µÇÿÛõŸ²þ½'ƒ¤¦¤ª¸2Í7Ŀݢ¶Æv@ [3€²ó?Ü:ÿÛðo·‚¡Åº Àø?Ëý[gú]ÖÐÙÑ ÍHÿ¯vaü×÷_’îÿD Ù¹{Ñ1±±èXØL,ìLFŸÿרÅÑÑÄÖùßðO-ÿ¥ÿ»±LLÜMŒaW—쌹ƒ-ÓšCË}EÏTü ™™2P¥Yb]žìºÖF±0ê\âÞ©àæ»ÁLñî\lKÞ5Æ™ÀOD…ë.ᾓT˜ùVo¾)¼Ö…ë]A¨áØËA\GÜ:g¬­*PO°fåžUK9”ò¢ˆ3E”µ÷þ@ÞŠ_™°§Ò¥@ÍÔú‚7Ù£©Bž1)¼2âÙ] ËÚ¿&ÃÑ7/,°?Ó*¨»´ÖÖT˜eú+•§½ƒ#²J¾åZñ;°(xwɲ ™¾á&¸t ó˜Ï04YøãšG~cËŸ_–@þ† *{áCº’G52nó‘—µ¢Ã'â`gÚõÙwJ]]‡Â!ÒßyŠBwKjN”6]îÀkØf—+wMzÑYÆa]ÀkŒs}tE•š Üâh&A“¸2ª52iîÊñ´ðÍ ‡×z›LH)Òy(wÎ^n ±Ì °=†—> …9im•J„/gyy dè1 Ôc„³3çO@œ¿6™rÛ‹þÉÙÑäĶ—B!>"lÁ’ÙÛï%ÈÆ/˜´Ì¾w¦bZAÆpîC·‚"q.@Sžá~Zq0ôç!@àš¡–¶^¬áC^Ú›_1Ä-¸¢Bx“×°Ç£^Tߕܯ5~‚p$¨gþÆÓ~tûÕùH>{PÎ{÷¯$¬Qû9U­{ mt9¿ ÔËŸ½ß@ìØb†týت%;HiTy-ÒVy[™¢Ù1\"Fà+ê&À”°ÄBgÞLÕÐ7>w"U}·(ÖIÁº 3›ŸT,gšãpšëáo\Š6šÅ–ñQ?N='!Ãä'èD²‹é¯ c-âzV5jšCêjð+B{óY­~õ¾bÄc›o_¸bb°ü|Œ¤Üksï‡ÍX?éιC/qQÓy\YD‡Ñ?pT¿‹ˆ9„}SÇêÍìÉè‹v)ÑäQ" Ø¾¦[„ßÇ .5%Ü냼$›Eȧ!©¨l«K«úÄÌ .“ìûp ë(Ãnƒg £mg5І@$íj8äiÄß‹¤O©æêcÒ¼s&µý¤ø Ûðïw«éÇA1'ÚU€¡ãòíslöTJÊ9ÓËæåí°&†(]%°Æd‘ßíÙ~ü2'×vëÀZðcÓh¬¨1ú‰}ë,!Ê–¹{œ˜ê êXÈ$Þ­mg9ÃÞê,“Q¸ +xol¸×áhßáæ ÈjÊ–Ú† ±Ñ”|ïø)õÛÉJýÛŽ è"L'²† I…®(¡6õº4ÌTæ§>œ¸›x }¹¡4;{séÓˆ@hÊî —‹[ ˆ)HaçÙ«~BŽ%2¼ï©`TqvUŸ·&ÄKö¹?êt†»{€Lî.8:ò1Ê ¦Sªû©wxѬmèà¢7øTF¶L@žÏÊQ!”γ÷2 c£˜$áN™V'Í¢)Ö%Gëá³É©œ¶0/5†-dwô1ïÌÝ(:6¸BuŒ~ðïë9h܇ùŸ¸˜='Ð¥aþ7Eõ<}„ðI΃娣t7OlJ _ì~w)H/ °Ñ#5îG¢Lò;Ïþ=Ê!MAáP¬Âõë¯Ö3ÄŽ’µ§D|}Ža»{GõóÜâڠб7„ØúG…/»ëòkêWæ.º6{Z¡àcúÏa:Év}BVØîçc;_Üø6ïlàÝZ#€e§Ž™dr”зRÞžBÃ{_Ë8-ö…& KÁè|9e©yÛXLßþ6Ï®wß]МqD¯æ‹%%ÅË*÷‘ϲ "`íÊÖã3ÛrVžº)$ª²ˆÓ¢ÅK_bL\^<ë S%xcT@‘©Â$9Î £v5Å¢PÑâ•â©NÂs”qœU†¡³¼W•cÁ`~}t~…Ä%z>ڹΰ äóë,•Hvõu±\@_Áõ»™¼u•b›sf¯«k›x …¸†xÀ.<€ ½Êx·o'ô„ðAlN´yÝÐ3ç€îwÀÛñ=äÛïÙay•),ts¯\6Û¶R‹­§±¼58"#„:”Ùa©S´¤o] ßïJÔfpÝ[Z!ð“Z—bÍf7ÞíX¿‚NÏ\ò8õÔüÎÊjì‹=âûÍ)‹T‘¨Y(ˆ)´_ª ~xØcs]w ÚƒkŠ Èxc I‹qj”zÒZ)å™o•Ü1J†!`±Çú»"KŽéOɳmÃÖ.M%ç'úº‹Ñd¹ÅU»¨˜‚(ŒŒëÇWûù îáY…>ëï:®vÓî™D…P8ëØÍÆ+¬i|/É5¸ùlZ†¡$eF„¹+m¸A©=”…U²²*^mâÀ’ËžÒ"OÙÏqÅÿbÚ×þ¼ƒÁË”îi&0þäËÀ i>3™\u ©ZïH;”îãN¤ŠêùFß.Ài’ß«ÍO.æÏå<8Q(£®°=Û}¾3ä}ö+ŽÇ.0‹IïÜÀ wSaé~ĺª…”Ñ!çIø]Àù—‰7{߉@ÌÈ9C–YÛ9[Ìò#–ÌK”‚Wª=jô<¦;Þ22[gÁâúc„íªx¿ÌÍrY/sy¿ý®Q’—æjí@ AþØ6‡QÈ~W±„ÝÊÁ9•ÌÀ›Òù'µû$m3ôM›q69|¿pc´ùo¨yY騍ý ÄpZé}ÈŠq-uoi2Z£² h"z?å÷E=/É- éšhm€˜¦š AòÉA-’­C|šœË1 šéTãOÛꀟ? ÔµÉâ…ýFzDÕuV1 `Àä•dßä'Mj¡÷-Ô…izÄ­æâ¹†¿åÎ’†A úܦ¦IèVâÊy®$š×0£Á`9÷d¶cçM'Ô¢øÍ$€ìUï¶…O¨”Ê©=­s¹2ÇŒ£²‘ð‚^ÂÆï ™ëïß—8GÈ"çq¹‰†øðHVO¼©;¦†ÐôbñJp¼‰W™›Å"ïûï@/Ù|ýŸ†œ¨OzÔ$@¾òÛÇ.‹ÝXöâ,VÚâ7v5ï­3ݰ{ !pfbæø¯b¦{ˆA +ûp¬²biªÖ™`W=Ci6/ U?aÜ¿HÀ ½,òôÉÙµÐĽmµz ç œ u¬SøŠÂœ)e¤)T™vì@ÑàxÓäÛú36Ÿœrñ…ã<Ž]Æ?”FY‹‚¸—l:ž5¾MI\ê#+kNxvœI~Ô³å~Ê;ÛÅ&QÿFÑñïÌfŒWå܈T«¿c‹â%¨ýÔ¨9coK¯D4¦Pf6ëR›µâ”ü×O[ –©ËIÑÇP” L¤?ɸ"Ö·üªŒ«w1_û½WâÛÎO<¤¸^r+j~¾Çú2+ìëf8ù//5»‰ÃYL™¹AÞ*½üëºâ·xl«nxšÁä¥Õ×"¡·ÏÍ›6ÃB`Ó`,wS¤o–`½¬½Òý<ÕÒ¬€ÙWa™­ËAyÞÂ)iCvžìÈôBa` Ú± ZÊ(fW †[Ùl‰)íR—;»ÿfHà´è– Ožõ9ešïæCr~ oÓžÚùýsãjæ+ƌ½D 5ÿ:y%~a;æú2nŒöVîÂÛMEmzï,}è –g”àS8@?-8v._ÔP}kˆôw—ÙI¨úÃÍO©ç'InÛòº _›@ÇKPù¿Ë#s:ë7DV–§î½ðÕ"0U{{Ì$’CÆCA)3ôZ&×ôÚy[ÐSÄŽa)”n`àœ4²ajú¨ÚqÓmÅŠ*’6LÀì¡ Ü¿‚ªéì¤EF¬ôÂ-ªø"‹¨ûÅöpÌ×ò¯ë“vµnËhß©¬Š`ÔíCÉf¾†Izþ÷ ©ä4³ð•e‘Õl›9‡»²ÆÃÙ#p0L¹ m†Á]ë¶gNYx6:=Ó2a˜|g»³ö{‡ }‹Øct`ùøxºéØ.Ìs íP nItïA'4g޶•äi'ÎOùÃ_º}ð"I²}2 >c¡³@5f¦òÆÃûµÛ¨m\!í„33\ º×!ƒxäJÍî¿  þŒN»†¿â>ܧuój0Ÿ”Ÿ=\×j +P¡»àðõ‚m§ì­õ-슚Z µ·…£ ùሠ®sÈÉïµ ‚¸Ÿ[‘EKxöþ¨É$ä{… Ï ’ P%¶yLuö…îçû¤ª=Ÿ/M˜Z*w­F/îÔWœíWÑ.ÍKƒ òŒ%¯m— ÿ­eÍ…Po~eÆ!ˆúÕ:N©í:,áËŠÑj½j†MÕËYøâMŽ>¯×zæÚ¯4 ³EØ«ˆá1‹D9¬Å­ÿY^wíGWîk¶êC€éÀÑ€?OVq{®¢ÝŽXë}Þ_[¤@s•ê•rÝIø÷šÜÄ©}YS#ŠKüû=Ǭ-)ÓD›Á^ ÔRƱ¤` /©»ÏDxi6ÀÙºz¢®O§–_©DÒ2Ö4m01ç³áS‹iJ`ïòù®$¸à¸ÎÖ¶} =û¦öÍã@ØÚ8ŒøøÚäcKA>)DN’ŠÇŒóÃ%Å3èU®2$ Åy‚;¼7P²¹M ¨n=ÃÇ úQí^˜ŸŽ(¼´þÐ3ìÄe-­¸;Î'¨¿ån,˜Ò¢ “Wƒ¯Šb*gš(-r€½öy4„``TÑîz¼£=Fä"…`WbFžÆ%õa …ÖW¼8#W]§¿È.ò~’ÀÇc É?¯MŸ±½Åá2y =)µ™yiÖU©Ç)n)= ÛÀ=ÅÓH>ß±Ùÿ?mUû2ÈÎL˜ÌÙæõ:Å “»Fý54Ú‡'U,¸2­ÚÈÕöVå0SXüqËG!J—^:ïIâSjO˜ß†°ærhO}ÐíŠÁ½·Ó ÜqêðQSÞyLAU0èµå÷‘B\ÄSÉÊòdìWpWâŸÊ¿GƒBx˜Û³Z´ÂÅ´G2Iú&ᇰo JÁ²o÷ë¢×ÞÁñê1λã†ÁLXO§Éᣯ$«dã›VgFO4LDb„é–™CH7'¿9Æ? 處éd—È®7K~Be¬ðf'(j\uµêëË@›Ó9Íÿüt“÷„Éæg\-ÇÝUÞ¤ Û~ÅFPD‡†Uj§ÎÏ]´Þ\çŸþ1¥è¨¾¬‡Öô¤ç ÀNóÖUÿXÑTÏ_L®y̶¯Ùƒà4:(Œ§R(²åßôÜ8i5uR¢ÂùCÅÂC4=€(g„÷¡žYÏ “ÎvUÑûE^ÿèÙµYè=@Ô”†Ýò£ÁÕagªj2p<Žè,;ù¥A³G‹“Îr™YCÊÝŸ uq•©úf Jô%úãz"v†l%{=CÇÒ<Ĭœ‘9LF|¶O¾ïæ›ê­¯†Ö-°‘KTß‘æÉL‡õ/Z‰¸e-m(ê­h…ä¹Ñ<Íæ˜–{ò²½8 –jV· US(ëŸcšk’íà]$×pË,„Q/¦ Ÿ;zš˜ $‰¶-’8¬ Ä› Q~t¯íß ‰‘¿MqÑ»­¯- ?ÒŽlYmÇRÐæ»â;½úûÆ“ÚDg³0KÌïB„it«\¨Ï%ªK«Ðóë_(îZ¤ iö˜¸¬²¡ PG¡fž³)ñ+rÉ>Ö©-ÊØak—“ß9þ â¾ÊÌËAÛƒ| Í|Ðì}Ûw¦{ ‡Ì°ä^®±Ec¨ö‚R=ØdÖìÐÀÕTj,`a¶!|I´¥ÕÊùj(C Я%±Ù|—ôæ¯öw?bhrï…z§CWÊò¸y4—üêÏ ]ÑsÊ"É#HÁ­º„Y¦×˜Ã•-òÅž®œ²:T=º1Íf p”þ+L/+ cmi춪ž.ŽÐž„.l±ó©‘'Q’0kó{æÝY"š‰ªâXº¹w;Ádꇔ&ÊT§ú§˜ þ‹Ò=zý¶0ÆÉf˜ó ¹•6–þ`ý廵,I_1<“ÉÁ¸äšåYuŠR~Ô'Dîøƒ"᜖ÎAž¶i @;Þ—ü8¨t—óä%S_êM}K^NF9¤%•8:ñ%ÅìÍV~5Ä<­—j×ÄKŠç •=öñPhVËŠ¶Âqˆ{"Ex…d– :g*^Å<%K"™ñ¡½”qô¯)Ìå+sÚKi'”Ëoy†nÙù‰ðËlG2B©†#PäåÏ)ÕlÞÒ0Ê¡¼Ô“wå5ϳõºdÏ:nHûu5p’›.'èSA÷¾±ŽB!³h•…’hj ‰ Þèe[RÉ‹€Cüf÷©[ Ëú=œBV|¼dóع€E¼óÌ*4*H §R5`Ñ„”C<8—h®n‰bõ¾çîÃì‘Oðýð\ÖºËò™eJÖ–@Ì®À>¦P¸éj³4”_s”/ %„âᢠÐĈҷªÃË!¬‘ÿýEôt¿X5Ûâ@lðôT\z /txé)‹wFw“´ ÝÒl 6ת¿EÖßÏçÈ}rò·ÿYšv&Ò¡—yó}°Ãµð^ ñ ÙðþeòDEd•NA÷¶Ù¹#ªb˜í6Œ`øéåõÒ )~ØÞ¯ê¿œ¶,ò™J;ßH>»,²EÌ Fóy³mÏy˜ç/>žñBS°§qžºhQ¶Vh*€Ê‘˜Z[¦ý=Hû¿—w‘CJÈý&t©ìv¿úxâÎ`۔ʊ4ðå† k¬ ‰é2ó”6…ü [ˆâ˜S RF*âî¡`жx4—þAÁ¾E>¾ÂÆAÝÿ0²iφNW=8•c¥–,q’ÄÝOé  Âpl9yÏñgƒÝ%¨¨¢¬Ø<ûÎP±¯=| æp»,­æÂk °~ä}¨ÉΕñN”—›õÏF´¦úÉEÄ;p÷‹#¥àt29ÒÑsà³Ò¿Zµœ# àÌí‘E|¾‡-5Àä$-/ßi8Tq3Ch.)0Ô×37Ô©ßBTZ"ç7Â#pPžSíÔÕf ‡$šÊE猌(þXñØtõ±RPŒ¹uìÛ@œØˆ‘¨oì—Rt˖ל…ðíB`ø»¼5Ëâ!AµÍ2¨s%Ê®Ý-0WµŒ($ž‰.¶@ÿDû Î3‚çG¢nÇ`mææf¤ˆßʽEV`ïšÝdõ-¬?pÜ‘~zËÏ:!¹E0l†¤$p)¥|}F¶ÁÃW›ýóxC§ŽLi{IÍp’¦­.ßùí{WNvƒî&T˜c÷ÈÐ~ª~Æ!\Þ´ñ²:¹“ b‰L1–*~Eߤue­PIÜCçoêÖâ#RF[:à˜¹è–Ô¥<à·¹Ì×êD¼Aç‘Tàrĵ4ÌQÜßU*˜{r…Þ¤‰ëG=céó¬ê@9y û*§wÌÑκ©©|ì‘d!}P\ò ©…H­K]XÄmè?›í¡Tç%.ªŠôFGB¥fS«^¯§5Ù±ò‡U z^¢yf-£jO¸‚¥Í6¢i“QÂf‘­4 ÅÎÍÆÿ*÷ õ8 LA£kož’[ür˜l•ªb<`Ü(Ì»µyûl£ÆeÇ5ÀÁëRîK}þ›¾Jÿ³ó†ùãB©+WY”©¹Š [/¥´çÎ÷ÿ´Ô F{‚úl^^ÛR­Îù[«Ê&@4.ù÷n§õK‘W¼®ý6V¤Y~W¿ÈõöY²Åe†ÿ• ÀÄð–Oœyõô:ñ} oíá ÙX \‚½©Ù<†°€¡Zºy¦àa™;|Çy4EªQzBÑzfLh«r'–L©foO„R{x<ùélËÐg¡Ž^¥š.¶yü‚ìâŽtv}•g¶_܈ñ¹"‹¬d€÷½€ÿmÄ ~ÓîñõòËùŸÒŠ¡ÄŸ-×;nc¦½V‘âà°šw%[›ÓvÙ7‹õ¤J‹È!Øh¶/uÿ³šá]†½¬€ÿñ„åò+0}ʄ챦Yë—‚UŸÓ7C¥0™MHÌ_¶ªÕüdBZFN ÃcùÐöÅ­»žÓáRâг÷¬~lp§Ì¨Äu<úKH[óÇ)ÛÃ/»F'cÏâÚÀÓYj?rÿJx2£ŒÛc>e¥U×Û? ¢V©Ô¸øƒ±~c_réÊ÷™qr#ÌÀçN¯¬5h)d!¤Ëõ%m;æs•ŸðBŒ6øó±vîh˜&Æîl`j‰Á#⑸µn¤÷2%o:jVø@¾¾6¡Ì錡³Xà‰}¢`Ýo61Å%ÝóÙpÈ`h¹çµþŸ§Rqö‹ÞS y¦žj*nó¿Àëc¶LÂÌîÊEßF3ßÛîÁ÷"Wº¨&Ö=Ïk°¹ S°7DQA€²ÙZAÿË¢b\œ{­™v\ßï¨ä·Âvóóyo‡©EUõé #ôb{Þ”oÒ¸»3$,Òêd‹ÛÈXAÌŽ„q¡¨?žÒWï‹ I¿ð´˜é$L÷Ê(‹ƪU²X:$1ÆÔ"ñ©z‰³åVZŒ-FÉ!“T=Ó;±éØØú†ï¾2é3¶­¦à_•‘-P«ç8í™jfl‡£ÙOÔ®Šù¦&_÷ÕáûE3îh6†å݉ðQV÷tO¹'ÎÛº¾óf‹sk¹æE AqÓ:&8£Õšw:¯½˜m½Y¥ÆL½,¯£öTúœoø¼íÎ+‡Í!³}þÌ95Áõy¾ìØäSK1Þ<²Iïê-o^&^BK‹°~¦;ýî^^¿EƒÚs#r çoå/´.4Ëz/þ ` ƒ„yoêT_Óò§µ­°Ø” ò€ËÕD‡Ûd¦iq³Çén L±ù>,Nf;WÛ°!É/)¸È,bœâ î&'Ç~ÖJæš"*^cºra[û7‹0½“íW…fúíBᡨ„(ñiú ”lÈ/\´1/|>|ð”ÈBf=å蜘 VnŸGPk`µ#/ïøúž4÷\‡ïúGˆ$ŠèbHc{Ó­-_­õÚæ 5UãXJZÈU-vp±Þ‹c‘ÖT“jP@Š2ký†.6,w{YsE©bå—?‰ä[%RñrOõrŽø¬ˆŽ&..úªõv™ºqZx¥]Üi¨‹BOç}ª»˜i®îσ¶ˆùÂàcÅ96|Hï™®óÓ½Çü­]pÇÎ } 8u' ×·½Ù9¿Ùú5æÊ l”v¿ ¡5ž_&C]¼ñeVÉÀs#Wz5« Äû®AMéJZ¼MnG'9WðD¾…™d™Ã^(ú¢qŒùdÈêï•“b ãu_½ZU‚`M «Ymd¦ƒSÒÄ»+ˆÈv[^ ð²½l‚­‰Öyš¶xöÙO´ˆL€Âfÿü°–*½a⟞äíu@\&$9è¥î&ÉÎóíÎèl%ÝÔ¯”óAI™oek[xœSðù‰%9í€ìÕ»àúýT„Jì%ˆ±xUˆYéJô"Fm^ q”M ¹ne’ºnÝç{WÏEg3-{í$i|À.Ô.„ôiÉÖ¶ÚgUºØÆc.•5(À6®¯¹Ìx7D …gô1óøù²¼·FÓ€½R³h«XboNýHóàÌ3zvê)t6N]‚#‡ßíÊWâÍÎõ'ímÑÆ‘‰‚ôÔ¥â®GSwlÖ|ܼ ré¸á£K@C4ÕÂJÜì±{È Ô'²ðd}ŸÁƒvLèF*ØÉÀ UaO0ì)"œuG@×Ñ™{ñƒÌùkˆ+ØTÜÙÄ•¦TdÑš¾2>¢ÒùµÙ?óÉçßú¤ó³²8LêÌZN«(ðË}Û`:c†÷(Ì%ÿÊ2Ž šØw¯)É%ÝšʾÒEÅéɳTjBt¢q®Ò¶B][.(þŸù”|W‘ºÖR]ÿø“Éhš”a (ÄòIÄŽ¨Ë^1ë?ù Ø¡³©Zw-ˆÑ:eD”VX’m–ø‡¥a ã £ ­Œfmû!mÈŽ³Äö k¨vwô™9<ìçFöÎR~rh/AmG Jªc£dBò¼”ŽQÈøúîYÛçC4ÇÈ3פ€™»£ñ½Y©‰+¸làêÛÔV£Ê°Ùn[béGÙ-‘m#NÂ@=+jQïÏeóKÝ÷¿îã%…k5ño°¯ÚBOÀÊ…Zz&*qL¹:”Œ9âWõ¾!„X!ê¥П@<§Î„ÈnRµåÏõÑ©_øbóüj#ìVhñE¿š·>y|ýº©×n›_šúÄͳÜ6grȈ%ø! cÅ*0<°ïùü œ¿Æý—K·!”ß™eÝPç•ô½[éã±té6¹]›0»¼¬¹¹uRÝòô)" ã?LÖö/Û¶öÍ*xd„òHTÕ‚ªØ'!îàÕDö°Y1Yc} ƒÂXÄÀ”hÅ±Š»×ƒj"4¹ ûvdù´W^¡T*(±uGÏ–]7†:ê æã°Ÿ=™æƒ4,fo9Í“ØÎèµß¼>s*NU© ÂKœÑÔ$kXï€TºÄÄÖ IÙ83ŠÄÅEŠ1tdõÜSS#SiA„ï©*JWCïb‚ÓUØá5DiüV© …$+i>(/åEèàxt?œ­KÑøÓo±Hf—ÄYv´FLäwIª‡Îëªháx|˜ ¾îÝpib݉m„úDZL›Ÿþn}^4Ó¸˜!WB}[Ù`¥S#VïñZTF¸Î m6UqøêF©ûYJÂr¹räO‹G”>2p³ž¡¹.œiΚ¿nž-Ý;~ôbêÞQaûóûûÍT±ø&X¹jRÇû¶¢/2ågxtü@«Æ’=“9mlD¯n3L±ÎL/¤©¦ªÌ(T´›ºWTj™Q°¨Ù“Âé/ Í­wn¤YŽG"x¼[aí_ÿ-’+Úî*¯cå<×$ËôT#M1’ v^µ½´!‘¶µŠ\¾” Ž,.f nµá¤v#‡ô ÎþË8—üö&ÏŠFï÷’ÆFê¸u2z&?Ž• ë1ž²Ìoƒ*Ô‘pmiS“ñY¿þ )WõháRä‰nKíÁŒŒ˜ð÷Ó¾üÄÞIÇEdØãi5òxaCx?ó2š?e¦) …ZÍ‹cITYP·nµlµýå>©c»T=™§=ž°àg_*º0ÙÒSÿ¸cq°ä¾ªÇµG4àv7m®˜ß9®.&*=C¡8Ûb£¯=ÈO&ò6”dëS3tQê¿yzF_ W@ƽ"ÙÓÏÝoA¾®h€¹d+ˆ2 ¹O‡ü†…£C —[Ó%/4 ª5qB³Ëçõˆ§š™@$ÞÂÌJ> èÀIìíâ4 5àû‰†O§‘ÏŸé3Ç¢—çF))0kG›­½ YcŒXzJò¸vá÷‹¾Æ#{öc²Ð2 ÑÌ59‚´ŸÕÇïHÜY‹Æ ¨E1,ƒ€Âxãá6ŒqÀ0KßǬ[<­êÀeúD­fî~rr ŠÂŠÅÖõá(V[‡O‚EuZUÉDÿÄQb¯÷l Pé,—°øÍoUî‹„¨!{´PÌ„®ÛZf=\hwœÀÇ9¢Ù:ª…,$Ûcwíø‹¶G±|îfïs=¡M\*’\ÒF];ì!Ò!ý3»1àä@»¡§…¹¨Ôà²oM”{ó‚xwÜ0ï‹«ÈZè©…lºŸRuøƒK'-—C6êíÐöxÉl³*;ø„óþbÂïÆ–ÒÉšïSú§b!’ JšÐÃ>%î7ë”T‹,¦òû)™Ágßhμ™@ †ÇÑ+d áõöûrECÙ&ŸÏ¿ñ+ Æ)’¦âˇ*Q”ñÞo‘Kò§Io*å_,8³Ösf[ {úP‹E º5·r‹ek>ÙÛ‚ž®v¢üWP¥ÖÆt:|0zhd‰¬ú)±?ÎÀ‘*Å•ÁžUš¾pDß7²?VÆÆ‘×@K'ŤÐ ?ÃÂŽYIW.ªxJä+¿žP.Ç©²ÈW†–d= îŒá=ÇÞÙF8½‡04¤¤µQc:1ȳæ»H]âíÞ;Bò¢öÿ°­œi’Ýš<å&Ü’Ü^~ #®äuÛÎêúr®î*…Cý/`¥ôB¶uKåö^&Ü^®} ·òz ÔFúÅÒ 7£4²ˆËnwu’4M×óÉxÏŸ´J»2¬Þj.xY7¶ª‘›h¼íózG€YR''öók% /MˆP%¡ @1ÁŒûdªæ!ëÿà0„i±”(±E¿ï+¸ÃôC…мyDA­èlž$WÍ|>¶6ä›ZÐÇ~÷{»s™®ö)Ü6Ãw®·p‹²quá:6IVÕOE~3&s ¹Å MoÔ††¥Â¡Š¬fTð!·ê«ÔÞ¿48ù_GhEµz°;öŽOMHD]šùœ?ÈôÃ˶-ô¬é'Á™£}RÜŽ©Ž?vëïrb”\Ë>Ë|#Ù_Ö­ “P%3¡Dó+.¨¹º¼‚!¶B­˜‘®[šl¶¦¥„íªr=šÅ‡/pî xW¶³ž°­Ïu LˆG»^Ô`ØkË#$jÐäG°?¹Ô#{óuÒià7˜vOÍr\Ë çQé[LB˜  þ.T¤?üþ`»| Ò/ZóÄOO©à·ÆÕh|u"è5¥ ÎõžLûù¿‡Þ?g)9Š9«ìõCn*QÓʾÁj}ï^%¥|ü…2i7gݶ—J€MŸÝPoÏ» Jôä>õsɘ3ù2K žÀ~ůó%àAà‘Ë@*劕ÜûèjQ:õ𔄍íP8‚oýâŒ%O)îJº¼Ãß+U1ס0þÐ éMkK½‡¯×“ý‰žfõXÌ 7:™|adÝ#økŠŽþt;=¥à„[´éGlÆ©F™†ê0d_¼×ÊôÉ£d/H¼ç4’e˯Åtà‡’lg³YŒ{ ›Å ¤Mÿò5/‹jîM~ŒW:sIíIFÀ sA§¦¿à<ÔËE×çÄÅ›ù¼´ ‡É Hì‰<Ñsé©äZ‹ön4BÛáøçM µ ¨¥Å”å*¨›®•~Ö´œþð[4ªh;¼8Ì]‹8¼ 6éƒM{@rÞ¤ÉL*¿«ölRž¿—FzýØ–a·@ò‡¼NPb„ÃЬI' ãâGJf.SÏmH½Ínº”h?È­WðkË ñèl,ny[ÿ7’@F]ãî ô;C»ô({íþ€^ðÓÑÜÉjäŸ'd£Ÿs’Å ¨Q°f;bž<©—ñþn˜Åº…–q ]] ›#¿åº'©%Bïë< º-É•NùÐÆÈ¤»T“'}90[!ŠÎ<(\꫾õò®øè—Šh/´PhrŽàiÅM&{ª7¼áX &‹æ`ãéëƒ-h‹7Àë’äîÁ‡;\è9~eºé¹9¸÷+$=ý›£¾[LóC® +BLŠª¾ÆºvÑÕ™ô)Hydä ‚]^ÀÁ1׬·+‚Eж!UÆ> AúºyW(ø²Ë›ô!-éPq’8QZ¯àŒ—Ï%Ù\Dè°çÃ_aJú»2‹¿PJEˆ kÀ¨OÐQÿæ8=>`öxT„®IfŸó¤?øØmz=›êjÏw4~„„(XðËØç°.¦ò‘Eþì”ñ¸‘aõ |¸ˆ‹i+´s§¼h§³Tc |„ÛAX‹Gáh,jy/“¤ðjG¢Ž^wbUÐuäÐ̹<Ò˜Ç?K}ƒfË DÍ&aÎ>­g(¹öú] Ù&T·¤Ý·Ê¾è9—­!2¥ ·ƒ?VÊsöUtËçPŽríÄ9×W·&â+†ÎõgeGJKÌý¡uz™Ûžï¨VÞG`ÿÞb¬a05±ºDã«”ŠŠÇº?‹[PºV”+©ë5-%+Ëöã46åNƒ¢ÈbʰD<3eaÀ4çX8„ã¶Ð¡°6€çV·ÝBêý¤å®â“‡{Ïo2ø…Òi{‹€Em~¦:Ï” 6DFp¥÷5[ ¾"=E6Ø7þ“¦|š„qqrP}k°?KõtÑMHœ¯ƒÇ#p¡M%nLVFÛÏpëb®}RµyÕ`>óÙò:2›?.;¨”äŒôêïY4Á–‰³ççkþ/øæ>Á)–Ë~qØRÉ7}r1?AVÙÅên]N»¤¡UÆ Mtñø&ôdìÒ´ÄÜà˜/—Üž)ôƒÄh‰Õ…ÔÃz(jhBR¤"HÚgF Ù±t¡ß’ƒ—êƃ/8».Í,¼£äg½”;ߊˆrõhÇA¨î1.+¼‘NŒ ÓªµØšËã;ÔÅÓQ9”ýÉÆïÈÔNü'íòmŒ 9˜Ë·žÕÉ@œúîù:sG‹ /»ÏÜûº`„™ÓüÈ€ wÿ/yå‚eØí‰;ÅQ=U1ÑŽƒ¹õeç–¢\0ñ>‡|4¥ìC0gœSºNØ…+]vªþÁ±NèÌLoæ÷ B©¸0ñ‡ãf;(O¹@4šÒ;aå×E÷Yk¨nŽƒWÍ(ؾà|VáÚÀ¤¿byP³ë(E—O¶`àG]Êï½¥úŒNÓ©uS*m±&þ^´ÞË-ƒîéÍÚQ¬ÿýhúÔO=–—~ Bnµ bH\Å‚µMlÖýsëÜ%b-§ë™ò7+ ß#Û)¬<;G ãMÕXÊøü²õŒyµ;¸šNµêºmµ¯Ý¥¾:WjµYßµG!ÚŠ´EÔˆ®¢ Iœ¿6Ê\Ã&P¸“à»ññ7=¯åq”ÕŸ~hì?'Ë1S¶¬ÿ€S{–î½iÒíŸVó9Ä.Kµ‘jÂX‰‘‹¥µm Ђ‚ï=çôJõGfC®JH»¼æ0i9îÓÛ¹p†¶G¢'[W¼\L_ ¯+Âm0ì7È5n÷ÏøñUl˜ß…")vS\#ìšnˆ·aìlòX|6æâ1Á·Ë}•V!—Å™Ôñ­Œ&’Gs’+$^Ü~²Ã ¥ëØ”Bÿ¡Û¾O$å`ú fàçªMX  ¶ 1ÆãÿÖ_ëûýI+ŸÕùYŸ¬l7CÑ{ýüä‹Ýõöyá>Cõý[†ç[^®°$AÞæcßJ7”ÑšŒëaž3Òy®¡öU(ÖÓâ¡Ìª<|é—òñÛ£¿¹W‰2ìÞÆÕhT¢ç]¬KúõÝWkT_>%½À¯‰Söyf_h Ò啇¹¦,a}.1›z½D{TOÖ;sLõÔKbZçÂã¹÷ 9Î\î´†ó `ºw<¥›H¯Ü§ÝbAбŸ­á[ùœˆÒN(?®qæLݽvY<€¶kåBŠM » 8B1%¸Aš¸¿úÚØB$îOÂ&RlØe§ºÔö ÍÚu O!pþ&#©ÉÜW|¼¼2x/ýÍ+W+S°„9=2n­bW ØÉ4ˆTDA%üqf£¬ÕU(ÓEI©šÎ·7÷dmbUaœ+Z¥ùAW¡ˆ […ÿ–$ß\o8@(¢ï&4VC–dÛ¨ÿt÷›^Fl”-ÒvF MY?OœJV"¯Jè˜.n@f=^8Rë 1¡'\-„¯Ãkº8ÝÍ%Iåýúøx•*&Ï*ëKÀ”r{U{`øÇŒ3„WÃMgºÄqq1åËÔïKãf” ýL±€ö_¥ÂpÚ$EZœ†=ÊLpüP"Øs€:±ëÀÐâ±,7 VZšß>ÒÅ9©„Þ•-À,å[ßjªU,¡T£yl6®ý+¢—*'Àww?V]áàÒF k‹u¯uéÉÌhGï¬{¡ajæ°ó)nE¬]6b’ccsÀÈü•¾ã::¯•Ü}Ž]ƒ Š¤î“›³lð´õý0›Õ©<rº]sn •ËKÜ#ÄiIûKÆ|òXÏ t´¸lIÕŒØ\s'|­ÃÉbp-÷÷äž62º1 Æ0Ùõø!ÌÁ‘Ä[^³}šÃyƒr«Œ¼_'=A¾Cè&òŒ»'zÜ‚8Œ :Ÿ=']êöº«0Ÿ:À‡ê¡k¤ûLË3ãc,¹s·D9†€*ÍÌšü’2(#.!—Oa߸Ùûf¨¥ÒŠ–$*TÃIµ±%dmìÏñC»¨B‰ñåCˆ?¢&QŒWLfù¸÷Ÿæ2âo{ A°¹ãÍD—ônü4§&-ùÒœfÉÅ”r™L§wú3å¥tAOÒùÜ'’Úƒï‡ VaS¨Õݚʓ­¥ÑÆ N·-mm^3jw„ìï´%RâÓGyIFp¦Ý/9¼Ñoqá(]ÁîgôxHçu¸ÓÌÜÏþã!lÌEï’‘`6û]oD!XŠ|ÇŠg»lÊbü… €\”×OÎÙrëÑ¿19áKŠ=Cý‚jXkŽ:\ÏémÁ8¹Þsí¶F¹ÉõÃ¥EÅêfÖ÷k°@.lÞíᄄ>q´åŸÚÔ¦äæ–ª¾DVì¬ls ¾MrÙòýî õMäLë” ÝÝ.ª"3Êoes%³†´ìÕ_±ðÖ(­‰œÒÕŒÊ^þòV¹¤£ ö¢gš<à~1GSζ+=d„·¦‰eoÖvzLãO»iñRƒ&Üùªâ$F¸¾Î]ˆnÈTŸ„‰½ ×í¡fÖ6@èeö ïŽw—­ÄéçˆøléG uE¹±ÖÏ—Cð8±ù$ìõaÕ—Ì>Âþ°³³–¸>.›û™cŸx1Ë…»Ý=Ξ‘{Ú¹1’ÃZIÄžu¢‘—d˜c„pÏ¡_?T=«hx^Âorа2ÍU“÷Ò]î"]ú1¿^ØK¢”žnò»ý=±»¾–è¾¾èÒ–>•¸vzWAŸdû¡‘,¡ß¢ê|hkRHoiÅ}¡Orâ!Fà ¯œš·J€šÁþ@ÛL„rEµÂ¿Å8^…gqӯъªÒM +N\ï,çåWÚ,lØï·JûeÖ%d=o¹·¥üˆ0Ó-oYí·NÃEc34²‰Ü•L¤S„ãŒM¾WF¿É¯ Ì’³¾O„©c·¯5ŽAäLÚÅ ÁÂ}ÿ¶QÕoÏ#× þ=oÖ´WH_Ð "áAëa‘÷«í¹å€Íw.àä{5XWµ®2UoªŒ ºÈÌxÀ>A| uîMH†qüe‚ÔiáP6êsZÈGÓ ƒ\>¢ký»7Ÿ—E³3÷¬Þ¯'›bdÎe"·Û¢~@'!§´+tÑ’^de1ÔŽ¨[Nnsg )ÀújM–Åú“E*P9¯<ÛEX  îa2­0V%>¢×%¾ZàÒ+ü¬¾ µQL²zïz{XqÔ° )EX¶rª)ˆTÏÈdÈáŠîÌÿëwX ‚Jþô3gYŠ®P2³°˜Ù¿+†ŸeȽyp–d»¤K¨`{–ƒþùb{QÖµ+» s™BÚ÷}–ôÜ3Á?Vá™Cîü§nyoâ/ØåcÎôbb‚x_Êß7JÓÑðdžìàÆÁ„ðÀ5›"§¨¢I5f˜Ë$ç 7Ç» ™~ZcªáƒÄ‚*ÞW`{’]\4¼Š0®S°Œ£ †“Ç«efÏåÄB->婆±-,š/6Ù-f*ž‚ñ~ƒXXHÒæÖŽ²û…3Í€ìȶõÿ#ïÜ< =óu0IfýϦñ¬NòúÜ–D˜3Q/Ò%åÔ²L¥;¢ÀÒ—ô\¡kß>§!ÒH@:õcmhõÃ3ãÚªåIN£óívlÑ$ýhÏyði·2dooÎV*1³?ñ€™`,h0qÏ}Vˆç<2 mÒn¶Á·J ®¬YÌÕø,hI ޽½³ Ë}ŒØYè`?ÿ÷ ¾•J *¼žŽ+^}STù/ó©iŠèL±è%‚#8c<zqüÞ¿€ú°Çÿûˆ•´º3¯Þ;cø © Ý—DØðò‡ç¸"¿ˆÑ—8žvΰ41ñC^ÖdÖ™pqG‹PgËA÷cüÑD_S¦žö^JÏá1=x<+Èa¾ŒÚEâçfÀÍ3!~&BNk¢dœ4®ûÑ[ҙ߱­é Ô+Ȭó?£°Zí FFŠi‡RM¸âÃ(\+u>{SYû¢—!^÷fTÌú†8¶— ©e}¼~üÚx«}dÁI PRPðò€Äg`Þ uYÊ`ê‚Åþwû?öswQY¼;Ve ;ì¤%­kÒ µ¢äÄç¨ÁV†ÝèûLÅ“¡,¤îçì‘)òÜ)e•áýý­äË¢€÷3v#³¨ý3¢­žt¹¾¿BüsŠ-ªÖ5ТÎÑ ¤ùãÜ¿ëgŸ¼fâò_¼¥Ôò'Wÿ]Ö¹gj‘tèÆ[áàƒÏPQ¡¯X/Ù†¼-”¶ÒË9`ÖðÖðð…x£ k&Ðf(×%\ÁÏý•ø§Ô^_ÃŒ>:%©Ë£ øåxŸ{#4€,d¾‘ÉþqFé`SÎ'‹ð ÕäR/ð³*ç¿– ®ô/Î ÂHQ~ÒÑH·©9áû@];R36ŒDG©ŠJ°V;´®`Ëfmœ‡Å;ì.÷2•+™`tä¼aogõ ÄKÑmÍ·:å©Ý †û ‡ó|?ª“Û͘à D&;.ÜKÄRå¬ã:3üíÓ¸|öþ‚9\Ø!²,ÙOƒ ÍÕÑ ;S>×m¢Oìdª[<|­½½ Õª{ãIö¸)ÀŠCôk@Ùr`¶·ÀCn`]¨B úzì¯ñ×¾m&P¬¯Sš©E¡xÇB1¸(Ow&Úsgpîò¥t΋rÚÙ!sÿ7y¡|M’²†U¹EÆ\{wÇ»Ý'm{’Ôm±@™©çƒ4Y7” Á¼æfÔ­€þÊÞe3± Ö¬7ÀÕ Èþ®;*ý°XŠŸíXWÜ$¨p•ŽRÒ†–ó§’H ³ÄE!pe¼²–ÞÇgRJÚ<Í.àAÌA(TryeÇP`ä]Çfâ”SÛ!. ÙJL…”z/(¨²×™øX;Æ·þ°Є?‰¹Žð•ÛŒ.6Š¾Ì§&¦R¿l=ºM;øëÌhŸìEÙ€ÄÄ+¶¹W÷âúÉw¶Ô´Ï•ÛÛ“è41Zl‘Þã Äb8Üg ]KI¯]x/¸¨ÎGŽ%lIóy‹ap‘–8"kV7¾·YÑÊÞLcï×rØF}ĹõrµmI͆N0h§›€-b²+ˆFÊJYè!ç,IëGeq$…~‡‹›½¨Äc ÕEx×A±[ô‡ 7‚©š×ºabåAQèmpݲgþ¥dhŒ™ç[ýéœÜ`œŒÑ¥àjåd{w÷”ÚZÑ1¯•›{£üLùX]þÊhqïkÇéÊ~Ãy‰­LHŸ;[,E/èÉrŽdŸm¨³-‚–êÛwI0`¦`–¦&]Nÿ¹:#ñ‹¥=Ó‹Å¢š°æJG€È‚y†ñ ›:ô†‡MAìâë;G)‡8±àÛ•ºtìÓ¿4rtäóiÝg¼s/Ë:ÍADåz‰“е¹ÿåÒ`½ì+ ^.“5áî,ŽãÆ|·Ù‘F½`ÍPþ Ѳæ¦\å¥ ÝR"¿w„/ ¼Ãʺü2¯ø·‚}O|°ûy#lãÛüH}¬ŸB $±¦ó‚ÔóS.Å ¿„;>ï<9 Ÿ&ødïÃnht^„p5š‹§m/Í(;õ$ÔÕnPº¹aŒ]jx×ð¼¦m~Xö‚#¬šäší÷¼1o_œøö·þì@š­ºòæ œÃ~ÍúóÇoÜá#ÁÖ¼áõ»ËZ=ÍN1€þÝ¿áâðÈF²ƒT2vŠ: Ïbͺa×`j˪èãëÄþiÝ¥8ݵ ö£T{û}ÃíG2Ú¶án”´¥[h»úR9~îò~›·ʸg9Ѱàül‰º26ñDÙñMfD…‘Êäs±çíE‡û«^ŸöÁ‹â "^ë ƒÓ ±±“‹®;×Ö'ÖÚ\h›Ü‘•ð=/ƒí+$ˆ÷3Í<ý„P"äóåÓ”¥¿§@=ö¼…Ÿ-{_‘XÄUÄÌϲœ&4 2?X>.:«šŸ«Š•€ÍqôӤŔ«*΋ô¼k ­H¥ª²™m¿ž"1¥/H¢ÿÅ’Ñ™¸¯"÷î€LþZªAr½Õ´‚Ÿ å¡M£"mãÛÏÏH*³œO/Þê´·“$־ʚg4à]›­³:ÄiÂst’:+%«»Ñ?Äà•Êï–[™Œ;cˆ.~ "°ç^ÞïÏ@væžïöæ ô@âæl¼"¥.e!}Û qèinÃë» C“k uZ¸Š¡®qÅPäBa«%¾ÍUîi§lj0ï]|•ªÊÊjQøÐ/m“ ru>©1GíøïtÎú î ®ÔÖ‰YÒ¤ ;†º,u‰Õ‰qˆ£° ÑÓîÁ~aQï(.[9¯=vÑ[êUavž´Ýhš|:çšI²óoX6ÛгÉäÊ©Sáç–´F샙·.õÑÒ¶ì¯{ ®†ÆÞ«H(O¯» c:dT>¿ÆÔgûI4TÄ*Óóq~"Ж^IŽ=ì ;àçU3”Ú0ÅÜÈÑâtï ¯aëäxá²CÔ‘‹ÙØâÁÓ:ø5µ XæÆj 7'µ7عsN/̤¯Cãäk‡>è»Îôê4°å[áðtY„)Î !ÝW>E¸à,|Ó^6ôêYø>:‚ •Ëa'Y¸oï+¿¢[Õz{QÉl3£ØÊµc·òûܼ£ ¼tAŒ¸“¥ŠEN¬øGùC Njç”Ûôp2ãŸSµ¸¥¤ìÇ” ft½BÍž¹<~ßüŽˆ=—z|"pßõ ,a|ãïã–ŽyMëÔŒ¯ÜlªÓyØÝ¹;ÝÃDߢù¢!n;tý¼•íÆâE sÝ3ò½Ò‘ã#ëö­Zc;âzž÷tA¥+†Y 6´J%€\Êòˆ@”ÉËö<†4‹ÖYu¬Llu Jª(!}„×Õ& (rRßDN…÷èè@-eêk¡½ïiôOá»B•Ì"Åi–6½'wL¦ ÜH˜‡‡¼ŸqæEU@J™“që›Bqe“¸) UÐÓ¡4¨\X,½ IF·êA …çº; ¶EÅ`sð&²_Q¢Dû”þ|G?),p½¯ÿ˜?¡|èx“DDsÖk×ûjÉ —–Ö)ìJªÈÚ™ÂâQ«ârÙÔbðÚc¤¯YšRê¹+Ýïãz(<ËßMÒÜ•D»oÆ`ëÿ¾?õxôž­àiþ„ÏûMIš½Ÿ=ÈõΗ Ëó«¿L¦×§ñ²f0ØA¶¢Ø8ÊjúÁß ×Z åÉ;®Ê%ØÔÁ­€ÑåW˦×FD%s ȶ{â¢AÎÚ¾óDLpãEž8“ZÿXº€GûGÄ–9ÃÖ;¬p!ÕÑÁJ×â}˜ìg®ÀšP…k#âœâ¬“ÚÜç<.ÅÊå¬ÂÅ íKOòÎîÖhéýáëJ@q§î =~<'œø-&T†ðd³wÖP c[ï.Í©EWµ>s´erE8o·ü>nf›zƒÄy•·˜Ë æ:Ã$_çãH}U4iá€Mzž¦wDž+F Àµ6ûq* n:Q‘8HIz ²H ³",t£èÌëã—³sÍh4®˜g6ÅëádšG£ô-ß3!(ìk_å}ô¢‰}šaÏ:d~é.ü^Ú™¾ àÏEj³Ïß&ƒÃõVh ¬…<’Ðdý jî:Ús~¢72_œ7ÖÁñ\–3 `þÏMÞÜô"“ýåòOTûú ÃRÚ0|‰aÅ!«·84jËÿi‘Â˿̸oêþô‰9Í,Zn³ÔŠC]|´xëIP ¬E*Îá*‡Ãº¦¶ótÜsÒuÑô$ÿ#>Ô„ªFi»ÉÿÐáId€kd}餜ð´® ’ŠY¢†û€*xôßr/AYß„Cu—=Ë#ÁkÑr„åXï½—#tcV·rïf‰:/ÿ%Øé˪wåF“±¯lMÀ„›[šÈï´€nþš7ø¾¿¸BrÏqM^š"èÆ^7Ö§üº_™Û’²ù¢ŠWà¤îÈÿx&*áÂêÚ„…ÿ.l)u™Þ5ð&Åãš0, Âu“ÿ=Ž-à™uËLÒÈ2‹xßZEæÃN;­.ßRû´PÊz×´B@Žññ)G„6;sa06j Ynÿ|³¸Ãì&¤œýõjŠ:ôpxûLð»j5!Ò(ß´ŒôÛÆ;&mˇ¾®;ñE„jí`h,Æü}çz€”[8Ë»£ÙÆåŠ"àt‹PϰõÓ+kÈ·ïhd(•Menýp,dE,¸·täEd/”ñbIÊ‹<óKu”¦³î|\6ÐÐLgÎ(.Á6-±7ÿ¬¨>!˜z%E2ùG䤯ՆÏʘ·`¿; ¶ yDSú”J·Þ2â·#ÛŒ0fY¼² Žg!¾×Á©Âƒ—hâÒQi8œKÃ#_š3îM&.¡bE^è >Y]nWïôVºø›$pz–ôÚ ÍÜIƒd¡ @œ!ç{$³m&¡\H:‰8$W®L%Æ/CKÔZª]éÊ—æûµ„ö©YðR)!@±{±1×èÄÛɵ¥yˆÕQ/˜Ò“Qék1é0ö‚€E*HÁi ©ód§2óÞ_^§¦}Q‚Û‡9þ—Á¿zsþ;"›ácrÀ¬êh“"1üãmÍ›.{G>l½Æ6ðL·”àc¯‚„‹rmÙ–• â³(žvÙP:~ð[€ðm‹ñ+›^ãâ$Ï ³ZCS”ÀU@òÞšÅ_øömb9ÏÐ÷µ_†PJ†O,>†aúÙP'؇£$OMs`4«Ê½Ê·Kxç{’rÅ :fÓF<&«µ;·.<}X>?ÝÆ¯ýÅ hˆ¡HBÑe Óó(sçòŸ•}=ù¢#a~­Í;¡ìOÍKw뤛)Q¢šÏþŽÁoÊÀm-^7È?o¯‰ ÏiÉa»'Ö™ÇEŸ€õý¢L¼Ôe‰M¥qïŸqZTiö߯9jÉTÁ¯³T½Æ õçn-Ž_›(€q8S°ZÖ,¹·–p!p´e?¶# #—:ËÜÝg?Ò»ýƒw£$£š^_vä>SZàºÊÃ{p•JÀòðl,J¦÷µsúÌOVùB~x„wÔ86Ô2¯éeäè(ñ•1KÝ٭Γü*m¹¥JYjOÇ»‘Òs* e8ÌàO‹pyƒÌµVòwÞ‚l³Å=³rõ‘ZWtQfå ãmz\(ëÛÙ‹Ýݬf[Iœ"ÀÔS_¡TÑ+hŽ €šœQ)§Ê_ ZPÜ„Bõ9f_Ñ÷e°ª„W)G>£Ÿµ—õÑ 7´˜—Ò—Å‘·RxØd‚Y½eúÇlmQcØ3ÅhbuØT[R96óž©áú€*Èh`«Œ*ÀÝóÚfÅPìÚ@’ÿêÓ<×GŸdÈÜàí¿-Â*Ž‹x\ïÆ¸õ݆]ã;¨èZ%z‚‹ÙI+DÒŠæêÄ#Ætb½-¬'̧|nÛ5Ž›$˜‹n‰)#àm²ÌµnE…¹§1œ4@Íø£¦áâ0m½åY\@?!OÊy%jfE‰Fµ àóö o?YË”g—£‘$Öas):è3{4¤K1¿~¾dP@ø7 €¤›Aà Ýdð‹\¥®­®Q0”ÝXP{@²*[X4„RëÛâ[‹÷,@Ì7>ÈÍ84Üùaú#ÐïìGáÓEº®Ô'Ym?sBí¿sÊu7z9êéãþ¶Ö ïñÖe݉*£&éô!IfuËr´^ž¦.¦©UÄkrž¢Î!pyM5 µYƺá–ð[Š>s3Cw¨8XMÀRÙ’³ó¶÷B7ðœŒß‰‘›p¥: ìÂÊJ高Y ŒL! ][þ‰xt­'RK‹‘“¸”Ÿþá&̲LÑŸ î%pÒšÔ·Ÿ·LÄV}M‹© «šÎ¸çg¹îŒMÕ1ïïŠÅ¶sk±;r¢UÖÓ«~’é™EɳI‹V»yIþ"Dù¬L;Õ4¸ú•±sÇ>àx-™Ô/D”*ø WGŸñð\’>-¢Ñ»Ô3éýÈv ¬NVp+`çeÿ5©„Ð2Siº6¤,…ð¤ÔÈ€6òv˜^EàHtèì•${G3F|4îr¹P/š®ÇK ã+[Çb,€&ü¨KF·œ“¾AŠ‚;TÑ0ô1‰@Ý—€ràÿÊ*èŽùˆx9`dŸ­þ&|Ð6gáN8÷M®ã‘"Ǧqcd.(±éÓìvJ) û|4ù>y/ŠIEˆÖͰWw倆·Ñ0Ó#Q(ƒ XËü+¢ÊM˜M.òŒXàD¡ö>©øÿ⥎‚ä ™°TÐÂ5®eŸ(ÏZ¡]íŸEcäÒ ˆÔqˆì”åûïMöOBDV½²]ÑA¯ÁM못+Ña笾,ðøÙ_¦1¯]µï.œÍøˆ:åKdd?Xþ<ÊiZKEÞ#Txbcjû4WZ[gY…ƒ².P|›1’pW–,|–pöèä?˜å™¶¤Þ‰Ùa‚ò`,Œe›C)Ÿ—X’…²m×Il )×^„²^Ë¥IuB=0Ô-L1šÊŸø=P¥p8oÏq˜«*¤ÈÕõ¥àfÙvTx½û9@k^ѱR$mMò"¿#Ä_õAKÖž³æÓnbZÆÆ•îRÂú9Tû~ÛÏNª»§Õ®f’fšç_Aƒ²‹þ rDGIÓ UX¸Çd„ÅŠâKw}ƒª€[i ž$;cÝ’xo1ð¡:Mk¡Ç®Sm…l idP˜Š©Dš{§‡À=b?ÆXÑ—Ê‘–q‚Tß0~V5ç@qì¹öÇSüº$!cZ/dt²!îJI¾ERV <°£•qè&NÁ¨|ÿëɪF£UÉk!Ç.ŸšOaЇñyc3VÙD7$žŽ)´öO‡›âäÑH}Á}~= êê%)OÀ—ç¬4êXK34Xº:ð£Žu€F»Ê*gÔ*˜â9s iWïÉi#c$QÐüÐüGÔ¬“™íÔÖÁMÖã@× ÏcÒÜB¢sÓNÄ2J¾#swBÝÐ^6ÌcÜh%I8óo\Zëa †óKWDzìÁó½^FjªÛ4qF+˜Ó4•ã~na*5ñ &£OŒ+dáœøÌï“ÌÃYtW¯÷`³·­Nž±·–ƒ}‚Q> Z‡v©ÈB¬¢É$E[¦±+%Á¶«Lß:Ò5ŒZTí—n¥œ ìâô[6ùÛ(¸?AídYIE抸KÎÚKa±èµf~ú”Z»ÛÒ¤_ä}ÍiCÌ'(!Eº{â;v>ŽD¾§ªÛT•ÓÇÍK «ßr»~Àl1³÷Ï-Ž"åÜ' ¢ªòбæAúŽ\F­ëÍ6ÒŒúø Äû¬¿W€}ÖÉc\+tDNø»wx_ÏWmÊ[BŸ*^í/ËÇÐé§ÇMPzqÅå¡çTør%®œ»}QàLkr±¥K,<™¡[GÙfâ @Ä}™vkß‚I¹0–µX̽’.ÈÈn—ÌûX™¼õ†TÞFþÏ18^Þk…ªuåö¡QXN€¬Å+-^žãw{+ûˆµSvåtÈ3ÐodÖLñ5ŠðÕàÑãS¦ž|€Ç@!v¾âò^tŒ6ÁY©ßËšXè%˜NÎzHuK ÑËÇ•äaõ»Dßê5ËY©­·µ‰§²Ri‹ø‘Užºa· Ø& ã^©˜õ†s=ê× ó)k¤1Ú[<¹F‰ŽÃÏs )>ªac, Öb 8v.˜ØpÈ:‹ˆ+޽às,›³ îº(îù8tr¥œ8ùÿ(O ¦îпiÎOÔ\Ðù—iy£.S“+ª#ùjÍ5ÄWY>sœz?L^Oƒ<\­GÄI3$¨"ÖÅ”{ _7(X’ãÑ Qê}Ê%Ì~Üà=ÈUqù܇•ß‘ûéŠsž§.ˆB½ÜEv8ø!ð¸­é(ar¿Ž9õ ‚þ}µš¢ò oKÊ6BRU&l5”qG|åyêßÇ“¸¨‰„Mç=ÌÊÃv»ªg€gÞC¶«v³&~{p<0RÐüP>È£e™`I¿U²Ú ’TÒ)¥]”¢¢‹Ñ™”ucç½=¢c¦¸o$QÞÿÃÙ èûr¶üŽ[Q fÍ$ T¶uv75ÿ/œj04ýg¯õ‹hö1Ã.4éP {¬«å~š™Ä½MÀ)nÍbÛA æMôgG6#í’¢¡1dÝ-}Œ-ïB”r—ºÏ{†6!'òÄb ï^ƒ SB¶Hûöüâkx¥:OŠ©\‡Ìþ•mAÄsà­•rô^íÏÔ@ɘ:CÂ5·ð÷õ\ÏŒ×s‘± Ž$â`ñ™·§:™¦vçÁXñD‹’ÆÝYƒU$¼Íï€×2*¼Lìf`©‹À$7j͈ô£l9Ã/Â+7`d‚7Æ–Ý UJ~ô7ü×;&(%ì¥ÿ±‰€÷vI°‹9Ÿà@6£Ä´ÌyáFïwøÁ?ì–™WÒFæèÀ:]zNÍ=À™C-ŸÀúŠðƒ#Q›É k×`† ~ùâ°ŒÓöz<(Ë^§)xÐ…ë&XW„‹7©Ò¨ Ê£ eÛhL+÷.1æ®– x\ñœPµ–QÒØpjK$Yÿ·¾ˉ¶Èµ> ÏÍ2²? µ3Ž…þ'ÏH!x=b6ºï©å–[Çѧx_Hv™¨üénÊ4ògÀ ðØvjs(I0—sÐ(¸]…fJH5~78uNmc^"3Xį;Í«²P\ãôº’8‘ Ú^#*d4‡†à Ú™QÓùjù¡ë µ64*pŒÙÝFÓ^vèdéÓ9>¶óà± ”Þ½»ÒO} dHq,ù;žo%ê G“ïAǘÄ#wˆöìû–çujœ¶êv­"”MîIgl¦reÖ*ð§< ¸“éÔLå¾p…{4á|¸\SÉoœ«’S’Ì:—×qÕ9 o€,iÏdùn7ȱ ï™a{Õ>K+B“ºåöj6]Ô×þʰüðÒ hu}Ý÷?|UïÞ^I1‹ÀDÝ£ppžlë5xÅÝ»‰É^{\Ê‘ãßúÐÿ³ç}2)÷h´¯‘6a÷'®>û9”`½ ë6†X É Ðø=£ØtMÓ1B-6SmšžÃD´] ÚÌ.þñ:ó© Âjšf:u²…Mâå‘‹E2ñ¦'/¡xã‚ÑéñŒ©ð"@ÍÓ‘$»Ák†êFÞ¬¿É?ø‡ Á€‘¢0&›F¸Kûã:ó^¼WŸòb dDcF}ËQ•Ý+’u[óOoãx.¥£åL¼ƒ‡f‚ò_j'I›äiøGÉÃc;c#N6v-åÿÜ|I?vøMŠÇj?KGuÁ›#I+<ˆ«ÞQ¢"ò×y”Ȳ™g)N/ _Ô¿–Š’o endstream endobj 124 0 obj << /Length1 726 /Length2 29757 /Length3 0 /Length 30296 /Filter /FlateDecode >> stream xÚlºctnͶ-Û¶mÛ¶­=±mÛ¶¬Ø¶WlÛ¶Wrß½÷=û~§µ¯Í?½ªÆè5ZUÍY“œXÂÞÎEÍÓÀLÏÌÀÄC¤*¡&ÍÌÄÍNÄÄÀ CN.ê0r±´·3rðiL‰TDÌ,D,LLÌ0äD¢öžN–æ.DT&Ôÿ6iØ™ZÚZ:iØÛØ»YšXñ¹¹¹ ¹9»28¹ 0üÓH r±™YÚˆD•´¥$‰¨$Ô‰$v'#"%WcK"9K€3€šÈÌÞ‰Èæ? ‘‰½©å¿brføwvn'—3s²·%’W–PTP#eT%2²3%’“ú§¹‹3Ï?h€‰‹å¿²£#rù¯dcô_Éø¿Ò½Nÿ‘`˜™‰L-M\ˆŒæ–v0ŒÿâMÚÎÌžˆã?fSW‡ÿqýó?ñQýÃ!5‘)Àì´«‚‘-€ˆJÔÞÖÁÕàD$oo p²#’v1ú'³ÿÙZÚxþÿÿ/â?pa;ó¨£gfc`bûÝÒYÂÒ`ªdéòáfF6ÿõo»&à?³#0µtµýO'€Xùg^Mííl<ÿï¸ÿ$óïQu”U%µ”hÿ[ÿv‹Û™Ø›ZÚ™©ºüè‘“é ÿv+Yþ§†ˆ˜þ_wÿÖ™ÿŸ.oäâdéA¤ËÄð¯"aú×ó?’þÿC‰ˆØ{xÓs³ѳ²01³ýàfåôý_Ñš¸:9ì\þÍú?©üþïj<&0ë+ö&¼!Vé­a•~âÅsUà´s3†ê´+ÌÀ«Ó½·ºÈ–濎®qÔp Üáðgù÷®v¥ZHŒâp'¦bˆ#ôWpÞJI«Ìýj·?” ß" nÀ4°mMÀn#ï]27Öƒ”h¦Ø²ó.jeËù‘%™c +:À‘vÖŸì$©õ)Q²t¾á´5Hs¡E7Æ‚oûë áÁ{p·äØ¿,Š*ŽÂµÙ©%{Ä`)o•ÐN²éel[zïáe'šõ®:÷—¿N)eйoùy<(‰ ÜÔ‰Lߥ\½ì_½%¿|Ÿd¼éëߟ^ì¬Æ¡ÛQcP½My€ÝY/vF$sT[mî·­€LáRÛgO¦ï°n5z®Üuòg¹ ]@èµEt§Mn¦ùW,ƒÚ„ªÏâÊž·”Ö´{ÁqÚÁ¤x—Í8 aÓs¾‚7ãj$Þ×¹Tí§~2½›TÆUïì&tˆ™›X4æ(svhúÏ­‰Ô·RÜ4Ã’:[ʨ7*Gß A‘và… ÿ©ÍQ!‹Xþmˆ‹NÙ§ÛAîj<Åç›t-.4rAŠÐAcæŒLÁ4òë+UÔ–rÑï%˜J4tX†Ã¦6`ªZQß?б·ŸZ²`dïõÖ„ †®5aSéNá£4¸KåÐ5òî[ÔõãñÇa_¬Íùkž±ÐKï¥O•²qqB¹×Áå ‰“ù–ªEwˆ§D,Ô¸ÿiF‡ÉÓ÷Œz²ö8>šÜ ü,Ÿãh£õÒÁ¨8'ëWAš¯z\» ó‹™>€ç¸;¬ÚŠ/ ³-Œ> vmëIÌï´êaœ2Ûñò9ÄÙì~ñ(ýÂä v¿GK®z²Ë ¼Šf8q4=†¡ºwÅ'ÑÚM›&Ö,I*ú·a˶dP5riZÔžñ†tǽ£AãMµú+ƒ¹úpß¹ãŽ<ÃgY#ñGó‡÷fÚù„ŽËX…/e[»ºÝ®âÉC²‚0æÑýð}p2‹–b«ÞLq_ã¦"ø™[¡¹(à÷•wúÛ_Iàw¯Ø°"ŠýØ’}9[b l¸8>ˆp >!SÃ쥩%²‚'hu)¸ÇW»'TV?Òˆe}q ºƒg?(HZAƒ‰O–×6龂S\‡)jÓqid—†Ž#“5Ÿ<öVˆ¼Eåé©òðÄ’æèv9šww¼D”k€Ì%bâÓ¦Šì&X9D‚Ó±RÞWýÕj:Ð×JãEzyíc“× UÍ=4£ˆjá}Á=™j«ªLGEu0Ä´¼œÓB³ÜuÅã¦Yo|²äádꋊ*vDj®r[Ì.ÑyBÝ,<Ìï?ê }š7^óHæ˜ ’=d!¬)XÚþZÍGåoýA] ¶¾Kb0b%È;*G9ê.ŽÉïú'žõ ­øw«·ßù¸W‡rðJúÏsJ󇮗ºi°Ÿ¼Š¶(¸Ùõç{¾ëW\g©µu—+  †å²éá [`½™åº•’Æéôdù)•” L7ö v»kÉ€ÔÍ“À|ÁÃr5R߻‡S.*~áP”Ù¼dœ}øšMÌ+ò.0Vsý¹ÒºV𿉯–‚ªÕÊ«¥Io´Ø•dË^*Pa|#«–.´É;Ç’·Q‘oƒL™iT ±ùrÆÍVÌïnÉ•ôéÁ :›ÊÕ°—/£¶‰c7ä÷Ar9¹‹h›äBCM £g½1ÄâD_ôåm^>§ \ Þ­rˆ¸x›.+¯îKåÓF*ÖgÙpÛ5O_Ì«´õ[Û|Žvå¥Íœ‘›°%뎈ß)8q;Ì‚›pñhpõÐE2”@œ-ǨlzS ÙÍ:o7ãGÓ{—a÷ÄúÇ p¢— êšíúLzÃxš¥YâDÉ“Ó" ÷y0ú|”Ù`s4Ý?£æ;ècöÌG>èɵµ-/YˆEÜÇêtHʰî-ªÅ‹9/OÌ_J¤·½ç“¾.ïaFVcTôony½·ºb°¹n©<ú+žY錌¹¦ÅaʹYd@üˆ`™“Ûý}oøÏ:Gk˜ý=ö³BQ!_Fq³*Ð_q+§‹¦(C¡(/"G„D‰R]µ ‚/=êÕ$a %÷*QÄ#ðÂŒAgþzt‰¥³õÐPvaÒ‰l„¡Þ4N–¬u#ÓÓ)¡³T¶ „2üZry5Þ‰˜O3AmY£ºW¢ÇˆÏ`˜Ô´ ûe¬EÙjxz³ç¥ùQ-ÚP8¾ìlè‰1­%"á~vÍËßÏn1ç ÚaINhÞÛ½Ä"qµIóë£ElȾM‡!µ;H‘Žï‰+ª½èW‹ŒW„4ÝjÄÏLö„¹áK¡ç# Fdä ÊÍ þfßïce2‹ÿ·kubݵº÷³òCÕi µV$×gÀŒ•Zúê·‹Í«lX¤«"ü›œç5AH¬ÌO¤7Npmrf£1~ú÷„›ÎªXÊ'ÚŽ™šÃÏP “æíj~¸)ŒÀ9k¢RTêÃì ãeN’Cgè—‡‹)Fº Õþݦɟ9Xž»ÔÁ>±‡¸,»îa©¿Ã:IòüXRô”ðÛ‚`±²- ZÌCy¢Òªrl´h½Ä"Ê™­ƒÏÀ.çŠ{QS†â`èåó’ÝÛüžÚ…ÅðͳÎt ½áØÜsTÎ`= óïHçm[g;¼ üÂÌ= ¤Ý{C¶Ð³yŒs΄Rö`6IëÙƒWÃÂýfÖ 9Ó¶‡‹°ø iNx+üÊ5ôÚd†¾ òéÔyæøy/¥Eð¿œè+lÆœp8pϼ49 bžzrö<ýo…ñ^úÔÚEv^OçiòЂ°ª@ä.S¿w‚2)â ý xèÚU!³'%äY¦æBòé¡äC?,Lë/³Áua«>šêí^€{äeV‡ìyŠ-ó<ä‡ÑÈMí<c+ôÊÑÓpn9†×$ìL¿/ÙIö‡yÙ$cËý &DpÎ^$òˆì}i0žíOh’nMeR„âÅL•:c‰§0N†èøŒTã6q/VG«Ã39ëe¾Bjhîàô÷ßûîH©AE¢.ŵ‚¹“Å5©:Æ7Sw"ëò2F®‚÷ÿ}¥ÂÎÞmÅz^Èìg{xƒÿÏ Žê@îLQ\)ã¶‹<à|yÄ/UL+kÙΘԒ¸“<ù=uÚëASq þÉÊ0J «ˆ‡©âÏí2SoÓD¡0J‚°·.Y«4f¥òNÛÜu5«!C#sm«‘(2qŠuîßNˆ†'•9¶ËCå -ß4ëÄR õÆp.|=U}ßYB[ý[P9›…djbMÔßX6|8N¦iÑOáÔE8œ×6éÚFÈÑYì n©•²-7 ·ÔØ$+½¢À¹FBjø ü|§ÇßtlOEÙçÒÑÑ{ø€EBê6Ðs›?Þ_¶ÇƒTg½"®Æmè‰è fÐúÜHR¬iuˆ gŽvávuQŒ«ŽF²&”êç£Í˜düûePÝå‡2a•¯!tV›ûDó‹îø¾ ¤:‹i'=*x™0Ù$í:ju _ÿ$‹ÙõüÞJc”ŒZËÂÉ—ìE‡ÂÎÖ—†%J{^ò\|úÔçˆlw4PÓ¬×@¶·WŒŒåæ=Eí_ªÅµÖ;†ßqJ¹@…C] g< ·ki)U7úÄ›‹ Ý”Æ4ØE¸•ó¤ü”-Eß"ÓJä€S6Œ}ƒ™ ©®}Ç'–ÇçÍãµ›¢æTíͨÎ,7Ù!\Îü:¡$‰ÁMÐÇ:ÎífŠq%``ÖJ–H °™÷;Ù0îÜÍbŠ]þµèÖ/´Ù5’¤x—vŽb¢!ß7ê‹n9¶6É´^k1S#9°Ów× ­xÜN¦¢æúG)-KÆÞ0†ÍbŸ‹4‡° ~˜”Sʈú¸QËòè„>ÒÈm—¶Þ Àn%âîhʪ!šÑIsáã>.p– Öö§éYÃf¡fc_=¥É°ÕT Ö2Œ\6œT5å“ÖYã"Ôkýn û­ä¹ÅG$@QïË‹ºPۻ˱Êåwï@šÕ <|Ñ¥(+‚! 7éô…JaÙ,€îªóqÎ ?á§EÓC«i.:Ó)v¦²”h€‹ôFŒôæ.Ft–óI=qw¨/N kSÕ1ÆS¹ŽTd[­GÅ¢äÚŒênûwiÊ §å²Y ƒ¯tÚ\º Åú‰´ó°0-¾&Êñ»tg˜‹•Ï!G¨à§ã?ËbeÖkûhyޝ®õÅòQ{ž[§ ·)Á<ɶ†}¼Øêþ¦à¯àJQzTÉ ´¼õ6s€?-]Û–|mãÃ9 š?žD66®bïíýV7šÜ1«“'z=š >8înS¬ Ã͆EäÓ@³ÓëfL ñòЇ±Ž ‚”­ h× £ñ8ëd\ôàÕˆv£¸’­ëØ9<¤¯Árv郇EÆÕ(ð¤Àh¶ÃuEö-¤fG¶©v<ÚèûÒLÝô{÷jôY¸GAñ;œÔ¢Åý˜…hãÆ¿Ù`bù_q6åiÏLI6ú®(gß[ü†‘µ9("ò Ê•ðþˆã/0­îÿF¤æ Y‡¹=ŠÀmTQÂÔŸ$b–!czè¤T÷Ù·ðÌséë!¤ê¼Ë9Ã@/äj¸Ç!@G/'58åû¢\UØ öÊ¿–<¬ý3fnü¬cW¢…í©r-F°ó«ØÏÝêl(·4’˜Ë!C°…)¦Ó;g€=ÞÏ-¦Û¾£3%§넲åaž…­Èl)4Ï Z¼áK£ ¦L?ø[Â(¬j5tRˆ×“Ix…”æ“~øÕ¢7J’è e‘©*f[“Ó ˆSžÿýD~2ãÜù둉+¦*+ÞØþö)˜Ì¯Žø!Zȼ]¬M@Ëü"ùõgRÞÑJ‹j®"ý\íV¯Ýêôö\j$Ô KÅò˜ŒÎìÙ‚™=¬µ=5*tÆ>ûìþ4„ bήê/|à%‰›0±4ÌÊ–ÝÝG+öÜ5,'Í}êÍ|MU*G§p¸¨PlH–3̳[@Ñþ(nÌ;´%L™F*m»ÆH¯/ä°b,å¹£Éaüýy®ÞÖâŸ{šä;ÌN¤%´˜R=/|oFaLö¦}ÍIè¡/Lÿlu™¹Ö‚25 IÙ¯š3³å6ì4[Ò-ð½ä+ûÔBxô¾¿Â¤ƒ^X° )üÐù“GÅ!ÆGµ¯M"<˜T›Ç›™‡c¿eJ…-BÍŽvåéÁEšdÈͬFbûÇmÙTÃ÷†K˜5νÎkê+ ?™œ5ïU× ± Ý Ã(íl>Ê*Ücf…M…u{î>À"{íkÑWüÏ}ä“¥GÃÖ—P)ŸR>î¯U‘Pü) Šã`?_‡ (ƤnY½CzlÜ`ÄÔø[ަè$GVÒbH7ÓQ&ÐXD*Ï}}%‚°Øòê—PêØ¨R^àÙv£/÷,ŒÁÂ,ê»`´IX@Åj(j6’uœ÷CÁ†›€žÀØVþ÷»NŸL¥í¡ø¡»4€-¿D‡Ìæ’q@m{ Ôÿµ„Ï"}´ ·œ¿µÕJ-¼XÜäôÖ¦úÂDÏLÏŽñWþÚv/h:µD¹ï£²èÓ[ù’÷v[¶(‡ÝE¾ioM7yðk>§ÖTˆ™Û½Ín,ËÖ§kOï‚(KàQ·ºÂŽž@¢Û¢$(—s÷ÈìmÒùƒú°(6´HeÞÔµ¾ÀãSO:‚(ð}.Ro<‘h_Éô{áÁ%“ÉW)òNƒ·4}rpÜ@>HY|‰{êîâOˆÁ‘Wu`w~äÑKhî·~’ºš™“ÄgºQêdFórÂg„y L$ú!7§¾aQX>:÷ÖÏtŸ‹°-yGUÅ4÷àd»™ïçÑ(έrQÇ­6n÷áL‡˜ Ī*}±5·ägYTëë÷Ýt(ÀDP:àR#!©dr+ }l¬ef§Dù êX‡6ÊËBžo ¢¯MT½mý =½‚¢ë„IßF9jÒƒ¯2´X°qº¼«FÒ1,F·ÇT$[¤“ÝjO}²°/œz”Ñ¢ ?¬¿’½Ë ]ÒCB¯g’ᵕ˜s ˆ/Ù3Š¿&éþ¸9R •ú{fH vÎ%VÜÜ)^ø±V ‰^†ÊçI݇õV¡K%8 ‚þêóÛ™÷ˆ•÷s¿‰Â¶gTZ6›i< þ£äé&âÄ.Ú軎˜Anªý¡w¦¿Ä/fÛΆ?á©U{º ÔÙÊëó°Z†ÞáúÐÅ—ý3t;å ™‡d+MBHôÓšËntœÎ±QGÕTß^ßÂЕ›¡0pµ9Áœ“è¯UHClú`Zfܸ°‘XÒ ±X¶K¨î}yLcÝÛŒV›wƒqŒÍDåãíboÿÿwÅúuáÔÛ´ù±Òî,˜9Í;ªé3›^Yrê×ÔË“W 1èÕAR>6hØÝþÝF7Wë ÕÌP˜úž,Ýø›Ž}K|gx«äíG@мøÕè-j¹0€¢/ü‹Âš”´ÀæR¾‰¬YzŸô`jýúªK­Ý) Ï—ø%+#Ek;½í\RüžŽË:S,ßT‡ët- 'Í 'ë+ó¨õNe·7€4!2o*^¾šŽ€Ó¦µ>9rAC9sŒ;ØžÝ̱¬fÜX¶e®TÉMÿãNØÛ>; ¥ êºÝ­Ë¥iÐUª¶_­Ý–ÿØ(ä  f3PCaR'PÚQê˜ÕCX82¥³°ÇE†.Åy’^)Ò‰\ÆoʇžÔð²‡Ê)Ð {•«qµ‡½taƺJ¨ÎȹïÇÆ‚Î 4ðÄÏA¯ŠæÔåÓ´EƒBaÝ÷…|†pH´»×¡vh?GJÆ.áΩÒÍR@Dä;ýþFê œ½jfYÒ ï¨Y§‹Ë@5Â[î…Ú„ßç"ûü”æB††Â‚ŒNŸ‡Ýß;ƒ’(nf[³ctrC*m$sÈç€drƒzºoÜÃ Ž†¡¸W¾é'ÞLײ˜k»GÀ[èö)5úðì <¾·>¾7%b׎¬–P| ¨ÌsšEN•ñkb$¼oæ«UQ­ÂK¢O± ÐírlÍÿ«*·°¬8ÎF?ÿg HäBVñÉK @\â°šhÿ\™sÍô… (à­•J\f™{ÐSŸP¡ŒúíÿMO)l{ÓqI–z¤é[p± ×ÍŽ•Šïï¦ëY¶äQ¥\¡zëŒ*¼ÙºþRbÓa®6íýñ'ÈEç±gŽÀ<õë:z&.Ί*0ÌI0ÑÖ-ä±³‘S|Í48¢mos}Ä7£Þ×ø`y\pºâV»‹O Êg®#P÷Ðrø)™}]ÿ©œA/ôøióçãapç9:*-¬¥©vÁœÌýi´—ïøâ–΋eîbšúu¢,p^OËÓDÍÎÓ‰Š¿c¤èÚ_”ÞÀoüj¾ Ä£í+3Õó¿»­E« xo€%¦·É¦ÝÞ&ÌV{êhp]LCÛ÷‘5¿Ù€a¼÷ŽÒK™{\^Æ™òÀzCç”Î|i•2²Y2—b†ä.æ ½‰Tèœûo9Ó+Á»ÅÏè¿]¶Û)éqÌØdy§<½Íˆ²y8•h)×n…ÃY”mw»Fž d·³¨qSß·ÿùņá”WüÐÕ«ów kÂÏz` Î'NIŽRH‡‰ä?$˜3tÈt¾è™Š9÷µóÑ®I&º üzîl)4:Ò7F Ù6ÕòÍÔ »ˆ-EJ Ô&"Äן—©n{Ö²ÅË‹R4Âúˆ¢¯Ž•ö熔Ë9hl~àø:;ÛJµÅ;u—„¤Ê;  1˜(̧DÈÜUp°&èðó—.s_ߎ9õŽ–î¿OÅinA!ýÒ­qˆ¥¸¢m¶šB¬cñ¬^Yß?¡RÏÛYnªïkÌ€ÀÓ«ÃÚõ¸«yÏ5Ä|üCÔùS©åº–Eà-1Ђðë5r3™Æ!§ÅRïíR ?>8i1òÛYÔYçqŠdIŪ@®w¼æâ3~ÄY $Œe MÍ ‡A ‡–Ÿ—Â=|Xš´¼·°Æßµ9 Tc¥¿ß‘)ßõ§ñFžôí| ç/>PÙaXvƒŠap$ó|Moê­Ek´\ÂÌ òÖù̓C;‹G0J;kçTŸ>èr¡;—ö*?ì,N´ŽÏ³\çÖ—¶i?Ô‘g;¯ÃT`k‡~58r'_oºÑ0Å›~niò©)æoë520ì·\Ò-<^©LÕªv+[½Û4þüÊÍ‹µ.]Ù&ºC³4¸Ùc6Ÿ­!ŠyȪëâ¡ý”¸M²¾‚(ðüƒCuÑFþçz@÷ ê1]ã)2C‚CnírŸ£ë¢—5ƒ„}Š‹ïx¶,³“JaÜVýj¯±Y²”{ÍLÍ=SdÌã4g“b/5sÓªÓ P~gÚ´è.7ÌœáÒën)‡vêZLÌqä帊úîRh‹æëЪ1£›°b-±÷(ÖýfÀü=Ù¢nÓÀR¿xyç«·:«­lßx0ÕO:>ûw Œ£÷†³›*•07—i6Q÷ñ}ê eòwØhéy—½£S¬Ocä4‡‰ÂgD }=bÐoh-<™êb‹Ô!`>A,Zx9®8óŠXˆѧ哳#Ê›:~>Y§çÎ#,‘¡#ïfbz¡þ/üû¨ñ×áâ*ÿ­¡«– èû×"§š*ëeq÷µÊqRô ÁµFÏW˜uÓU^èBg–¸à,ʨ …JUô1´ùiž€þß©€2‘V4Ü©ÐvýÅ(—éRxŠõo~ɽÁçñ6É0ç­æ1°|wñe ¯¥âÞq,…ylÇ%ä ½ƒA“^&îuæ{>]˹S/#‡‹û¥Æ"Ö¦ùÉU ë†ë#*«.H…wÒŸžâKJ¾K¡µñ“5aº;Ý_9Xm«ä®XÄãß{'-ô©y²ÕóÀhuB:?‡<¥ìÊËU«Ó5„ ¡ˆf™îf4*5¬àA0äÞâ”ü2Ñ—mHÌÛþzëÄíÐk‹UµúH€c«£!“!áû;¹ÌÇpç;„Ã*Xœ•y Djo«^嫳¿Äؤx¼™«ƒdÒ^HÉzg/»o^D v[^"·¹ŠÆJN N¸œ“¥’Ç‚=@'ÞHXˆ F$[\ˆ«4ðuOý›?Rh¶Ï×Jù¿¾©mXîQLT¬j‹ÿ^ ˆ…v$?[Fkâ€2Á•Y|\D‚þö¹cˆ °d’f»A>ö'ÛÏó<}i üU¬ãÄ*ê(„²ÿíq;P˜pâuVÔá’œay²áýȸÖ~~¾é•Ê Õ”ÿ`FjÙó |°rÌv÷Ž2LÐà ³0]VµÊ\ 3’ ªÈ±M${P¤ÜŒB¬]&J£ìvÓÆbÚBóEIàoì7î‚)~jÿÆúp09Ùç¦õÝÐÓÓ“Ùt¿¯ "L­-áÆky‚Ëãs¾×’:4ŸxWê'åDƒËñɲZZûŸ'¢k`hü ͳ9þ?ÿÚIJ*–¥‰#ü‡L_?e•e¹±~ãNw<@“6Å2øÉÊ?Ëû7‡‘ŠÚˆÂFÞ)Ÿ¨Vµ0IáJÍåÃò'8 éômœ\Ÿ§%¬š¿¦«~¨od`2×­Ańޚ ®&¼-÷Ô{…OÆ’½U¦ŽU°gW®àg<öIà ð¹ Öµ:—# FŨÎÄ™÷¼^w†6¢%€ºT U­_~·|,Ç ` v>[ò"}áá•J‹VÑñœ¸gİËÿ:K±>èÉ5 FÐãh3ßÕ›Á¿9ãÙpSÌÉ©1°iòûfÿkïnp¢AM¼ð¯ïLû¥ÊÔ-¹Šû¦]rg6q9ÙΓõHi¦ãŽ_&ÔšÎèYÃù‰þëZ«^”6B6?£3ÃÈ'cü_V¦|æ»ä+¿~ÊžF zb±`ãh‚ÍZâW…^É)Ôrüícq‡xþrn‘>üv‚ì^?.CáÞá¹ SƒZÝâI†Ê|‚€NošÙHZý¨"eÅ¿ }åªÍö]ü…½KMüç4ÌÍ!¾e³ï Ä¥¤TÓ9•QîokdGxµÄˆ¢…²TK!.äòD1–¿G¦„x’¦¤Yû±¼Š¡_Å£S»—Äë #€uÇ9y—V¥u¹PËà<)öå'xl˜Hfœ˜ 1îÌ!Ò{)þ|Ezî ÛÑ™ˆñôœŠAÛœ/S›°†¢x!¸¸ý(p™¤"RØDÞ§/ƒ69Ek}kEo*Åöv …«RÎùK«8ºæV<Ì·¤à‡ÍW†¯¢LLÕU2¶KØXAóþnE  …óë=þ2hÀvG)gŒ¹©{N²ÉÍœñ){G1”Œfæ“®›”Ü ÙØð§´ò“f4}Né®H8¾\Zx;úè=è©C…Žë€F]@žDeýflÃ^ÅþD ˉX4Š&³.‘¹^Ð «§ùât¹S#Ó†»Ýr*õ@7—ž²°_ž6^²·c[’ |_:f/Æ"ÝQRéœ×˜¨Ÿ‡’tÕã=ˆ,Zln;à5ÙM ý7Çá89T)UŸ7Úwo,]ûäKSÖ˜g°è<€ê‡«y­â; L±õOÆ¢åÑøH§ NùQo[òœB¥‹œžH^¡=cÌX± .Nôœ-œ^Pâ©/ßÝÜ€ìdñQ¶¸§ž<]¨û ä™ÅlÀ<äÂxC22_ØAÊÅúWØ]=sjö³öBGÒq´¯KÄåib"=wôáоÝu_Λ¼Cÿ¦Ë¦ v:ìã­«cJ»eÕ~…1†Ü? ¨ÇL줶íl»’ÈY-Ö4¶H0jÇY¢’Ï;¾JÑú ²Ùå&Ô´K2[ÂïþÄ×øMçó1p ƒ¨àá@*[ìî§UÞ­7'5=¢…]íwÏÆ¾¢’Äåç50Õö¸*ŽÌ§” 7UT'‚”R3-qÅñÖJÛÈ£@gœ{7Zñ%Û ¦6Žk¹$ï¬ýi32•–6Þ†qºÕkPD_ˆjܽý¹û×kä•Yö|Ó é}%úAȦ·†285fà5«Vbñßã„Þè®À• %au‘ÄÏ™£óts~a ùùr©døQÞXøÄÕ{2&宺Èr–Ùþu»7°7ɇ•¨Ã|¼ÒnB’C]†fÎCSnø ]Õ‚m!£¶íµ’ä@=Ò3K$¡“eèÂ,×Ò¢®@HËÐ4†›)z /‡‡.o…·'Dÿb^ÇA®\o vD¡¡ôT`Ó²9î¸òDk¼S\©"Ƙ‡Œ94êbBvèù홲¶ö˜'îEQbÿ>eb¶ð[û =•™‘.舴L“8÷2 þÁ¹Å ùŠ †ÓFkqY)ôW§ºíí2/kCËè߀2ÒuA¤»ŸX%¢è×d™ 1¢^a3QÈ'Ϲ©žë¯_–6Zžc ç¿AGª¸Ø(÷mH:C”$ÚâCµÕÔH[ö&‘ÜKEwÔÕÙNÕ9ib­´cSâ> ž«V(›“MO§IV“¹Cuý$Ùj¢ø"634Ý´XX®qñf”!¤c$îúhQ 8ߨ¿bžìœ‘5Ñ tÌÍ1=-¸ò}Óà»OCOC³9õ¿MgsÁ¶ëpqr¼ÖL½Pç¢5¹¥Ì¯A§Xýn§&/u}p€€ñµjq–܈9íiÏÔc6#fcSP]…â¨QãV#› =±‹Ö—µÅ`k°l$ŽUYM½¦‰#©½oöº=AâÙE‡™ÂØ»k_ SÅ-·ÉôÌè½°÷d°QîÕ¬¡‚¿”±ïéÑ^àäg\LÓB;MÊj ñ¦øWù“vÿÐMh©‡l§äËë,½¼`º ‡uî©{wZ§ç3’FYI`–,,ìþ陕;Q¶šù¼¨Ë&ñc[-5—XÒhÛ¹Á¼fj& 〣ßmeýhºV^$Î[ž³“×HÌ<$+Ôµ©$¹àÑ÷‰®—˜eÄ™¨Ûx²h(+£+÷‘†GÝ¥šXÇ æ„ôi"…*êPgf˺þm…gyó ¥ûn+pä"ö~ä7Œ}fÝž’ÓZ.ÅÀ{ôT~Kð—¶Øaò˧5š6õ §žÆ½´›¤F¬å)Ôo¨E{kÎÆÊ:øùpDS³sðÍÀ¼¼-Ë’}®ú³ö÷€ëT2d OÑ.çÉÈ%býÏfzi¹Æp‚Ëï ]¯ÉÖ±#"’1jDiü ÁSu‚› ´µ7þ°|ÙUf¾e8<_^ övZà2R£åÞ†Ç1ÿd«Y†Ó І&ÙÂqcÞl‡J¨±ôÕ-F'¢¥`Ë8skÐ&„<8 Mô†]cíúÐW-é:E·ŒUGᕪ¨ÏU‰–g3ýx‘šÉu\ØìûØ‘@õ$­ú$ ðÑvæ½¼¤b=֗ᜥIcnÕ=]ëæqûC‚/œbÏ™¼ò%’?Zãx2=‰¨P3™a+ è~8JðÍ(°í.ÊYen âÊ÷ mã–2%¿N¢#eôÖ¯âtÛmYYþÐÂØ. bÌXÏÝ0Ó?NÂarl]fŽfŸhWP˜Å½ñ8ÙGQZQžF·K ˆ¬N‡yMÑË+ 8¼?²ÿr±I¥•&BK¦%Hß*bÓäçaF" ÉS2ï”jÑÔ—p}È+Û§¸"áÜ£©‚qå†Hº¦Wæ:;|%öAÂ_j =û¾ոî“aboŽl ñ•È$ÃäâãÔg(…VËõ¥¸¢AÈ: ôØÊw½eœ‘tlùúOøýÜèn[îì4úBJ$êC*}ý¿Ê"úª‘$ä´g#»d`/eGz­¡ëœ¬”ÖÕò~Á¬˜²Me†žnrƒ:³U¸ÑãŸÛzÓÔ±·‚íÆï5¹h\=õçž#I}'òÈ0Ÿ,ñÖú–ó÷ÜÉìM ºÞóqŸj@[´žªEý‚RüµŠáÇ,Ç€_âˆíŽB¦¡càN¬îzl(Å•%ʦ… ! 1 õ´¾$YÙ§Áà4İØ0ØâüƒCUepÒžwZÑ’¿Ç¡Lä®ÙïÅÀú_ðij~5]?Pæl°7á¿©9Mäç:ZQ°ó)ý§ŸŽ'l¨fo‡7€. •|D”—By~;¸¶Û„Ї0†˜£1®Ø°¿É˜c%ß±†Ä–:b°æˆwå+[…8§¯„‘;À,Îpø«N̤+EØož˜^Æg‚ tãé&|p¸5|úO„¿üш¼¯üiÃ)®²Zà#Øö$-6#Y`¶¹lª¨Jêã®>«@•ÈÚNÊA0c‚Dqæ—|ˆ°jÍÃ×d«}ë+Nò ”ðxŸ„ï8MÐ+†Ñ ÄS?]‡hÐtÙñ#ö«°ûíXD˜‚á~ºÈ/}r®œºá¿ `A/¤Sà]¸À—©a¿².ÔýqÁ´°yˆèç÷`«KÜÃ…²¶×Ö%^‚üïÕÃä¢,6k&79qHŠa*Þ28‚æ]ÀbüÙd­ãV.ÏÑüa"ÄkêÉ<µ¬–8¬V|d%zÊ ·á”hÝ ¡ƒœ‹ùf·`}óQ¿»NÄÏAV4ö9*çr®i1Ù!¬Ewëw$›%(ksKÂÌ£ e=˜Í³œ§KÕ£Øû Väæ‹°JÌ•±zÌ¢`%"ŸÒJ–×E¥¦œµº¡(yYµ•i)vºµÑ„V¸Šy4A_aöEi÷i]ö-VUm€)še?’4²ÿ}7ô1®º;éûƺIŒ ”!Bü‹’ óSfÃRËž“¼&t?ê›O–^qåÒJ Ll­uª½—Uâ· ÝZˆZÓ œõÑÅS†ÒÂÔ-௩Å|ž*ïr’ÖƒÎy5½ï‹mÛ‚š;?þ wìÍ[…U,]†G¯à 푬z#ðuiåhÔî'òÙZ¡þñÏã´ÞôŒ(è?Îm³àAŠèÏ„*êá‹=óÖ¥ÿt¶f »¥j‰ô[ƒ [‚6]þ¾6‚Ķ–¥ ‚(ÁL2|-oUnÓ8×c!5á@ëBú˜µ±få_”ΜÄöU’F E‘ËãåC­xÚ*ùjj|PÜÆ¿ø„äìÍøà³ Ð$3FÃRž‚pP±â™ e¸ôK×áJcöŠG xÆ æÿðÙ™øßf§EŸ½¯('`]Ÿ4xhþ-_¨HG“Õ·¿&ðªðè§Gù6>_y‰C–Ú£CwŽÇ§v•M¬kEÑYñ Ô»O+þ¡ðoØ;9ÅðÃöW-Å??ô¬·~Wºë ½R,´=çNÂlÏu–?¡¦yûrŒžRazû9`V_¹jhóöÈuWSL+ƒ»UöÌx8î> ãëYhÁ¿¸Õâ«[ Ž02E“Ü)Ñì¹`³Á\¨ i£,¿§ h•Ìj‹w㺆><-xÇ þ­ççýªní %aðY³“Ke{éÓ螨<±¸³t’™åýK¢cÄñýÑPœ±•æâR(³w  ¯.KTH*­ !Šr’)ÆZ{MCÓ’ðh£[7Í‚ÜßG˜ôVޝfð" ¬i«›³®°‡‚o!š'Ç2¾µu Ô×ÿlV³µ ~iûyÿdÿåÄI4Þ|œ‘³‹ù¢ðƒýÒî-{dÖEˆOæY|H•kÂŽlmxi…r_~7§ÉkTFŠ?&PÍ•~ÊXkæ |ð†dŽÅÓÇÉdxs‘ÇXñµ˜.É¢•\å1KW¿2U„nÃÃßµ÷ÈÅÖŽ™›âñ—4¸Â{¦œóÒÁ„Ý·„|G“/£^;Xm&·o¿–ceÓîlü®ß*þ•\#ÍâAä1Ã=OƒË'”4SX¦Í˜mÑ ‚¸cðÅ=wʋʡÃɶß¼/GK²8V½Ú%Tåãë*îd¢º¡]ת¢8:ô›Uø[Ñí,Õûš{WõO\úc¶tâŘ®Ê…ÅPɲܼò46éu„¯1vkÿŠgÑŠÍùß+^¢%r®ÑX÷K‡½·Ðì{?‘¿EÁõècº_þn‘•’ ô7b±eƒ}ǽxç§çH#»’-¿ß÷µÈŸ_ÒÛÑ7±†„£Uò²Ø‡—ší-N\wšb`b=(ËMÙ™Õ¯\çðSïËT{*¶•¡¹ ƒP¡rÅ‹$0?lj/9 ¯*œ8¾Û]ÉuD{,}.Bi6ÄŠL ýÀ>ÄçXæà~î*¹ü'Mø†K"]qÏ­Ú¿œÅñÒ|0BX(y‰XÂd§Uyþ2|.‰ Ú;‰†”rIiIžkÅ ”^JáSÖSGŒ¥ HŽ—wÉ­Ë@Ë¡l­+m,<I¦wkðêäÒ•Ñ3S´>L¨_CrÆófØn— õöôhdÕ=zƒ>ÆõD¦¶„u+”Šš¸¸654”û±ªY–¦¶°† E^ý-O¦(µB|h§Á1Æ#¹ °Ü‡ ‡Œk¼§í¹ÚÜ lÝçÒšGYæ±G¨¾…€P[šú«>R$¬rãÂ;2™{õBéœéËuˆœŒ(ØÛ¦by á²Q³î–wÓž ðFç,̪ }Í««—p6ÓY9vR>áÇ‹Äkuw*XïADµÞº#¶¦‡úEwšj¼[Ó+ðó%ƒ³J±G¥^H«*5p;±HˆÔ.õÚ`“Pa¨ܮYAÀcÙ~ïvÞV&+ØÙ]£™Pòn •²Ã‚Ç "Ÿ/£ øÝ— ÁùêÃ…³ü¨3º=ãB:ëþ0¤ âŒWɵåøèÀEøB:X03$µ<”‹ü#Ú#?a‘ã-ºHš|ï’#ÅÕ!òlásq |wµ4j:NÅqÿâ?s÷»ÐÝ6þU£eK¡füWñ¬—¶%îKj`9K…=¦„É|åÛ&Á&ý >‰lKµ[°~Êá·3ÓÕj…’2À9©ÄŠ‹唡€# öî`ÝWõäîû ­Ôyeü¤™ÑíÚŽƒk2vkWßPÛtœžO­£Ãø-˜úþ»¬>Y k5» Õ/eôˆy´¾°5õ‹ š+­pü>Wî䤖FN↓>h©5„”feÓÚÏ&¨©ä~¶ù¦_^ÿ—káûY2ئí@_½ÄxpÆ-ÔÅTx?-‚ݱ†ôHp#’øåÖÛ!ÛDÊ ÍZøÂÒåÑK”¬š;âú©Ù†õ—„—Ü5Ñ'\-ÎO4"ü}…d[á2ù‚qNB³Ì4V ufî'SŽ[8å1P«ÓЗÎe'ìß,H«4ÕÈ^Ó·ùåBdòP­;?:JC+õ»<û…!/ã/€Å[‘HÒi©ë€] ¦Œþæ"‚û§~qèe7ĈìÖ÷¦÷ù Î ¢~«¥šâñ¹}V‡b£Ð¼ õ8lÄ0ÑÐ'„@nt5%<™àúd#@Ã4”ñ8%P.$”ÀÏ Fè-¼º·Eðó®wž`ÝÎn_Q9õû¯(±ºòÚºêx[×mѹ>áKÊÇίIŒÚ¦_[¸Ic߇råØÏ–òRa8r?t%äøZÛCA•7õFþz=f,¨/o÷aÝGO£ ä X«è-ÿ³•KêMOq-:B 3}0°iÏÇ2”y,˜+s­ùž¿¿„²ÔÞ¥{9Pncqƒ€UßcÖÎò¬j…õÆAk ýÐ`GqQk"储»*þünJþÃðEƒguîß«P£{݉Ú˜™\%û&NMU×c0Md8Í£>æpÝkjf^–b,ÜÑEúa“ä@åÔZ>ÉâIw@¶þr-„´.ì. 0§Ðôè”A>R|%V \gL“––]PË¡-‡]2ª,šëõ¹¯¬»aRsr[z§¹3o‹©YðÇø”aÈÿ%­$òKhÑ™ÒBk5ñ? ßúBO‚uˆ£[Ù¸@ÿˆÌqLM½ð’°èu/Žþ@É M°««£æãD»"U&©*Qèû¦'³L ë—ÚM’ÇíÎd8)ì¸Fç#€>ÓMÉZ'fÑ:¶ÅÏz8¥~†ø#Û=Ñ7ÇGüøöֽ໑ X»ˆdF'']…ÅF÷¦^eŽÖ}hËýùjôŠ]X‹uiä#>¶Ú¼®`¿Þ ,m ñiï¶@{Ò BTâÁ }ÚU[˜:‚‹Uyq–"•„„ aNæýŠƒ^¸rÐö^ ‡1J¹.Çâ·Ãæû!¿ÃG|”}]oY;a^»{O {hþ±ðrÃuî †…(¸E¼;“D Õ‘Ü…¼÷¢<6'G"EÈÂ¹Ò jÁí»ûU¢Xj ØQ–H÷,˜àÁÔ³öÝ÷ò‹£÷¿Ü=QGR´ô1F±òqKh;Îõ\`Å¡ÒFí–-VšÄHfrÍhª£ú=ì—«!;g8ŒïOʇm®çPöÙ­Ñ6œB ˽(UÀ»!Û7©Ä?&lºEü@1žÒá ON¬¦ ú“[ñ°Ã| xm³‹ GBÝ’9ku/Ht}L%¸Q` Z3­óabCl#}¢NX§ú²kp¾§Mqéö–f¥™kãhZÔÀOaÔ:Z{V 1•—¤• FKk®\ÆomÚ¿$fGmmÇŒ…úþ.À^Óëx¿7QË)ŒÜn6ù&9€Ö,8b•Ý-­?F‘cZnéóT.ÆûëìYƇxC>|I\À’ˆ±Sþ¿>àèÒgU»i€«ZÈÕØ»¾À>2ž×T`öï6  Û¬9÷%«äs”gå6¾p¨XàSScmÜÊe»$™©KâFDsô:¶ÈÈÂmê–ó¼ŠYsKqÞgzìÓË£h ª|6éŠظեš5‹¦ÜÈ"N&#t˜D§ 1Ìå„)ŽêX#"#·®"‘Šà¹K6Z9š€ªÉ£i!Vmv7ûjS,.Î2=<&®¤TÏB=..~R˜Þ%‚:8|䎖›œÐ?ªy8iòC-ΆºØ]y¯t7Ä–!Ÿ_±ôui~©÷Ð’ N¯ ÕA†ò¿Nâý[ŽŸ_½A^x›ú'ø™à­ã´G!vö!«€ šzjk`î›-³ú3õØ‹ü!dC¹™†7_Êr¾Òù÷ç/0û¨h¿|^£Ïæùûqa­RoÐjÓØDA†¸nuÍà¥ÖVüÃLþQ¹­ˆ| §¹’ŽéŒ¸wBšŠÁÎA;2I­üŒZhš-Õ•În½y-;r-u«Roó±‚üNÿ|»s*Œþï)Û& šÎ-‚\‡)$œÌ1ì,j‘X¾üÀ<®j†-CRô&E.vòPMâ*®ðÅ#Á¤­J2Û `¯ïy…Կ̾R/Õ$¿ô“RîX¡ÒÙEšî‘KqVÇx‰7´—u§¹²»kHU 0w ÒmÖvHÍ„°Î\¢¹h÷¬‰ZÏ,Ýþè ¯E-±kÏl”Ú쓞v¥ÄCù;ÅTÙÀ^ÿ*iãø‘Ä#7„öóƒQO¸Õù|-iÒXf±ß¡p6“IÊÉØ‰·.Ÿ¿;¤®b‘:ªÍ-ùµ'W{ Ûµ,þçÐÂsuièw!•ÉJ›”ÇÅ›Üó¿Å¡jRÆúD 8‡~@b²_±òü9ÑÄ~Y¥ºO—äkñx ¬; ŸtR%fê[  kØ*LÁU¤ øêã“3Ów‰Ê©åXbsžOö ÛÏÔc…B¯2ïä:diÀÚs>;®A>Ó4 >Ç‘4 M´¿¤½Ù³úͤ;‰,4Ó'9êwDÙhÏÅ/ÃÊ€’ ššOR å{¥µàÍ{÷¦ºÍ‘‰ ýþ7AlsJOFSwÀg‹;„x6UP7+&ö|¤é7Øo¼ÄÚ»¥ øH[k,鬊HrøPFTÿj1£’T,:~—H®Ô.µ›4uf§èѰÈÎïr2³ QŽ´lÏÜiê'bL¡g¶Èê­„eµ>±ª‘†{.„3V9À1oî䑎Òxûµ«ÈµæiD Õ¼Ó¬à‰É*寠1[ÄôGœ6(~¢Ý´"qEAïÌ™-æ]y’}» ªZ7zјkÄÒEü>„ãBœ¤Ù¾‘Ž†Û£ã`ÜD”Ûa.¢/huKY²9o:±"FRܬv=»'µc³ôÕJ - ’”c8Pú“ðp‰kw¡Yj—ÁêàNݧRñ›(¡ŠG`×ñ,a£6â BOýøÛ*-‰XáÁ‡ü“Gü%16Àž,|Ãù ZOþ£Ô½4áÙ^>úß6ÿú—NX|NÔ]Ì»hÖ÷cè²:Õ#z……ÿ¦l›Ï^kÑŠà@ÝqÃÌ¿Îe¾Z »…—còÖ*q<ÅàÝø½bŽH_§‚ÑÍ7ÌdAnþkôXš¦Ë}æþ;S”)¨^:„5c±³JôJÆí—Õ,ÆØbnï!ô-‘)U•zYÝO‘èTA×|2¸e£64qPñžRàËIJí`CLJw‹HWßçšaÛ,¿6…‘¦Íc[=‘ä,Z(N1s-SžìÈ šÜ䢦íjWs¥2kðJÔœ¡Ë}g “‘½aHúkj"^[ˆïŸpÁ„ XuXÿDNwâ•ôDA­ªêVÿ)åü ,jiR¹×I”쪭á¾+;G<·¾›• ô|yÁŠ]Jï GÁÝàgk^¯Öý{{®\VtÇCPËà\o±ñŽbîŸ÷RÙo;ÂD‘NAäè;k"›Sôò=„¶K×»Me`~ql.B©¯oî÷œl@Ç­„›ŒfRì+PðÌ<ù‡3J'y¿Q››Ò»y¸ìw×ÞA…ö’¬kvÌ7¯¿u¸ÌT ½nÈ7WËŒðl,Ù§yÏœüóTÆà=h7öEñ™Ë$¿i¿ÏIaøïÝTÜ%OoùžHbÙë©gìR,Z¡ª“¬êrN"B«¢U…Ø ¶¯¹¡¤ ÿŽ¿Ñê3ÒEü1S"p‰A‡ê;Ý ÞfHOd:Ê·Ð=Ì:H#cEø„jÜM6¤£§ “ÒM0ûŸÒáè7¦o1Â$áL¬x,¦Ckù’º¨h÷=ãAñðÀ¦Ü¤6t^²…Ê"»»x2©¶‘¸‰ƒM¿%ÈÑv„u¯‰ÓûË›Sn×+5ÛA²ÑÀJô_2£Eà5ñ;Ã*þ‡¡ì_N¬ÐvÒ=ö¡û©NæroÇóö/áà\Í ·Ùkƒ‹XçÌì{TM«ln#DÃa¡»f*@ÿãaIÒ–g™ApXÃÒ1‚#Ü‘Ž‰ß HOàR#k³J1ÉËÄl¢<öª¾v¥LÊ$¼^2KæÿN¤ó§lŽþ K?ó Ù}$SÆê¸ÐÑ‘ 0©‘k§ ˃xÁ)xÙfÙ°‚½&´@I·yK®éŸ®j¦|œBƤEÎ¥ä‘17ƒ@ôm»pkºà¶÷:¾*Ý^­Ϭ¦I RSˆØmVíÙìv†reË„„ËåŒø=¤µã¶ÆXV;¨EúˆdžxZd`…Ô«z&7ìE–ÔçÍ*8ïLeAÏGÊÅÆ†åwuDû4,Ü댯ØÊŸ‹ªa8zÐ` ¹WþŠ„N{èBuX3$b “ÓSV—B­þ¿V,Ý ©ñfÅoõÕàµTO–‡ûqTyh:­Îú»!܉…fn–Èëç^¶k» €õ ^˜ÒP7Éÿ¦çO^r©/<,,9áÙq&ùQo–û@]”¸Öç¦<÷áÞöûékù¥dSûTö‡Šhv8œ|p?ûŠï½˜†¬Z ü­v¾„kè=nâìƒØ)áŽÏôZ_?!»8î9‹rÜHºÊ5î¡Ñtнˆ¦4B–g™ ,dî$•áë|‰"s)Áyãô¿{S,|ÔV†tmÈØ8u/–X6Yáѽ',@î–Üœìì½pv- ¦Ò¬„oÂ7‹$Ab©¾+(Oš0æ¾K;°ŸªýØ(ÉÇ68ÕÛp\L±¸“hdÝÕ™&V±2šxë8håçÍÈAÆý®[§îx'Qi|Gë>ö– ¡-Ãëv9&v ²iÂl¶—8)vtµY`%OÛHè”ÇÀg0؇ˆ™Ï»=êµÛöý™«±]°nù W§úœA¡ OzP/ëÉR¶‡.HæÕ£]ð&@ŠS,ÌÙL‚®=ßD Õ†§=ýP›¬…É$$l=·÷¾0”¢ TgKŽMEˆ bÛ‹Žd4€ÒJÕáËý$WœÎƒ&á†á?} PÇ(‹n€ógdšQð¹sèJÞ"qÕ)Êþ‹¿±Ewr”$ÍxxEBçˆÇÙ¨]MtèÞpí>¥ÆD#ÇŽ+q“þ·!6H~:oð€WˆR.‚*~ìï,®†1ŽŒM¡>ŠÞákNÙ0çþ;¶ùC4²ŒA §kƒø2ÜÛWîûJAÃ’±­õLúçv)7ÀPý­ kO¼ZRèõô*­°iƒ¬«Çkç€Õ¾ Dw¤†7[ A}*Y#âš‚TÃ0>8ò*õ<Ãfƒú@ßß¡/–‡h„f%ÉJð¡“}@÷ ^,®s-!sÅð\Ä·÷¬w›Ü8rÍéìÕã>‰Aß0r•§pÝþ%ºÅ9™«l•›5°àƒÕæÉ÷<ÑÇaDÅê €Kö2­‘£QÁWñÊ¥<]ž_. nõ}ËŒǵ¢—ˆ®±gÅŽ©Ag×fìÎWÜÏi¤Í%ÇÖz÷êaÈåøŒI{Lÿ®åúŸRdÄ`X¯ì3 dˆQܽ“ÂÅ1¾94ŸP`P8n•ÜÌ%cÛcˆÔuC>eQ\‹æHbFLxÜhÁÖáåq!Æ;5x7 =¾’q“y™©Œ}[Ýiˆƒ©6âÖÈ`gÞX`¦ÅhôæµçˆXû-ZfAt-,:»îñf“IÍÞÛN2Xåã¦XÄÈ#Ûì×\ä ñáäÝ6’eŒR¤†ú „/æï,1gÄ 2nßðæŽú°ö®KmÁ”ª$§ï²7$Û\Ž9Ò¿Ï/òš·–ð5¹@úÇV7% zøÑýÝ×ðEA=Ô“œ¸ŒxÝ\TÙ*\wEã°[hƒÂß¡ahÙ”6b¼¾ÇŒælÓ!‘V¬Ì›Yu‚/]³€n0´õˆ`+™ì2ëæûý&è3¦bk®áÁa‚¢BbŸâA)Ñ’ð:WóqÌ3ʺÀˆc]Y«œ¸Bû ÷¿ˆc;Ñ ÔÎ0¼ÍÈïc Wj·°$‰©Ï8ëॵ4˜5°2dVûXižD†­9í‡J_92úTCén2'npôŸl8åë´\ej6|^„‡ñy²*Rº­ªW“KSH ]åhÅqÝóqì™™p®ò)#—GïUáŸ1jHÌH¬µÊFZ¬Zz—Ù˜°£K[èsÃ]Jfp»Võhq÷÷mÊ Ëú4JÕï§v<™Ì†žfyL÷Î)9ÝHèâ¥=ÓÆŒhWÞ|LH²µÇ5cfÆ!\Ì}s8¸¶rrüy!D¶ &rç]S|µ<k¥Åk­^Y-´‹£¥‹Î¤P>¹#­^æœ]åfDø±Ø–$`Þï3&×:¿-*ˆ#Y^¤ÑÓbœ»Q{áìc¸™r‘«ŸUÎ m5âq/Јð2ü Ñô_ȹ|a.Tqñ›²áY²tâ%v®Ž°ç=ðEš¾ ÿ-+jZóæ ¦¿8­ ¦Y»Ã—õ•ë•J“ªõôü¼„F€Ã¶vtod¢#èBlÄß‚ªš¼H¸nm÷iÑw÷…síÒ›1wœã‡I¹"Y‰TO¡°æ»§‰‰G鳇•ØMkx($ëd^jeÙ…ž`(È£‹K„~  ˆ,™€ÜoòØo1ÎÒFÎåL0_Sï¯i‹ñ=2(*£¦•]LY$,’D~‘M|6 akÚÂõÒºþÙ·MdÇFLE9ot)¸ð̾ ÿ-ò©úÙÕ€>;ÙÃÞ÷s’¾]%庿l“9¨çj,,ð‘ÜRìÆáž~â›1ÜvÜQÔŒ6ýLm8 <ز›ÄŒ¾1ô>ß¾6’®b·™â¸®›ŽõÏ0F–ŸŠ'̺ñ­ñã«h·‹¼ uüÉQ‹†K"HÑ2Þ²yØÑ.CCÌõ‘¨8‰˜^²®u*‡! ìÜØX„¹¦a½þ%ˆ³¥ç”Zó‹·G¿DŽDüõþ‘ðŒ/Ò>þÃvx0þ©]ó0C:zÆÓg¹þwKmÅÎ"Þd¤Þ†úmÝU^X·û]uvnn‚SÿFV+ÉÅ\&æÈ-äH¨œ=·"ßUcš6cÄa€ ºüe¶ÀÝrànîÏÙÜd>½T^êt[/ Hwé¡uéª0ò¿Ðì¶EØXx…"`âˆÎä²¹À9j®ÔòÃ[?¶“ÑÊ^¥šºè‚»Ûx¯oím#ä|äKÿv˜ÆùÝë0£m2yÅ3óádm§*oZq“_LÕL#ù–´EZ°0ö2ÜÆRõ,‡³I3v°°õìGÏxº¾‡s(ýÛû~aHL¢“ëMÁ‘f%;–ìtø‘+%¯ Xø~bǿɸÔÊ9u²Df¹Œ8m%Ôv!9ï¡X|Á¦°ž*Zõä˜sê}6|ýí„{ïé6èt¤è+Z¨Ã)P8²`S¦äŽT‡)`Ó?‘§€Af¥N¾ËH WJIAyÿÚçðÀž_|×Âç)±º{å5ÿn¹jÍÔdô" ò²b—EO$_¿£Ç®› sëRw²ÊŸ•0ÍT¨çOÚBíÐ_ßEjŒfz .^?F.Ó2Ïm‡|/l>° Ù®æ:âÍy€àwýãÐcãË™söH5'^ua+$eW´¹Â<¼WǪ̀A>(ùK¶SIjR.£}„‰a*­µfÀÊòîo$.e¬Qϰ¬õuÉ©±†s& r€¬·ªÇ t¥Ò3+ÆÌ£¹ôÄò3«½¼æôEuà–Zz/„}×A*Êþ¤8±ê.2ã`É$[ègÙÅØþ{ËôÖF“—‡°+I¡vAF0 ò ѵÓ]*/r“œË¥þËyH?üªµP‚ÙþŽ*£ÆFšm cÜQ-¡·ÁEm$‰|Y& g ßHöŪA‘ÄÅó9Õ¾¶9Õ§¨­f8_ßÑ 7&u×€pw¥Cáx:Ído Sz€£“Ê0Û‹8÷GõoÝMî¾1ß»ôdeÜ¿¾ZN{VÀGbäbÿï=4kxšùhà5G·æ-—L-ÛÆN,ªœµÉÐŒÄÛå ]‘ûif,DÈËÿô²@QáE5/‡}Ãwx!»ÃàŠ±h¶h.5ÚÀ–ß™ÿÌdùŒZ?×®}LY‹òÛ›¯õÈfVäJ“I7rpxbxgE:° ö˜‹Ü?6‰ÂØGdë<‰:¯Çñ«Sÿ*ÕG¤¥¤ÒÃß_Ô‹ÅóŸ“ã×î–>š*IàÜ5Ç‹[°Ò1á}w˜ê± G™(®$J3Ä÷/$°s`@Bíž ÎΤsKkÏ¢0ŠÓ ¾æ»JW–誩3Quc+ò‰Ä ‡+Èh¯ë…)w¥!–šFæ—ó:¼jCÑ‹Û=éeÈC9W ý¸>ÁÊ$c`â¼hõß(&T÷P8¥¼)/ê\) @¢qF$÷ó¢ÊËŽË{c¹š/²ùaÌ Ag¼¸3À^?M§ö"ïûåÈuÇ.ßÙXñØ"°©äI"š§ÙEî9-õ ѸwÃ"SÙ„"-AD!wG‡ ¾n*éºnM>mc3à9ýG{®]ôy:¨÷4½LîÚ®QL y½@DÐKÝÏÁh<,7›íÀhš;”ø4JüØgNÌjîŽæ‹¼áN$eƈo*•©M[Ãìfùøg˜åÃ)K»gzU‹¸²a»ÄOðw7Úï¦Vq¾ÞØìd4tS$ûR«ù¦«TÏ™Ùôˆ‰ÎôNûqüÆ€IŽÝ°À¥!oy|g¢‡ô3 >ÖŠ]ÄÑý#ºË2IsÌ ˆŽ,Á±°ÖX—ˆmÚŠPÿò`0`RF¬RôÔd”ÙV©ü÷h+Ö¦T,ã­5ÉŒ‘uÖ½:Id†}OçÐ .oáÙ*õ?žt"1Ȭƒ÷‡FxÉk3¦¼\L¿)„Ô¥¡¿rq—M‚©âcþ¥¼òÓ¶}ÖY)x½n‡Ñ—¥ ‡ì&\d”Õ!ºœ3Éã ÍüXzߤ]ÃJ¢Ð ûóïk€t«(«Wã)Sv´ÈþÜú:c t½wCtÊi@Í¡»¸÷½#ŽV×¹=܇ɧWΟ‘ºÊÆà˜¾"q7»vä Y¡‰ p[ymÞ´Æ,TÏÀ ½È9¡ ¯± ¼y¼îûýÒñ }‹¼3WŸæô;uRÉÂ}«CZè:c:#€n·o2wú{ϧÎÙH—¡cÀ-…Z_bõ)ºË7¸œ&‡Ù9݉£i ÇÖ|˜.uý§×¬Ô ‡SÖ&6`Uœô³8B¶ü,7“X/÷Ë;iÌy²â<ë““N°ÖÑ÷ÂÖüóåÝ$6ãc§öƤè’"ìAüç 󣌦ԑ+’-®/õj„ü»uIõþ)¹õê㿞þ/f[Z¶x±ø/'GäM•{#aZ¾¾ŒôÓ_íÉ%¹ HÄN2MÇr´ž&mÊ|ɽ™¶“E”]Au$Ì¢±BÖ­Ѐ"OLMÿ:sÙñNçx󒃩kÌ.„Ž: Û”ͬaÿ¹A¸ø"Ë(ß Éµ2Í´¨„JÙ]=€×*ë_$ÊgMPšåÊ€Ööc½w˜¬§ªñžò9JNÈ™c¬¤#­'cß,-¶©Á@¥ýrÇmsûörpÇ©ì);¤vÚó xû¥™”×3e²Ç-Xì{²me ý M M¦†Ãüƒœ¶Ùü ¯tÆ„ ÈÌìsÍ–.æh8—‰;û€ØïÎOg›+ºdÓ*ôÀÒUîÕ~ùª%xC†Ã[ȉŒ¢4æê è/ù[Oî¦Ð[j•:µƒ{B"1î_²N²¸¢ÆxÓ(Ñ@ì< ôùîŽ oðÖ½–ŸS ø«€ì"ò]¡®´±Ý¨A{27Äúä¶B˜S×u„mÍÌ+Ùïè'ØP±­ *¨*f€‹üÃɽݡ„â-´ Æ9MÈtL¸TߪY™"o¡k—wìÙIo¸½æü¥èÔ(ÆtÍeÊ’_ñP‚Ä rS^+ÔY)´£Î R[ß<#¦ ãmp»¸æ?jEØx©`Ê3º³VÁ¥vZL øµòíDÜ9øùБyoÿ£…VáçZÏ”¤n ë™‡âPÏ1ìÔ!x|gY/QoÇsE©xUì0rõ7¥#ž)>ÁÍ"úš@[_ ð×O îwçr¦µº Š ·Ê)A$æo÷“yƒNÌËáÌ Õ–¦îPœ·@ù”ܹ¸îÈg¬?£ÃÕé§;8¸bQ@¥È¡®Ë[™µç2ƒžöò¬Ï)8mZê@±owB?et¾!tLlìF‡Â°?²å>«+ë"""@*ä³ußp‚âüssï<6¡î·á]÷K's`8€Ûî+ØŒ³"Zï–T]BVÒ(‡‹A5nŒÇýI@й[µdŒžfÖKé½CpÝÛ‘Ì#®”ç¡'4å_'ˆZ éy2‘–/L%Tfac` ÜslJGÀ£¾ª{¾Z91þmÿÓÓäMUëFI[1Ë9>¨BÄH$¸ ²§Éþ‡ÉDiÆPçcV¿j¾eûå„Sbf˜Ü•œƒy…ÅÁë©6‰,æ%ïNß0¬è…q…ƒ¾ÔÔ MÊ+IÄøKÁšé»+{žn¾ÕfdÅr̘‚«IOE~ýd(Ü{«~=c©ã£Bâ3B”es\|¸= ua§RóænŸë NCéáælœN'ö^è?8ç´,ò@”†éx1îÈÑìIJRÃø>}‰Â+PÝ‹–Ãk_¢ðN7³HÁŒ¨†6‹Ó†5ˆõ&åÒæ0¥Y@Æû•r´§ð|—ÐïxÝ ¹aÇkJÙ•N3Z÷ݨÎÜ©‘èAç¦ ½ÙÍÄáŦ½½Àˆ[Åx¯ˆ‘AdÁ‹eJÏûÉD1¨†’©Mò5µF­7m™ ¥œ'30E^ª }lz_ÊJóõ0×\R`Í€çªî¨bëÚÏØ‚Åïá.¤Ç§˜ÞÅ»¥àfýS'T‰É¼}–ÓWBÁÐt¡b§©@ È?ˆÙ%aÞ6e´C/‹2÷à ´e;E?\&-ï*<Ìwý,dafgwo«Ù]§lLÙ >[pª‰Êõ%B{>þß ¥ñÓÕã±ÊŽò’’6ãrÓÖþÁ‰ ú€"Ó³jH³È¶«¨‘Î,þ`ÿ0ã¯;ŸZŽš>­5üìŸr„c°fÏÖ=;¥+é ”:g[ÜIÍÌD‹qîßPÜÊÅÅhÕ 5eíéG àï]ºIp«™\ G7–€|ƒfY&ò,,ïs1àÿ gb1h¶ÆÍã;jÊÕORä |ˆ n…Øþ^ñsôç UM–OÏÈÊ?¼Cõ|Ï­vó؇ÕÅEÀ¨=%öÆ.RíÙÒö»v£‰¡%ØÌ=¨(Ðc‹ø©oÿlÔ}Çq¿Pé±T$C©™­[D’{#Y‡Ú,}§ï\üR+U hÑ€…Ú}wN‰iĘãË>Š;—‹B`új²AËžð‡ŒoŒž¯Héâ\£Þ+ú³xâkHÐàgJ²„´—Á©¶J»ôH÷klãMQò¯¼]î«ÏLº€&gORl »Âê²H;¥Ä{zY®¥4è@#‹ºZ.ÆÆ^ɶ¾× †ß>–™Z®@ƒ±6DÀ½-í­¢ay=s½±Î:Ç6ø B$ËgRZ‹ãa²V3SKhËÃ$û& ¶»ï†#‚C'IIQ³Åyèȷºì-çí÷ šÉîHÙÉ  ˆÝD€ ±Q·zfÛåT8sp+%s¼þÕ”Ö矽H˜€»Ý6‹¹ÀÿL#^7™DȨ"€½ŽÝØ‚åX #P¼E¬˜+kˆˆ¬lÁT"ÛÃc )(“$œwTôt¶ \9ÄTé;ï!«¶€F‡n=pUnqè[*—8Ƨ… ¡"ƒÞPĉO0ZƧ±l¶;Í)/ªW®Gëñy³;ž¼•ñ2¼,+9âüÂy¹×ô«þ{û $§ »Ÿ'<_x¦à³@¸µ|çÐ]x¿¿XïÜ$ŽžUÎ[.ª«»f’>Ÿ¸E½â¸sA:“Î:b(¤Öt„ ©PWHökœ³øóÁù ÌMfÁX0°ÖŬÙ?ú>÷ùu98;ßø2·µÑ–¢-77Hx÷:¦à†Eø–+En)s7m&ë'Þ‚eÿ~~íðr$+†XÛ Î¿ª W2[vz½^ÞØ†©\é¦êº—ÚŸQÄŠ'xuíÀŸ–™‚ˆÀt²†Ne°ÛÆB›gSvÅïwÁNÍ·­Eªá}ÓþÔ:X•×Ri£Ku›è¡ˆB ~:„ê±.÷PtžØäý» >ô°û”ƒ‘Kk )õüXâÿRƄ殟³7ˆÿ7“j8ǼŽ`ÒBJèÛhëo.E’O;#¾Å"Æ›vÑpcm"wžæíƒ"/®¨ÕAHdƒìÇŸ†Þ@o@W/=¼ÛÅ!‹Fî)zÌ}ï[ü­°×­+d6lÛì…J(±%˜·èÌBµÔPy}ÙÊþ±B´rCó«-`æ›àÇ ”ûèçƒ$ú1­õü‘â‚ÅúœgbË×zDmeQå¥EѪændܪîbëý ©ó®Ç-R(ãa…ÜÀmðUñ‡¡%k‚#zN‹k®NusVè>ÍyÝ@Ê#(nxÐü-")[Ùè ¡¾Çš©}\ºÝüoqäÕŸ?ƒqXOžÛm 7bÇ:Òq¦­-UgiK·;ÅöVÄ@˜ ގӣͮÖ}Ò÷ Ä#`¶ñ¨@¯/,¢ü†oùbXq‹]Aê×Uºü“ž`/©¹ Jãa©>]¡Í½‹mdd黚FQ «ä÷„™J¡%Æï¹zýÌ * Q’àÁê¢(ë­f¥øVe/ÜÖa2÷^Ü ­ý úBÝvˆð){O¿±=KM_¶«NOø¬#C¾Çzµ©†NåWâ@4Ò2œU@0³½ñ•Ɦ-f˜ý=³˜ˆÆêà[Œ2ea®°"7•5øŠ²Ø Ìe¸M3O Þe|£ÄrHOÕ…ÌØñÞ‚(œ÷¦û½’aGïØÓ 5×𤋻IÌnÛwþ9‡´l¤û¬ÛF#¥P~…ÜYœv}f7ÙPuËsÝ)=‹²ß•W°â¦ðxÈ{ Ë-ö5IùjQlúèˆÒ²C‘`%ÜMu"s¤EÄYlO‹Y²E–`Q¯ê G€w¬nšžÖ¢M}Ymqù€L:„+i& —)eÍûdKovÎ&w$ñF£¸Ç‹/6JŽ¥©¢£É4«À¦ö”6å¥!#^r§+X¿v€+™)U×C PÕÔÀ kŽÚÿb•¡ìÆaäÈ ŠÉ±ÇµAK* °iÛ†²P¾¿‹Àjym Á­ „m2]ô XaîCßÛ&ž«¦î”J4¡ÚàÞe¥x'_ŸŠæÙj=á\Ô££Ÿ£è©šà•˜–½d]wpT‡ &è“°’€¯øÖ¬|9KÐG£ø9Ìj—ÀE ƒŸ–“ËǙƓ‰ìé\±»oë…÷Þés)¤û®']fü÷'²U_ŒàÞ׿ˆøã¬Á˜%n>Üçž\| †7×s·Z.Â3Ñ"ÅÏ¢AÙóçA\øüÓ²½ö–œY«ª,±%áȘ¡©#ýçyÏ"Ñôý™ì/à¤CÔ¢ý «oq­ùb"ùâ²o…ÑZvÂÚL³”l£b‹$nemÇBFCzÿ>Y½h|)ÞJØÒk~ d>%q úzRûÅAMÓn ˜J0ôæŽ3)£}µ"‡7ëÂtÎ&Ù‡›vi¿aìðI€…Ú‘ÚXg ÿ@æfŒùx·d 4¦¼–ñ7ÇA’°|̆lwQý‚ üIvÝ:_ûK«Ït· mñ÷lÙ^ÔºTRô+Àé}ÊhÔe˜à×áW¦9ݤì•w)Z0E^ï½ùÁytï MÑWÐ àÊ“ žeÍÍtx»¹ÁZNOçt7ÏGjb£ä$ïH™¬9cüºdFÔmš·õ»µÅ±'šù[‡|šãƒZ·YÉÜãV¡jî8$;p£€x’SË®gض–rù¸qNp=Yöæˆ?éß=‡¸†z¬+, nÿCî©J™~ÕúA‰}ÂŒ…Õx×¶1>NsívÂzël8ó‰Í¸Å¥"ºLZ~Wñp :”ž.„…î• éŒ†15ókÜIÔ€¡àÙS3ÑBFŸ9Ví{ÿºªQÌ­ççí\\\Ëÿaj»Ý²ÍVà ˜z§v+ýt6a/ ɬøÒÁiËñ^xßë_¯ ÞeSïåTÉà‘ |O»…Pe¸;ø4) fNòÏäÅ¥k·ÍÉw‡Ë#µÜi®½6ät‰Ä®GÕý «Ò"Çä~K²k(µ ±RÍ7KÉ3ªg‚ksCÅ(3©¸¹$3Û@ÆŒÕ'E=3Ù˜6)1G˯åó|OØ€ƒ7ñïb’‡ŠûI7æ`¾ãõF‚â¬@qÚ{cq§÷JŸ’™(2>‡_´DZ†—"û×Xc.½hŒŠ;\ K°Fd¢ÓÀ<Xa}ª8j#ÔC#EÕ¾2ЬOq—YqQUÜúÒœ![®—y/ÀàÕ«|•}Ó8}ŽÃù›Ù´.XƯ²m -¸Bx)ìc¿ÈàôLôÛ²ÌYëO4tâ¨Øqþ¶ñ†<ÁF£¤ä6Y°€Æ,¿*våÍ#\A/ûÑ.W´¤ãd7†OðTÊ¥%£]+vüÅ ¿®ÔíšiYŒðŒåÄ>à m÷gÙÓF%aãÃEÓ2Ç~'֙ȻI½A‚%1™ã¾rP&=Ø‘Yt³Œõãc0ÌÏÓrêŠQ*LõaéŠ7åÄ–±Å©öÉ v<„Á zSì0éäþÄÓ/ÂNöÛKè ÄX”©ÂBÁŠÕ}CWâ6J²¸> stream xÚlxst¦Ï²nœ‰m}±mÛ¶/¶‰‰1±=±m;Û¶y{ïsÎ]ç®»Þª «ŸzºÞ^UMN,á`ïªæåd¦gf`â¨J¨©13q³˜XaÉÉEÆ®VöbÆ®@€&Ð   t0³X˜˜˜aÉ¢Ž^ÎV–®*Sê+¶ÆfVvVÎ [w+SKŸ»»»»‹ƒ³›Ã?Aª@ ÀÕ0·²D•´¥$T’ êI =ÐÙØ äfbke ³2Ú»©æÎÛÿ,¦öfVÿÂäÂðoöî@g×€™;;ØäÅÕ„%Ô⢌j¢c{3€œÔ?áö®.<ÿxM]]ÿ•Àõ$[ãÿ‘LþGú«³ç$Xff€™•©+ÀhaeËø/Þ¤íÍÿQ›¹9þ·é@.ÿàPýÃ!5À hþ·›­­‚±@%ê`çèæ tÈ;˜íÿâÞÃÙêÍ»ÛYÙzýÿËCÚÕø2„í-þ¡é?*+ +O ™’•ë?|»:»ÿ£ÖþçläfVnvÿ þÃÉ?§jæ`oëõ_;þ“Ê¿÷cÔÔÒÑVQ¤ýŸ"ø·YÜÞÔÁÌÊÞ êúŸÆÎfÿ£ø·YÉØê?ôßXþ»¢Ìÿw-oìêlå ÐebøW‰0ýëûoIÿÿz‰ˆ8xúг0±èY9˜̬ìÌ.V&¿ÿ×ÔÍÙhïúoÒÿÉå¿×ÿ.& Ðh »ºä`ÊbÖVî/^8[I;;m¤N»Ä º<Õ}­‹b5hifxp‰{§†›çO0h¿s±-}×Dbü 41C¡¿„ûRLZaáÿgóM9ÿµ.Âà BÇÎTâ:òÖ5cm5H‰f’-3çìŒS)?Š$sŒQY{$òVüʃ½$µ>%êo/xàm5ò¬ehÁ•‰àËî*Xx^ðü59Ž¡eAy ãA¸N~Ý¥µ¨®¶Òs§V®î>ŽØfÅ–{ÅgÜàFðÝ%ë.Tú†‡ðÒÞcãðõ¨_›ül~‘;3îÏþÓššz”NQ?]§)õ·dæÅI`ƒ8„à|FìvyrÖdÿºÊ9­ ùŒs¯­h¢Ò”[Í&n`’TF·F%Í_9Ÿ¼¹àðÛn“‹¨¤C¹®›)œs€ÛC¬›åÚÃˉž„ÁŒž´¶Ê$—³¾<1ŒZ0ÁÙÿ²L@\¸N{ìÅrw4¹°ï¥PJŽŠZ±þnŒðfNZæØ;S3¯‡ g<÷£_A‘:WD -Oô<­8®zH¸gh¤­kyáP”öåU &ñ ¯¨Ýä6ìñi•ÀÄw'hBOœ  X¾qǶÝ~u=O ÃBô.Íž+›Ù0Yg… ßšª©”£T‚ñÌBpcžêã8<_{X×d°Xx×ð8¬óæ´T.†·ˆ¡ê.`rä±BËåg¼ÂßAZ¢ =õÊSà f1ñ×<)€š<ã'Oöád¹bŸJ &ŠÏÄ[çAvHztkîõgýÆeŠE.ÏÿRiu÷úH:«ªD—n‡nØ }b޳Íú{D~’ü}¦Í ãEûµÒÆQoNБls8Ò%Q£o_‰À ‡jKÞd¨ÚQÆ&–óKÍ9׉°ŠœÑNhû=‰r,ÃážÓxÞÆ“ÄÎÑVé÷tzØ’oÄ|ý´¢dvGžôªˆ)Ò6¸ˆ=ÙX¬·{t¿GT“ÇPmAº˜d…F_˜ÒÙ—ܾ>cê.d*4¸ÜÐÂô¡×ƒr"h\„Ñwd;MÁ³~ÿX0ÈgÛéC¢ešIнõÇ3¢Õ†ùö•WÀõoñ2¿~’}“6Ü ¥‚“Ъ(é:ì<±?™!·úÓ òUÁ+À_µˆwmH }šêR¿Â8˜2ýŸóI,[äCÄÉÜ]WBùb’,ìÁôßËlð:¯ù‰ jnr%j}‰‰ó¹€'©®$Ìt£ï>äJ1ÈdzYTœ)ÃÕ6ÊO›À Wa£‘Å#¯¹B5lh¾š‹4mìm ¸\9õH‰ŽPœÐ˜¡, På?&B,›û¹zl‘n%mê¥R;S³½/ˆÖ{Ê·ßÒ<-ö"×ôÉÔÚ éC)Ž5ˆU8ÒÅsÒÑÛ Gó6aT`PÐwÉØâÙÖTÎZ%Lg©/1aÙ“.õLéý`}Xɵl½‡ÆQºW"9Ó±=éü˜#ÅX|*- 1 T^¬¯}sö5ÀÿJá2Xÿ­:b|OCr¶C[ц¥ <é Å37<ÌI˜o<ØÆÿ+'K@Ê‹^<ë}í¸"ÃHjçiº©ûL²noŸ;é:“tÛ ]Á%ÈLÖ¼,ØâWœg˰“XPhôIôûÁk4ÍÞ¬éÚ†¢îhj}ã°€C©ªþ„#†@üùâ­ºòŠ˜wï¾°x±°ÉzåÖ@%âäà¤æ6¾­R Ì9“' ƒ±Ì—†â‹Ú¨¹—ز"¼ø E! ‘yGÃÑ©ŸELvžÛ/RM®ŸžL¼>Mùq\ô:/‡;QØŒ» £­ÛŶÖj\™-Íp*)ÆyÝÛ#¼\Zÿ}‹Ò¨ž¤(°µç#zÇ”§¼æDt:\‚2]ZçÓ«A}å·ƒ ñ$Hw;Gt˜Ñ¤ Fgfk&rí=Öβp)•ÆÁLÿ¾ËªSMî!ÍÎ8¡Fäœ.iDn;Q0Zûõe½#))S7‡¢*"c aS¶Ü¦¬=À$à© ;”íQF‡xQ¥É¶v»ÎýÅù'c<æFž½„£N–GùÚk÷z3(Ì"Ù !ÔÔ^Ù=4ŠË‰Sþ8;·ˆu—Y[¿žéNIu’mcp‚ˆX®ºÝ}róíü¼÷GcûÊ;d\¯qnyâW iâj.>§‰Ê»$Ñýò!Ψî¡df··=0])—±1é‚Å[öeÑÛeofy½4zæ{žOxÕ Jô¿Á¤Ú(–/¨¸–ë3KE¿ÇãÏH `GmŸ´ºž¯VŸðQxåë .Øú%áÒQòž¨.–`R¦-¹«SÙP΂4#mÆÄæGQ@ }^×ôDç?nY§Ú²ŸÜoqå™fј;ÂiÔÙõVlRew`Zuú\Jøö¹,ÃÈBÜîƒÞ,Þ×C·‡+ì0C+ñ‹£D²»t™Ø¯¶ˆ÷l:Í‹¾?‰,‹©¡ßµË¹È¶ 5'³ûû™³Z÷–òþ¡ê3+™ûSAgô2ü­Ü}‹WTºBz5›Ý”©~e¨_—’‚çË7£DsG2¨¸ò3n¡?Œ4•3Nn!˜&ÃEGµÐÖÃCŒzLCL[ûŠuóúS)ù[Ìëb Šruâ4¹Ìˆ.AÚ¾ájS)Z DfôYVµ·UŒý³™‚ Ï°|éã\–Áî~„jó ”íŸ »jÞ赫Òè³å š&·¸CËê„mI3Ê)w›¹Õ_;=VËç´²Mäë°'تBÏá°ØŒ6•\D•²«hïemÊe 8”?èÔ\…|ü­ w">øš"®»Q@õ2ûPP¼þîˆÏ:4dn>›\(uzäâcðäÑÈRý!(k'´þœCoßH‘Òš{噲WÇ€Û¤/Ï Díã°$©ñÞN&bƒIÇMOºWÌœèϳSDÒ=.ÅZç » –¡ãgÚ"´ü'Í2QìNêþÅŸP2*þ*sI‚zD¶ÜcâuaYYz½|cÕùÈ㢋QS<"ê¨Îò×˼HêÕÓ‘((ý‚ŸB”qÊ·Wrµå•øÙRóþB^cÏsg™ù†©qòfJÜì®=4Av¡]RZW%À\Öcðð>£5ñz~?ÄbMÿxTd†Y¥5g3~ÓgÊõ:9*¤×ˆ_ŒÈÍ€†ŒøäÈ}`ËžÞø.ˆMÇ®÷qÌp³ z«†î·bAݘŠ\[é”5yC~Æ•×!+ŸÉlt3S:Øh½û)k„؀˓¸û^°×8û4Dý¸äÙ€g&>Òèþ‘ß!yùÂdÕÿû´ñ!—‹höÚSw¹ÿ«&Eü)Çt5ӮݟGèDŽ‰ÇØóTΟº%Þ>ë¯=Kg‹H7ß3«AÞ)S–å:hÖJÏv)•'CHŠ)Ð(ר|Ç£áïK¹ð aÒÓ §ᙟÈú­õªYbÕ[1E¾>V÷ÆÍ]3?õ0VÍ¿Œ{ßQbTK‡’H]wAb÷§:_꜉Æ9sMƒŸñ¦Ûp¢4MÈŽÅçPál55s¯6¡Ž™æ7A{ºÄ̪ o߬¯¬‡?·ÈuèÄJý×åP;àŒ›ê?㋯-ÃHˆJrVU­æ»÷’nFâ®båÒRªŽX°c{œƒÆÐ“>|l|›œ^[=Þ„<ⲃJ‡¬~c¡ø4½ÛÀ0—ïØiõbI£N³8þé¨ÍjÆ›7"o„ ØhK¼s8?d´_ÚB– §üì^ªbæ‘ÀË™S§žëѶ M·$Ç efP\„mÚðe̵\º÷©›wT.³aê1.‘¥¦ŠqØ2Õ¬ôGJ 81r1‚­VöVÕ†Ï T»£3ÝÎÐë çéÝÂùãûA§^.ÖqyÓH«!·|6# )H „ä×÷rÃáÇì_.xŸš±0$cTÞP&66¥’Ÿ90€òžT ^Ž5ý¦N£Øp»´gÝš?e±5Î œó %ck:Ôó¤Žp^—PCIÇáÀµªxšÁ!‹ž×“]¨é£]Æû!"DXÁÅÊ®;§i¡VÝèvÄl0~ñè]ø šÔçÝFsM$•Ô店•-3•Rý]Àb”ߘƒävw¯^ä}$¶ ]çòÕÆ2œ“/ù„»`ª÷3²l­ö& 2Ëu*°bqêý[;È`ù÷7 Crñn…€šûÐðm-fGñ¡o6å8y…Ü NQ•ýòKJGaÕÃÌù1~ñ–Ùæ]yUΦh(Ün´ôBTcLљãßçÃæ‹Šé„xŒf÷o¯ßôkž/o;ð ¡>Ôó¾ƒÅ¶e|¾/§È§ßŸ#Ù ·n‡M¢+V~]ÆÕk}oÜO°v_Ì¥Ú´€ ë‰w€º¶±%B>öt@z〜âõstÚ¹5Q1'’œF¥ö(£4#d§³Rñ™QiuÀ’ ŠÐ²æYuæÂvæEã%©ÇŠUÐfs·-v±?Öhš@Ñ sP‰ÒëÙäì‰òR¬iÕs ½€¶{ìú°W©9©õ7¿á»%)ÓÌeø¥qnKùnLÝ5e– |ÔÆ‡EØ?²óÿ®`1 U{#bùa»-ŠŽû8DãÒ’Y*K°Jg›%‹¸´õwð¦htµ»å²‚”¯hhuû‡Ô´]Æ_Äe xO^ø?Ùk"œŸKŒ_h`ÖuóI'J¿8Êr¶u©í*vêœÜ™[wÔûé4©ÙÆFGíûôÞ¼o%š‘ ÒMw¯Òðš­s 5KPT‰ÄPᦂ@¹À©%.—¯¯E“¸¹¿_ȆîQ%â¦J p…×›¬Åõ'E^“ßuÂtž KäŽ?ŒXe¥»9þíÖtö/•¹xŸËY øE½§YÅYòÆœ %RàÁàQ˦NËæ:9 &äa¨n¥ûüyDhˆhޝ™ÈQ‰9êFl5Çá¶Ã"sxnHÞ:tÔÔ¢Xõu(E;h­Y{¾47üw3ÅŸPK5Çh•ÖVEm˜A+JKåSž£: ¶Ê,ûUÙÆMºŒiŸ'ú¾ã5vÄsU¿ÎõÎ#ÏÖŸ‡>O,ྯ&pËÉÊKÑÑ3=á{j,èÕñ*Êïó¦)¼ø9¸-XN â%êy‰fcpwʘQtÉŸ-‡XÎ3 kßÍŸ!ú©6ÞߨRrvä7¢â>Å ! û~î‘8…þ®5¡LÙ×ËeâLÂýšó&äiz¬„·Æhk^Äñ@–y¥±?NŠâÖ¾Õ uÎGÖÝUü韑¤†Ë£@ò%NráMrüp´|¶"+Å—–Æï×Ö¿’gìÏ”é Ç¿ Yˆeý +ÖÞ_"{}~§7ɀɿùÙ)Øý±O߸‘KÚÒÀ2…c¨N{¾ƒ”ÖÚ—¹-ùgÓø–lâO:Ë¥ 7eón¶ ±^l 0R 3N¯Ftiž®| Õ÷¥Û3¤ñã±—õ˜ \CKü}Ù„}>Ä®L1QzŒ/cB@tíð©Ì*Ñ_ãqi>ÓὑûØ›TŽuÌÊðÄpçŠzê‚ìdèuùýd-ÛíkÞÞ `d¸Ç‚›)cÄx¿.>ÏÄ hºÇ‰ð)²íãPó„Oô ]Iÿa‹ö!Æ­Ý‘0÷Äö¸ÀovE±ç-I8㌿#~ˆë±ÑL4£ºJvî§úR"uþ„ÑuInM¶$Ùöh¥ýšÄòíÊ4¿„ÑA©ÎézñR„y… üTÊ&ZU"ÐQìÜ ìp§u³ÝY:Y_Y-–ìÙÍm¾Ë æéËâ¼N|êò±wm:AËK]cx yL?On4†ôxäÆ`TDsêò‰é™æ™}[ïOÓ®Æ]Ìy­ "åØàWÁ”o ò°ÄŒcþ Èèxþ•±zAŽG­ÚÚ+žè•+vÐ~ò{7¤‹=C {l•©Ðµ‡0/‹†¸=”7ó÷ž@y§§EŠ A_jTûÄX¡,D$¦›ð_=ìWÈŽÉž³~u2>€a}ç0ÂÅR¬.^”5 âå§Å %¸­z¯EE&{1^³jWT½í&CýºÌŒ‰°—·.W$ˆ˜Sö> .®Eš'2æ@ðïcþ‚ó'H\7K5ËDõ¨Ó€ƒLð—øáîÊ5™ÿ€îCš Ý{WŒ ǽ¶Tz‰=úÎní:±ünÃ43ŠRõÍdì"þk cÅ\ðíAL¯²'Ú~ó¨Ö*÷[Ôå!¥÷Õ®Ù3@)«½'FB§*’#—¨ðÕ9üÞ‚ªúƒD[n±Î8–'Âv :¦gÈ‘S¢KÚ5úåuÔÚWtS–{ ôk%›® ̧±9J ܼ WâÛ±wÏ“¡û+F‹VÏ nµ.VÐwm߸ÁIÈ£§<ƦÛM=£âWt¾þ!¥“Je$Iù¢ ÚÆj°~,b¾Ì·]‘Š`Pú4†/I~½§ŠÎlœunjÉšt3ìy-¦,}©“¡—*¡~Î*…U;rÙÙ”šžÖš~v“–³Ì ìAo½X6ŠYmt8™FбH˜+H¦´s@Y ™¤à%«²Ú@$¯º"jÌ_ú­þ嵚¶È ffR=Güuïë^.â6¬¡Î­;pÂWÐÉWõ¾½ ˆ ‹ bœh/9†#, 2×E¬ï¥$öE‘öÍå>ª%Zdv·MZÜ·Ráô–õ»ÂïÌÖ[i ¶.V—3™è–kq=2*Tq¼òÝûP'K¥bwQ{Ž£ªš‡bkb°AüNu‘xJnß;£EäºíÆø50ÒÇ‹*­Ç»J»-;ê§>gÄ‘^X&ûzÅèeÓA\—ƒAwµqDÚô²ùá*+>l}t>³.<œ]]ë¹Jdªßp_‡Î®óSx„åÇ ˆÛå µa~^>Ý]9Q–0nfe"YõÌËôAùÕ(Wz^¥ÖzUt‡›÷ê$.($€&îuÃE^2<Í-›Ýœq³}¿ätè:ž‡Q##}­èê3¶ÖBƒYäÓ]a>²ŒQ? [æ† tmwG‚üqšª*,S&*tøã¤ÆHÛçMºï8ïHKç?šú©©6Ÿ=.ä9—ò:ÒpÌÿòDÇ;L"ˆd‡Ó7Zýñ–Gö57ßmP‡9’…R«ØÓ]D)¨ ½éƒàIý”%Ä–ÊP2~¹9¥9†Ç3ÉÁªÑû'U«QpxàFEf›£2–.õõÂl àã¿?ß,~‡ÕÅ\qõJ&BLÏÜ¡z€ G_êž?TÙR¼?̶|OgÛ&Œ¸ªêßkçšó)GçYVå*;`†í|ÅŠ}´¦4¸§MkÖ$u•ƒ˜òfÔnÚ=£BçžVl6_Ÿl7èMŒé|ϸ ÐÛçª 7(–ýË»âF ŒžŒþÌô¹6à»Q˜ò j¾ÅÈ¾Š¬ŠÎ\‹-õoÞ$â‚ÕÎA…`?à‚ø›³s=9죞¤ÇÝM¥™ØFj:)0Ì#˜©ëb†p·KåpéÙhJm±xý Šž1éÁ!_yÕCCëšÙíÌ\,ÏٺݳT y#°<®O× ÿ|‡ªŒ€pK³s5E—Ê\Mg;¤Ø¹H§Û)¥Nõ³2LŒAC[k™ cÀ;ÇÎnE3ÃðñÁ±Ú* †º°Ò%å\Z­oìT”–¿µtŸ§ì¹Üˆ[§“ÔbÕÓYqñA¤Ë^ºÙ`6Èÿ†Uï»?ô„`íŸËÀúV~y¡ CƒÏ…ÙŸ¨¡¿À¹ÅèŸÁì©Î;5¨ŒØëDÞköCÌÞ)qêbæÕÌXQ9ò&µ3½gV‚E®c¿뢘_ú˜Þ5»é·£GÄG a§.±äø'UÓKôÚôJr’4#qV~/Ü$ç¹Ï ?ÙõÅÙïI,ÀWñS|:ÒJíñVˆÛ. ¦*vE¤ôKìÉms+¼½úÙ ²Mù^ñçª:ÈÍû¯BÞ]¸G¬½Ù´3Æ©äÚ½õš?ËÊ9ñ_æ•& ÌþÚÇÈ,Û1r¹*G¡ƒK»Ÿ`6ÌÄ¥Ãr‰rc(D¨æ£UÔÆrïÁ ðoDâÏ@A#7§µÔ‹€è"oß­§h°v“æò®Ç²Åù6À «[¶çúZ½m‚~lõ²ÌM Å¡H3NlbŒ¨õüzä w‹ž÷~¶då¸ÎøÀ'õ¨³»‘‡SÌ ±'øMÆ—U/¹,t3ûº#§à›2ÅÎ$)†æBß*·×l?üâv2jáÔ’òòîÌ(/B©Nl”È]ãlîúÁö5„äÚ[¯>ÏŸpKéŸüöÂãÙuàƒùLÊÕ¡à@(kBó½_Împ‰c›y fP-ø”‘º© øÓ¢™¼¹Kçžõ Cž-}Ê*öÍEõ ®s`†ö|îçRU}¢„ }ôÎY³\>ÅØ©·%Žžá°ö$Ë6&Dí²Vè[cXú5šœ¯â7ð âîýWÿ™›ÇgÐa›“ õDÕ$ ,÷+© ×çŸwmŸÍDš*ég¬­p~ DÇ¡˜ØïÁEX&¦í’2{uº°˜ÅzÑòBš-¢œ/Rµpƒ×°O®ôv¢r%OpÁÿ©aïã”ÜjUq†@òV^°¹ˆÒ´ÍüÌÔ_°HìŒÚÝOÞ¥T.ËòdGTõ ׄÖé"„,L‹Ÿ†Óu×òÛÊ ²aîî¹Jc⤲ÜÈÇQÀ›Æ ¥òZx\™¤ÐÓ«æÀ9¨+öø7EO­6¯î R›­§¬™jøukë•õÞn¾uyý\ Uó˜tZYØ ó.ÌðÑ;í“[Ó ‚­“¡eè¸÷é}§iZ˜¼Ò I·&Bs¶B‘½_gÓš¿ó¡ßùorvĽŸº¼åM8 ŽÜße•«K]oà )UêsÚâõÇ{;Ãpðø Ú?~StqR¬7>¤ˆÀ¨h_t ¬É‹… s!ÔQöWô®ø×§û¦²ý S;Ƈ8(ðŠüÑ6’£[bX7á|矩 Aø7»«sÔ›[é´]ûdÙ!¦Úƒ´Â­’-!JhÍoo„  „Äá+)MÔV•yÈ ivÝq‰Â•3šgBõ!W\O[j¥ T>9`/ÊW[7ïT凔,í rC›ÔàŽ»°³å›Y嵿X\ûàu~‹#E‚×u3ýúu÷õ¤«»I›‚”<¡\‹}ÿVt= ’Ì'vË‚²{d~a&ª¦ô .=«8Ç œ*”hÅ¿^º{Ã¥ÈøÅúæÌ$óÍYræúqvõÖcÒ6¡”vÂý;B(¦×í²%Jã1âÛ=ŠW«À²9šw¿ ÉãH_¢?±eQk ûûT@ÁÖQ„R„ù¶« ¿«$„ʒψNM˜%’÷Ú ÉM‹!þ0¥¾ÈØÃJ?ÈÊÒ)ÉÒƒhôV±÷éé°•Èð³žYÙö/?` >ÈHÓìŒÆ¯dGŠ·®‘·T_MRd‰Û-3‘WhRR™ËüBN†À/{M@Øõ_Ø!{ÀˆÃf"†¸3þ›)4M^.ªÞ0Í…;K;½¼ ÁUC+ªÖæ† 0E“}¤pì Ú’× aaS²tÎ-‘eî¼¾0†JO›EË ›“öàØ S¢“Xn¿çYêµ–&Ôß¶ÒÛœ¯$kGZÉÕ£â~1¶Ä/ßÒ°kNK—p‘±-¹p¤aEÀ‚fôï0[º–ò~¼ð6ê0gküò>(œ!‹òyÐC­ 9Ëöm¬FJ}œå“p‘U¦Vfz}xèß6IlJ;ý˜=iàI«!ÁévÌ ®¡šŠ;Eâÿñ(P! 23°›ç/¿?Ü­²CŒøMù×¥ÿ 9\DL›U6ãH_E+'zÌÏÅZ Ã~½zBé(ª}Â鳤j—£¸ˆezŽGg¯þ¤‹œ"£TVß³õ£Èó Èó—_dENÖ^˜Ý-êÏXF…6Vñ#!ÉÍ¿+ñórŒnVÇZ²"4ÔÏŽ‚Ç_OmbƒŃÊð^·˜ê*ŠÈ½‘ˆ{–¹”¥· ¡ŸjåaFj‚l޾õ`¤wvœGà±i¶Gû8&P+ØÓÄø¤.†¢CÓJ¤ 9ò4˜‚†zá_¤mrˆúËõüy}¾FºAõ‹2CÕ‡5›È8(ƒ¨«¬§êYþL\~—M?y:.î|a«Už rx,¤“™-=ú« ó´½û.Uª“ï_Ôâã{ç?¼}‘®çY6œý›@\û?8Ž—õVÑ·0Û´ñ›'•Á ¬$‘bÇ 2‰ Iö+‰i7p†e½fƒÞÃ\7?`æ'l¶Çf½î(šÏÿ’[mÆÂäâeÅ=ûRò'Q¤ jþY_³›b'’··ya ˆ±Üƒ¥X1\l|½pVvfmZ7óçw³èfxþ‡´tUDºò8tÐq°ò¼ŸG‰ŸŸSŒu/øy›¹*Y¼Þȯyž%‰øjV¯æe'¢Ày@|d½x!Uéº}[çëƒgº@ÇÚ{F>âûÖÂI·öÏ ¨£„×®›€•»Û)©³á,­0IJS:DEÙøÒî ÍE˜½ ?«%ifsý8ÞU 3­2>e[Üpu2zäjÓÒj›Ô1'uÓJ*‘@H f%[¤½°­{H¸±¾¦¦S}€ÎêœH|sìàE5^ÐM¥äkÁ°ƒOÛS†Pí™ý¸Ýyì°˜8° i™£¯N­ÎfªÜ£¿ èˆÌçÎÚØäÜYÙl'þ¤+â–£:FÌŠ6+-fÈøÉ„7}s¤Ò.;ž›Fý÷ƒã‹š8áPßÍ ‚š0†3Á™0mÚ+’M: #àÞx¹ÂvG0Ýÿ…G!qb•¾®7Áéš^)× Zu¬ í uìuªˆ=Ù¬›•$Õ€±¸æ/ŸÉ>œ b¿¥ÀÔéÒeÕ\ë‚N”3ÎBg¡F–ö¬æô6‚·êŽ‹7 oÁ»Ê¹³g¡ êp&Òv¥–â;|EÛhÌÙE®ÄÙZQÐ >óÑʲòð"vÉCÎàJð@ÃÜc™¤xÜåÖÏŠr~~ôÛïËa U»<ŸýL5Zó9‹%oœ‚µIM{z¦5ÚcäÖ¸oô_:nSÍ 7¢e–ÔËnÜç¹sÂ#jgaAÎü>Ë™3Kê¥1ÁÁ¿WÅðÊÔ#j_ÙFYÒ½?`×SV2XóZäb²ýä-*‡>QsWBzàp£õânê €*ë&\§ `íg”7»ÖÏ^ë†òŸ%»²P§R‘²LXùw—TdnÅç±Wtøñ¬ÆIÿtpmïBÇk¶¾'ÀN½>!¿µµ$(ž+o8Ю;™Àb1._24;Ä\sv‘Wƒ©¤gm ÷ÅLœ&S1¡@Œ Ï]ø¥Nˆ’]o­C Ù-ŠŸïp ÝÏü§ii0ÿölIòvΈ nÌüvõîårúh}`ñ-f;qìõ銕ѹµ`…TDy\Œ¢ÙHކ5ÍËäÖÀyŠJä¼ÎxÍž0—Ç•øèPå:^z>¢C LÅ¡´â¤à®zzצ‘&±xép­þ¸z‚ØÌz7FQÅ7ÛÀ¿owÈ}Š]Çʼ?õ¸u¶(À´Ú-6ØØÕ‘ø^œ {›r¨`ÉÙE¨ƒ­rˆ§R†²ìÛAÄùµ )+‡ÕÄ,NÜVbVí¹_øù¬Gá¶-¤2¯}-³1—qôüΗ(ÊNMo-8íæ(r±^Rܘ>8(ÝLeÿfÛ‰“*Øäë ~ÞO¡{ùŸZDÙ–7lY£å{ò„7ôþ{à(7ÈI"^Ÿ2÷”DQÖ>2oÑ[tŸ—J¾\‘)ï >g––v&:ònÊIXc‚Råa½lð{y¿ý®QšŸöjCÂéÂZ%ѽýgèáìø/Fµ¡æÑíCàV€º£B—`×Î8„ø9sЇ˜Þ%ì=J¥x~¤ Îh5šÿ€â¶ðÓD怟=žUnnñùÐq³ G “ ×¹V G‰ôzÍ+ÿÌsyvw.üÕ¢ðaã ’u¥¤Có´ #ýA^¦¢)ø,×ÚµÐ&Rjäàš7½| òKi˜¦saO¨ àÕmr~ð­ÕûŸ±©Ýáê`:à^жç»å4÷W‰>n§ .hÌ Îzž3×ÒËÙR\ðäĸJç]½˜8¤þ!Û‚m­548cqFøŸ¶t®à(5µ™çÒÍ6%«¨¼¼s5º£”êwß dw««âA›_¶†Í}à÷o;ö 7°g=‚›øCOíX.zˆ›î£ûŠSÅøÜV›¯ô|?4¡+n.KŒ´%{:º£îw:šÏ¼ÂÚG÷Z2UÆ?¯诡dN9Iµ½2YÂÖöå^7úÉ1þEÈ:ä <çªiNs;‰Ç!øìrXX ì©ëÁMƒÄøüòkIÞò¹&=B ¿E$5Wž2->îæÆÁˆd'ê*Îw,k&Àãë¦nM5sö+$IMv®xð˜2fæW8lIIXMJ,ïBEiõ«ÝhÎ¥ÿT‹QC΃r.5@{ îc¦r´l‘#d48œÌ8#Oö“×@ÿhG Zøá”™ŒŠfÔWXI–ê‘i±ÐËg¶§ßo†ˆxk 7pq¿“B̶ $§m!õÅüúÄÉsûÇOáÛ♵FzSýéÃmʈ HLâk¯£°Uì¹Sr±)õ:™ËâW£‘*+ï ñôÛ±úÑ15Ÿý Û›D¦-ðõ·ý/ŠèŒ¥¨ÁÏK™þeƒcVò ¢‘ë f/º4}>Nê¤ÎØÛ({ìTuDÜ ñ»Tzfþ(æ«›Ùˆó3ÍÕõW¿—×£v?òMxiÆFMOÓ*3ÑvœÚߌ€§>¢æ­U3{/È¥aÝ’‰´³[Õ±–|¢~™½Aú˜¯åŽB.£„†YùzÇù}kyÿ?ˆƒ°ïÓȹ²½‹;f¨g—‘±s´Â̶Zéæäw² œ»Ô,Li½+ƒßˆÝ~ɶq²Q¦e)ŠC—ÓJýRÛ #)Ž ~óŠÇâÒ¶8yñvQ¿fYèÏ3tyÆ×KyCAÖÆÑ\ÈuÎ[IAÜÕ)#gw‡òC¼¨y÷£ £X>òÃí2cÝ6úð¹¿ZzýîŠPHt¡cÞ|¯6«ÐW`òîƒÞN~P"½ÅOÇ(ÛP æ“ÄÅ‹Œ‰Lg}ÖF´›,BŸ5Ocs d¸mC†n˜á:VᛡO^übh]"è 'kÚé0º¢0 ¨”ãl.įyÚI }$“]eÒ~¢9GîžyꩉÊ÷:Zö+YÅ>Sòî~ÍÈ]éô2xÖ5æ=KnG4"Ø>e8“;}T÷X¥ÞÈ$zbŒKûï`ÔÕÑ•ˆâ”E.—f*µ0£ÌHãIÅrµŸÇ‡:¸ÁŽ\Çœ÷ 7q öóc®œÙ…³öæ$xÙW9D£©i;DqÃQƒ‡v”G¬Q>© Rg{µë/·ŒcwϾ¯À³Ö#ˆõ¶= ­'רraŠSåoN¤“%÷®T•˜[{hÃD$iƒ. ÀùJx"0rÍäFçX™,Öì –(E(¯¶¤)åc•¯åyrICÖëÝY[»5åR,qªÉ÷UËã EŒM}a\½_P™áÂTw*‰ü4pâ/¿»OJ¶úˆ\Å¡¶"qŒTûøÙóV+×Ò (–þ…/Þ[G/ï¿t¸(õ|ôßsÃëMg#1š™"p×z ̉,^!—Ø9 ô]Åu5ÞC…Žr¸+tÁ ˆYД pgœÈSºD²Ýs^â'Œ~0]Wº}N~y_Iwû?]œÓ’(bÇ6îØ¶mÛ¶mÛ¶mÛ¶mÛ¶­lªò²ù‹îSÕgšNxÊÓO‹ž` ÷*âÔ¿}¿„¢ú¼õÅ cÅ™'8RÓ ¼úO ‘ŠÄXˆ Òk}kƒ`Ðöþz‡ ÷P—rÆ¥&=Ú ô>kV3{W)S_tgKÔt{UCŒíY1@i¶émÏ3ê–·ÿé…!Rÿ rDVÛþû…ÄÆË™atàAÐà û'cîCYÝÝ5âênÑ{¢MÌe[ÃÅ€òÄ”j‘’DÐxÖÞ¤¥*)ñì-ÞLäåê4”P•¨zÒ:|ò ÃÄsé—2**uº (ú\êž&¡Ü–œÏ0ш0úFA.4l¤¡Âl¢´ÐháÅlÏ^:½˜2Ú‹Þ¡JtšÖp=5”sÒ($sÂlUÆb»h\ØÚ£°Ô×¢m­ÞåeÇ5òE‚?T¶\É÷ï¾~U"‹#$ƈ”Ô%°fJH}q¹6žã¦/»U„8—®°'[;ªÿ½h¹Vi¬ÖÉJYki+Z¢àg\ð<·Ís´œNøo¾øÑcyÐcy§6 /Œh/ï}™¾ÓÍ Ö°^-K’ Ö芚Ԥ|ê’¡ÁA„¯#þaßMÕWW’«°z²*•nJk6ùê™Õ[£ó—?@ÏÜTBÞP‘¸²Ò-Ë]kím~BÏ­yo’ Vq³œeð¥FÒÜrï¢ÊWݰ¢á¡à˜F cÄ÷±jRáQÛæ4ªƒËè¶«^Q§MîÈj²àºªÃQ6Óo‹Áë*a™iõ|³³c=ËÞl³îzø¯~¡>%¾e7ðÁ>/`«TjÅ+¹‚ü:»Höþ@ˆ¯ey—•Y­¨YX²¿kŽûö='9›H2Ç‚üæ¢ZöÄɨ>üL˜î {Á®’@>ñwmO Å_`QÓeǸÿ7Î )ú(’ëÉCú¤`Î3·çãŠeñ C hñc¡Š<öÌjlÙB•Úm4W“¤ÔV&nºw™Tˆp‚œøWbÂ3¬‰ŒòRn<±›s×qU ¢&ÈåšÆõþ&$¶&ý y.fÇÜH¯›)ó&’ôQ£A†kˆMN©` joúQaBW˺[–æÏ(f·–•<ƒ§Ý(Úå-·…É4µv3M…  ÐÒfcFki[ü¶â,MÒ±z¥EBe9D|«µã›v >ÍÄOa¿\èDàŽÐù‡¾%gåEÅ6„ñ~jr‚ < œIYLE•T¿Äª@Ͼ»?fävúºl/•ÜÃB¢£ˆ]IÅß"ôÓ“æ—Ê ÷_Gà 2e€ËÏð†öõúÅ,ٸ攣"ÊQ†S0–n=¥´e%™ªŠÝ4U'«e3uÞ—€—šz2a¨ƒn:³±¬–ñ:±x¹çËÌa·ÐƒÕŽÄ;jÔÏyc=U‡œßa1þI7K3 œI Ax"ó€Óë(Øú$lïi„Õ__’älDX~ÅÒløwO!EÒ7W³ Q ½tf-‰^oBÎLøk|ŸQlHN¹k¡·¿bíò&nôŽÒ£¸éEµÏK´š¹qèãæŽ?Ê Ðýfr:ˆ¬.hAXÉ/Z’9|öQ¼g!öÅ+¹ èp%çQ¶\!W b8¨_ÿŽL1—S„\)à6]»$Ç{;0FJ\Òünæ;1uèå¼ #§Q›«ñÐIâ^ÂlÔ†ï›D¾=ÇóŸ|5ÂhLÿ5õÁ­™®³O´ ÷¤/í?j@lw½ –| ¨ âÕ{¿Sæñ;™‰ºp¯ooœ ïžqôÀy‹³7Š¢:}d›z¥¬¦5¾n«®LZ-…êÕðH¦"r$+*õB(ú:/:”×iȉøYBp È››áOÕ8JßÚ{ýÞz_ç|<‡ò\· ›Y ”:÷& MÏ!¾û2aÐ/åtРÕyVÈÅXüUiÅ0­nîÃöA.1KuGåÈë¼Ù.ë$ŠäjNÖô I>6]áª_ç(DKûãI\xªÝVöm_£Ì»¼ð„LO¡¶„4':Ù2y(=ÜÔ Ñ¹Zy[fñäÂú­,Òz¼(±m—Ýz¤bÞ‘að5É'SI¶Ÿ(hb5qáósªÈå[6æI‚‚lÿÇ ´ùƒ÷íƒÍ?´Â0Âè¡!(”âk—(«óÖ•\ýÎ µ„ðº¾ôdø]‹|ÛåÛéд$%*´b6›e»úG Ȭõ“ç•Ø;HE†ó´< RÚp1çÞø®pñnL[¸åÙƒ1Â/«8›öèÐn‡/áµ%gÊ ˆ¾á¿Ëê“%1W³ BP=´`‘…0Ó¦j4¨ttþexLU÷Ú‡(ìÏ·—‘‚®ƒ“tõ]ËÝVÌÔ²XëìâÌ䩜?qˆÒòÉâ{£tùWùú‡UdpJý¬4uZ0·Pc8' [¼kì踾ˆ¸›BûÜš>üÈ *êùpİÂZ 3^æöpÂëÜH¼Êƒy„ õ«iÔ~*Ø «´QIµÂ£’iø vhŽkÙÏ(9®Éo”hEwõUØŽoU84 qkP+§à”nKÓ*õíj’@SMf°ì šô¿´Y`Ï%‹; òRí÷Yšëå×î#—ÙR~íËoYyîå{GîÙzVF]b–­}ÈÒíV )L¹•+!KúŸ*ëM7‡ý\5=Œ Ž¡ÀÀ»¥¿¬ú±b: f,¦#ù üEwìy”C˜)dKŒ'±òrMÏö’­(EBÈ–>>-Ô~­P›Ü@—8ØÈɤ}4¨Мn땱û¡Þ?S†³±£¥$öÞk±Zÿû5¾Ù@Kª ¿Ý@Ì)᎛» ÝÍ”'5r}tù\ÞĖϵ»7Ñõ^­®»Sí/^¢›Šˆì¦Èˆ_!âož©{C³EÝØ” aªU=UN6”õÆnˆ…–ãT˜L“˜¶¢¯«§+æ…”«ˆøFô¯_¹Ók“Óµ=!1Þôaƫ؊2Ö&úûÓvsÞ¸J„Ý6¸÷„riY—߃_xƆßCq׳j&‡Õþ&voKuU“ÀÀ×3ƒE —¼wìM§IͶ²ŒQvbÿˆäÉI¥%Ô:‰k >ö…Júø–˜î¥O§\¬–±ÓÞí,|Ùn ö4†ÙcN±!uÓ]uÀCæ#«©(œ‰èì«-h5~ø;±|$¶CÎöDà‡±¬Â¦êDiÇgÒ3=»wªP!M>AF^í‹°·«¥¾&Ñ•ú|Gþ1•~.вõ*‡Çþ"þ2CûØìµ6Ö¾¬êÝ)í»zKï(îeŽÎRŒBÉSË·^õ%y9:oÂêl‰+d)‹¯‡©4C†{_=yàÕkMýèLÂ!Qû|ÎƤ¡Z#Ô·Œ'YÚ&%ã‡d§|gù‘ZCÜQ<ÅëÏêŸå¦ö³WÑé°ÇÇÆ ¥èŠxš·Àm­"ø“¾’gÑÛ ¨R¤•óÒQ2i‘(àÝ_I{Dص"Žîy(ÁéFu¶Nêï^¡c~ábXùµÅÍMÇ{;h->ŸN&&uðÖKWP²r½¶Ö<›HÃåQéîÈ»‘YLÓ¬µ±i¹‰S§’KÖɈI/|¢oÕu‚zâi}Bëx®†ÊÌX‘gµŠÛ*Þä#FácÖfnQwAAã}oñ•±ÍpNÚÁã¡=ÁË 4^üï»õ brÄ‘´Þÿßh)d‚µ:Œ>Fǹ`~äò–GÎ .5IS‹Ñú4è¥] Â²"ŽüH"V1´ÃêÜ"mž€ðàÇÓüËáÚ£H\HÌÐÈJ@,X-K(Š¥çûlœGXöï:£P¹4„)(¼‘Ú½ÀÉ"#NvØÎ·Œª?£’‰¿J߯ÎÚ‘„;XÄ€è°Lf°šoüm˜—HŽOÌr'²šAäOçXadLûÝ¢ðç …E6*/âYôËq”yç%<šõ¹Z2e±}7Vž@×jôoа÷âƒÝоPuyú)&rÒ Û/X Ȳ,4G?ÇÀÀ3ãÇ }¡ï”¬".‹=“Ö .‡¯yW¹B®ô¾ œ…Ü VáºC“@,9[S ‡6&rÜ!5:…Š?{í¬}Z)FR¹âsMüä”ôlíqªUyH¯"Ë| å~½`ƒ _¡1Í÷nF¾ïÒH_ºW}¿k´ùïErwûÄ™ì_š×³¼W°T&y¦„R y¬uHÂ1ùׯ1rÍ™v®›‚Fú€f8v¸€;ÙªnâŒ.KrF†—â¦.3ªë‰:J}årâ·Õ'Ç­”Áã0ÝØYEYú@šò^/ ¦ª†©}!C¦7°ÅŠ®P˜f–5ŸL%h…%¿@%@t!˜/jã3ÄVãQŸ01"ôVäf©RÜÖ„¢z™&Çÿ—P•Àó¯¯jÁÍÛ €–G/%ÒOdõ4`÷A‘ýÝbÙ¬“^5ÒY’~\ïV{a³dÀ5ã&°p"ü?òÌ÷öÙ±J£ÚcÜð¾¼?X-7qO¦ýúÞŸw%[<|—/¼tt÷8 ²ÛÝÏ!UH¬³`¤2‚€•¯kUª—#E™ 8•Ãf‚+um’±™r(…„hŸTH•J†OªËÖ©çÏÛ}º)ú¾2Êݪ˜u2íùDê°/ÓnGn3û B=ú®;ÈÖ6J²OÜñîÉae¶æŽitº€q#õû1÷8C:œº¶HaìÇ@3n }íþ‰€ÕôÌ¢¼D¬`‚ÃˇøgLOa¤ÍÎÛ}0åM.>ÒOØ¶Žž)}aXrñ0®«<{¦}»²ÍâwmWᯅaYKd¼Gb=T@Y®äú>´6rÊ/ŽzÝù0œG%GÔÿ›Ð^4§ ‰Y¿~ÁßU?§ž &Œ`ùÙóàms3ß"ôßú‘2œö(ÃVL¬‡|àö{dŽÚ;ªV©Ç‰¾ÀJjA—Owvò*š~T¸ÜU°·^ÀÆeà †{ÔÓ$S‘ë*`ëÄAH5þ ²ïæÎ8W9.~gKPöTo ÀØ­-–cvaŒ Ôà4ÅM¾ñ³áݸ·T>;ó¡|ªfÝ<ü65ý ˜Ì¹VªÕÉY·Ê§›w,| ^s†ñŒÒ–ëzn^èŽÚ— ¹"z&zCY ~Üylh4y(»i:10Ñ0­Ý[ÛgpäaìÿO(.|ŠVlÊw.^ÆE•׺Å+ ”Ž`,kÕ’Õ½91Ößõ3œ‹Ê»¬¹!« ¥·Õþ~L¸ñ¤Õ­RQhòà«O»\ÀZi*ù8|tOÚàYÑõÊn3w{ÁÕCÖÜPròÄ„¦¶…ï×RŸ I,šM: ÒïM­£=ˆIN1 ß$QzxŠ]‚#9ro·çãýâ&Ì"á&ŒÑô6ýb¹è׌§°…p÷Éö,å8¢ÃòÛè5·ÒÙ™¤¡ÜÙ=óVeǹ…Ë[{SÆI™v>VT ÚP¼pZèr¹¥Ãu;æË›`€Å5®îUF)ÄÞMã’öŸœ0ÁóKµ?@ýzøPh‚Ÿrd’9¬QOÈ ¿èeïË9 TÝU x_8¸;„!—q3o’’ןRY'r¤™Òà RwGçÑ¢’¦µ§Ù϶¥¾·Ó&õÊÿîWB\žý/å+µ1I’S¹zäJ¹À4?÷B@v?1Ĩ@®:ÑF­ê[ýàªã2o`’ý“µð»1Õ\ö³‹ç2v>‚úÉã'–/Û('>KøVTäUŒNDAáÙ<žÔ“ûRgÅ!1^ÌiƒGu²å8ŽR¨ý,éðNH£‹Q­.Èçq½Ç==¥3Émœ#7‚æ†tAy<1Õ¤Ša‘D!ø;¨ßBœ ±L!®Ì4z¥Bgs?«Í‘ÒCÞ( }¥ŸŽ¡À`…½±0Ä‚]æ&„g{ËN g1äxg:¬™ê7xCûwêŠ>Û˜ÑC…U1$A-¾$´ŸÔö¨L˜‘Ô.ÁX’mËw¿T1o%{ÀS[zAæ`”± ƒ),ý¤CWÀk$´³W#bÍÀ…FÍ·2e÷Qî׫¾ÞM0穤$„äÇgYOäÚÚè À¦§.öþûõª2øóßÛ/cßÓ›aUîé’3‚jý¾¡A~!+ Éåêø(£?DÆcÿP!«å%"Ù')ØáëVÂýð»´ôÞ>iû£Ê2{AÎÞÉQt†˜Êü¾Úȯ'jØÆîuퟩEíT˜G’ñ|œ-,-Àf«W±ÿâ΢ó`T'z¯ Ë,Ú¡Ÿ øe„É sœHv[êz¼nÄþ%ó`B ý@¿°kRå¹6þN„zÌž~ŒdšäãÚó Ú<¦è¡’Ü&Ý !˜ä6¬ÁÌY­L¨«eúãæHÏ:TG˜ìÂ튵ƥt¯p¬Êp½ý÷–«%—¡ë-ƒ½Ó mX'Ô±§á…ßxNy|¿"p` bÑâyFÅÛ†yñt€F?¹¹Õøf#³RžœäRÛµ›A~ntt"ºe& .”óæŸæI÷«X‡fxÀ,³¥dDÔ ®:¢uhEgN•MôÕóiÑ—ÐðJxÆ?¡¨_FBŒ,lé)å]ÉûÚ%~² œáP0Ú¡hyop'Óµ=Ptþt£ó\µÊZÒðvÞXo¡îEGy—}ðÁÆLÜ0}àÇT¼™tèý ò!6CFÜ7lcNŽ0e‹9ý/'®µÕ˜•ÖXóMÕ<[ëWȱ›Š¾ìžÎŠÍ×{¢O/ÿ亂Z'Ù§äéÛ9ÉÌ^¯h>ÕÌ V/¹I8L×,0tGèÞm®~B5SmK¨òޤ‘ÙW†¹jãs€·£¦€ñB"]å}fïz‚ÆÎˆŸýÛÞÃ=‹«öCìÖ‡”º!S(zïŸ~ŽŸ¼1.k/ºÄÙ vϺD%.tD"¹LÑÜaÙí Å&P–Wædb“«%É”«C‹¯Ñ9 ­¢éÞmRÊ v†ˆ¤ºd!Ú‘8º°EŒ]W#­WB|o-¾é¢ddàI¢²îûµCXúZ,2 ,Ó°âË~ÖüVª~bAÏ\ˆXñÍX?’¤£!=ýv¸G (dG÷ ¡¡i˜á!ðõ¶xèÁÊ­[= 5«ߤœRC÷Šl8Èו¹ð\ì3Ò¼Û߀©1’•¢¿tÇ…˜1CˆX%–£Ey²˜+q° *D_ F€Õ9âý€Ro3CD¿øÓê«e8Q6ÁB!0à~pYHº)F®¯¼pXiÜu·¸Ž£/Âý&:lçvÁuçg—§Tó”‘+âÚÐNÆÞœ ôÉç;à°¯Ô€Y°¨.¾ô"œwX?rd¹Ÿ?‘8 ì6,¬ù…ð@ø9B´Ì}hÇð˜FabG1uòû0¬?ó>UÊȹ¨¨Ñü²À,4=Ʊ¯¤Oسc ¡lÙ Lq &GS nÄ‚­™ÉQ«„þ-H«q×YdD Þ®½¤ŒÐz«IÞ‘”¶Ggc•GA·‚"vëÜh»Ulc¶úãP‡lÁŠÏ´Ås‰’÷aŠb9Ó++«"MPƒ–Ÿb2BKÉÄ3|³Ën`Ãb@èiøÂa^<¿÷(XCMÈ}1”—Áæ>ø` <îK¨®y…ò·Aæ–ÛöÜ‹”ãÛK;x«žM‹ |æõm_Ñ9L!çÝÜ‘àDñîuÏ¢B¯`òƒïê[;]I!ô×ÒÇÁ¨´CÑ`TRI„´X[RÕ$q³76‰^Xú<àMy¬”^BÚ.u´ð÷ÉÓvÏ€Á £€_¥]e §X+Áˆwö2sp`ÒÃŰÕ<âÄQ1¾nÁ[¿fG²\]Ú–y Fú"ŸqÉ,À£ õQ¹ˆÃã!"PÇ œJø£¾òécD%q`¬§Ñp®/)Œå’Ë¢†)éZXðz¦ÛrP Òöìy€¹øÒÞPO_Z“Ì;Ž*cÃÎŽB–N{SY‘0*Å­ÿ/–ÙKÊhÅÌŒ[‡±2ÅZžWƒev]¹“ VEPî^§Ÿˆµ—€¶ˆŒ•Ê™’ñÍ=FýÉp ‡*÷³¶“¸W¿ªf®æMA6U™°([<¥¬jO­ØÆ~¼ò„ öü×ò4º 79¸Þv©|$ÞŽ/ˆz}Å›Ú'*Y:ø¤Ì€c¹>^j)Üø#bŠ••n8/_Ü0JÑr“@äÑʆ+ÄK)jšÒ¢‘à‰Æ¡˜ä76s×=ŒüL‰«$Ôo–¯A'³ŽHž™›UFŸùÆtC>O{÷?Ü*Dø£†Éíý£;–*î®5ôÂâÞŸƒ|ÿ‡è):[AÅC~lž m±qèZª v‰6z¥ o²|°’€« â©^#à¢Ìx”&Ÿ?æ*­¸È‰œ±›6<Üá3°ÿ˜:C6ËÊFl®ÊìëžÍÍ2FQê~¶IHì0è!"Øã£: .Bqø©’þ%Àì`änw+&®ÄÚ¶ËqZéo×><1 õ3ÙØ^¥]ÛßÛ5önÂLš5$t@†*õwhê.å÷üiŒÐÜi¸ÿ©žj'x#yéÌz[rt«ù{¥ÃÎ9ÏjgW×/\þ {úK_Ó"ñSZ°®=>ÔD—£Ù¼ðLêJ'}øž€”ŽO…øìæ¶›É,œ<¦ßSõDÔ4Ö KOH©caT,Å+˜ÊAÿkb/MfU¹Ã‘#e}‰öŒ‚¼ ×£¾ìò'©û~‚ýÅ<„ƒCMN4)FšyL*NMélìíçÚui‰ê@Çát•Ýæòêmqµ¸Òç÷ÐÃsD{-ˆ• Ê¯ÚÒ¹î"ñ·iý¹Î)àד‚~Ëi•J–S=EÁ˜Ú®ÛåÑRk°vVú·š‘~¯¤¿Uת’l?/¦µÇt¼f«œC Ü:õ‚ü‘Ú·\ùûk³guˆ z Ôè6¸ˆ†Ðb<ÂÏiIÕËðoR²õç|J=ÌܽmÞ)…Þ¶RWÝèxKr„ 7b¦8wÆå®¹ Z½7ŽÑóhq ³f ,qþ;#üÇÓªßèÎÌæ4üx@È]¢EY+‘;J?~ð]¯S²I:Z‘Q7²óåKCÕcâ8ÁàEHNÒDðÑàK_õÚò„Æ` áÜÌüTjÀý/ùª¼ÖTÃqþÍŒž¬~ÎàÑ‚†Üfg\ÝÏ] Ë1éÙ˜¸"½ø¥=½”—=òf^ŠpÎàYoË£8ïNfuOH¥q¢É]0ÓT¡GçT5N›Á‘x!Íñgõ¹™s*ì™FÔþvןwÞ²G<2ž§*{æÀ%oJý]wå«ÛW1ðQR—‘ew篳–ñ,æ÷Æø˜J¿8ôþâbÚÙ“VÔÚß§è–Õ¶Œ[ °}°R.pÉqX&Zù\ÁÆ¡ÛE ™Çï9M;˜™EŽL®hóG¥Ì¼A‹"ó˜ŒÉ©¼™r^ü:2yp×”cŽz]M’)´Ù¨ÝF Á‹Svo&Ÿ‚#OÅn˨ýÂ7e] —Ñ‚>Qeœ›·×ìyxãmáö: û€„¸ÞÍõí$fZ@´Ñ£‹r8Å!vYÍd£ä·á» JúÅVÄ Ä¡ýÃ}˜Y¬'’²VÔœ¦! GŽ•#5=6ŸÂ39è 1 Q, ]æå¦Î€¥Ú%8  Ù%s3–­õÆ)Nð&§–÷UÝÇ‚›®úëùÂ{,èä°}QüO´uñ´" žS* ”ýÂÊ5_Ò®¨+ìþ×Y¨Z;«Ï±ä¬n®“Öèæ´É¶¶R~Š¡œÕö  ê¢\5¥^Ѳû¯ìíîçŒq3Õ’vZËÕB…|xëé/¼ä¾ZxÉ—¾BG«C‹¥+æTÖx‹n1[ד¹øÁ]ý‚™ß6 Yf’þ %VP…ä›#«Ý€ù¾8äµqÙÎîæ6¬—ÂY Åf,ò^ f½« èT1ÂJCÑŸ¢zÝ Ù˜ìëA`æ|ܬ½— rÃ5%Å/™×ìB´¡Od¿°üA¼Mƒ UÒWIª:—?#bûaŒ`SzAya¼Éx£Ï2!öš×`©-][[éïQ"o¨šTSÒŽO¦¹¹qx‰^#Ÿd0Y@7°ÐÆrNU‡ëV†/½âý²ƒ-œÉC^ùæöõÖý¾¨ãa½dÒ s*'(¹•\Vq—…ÅêŠûPâ­lbY4@ø5áLј¡?’‰¤¸ž*•ÜH-ÑjHZ»&®nŽmÁüå[ƒ›ñªù»‚–qßëa{Ê%ïp'ó’Ú “iÜæ}Þl\Ûpj›¹§]’Wë"rùìÕW¹&ª§¬\:'„ªõ)¥tÖ>Íï ¹ÂK'M‰”íˆ9˜½¦áRÙâÏ]¸hï춸´ï6…„è•nMˆ¯¥+©0vÅwPÙY›1±»=À*SåÂ'cŸ•r®@Gäwðo¹•Jd±þ*–úeסM7Èó?Æî«›;‡¯*eè¯ ,e«ëèš{ê5Înðþ,œ¼²ý.îiäM0’vŒ„ùAnE}io³Æ¦ÁÓä²aO68ÖUéŸ&¨xòkåóÜÖÕ²6¾˜=Ùñ €¿¯§ÅW Π3SÄÏ“Þ wˆXÕÁZf‚j‘+J‘nUc‹Ú¯;§àÍ5©ãÞ¬=cÎØ6J¬%q øqÛ»™ô³FßoxVì©¡-•­xÈ{‰Ž¥ØÆö5Ï):Q“%˜AMXŸ¹¶›Zk¯Z|r$—ü0>g!ƒôëyL0‘µ–G5ak\ïçD!|Lò]xn<²œ5Ù›Ö2€@%õ¬D=yXùeòŠÈqî°Ë.tkp} 6“€ •N3ñäñGÄ]8Ì8§9dâèB$¥©+£ÑQ€@þÊ 8»RL¦7 ÉCÌ# íße,L!ðEßêC‡»o E¦}„–’)äåú¾‹áBš݀Þ{IVÂÜÊîÚÐHƒé@vc7g“=“8‘t (\f¦éX¿ýmíÜ« »â—¨ÌP Yª†ÁàâmûRJö˜ÍÌßAX=}›dŸù›ˆ; ‚qrijŒÞŒQ‹Š DPCìRé» åž„È(z±¢ò“vß~éäm„åŠ ".Ý”hT¥ç‡:Ÿ¯W,à˜½5à3ܼ›0|å­›J$9cuFÂŒ\}£öšCòÖ!kY7â\âWûìù¹¹3Rñü¤ƒfh·¸µ$³ oó×å oÌ(Ó`!sOÛ––”m ol"ÐŽ;ܽœ$óLd²ÉB÷À°%±5Ϙþ­wùDÀ‘¨Š+Ö0+)#»¨½M‹žN~Òl°ý`º¯}&œ ”f†@q¶‹Ä%&Y¬ÛCy¯z©²ù´!À‰>Úk0ßްJ¸¥&ìaÝ4I+Í×võB‡—»”Ï!¬¼h\£ÚÆoÿ›)ÁKµªv­“ü‰k[Œ\:R,G¿ÂQøRÅ©·{”)2Ó@“GFÇã}¿Cf)™Ò½8¸STço*voF èÓ¡—¥gŠÛv§a¬C‰­:'^îë3âVnÍ3‡¢¸¸ƒ¿xk5ˆbp‹îo›`:‚¤eªd zËj¸q»¨B Ù?¬ŒV “äÄ,Ä5‰r}û ýw¾’Ó½ûR9rs–½ÒÕÊo[GlžFa—õ©¿%“!Õ³ãpü–úRF›hÎ{©úìj-Ä%mháÐÏMC¥)d‹mˆï÷z«_M—ÿcµ\“k=a–dF潈)pŸ•yæá–(¾Üš/÷ÐïYsX êϧ3¼ŽTÿ ]•™b0ÇwxÒ `ì7³¹á¯vQêw_aP2=fr¾õ«¹CNIqì-3µÊjß×\ë¾ÍüVdò%æ—›dòcù‡Ž6-~eâçRûwûꈔP‰\ibr}t<út]B}óHÃ-EÅNÅœM 9ÝåÀRÀ¶N,îÒeí%©káç –¹‰³3iMŽ—àÞ¬ÎÅwCZnò”2¯ôïx½1 Vùé“+ÊÜëxˆGÁˆKh‹óæT¯lÐy4å}÷~ŠÏì(é“Òƒ&Äðjy؉ckœù2{DPÆxÙßuºoÆû \PF‰ETÐÈóêâô¹C¬¥õàÞP±òÖ:Í@zGé}Ÿ-dsvÄw•¨…[ ²ü\yqãMd¨#ŠRkÁ5yЛßÍØu‰L…¾k”}¸Ý'L 5Ìm6>|† ýÚŽØŽþ´>Kí>A¢£µíoòˆ@+E‘iš®P‘ •z:2×Ú5iž¦Ð«6dG™tÿøÉ}‹¹Û·5* lMIg‡ùCŠöèâ BËx¦‘48æPŽŸŠwV¤Áœô Y ÂÆü¦³äWá¤4Ed8…•]Ó]r»0;*>´¿» ›ŽI(‰þúÝçäsœ7_Æ×О2lBŸÃ¦+Áxóú˜cH¼0Ki¨“ž±ÔÁÙÝàm€ˆÏXŒM7â«b•6=аŸ·¢±®»! ¹¹åžûÉOàÐ0M_M6^MNÔ-x=AT·ñSziÜ•©ìÉËfº”êßä-«Ë¢ÚŸc•@} ¾5”Øõeöf"R Á¡‘ª$ÿ(àçnV43š¿cü£]:|ÁqÞtç÷š6ŽPPÉ€ìÁ$´¥ '…K@u™ŽY,9Ñ:ñ~~(’" «]àÕà*¤:Á–=†ÒË2¼ú¥š¸~tÎ5e´P‘3ô òè½OÏñjbœv¤›î«¥vJúòÿÉ[e\tAxÙáqáØA´†ýP=Ø—u¢áä¬NˆÕñ~ŠF-úÙÏï¯>ÞðÉ“m…Å÷É·‹SF*–+¨ôÿÆ…¤ßQL>”ZP+[VˆÆšÜEê¸b®ëÍlÎæ±SC÷az£©¤È§t|sºXžÏŒ^Wÿ’kÅÕ J{¼à‡åÖaâ¡«ŽÍPÍÜ5ÃñÅô‘u¿YܬAŸ6ÝC%ßæ¹=Ñ[ pC¢š¨ÚÇpáhbÂûÏP›O‘Á3pº­îh€™úò‘µ•î›'ºr*=óYù¾L+HÖ~h± -XCWz…ñÅ®€ÿè@ ŸAzùv\ ¹r~•޽_ÚwçbeCÝ«Líø4Šwç¼ýUÁc¢)ÌòÅ$åä5UÐßW•Íê? ë6ªÊ k¹À,ɯ-p@ œÁ/éÃõFwº¼ƒ+K~â úRpLظá†<‘ñ 9/xÓéaìÜ“&ýÊý^Ñâ¬a·³˜3’k%Û!Ž"î‡Ö—Œ2ÊÙŸ¢©§»á¦Õd^^Åêã¦Ô±br{vÔº0êâÈï§-±Ûé3ÛF6Ï?Ây%ºp'\c¿ŠY4«÷‹(©{=™arר!±àŽN Ô‹‡s?;ó+—ÒkP½‹¦ˆçþ8÷OÍT?‚È!zþx!Ü,oNñºQn ­î¿Á1§žªÖ~:eЍe šñâVh“Ô¯I‡ROpwÿ&X@m÷÷މ²F¾¸Li¥hâ,߯6ës€}¨"¸EUIH23û4¾“öïñÅ ãò-ð¶#¿é ’%çDüeR\K^v¥ñ‘þÁJί|€-ÃBÕöÌ !pH Xò+Ô7ÈÝgz‘Úf»yüèÍÏÛí²³p/úM§4ðàSÆ\Â\‹Î·< ĬÓ%£BKSà+ó1vËÔA•t•{  ·ÂC’™ªû}iüŒ„±ô”PHü;•Û}ŠÿkU§Ðª*­d™ m¾±ÿÕK4ìêYüSÛANâ$: 8 Oƒß꥘þdì>Ì¢›…QÍ£éã5y"Iu?4Ø+1˜1Tµ›bÈ&!cøR… endstream endobj 128 0 obj << /Length 741 /Filter /FlateDecode >> stream xÚmUMoâ0¼çWx•ÚÅvHB²ó!qØmUªÕ^!1ÝH ¤í¿__‚—m ñóøyÆÌÝ·—ÍDUíÎLÂGÎ^͹½t¥™¤ß·§àî.kËËÑ4ýc*S³ç'öÒµåÆôì>]gë¦î,yÝ”‡KeFÖ×$mÞëÆS°»3¿&åq÷GðÉîRúº™pßêþ`I_Î3[d·Eæý4ݹn›'&9ç¶7UÚaãL)l:ŠÛ×MÕ zØê!YU—ý0rßåÑžo>ν9®›},—lúj'Ï}÷á4>Óç®2]ݼ³û[ivjs92V+V™½íhýÿØ ›~éñÊyû8&ÝX®²­Ìù´-M·mÞM°ä|Å–E± LSý7—ЊÝ~¤&–Êçø U´ –2´XÆ(p‹m“¡¦ÂÜÂÂ∠ËXXœ(W°8X&˜LR4â=z¨Åu«kTÌGEåïm7hçáË8KÉc`Iu(à!a <#œG´Ž »>ÃÎn-tJ!]O2Çø`œúñãÌSŒóø#§¸­'œâ,<Ø“L€%q¡O8\Ï€™:Žó 3ht ‡,ª+à9­uçgŽCwËpÞDÿ‚|ŽOžRÇɉ#ɇÛW ºmè—’®1NÃwH=8!õ Á éŒ4ôDCp&q"p¢œüBCT/ôŒ9ñ¡!ɨ~Bü }ÒéîRqÒ‰óTÂçFIŸúܨ™ÏŠ|nTìs£Ÿ¥|neEA¼;~æó¤òÛ<©â6OšßæI‹ÏyÒòsžtèó¤g>O:òyұϓN|žôÜçI/|ž´òyÒÚçI§>O:óyҹϓ.|žRîó” Ÿ§Tú<¥³ë¹_¾û¥ãmÂKz}öÊK×ÙÑ=·î¡ÃW7æú"ŸÚV¹{ÊÇÿŒž‹à/@̪X endstream endobj 129 0 obj << /Length 741 /Filter /FlateDecode >> stream xÚmUMoâ0¼çWx•ÚÅvHB²ó!qØmUªÕ^!1ÝH ¤í¿__‚—m ñóøyÆÌÝ·—ÍDUíÎLÂGÎ^͹½t¥™¤ß·§àî.kËËÑ4ýc*S³ç'öÒµåÆôì>]gë¦î,yÝ”‡KeFÖ×$mÞëÆS°»3¿&åq÷g>Ù]êC_7î[Ý,ç«ifkì¦ÆÜ’Ÿ¦;×móÄÄ#çÜò¦JÛ#<œƒé ƒMGeûº©ºA ÛAZ $«ê²Fî»<ÚÃÀâÍǹ7Çu³oƒå’M_íä¹ï>œÂ‡`úÜU¦«›wv£ÌÎl.§ÓÁ@ãÁjÅ*³· ­÷Û£aÓ¯ ^)o'ä RU¶•9Ÿ¶¥é¶Í» –œ¯Ø²(ViªÿæZ±ÛÔÄRù_¡ŠVÁR†Ën±-`2ÔT˜ÛBXXQÁâ` ‹å Ë“IŠF\ ‡Bµ¸îbu ’ù¨¨ü½íí<\`Gc)y ,©<$¬g„sàˆvàÀ1a×gØÙ­…N)¤ëI&âŒS?^`œùqŠqþä·5ð„ó Bœ…€{’ °$.ô çë°ó SÇqd­‚®AáEBu<§µŽã¼ñÌqèrxΛˆà_¯ÂñÉSê89q$0ùpûJA· ýRÒ5fÀ aø©§'¤ž!8!‘†žhˆÎ$"NN”“_hˆ ò…ž1'>4$ÕÁOˆŸ¡OB:Ý]*N:qžJøÜ(és£BŸ5ó¹Q‘ÏŠ}nTâs£”Ï¢¬(èQƒwÇÏ|žT~›'UÜæIóÛO:öy҉ϓžû<é…Ï“V>OZû<éÔçIg>O:÷yÒ…ÏSÊ}žRáó”JŸ§tv½3÷Ëw¿t¼MxG¯¯^yé:û ºÇÖ=txâêÆ\ßãS{Â*÷qùø·Ñsü'©‹ endstream endobj 130 0 obj << /Length 494 /Filter /FlateDecode >> stream xÚm“Mo£0†ïü ï!Rz ˜|U‰ÄAÊaÛª‰V½&ö$E 6ÿ~=HÕUAgÞ¿“ɯ÷½Ÿ«ú~üÌÙ´uo$ø›ßÇÆ›LD-û t÷  @Ùö…½›Zî¡cÓÍNìtÙ=YñNËk¯`T=­áRêo ÞæøôeCîŸúòÚ•Úç(>”ÝÕŠæ™ ²ŸAæŠþ€iËZ¿°ð™sn[­6u…c´^0XaÁhî\je?ì„î¼0bª”ÝprOYÙ÷Åû[ÛAµÓçÚKS|ØdÛ™›óøäoF)õ…MZ³©}ß4W@Œ{YÆœmG;ÿë±<œñ®9Ü`‘;‡äKÖ Úæ(Áõ¼”óŒ¥E‘y Õ¹¡ât¤ba¥bi<Îg®bÌÅw­ü:/]×åvYsäˆâ[ä˜â+䄘#ψ]íœôò‚â9ò’8D^osâyMìîÚGÈ‚X o‰ä‚îBŸÉà5Éà‰<øÇ»’ÁÿÂò k£(Do9Örá,Âq¼B?"tŽýEDqì)bbœW$ÄèYÌèM»>sb×gEìjqÞ(ŒæÃ×po¿$îÝ}IdoŒÝ·œn-p!J ÷ýmê«ÜÏ-þøOÃÓ[áýL‡ endstream endobj 131 0 obj << /Length 696 /Filter /FlateDecode >> stream xÚmTMoâ0½çWx•ÚÅ$ !Ù ‘8l[•jµWHL7IP‡þûõ¬V=Mžß̼ñ s÷ëu;ÑU··õÈÙ›=w—¾´“ì÷îÝÝå]yil;<[[Ùj<=?±×¾+·v`÷Ù&ß´õðàÈ›¶<^*;²~&ûQ·‚>ìþÝþ”MS >Ù_êãP·ò{=éÇsæ@öd”ôÇöçºkŸ˜xäœ;`ÝVY×`Œs4½JaÓQÜ¡n«þª‡í¡.’Uu9\ßèY6î>¼ý<¶Ù´‡.Z.ÙôÍž‡þ“4>DÓ—¾²}Ý~°û¯ÒÜÑör:-d0­V¬²WÑÍÿ¼k,›þ8ãóþy²LÒ»ðºÊ®²çÓ®´ý®ý°Ñ’ó[Å*²mõíLrŸ²?ŒÜÔqù¥ã• â5F8@ šˆ=@Šð)&°  È8Ô¹€ÂÅRx u€Dº\j2H—†ª¡ÐVÁ¹0CzL]ø Âb°ct‘I ©g$`htÑ‹0œÆ\F„áŒ0ä†sê‡á jd< —Iê6œ»õñzgóñºË»þê W ¤qÈ’£+—Ÿ#ö•ñÌÇkÄÞ .‰bªsré…¤šáæÄç†bïmŽXú¾„Kß7ǵHß7Géû„û¾nb§>&jÊØµäuœ¯¼ú•ñ1ÜV™÷•âÜãâµÇ‰Ou$ÕŸqWèS/%1{\øxB!€§ÔK(hH©—TЖ枃»J©Ïϯv×ÜëÁ=küÒ2ø¥UðKÏ‚_:~é$ø¥Óà—ÖÁ/¿Œ ~™Eð+7¿èË¢/ ÿlì¡ÛÒ(/}ïö -+ZXukoûìÔE?Z„ãæÅÛKýqíƒÄ endstream endobj 132 0 obj << /Length 695 /Filter /FlateDecode >> stream xÚmTMoâ0½çWx•ÚÅ$ !Ù ‘8l[•jµWHL7IP‡þûõ¬V=Mžß̼ñ s÷ëu;ÑU··õÈÙ›=w—¾´“ì÷îÝÝå]yil;<[[Ùj<=?±×¾+·v`÷Ù&ß´õðàÈ›¶<^*;²~&ûQ·‚>ìþÝþ”MSÏ'ûK}êvÂÁ}¯‡£ãütÌƾ`ŒRþØþ\wíœs¬Û*ëÌpަWl:*;ÔmÕ_Ű=¤EB²ª.‡ë=ËÆ]’·ŸçÁ6›öÐEË%›¾¹ÃóÐ’‡húÒW¶¯ÛvÿE™;Ù^N§£… Æ£ÕŠUöà ºÙŸweÓŸ¼QÞ?O–Iz^UÙUö|Ú•¶ßµ6Zr¾bË¢XE¶­¾IîSö‡‘›:.Ÿã¡t¼r€A¼&ÀT±H>Åä4€"÷€P¸X ¤H— BMéÒ@5Ú*08WfH© ŸAX vŒ.2I ##õŒ .z†Ó˜Ëˆ0œQæa8ƒtcÂpNò0œAŒ‡á2 @݆s·>^ïl>^wùo×_áj4RrtåÒãsľÒ žùxØ[ ¡Á%QLu®óQ.Ý¢T³ ¼ÀÜ‚øÜPì½ÍKß—péûæ¸éûæ¨#}ß‚pß×MìâÔÇÄAM²–¼Ž3ð•7CB¿2>†Û*ó¾Rœ{¼@¼ö8ñ©Ž¤ú3îï }cê¥$f Oˆ#ð”z  )õ’ ÚÒÜspW)Õá9ðùÕNàš{=¸g-‚_Z¿´ ~éYðKÇÁ/¿tüÒ:ø¥óà—‘Á/³~eâæ}Yô%áŸ-tÛå¥ïÝ:¡UEk ¢ním›º²èGkp\ºx{)¢ÿÔÞ‚÷ endstream endobj 133 0 obj << /Length 696 /Filter /FlateDecode >> stream xÚmTËnâ@¼û+f‘’a؆!ÍØXâ°I¢Õ^Á²–°Œ9äïwª3ÊÔ.WwWwA?üzßNtÕííD=söaÏÝ¥/í$û½;EyW^Û¯ÖV¶ßž_Ø{ß•[;°Çl“oÚzxräM[/•Y?“ŒýªÛ@AöøiÿNʦ©÷‚Oö—ú8Ô턃ýYGÇú™ÀÊîPFil®»ö…‰gιÖm•u &9GÓ«6õê¶ê¯’Ø#!YU—Ãõ‰¾ËÆ­ÉÛïó`›M{è¢å’M?ÜËóГʧhúÖW¶¯Û/öx§Í½Û^N§£…Æ£ÕŠUöàJº¼î˦?y#}~Ÿ,“ô,¼²²«ìù´+m¿k¿l´ä|Å–E±Šl[ݽ“ܧì#7u\>Ç—ÒñÊñš# PMÄH Eø“XÐdjˆÜ @áb)<:@"].5¤KCÕPh«Àà\˜!=¦.|a1Ø1ºÈ$ŽŒÔ304ºèENc.#ÂpF˜‡á Ò Ã9uÈÃp52†Ë$uÎm}\ïl>®»ü·ë¯Îpµ@‡ )9ºréñ9b_iaÏ|¼Fì-ÐÐà’(¦:×ù(—¶($Õ,/0· >7{osÄÒ÷%\ú¾9Ö"}ßu¤ï[îûº‰]œú˜8¨)cCÖ’×q¾òfHèWÆÇp[eÞWŠsˆ×'>Õ‘TÆý®Ð7¦^Jbö¸ð1ð„8BO©—PÐR/© -Í=»J©Ïϯv×ÜëÁžµ~iüÒ*ø¥gÁ/¿tüÒiðKëà—΃_F¿Ì"ø•‰›_ôË¢_þÙ¸D·«Q^úÞ:Wt&p êÖÞ.Ú©;!‹>t Çó‹§·"úª#…® endstream endobj 134 0 obj << /Length 739 /Filter /FlateDecode >> stream xÚmUMoâ0¼çWx•ÚÅvHU„dçCâ°mUªÕ^!1ÝH ý÷ëñ#xÙö?ŸgìÁÜýx]OTÝmÍ$|äìÍœºs_™Iöss îîò®:L;<S›zœ==±×¾«Öf`÷Ù*_µÍð`É«¶ÚŸk3²¾'ióÑ´ž‚}Øý»ù=©½à“í¹ÙM;áà¾7ÃÞr¾›f¶ÆnjÌ-ùeúSÓµOLg~¼À8÷ã ãâþÈ)okà çA„8 ö$`I\èÎ×3`çAfŽã<ÈZ]ƒÂ!‹„ê xNkÇyã¹ãÐð"œ7Á¿ _¥ã“§Ìq âH`òáö•‚nú¥¤kÌÂðRONH=CpB:# =Ñ%8“ˆ88QA~¡!*ÉzÆœøÐäT?!~Ž> étw©8éÄy*ás£¤Ï }nÔÌçFE>7*ö¹Q‰ÏR>7в¢ G]¼;~îó¤ŠÛ<©ò6OšßæI‹¯yÒòkžtèó¤g>O:òyұϓN|žôÜçI/|ž´òyÒÚçIg>O:÷yÒ…Ï“.}ž2îó” Ÿ§Lú> stream xÚmUMoâ0¼çWx•ÚÅvHU„dçCâ°ÛªT«½Bbº‘ A!úï×ãGpÙö?ŸgìÁÜýxYOTÝmÍ$|äìÕœºs_™Iöss îîò®:L;ü2¦6õ8{zb/}W­ÍÀî³U¾j›áÁ’Wmµ?×fd}OÒæ½i=û°û7ógRúùd{nöCÓN8¨oͰ·”of™-±Ï%æü6ý©éÚ'&9ç¶P´uÖ`àL/"Øt”µkÚº¿(a[è „duS —‘û®ö$°xýqÌaÕîº MÙôÕNž†þÃé{¦Ï}mú¦}g÷Ÿ…Ù‰õùx܈` étw©8éÄy*ás£¤Ï }nÔÌçFE>7*ö¹Q‰ÏR>7в¢ G]¼;~îó¤ŠÛ<©ò6OšßæI‹¯yÒòkžtèó¤g>O:òyұϓN|žôÜçI/|ž´òyÒÚçIg>O:÷yÒ…Ï“.}ž2îó” Ÿ§Lú> stream xÚmUMoÛ:¼ëW°‡éÁ5?$R. ¤d9ôMðð®ŽÄä ˆeC¶ù÷³k›m‘CŒÕp¹;;†wŸ~>Î|¿Ž3óEŠ_ñ¸?O]œ5ß¶‡âî®Ýwç]Oßcìc]=~?§}÷Oâ¾yhÆáô9%?ŒÝ۹׬“B|Æœ‚>âþ)þ;ëvÇw%gÏçáí4Œ3‰ä§áô–’>\ ‚‚6ý§ã°¿ õEJ™€õØ7ûÆ8ó 1¿’{Æ~ºðÏ`W(-ú¡;]¾è·Û%=°ùñýxŠ»‡ñe_,—bþ+-OÓ;qü\ÌL}œ†ñUÜÿI--=ž‡·B«•èãKª˜æÿ¾ÝE1ÿpÆ[ÎÓû! Mߊyuû>Û.NÛñ5K)Wb¹Ù¬Š8ö­iÇ[ž_®¹uÊ•MúÑzQ­Š¥Ò)V†€Ú(TØ€àx¿àÞ¢ žjy‹°°!ÀÐÔ•µZÔÀ2àP="¦ZdÔ0\ÃG©R\¡·”).–2*ÎШa!„U¼Ä,†³ÔÛHð° `+jÐÃ.¸5Nα@èâ°èÐVK-àxŸ%ô˜Ü3š% A°YÓ€z¡ÎšÔ>kP#¬³¦õ™5m0W£oš¦Ã¾žj­®§Üý·.†ÐZ¡ŽT$X/©)n)æ#W—„o(æ“oÀRZÞ $K¢p4’ŽZ¶-bâ\­1¦Ü°Jä æP"Gñ‘XÔQ¬‚i/8ºkÉ^€ÂZqŒ:ZsŒ½š9”d š­Bù Ž)ßsLù-ï7½æx˜ÏJ›¡¾Ò`¯ažÉ½)f¥É$†µ’1™¸ dÑŠcªCZCù<£7Ã3JÊgózÌnøþHȰíáÌYÉšäTœ¯a…Šï¯Æ,_»œ-Ÿ—Oë87Ë}êÛKÔ´Ü—Ll¹oKñšò+Êg­JÌâ.¾GZyóº‹Vðc­48¸’ï¼äØWtù]Í:P~`áŒñ±–rZŽq.nÍ1]Ç ÇàSÿæ/©ßP•ýïuö¿7Ùÿ¾Ìþ÷Uö¿·ÙÿÞeÿû:û?Èìÿ ²ÿƒÎþ&û?”Ùÿ!dÿ‡&û¿1y–¦¼ÍH·œn5þ¹ã)º½ÝyšÒ“Bï½x#†1Þž´Ãþ€]ôGoáõñÅ×Mñ?®Xê endstream endobj 137 0 obj << /Length 900 /Filter /FlateDecode >> stream xÚmUMoÛ:¼ëW°‡éÁ5?$R. ¤d9ôMðð®ŽÄä ˆeC¶ù÷³k›m‘CŒÕp¹;;†wŸ~>Î|¿Ž3óEŠ_ñ¸?O]œ5ß¶‡âî®Ýwç]Oßcìc]=~?§}÷Oâ¾yhÆáô9%?ŒÝ۹׬“B|Æœ‚>âþ)þ;ëvÇ÷zö|ÞNÃ8“È}No)ç£e‘0ñ&hË?q:ûñ«P_¤” X}³ßa†c1¿ðó+³—aì§ ñ j…Ò¢ºÓå‹~»]›ß§¸{_öÅr)æ¿Òâñ4½ÃÏÅüÇÔÇi_ÅýÌÒÊãùpx‹`!d±Z‰>¾¤‚iöïÛ]ó¼¥<½¢Ðô­˜U·ïãñ°íâ´_c±”r%–›Íªˆcÿךv¼åùåš[§\Ù¤­ÕªX*be¨-€@E€-X€÷@à-ê©–·xkM PY«…@ ,Õ#bªE†A Ã5rEqIø†b>ù,¥å½A²$ G#é¨eÛ"&ÎÕcÊ «Dž`%r‰EÅ*˜ñ‚s »–ì(¬Ǩ£5ÇØ«™CIªÙ*”¿à˜ò=Ç”ßò^pÓkŽˆù¬¸ê+ öæ™Ü›bVšLbX+“‰«@­8¦:¤•1”Ï3Jp3<£¤|6¯Çì†ï„ Û^Μø—¬‰ANÅùV¨øþjÌRñµ«ÁÙòy9ð´Žcp³Üס¾½ÔAMË}ÉÄ–û¶¯)¿¢|ÖªÄ,îâ+pp¤•70¯»hÿ8ÖJƒƒ+ùNÀKŽ}E—ßÕ¬åöÎÈk)§åçâÖÓuÜp >õožñ’êð UÙÿ^gÿ{“ýïËì_eÿ{›ýï]ö¿¯³ÿƒÌþ*û?èìÿ`²ÿC™ýBöh²ÿ“giÊÛŒtËéVãŸ;¢Û³Ñ§)½(ôZÑK7bãíA;ìØEô^ß]|ýØÿs‰éQ endstream endobj 138 0 obj << /Length 867 /Filter /FlateDecode >> stream xÚ}UMkã0½ûWh…öF²ã¯²CÛ–¦,{Ml¥hìà$°ý÷«73nw—ÒƒÍÓèÍèͳ,]}{\OlÛoÝ$ºÕêÉúËиIù}s ®®ª¾¹\w¾w®uí8{ºSC߬ÝY]—«jÕíÏ7ž¼êš×KëFÖç¤Â½ì» ÖQ×ÏîçÄ5ÛßFçñ¤9LN—£&g3ÑHyÞŸ_=õ –òSê³)E~¸á´ï»;enµÖ>°ìÚ²? ±S0qj:ÊÝí»v…j ½ U»oÎ2¢wsð!yýv:»ÃªÛõÁ|®¦O~òtÞHïM0}Z7ì»uý™@OX_ŽÇW1J‹…jÝÎ×õ¾ÜoNM¿èúùüvt*¤±aMߺÓqÓ¸aÓ½¸`®õBÍëz¸®ýoÎDœ²ÝÜ¥çêÚ¿Bå‹`nlB ˜„‘„­=öÌã¸æ@æ )UÖ 9yŽ€IÁ(±JÅ5<æ§T`,© M%5ŠÖœR£h”ºäRê ®á1ÚûÌgcßÍ¯Í yq(¬ ábŒÆuX&Àá &èq,–Ñ1Ç+à„±N97Î8Nüœsk`Ëq8­ ^—8%Ç àŠ½FMq.â†5„Sâhú@ìkO × Ápý$Áƒqù1¦/]}Œ©Îò/þÈ©ÿ»pÒ^`ÜD3F?©ìx”‘ׯ[ë±a ¯³1´ecÔÏfŒ—Àäµ!/²„1êg)câdÜ?4dâ­K^˜|É ÆÐœ•ŒáQV1¦úÔ¿‰±'²š1tæ¬?ƺ9ëÁÏY?í¡œõÇГ³þ„rY‚ÞsÖŸŸõ'Äg)4ç¬3Å7ÎYgD¹¬3¢\ÖièÇbŸ-z±â3z´âs ,>G|ÆZV|ƾ´â3Öµâ3ü´â3qÄgônÅgè·â3tZñ½[ñ¾Yñ™ê‹ÏÐoÅgè,Äg¬[ˆÏàâ3ø…ø =…øL¹â3z/Ägâ‹ÏÄÏød ,gz)ÄôRˆÿ؇…øO5ù_±T“ÿ“‚êˆÿàT¼V *ŽÇM2G˜çªZN(:‘pTã’y¿ šË0øk‚n":÷qâï;÷~Yû#²è¡[n¼h1z¨ƒ?¾çF endstream endobj 139 0 obj << /Length 868 /Filter /FlateDecode >> stream xÚ}UMkã0½ûWh…öF²ã¯²CÛ–¦,{Ml¥hìà$°ý÷«73nw—ÒƒÍÓèÍèͳ,]}{\OlÛoÝ$ºÕêÉúËиIù}s ®®ª¾¹\w¾w®uí8{ºSC߬ÝY]—«jÕíÏ7ž¼êš×KëFÖç¤Â½ì» ÖQ×ÏîçÄ5Ûßf6Ó“æ09]Žn˜œÍD#åy~õÔ/XÊO©Ï¦øá†Ó¾ïÕZûÀ²kËþ€ÆNÁTÄ©é(w·ïÚAª-ô&Tí¾9ˈÞÍÁ;„äõÛéì«n×ó¹š>ùÉÓyx#½7Áôahݰï^Ôõg=a}9_Ä(,ªu;_×ûr¿985ý¢ëwæóÛѩƆ56}ëNÇMã†M÷â‚¹Ö 5¯ëEàºö¿9qÊv7r—ž«kÿ u”/‚¹A² )`JbD>`´öØ2ãš™$`¤TY'`ä`ä9&£Ä*×ð8˜W`TœR±¤&4–`Ô(ZsJ¢5Rê’H©+¸†Çhì3Ÿ}7¿6ƒXäÅ¡°6„‹0×a™G„+˜ gıXFǯ€Æ8åÜ8ã8ñsέ-Çá´.x]â”/€+ö5MĹˆÖPN‰£é±¯<1\?‚Ãõ“Æåǘ¾tõ1¦:Ë¿ø#§þ7^ïÂYH{upÍý¤²àQF^o­Ç†1¼ÎBÆÐ–EŒQ?›1^“׆¼ÈƨŸ¥Œ‰“qÿЉ´.yafð%+CsV2†GYŘêSÿ&ÆžÈjÆÐ™³þëæ¬??gý´‡rÖCOÎúÊeý zÏYJ|ÖŸŸu¦Ðœ³Îß8gå²ÎˆrY§¡‹}¶èÅŠÏèÑŠÏ1°øLñkYñûÒŠÏX׊ÏðÓŠÏÄŸÑ»Ÿ¡ßŠÏÐiÅgônÅgøfÅgª/>C¿Ÿ¡³Ÿ±n!>ƒ_ˆÏàâ3ôâ3åŠÏ轟‰/>?ã“°œEè¥ÿÑK!þcâ?ÕäÅRMþO ª#þƒSñZ1|¨8Gxp4Éaž«j9¡èDÂQKæý.h.Ã௠º‰èÜlj¿ïÜûeuìÈ¢‡n¹ñ¢Åè¡þðç( endstream endobj 140 0 obj << /Length 866 /Filter /FlateDecode >> stream xÚ}UMo£0½ó+¼‡Jí! ᫊"R»­šjµ×œ.RIý÷ë73´»«ªÐóøÍøÍÃØWß¶3ÛÏnÝjõèNÃelܬü¾;WWÕÐ\®?ÿp®uí4{ºSãÐlÝY]—›jÓwçOÞôÍë¥uësRá^ºþƒ‚uÔõ“û5sÍx0:gÍavºÝ8;›™FÊSw~õÔ/XÊO©Ï¦øéÆS7ôwÊÜj­}`Ý·åp@c§`.âÔ|’»ïúv…êzª¶kÎ2¢wsð!yûv:»Ã¦ßÁr©æ~òtßHïM0¿[7vý‹ºþL 'l/Ç㫃¥ƒÕJµnïëz_~ìNÍ¿èúùôvt*¤±aÍкÓq׸q׿¸`©õJ-ëz¸¾ýoÎDœò¼Ÿ¸kÏÕµ…:ÊWÁÒ Ù„0%1 " 0Z{ì™ÇqÍÌ0Rª¬0r0ò“‚Qb•Škx,+0*N©ÀXSšk0j­9¥FÑ)uɤÔ\Ãc´?õ™/¦¾›ß»Q,òâPXÂÅë°L€#ÂLÐ âX,£cŽWÀ c œrn œqœø9çÖÀ–ãpZ¼.qJŽÀ{š&â\Ä k(§ÄÑôØ×ž®AƒáúI‚ãòcL_ºúSõ_ü‰Sÿ/„wá"¤½À: ¸‰fŒ~RÙ ð(#¯·ÖcÃ^g!chË"ƨŸ-¯ÉkC^d cÔÏRÆÄɸhÈÄ Z—¼0 ø’Œ¡9+ã¬bLõ©cOd5cèÌYŒusÖƒŸ³~ÚC9ë¡'gý å²þ½ç¬?%>ëOˆÏ:ShÎYgŠoœ³ÎˆrYgD¹¬ÓÐÅ>[ôbÅgôhÅçX|&ŽøŒµ¬øŒ}iÅg¬kÅgøiÅgâˆÏèÝŠÏÐoÅgè´â3z·â3|³â3ÕŸ¡ßŠÏÐYˆÏX·ŸÁ/Ägð ñz ñ™rÅgô^ˆÏÄŸ‰ŸñÉ@XÎ"ôRˆÿè¥ÿ± ñŸjò¿b©&ÿ'ÕÿÁ©x­>T#<8šdŽ0ÏUµœPt"á¨Æ%ó~4—qô×ÝDtîãÄïz÷~Y‡#²è¡[nºh1º¯ƒ?klç_ endstream endobj 141 0 obj << /Length 866 /Filter /FlateDecode >> stream xÚ}UMo£0½ó+¼‡Jí! ᫊"R»­šjµ×œ.RIý÷ë73´»«ªÐóøÍøÍÃØWß¶3ÛÏnÝjõèNÃelܬü¾;WWÕÐ\®?ÿp®uí4{ºSãÐlÝY]—›jÓwçOÞôÍë¥uësRá^ºþƒ‚uÔõ“û5sÍx0¡Ö³æ0;]ŽnœÍL#å©;¿zê,å§ÔgSŠ ütã©ú;enµÖ>°îÛr8 ±S0qj>ÉÝw};ŠBõ ½ UÛ5gÑ»9x‡¼};ÝaÓï‡`¹TóG?y:o¤÷&˜ß­»þE]&ж—ãñÕAŒÒÁj¥Z·÷u½/?v§æ_týÎ|z;:ÒØ°ÆfhÝé¸kܸë_\°Ôz¥–u½ \ßþ7g"NyÞOܵçêÚ¿Bå«`ilB ˜„‘„­=öÌã¸æ@æ )UÖ 9yŽ€IÁ(±JÅ5<–§T`¬© M55ŠÖœR£h”ºäRê ®á1ÚŸúÌSßÍïÝ(yq(¬ ábŒÆuX&Àá &èq,–Ñ1Ç+à„±N97Î8Nüœsk`Ëq8­ ^—8%Ç àŠ½FMq.â†5„Sâhú@ìkO × Ápý$Áƒqù1¦/]}Œ©Îú/þÄ©ÿ»pÒ^`ÜD3F?©ìx”‘ׯ[ë±a ¯³1´ecÔÏŒ×Àäµ!/²„1êg)câdÜ?4dâ­K^˜|É ÆÐœ•ŒáQV1¦úÔ¿‰±'²š1tæ¬?ƺ9ëÁÏY?í¡œõÇГ³þ„rY‚ÞsÖŸŸõ'Äg)4ç¬3Å7ÎYgD¹¬3¢\ÖièÇbŸ-z±â3z´âs ,>G|ÆZV|ƾ´â3Öµâ3ü´â3qÄgônÅgè·â3tZñ½[ñ¾Yñ™ê‹ÏÐoÅgè,Äg¬[ˆÏàâ3ø…ø =…øL¹â3z/Ägâ‹ÏÄÏød ,gz)ÄôRˆÿ؇…øO5ù_±T“ÿ“‚êˆÿàT¼V *ŽÇM2G˜çªZN(:‘pTã’y¿ šË8úk‚n":÷qâw½{¿¬ŽÃYôÐ-7]´Ý×ÁÎ%ç# endstream endobj 142 0 obj << /Length 867 /Filter /FlateDecode >> stream xÚ}UÁn£0½óÞC¥öƆ¦Š"R»­šjµ×œn¤"’ú÷ë73iwWU çñ›ñ›‡±¯¾=¬'®žý$¹ÕêчóØúIõ}sˆ®®ê¡=ï}úá}ç»ËìñN=ŒC»ö'u]­êU¿;Ýòªo_Ï¿°>'•þe×P°Žº~ò¿&¾÷&í¤ÝOŽçƒ''3ÑHyÚ^õ – Sê³)E~úñ¸ú;enµÖ!°ì»jØ£±c4qjz‘»ÝõÝ( Õ3ôF&VÝ®=ɈÞí>8„äõÛñä÷«~;Dó¹š>†Éãi|#½7Ñô~ìü¸ë_Ôõga}>^=Ä(-ªóÛP7øòc³÷júE×ï̧·ƒW1 kl‡Î›Ö›þÅGs­jÞ4‹È÷Ýs&á”çí…» \Ý„W¬“bÍ ’MLS!"q Ø# uÀ!`NØÈÀÈ©²ÎÀ(À( LF…Uj®p4¯Á¨9¥cIMh ,ÁhP´á”E¤4ÒÔp€Ñþ¥Ïbvé»ý½Å¢ …µ!\Î€Ñ¸Ž« 8!\Ã=#ŽÃ2:åx œ1vÀ9çæÀ–ãÄ/8·v‡Óºäu‰Sq¼®ÙkÔ4 ç"nXCI8'ަľ&ðÄpý ×Ï2<WcúÒõǘê,ÿâ_8Í¿1ðbxÏbÚ ¬Ã€›iÆè'—Ý,ym‚µÆðÚÆŒ¡Í&ŒQßÎ/ÉkC^ØŒ1êÛœ1q,÷ V¼ uÉ 3ƒ/¶d Ͷb l͘êSÿ&Åž° cè,XŠu ÖŸ‚_°~ÚCëO¡§`ýå²þ ½¬?'>ëψÏ:sh.XgŽo\°Î„rYgB¹¬ÓÐÅ>;ôâÄgôèÄçX|&ŽøŒµœøŒ}éÄg¬ëÄgøéÄgâˆÏè݉ÏÐïÄgètâ3zwâ3|sâ3ÕŸ¡ß‰ÏÐYŠÏX·ŸÁ/ÅgðKñzJñ™rÅgô^ŠÏÄŸ‰oùd ,gz)ÅôRŠÿ؇¥øO5ù_qT“ÿ“’êˆÿàÔ¼V jާ M2G˜çêFN(:‘pTã’y¿ Úó8†k‚n":÷qâïzÿ~Y†²è¡[îrÑbtßDÒÓçn endstream endobj 143 0 obj << /Length 866 /Filter /FlateDecode >> stream xÚ}UMo£0½ó+¼‡Jí! ᫊"R»­šjµ×œn¤" ‡þûõ›™´»«ªÐóøÍøÍÃØWß¶3ÛöÏnÝjõèÆþ<4nV~ß‚««ªoÎG×M?œk]{™ïÔÃÐ7[7©ërSmºÃtãÉ›®y=·îÂúœT¸—C÷AÁ:êúÉýš¹fœŒÎãYsœç“f“™i¤<¦WOý‚¥ü”úlJQŸn}w§Ì­ÖÚÖ][öG46s§æ¹ûC×¢P=Co`BÕšIFônŽÞ!$o߯É7ݾ–K5ô“ã4¼‘Þ›`~?´n8t/êú3ž°=ŸN¯b”V+Õº½¯ë}ù±;:5ÿ¢ëwæÓÛɩƆ56}ëÆÓ®qî{qÁRë•ZÖõ*p]ûßœ‰8åyá®=W×þê(_KƒdRÀ”Ä$ˆ| 4Àhí±dÇ52HÀH©²NÀÈÁÈsL F‰U*®áq°¬À¨8¥cMMh ¬Á¨Q´æ”Ek¤Ô%RWp Ñþ¥Ï|qé»ù½Ä"/…µ!\,€Ñ¸Ë8"\Á½ ŽÅ2:æxœ0¶À)ç¦ÀljŸsn l9§uÁë§äx\±×¨i"ÎEܰ†‚pJMˆ}à‰áú4®Ÿ$x0.?Æô¥«1ÕYÿÅ¿pêcà…ð.\„´X‡7ÑŒÑO*»eäµñÖzlÃë,d mYÄõ³ã50ymÈ‹,aŒúYʘ8÷ ™xAë’f_²‚14g%cx”UŒ©>õob쉬f 9ë±nÎúcðsÖO{(gý1ôä¬?¡\ÖŸ ÷œõ§Ägý ñYg Í9ëLñsÖQ.ëŒ(—uú±Øg‹^¬øŒ­ø‹ÏÄŸ±–Ÿ±/­øŒu­ø ?­øLñ½[ñú­ø V|FïV|†oV|¦úâ3ô[ñ: ñëâ3ø…ø ~!>CO!>S®øŒÞ ñ™øâ3ñ3>ËY„^ ñ½â?öa!þSMþW,Õäÿ¤ :â?8¯ÇŠãq„G“Ì湪–ŠN$Õ¸dÞï‚æ< þš ›ˆÎ}œø‡Î½_V§þ„,zè–»\´Ý×Á€ç‡ endstream endobj 144 0 obj << /Length 866 /Filter /FlateDecode >> stream xÚ}UMo£0½ó+¼‡Jí! ᫊"R»­šjµ×œ.R!‡þûõ›™´»«ªÐóøÍøÍÃØWß¶3ÛÏnÝjõèNÃylܬü¾;WWÕМ®Ÿ~8׺ö2{ºSãÐlݤ®ËMµé»éÆ“7}óznÝ…õ9©p/]ÿAÁ:êúÉýš¹fêŒÎãYs˜ÎG7Î&3ÓHyê¦WOý‚¥ü”úlJQŸn€‘Re€‘ƒ‘瘌«T\Ãã`YQqJÆššÐXƒQ£hÍ)5ŠÖH©K ¥®à£ýKŸùâÒwó{7ŠE^ kC¸X£q– pD¸‚ zA‹etÌñ 8alSÎM3Ž?çÜØrNë‚×%NÉñ¸b¯QÓDœ‹¸a á”8š>ûÁÃõ#h0\?Ið`\~ŒéKWcª³þ‹áÔÿÆÀ á]¸i/°n¢£ŸTv<ÊÈkã­õØ0†×YÈÚ²ˆ1êg Æk`òÚYÂõ³”1q2î2ñ‚Ö%/̾dchÎJÆð(«S}êßÄØYÍ:sÖcÝœõÇà笟öPÎúcèÉYB¹¬?Aï9ëO‰Ïúâ³ÎšsÖ™âç¬3¢\ÖQ.ë4ôc±Ï½Xñ=Zñ9Ÿ‰#>c-+>c_ZñëZñ~Zñ™8â3z·â3ô[ñ:­øŒÞ­ø ߬øLõÅgè·â3tâ3Ö-Ägð ñüB|†žB|¦\ñ½â3ñÅgâg|2–³½â?z)ÄìÃBü§šü¯XªÉÿIAuÄp*^+†ÇãŽ&™#ÌsU-'H8ªqɼßÍyý5A7û8ñ»Þ½_VÇáˆ,zè–»\´Ý×Á&vçU endstream endobj 145 0 obj << /Length 866 /Filter /FlateDecode >> stream xÚ}UMo£0½ó+¼‡Jí! ᫊"R»­šjµ×œn¤"Bý÷ë73´»«ªÐóøÍøÍÃØWß¶3ÛöÏnÝjõèÎýehܬü¾;WWUß\Ž®8׺vš=ß©‡¡o¶nT×å¦Út‡ñÆ“7]óziÝÄúœT¸—C÷AÁ:êúÉýš¹fÎãYsœ/'7ÌF3ÓHy:Œ¯žúKù)õÙ”¢?Ýp>ôÝ2·ZkXwmÙÑØ9˜‹85Ÿäî];ˆBõ ½ U{hFÑ»9z‡¼};î¸éö}°\ªù£Ÿ<Ãé½ æ÷Cë†C÷¢®?è ÛËéôê Fé`µR­ÛûºÞ—»£Só/º~g>½œ ilXcÓ·î|Ú5nØu/.Xj½R˺^®kÿ›3§<ï'îÚsuí_¡ŽòU°4H6!L‰@Œ@ÂÈBŒ€Öû@æq\s óŒ”*댌ÆTgýâÔÿÆÀ á]¸i/°n¢£ŸTv<ÊÈkã­õØ0†×YÈÚ²ˆ1êg Æk`òÚYÂõ³”1q2î2ñ‚Ö%/̾dchÎJÆð(«S}êßÄØYÍ:sÖcÝœõÇà笟öPÎúcèÉYB¹¬?Aï9ëO‰Ïúâ³ÎšsÖ™âç¬3¢\ÖQ.ë4ôc±Ï½Xñ=Zñ9Ÿ‰#>c-+>c_ZñëZñ~Zñ™8â3z·â3ô[ñ:­øŒÞ­ø ߬øLõÅgè·â3tâ3Ö-Ägð ñüB|†žB|¦\ñ½â3ñÅgâg|2–³½â?z)ÄìÃBü§šü¯XªÉÿIAuÄp*^+†ÇãŽ&™#ÌsU-'H8ªqɼßÍeü5A7û8ñ{¿¬Ný YôÐ-7]´Ý×Á¢çŒ endstream endobj 147 0 obj << /Producer (pdfTeX-1.40.25) /Creator (TeX) /CreationDate (D:20240920153231+02'00') /ModDate (D:20240920153231+02'00') /Trapped /False /PTEX.Fullbanner (This is pdfTeX, Version 3.141592653-2.6-1.40.25 (TeX Live 2023/Debian) kpathsea version 6.3.5) >> endobj 21 0 obj << /Type /ObjStm /N 93 /First 778 /Length 4326 /Filter /FlateDecode >> stream xÚíksGò»~Å|<ê Ïûu•º*°16Ä@üŠB^ÛJdÉуüúëîI³»’‘ T%9ªpk4ÓÛÓïîÙ]a˜`šEÏ”cZZ¦3Ò0™5’iÁœÒ°ÀœQLk\`Ú°-Ó–Iá}G;øŒŽ)ø®•‚Ë™4°h4“paÒ9É ¬{˜4€€°ñði"ƒJxˬ`J Õ±>ƒaV1¥½dVÃgÌsN:fa+¤cç›ÚÜJÅ,ЉȩDA<0Î4Ðé8à_;ç@ÌëÆŸ ¸‹¡DA¾g:‚0Xø "Hàß;`„ðž™ eÇ` öõ ¢Z ÀZ¼[k`2(ø„MƒÆÏÈ@D`.‘ø ¤†O$e脤¤dPD Ì%‹ZÃw4ð T£h+)$# t@*h(ø MG P·¨l¸V‚ h.À•&a?)a{\¿èi…I¥$®ÂÊ a%Ì«à@ÀÕ° Œà ­­èH29ò&ü+^£õAHÍxt hh©Z¡ $ˆ*ÉEÂÖX$'t´}Ðl@)áBX-àr®Ä<\å4˜vp”àÑÓÀ¾ˆú "(õ@¯d¼VèaèŒydoMèüðC‡Ÿþq[1þª{Uuøîh8­†Ó ³"Ç~\MF³q¯š@¼ÐÄQuÑï>}bïC°; |%¹âûPÃå¸bÿ÷¿´Á>e?üÀø>XÖá Á}¢_ä!øsÌCpQ)æÈÀ¹œÏ£]æóh‹9RdƱhŠŒƒŠšÏ£æc ÉùX¢ÊÓt¬Ò¾À>5õNª){ÚÛgü´ú4eïç’­TòMÝ){å펫î´?îu§û×Þ [•‚ŒñˆŽ.Ö®ö§X9fOÇÝÛë~oÂ^Φ·³éìb֫ƸjvÌŽ|÷áÔƒº„hÁ?™}˜ÒWœ”þ¢{ƒkªÃw'Y™TƒÕ´ßëvø“aotÑ^a&ͽㇻ»ˆ|i 'ß›Ìñ¹ÅÕüMøh8é/—÷ú——¨ UûR,¿égÿm6šVƒêr ®`øÅ¤ŸLúüjÜýXñno6­x¯?îÍn.Õ'>í.*~ÓíGCþa\\ÒíõÀ ü¢[Lú¾3M/ªK>†½y,8t“׳áUw<»tgS>º «_y¯‹ô&·Ý^õþŸdE·ÊŠþÿÉŠE"[™²ê©ãüå‡_ª^ºàð2LÂ:¼ÿ%ùŨf~1âå—x‡gÆ¿Œg_ ´E½’âÃÇ£ÁÅJ²…Ûð=mý­œÃêoâK²…sXóÝ9’h±Þ9hí/âñÛ8G\áN|wŽ¿•s8ûMœcI¶t÷ßG-›§¢§j÷QxcdÞHÁf>¶L‡ùØ1¿¤Ãr±Ùa9ÿ5Ž¿€êEûÐë=žÕó8àQ…Ö‹ñγ£Y‚ÖËÅB; éw ¨EìÄC+üŽdå*Þ J£´ö%0S$˜ö(a¹·×8v1àŒˆ; D:¯OüŽÎ˜:(äP Á;œR0B©N¢œ(ä]”Æ}­Ç™´KÂ$ã’“Ì þK‰Ò|©µ´j¬Cý,g²Þ‹ùR3i^Y ²IV#KMfÊ2,ì§UDûiµ àl\b9·1X?HÔœ“¨ƒ»Ë‚;xSF#±ô%AT2hQU&¢ëÑ8Ík‰ñå*’5Œ%‘]7.‹µ„KsàÈók£$–ñŠfZ7Æa0h½€u x æwøÏ‡˜¦<:¾"˜ÆèOÍÑ&P @·ùoóË6 J`õì|¿åL¼ü¾ŽÉ…àJTåmÆt¨±$Ä ©vA$º ‰ iPÍaÁdÚÍG,+RèS9LwP 0‘&¤  îÅ#ÜPäÓªƒÜ0éH%°"Eù2:ÊÜÈF cÈÖp‘ ö€°§,±€™8†´•ÆŒäfteB®……ʰNšp®}M*J\ +Z L;Êî“Ðä§ ½µ?)h«üá#fÙË«œÆ½Û0NÚP&ϤÌ<ÿ>£–ƒÜ 3~¦½Ø£ÐqÐ(‡”Iá^QQ"VÓ”´éã®o‚.ˆÉBD"Q]%™Ëx&~F!–æ´Eˆ Ôt¡Zä‡MÿTA£Ì6‰Ísô áErSšu• ÜÚRkÐ}O;”ÛJÄÈQ§R¥KëyŽ®ÍÔ“#’¢2ý…£©Åövž2–‚6G… ˆXŽˆb¬S<.0²Ó¬ZÒH¶ÈÉe04yºŽD/¤XMÐ ·–hYE=­-˜ ‚À´ã-ö%ÁSyL>Ød¸Ê£¯Rçfœ èÂÖ Ä¤&|<¨hÂûG}‹Æ€JýŒ NÆÐò Þn€S )W@66>«¨sÇÃÝ™",MuÜ •¨RoN}^OñÚL›ZkhM,°B²‹/ ÚÈÈt)FšO$ËÕDØhKéx9V‘â©ÀÉê ù“i&P ”¯-ðÞ~ôd.™Ç´ ¦Q½6AŸÊÁ2‘gÒm3‡÷Ž5zÚ ¨zYÌñ–›Î˜š”n¨t#,æØ6*› 'QNò8b½°JÍ÷Hx#•ñrÆzl“÷»sIÒ|¢X®ZzKÐçÒH_¬f}à\´‹«}ÐTçTI‡Öµ£ àé6"—kI‚4&ø¾ñc¯šôÆýÛéhœnP§G¯v_¼ú÷îÑãs)`aнš0“0Ómï‡Ö±‡Š*½ÃŽXüÑŸ€þs·{{Põ¯®ákpŽÛàÚC‰‹‡Óî ß{4¼T ÈŸL«›×@VÎóEpØ×Ý1Þêþ?xöÞïÃQ­~³Z”×»Ç{O^“(a$1K¢´ý:’(µ¡$fI~:Ù={û$y²Ö(Ê€(ߌƒ³‚¥,F”¢ˆBè£Öb|!‡–¥ÆÝÞ¯ÕŸ*}è_]Í¿—&nñÈb™¾-'³›z,xџܺԔâ¶QÊóƒ'ûÏ~¥®SŠVÙ¾Â„ÏØWße߇è I1^­UÌsþ‚ŸóÞä—üš÷ùßVãþè‚ßöù§š¤aI_í?Ù%IÃÖ§”âk Ö zJ¢MÝÉu)”¢m¤:y¼{øäIõx¥¨t_ …\Rêµ!z^·4BÛt~özï t¼N#æþ(â—˜i¦.®5Rñ\·â„Á«ßfÝ¿ÍÆ|ÄGÊß.cµˆS~;˜Mø˜Oª›~o4Àg·ÕÇjȧ×ãªâÓßGüÿ³êªÚ*¡<þUµÆ¡õ¼Æ/¾Š¦¼[«©¤ÔÈVj«„ôâôìņéÉÛµY %i‰þìíR,ñ~qŠÙ}]¦îŽÓ«u¡¶Ê=§G»{H¨u¦§Ž_»ÇWý}ü2­7V[¹UÖùñøÙ›ó·ÿ>ÙÇþ&Ú5fRž=Ôøæ½WšE]s¿ºLQlä~V¬M:ˆ«Â0G]Ã'åV9é§§ONOž&qYç•pnJâ:‹Gû-Å=æ=HH—P@ ÃÔ%Û*…Ÿ½}r’­7¤ (þ MWѹëη—L‹R²Gü1ßå{|ŸðCþ úùô/ù+úÊæÍßð·¼Ë?€RnMýÂéäþõáßÇ*eé+è"®ÿ¸½7èó_ø¯|Ào@qÃ>¸ÌݼÓK;¹ýømù"P%,0@v²IÿŸ ~Æ?òß!ÍÿÁÿl§z¹UVÜ;=\¯ähæÑ$ðI‘¿¯_%ù3yoŸòݨû (¼{*™t‡‹,—óf¶+:­Êe«t»Tê”ø;ÿ³®»­JÆ›óŸß¿DÝÝá J òðç`Ú~‘ƒŠüó9•†nyª®¨WÔ‰äÀ—ÛÖ‰–“BÎo;*ºiÛEWdü5¯Ÿ*ŸªÓ~Q…–îPŸUÖ†^!Ë™¥ÉL'6‘­y›ž9&éI$ É•ÔC²“æ˜)˜SÈçpgÎÎ 6ó‚ïÃÁ¾]¶YPeåf!ò·Ð•Æ‚ðªÙ«öÖmÎ=Éù“ßÿR$Ý[(Ì)ñéáý&¯/_2nÞ¯«9?þˆ–^ZÝï'St;ºñc·øò¦1½žàÏw ótt6„SßZ=lúöó’Æ=·7vSn|ls·ç¦y߬ÁŽo²#CÁNKåˆ;ZlÏNóŽUƒØdGÙ‚(ÆK~l›¹=?ûJuvð×Õ ~œ,ù‘?¦ÍÚžŸÖ¡&G-oa†|›#½=G[:M~ZþŒ'¬Õüè6?f{~ê÷Mšì´üÙÄ‚´Þœ׿ÆnÏMó†G“Ÿ–C×´#Ü2ÜC›·=?{ v¤¼‹ÂT²ÍŒßž™ö†&?úNï±v©ž^‹Ÿ€øÞtdݾ9Ðd±åàAÕR@Á£Û€Ç{¤ðö1¿ÉcË땯aQãâçy4÷Èëí#n“ÇV$˜Zâ’~Ácðx\ß>6xT­ð°õäºLQnÀã=òûÙäQ&á.yô~ïQÚç¯&­˜Ña?z»÷¨ísN“Çv¥05ŒË˜ ð¸¢vàï©&øƒª>ìðçý 8¤ßj¥ãKnŒÿ ýð­vùnNR£«N¢Süœêyg…Ó endstream endobj 148 0 obj << /Type /XRef /Index [0 149] /Size 149 /W [1 3 1] /Root 146 0 R /Info 147 0 R /ID [<088E8AB048C0FF46071898E1F340F262> <088E8AB048C0FF46071898E1F340F262>] /Length 431 /Filter /FlateDecode >> stream xÚÍËK”qÆñó¼óš£Ùh:Si–—LË™ÔѼ¤N^RK+/i£• +%Z´¨UýEˆ;³“ ¼€¨T4´h†EA‚ uÑÂ@BBÏs6¾çå÷ŠˆìCü‰DÅ €è”T›ý$Jn“»¤“t“ä&é"íä¹N:H€»1gï݃<Òç'zm&Äç³é!.I ‡ åËvH$^’D’!§ô•·-Ê÷?öII<©²y„øH*I#G!/—ìšN2ˆŸÈ1ÈÌ;»''H&É"'!Ë£vÍ&§Èi’Cr!¿Gìz†äCþnÙì#wH/¹F ÈYRHŠÈ9rž“ ‘ ¤„”’2&夂\$•íöË* ±ªÂ5V5@ç«KÀã«Z`,fUÌŒ[Õ¿þ[E€½A«ËpkV p™Vp”Y5Á+±j†³³huž`³U <=ϬZá™[µÁ ½°º ÷U…À}}_™Ž)³«Êܦ²°¥¼ÙUÞ‡”ÃÊÇJåS²ò9®¬>= library(ks) set.seed(8192) samp <- 200 mus <- rbind(c(-2,2), c(0,0), c(2,-2)) Sigmas <- rbind(diag(2), matrix(c(0.8, -0.72, -0.72, 0.8), nrow=2), diag(2)) cwt <- 3/11 props <- c((1-cwt)/2, cwt, (1-cwt)/2) x <- rmvnorm.mixt(n=samp, mus=mus, Sigmas=Sigmas, props=props) @ \setkeys{Gin}{width=0.45\textwidth} \begin{figure}[!ht] \begin{center} <>= plotmixt(mus=mus, Sigmas=Sigmas, props=props, xlim=c(-4,4), ylim=c(-4,4)) @ <>= plot(x, xlim=c(-4,4), ylim=c(-4,4), xlab="x", ylab="y") @ \end{center} \caption{Target `dumbbell' density. (Left) contour plot. (Right) Scatter plot.} \label{fig:dens-db} \end{figure} We use \code{Hpi} for unconstrained plug-in selectors and \code{Hpi.diag} for diagonal plug-in selectors. <>= Hpi1 <- Hpi(x=x) Hpi2 <- Hpi.diag(x=x) @ To compute a kernel density estimate, the command is \code{kde}, which creates a \code{kde} class object <<>>= fhat.pi1 <- kde(x=x, H=Hpi1) fhat.pi2 <- kde(x=x, H=Hpi2) @ We use the \code{plot} method for \code{kde} objects to display these kernel density estimates. The default is a contour plot with the upper 25\%, 50\% and 75\% contours of the (sample) highest density regions. %, as %defined in \citet*{bowman1993} and \citet*{hyndman1996}. These regions are also plotted by the \pkg{sm} library. <>= plot(fhat.pi1) plot(fhat.pi2) @ The respective kernel density estimates are produced in Figure \ref{fig:pi}. The diagonal bandwidth matrix constrains the smoothing to be performed in directions parallel to the co-ordinate axes, so it is not able to apply accurate levels of smoothing to the obliquely oriented central portion. The result is a multimodal density estimate. The unconstrained bandwidth matrix correctly produces a unimodal density estimate. \begin{figure}[!ht] \centering <>= plot(fhat.pi1, main="Plug-in", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) @ <>= plot(fhat.pi2, main="Plug-in diagonal", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) @ \caption{Kernel density estimates with plug-in selectors} \label{fig:pi} \end{figure} The unconstrained SCV (Smoothed Cross Validation) selector is \code{Hscv} and its diagonal version is \code{Hscv.diag}. In Figure \ref{fig:cv}, the most reasonable density estimate is from the unconstrained SCV selector. <>= Hscv1 <- Hscv(x=x) Hscv2 <- Hscv.diag(x=x) @ \begin{figure}[!ht] \centering <>= fhat.cv1 <- kde(x=x, H=Hscv1) fhat.cv2 <- kde(x=x, H=Hscv2) @ <>= plot(fhat.cv1, main="SCV", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) @ <>= plot(fhat.cv2, main="SCV diagonal", cex.main=1.4, xlim=c(-4,4), ylim=c(-4,4)) @ \caption{Kernel density estimates with cross validation selectors} \label{fig:cv} \end{figure} The unconstrained bandwidth selectors will be better than their diagonal counterparts when the data have large mass oriented obliquely to the co-ordinate axes, like for the dumbbell data. The unconstrained plug-in and the SCV selectors can be viewed as generally recommended selectors. \bibliographystyle{apalike} \begin{thebibliography}{} \bibitem[Bowman and Azzalini, 2007]{sm} Bowman, A. W. and Azzalini, A. (2007). \newblock {\em sm: kernel smoothing methods: Bowman and Azzalini (1997)}. \newblock R package version 2.2. \bibitem[Duong, 2007]{duong2007c} Duong, T. (2007). \newblock ks: {K}ernel density estimation and kernel discriminant analysis for multivariate data in {R}. \newblock {\em Journal of Statistical Software}. \textbf{21 (7)}, URL \texttt{http://www.jstatsoft.org/v21/i07}. \bibitem[Simonoff, 1996]{simonoff1996} Simonoff, J. S. (1996). \newblock {\em Smoothing Methods in Statistics}. \newblock Springer-Verlag, New York. \bibitem[Wand, 2006]{KernSmooth} Wand, M. P. (2006). \newblock {\em KernSmooth: Functions for kernel smoothing for Wand \& Jones (1995)}. \newblock R package version 2.22-19. R port by Brian Ripley. \end{thebibliography} \end{document} ks/build/0000755000176200001440000000000014673274557012012 5ustar liggesusersks/build/vignette.rds0000644000176200001440000000027214673274557014352 0ustar liggesusers‹‹àb```b`aad`b2™… 1# 'fÏNIÕ Ê+Gf cSY’†&Ì 6h‘0X" ,LÈŠòsS‹Ñ tI-HÍK ÿîŸñ?šïÔÊòü"˜5lP5,n™9©0{C2Kàæ7(“1Ý 棸Ÿ³(¿\æ^P°6‰ÿ@€îÑäœÄbtr¥$–$ê¥õƒÜ òÍá ˜ks/man/0000755000176200001440000000000014606000765011446 5ustar liggesusersks/man/pre.transform.Rd0000644000176200001440000000210014336531532014527 0ustar liggesusers\name{pre.transform} \alias{pre.sphere} \alias{pre.scale} \title{Pre-sphering and pre-scaling} \description{ Pre-sphered or pre-scaled version of data.} \usage{ pre.sphere(x, mean.centred=FALSE) pre.scale(x, mean.centred=FALSE) } \arguments{ \item{x}{matrix of data values} \item{mean.centred}{flag to centre the data values to have zero mean. Default is FALSE.} } \value{Pre-sphered or pre-scaled version of data. These pre-transformations are required for implementing the plug-in \code{\link{Hpi}} selectors and the smoothed cross validation \code{\link{Hscv}} selectors. } \details{ For pre-scaling, the data values are pre-multiplied by \eqn{\mathbf{S}^{-1/2}}{S^(-1/2)} and for pre-scaling, by \eqn{\mathbf{S}_D^{-1/2}}{S_D^(-1/2)} where \eqn{\mathbf{S}}{S} is the sample variance and \eqn{\mathbf{S}_D}{S_D} is \eqn{\mathrm{diag} \, (S_1^2, S_2^2, \dots, S_d^2)}{diag (S_1^2, S_2^2, ..., S_d^2)} where \eqn{S_i^2}{S_i^2} is the i-th marginal sample variance. } \examples{ data(unicef) unicef.sp <- pre.sphere(as.matrix(unicef)) } \keyword{algebra} ks/man/Hlscv.Rd0000644000176200001440000000575314336766344013042 0ustar liggesusers\name{Hlscv} \alias{Hlscv} \alias{Hlscv.diag} \alias{Hucv} \alias{Hucv.diag} \alias{hlscv} \alias{hucv} \title{Least-squares cross-validation (LSCV) bandwidth matrix selector for multivariate data} \description{ LSCV bandwidth for 1- to 6-dimensional data} \usage{ Hlscv(x, Hstart, binned, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim", trunc) Hlscv.diag(x, Hstart, binned, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim", trunc) hlscv(x, binned=TRUE, bgridsize, amise=FALSE, deriv.order=0, bw.ucv=TRUE) Hucv(...) Hucv.diag(...) hucv(...) } \arguments{ \item{x}{vector or matrix of data values} \item{Hstart}{initial bandwidth matrix, used in numerical optimisation} \item{binned}{flag for binned kernel estimation} \item{bgridsize}{vector of binning grid sizes} \item{amise}{flag to return the minimal LSCV value. Default is FALSE.} \item{deriv.order}{derivative order} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{optim.fun}{optimiser function: one of \code{nlm} or \code{optim}} \item{trunc}{parameter to control truncation for numerical optimisation. Default is 4 for density.deriv>0, otherwise no truncation. For details see below.} \item{bw.ucv}{flag to use \code{stats::bw.ucv} as minimiser function. Default is TRUE.} \item{...}{parameters as above} } \value{ LSCV bandwidth. If \code{amise=TRUE} then the minimal LSCV value is returned too. } \references{ Bowman, A. (1984) An alternative method of cross-validation for the smoothing of kernel density estimates. \emph{Biometrika}, \bold{71}, 353-360. Rudemo, M. (1982) Empirical choice of histograms and kernel density estimators. \emph{Scandinavian Journal of Statistics}, \bold{9}, 65-78. } \details{\code{hlscv} is the univariate LSCV selector of Bowman (1984) and Rudemo (1982). \code{Hlscv} is a multivariate generalisation of this. Use \code{Hlscv} for unconstrained bandwidth matrices and \code{Hlscv.diag} for diagonal bandwidth matrices. \code{Hucv}, \code{Hucv.diag} and \code{hucv} are aliases with UCV (unbiased cross validation) instead of LSCV. For \pkg{ks} \eqn{\geq}{>=} 1.13.0, the default minimiser in \code{hlscv} is based on the UCV minimiser \code{stats::bw.ucv}. To reproduce prior behaviour, set \code{bw.ucv=FALSE}. Truncation of the parameter space is usually required for the LSCV selector, for r > 0, to find a reasonable solution to the numerical optimisation. If a candidate matrix \code{H} is such that \code{det(H)} is not in \code{[1/trunc, trunc]*det(H0)} or \code{abs(LSCV(H)) > trunc*abs(LSCV0)} then the \code{LSCV(H)} is reset to \code{LSCV0} where \code{H0=Hns(x)} and \code{LSCV0=LSCV(H0)}. For details about the advanced options for \code{binned,Hstart,optim.fun}, see \code{\link{Hpi}}. } \seealso{\code{\link{Hbcv}}, \code{\link{Hpi}}, \code{\link{Hscv}}} \examples{ data(forbes, package="MASS") Hlscv(forbes) hlscv(forbes$bp) } \keyword{smooth} ks/man/cardio.Rd0000644000176200001440000000137614336531231013202 0ustar liggesusers\name{cardio} \docType{data} \alias{cardio} \title{Foetal cardiotocograms} \description{ This data set contains the cardiotocographic measurements from healthy, suspect and pathological foetuses. } \usage{data(cardio)} \format{A matrix with 2126 rows and 8 columns. Each row corresponds to a foetal cardiotocogram. The class label for the foetal state is the last column: N = normal, S = suspect, P = pathological. Details for all variables are found in the link below. } \source{ Lichman, M. (2013) UCI Machine learning repository: cardiotocography data set. %\url{http://archive.ics.uci.edu/ml/datasets/Cardiotocography}. University of California, Irvine, School of Information and Computer Sciences. Accessed 2017-05-18. } \keyword{datasets} ks/man/kde.test.Rd0000644000176200001440000000607214340025574013463 0ustar liggesusers\name{kde.test} \alias{kde.test} \title{Kernel density based global two-sample comparison test} \description{ Kernel density based global two-sample comparison test for 1- to 6-dimensional data.} \usage{ kde.test(x1, x2, H1, H2, h1, h2, psi1, psi2, var.fhat1, var.fhat2, binned=FALSE, bgridsize, verbose=FALSE) } \arguments{ \item{x1,x2}{vector/matrix of data values} \item{H1,H2,h1,h2}{bandwidth matrices/scalar bandwidths. If these are missing, \code{Hpi.kfe}, \code{hpi.kfe} is called by default.} \item{psi1,psi2}{zero-th order kernel functional estimates} \item{var.fhat1,var.fhat2}{sample variance of KDE estimates evaluated at x1, x2} \item{binned}{flag for binned estimation. Default is FALSE.} \item{bgridsize}{vector of binning grid sizes} \item{verbose}{flag to print out progress information. Default is FALSE.} } \value{ A kernel two-sample global significance test is a list with fields: \item{Tstat}{T statistic} \item{zstat}{z statistic - normalised version of Tstat} \item{pvalue}{\eqn{p}{p}-value of the double sided test} \item{mean,var}{mean and variance of null distribution} \item{var.fhat1,var.fhat2}{sample variances of KDE values evaluated at data points} \item{n1,n2}{sample sizes} \item{H1,H2}{bandwidth matrices} \item{psi1,psi12,psi21,psi2}{kernel functional estimates} } \details{The null hypothesis is \eqn{H_0: f_1 \equiv f_2}{H_0: f_1 = f_2} where \eqn{f_1, f_2}{f_1, f_2} are the respective density functions. The measure of discrepancy is the integrated squared error (ISE) \eqn{T = \int [f_1(\bold{x}) - f_2(\bold{x})]^2 \, d \bold{x}}{int [ f_1(x) - f_2(x)]^2 dx}. If we rewrite this as \eqn{T = \psi_{0,1} - \psi_{0,12} - \psi_{0,21} + \psi_{0,2}}{T = psi_0,1 - psi_0,12 - psi_0,21 + psi_0,2} where \eqn{\psi_{0,uv} = \int f_u (\bold{x}) f_v (\bold{x}) \, d \bold{x}}{psi_0,uv = int f_u(x) f_v(x) dx}, then we can use kernel functional estimators. This test statistic has a null distribution which is asymptotically normal, so no bootstrap resampling is required to compute an approximate \eqn{p}{p}-value. If \code{H1,H2} are missing then the plug-in selector \code{\link{Hpi.kfe}} is automatically called by \code{kde.test} to estimate the functionals with \code{kfe(, deriv.order=0)}. Likewise for missing \code{h1,h2}. For \pkg{ks} \eqn{\geq}{>=} 1.8.8, \code{kde.test(,binned=TRUE)} invokes binned estimation for the computation of the bandwidth selectors, and not the test statistic and \eqn{p}{p}-value. } \references{ Duong, T., Goud, B. & Schauer, K. (2012) Closed-form density-based framework for automatic detection of cellular morphology changes. \emph{PNAS}, \bold{109}, 8382-8387. } \seealso{\code{\link{kde.local.test}}} \examples{ set.seed(8192) samp <- 1000 x <- rnorm.mixt(n=samp, mus=0, sigmas=1, props=1) y <- rnorm.mixt(n=samp, mus=0, sigmas=1, props=1) kde.test(x1=x, x2=y)$pvalue ## accept H0: f1=f2 data(crabs, package="MASS") x1 <- crabs[crabs$sp=="B", c(4,6)] x2 <- crabs[crabs$sp=="O", c(4,6)] kde.test(x1=x1, x2=x2)$pvalue ## reject H0: f1=f2 } \keyword{test} ks/man/hsct.Rd0000644000176200001440000000207014336531304012673 0ustar liggesusers\name{hsct} \docType{data} \alias{hsct} \title{Haematopoietic stem cell transplant} \description{ This data set contains the haematopoietic stem cell transplant (HSCT) measurements obtained by a flow cytometer from mouse subjects. A flow cytometer measures the spectra of fluorescent signals from biological cell samples to study their properties. } \usage{data(hsct)} \format{A matrix with 39128 rows and 6 columns. The first column is the FITC-CD45.1 fluorescence (0-1023), the second is the PE-Ly65/Mac1 fluorescence (0-1023), the third is the PI-LiveDead fluorescence (0-1023), the fourth is the APC-CD45.2 fluorescence (0-1023), the fifth is the class label of the cell type (1, 2, 3, 4, 5), the sixth the mouse subject number (5, 6, 9, 12). } \source{ Aghaeepour, N., Finak, G., The FlowCAP Consortium, The DREAM Consortium, Hoos, H., Mosmann, T. R., Brinkman, R., Gottardo, R. & Scheuermann, R. H. (2013) Critical assessment of automated flow cytometry data analysis techniques, \emph{Nature Methods} \bold{10}, 228-238. } \keyword{datasets} ks/man/grevillea.Rd0000644000176200001440000000175114673274403013721 0ustar liggesusers\name{grevillea} \docType{data} \alias{grevillea} \title{Geographical locations of grevillea plants} \description{ This data set contains the geographical locations of the specimens of \emph{Grevillea uncinulata}, more commonly known as the Hook leaf grevillea, which is an endemic floral species to south Western Australia. This region is one of the 25 `biodiversity hotspots' which are `areas featuring exceptional concentrations of endemic species and experiencing exceptional loss of habitat'. } \usage{data(grevillea)} \format{A matrix with 222 rows and 2 columns. Each row corresponds to an observed plant. The first column is the longitude (decimal degrees), the second is the latitude (decimal degrees). } \source{ CSIRO (2016) Atlas of Living Australia: \emph{Grevillea uncinulata Diels}, \url{https://bie.ala.org.au/species/https://id.biodiversity.org.au/node/apni/2895039}. Commonwealth Scientific and Industrial Research Organisation. Accessed 2016-03-11. } \keyword{datasets} ks/man/kcde.Rd0000644000176200001440000001021214340025164012632 0ustar liggesusers\name{kcde} \alias{kcde} \alias{Hpi.kcde} \alias{Hpi.diag.kcde} \alias{hpi.kcde} \alias{predict.kcde} \title{Kernel cumulative distribution/survival function estimate} \description{ Kernel cumulative distribution/survival function estimate for 1- to 3-dimensional data. } \usage{ kcde(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, verbose=FALSE, tail.flag="lower.tail") Hpi.kcde(x, nstage=2, pilot, Hstart, binned, bgridsize, amise=FALSE, verbose=FALSE, optim.fun="optim", pre=TRUE) Hpi.diag.kcde(x, nstage=2, pilot, Hstart, binned, bgridsize, amise=FALSE, verbose=FALSE, optim.fun="optim", pre=TRUE) hpi.kcde(x, nstage=2, binned, amise=FALSE) \method{predict}{kcde}(object, ..., x) } \arguments{ \item{x}{matrix of data values} \item{H,h}{bandwidth matrix/scalar bandwidth. If these are missing, then \code{Hpi.kcde} or \code{hpi.kcde} is called by default.} \item{gridsize}{vector of number of grid points} \item{gridtype}{not yet implemented} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{vector or matrix of points at which estimate is evaluated} \item{binned}{flag for binned estimation. Default is FALSE.} \item{bgridsize}{vector of binning grid sizes} \item{positive}{flag if 1-d data are positive. Default is FALSE.} \item{adj.positive}{adjustment applied to positive 1-d data} \item{w}{not yet implemented} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{tail.flag}{"lower.tail" = cumulative distribution, "upper.tail" = survival function} \item{nstage}{number of stages in the plug-in bandwidth selector (1 or 2)} \item{pilot}{"dscalar" = single pilot bandwidth (default for \code{Hpi.diag.kcde} \cr "dunconstr" = single unconstrained pilot bandwidth (default for \code{Hpi.kcde}} \item{Hstart}{initial bandwidth matrix, used in numerical optimisation} \item{amise}{flag to return the minimal scaled PI value} \item{optim.fun}{optimiser function: one of \code{nlm} or \code{optim}} \item{pre}{flag for pre-scaling data. Default is TRUE.} \item{object}{object of class \code{kcde}} \item{...}{other parameters} } \value{ A kernel cumulative distribution estimate is an object of class \code{kcde} which is a list with fields: \item{x}{data points - same as input} \item{eval.points}{vector or list of points at which the estimate is evaluated} \item{estimate}{cumulative distribution/survival function estimate at \code{eval.points}} \item{h}{scalar bandwidth (1-d only)} \item{H}{bandwidth matrix} \item{gridtype}{"linear"} \item{gridded}{flag for estimation on a grid} \item{binned}{flag for binned estimation} \item{names}{variable names} \item{w}{vector of weights} \item{tail}{"lower.tail"=cumulative distribution, "upper.tail"=survival function} } \details{ If \code{tail.flag="lower.tail"} then the cumulative distribution function \eqn{\mathrm{Pr}(\bold{X}\leq\bold{x})}{Pr(X<=x)} is estimated, otherwise if \code{tail.flag="upper.tail"}, it is the survival function \eqn{\mathrm{Pr}(\bold{X}>\bold{x})}{P(X>x)}. For \eqn{d>1}{d>1}, \eqn{\mathrm{Pr}(\bold{X}\leq\bold{x}) \neq 1 - \mathrm{Pr}(\bold{X}>\bold{x})}{Pr(X<=x) != 1-Pr(X>x)}. If the bandwidth \code{H} is missing in \code{kcde}, then the default bandwidth is the plug-in selector \code{Hpi.kcde}. Likewise for missing \code{h}. No pre-scaling/pre-sphering is used since the \code{Hpi.kcde} is not invariant to translation/dilation. The effective support, binning, grid size, grid range, positive, optimisation function parameters are the same as \code{\link{kde}}. } \references{ Duong, T. (2016) Non-parametric smoothed estimation of multivariate cumulative distribution and survival functions, and receiver operating characteristic curves. \emph{Journal of the Korean Statistical Society}, \bold{45}, 33-50. } \seealso{\code{\link{kde}}, \code{\link{plot.kcde}}} \examples{ data(iris) Fhat <- kcde(iris[,1:2]) predict(Fhat, x=as.matrix(iris[,1:2])) ## See other examples in ? plot.kcde } \keyword{smooth} ks/man/contour.Rd0000644000176200001440000000600414336531235013427 0ustar liggesusers\name{contour} \alias{contourLevels} \alias{contourLevels.kde} \alias{contourLevels.kda} \alias{contourLevels.kdde} \alias{contourSizes} \alias{contourProbs} \title{Contour functions} \description{ Contour levels and sizes. } \usage{ contourLevels(x, ...) \method{contourLevels}{kde} (x, prob, cont, nlevels=5, approx=TRUE, ...) \method{contourLevels}{kda} (x, prob, cont, nlevels=5, approx=TRUE, ...) \method{contourLevels}{kdde}(x, prob, cont, nlevels=5, approx=TRUE, which.deriv.ind=1, ...) contourSizes(x, abs.cont, cont=c(25,50,75), approx=TRUE) contourProbs(x, abs.cont, cont=c(25,50,75), approx=TRUE) } \arguments{ \item{x}{object of class \code{kde}, \code{kdde} or \code{kda}} \item{prob}{vector of probabilities corresponding to highest density regions} \item{cont}{vector of percentages which correspond to the complement of \code{prob}} \item{abs.cont}{vector of absolute contour levels} \item{nlevels}{number of pretty contour levels} \item{approx}{flag to compute approximate contour levels. Default is TRUE.} \item{which.deriv.ind}{partial derivative index. Default is 1.} \item{...}{other parameters} } \value{ --For \code{contourLevels}, for \code{kde} objects, returns vector of heights. For \code{kda} objects, returns a list of vectors, one for each training group. For \code{kdde} objects, returns a matrix of vectors, one row for each partial derivative. --For \code{contourSizes}, returns an approximation of the Lebesgue measure of level set, i.e. length (d=1), area (d=2), volume (d=3), hyper-volume (d>4). --For \code{contourProbs}, returns an approximation of the probability measure of level set. } \details{ --For \code{contourLevels}, the most straightforward is to specify \code{prob}. The heights of the corresponding highest density region with probability \code{prob} are computed. The \command{cont} parameter here is consistent with \command{cont} parameter from \command{plot.kde}, \command{plot.kdde}, and \command{plot.kda} i.e. \code{cont=(1-prob)*100\%}. If both \code{prob} and \code{cont} are missing then a pretty set of \code{nlevels} contours are computed. --For \code{contourSizes}, the length, area, volume etc. and for \code{contourProbs}, the probability, are approximated by Riemann sums. These are rough approximations and depend highly on the estimation grid, and so should be interpreted carefully. If \code{approx=FALSE}, then the exact KDE is computed. Otherwise it is interpolated from an existing KDE grid: this can dramatically reduce computation time for large data sets. } \seealso{\code{\link{contour}}, \code{\link{contourLines}}} \examples{ set.seed(8192) x <- rmvnorm.mixt(n=1000, mus=c(0,0), Sigmas=diag(2), props=1) fhat <- kde(x=x, binned=TRUE) contourLevels(fhat, cont=c(75, 50, 25)) contourProbs(fhat, abs.cont=contourLevels(fhat, cont=50)) ## compare approx prob with target prob=0.5 contourSizes(fhat, cont=25, approx=TRUE) ## compare to approx circle of radius=0.75 with area=1.77 } \keyword{hplot} ks/man/binning.Rd0000644000176200001440000000265314336531214013365 0ustar liggesusers\name{binning} \alias{binning} \title{Linear binning for multivariate data} \description{ Linear binning for 1- to 4-dimensional data. } \usage{ binning(x, H, h, bgridsize, xmin, xmax, supp=3.7, w, gridtype="linear") } \arguments{ \item{x}{matrix of data values} \item{H,h}{bandwidth matrix, scalar bandwidth} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal is [-supp,supp]} \item{bgridsize}{vector of binning grid sizes} \item{w}{vector of weights. Default is a vector of all ones.} \item{gridtype}{not yet implemented} } \value{ Returns a list with 2 fields \item{counts}{linear binning counts} \item{eval.points}{vector (d=1) or list (d>=2) of grid points in each dimension } } \details{ For \pkg{ks} \eqn{\geq}{>=} 1.10.0, binning is available for unconstrained (non-diagonal) bandwidth matrices. Code is used courtesy of A. & J. Gramacki, and M.P. Wand. Default \code{bgridsize} are d=1: 401; d=2: rep(151, 2); d=3: rep(51, 3); d=4: rep(21, 4). } \references{ Gramacki, A. & Gramacki, J. (2016) FFT-based fast computation of multivariate kernel estimators with unconstrained bandwidth matrices. \emph{Journal of Computational & Graphical Statistics}, \bold{26}, 459-462. Wand, M.P. & Jones, M.C. (1995) \emph{Kernel Smoothing}. Chapman & Hall. London. } \examples{ data(unicef) ubinned <- binning(x=unicef) } \keyword{algebra} ks/man/kdde.Rd0000644000176200001440000000727114336531336012656 0ustar liggesusers\name{kdde} \alias{kdde} \alias{predict.kdde} \alias{kcurv} \title{Kernel density derivative estimate} \description{ Kernel density derivative estimate for 1- to 6-dimensional data. } \usage{ kdde(x, H, h, deriv.order=0, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, deriv.vec=TRUE, verbose=FALSE) kcurv(fhat, compute.cont=TRUE) \method{predict}{kdde}(object, ..., x) } \arguments{ \item{x}{matrix of data values} \item{H,h}{bandwidth matrix/scalar bandwidth. If these are missing, \code{Hpi} or \code{hpi} is called by default.} \item{deriv.order}{derivative order (scalar)} \item{gridsize}{vector of number of grid points} \item{gridtype}{not yet implemented} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{vector or matrix of points at which estimate is evaluated} \item{binned}{flag for binned estimation} \item{bgridsize}{vector of binning grid sizes} \item{positive}{flag if data are positive (1-d, 2-d). Default is FALSE.} \item{adj.positive}{adjustment applied to positive 1-d data} \item{w}{vector of weights. Default is a vector of all ones.} \item{deriv.vec}{flag to compute all derivatives in vectorised derivative. Default is TRUE. If FALSE then only the unique derivatives are computed.} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{compute.cont}{flag for computing 1\% to 99\% probability contour levels. Default is TRUE.} \item{fhat}{object of class \code{kdde} with \code{deriv.order=2}} \item{object}{object of class \code{kdde}} \item{...}{other parameters} } \value{ A kernel density derivative estimate is an object of class \code{kdde} which is a list with fields: \item{x}{data points - same as input} \item{eval.points}{vector or list of points at which the estimate is evaluated} \item{estimate}{density derivative estimate at \code{eval.points}} \item{h}{scalar bandwidth (1-d only)} \item{H}{bandwidth matrix} \item{gridtype}{"linear"} \item{gridded}{flag for estimation on a grid} \item{binned}{flag for binned estimation} \item{names}{variable names} \item{w}{vector of weights} \item{deriv.order}{derivative order (scalar)} \item{deriv.ind}{martix where each row is a vector of partial derivative indices} } \details{ For each partial derivative, for grid estimation, the estimate is a list whose elements correspond to the partial derivative indices in the rows of \code{deriv.ind}. For points estimation, the estimate is a matrix whose columns correspond to the rows of \code{deriv.ind}. If the bandwidth \code{H} is missing from \code{kdde}, then the default bandwidth is the plug-in selector \code{Hpi}. Likewise for missing \code{h}. The effective support, binning, grid size, grid range, positive parameters are the same as \code{\link{kde}}. The summary curvature is computed by \code{kcurv}, i.e. \deqn{\hat{s}(\bold{x})= - \bold{1}\{\mathsf{D}^2 \hat{f}(\bold{x}) < 0\} \mathrm{abs}(|\mathsf{D}^2 \hat{f}(\bold{x})|)}{hat(s)(x) = -1(D^2 hat(f)(x) <0)*abs(det(D^2 hat(f)(x)))} where \eqn{\mathsf{D}^2 \hat{f}(\bold{x})}{D^2 hat(f)(x)} is the kernel Hessian matrix estimate. So \eqn{\hat{s}}{hat{s}} calculates the absolute value of the determinant of the Hessian matrix and whose sign is the opposite of the negative definiteness indicator. } \seealso{\code{\link{kde}}} \examples{ set.seed(8192) x <- rmvnorm.mixt(1000, mus=c(0,0), Sigmas=invvech(c(1,0.8,1))) fhat <- kdde(x=x, deriv.order=1) ## gradient [df/dx, df/dy] predict(fhat, x=x[1:5,]) ## See other examples in ? plot.kdde } \keyword{smooth} ks/man/histde.Rd0000644000176200001440000000333714336531250013221 0ustar liggesusers\name{histde} \alias{histde} \alias{predict.histde} \title{Histogram density estimate} \description{ Histogram density estimate for 1- and 2-dimensional data. } \usage{ histde(x, binw, xmin, xmax, adj=0) \method{predict}{histde}(object, ..., x) } \arguments{ \item{x}{matrix of data values} \item{binw}{(vector) of binwidths} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{adj}{displacement of default anchor point, in percentage of 1 bin} \item{object}{object of class \code{histde}} \item{...}{other parameters} } \value{ A histogram density estimate is an object of class \code{histde} which is a list with fields: \item{x}{data points - same as input} \item{eval.points}{vector or list of points at which the estimate is evaluated} \item{estimate}{density estimate at \code{eval.points}} \item{binw}{(vector of) bandwidths} \item{nbin}{(vector of) number of bins} \item{names}{variable names} } \details{ If \code{binw} is missing, the default binwidth is \eqn{\hat{b}_i = 2 \cdot 3^{1/(d+2)} \pi^{d/(2d+4)} S_i n^{-1/(d+2)}}{b_i = 2*3^(1/(d+2))*pi^(d/(2d+4))*S_i*n^(-1/(d+2))}, the normal scale selector. If \code{xmin} is missing then it defaults to the data minimum. If \code{xmax} is missing then it defaults to the data maximum. } \seealso{\code{\link{plot.histde}}} \examples{ ## positive data example set.seed(8192) x <- 2^rnorm(100) fhat <- histde(x=x) plot(fhat, border=6) points(c(0.5, 1), predict(fhat, x=c(0.5, 1))) ## large data example on a non-default grid set.seed(8192) x <- rmvnorm.mixt(10000, mus=c(0,0), Sigmas=invvech(c(1,0.8,1))) fhat <- histde(x=x, xmin=c(-5,-5), xmax=c(5,5)) plot(fhat) ## See other examples in ? plot.histde } \keyword{smooth} ks/man/kcopula.Rd0000644000176200001440000000703114340025575013375 0ustar liggesusers\name{kcopula} \alias{kcopula} \alias{kcopula.de} \title{Kernel copula (density) estimate} \description{ Kernel copula and copula density estimator for 2-dimensional data. } \usage{ kcopula(x, H, hs, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, marginal="kernel", verbose=FALSE) kcopula.de(x, H, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, marginal="kernel", boundary.supp, boundary.kernel="beta", verbose=FALSE) } \arguments{ \item{x}{matrix of data values} \item{H,hs}{bandwidth matrix. If these are missing, \code{Hpi.kcde}/\code{Hpi} or \code{hpi.kcde}/\code{hpi} is called by default.} \item{gridsize}{vector of number of grid points} \item{gridtype}{not yet implemented} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{matrix of points at which estimate is evaluated} \item{binned}{flag for binned estimation} \item{bgridsize}{vector of binning grid sizes} \item{w}{vector of weights. Default is a vector of all ones.} \item{marginal}{"kernel" = kernel cdf or "empirical" = empirical cdf to calculate pseudo-uniform values. Default is "kernel".} \item{compute.cont}{flag for computing 1\% to 99\% probability contour levels. Default is TRUE.} \item{approx.cont}{flag for computing approximate probability contour levels. Default is TRUE.} \item{boundary.supp}{effective support for boundary region} \item{boundary.kernel}{"beta" = beta boundary kernel, "linear" = linear boundary kernel} \item{verbose}{flag to print out progress information. Default is FALSE.} } \value{ A kernel copula estimate, output from \code{kcopula}, is an object of class \code{kcopula}. A kernel copula density estimate, output from \code{kcopula.de}, is an object of class \code{kde}. These two classes of objects have the same fields as \code{kcde} and \code{kde} objects respectively, except for \item{x}{pseudo-uniform data points} \item{x.orig}{data points - same as input} \item{marginal}{marginal function used to compute pseudo-uniform data} \item{boundary}{flag for data points in the boundary region (\code{kcopula.de} only)} } \details{ For kernel copula estimates, a transformation approach is used to account for the boundary effects. If \code{H} is missing, the default is \code{Hpi.kcde}; if \code{hs} are missing, the default is \code{hpi.kcde}. For kernel copula density estimates, for those points which are in the interior region, the usual kernel density estimator (\code{\link{kde}}) is used. For those points in the boundary region, a product beta kernel based on the boundary corrected univariate beta kernel of Chen (1999) is used (\code{\link{kde.boundary}}). If \code{H} is missing, the default is \code{Hpi.kcde}; if \code{hs} are missing, the default is \code{hpi}. The effective support, binning, grid size, grid range parameters are the same as for \code{\link{kde}}. } \references{ Duong, T. (2014) Optimal data-based smoothing for non-parametric estimation of copula functions and their densities. Submitted. Chen, S.X. (1999). Beta kernel estimator for density functions. \emph{Computational Statistics & Data Analysis}, \bold{31}, 131--145. } \seealso{\code{\link{kcde}}, \code{\link{kde}}} \examples{ data(fgl, package="MASS") x <- fgl[,c("RI", "Na")] Chat <- kcopula(x=x) plot(Chat, display="persp", border=1) plot(Chat, display="filled.contour", lwd=1) } \keyword{smooth} ks/man/Hpi.Rd0000644000176200001440000000677514336531301012467 0ustar liggesusers\name{Hpi} \alias{Hpi} \alias{Hpi.diag} \alias{hpi} \title{Plug-in bandwidth selector} \description{ Plug-in bandwidth for for 1- to 6-dimensional data.} \usage{ Hpi(x, nstage=2, pilot, pre="sphere", Hstart, binned, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") Hpi.diag(x, nstage=2, pilot, pre="scale", Hstart, binned, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") hpi(x, nstage=2, binned=TRUE, bgridsize, deriv.order=0) } \arguments{ \item{x}{vector or matrix of data values} \item{nstage}{number of stages in the plug-in bandwidth selector (1 or 2)} \item{pilot}{"amse" = AMSE pilot bandwidths \cr "samse" = single SAMSE pilot bandwidth \cr "unconstr" = single unconstrained pilot bandwidth \cr "dscalar" = single pilot bandwidth for deriv.order >= 0 \cr "dunconstr" = single unconstrained pilot bandwidth for deriv.order >= 0} \item{pre}{"scale" = \code{\link{pre.scale}}, "sphere" = \code{\link{pre.sphere}}} \item{Hstart}{initial bandwidth matrix, used in numerical optimisation} \item{binned}{flag for binned kernel estimation} \item{bgridsize}{vector of binning grid sizes} \item{amise}{flag to return the minimal scaled PI value} \item{deriv.order}{derivative order} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{optim.fun}{optimiser function: one of \code{nlm} or \code{optim}} } \value{ Plug-in bandwidth. If \code{amise=TRUE} then the minimal scaled PI value is returned too.} \details{\code{hpi(,deriv.order=0)} is the univariate plug-in selector of Wand & Jones (1994), i.e. it is exactly the same as \pkg{KernSmooth}'s \code{dpik}. For deriv.order>0, the formula is taken from Wand & Jones (1995). \code{Hpi} is a multivariate generalisation of this. Use \code{Hpi} for unconstrained bandwidth matrices and \code{Hpi.diag} for diagonal bandwidth matrices. The default pilot is \code{"samse"} for d=2,r=0, and \code{"dscalar"} otherwise. For AMSE pilot bandwidths, see Wand & Jones (1994). For SAMSE pilot bandwidths, see Duong & Hazelton (2003). The latter is a modification of the former, in order to remove any possible problems with non-positive definiteness. Unconstrained and higher order derivative pilot bandwidths are from Chacon & Duong (2010). For d=1, 2, 3, 4 and \code{binned=TRUE}, estimates are computed over a binning grid defined by \code{bgridsize}. Otherwise it's computed exactly. If \code{Hstart} is not given then it defaults to \code{Hns(x)}. For \pkg{ks} \eqn{\geq}{>=} 1.11.1, the default optimisation function is \code{optim.fun="optim"}. To reinstate the previous functionality, use \code{optim.fun="nlm"}. } \references{ Chacon, J.E. & Duong, T. (2010) Multivariate plug-in bandwidth selection with unconstrained pilot matrices. \emph{Test}, \bold{19}, 375-398. Duong, T. & Hazelton, M.L. (2003) Plug-in bandwidth matrices for bivariate kernel density estimation. \emph{Journal of Nonparametric Statistics}, \bold{15}, 17-30. Sheather, S.J. & Jones, M.C. (1991) A reliable data-based bandwidth selection method for kernel density estimation. \emph{Journal of the Royal Statistical Society Series B}, \bold{53}, 683-690. Wand, M.P. & Jones, M.C. (1994) Multivariate plug-in bandwidth selection. \emph{Computational Statistics}, \bold{9}, 97-116. } \seealso{\code{\link{Hbcv}}, \code{\link{Hlscv}}, \code{\link{Hscv}}} \examples{ data(unicef) Hpi(unicef, pilot="dscalar") hpi(unicef[,1]) } \keyword{smooth} ks/man/unicef.Rd0000644000176200001440000000127014336531550013207 0ustar liggesusers\name{unicef} \docType{data} \alias{unicef} \title{Unicef child mortality - life expectancy data} \description{ This data set contains the number of deaths of children under 5 years of age per 1000 live births and the average life expectancy (in years) at birth for 73 countries with GNI (Gross National Income) less than 1000 US dollars per annum per capita. } \usage{data(unicef)} \format{A matrix with 2 columns and 73 rows. Each row corresponds to a country. The first column is the under 5 mortality rate and the second is the average life expectancy.} \source{ Unicef (2003). \emph{State of the World's Children Report 2003}, Oxford University Press, for Unicef. } \keyword{datasets} ks/man/kde.Rd0000644000176200001440000001374614615043612012511 0ustar liggesusers\name{kde} \alias{kde} \alias{predict.kde} \title{Kernel density estimate} \description{ Kernel density estimate for 1- to 6-dimensional data. } \usage{ kde(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, compute.cont=TRUE, approx.cont=TRUE, unit.interval=FALSE, density=FALSE, verbose=FALSE) \method{predict}{kde}(object, ..., x, zero.flag=TRUE) } \arguments{ \item{x}{matrix of data values} \item{H,h}{bandwidth matrix/scalar bandwidth. If these are missing, \code{Hpi} or \code{hpi} is called by default.} \item{gridsize}{vector of number of grid points} \item{gridtype}{not yet implemented} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{vector or matrix of points at which estimate is evaluated} \item{binned}{flag for binned estimation.} \item{bgridsize}{vector of binning grid sizes} \item{positive}{flag if data are positive (1-d, 2-d). Default is FALSE.} \item{adj.positive}{adjustment applied to positive 1-d data} \item{w}{vector of weights. Default is a vector of all ones.} \item{compute.cont}{flag for computing 1\% to 99\% probability contour levels. Default is TRUE.} \item{approx.cont}{flag for computing approximate probability contour levels. Default is TRUE.} \item{unit.interval}{flag for computing log transformation KDE on 1-d data bounded on unit interval [0,1]. Default is FALSE.} \item{density}{flag if density estimate values are forced to be non-negative function. Default is FALSE.} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{object}{object of class \code{kde}} \item{zero.flag}{deprecated (retained for backwards compatibilty)} \item{...}{other parameters} } \value{ A kernel density estimate is an object of class \code{kde} which is a list with fields: \item{x}{data points - same as input} \item{eval.points}{vector or list of points at which the estimate is evaluated} \item{estimate}{density estimate at \code{eval.points}} \item{h}{scalar bandwidth (1-d only)} \item{H}{bandwidth matrix} \item{gridtype}{"linear"} \item{gridded}{flag for estimation on a grid} \item{binned}{flag for binned estimation} \item{names}{variable names} \item{w}{vector of weights} \item{cont}{vector of probability contour levels} } \details{ For d=1, if \code{h} is missing, the default bandwidth is \code{hpi}. For d>1, if \code{H} is missing, the default is \code{Hpi}. For d=1, if \code{positive=TRUE} then \code{x} is transformed to \code{log(x+adj.positive)} where the default \code{adj.positive} is the minimum of \code{x}. This is known as a log transformation density estimate. If \code{unit.interval=TRUE} then \code{x} is transformed to \code{qnorm(x)}. See \code{\link{kde.boundary}} for boundary kernel density estimates, as these tend to be more robust than transformation density estimates. For d=1, 2, 3, and if \code{eval.points} is not specified, then the density estimate is computed over a grid defined by \code{gridsize} (if \code{binned=FALSE}) or by \code{bgridsize} (if \code{binned=TRUE}). This form is suitable for visualisation in conjunction with the \code{plot} method. For d=4, 5, 6, and if \code{eval.points} is not specified, then the density estimate is computed over a grid defined by \code{gridsize}. If \code{eval.points} is specified, as a vector (d=1) or as a matrix (d=2, 3, 4), then the density estimate is computed at \code{eval.points}. This form is suitable for numerical summaries (e.g. maximum likelihood), and is not compatible with the \code{plot} method. Despite that the density estimate is returned only at \code{eval.points}, by default, a binned gridded estimate is calculated first and then the density estimate at \code{eval.points} is computed using the \code{predict} method. If this default intermediate binned grid estimate is not required, then set \code{binned=FALSE} to compute directly the exact density estimate at \code{eval.points}. Binned kernel estimation is an approximation to the exact kernel estimation and is available for d=1, 2, 3, 4. This makes kernel estimators feasible for large samples. The default value of the binning flag \code{binned} is n>1 (d=1), n>500 (d=2), n>1000 (d>=3). Some times binned estimation leads to negative density values: if non-negative values are required, then set \code{density=TRUE}. The default \code{bgridsize,gridsize} are d=1: 401; d=2: rep(151, 2); d=3: rep(51, 3); d=4: rep(21, 4). The effective support for a normal kernel is where all values outside \code{[-supp,supp]^d} are zero. The default \code{xmin} is \code{min(x)-Hmax*supp} and \code{xmax} is \code{max(x)+Hmax*supp} where \code{Hmax} is the maximum of the diagonal elements of \code{H}. The grid produced is the outer product of \code{c(xmin[1], xmax[1])}, ..., \code{c(xmin[d], xmax[d])}. For \pkg{ks} \eqn{\geq}{>=} 1.14.0, when \code{binned=TRUE} and \code{xmin,xmax} are not missing, the data values \code{x} are clipped to the estimation grid delimited by \code{xmin,xmax} to prevent potential memory leaks. } \seealso{\code{\link{plot.kde}}, \code{\link{kde.boundary}}} \examples{ ## unit interval data set.seed(8192) fhat <- kde(runif(10000,0,1), unit.interval=TRUE) plot(fhat, ylim=c(0,1.2)) ## positive data data(worldbank) wb <- as.matrix(na.omit(worldbank[,2:3])) wb[,2] <- wb[,2]/1000 fhat <- kde(x=wb) fhat.trans <- kde(x=wb, adj.positive=c(0,0), positive=TRUE) plot(fhat, col=1, xlim=c(0,20), ylim=c(0,80)) plot(fhat.trans, add=TRUE, col=2) rect(0,0,100,100, lty=2) ## large data on non-default grid ## 151 x 151 grid = [-5,-4.933,..,5] x [-5,-4.933,..,5] set.seed(8192) x <- rmvnorm.mixt(10000, mus=c(0,0), Sigmas=invvech(c(1,0.8,1))) fhat <- kde(x=x, compute.cont=TRUE, xmin=c(-5,-5), xmax=c(5,5), bgridsize=c(151,151)) plot(fhat) ## See other examples in ? plot.kde } \keyword{smooth} ks/man/air.Rd0000644000176200001440000000223614673274534012526 0ustar liggesusers\name{air} \docType{data} \alias{air} \title{Air quality measurements in an underground train station} \description{ This data set contains the hourly mean air quality measurements from 01 January 2013 to 31 December 2016 in the Chatelet underground train station in the Paris metro. } \usage{data(air)} \format{A matrix with 35039 rows and 8 columns. Each row corresponds to an hourly measurement. The first column is the date (yyyy-mm-dd), the second is the time (hh:mm), the third is the nitric oxide NO concentration (g/m3), the fourth is the nitrogen dioxide NO\eqn{_2}{2} concentration (g/m3), the fifth is the concentration of particulate matter less than 10 microns PM10 (ppm), the sixth is the carbon dioxide concentration CO\eqn{_2}{2} (g/m3), the seventh is the temperature (degrees Celsius), the eighth is the relative humidity (percentage). } \source{ RATP (2016) Qualite de l'air mesuree dans la station Chatelet, \url{https://data.iledefrance.fr/explore/dataset/qualite-de-l-air-mesuree-dans-la-station-chatelet-rer-a}. Regie autonome des transports parisiens - Departement Developpement, Innovation et Territoires. Accessed 2017-09-27. } \keyword{datasets} ks/man/Hbcv.Rd0000644000176200001440000000342614336531244012625 0ustar liggesusers\name{Hbcv} \alias{Hbcv} \alias{Hbcv.diag} \title{Biased cross-validation (BCV) bandwidth matrix selector for bivariate data} \description{ BCV bandwidth matrix for bivariate data.} \usage{ Hbcv(x, whichbcv=1, Hstart, binned=FALSE, amise=FALSE, verbose=FALSE) Hbcv.diag(x, whichbcv=1, Hstart, binned=FALSE, amise=FALSE, verbose=FALSE) } \arguments{ \item{x}{matrix of data values} \item{whichbcv}{1 = BCV1, 2 = BCV2. See details below.} \item{Hstart}{initial bandwidth matrix, used in numerical optimisation} \item{binned}{flag for binned kernel estimation. Default is FALSE.} \item{amise}{flag to return the minimal BCV value. Default is FALSE.} \item{verbose}{flag to print out progress information. Default is FALSE.} } \value{ BCV bandwidth matrix. If \code{amise=TRUE} then the minimal BCV value is returned too. } \references{Sain, S.R, Baggerly, K.A. & Scott, D.W. (1994) Cross-validation of multivariate densities. \emph{Journal of the American Statistical Association}, \bold{82}, 1131-1146. } \details{ Use \code{Hbcv} for unconstrained bandwidth matrices and \code{Hbcv.diag} for diagonal bandwidth matrices. These selectors are only available for bivariate data. Two types of BCV criteria are considered here. They are known as BCV1 and BCV2, from Sain, Baggerly & Scott (1994) and only differ slightly. These BCV surfaces can have multiple minima and so it can be quite difficult to locate the most appropriate minimum. Some times, there can be no local minimum at all so there may be no finite BCV selector. For details about the advanced options for \code{binned}, \code{Hstart}, see \code{\link{Hpi}}. } \seealso{\code{\link{Hlscv}}, \code{\link{Hpi}}, \code{\link{Hscv}}} \examples{ data(unicef) Hbcv(unicef) Hbcv.diag(unicef) } \keyword{smooth} ks/man/kdr.Rd0000644000176200001440000001242614336531372012525 0ustar liggesusers\name{kdr} \alias{kdr} \alias{kdr.segment} \alias{plot.kdr} \title{Kernel density ridge estimation} \description{ Kernel density ridge estimation for 2- to 3-dimensional data. } \usage{ kdr(x, y, H, p=1, max.iter=400, tol.iter, segment=TRUE, k, kmax, min.seg.size, keep.path=FALSE, gridsize, xmin, xmax, binned, bgridsize, w, fhat, density.cutoff, pre=TRUE, verbose=FALSE) kdr.segment(x, k, kmax, min.seg.size, verbose=FALSE) \method{plot}{kdr}(x, ...) } \arguments{ \item{x}{matrix of data values or an object of class \code{kdr}} \item{y}{matrix of initial values} \item{p}{dimension of density ridge} \item{H}{bandwidth matrix/scalar bandwidth. If missing, \code{Hpi(x,deriv,order=2)} is called by default.} \item{max.iter}{maximum number of iterations. Default is 400.} \item{tol.iter}{distance under which two successive iterations are considered convergent. Default is 0.001*min marginal IQR of \code{x}.} \item{segment}{flag to compute segments of density ridge. Default is TRUE.} \item{k}{number of segments to partition density ridge} \item{kmax}{maximum number of segments to partition density ridge. Default is 30.} %\item{tol.seg}{distance under which two segments are considered % to form one segment. Default is 0.01*max marginal IQR of \code{x}.} \item{min.seg.size}{minimum length of a segment of a density ridge. Default is \code{round(0.001*nrow(y),0)}.} \item{keep.path}{flag to store the density gradient ascent paths. Default is FALSE.} \item{gridsize}{vector of number of grid points} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{binned}{flag for binned estimation.} \item{bgridsize}{vector of binning grid sizes} \item{w}{vector of weights. Default is a vector of all ones.} \item{fhat}{kde of \code{x}. If missing \code{kde(x=x,w=w)} is executed.} \item{density.cutoff}{density threshold under which the \code{y} are excluded from the density ridge estimation. Default is \code{contourLevels(fhat, cont=99)}.} \item{pre}{flag for pre-scaling data. Default is TRUE.} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{...}{other graphics parameters} } \value{ A kernel density ridge set is an object of class \code{kdr} which is a list with fields: \item{x,y}{data points - same as input} \item{end.points}{matrix of final iterates starting from \code{y}} \item{H}{bandwidth matrix} \item{names}{variable names} \item{tol.iter,tol.clust,min.seg.size}{tuning parameter values - same as input} \item{binned}{flag for binned estimation} \item{names}{variable names} \item{w}{vector of weights} \item{path}{list of density gradient ascent paths where \code{path[[i]]} is the path of \code{y[i,]} (only if \code{keep.path=TRUE})} } \details{ Kernel density ridge estimation is based on reduced dimension kernel mean shift. See Ozertem & Erdogmus (2011). If \code{y} is missing, then it defaults to the grid of size \code{gridsize} spanning from \code{xmin} to \code{xmax}. If the bandwidth \code{H} is missing, then the default bandwidth is the plug-in selector for the density gradient \code{Hpi(x,deriv.order=2)}. Any bandwidth that is suitable for the density Hessian is also suitable for the kernel density ridge. \code{kdr(, segment=TRUE)} or \code{kdr.segment()} carries out the segmentation of the density ridge points in \code{end.points}. If \code{k} is set, then \code{k} segments are created. If \code{k} is not set, then the optimal number of segments is chosen from 1:\code{kmax}, with \code{kmax=30} by default. The segments are created via a hierarchical clustering with single linkage. *Experimental: following the segmentation, the points within each segment are ordered to facilitate a line plot in \code{plot(, type="l")}. The optimal ordering is not always achieved in this experimental implementation, though a scatterplot \code{plot(, type="p")} always suffices, regardless of this ordering.* } \references{ Ozertem, U. & Erdogmus, D. (2011) Locally defined principal curves and surfaces, \emph{Journal of Machine Learning Research}, \bold{12}, 1249-1286. } \examples{ \donttest{data(cardio) set.seed(8192) cardio.train.ind <- sample(1:nrow(cardio), round(nrow(cardio)/4,0)) cardio2 <- cardio[cardio.train.ind,c(8,18)] cardio.dr2 <- kdr(x=cardio2, gridsize=c(21,21)) ## gridsize=c(21,21) is for illustrative purposes only plot(cardio2, pch=16, col=3) plot(cardio.dr2, cex=0.5, pch=16, col=6, add=TRUE)} \dontrun{cardio3 <- cardio[cardio.train.ind,c(8,18,11)] cardio.dr3 <- kdr(x=cardio3) plot(cardio.dr3, pch=16, col=6, xlim=c(10,90), ylim=c(70,180), zlim=c(0,40)) plot3D::points3D(cardio3[,1], cardio3[,2], cardio3[,3], pch=16, col=3, add=TRUE) library(maps) data(quake) quake <- quake[quake$prof==1,] ## Pacific Ring of Fire quake$long[quake$long<0] <- quake$long[quake$long<0] + 360 quake <- quake[, c("long", "lat")] data(plate) ## tectonic plate boundaries plate <- plate[plate$long < -20 | plate$long > 20,] plate$long[plate$long<0 & !is.na(plate$long)] <- plate$long[plate$long<0 & !is.na(plate$long)] + 360 quake.dr <- kdr(x=quake, xmin=c(70,-70), xmax=c(310, 80)) map(wrap=c(0,360), lty=2) lines(plate[,1:2], col=4, lwd=2) plot(quake.dr, type="p", cex=0.5, pch=16, col=6, add=TRUE)} } \keyword{cluster} ks/man/ksupp.Rd0000644000176200001440000000324514547760205013111 0ustar liggesusers\name{ksupp} \alias{ksupp} \alias{plot.ksupp} \title{Kernel support estimate} \description{ Kernel support estimate for 2 and 3-dimensional data. } \usage{ ksupp(fhat, cont=95, abs.cont, convex.hull=TRUE) \method{plot}{ksupp}(x, display="plot3D", ...) } \arguments{ \item{fhat}{object of class \code{kde}} \item{cont}{percentage for contour level curve. Default is 95.} \item{abs.cont}{absolute density estimate height for contour level curve} \item{convex.hull}{flag to compute convex hull of contour level curve. Default is TRUE.} \item{x}{object of class \code{ksupp}} \item{display}{one of "plot3D", "rgl" (required for 3-d only)} \item{...}{other graphics parameters} } \value{ A kernel support estimate is an object of class \code{ksupp}, i.e. a 2- or 3-column matrix which delimits the (convex hull of the) level set of the density estimate \code{fhat}. } \details{ The kernel support estimate is the level set of the density estimate that exceeds the \code{cont} percent contour level. If this level set is a simply connected region, then this can suffice to be a conservative estimate of the density support. Otherwise, the convex hull of the level set is advised. For 2-d data, the convex hull is computed by \code{chull}; for 3-d data, it is computed by \code{geometry::convhulln}. } \seealso{\code{\link{kde}}} \examples{ data(grevillea) fhat <- kde(x=grevillea) fhat.supp <- ksupp(fhat) plot(fhat, display="filled.contour", cont=seq(10,90,by=10)) plot(fhat, cont=95, add=TRUE, col=1) plot(fhat.supp, lty=2) data(iris) fhat <- kde(x=iris[,1:3]) fhat.supp <- ksupp(fhat) plot(fhat) plot(fhat.supp, add=TRUE, col=3, alpha=0.1) } \keyword{smooth} ks/man/kdcde.Rd0000644000176200001440000000565514336531332013021 0ustar liggesusers\name{kdcde} \alias{kdcde} \alias{dckde} \title{Deconvolution kernel density derivative estimate} \description{ Deconvolution kernel density derivative estimate for 1- to 6-dimensional data. } \usage{ kdcde(x, H, h, Sigma, sigma, reg, bgridsize, gridsize, binned, verbose=FALSE, ...) dckde(...) } \arguments{ \item{x}{matrix of data values} \item{H,h}{bandwidth matrix/scalar bandwidth. If these are missing, \code{Hpi} or \code{hpi} is called by default.} \item{Sigma,sigma}{error variance matrix} \item{reg}{regularisation parameter} \item{gridsize}{vector of number of grid points} \item{binned}{flag for binned estimation} \item{bgridsize}{vector of binning grid sizes} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{...}{other parameters to \code{\link{kde}}} } \value{ A deconvolution kernel density derivative estimate is an object of class \code{kde} which is a list with fields: \item{x}{data points - same as input} \item{eval.points}{vector or list of points at which the estimate is evaluated} \item{estimate}{density estimate at \code{eval.points}} \item{h}{scalar bandwidth (1-d only)} \item{H}{bandwidth matrix} \item{gridtype}{"linear"} \item{gridded}{flag for estimation on a grid} \item{binned}{flag for binned estimation} \item{names}{variable names} \item{w}{vector of weights} \item{cont}{vector of probability contour levels} } \details{ A weighted kernel density estimate is utilised to perform the deconvolution. The weights \code{w} are the solution to a quadratic programming problem, and then input into \code{kde(,w=w)}. This weighted estimate also requires an estimate of the error variance matrix from repeated observations, and of the regularisation parameter. If the latter is missing, it is calculated internally using a 5-fold cross validation method. See Hazelton & Turlach (2009). \code{dckde} is an alias for \code{kdcde}. If the bandwidth \code{H} is missing from \code{kde}, then the default bandwidth is the plug-in selector \code{Hpi}. Likewise for missing \code{h}. The effective support, binning, grid size, grid range, positive parameters are the same as \code{\link{kde}}. } \references{ Hazelton, M. L. & Turlach, B. A. (2009), Nonparametric density deconvolution by weighted kernel density estimators, \emph{Statistics and Computing}, \bold{19}, 217-228. } \seealso{\code{\link{kde}}} \examples{ \donttest{ data(air) air <- air[, c("date", "time", "co2", "pm10")] air2 <- reshape(air, idvar="date", timevar="time", direction="wide") air <- as.matrix(na.omit(air2[,c("co2.20:00", "pm10.20:00")])) Sigma.air <- diag(c(var(air2[,"co2.19:00"] - air2["co2.21:00"], na.rm=TRUE), var(air2[,"pm10.19:00"] - air2[,"pm10.21:00"], na.rm=TRUE))) fhat.air.dec <- kdcde(x=air, Sigma=Sigma.air, reg=0.00021, verbose=TRUE) plot(fhat.air.dec, drawlabels=FALSE, display="filled.contour", lwd=1)} } \keyword{smooth} ks/man/quake.Rd0000644000176200001440000000370214336531536013052 0ustar liggesusers\name{quake} \docType{data} \alias{quake} \alias{quakesf} \alias{plate} \alias{platesf} \title{Geographical locations of earthquakes and tectonic plates} \description{ The \code{quake} data set contains the geographical locations of severe earthquakes in the years 100 and 2016 inclusive. The \code{plate} data set contains the geographical locations of the tectonic plate boundaries. } \usage{data(quake) data(plate) data(quakesf) data(platesf) } \format{--For \code{quake}, a matrix with 5871 rows and 5 columns. Each row corresponds to an earthquake. The first column is the year (negative years indicate B.C.E.), the second is the longitude (decimal degrees), the third is the latitude (decimal degrees), the fourth is the depth beneath the Earth's surface (km), the fifth is a flag for the location inside the circum-Pacific belt (aka Pacific Ring of Fire). \code{quakesf} is a WGS84 \code{sf} version with a point geometry. --For \code{plate}, a matrix with 6276 rows and 3 columns. Each row corresponds to an location of the tectonic plate boundaries. The first is the longitude, the second is the latitude, the third is the label of the tectonic plate. \code{platesf} is a WGS84 \code{sf} spatial version with a multipolygon geometry, where the individual plate line segments have been merged into a single multipolygon.} \source{ Alhenius, H., Nordpil and Bird, P. (2014). World Tectonic Plates and Boundaries. \url{https://github.com/fraxen/tectonicplates}. Accessed 2021-03-11. Bird, P. (2003) An updated digital model of plate boundaries, \emph{Geochemistry, Geophysics, Geosystems} \bold{4(3)}, 1-52. 1027. %Data set accessed 2016-03-24 from %\url{http://peterbird.name/publications/2003_PB2002/2003_PB2002.htm}. NGDC/WDS (2017) Global significant earthquake database, National Geophysical Data Center, NOAA, doi:10.7289/V5TD9V7K. National Geophysical Data Center/World Data Service. Accessed 2017-03-30. } \keyword{datasets} ks/man/kde.local.test.Rd0000644000176200001440000000750014336766532014564 0ustar liggesusers\name{kde.local.test} \alias{kde.local.test} \title{Kernel density based local two-sample comparison test} \description{ Kernel density based local two-sample comparison test for 1- to 6-dimensional data.} \usage{ kde.local.test(x1, x2, H1, H2, h1, h2, fhat1, fhat2, gridsize, binned, bgridsize, verbose=FALSE, supp=3.7, mean.adj=FALSE, signif.level=0.05, min.ESS, xmin, xmax) } \arguments{ \item{x1,x2}{vector/matrix of data values} \item{H1,H2,h1,h2}{bandwidth matrices/scalar bandwidths. If these are missing, \code{Hpi} or \code{hpi} is called by default.} \item{fhat1,fhat2}{objects of class \code{kde}} \item{binned}{flag for binned estimation} \item{gridsize}{vector of grid sizes} \item{bgridsize}{vector of binning grid sizes} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{supp}{effective support for normal kernel} \item{mean.adj}{flag to compute second order correction for mean value of critical sampling distribution. Default is FALSE. Currently implemented for d<=2 only.} \item{signif.level}{significance level. Default is 0.05.} \item{min.ESS}{minimum effective sample size. See below for details.} \item{xmin,xmax}{vector of minimum/maximum values for grid} } \value{ A kernel two-sample local significance is an object of class \code{kde.loctest} which is a list with fields: \item{fhat1,fhat2}{kernel density estimates, objects of class \code{kde}} \item{chisq}{chi squared test statistic} \item{pvalue}{matrix of local \eqn{p}{p}-values at each grid point} \item{fhat.diff}{difference of KDEs} \item{mean.fhat.diff}{mean of the test statistic} \item{var.fhat.diff}{variance of the test statistic} \item{fhat.diff.pos}{binary matrix to indicate locally significant fhat1 > fhat2} \item{fhat.diff.neg}{binary matrix to indicate locally significant fhat1 < fhat2} \item{n1,n2}{sample sizes} \item{H1,H2,h1,h2}{bandwidth matrices/scalar bandwidths} } \details{The null hypothesis is \eqn{H_0(\bold{x}): f_1(\bold{x}) = f_2(\bold{x})}{H_0(x): f_1(x) = f_2(x)} where \eqn{f_1, f_2}{f_1, f_2} are the respective density functions. The measure of discrepancy is \eqn{U(\bold{x}) = [f_1(\bold{x}) - f_2(\bold{x})]^2}{U(x) = [f_1(x) - f_2(x)]^2}. Duong (2013) shows that the test statistic obtained, by substituting the KDEs for the true densities, has a null distribution which is asymptotically chi-squared with 1 d.f. The required input is either \code{x1,x2} and \code{H1,H2}, or \code{fhat1,fhat2}, i.e. the data values and bandwidths or objects of class \code{kde}. In the former case, the \code{kde} objects are created. If the \code{H1,H2} are missing then the default are the plug-in selectors \code{Hpi}. Likewise for missing \code{h1,h2}. The \code{mean.adj} flag determines whether the second order correction to the mean value of the test statistic should be computed. \code{min.ESS} is borrowed from Godtliebsen et al. (2002) to reduce spurious significant results in the tails, though by it is usually not required for small to moderate sample sizes. } \references{ Duong, T. (2013) Local significant differences from non-parametric two-sample tests. \emph{Journal of Nonparametric Statistics}, \bold{25}, 635-645. Godtliebsen, F., Marron, J.S. & Chaudhuri, P. (2002) Significance in scale space for bivariate density estimation. \emph{Journal of Computational and Graphical Statistics}, \bold{11}, 1-22. } \seealso{\code{\link{kde.test}}, \code{\link{plot.kde.loctest}}} \examples{ data(crabs, package="MASS") x1 <- crabs[crabs$sp=="B", 4] x2 <- crabs[crabs$sp=="O", 4] loct <- kde.local.test(x1=x1, x2=x2) plot(loct, ylim=c(-0.08,0.12)) cols <- hcl.colors(palette="Dark2",2) plot(loct$fhat1, add=TRUE, col=cols[1]) plot(loct$fhat2, add=TRUE, col=cols[2]) ## see examples in ? plot.kde.loctest } \keyword{test} ks/man/kde.boundary.Rd0000644000176200001440000000557414336531342014335 0ustar liggesusers\name{kde.boundary} \alias{kde.boundary} \title{Kernel density estimate for bounded data} \description{ Kernel density estimate for bounded 1- to 3-dimensional data. } \usage{ kde.boundary(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned=FALSE, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, boundary.supp, boundary.kernel="beta", verbose=FALSE) } \arguments{ \item{x}{matrix of data values} \item{H,h}{bandwidth matrix/scalar bandwidth. If these are missing, \code{Hpi} or \code{hpi} is called by default.} \item{gridsize}{vector of number of grid points} \item{gridtype}{not yet implemented} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{vector or matrix of points at which estimate is evaluated} \item{binned}{flag for binned estimation.} \item{bgridsize}{vector of binning grid sizes} \item{w}{vector of weights. Default is a vector of all ones.} \item{compute.cont}{flag for computing 1\% to 99\% probability contour levels. Default is TRUE.} \item{approx.cont}{flag for computing approximate probability contour levels. Default is TRUE.} \item{boundary.supp}{effective support for boundary region} \item{boundary.kernel}{"beta" = beta boundary kernel, "linear" = linear boundary kernel} \item{verbose}{flag to print out progress information. Default is FALSE.} } \value{ A kernel density estimate for bounded data is an object of class \code{kde}. } \details{ There are two forms of density estimates which are suitable for bounded data, based on the modifying the kernel function. For \code{boundary.kernel="beta"}, the 2nd form of the Beta boundary kernel of Chen (1999) is employed. It is suited for rectangular data boundaries. For \code{boundary.kernel="linear"}, the linear boundary kernel of Hazelton & Marshall (2009) is employed. It is suited for arbitrarily shaped data boundaries, though it is currently only implemented for rectangular boundaries. } \references{ Chen, S. X. (1999) Beta kernel estimators for density functions. \emph{Computational Statistics and Data Analysis}, \bold{31}, 131-145. Hazelton, M. L. & Marshall, J. C. (2009) Linear boundary kernels for bivariate density estimation. \emph{Statistics and Probability Letters}, \bold{79}, 999-1003. } \seealso{\code{\link{kde}}} \examples{ data(worldbank) wb <- as.matrix(na.omit(worldbank[,c("internet", "ag.value")])) fhat <- kde(x=wb) fhat.beta <- kde.boundary(x=wb, xmin=c(0,0), xmax=c(100,100), boundary.kernel="beta") fhat.LB <- kde.boundary(x=wb, xmin=c(0,0), xmax=c(100,100), boundary.kernel="linear") plot(fhat, col=1, xlim=c(0,100), ylim=c(0,100)) plot(fhat.beta, add=TRUE, col=2) rect(0,0,100,100, lty=2) plot(fhat, col=1, xlim=c(0,100), ylim=c(0,100)) plot(fhat.LB, add=TRUE, col=3) rect(0,0,100,100, lty=2) } \keyword{smooth} ks/man/Hnm.Rd0000644000176200001440000000277314336531271012471 0ustar liggesusers\name{Hnm} \alias{Hnm} \alias{hnm} \alias{Hnm.diag} \title{Normal mixture bandwidth} \description{ Normal mixture bandwidth. } \usage{ Hnm(x, deriv.order=0, G=1:9, subset.ind, mise.flag=FALSE, verbose, ...) Hnm.diag(x, deriv.order=0, G=1:9, subset.ind, mise.flag=FALSE, verbose, ...) hnm(x, deriv.order=0, G=1:9, subset.ind, mise.flag=FALSE, verbose, ... ) } \arguments{ \item{x}{vector/matrix of data values} \item{deriv.order}{derivative order} \item{G}{range of number of mixture components} \item{subset.ind}{index vector of subset of \code{x} for fitting} \item{mise.flag}{flag to use MISE or AMISE minimisation. Default is FALSE.} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{...}{other parameters for \code{Mclust}} } \value{ Normal mixture bandwidth. If \code{mise=TRUE} then the minimal MISE value is returned too. } \details{ The normal mixture fit is provided by the \code{Mclust} function in the \pkg{mclust} package. \code{Hnm} is then \code{Hmise.mixt} (if \code{mise.flag=TRUE}) or \code{Hamise.mixt} (if \code{mise.flag=FALSE}) with these fitted normal mixture parameters. Likewise for \code{Hnm.diag}, \code{hnm}. } \seealso{ \link{Hmise.mixt}, \link{Hamise.mixt} } \references{Cwik, J. & Koronacki, J. (1997). A combined adaptive-mixtures/plug-in estimator of multivariate probability densities. \emph{Computational Statistics and Data Analysis}, \bold{26}, 199-218. } \examples{ data(unicef) Hnm(unicef) } \keyword{smooth} ks/man/tempb.Rd0000644000176200001440000000162214336531545013052 0ustar liggesusers\name{tempb} \docType{data} \alias{tempb} \title{Daily temperature} \description{ This data set contains the daily minimum and maximum temperatures from the weather station in Badajoz, Spain, from 1 January 1955 to 31 December 2015. } \usage{data(tempb)} \format{A matrix with 21908 rows and 5 columns. Each row corresponds to a daily measurement. The first column is the year (yyyy), the second is the month (mm), the third is the day (dd), the fourth is the minimum temperature (degrees Celsius), the fifth is the maximum temperature (degrees Celsius). } \source{ Menne, M. J., Durre, I., Vose, R. S., Gleason, B. E. & Houston, T. (2012) An overview of the global historical climatology network-daily database, \emph{Journal of Atmospheric and Oceanic Technology} \bold{429}, 897 - 910. \url{https://climexp.knmi.nl/selectdailyseries.cgi}. Accessed 2016-10-20. } \keyword{datasets} ks/man/kroc.Rd0000644000176200001440000000747214336531411012702 0ustar liggesusers\name{kroc} \alias{kroc} \alias{predict.kroc} \alias{summary.kroc} \title{Kernel receiver operating characteristic (ROC) curve} \description{ Kernel receiver operating characteristic (ROC) curve for 1- to 3-dimensional data.} \usage{ kroc(x1, x2, H1, h1, hy, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, verbose=FALSE) \method{predict}{kroc}(object, ..., x) \method{summary}{kroc}(object, ...) } \arguments{ \item{x,x1,x2}{vector/matrix of data values} \item{H1,h1,hy}{bandwidth matrix/scalar bandwidths. If these are missing, \code{Hpi.kcde}, \code{hpi.kcde} is called by default.} \item{gridsize}{vector of number of grid points} \item{gridtype}{not yet implemented} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{not yet implemented} \item{binned}{flag for binned estimation} \item{bgridsize}{vector of binning grid sizes} \item{positive}{flag if 1-d data are positive. Default is FALSE.} \item{adj.positive}{adjustment applied to positive 1-d data} \item{w}{vector of weights. Default is a vector of all ones.} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{object}{object of class \code{kroc}, output from \code{kroc}} \item{...}{other parameters} } \value{ A kernel ROC curve is an object of class \code{kroc} which is a list with fields: \item{x}{list of data values \code{x1, x2} - same as input} \item{eval.points}{vector or list of points at which the estimate is evaluated} \item{estimate}{ROC curve estimate at \code{eval.points}} \item{gridtype}{"linear"} \item{gridded}{flag for estimation on a grid} \item{binned}{flag for binned estimation} \item{names}{variable names} \item{w}{vector of weights} \item{tail}{"lower.tail"} \item{h1}{scalar bandwidth for first sample (1-d only)} \item{H1}{bandwidth matrix for first sample} \item{hy}{scalar bandwidth for ROC curve} \item{indices}{summary indices of ROC curve.} } \details{ In this set-up, the values in the first sample \code{x1} should be larger in general that those in the second sample \code{x2}. The usual method for computing 1-d ROC curves is not valid for multivariate data. Duong (2014), based on Lloyd (1998), develops an alternative formulation \eqn{(F_{Y_1}(z), F_{Y_2}(z))}{(F_Y1(z), F_Y2(z))} based on the cumulative distribution functions of \eqn{Y_j = \bar{F}_1(\bold{X}_j), j=1,2}{Yj=bar(F)_1(Xj), j=1,2}. If the bandwidth \code{H1} is missing from \code{kroc}, then the default bandwidth is the plug-in selector \code{Hpi.kcde}. Likewise for missing \code{h1,hy}. A bandwidth matrix \code{H1} is required for \code{x1} for d>1, but the second bandwidth \code{hy} is always a scalar since \eqn{Y_j}{Yj} are 1-d variables. The effective support, binning, grid size, grid range, positive parameters are the same as \code{\link{kde}}. --The \code{summary} method for \code{kroc} objects prints out the summary indices of the ROC curve, as contained in the \code{indices} field, namely the AUC (area under the curve) and Youden index. } \references{ Duong, T. (2016) Non-parametric smoothed estimation of multivariate cumulative distribution and survival functions, and receiver operating characteristic curves. \emph{Journal of the Korean Statistical Society}, \bold{45}, 33-50. Lloyd, C. (1998) Using smoothed receiver operating curves to summarize and compare diagnostic systems. \emph{Journal of the American Statistical Association}, \bold{93}, 1356-1364. } \seealso{\code{\link{kcde}}} \examples{ samp <- 1000 x <- rnorm.mixt(n=samp, mus=0, sigmas=1, props=1) y <- rnorm.mixt(n=samp, mus=0.5, sigmas=1, props=1) Rhat <- kroc(x1=x, x2=y) summary(Rhat) predict(Rhat, x=0.5) } \keyword{smooth} ks/man/plot.kcde.Rd0000644000176200001440000000304614340026013013610 0ustar liggesusers\name{plot.kcde} \alias{plot.kcde} \title{Plot for kernel cumulative distribution estimate} \description{ Plot for kernel cumulative distribution estimate 1- to 3-dimensional data. } \usage{ \method{plot}{kcde}(x, ...) } \arguments{ \item{x}{object of class \code{kcde} (output from \code{\link{kcde}})} \item{...}{other graphics parameters used in \code{\link{plot.kde}}} } \value{ Plots for 1-d and 2-d are sent to graphics window. Plot for 3-d is sent to graphics/RGL window. } \details{ For \code{kcde} objects, the function headers for the different dimensional data are \preformatted{ ## univariate plot(Fhat, xlab, ylab="Distribution function", add=FALSE, drawpoints=FALSE, col.pt=4, jitter=FALSE, alpha=1, ...) ## bivariate plot(Fhat, display="persp", cont=seq(10,90, by=10), abs.cont, xlab, ylab, zlab="Distribution function", cex=1, pch=1, add=FALSE, drawpoints=FALSE, drawlabels=TRUE, theta=-30, phi=40, d=4, col.pt=4, col, col.fun, alpha=1, lwd=1, border=NA, thin=3, lwd.fc=5, ...) ## trivariate plot(Fhat, display="plot3D", cont=c(25,50,75), colors, col, alphavec, size=3, cex=1, pch=1, theta=-30, phi=40, d=4, ticktype="detailed", bty="f", col.pt=4, add=FALSE, xlab, ylab, zlab, drawpoints=FALSE, alpha, box=TRUE, axes=TRUE, ...)} } \seealso{\code{\link{plot.kde}}} \examples{ data(iris) Fhat <- kcde(x=iris[,1]) plot(Fhat, xlab="Sepal.Length") Fhat <- kcde(x=iris[,1:2]) plot(Fhat) Fhat <- kcde(x=iris[,1:3]) plot(Fhat, alpha=0.3) } \keyword{hplot} ks/man/plot.kde.loctest.Rd0000644000176200001440000000421514606000765015133 0ustar liggesusers\name{plot.kde.loctest} \alias{plot.kde.loctest} \title{Plot for kernel local significant difference regions} \description{ Plot for kernel local significant difference regions for 1- to 3-dimensional data. } \usage{ \method{plot}{kde.loctest}(x, ...) } \arguments{ \item{x}{object of class \code{kde.loctest} (output from \code{\link{kde.local.test}})} \item{...}{other graphics parameters: \describe{ \item{\code{lcol}}{colour for KDE curve (1-d)} \item{\code{col}}{vector of 2 colours. First colour: sample 1>sample 2, second colour: sample 10 \cr "dunconstr" = single unconstrained pilot bandwidth for deriv.order>0} \item{Hstart}{initial bandwidth matrix, used in numerical optimisation} \item{binned}{flag for binned kernel estimation} \item{bgridsize}{vector of binning grid sizes} \item{amise}{flag to return the minimal scaled SCV value. Default is FALSE.} \item{deriv.order}{derivative order} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{optim.fun}{optimiser function: one of \code{nlm} or \code{optim}} \item{nstage}{number of stages in the SCV bandwidth selector (1 or 2)} \item{plot}{flag to display plot of SCV(h) vs h (1-d only). Default is FALSE.} } \value{ SCV bandwidth. If \code{amise=TRUE} then the minimal scaled SCV value is returned too. } \details{ \code{hscv} is the univariate SCV selector of Jones, Marron & Park (1991). \code{Hscv} is a multivariate generalisation of this, see Duong & Hazelton (2005). Use \code{Hscv} for unconstrained bandwidth matrices and \code{Hscv.diag} for diagonal bandwidth matrices. The default pilot is \code{"samse"} for d=2, r=0, and \code{"dscalar"} otherwise. For SAMSE pilot bandwidths, see Duong & Hazelton (2005). Unconstrained and higher order derivative pilot bandwidths are from Chacon & Duong (2011). For d=1, the selector \code{hscv} is not always stable for large sample sizes with binning. Examine the plot from \code{hscv(, plot=TRUE)} to determine the appropriate smoothness of the SCV function. Any non-smoothness is due to the discretised nature of binned estimation. For details about the advanced options for \code{binned, Hstart, optim.fun}, see \code{\link{Hpi}}. } \references{ Chacon, J.E. & Duong, T. (2011) Unconstrained pilot selectors for smoothed cross validation. \emph{Australian & New Zealand Journal of Statistics}, \bold{53}, 331-351. Duong, T. & Hazelton, M.L. (2005) Cross-validation bandwidth matrices for multivariate kernel density estimation. \emph{Scandinavian Journal of Statistics}, \bold{32}, 485-506. Jones, M.C., Marron, J.S. & Park, B.U. (1991) A simple root \eqn{n}{n} bandwidth selector. \emph{Annals of Statistics}, \bold{19}, 1919-1932. } \seealso{\code{\link{Hbcv}}, \code{\link{Hlscv}}, \code{\link{Hpi}}} \examples{ data(unicef) Hscv(unicef) hscv(unicef[,1]) } \keyword{smooth} ks/man/kfs.Rd0000644000176200001440000001056414340025641012521 0ustar liggesusers\name{kfs} \alias{kfs} \title{Kernel feature significance } \description{ Kernel feature significance for 1- to 6-dimensional data. } \usage{ kfs(x, H, h, deriv.order=2, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, positive=FALSE, adj.positive, w, verbose=FALSE, signif.level=0.05) } \arguments{ \item{x}{matrix of data values} \item{H,h}{bandwidth matrix/scalar bandwidth. If these are missing, \code{Hpi} or \code{hpi} is called by default.} \item{deriv.order}{derivative order (scalar)} \item{gridsize}{vector of number of grid points} \item{gridtype}{not yet implemented} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{vector or matrix of points at which estimate is evaluated} \item{binned}{flag for binned estimation} \item{bgridsize}{vector of binning grid sizes} \item{positive}{flag if 1-d data are positive. Default is FALSE.} \item{adj.positive}{adjustment applied to positive 1-d data} \item{w}{vector of weights. Default is a vector of all ones.} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{signif.level}{overall level of significance for hypothesis tests. Default is 0.05.} } \value{ A kernel feature significance estimate is an object of class \code{kfs} which is a list with fields \item{x}{data points - same as input} \item{eval.points}{vector or list of points at which the estimate is evaluated} \item{estimate}{binary matrix for significant feature at \code{eval.points}: 0 = not signif., 1 = signif.} \item{h}{scalar bandwidth (1-d only)} \item{H}{bandwidth matrix} \item{gridtype}{"linear"} \item{gridded}{flag for estimation on a grid} \item{binned}{flag for binned estimation} \item{names}{variable names} \item{w}{vector of weights} \item{deriv.order}{derivative order (scalar)} \item{deriv.ind}{martix where each row is a vector of partial derivative indices.} This is the same structure as a \code{kdde} object, except that \code{estimate} is a binary matrix rather than real-valued. } \details{ Feature significance is based on significance testing of the gradient (first derivative) and curvature (second derivative) of a kernel density estimate. Only the latter is currently implemented, and is also known as significant modal regions. The hypothesis test at a grid point \eqn{\bold{x}}{x} is \eqn{H_0(\bold{x}): \mathsf{H} f(\bold{x}) < 0}{H0(x): H f(x) < 0}, i.e. the density Hessian matrix \eqn{\mathsf{H} f(\bold{x})}{H f(x)} is negative definite. The \eqn{p}{p}-values are computed for each \eqn{\bold{x}}{x} using that the test statistic is approximately chi-squared distributed with \eqn{d(d+1)/2}{d(d+1)/2} d.f. We then use a Hochberg-type simultaneous testing procedure, based on the ordered \eqn{p}{p}-values, to control the overall level of significance to be \code{signif.level}. If \eqn{H_0(\bold{x})}{H0(x)} is rejected then \eqn{\bold{x}}{x} belongs to a significant modal region. The computations are based on \code{kdde(x, deriv.order=2)} so \code{kfs} inherits its behaviour from \code{\link{kdde}}. If the bandwidth \code{H} is missing, then the default bandwidth is the plug-in selector \code{Hpi(,deriv.order=2)}. Likewise for missing \code{h}. The effective support, binning, grid size, grid range, positive parameters are the same as \code{\link{kde}}. This function is similar to the \code{featureSignif} function in the \pkg{feature} package, except that it accepts unconstrained bandwidth matrices. } \references{ Chaudhuri, P. & Marron, J.S. (1999) SiZer for exploration of structures in curves. \emph{Journal of the American Statistical Association}, \bold{94}, 807-823. Duong, T., Cowling, A., Koch, I. & Wand, M.P. (2008) Feature significance for multivariate kernel density estimation. \emph{Computational Statistics and Data Analysis}, \bold{52}, 4225-4242. Godtliebsen, F., Marron, J.S. & Chaudhuri, P. (2002) Significance in scale space for bivariate density estimation. \emph{Journal of Computational and Graphical Statistics}, \bold{11}, 1-22. } \seealso{\code{\link{kdde}}, \code{\link{plot.kfs}}} \examples{ data(geyser, package="MASS") geyser.fs <- kfs(geyser$duration, binned=TRUE) plot(geyser.fs, xlab="duration") ## see example in ? plot.kfs } \keyword{smooth} ks/man/plotmixt.Rd0000644000176200001440000000315714336531523013624 0ustar liggesusers\name{plotmixt} \alias{plotmixt} \title{Plot for 1- to 3-dimensional normal and t-mixture density functions} \description{ Plot for 1- to 3-dimensional normal and t-mixture density functions. } \usage{ plotmixt(mus, sigmas, Sigmas, props, dfs, dist="normal", draw=TRUE, deriv.order=0, which.deriv.ind=1, binned=TRUE, ...) } \arguments{ \item{mus}{(stacked) matrix of mean vectors} \item{sigmas}{vector of standard deviations (1-d)} \item{Sigmas}{(stacked) matrix of variance matrices (2-d, 3-d)} \item{props}{vector of mixing proportions} \item{dfs}{vector of degrees of freedom} \item{dist}{"normal" - normal mixture, "t" - t-mixture} \item{draw}{flag to draw plot. Default is TRUE.} \item{deriv.order}{derivative order} \item{which.deriv.ind}{index of which partial derivative to plot} \item{binned}{flag for binned estimation of contour levels. Default is TRUE.} \item{...}{other graphics parameters, see \code{\link{plot.kde}}} } \value{ If \code{draw=TRUE}, the 1-d, 2-d plot is sent to graphics window, 3-d plot to graphics/RGL window. If \code{draw=FALSE}, then a \code{kdde}-like object is returned. } \examples{ ## bivariate mus <- rbind(c(0,0), c(-1,1)) Sigma <- matrix(c(1, 0.7, 0.7, 1), nr=2, nc=2) Sigmas <- rbind(Sigma, Sigma) props <- c(1/2, 1/2) plotmixt(mus=mus, Sigmas=Sigmas, props=props, display="filled.contour", lwd=1) ## trivariate mus <- rbind(c(0,0,0), c(-1,0.5,1.5)) Sigma <- matrix(c(1, 0.7, 0.7, 0.7, 1, 0.7, 0.7, 0.7, 1), nr=3, nc=3) Sigmas <- rbind(Sigma, Sigma) props <- c(1/2, 1/2) plotmixt(mus=mus, Sigmas=Sigmas, props=props, dfs=c(11,8), dist="t") } \keyword{hplot} ks/man/vector.Rd0000644000176200001440000000210514336531560013237 0ustar liggesusers\name{vector} \alias{vec} \alias{vech} \alias{invvec} \alias{invvech} \title{Vector and vector half operators} \description{ The vec (vector) operator takes a \eqn{d \times d}{d x d} matrix and stacks the columns into a single vector of length \eqn{d^2}{d^2}. The vech (vector half) operator takes a symmetric \eqn{d \times d}{d x d} matrix and stacks the lower triangular half into a single vector of length \eqn{d(d+1)/2}{d(d+1)/2}. The functions invvec and invvech are the inverses of vec and vech i.e. they form matrices from vectors. } \usage{ vec(x, byrow=FALSE) vech(x) invvec(x, ncol, nrow, byrow=FALSE) invvech(x) } \arguments{ \item{x}{vector or matrix} \item{ncol,nrow}{number of columns and rows for inverse of vech} \item{byrow}{flag for stacking row-wise or column-wise. Default is FALSE.} } \references{ Magnus, J.R. & Neudecker H.M. (2007) \emph{Matrix Differential Calculus with Applications in Statistics and Econometrics (3rd edition)}, Wiley & Sons. Chichester. } \examples{ x <- matrix(1:9, nrow=3, ncol=3) vec(x) invvec(vec(x)) } \keyword{algebra} ks/man/plot.kda.Rd0000644000176200001440000000450514340026271013450 0ustar liggesusers\name{plot.kda} \alias{plot.kda} \title{Plot for kernel discriminant analysis} \description{ Plot for kernel discriminant analysis for 1- to 3-dimensional data. } \usage{ \method{plot}{kda}(x, y, y.group, ...) } \arguments{ \item{x}{ object of class \code{kda} (output from \code{\link{kda}})} \item{y}{matrix of test data points} \item{y.group}{vector of group labels for test data points} \item{...}{other graphics parameters: \describe{ \item{\code{rugsize}}{height of rug-like plot for partition classes (1-d)} \item{\code{prior.prob}}{vector of prior probabilities} \item{\code{col.part}}{vector of colours for partition classes (1-d, 2-d)} } and those used in \code{\link{plot.kde}} } } \value{ Plots for 1-d and 2-d are sent to graphics window. Plot for 3-d is sent to graphics/RGL window. } \details{ For \code{kda} objects, the function headers for the different dimensional data are \preformatted{ ## univariate plot(x, y, y.group, prior.prob=NULL, xlim, ylim, xlab, ylab="Weighted density function", drawpoints=FALSE, col, col.fun, col.part, col.pt, lty, jitter=TRUE, rugsize, add=FALSE, alpha=1, ...) ## bivariate plot(x, y, y.group, prior.prob=NULL, display.part="filled.contour", cont=c(25,50,75), abs.cont, approx.cont=TRUE, xlim, ylim, xlab, ylab, drawpoints=FALSE, drawlabels=TRUE, cex=1, pch, lty, part=TRUE, col, col.fun, col.part, col.pt, alpha=1, lwd=1, lwd.part=0, add=FALSE, ...) ## trivariate plot(x, y, y.group, prior.prob=NULL, display="plot3D", cont=c(25,50,75), abs.cont, approx.cont=TRUE, colors, col, col.fun, col.pt, alpha=0.5, alphavec, xlab, ylab, zlab, drawpoints=FALSE, size=3, cex=1, pch, theta=-30, phi=40, d=4, ticktype="detailed", bty="f", add=FALSE, ...)} } \seealso{\code{\link{kda}}, \code{\link{kde}}} \examples{ data(iris) ## univariate example ir <- iris[,1] ir.gr <- iris[,5] kda.fhat <- kda(x=ir, x.group=ir.gr, xmin=3, xmax=9) plot(kda.fhat, xlab="Sepal length") ## bivariate example ir <- iris[,1:2] ir.gr <- iris[,5] kda.fhat <- kda(x=ir, x.group=ir.gr) plot(kda.fhat, alpha=0.2, drawlabels=FALSE) ## trivariate example ir <- iris[,1:3] ir.gr <- iris[,5] kda.fhat <- kda(x=ir, x.group=ir.gr) plot(kda.fhat) ## colour=species, transparency=density heights} \keyword{hplot} ks/man/ks-internal.Rd0000644000176200001440000000072414336531417014172 0ustar liggesusers\name{ks-internal} \alias{getRow} \alias{Lpdiff} \alias{matrix.sqrt} \alias{mur} \alias{nur} \alias{nurs} \alias{Qr} \alias{rowKpow} \alias{Sdr} \alias{Sdrv} \alias{symconv.1d} \alias{symconv.nd} \alias{dwsupp} \alias{reg.ucv} \alias{kr} \title{Internal functions in the ks library} \description{ These functions are user-level but which the user is not required to use directly. } \value{The user is not required to use directly these outputs.} \keyword{internal} ks/man/ise.mixt.Rd0000644000176200001440000000466714336531314013511 0ustar liggesusers\name{ise.mixt} \alias{Hmise.mixt} \alias{Hamise.mixt} \alias{Hmise.mixt.diag} \alias{Hamise.mixt.diag} \alias{hmise.mixt} \alias{hamise.mixt} \alias{ise.mixt} \alias{amise.mixt} \alias{mise.mixt} \title{Squared error bandwidth matrix selectors for normal mixture densities} \description{ The global errors ISE (Integrated Squared Error), MISE (Mean Integrated Squared Error) and the AMISE (Asymptotic Mean Integrated Squared Error) for 1- to 6-dimensional data. Normal mixture densities have closed form expressions for the MISE and AMISE. So in these cases, we can numerically minimise these criteria to find MISE- and AMISE-optimal matrices. } \usage{ Hamise.mixt(mus, Sigmas, props, samp, Hstart, deriv.order=0) Hmise.mixt(mus, Sigmas, props, samp, Hstart, deriv.order=0) Hamise.mixt.diag(mus, Sigmas, props, samp, Hstart, deriv.order=0) Hmise.mixt.diag(mus, Sigmas, props, samp, Hstart, deriv.order=0) hamise.mixt(mus, sigmas, props, samp, hstart, deriv.order=0) hmise.mixt(mus, sigmas, props, samp, hstart, deriv.order=0) amise.mixt(H, mus, Sigmas, props, samp, h, sigmas, deriv.order=0) ise.mixt(x, H, mus, Sigmas, props, h, sigmas, deriv.order=0, binned=FALSE, bgridsize) mise.mixt(H, mus, Sigmas, props, samp, h, sigmas, deriv.order=0) } \arguments{ \item{mus}{(stacked) matrix of mean vectors (>1-d), vector of means (1-d)} \item{Sigmas,sigmas}{(stacked) matrix of variance matrices (>1-d), vector of standard deviations (1-d)} \item{props}{vector of mixing proportions} \item{samp}{sample size} \item{Hstart,hstart}{initial bandwidth (matrix), used in numerical optimisation} \item{deriv.order}{derivative order} \item{x}{matrix of data values} \item{H,h}{bandwidth (matrix)} \item{binned}{flag for binned kernel estimation. Default is FALSE.} \item{bgridsize}{vector of binning grid sizes} } \value{ MISE- or AMISE-optimal bandwidth matrix. ISE, MISE or AMISE value. } \details{ ISE is a random variable that depends on the data \code{x}. MISE and AMISE are non-random and don't depend on the data. For normal mixture densities, ISE, MISE and AMISE have exact formulas for all dimensions. } \references{Chacon J.E., Duong, T. & Wand, M.P. (2011). Asymptotics for general multivariate kernel density derivative estimators. \emph{Statistica Sinica}, \bold{21}, 807-840. } \examples{ x <- rmvnorm.mixt(100) Hamise.mixt(samp=nrow(x), mus=rep(0,2), Sigmas=var(x), props=1, deriv.order=1) } \keyword{smooth} ks/man/kda.Rd0000644000176200001440000001430014336531327012475 0ustar liggesusers\name{kda} \alias{Hkda} \alias{Hkda.diag} \alias{kda} \alias{hkda} \alias{predict.kda} \alias{compare} \alias{compare.kda.diag.cv} \alias{compare.kda.cv} \title{Kernel discriminant analysis (kernel classification)} \description{ Kernel discriminant analysis (kernel classification) for 1- to d-dimensional data. } \usage{ kda(x, x.group, Hs, hs, prior.prob=NULL, gridsize, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, kde.flag=TRUE) Hkda(x, x.group, Hstart, bw="plugin", ...) Hkda.diag(x, x.group, bw="plugin", ...) hkda(x, x.group, bw="plugin", ...) \method{predict}{kda}(object, ..., x) compare(x.group, est.group, by.group=FALSE) compare.kda.cv(x, x.group, bw="plugin", prior.prob=NULL, Hstart, by.group=FALSE, verbose=FALSE, recompute=FALSE, ...) compare.kda.diag.cv(x, x.group, bw="plugin", prior.prob=NULL, by.group=FALSE, verbose=FALSE, recompute=FALSE, ...) } \arguments{ \item{x}{matrix of training data values} \item{x.group}{vector of group labels for training data} \item{Hs,hs}{(stacked) matrix of bandwidth matrices/vector of scalar bandwidths. If these are missing, \code{Hkda} or \code{hkda} is called by default.} \item{prior.prob}{vector of prior probabilities} \item{gridsize}{vector of grid sizes} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{vector or matrix of points at which estimate is evaluated} \item{binned}{flag for binned estimation} \item{bgridsize}{vector of binning grid sizes} \item{w}{vector of weights. Not yet implemented.} \item{compute.cont}{flag for computing 1\% to 99\% probability contour levels. Default is TRUE.} \item{approx.cont}{flag for computing approximate probability contour levels. Default is TRUE.} \item{kde.flag}{flag for computing KDE on grid. Default is TRUE.} \item{object}{object of class \code{kda}} \item{bw}{bandwidth: "plugin" = plug-in, "lscv" = LSCV, "scv" = SCV} \item{Hstart}{(stacked) matrix of initial bandwidth matrices, used in numerical optimisation} \item{est.group}{vector of estimated group labels} \item{by.group}{flag to give results also within each group} \item{verbose}{flag for printing progress information. Default is FALSE.} \item{recompute}{flag for recomputing the bandwidth matrix after excluding the i-th data item} \item{...}{other optional parameters for bandwidth selection, see \code{\link{Hpi}}, \code{\link{Hlscv}}, \code{\link{Hscv}}} } \value{ --For \code{kde.flag=TRUE}, a kernel discriminant analysis is an object of class \code{kda} which is a list with fields \item{x}{list of data points, one for each group label} \item{estimate}{list of density estimates at \code{eval.points}, one for each group label} \item{eval.points}{vector or list of points that the estimate is evaluated at, one for each group label} \item{h}{vector of bandwidths (1-d only)} \item{H}{stacked matrix of bandwidth matrices or vector of bandwidths} \item{gridded}{flag for estimation on a grid} \item{binned}{flag for binned estimation} \item{w}{vector of weights} \item{prior.prob}{vector of prior probabilities} \item{x.group}{vector of group labels - same as input} \item{x.group.estimate}{vector of estimated group labels. If the test data \code{eval.points} are given then these are classified. Otherwise the training data \code{x} are classified.} For \code{kde.flag=FALSE}, which is always the case for \eqn{d>3}{d>3}, then only the vector of estimated group labels is returned. --The result from \code{Hkda} and \code{Hkda.diag} is a stacked matrix of bandwidth matrices, one for each training data group. The result from \code{hkda} is a vector of bandwidths, one for each training group. --The \code{compare} functions create a comparison between the true group labels \code{x.group} and the estimated ones. It returns a list with fields \item{cross}{cross-classification table with the rows indicating the true group and the columns the estimated group} \item{error}{misclassification rate (MR)} In the case where the test data are independent of the training data, \code{compare} computes MR = (number of points wrongly classified)/(total number of points). In the case where the test data are not independent e.g. we are classifying the training data set itself, then the cross validated estimate of MR is more appropriate. These are implemented as \code{compare.kda.cv} (unconstrained bandwidth selectors) and \code{compare.kda.diag.cv} (for diagonal bandwidth selectors). These functions are only available for d > 1. If \code{by.group=FALSE} then only the total MR rate is given. If it is set to TRUE, then the MR rates for each class are also given (estimated number in group divided by true number). } \references{ Simonoff, J. S. (1996) \emph{Smoothing Methods in Statistics}. Springer-Verlag. New York } \details{ If the bandwidths \code{Hs} are missing from \code{kda}, then the default bandwidths are the plug-in selectors \code{Hkda(, bw="plugin")}. Likewise for missing \code{hs}. Valid options for \code{bw} are \code{"plugin"}, \code{"lscv"} and \code{"scv"} which in turn call \code{\link{Hpi}}, \code{\link{Hlscv}} and \code{\link{Hscv}}. The effective support, binning, grid size, grid range, positive parameters are the same as \code{\link{kde}}. If prior probabilities are known then set \code{prior.prob} to these. Otherwise \code{prior.prob=NULL} uses the sample proportions as estimates of the prior probabilities. For \pkg{ks} \eqn{\geq}{>=} 1.8.11, \code{kda.kde} has been subsumed into \code{kda}, so all prior calls to \code{kda.kde} can be replaced by \code{kda}. To reproduce the previous behaviour of \code{kda}, the command is \code{kda(, kde.flag=FALSE)}. } \seealso{\code{\link{plot.kda}}} \examples{ set.seed(8192) x <- c(rnorm.mixt(n=100, mus=1), rnorm.mixt(n=100, mus=-1)) x.gr <- rep(c(1,2), times=c(100,100)) y <- c(rnorm.mixt(n=100, mus=1), rnorm.mixt(n=100, mus=-1)) y.gr <- rep(c(1,2), times=c(100,100)) kda.gr <- kda(x, x.gr) y.gr.est <- predict(kda.gr, x=y) compare(y.gr, y.gr.est) ## See other examples in ? plot.kda } \keyword{smooth} ks/man/plot.histde.Rd0000644000176200001440000000473714340025760014202 0ustar liggesusers\name{plot.histde} \alias{plot.histde} \title{Plot for histogram density estimate} \description{ Plot for histogram density estimate for 1- and 2-dimensional data. } \usage{ \method{plot}{histde}(x, ...) } \arguments{ \item{x}{object of class \code{histde} (output from \code{\link{histde}})} \item{...}{other graphics parameters: \describe{ \item{\code{col}}{plotting colour for density estimate} \item{\code{col.fun}}{plotting colour function for levels} \item{\code{col.pt}}{plotting colour for data points} \item{\code{jitter}}{flag to jitter rug plot (1-d). Default is TRUE.} \item{\code{xlim,ylim}}{axes limits} \item{\code{xlab,ylab}}{axes labels} \item{\code{add}}{flag to add to current plot. Default is FALSE.} \item{\code{drawpoints}}{flag to draw data points on density estimate. Default is FALSE.} \item{\code{breaks}}{vector of break values of density estimate. Default is an \code{nbreaks} equilinear sequence over the data range.} \item{\code{nbreaks}}{number of breaks in \code{breaks} sequence} \item{\code{lty.rect},\code{lwd.rect}}{line type/width for histogram box lines (2-d)} \item{\code{border}}{colour of histogram box lines (2-d)} \item{\code{col.rect}}{colour of histogram bars (1-d)} \item{\code{add.grid}}{flag to add histogram grid (2-d). Default is TRUE.} } } } \value{ Plots for 1-d and 2-d are sent to graphics window. } \details{ For \code{histde} objects, the function headers for the different dimensional data are \preformatted{ ## univariate plot(fhat, xlab, ylab="Density function", add=FALSE, drawpoints=FALSE, col.pt=4, jitter=FALSE, border=1, alpha=1, ...) ## bivariate plot(fhat, breaks, nbreaks=11, xlab, ylab, zlab="Density function", cex=1, pch=1, add=FALSE, drawpoints=FALSE, col, col.fun, alpha=1, col.pt=4, lty.rect=2, cex.text=1, border, lwd.rect=1, col.rect="transparent", add.grid=TRUE, ...)} The 1-d plot is a standard plot of a histogram generated by \code{hist}. If \code{drawpoints=TRUE} then a rug plot is added. The 2-d plot is similar to the \code{display="filled.contour"} option from \code{\link{plot.kde}} with the default \code{nbreaks=11} contour levels. } \seealso{\code{\link{plot.kde}}} \examples{ data(iris) ## univariate example fhat <- histde(x=iris[,2]) plot(fhat, xlab="Sepal length") ## bivariate example fhat <- histde(x=iris[,2:3]) plot(fhat, drawpoints=TRUE) box() } \keyword{hplot} ks/man/rkde.Rd0000644000176200001440000000320014336531541012656 0ustar liggesusers\name{rkde} \alias{dkde} \alias{pkde} \alias{qkde} \alias{rkde} \title{Derived quantities from kernel density estimates} \description{ Derived quantities from kernel density estimates. } \usage{ dkde(x, fhat) pkde(q, fhat) qkde(p, fhat) rkde(n, fhat, positive=FALSE) } \arguments{ \item{x,q}{vector of quantiles} \item{p}{vector of probabilities} \item{n}{number of observations} \item{positive}{flag to compute KDE on the positive real line. Default is FALSE.} \item{fhat}{kernel density estimate, object of class \code{kde}} } \value{ For the 1-d kernel density estimate \code{fhat}, \code{pkde} computes the cumulative probability for the quantile \code{q}, \code{qkde} computes the quantile corresponding to the probability \code{p}. For any kernel density estimate, \code{dkde} computes the density value at \code{x} (it is an alias for \code{predict.kde}), \code{rkde} computes a random sample of size \code{n}. } \details{ \code{pkde} uses the trapezoidal rule for the numerical integration. \code{rkde} uses Silverman (1986)'s method to generate a random sample from a KDE. } \references{ Silverman, B. (1986) \emph{Density Estimation for Statistics and Data Analysis}. Chapman & Hall/CRC. London.} \examples{ set.seed(8192) x <- rnorm.mixt(n=10000, mus=0, sigmas=1, props=1) fhat <- kde(x=x) p1 <- pkde(fhat=fhat, q=c(-1, 0, 0.5)) qkde(fhat=fhat, p=p1) y <- rkde(fhat=fhat, n=100) x <- rmvnorm.mixt(n=10000, mus=c(0,0), Sigmas=invvech(c(1,0.8,1))) fhat <- kde(x=x) y <- rkde(fhat=fhat, n=1000) fhaty <- kde(x=y) plot(fhat, col=1) plot(fhaty, add=TRUE, col=2) } \keyword{smooth} ks/man/Hns.Rd0000644000176200001440000000211714336766431012477 0ustar liggesusers\name{Hns} \alias{Hns} \alias{Hns.diag} \alias{hns} \alias{Hns.kcde} \alias{hns.kcde} \title{Normal scale bandwidth} \description{ Normal scale bandwidth. } \usage{ Hns(x, deriv.order=0) Hns.diag(x) hns(x, deriv.order=0) Hns.kcde(x) hns.kcde(x) } \arguments{ \item{x}{vector/matrix of data values} \item{deriv.order}{derivative order} } \value{ Normal scale bandwidth. } \details{ \code{Hns} is equal to \code{(4/(n*(d+2*r+2)))^(2/(d+2*r+4))*var(x)}, n = sample size, d = dimension of data, r = derivative order. \code{hns} is the analogue of \code{Hns} for 1-d data. These can be used for density (derivative) estimators \code{\link{kde}}, \code{\link{kdde}}. The equivalents for distribution estimators \code{\link{kcde}} are \code{Hns.kcde} and \code{hns.cde}. } \references{Chacon J.E., Duong, T. & Wand, M.P. (2011). Asymptotics for general multivariate kernel density derivative estimators. \emph{Statistica Sinica}, \bold{21}, 807-840. } \examples{ data(forbes, package="MASS") Hns(forbes, deriv.order=2) hns(forbes$bp, deriv.order=2) } \keyword{smooth} ks/man/kms.Rd0000644000176200001440000001120214340025672012522 0ustar liggesusers\name{kms} \alias{kms} \alias{summary.kms} \alias{plot.kms} \title{Kernel mean shift clustering} \description{ Kernel mean shift clustering for 2- to 6-dimensional data. } \usage{ kms(x, y, H, max.iter=400, tol.iter, tol.clust, min.clust.size, merge=TRUE, keep.path=FALSE, verbose=FALSE) \method{plot}{kms}(x, display="splom", col, col.fun, alpha=1, xlab, ylab, zlab, theta=-30, phi=40, add=FALSE, ...) \method{summary}{kms}(object, ...) } \arguments{ \item{x}{matrix of data values or object of class \code{kms}} \item{y}{matrix of candidate data values for which the mean shift will estimate their cluster labels. If missing, \code{y=x}.} \item{H}{bandwidth matrix/scalar bandwidth. If missing, \code{Hpi(x,deriv.order=1,nstage=2-(d>2))} is called by default.} \item{max.iter}{maximum number of iterations. Default is 400.} \item{tol.iter}{distance under which two successive iterations are considered convergent. Default is 0.001*min marginal IQR of \code{x}.} \item{tol.clust}{distance under which two cluster modes are considered to form one cluster. Default is 0.01*max marginal IQR of \code{x}.} \item{min.clust.size}{minimum cluster size (cardinality). Default is \code{0.01*nrow(y)}.} \item{merge}{flag to merge clusters which are smaller than \code{min.clust.size}. Default is TRUE.} \item{keep.path}{flag to store the density gradient ascent paths. Default is FALSE.} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{object}{object of class \code{kms}} \item{display}{type of display, "splom" (>=2-d) or "plot3D" (3-d)} \item{col,col.fun}{vector or colours (one for each group) or colour function} \item{alpha}{colour transparency. Default is 1.} \item{xlab,ylab,zlab}{axes labels} \item{theta,phi}{graphics parameters for perspective plots (3-d)} \item{add}{flag to add to current plot. Default is FALSE.} \item{...}{other (graphics) parameters} } \value{ A kernel mean shift clusters set is an object of class \code{kms} which is a list with fields: \item{x,y}{data points - same as input} \item{end.points}{matrix of final iterates starting from \code{y}} \item{H}{bandwidth matrix} \item{label}{vector of cluster labels} \item{nclust}{number of clusters} \item{nclust.table}{frequency table of cluster labels} \item{mode}{matrix of cluster modes} \item{names}{variable names} \item{tol.iter,tol.clust,min.clust.size}{tuning parameter values - same as input} \item{path}{list of density gradient ascent paths where \code{path[[i]]} is the path of \code{y[i,]} (only if \code{keep.path=TRUE})} } \details{ Mean shift clustering belongs to the class of modal or density-based clustering methods. The mean shift recurrence of the candidate point \eqn{{\bold x}}{x} is \eqn{{\bold x}_{j+1} = {\bold x}_j + \bold{{\rm H}} {\sf D} \hat{f}({\bold x}_j)/\hat{f}({\bold x}_j)}{x_j+1 = x_j + H D hat(f)(x_j)/hat(f)(x_j)} where \eqn{j\geq 0}{j>=0} and \eqn{{\bold x}_0 = {\bold x}}{x_0 = x}. The sequence \eqn{\{{\bold x}_0, {\bold x}_1, \dots \}}{x_0, x_1, ...} follows the density gradient ascent paths to converge to a local mode of the density estimate \eqn{\hat{f}}{hat(f)}. Hence \eqn{{\bold x}}{x} is iterated until it converges to its local mode, and this determines its cluster label. The mean shift recurrence is terminated if successive iterations are less than \code{tol.iter} or the maximum number of iterations \code{max.iter} is reached. Final iterates which are less than \code{tol.clust} distance apart are considered to form a single cluster. If \code{merge=TRUE} then the clusters whose cardinality is less than \code{min.clust.size} are iteratively merged with their nearest cluster. If the bandwidth \code{H} is missing, then the default bandwidth is the plug-in selector for the density gradient \code{Hpi(x,deriv.order=1)}. Any bandwidth that is suitable for the density gradient is also suitable for the mean shift. } \references{ Chacon, J.E. & Duong, T. (2013) Data-driven density estimation, with applications to nonparametric clustering and bump hunting. \emph{Electronic Journal of Statistics}, \bold{7}, 499-532. Comaniciu, D. & Meer, P. (2002). Mean shift: a robust approach toward feature space analysis. \emph{ IEEE Transactions on Pattern Analysis and Machine Intelligence}, \bold{24}, 603-619. } \seealso{\code{\link{kde}}} \examples{ data(crabs, package="MASS") kms.crabs <- kms(x=crabs[,c("FL","CW")]) plot(kms.crabs, pch=16) summary(kms.crabs) kms.crabs <- kms(x=crabs[,c("FL","CW","RW")]) plot(kms.crabs, pch=16) plot(kms.crabs, display="plot3D", pch=16) } \keyword{cluster} ks/man/plot.kroc.Rd0000644000176200001440000000226014340026165013645 0ustar liggesusers\name{plot.kroc} \alias{plot.kroc} \title{Plot for kernel receiver operating characteristic curve (ROC) estimate} \description{ Plot for kernel receiver operating characteristic curve (ROC) estimate 1- to 3-dimensional data. } \usage{ \method{plot}{kroc}(x, add=FALSE, add.roc.ref=FALSE, xlab, ylab, alpha=1, col=1, ...) } \arguments{ \item{x}{object of class \code{kroc} (output from \code{\link{kroc}})} \item{add}{flag to add to current plot. Default is FALSE.} \item{add.roc.ref}{flag to add reference ROC curve. Default is FALSE.} \item{xlab}{x-axis label. Default is "False positive rate (bar(specificity))".} \item{ylab}{y-axis label. Default is "True positive rate (sensitivity)".} \item{alpha,col}{transparency value and colour of line} \item{...}{other graphics parameters used in \code{\link{plot.kde}}.} } \value{ Plots for 1-d and 2-d are sent to graphics window. Plot for 3-d is sent to graphics/RGL window. } \seealso{\code{\link{plot.kde}}} \examples{ data(fgl, package="MASS") x1 <- fgl[fgl[,"type"]=="WinF",c("RI", "Na")] x2 <- fgl[fgl[,"type"]=="Head",c("RI", "Na")] Rhat <- kroc(x1=x1, x2=x2) plot(Rhat, add.roc.ref=TRUE) } \keyword{hplot} ks/man/kfe.Rd0000644000176200001440000000500314336531376012507 0ustar liggesusers\name{kfe} \alias{kfe} \alias{Hpi.kfe} \alias{Hpi.diag.kfe} \alias{hpi.kfe} \title{Kernel functional estimate} \description{ Kernel functional estimate for 1- to 6-dimensional data.} \usage{ kfe(x, G, deriv.order, inc=1, binned, bin.par, bgridsize, deriv.vec=TRUE, add.index=TRUE, verbose=FALSE) Hpi.kfe(x, nstage=2, pilot, pre="sphere", Hstart, binned=FALSE, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") Hpi.diag.kfe(x, nstage=2, pilot, pre="scale", Hstart, binned=FALSE, bgridsize, amise=FALSE, deriv.order=0, verbose=FALSE, optim.fun="optim") hpi.kfe(x, nstage=2, binned=FALSE, bgridsize, amise=FALSE, deriv.order=0) } \arguments{ \item{x}{vector/matrix of data values} \item{nstage}{number of stages in the plug-in bandwidth selector (1 or 2)} \item{pilot}{"dscalar" = single pilot bandwidth (default) \cr "dunconstr" = single unconstrained pilot bandwidth} \item{pre}{"scale" = \code{\link{pre.scale}}, "sphere" = \code{\link{pre.sphere}}} \item{Hstart}{initial bandwidth matrix, used in numerical optimisation} \item{binned}{flag for binned estimation} \item{bgridsize}{vector of binning grid sizes} \item{amise}{flag to return the minimal scaled PI value} \item{deriv.order}{derivative order} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{optim.fun}{optimiser function: one of \code{nlm} or \code{optim}} \item{G}{pilot bandwidth matrix} \item{inc}{0=exclude diagonal, 1=include diagonal terms in kfe calculation} \item{bin.par}{binning parameters - output from \code{\link{binning}}} \item{deriv.vec}{flag to compute duplicated partial derivatives in the vectorised form. Default is FALSE.} \item{add.index}{flag to output derivative indices matrix. Default is true.} } \value{ Plug-in bandwidth matrix for \eqn{r}{r}-th order kernel functional estimator. } \details{ \code{Hpi.kfe} is the optimal plug-in bandwidth for \eqn{r}{r}-th order kernel functional estimator based on the unconstrained pilot selectors of Chacon & Duong (2010). \code{hpi.kfe} is the 1-d equivalent, using the formulas from Wand & Jones (1995, p.70). \code{kfe} does not usually need to be called explicitly by the user. } \references{ Chacon, J.E. & Duong, T. (2010) Multivariate plug-in bandwidth selection with unconstrained pilot matrices. \emph{Test}, \bold{19}, 375-398. Wand, M.P. & Jones, M.C. (1995) \emph{Kernel Smoothing}. Chapman & Hall/CRC, London. } \seealso{\code{\link{kde.test}}} \keyword{smooth} ks/man/plot.kde.part.Rd0000644000176200001440000000625414336531467014441 0ustar liggesusers\name{plot.kde.part} \alias{plot.kde.part} \alias{kms.part} \alias{mvnorm.mixt.part} \title{Partition plot for kernel density clustering} \description{ Plot of partition for kernel density clustering for 2-dimensional data. } \usage{ mvnorm.mixt.part(mus, Sigmas, props=1, xmin, xmax, gridsize, max.iter=100, verbose=FALSE) kms.part(x, H, xmin, xmax, gridsize, verbose=FALSE, ...) \method{plot}{kde.part}(x, display="filled.contour", col, col.fun, alpha=1, add=FALSE, ...) } \arguments{ \item{mus}{(stacked) matrix of mean vectors} \item{Sigmas}{(stacked) matrix of variance matrices} \item{props}{vector of mixing proportions} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{gridsize}{vector of number of grid points} \item{max.iter}{maximum number of iterations} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{x}{matrix of data values or an object of class \code{kde.part}} \item{H}{bandwidth matrix. If missing, \code{Hpi(x,deriv,order=1)} is called by default.} \item{display}{type of display, "filled.contour" for filled contour plot} \item{col,col.fun}{vector of plotting colours or colour function} \item{alpha}{colour transparency. Default is 1.} \item{add}{flag to add to current plot. Default is FALSE.} \item{...}{other parameters} } \value{ A kernel partition is an object of class \code{kde.part} which is a list with fields: \item{x}{data points - same as input} \item{eval.points}{vector or list of points at which the estimate is evaluated} \item{estimate}{density estimate at \code{eval.points}} \item{H}{bandwidth matrix} \item{gridtype}{"linear"} \item{gridded}{flag for estimation on a grid} \item{binned}{flag for binned estimation} \item{names}{variable names} \item{w}{vector of weights} \item{cont}{vector of probability contour levels} \item{end.points}{matrix of final iterates starting from \code{x}} \item{label}{vector of cluster labels} \item{mode}{matrix of cluster modes} \item{nclust}{number of clusters} \item{nclust.table}{frequency table of cluster labels} \item{tol.iter,tol.clust,min.clust.size}{tuning parameter values - same as input} Plot is sent to graphics window. } \details{ For 2-d data, \code{kms.part} and \code{mvnorm.mixt.part} produce a \code{kde.part} object whose values are the class labels, rather than probability density values. } \seealso{\code{\link{plot.kde}}, \code{\link{kms}}} \examples{ ## normal mixture partition mus <- rbind(c(-1,0), c(1, 2/sqrt(3)), c(1,-2/sqrt(3))) Sigmas <- 1/25*rbind(invvech(c(9, 63/10, 49/4)), invvech(c(9,0,49/4)), invvech(c(9,0,49/4))) props <- c(3,3,1)/7 gridsize <- c(11,11) ## small gridsize illustrative purposes only nmixt.part <- mvnorm.mixt.part(mus=mus, Sigmas=Sigmas, props=props, gridsize=gridsize) plot(nmixt.part, asp=1, xlim=c(-3,3), ylim=c(-3,3), alpha=0.5) ## kernel mean shift partition \donttest{set.seed(81928192) x <- rmvnorm.mixt(n=10000, mus=mus, Sigmas=Sigmas, props=props) msize <- round(prod(gridsize)*0.1) kms.nmixt.part <- kms.part(x=x, min.clust.size=msize, gridsize=gridsize) plot(kms.nmixt.part, asp=1, xlim=c(-3,3), ylim=c(-3,3), alpha=0.5)}} \keyword{hplot} ks/man/plot.kfs.Rd0000644000176200001440000000226314340026135013472 0ustar liggesusers\name{plot.kfs} \alias{plot.kfs} \title{Plot for kernel feature significance} \description{ Plot for kernel significant regions for 1- to 3-dimensional data. } \usage{ \method{plot}{kfs}(x, display="filled.contour", col=7, colors, abs.cont, alpha=1, alphavec=0.4, add=FALSE, ...) } \arguments{ \item{x}{object of class \code{kfs} (output from \code{\link{kfs}})} \item{display}{type of display, "slice" for contour plot, "persp" for perspective plot, "image" for image plot, "filled.contour" for filled contour plot (2-d); "plot3D", "rgl" (3-d)} \item{col,colors}{colour for contour region} \item{abs.cont}{absolute contour height. Default is 0.5.} \item{alpha}{transparency value for contour (2-d)} \item{alphavec}{vector of transparency values for contour (3-d)} \item{add}{flag to add to current plot. Default is FALSE.} \item{...}{other graphics parameters used in \code{\link{plot.kde}}} } \value{ Plots for 1-d and 2-d are sent to graphics window. Plot for 3-d is sent to graphics/RGL window. } \seealso{\code{\link{plot.kde}}} \examples{ \donttest{data(geyser, package="MASS") geyser.fs <- kfs(geyser, binned=TRUE) plot(geyser.fs) }} \keyword{hplot} ks/man/worldbank.Rd0000644000176200001440000000200014336531573013716 0ustar liggesusers\name{worldbank} \docType{data} \alias{worldbank} \title{Development indicators from the World Bank Group} \description{ This data set contains six development indicators for national entities for the year 2011, which is the latest year for which they are consistently available. } \usage{data(worldbank)} \format{A matrix with 7 columns and 218 rows. Each row corresponds to a country. The first column is the country, the second is the per capita carbon dioxide emissions (thousands Kg), the third is the per capita GDP (thousands of current USD), the fourth is the annual GDP growth rate (\%), the fifth is the annual inflation rate (\%), the sixth is the percentage of internet users in the population (\%), the seventh is the added value agricultural production as a ratio of the total GDP (\%). } \source{ World Bank Group (2016) World development indicators. \url{http://databank.worldbank.org/data/reports.aspx? source=world-development-indicators}. Accessed 2016-10-03. } \keyword{datasets} ks/man/vkde.Rd0000644000176200001440000000601714336531563012677 0ustar liggesusers\name{vkde} \alias{kde.balloon} \alias{kde.sp} \title{Variable kernel density estimate.} \description{ Variable kernel density estimate for 2-dimensional data. } \usage{ kde.balloon(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, verbose=FALSE) kde.sp(x, H, h, gridsize, gridtype, xmin, xmax, supp=3.7, eval.points, binned, bgridsize, w, compute.cont=TRUE, approx.cont=TRUE, verbose=FALSE) } \arguments{ \item{x}{matrix of data values} \item{H}{bandwidth matrix. If this missing, \code{Hns} is called by default.} \item{h}{not yet implemented} \item{gridsize}{vector of number of grid points} \item{gridtype}{not yet implemented} \item{xmin,xmax}{vector of minimum/maximum values for grid} \item{supp}{effective support for standard normal} \item{eval.points}{vector or matrix of points at which estimate is evaluated} \item{binned}{flag for binned estimation.} \item{bgridsize}{vector of binning grid sizes} \item{w}{vector of weights. Default is a vector of all ones.} \item{compute.cont}{flag for computing 1\% to 99\% probability contour levels. Default is TRUE.} \item{approx.cont}{flag for computing approximate probability contour levels. Default is TRUE.} \item{verbose}{flag to print out progress information. Default is FALSE.} } \value{ A variable kernel density estimate for bounded data is an object of class \code{kde}. } \details{ The balloon density estimate \code{kde.balloon} employs bandwidths which vary at each estimation point (Loftsgaarden & Quesenberry, 1965). There are as many bandwidths as there are estimation grid points. The default bandwidth is \code{Hns(,deriv.order=2)} and the subsequent bandwidths are derived via a minimal MSE formula. The sample point density estimate \code{kde.sp} employs bandwidths which vary for each data point (Abramson, 1982). There are as many bandwidths as there are data points. The default bandwidth is \code{Hns(,deriv.order=4)} and the subsequent bandwidths are derived via the Abramson formula. } \references{ Abramson, I. S. (1982) On bandwidth variation in kernel estimates - a square root law. \emph{Annals of Statistics}, \bold{10}, 1217-1223. Loftsgaarden, D. O. & Quesenberry, C. P. (1965) A nonparametric estimate of a multivariate density function. \emph{Annals of Mathematical Statistics}, \bold{36}, 1049-1051. } \seealso{\code{\link{kde}}, \code{\link{plot.kde}}} \examples{ \donttest{data(worldbank) wb <- as.matrix(na.omit(worldbank[,4:5])) xmin <- c(-70,-35); xmax <- c(35,70) fhat <- kde(x=wb, xmin=xmin, xmax=xmax) fhat.sp <- kde.sp(x=wb, xmin=xmin, xmax=xmax) zmax <- max(fhat.sp$estimate) plot(fhat, display="persp", box=TRUE, phi=20, thin=1, border=grey(0,0.2), zlim=c(0,zmax)) plot(fhat.sp, display="persp", box=TRUE, phi=20, thin=1, border=grey(0,0.2), zlim=c(0,zmax))} \dontrun{ fhat.ball <- kde.balloon(x=wb, xmin=xmin, xmax=xmax) plot(fhat.ball, display="persp", box=TRUE, phi=20, zlim=c(0,zmax))} } \keyword{smooth} ks/man/mixt.Rd0000644000176200001440000000423314336531435012723 0ustar liggesusers\name{mixt} \alias{rnorm.mixt} \alias{dnorm.mixt} \alias{rmvnorm.mixt} \alias{dmvnorm.mixt} \alias{rmvt.mixt} \alias{dmvt.mixt} \alias{mvnorm.mixt.mode} \title{Normal and t-mixture distributions} \description{ Random generation and density values from normal and t-mixture distributions.} \usage{ dnorm.mixt(x, mus=0, sigmas=1, props=1) rnorm.mixt(n=100, mus=0, sigmas=1, props=1, mixt.label=FALSE) dmvnorm.mixt(x, mus, Sigmas, props=1, verbose=FALSE) rmvnorm.mixt(n=100, mus=c(0,0), Sigmas=diag(2), props=1, mixt.label=FALSE) rmvt.mixt(n=100, mus=c(0,0), Sigmas=diag(2), dfs=7, props=1) dmvt.mixt(x, mus, Sigmas, dfs, props) mvnorm.mixt.mode(mus, Sigmas, props=1, verbose=FALSE) } \arguments{ \item{n}{number of random variates} \item{x}{matrix of quantiles} \item{mus}{(stacked) matrix of mean vectors (>1-d) or vector of means (1-d)} \item{Sigmas}{(stacked) matrix of variance matrices (>1-d)} \item{sigmas}{vector of standard deviations (1-d)} \item{props}{vector of mixing proportions} \item{mixt.label}{flag to output numeric label indicating mixture component. Default is FALSE.} \item{verbose}{flag to print out progress information. Default is FALSE.} \item{dfs}{vector of degrees of freedom} } \value{Normal and t-mixture random vectors and density values.} \details{ \code{rmvnorm.mixt} and \code{dmvnorm.mixt} are based on the \code{rmvnorm} and \code{dmvnorm} functions from the \pkg{mvtnorm} package. Likewise for \code{rmvt.mixt} and \code{dmvt.mixt}. For the normal mixture densities, \code{mvnorm.mixt.mode} computes the local modes: these are usually very close but not exactly equal to the component means. } \examples{ ## univariate normal mixture x <- rnorm.mixt(1000, mus=c(-1,1), sigmas=c(0.5, 0.5), props=c(1/2, 1/2)) ## bivariate mixtures mus <- rbind(c(-1,0), c(1, 2/sqrt(3)), c(1,-2/sqrt(3))) Sigmas <- 1/25*rbind(invvech(c(9, 63/10, 49/4)), invvech(c(9,0,49/4)), invvech(c(9,0,49/4))) props <- c(3,3,1)/7 dfs <- c(7,3,2) x <- rmvnorm.mixt(1000, mus=mus, Sigmas=Sigmas, props=props) y <- rmvt.mixt(1000, mus=mus, Sigmas=Sigmas, dfs=dfs, props=props) mvnorm.mixt.mode(mus=mus, Sigmas=Sigmas, props=props) } \keyword{distribution} ks/man/plot.kdde.Rd0000644000176200001440000000442114336531461013624 0ustar liggesusers\name{plot.kdde} \alias{plot.kdde} \title{Plot for kernel density derivative estimate} \description{ Plot for kernel density derivative estimate for 1- to 3-dimensional data. } \usage{ \method{plot}{kdde}(x, ...) } \arguments{ \item{x}{object of class \code{kdde} (output from \code{\link{kdde}})} \item{...}{other graphics parameters: \describe{ \item{\code{which.deriv.ind}}{index of the partial derivative to be plotted (>1-d)} } and those used in \code{\link{plot.kde}}} } \value{ Plots for 1-d and 2-d are sent to graphics window. Plot for 3-d is sent to graphics/RGL window. In addition to the display options inherited from \code{plot.kde}, the first derivative has \code{display="quiver"}. This is a quiver plot where the size and direction of the arrow indicates the magnitude/direction of the density gradient. See \code{quiver} from the \pkg{pracma} package for more details. } \details{ For \code{kdde} objects, the function headers for the different dimensional data are \preformatted{ ## univariate plot(fhat, ylab="Density derivative function", cont=50, abs.cont, alpha=1, ...) ## bivariate plot(fhat, which.deriv.ind=1, cont=c(25,50,75), abs.cont, display="slice", zlab="Density derivative function", col, col.fun, alpha=1, kdde.flag=TRUE, thin=3, transf=1, neg.grad=FALSE, ...) ## trivariate plot(fhat, which.deriv.ind=1, display="plot3D", cont=c(25,50,75), abs.cont, colors, col, col.fun, ...)} } \seealso{\code{\link{plot.kde}}} \examples{ ## univariate example data(tempb) fhat1 <- kdde(x=tempb[,"tmin"], deriv.order=1) ## gradient [df/dx, df/dy] plot(fhat1, xlab="Min. temp.", col.cont=4) ## df/dx points(20,predict(fhat1, x=20)) ## bivariate example fhat1 <- kdde(x=tempb[,c("tmin", "tmax")], deriv.order=1) plot(fhat1, display="quiver") ## gradient [df/dx, df/dy] \donttest{fhat2 <- kdde(x=tempb[,c("tmin", "tmax")], deriv.order=2) plot(fhat2, which.deriv.ind=2, display="persp", phi=10) plot(fhat2, which.deriv.ind=2, display="filled.contour") ## d^2 f/(dx dy): blue=-ve, red=+ve s2 <- kcurv(fhat2) plot(s2, display="filled.contour", alpha=0.5, lwd=1) ## summary curvature ## trivariate example data(iris) fhat1 <- kdde(iris[,2:4], deriv.order=1) plot(fhat1)}} \keyword{hplot} ks/man/plot.kde.Rd0000644000176200001440000001327314340026065013457 0ustar liggesusers\name{plot.kde} \alias{plot.kde} \title{Plot for kernel density estimate} \description{ Plot for kernel density estimate for 1- to 3-dimensional data. } \usage{ \method{plot}{kde}(x, ...) } \arguments{ \item{x}{object of class \code{kde} (output from \code{\link{kde}})} \item{...}{other graphics parameters: \describe{ \item{\code{display}}{type of display, "slice" for contour plot, "persp" for perspective plot, "image" for image plot, "filled.contour" for filled contour plot (2-d); "plot3D", "rgl" (3-d)} \item{\code{cont}}{vector of percentages for contour level curves} \item{\code{abs.cont}}{vector of absolute density estimate heights for contour level curves} \item{\code{approx.cont}}{flag to compute approximate contour levels. Default is FALSE.} \item{\code{col}}{plotting colour for density estimate (1-d, 2-d)} \item{\code{col.cont}}{plotting colour for contours} \item{\code{col.fun}}{plotting colour function for contours} \item{\code{col.pt}}{plotting colour for data points} \item{\code{colors}}{vector of colours for each contour (3-d)} \item{\code{jitter}}{flag to jitter rug plot (1-d). Default is TRUE.} \item{\code{lwd.fc}}{line width for filled contours (2-d)} \item{\code{xlim,ylim,zlim}}{axes limits} \item{\code{xlab,ylab,zlab}}{axes labels} \item{\code{add}}{flag to add to current plot. Default is FALSE.} \item{\code{theta,phi,d,border}}{graphics parameters for perspective plots (2-d)} \item{\code{drawpoints}}{flag to draw data points on density estimate. Default is FALSE.} \item{\code{drawlabels}}{flag to draw contour labels (2-d). Default is TRUE.} \item{\code{alpha}}{transparency value of plotting symbol} \item{\code{alphavec}}{vector of transparency values for contours (3-d)} \item{\code{size}}{size of plotting symbol (3-d).} } } } \value{ Plots for 1-d and 2-d are sent to graphics window. Plot for 3-d is sent to graphics/RGL window. } \details{ For \code{kde} objects, the function headers for the different dimensional data are \preformatted{ ## univariate plot(fhat, xlab, ylab="Density function", add=FALSE, drawpoints=FALSE, col=1, col.pt=4, col.cont=1, cont.lwd=1, jitter=FALSE, cont, abs.cont, approx.cont=TRUE, alpha=1, ...) ## bivariate plot(fhat, display="slice", cont=c(25,50,75), abs.cont, approx.cont=TRUE, xlab, ylab, zlab="Density function", cex=1, pch=1, add=FALSE, drawpoints=FALSE, drawlabels=TRUE, theta=-30, phi=40, d=4, col.pt=4, col, col.fun, alpha=1, lwd=1, border=1, thin=3, kdde.flag=FALSE, ticktype="detailed", ...) ## trivariate plot(fhat, display="plot3D", cont=c(25,50,75), abs.cont, approx.cont=TRUE, colors, col, col.fun, alphavec, size=3, cex=1, pch=1, theta=-30, phi=40, d=4, ticktype="detailed", bty="f", col.pt=4, add=FALSE, xlab, ylab, zlab, drawpoints=FALSE, alpha, box=TRUE, axes=TRUE, ...)} For 1-dimensional data, the plot is a standard plot of a 1-d curve. If \code{drawpoints=TRUE} then a rug plot is added. If \code{cont} is specified, the horizontal line on the x-axis indicates the \code{cont}\% highest density level set. For 2-dimensional data, the different types of plotting displays are controlled by the \code{display} parameter. (a) If \code{display="slice"} then a slice/contour plot is generated using \code{contour}. (b) If \code{display} is \code{"filled.contour"} then a filled contour plot is generated. The default contours are at 25\%, 50\%, 75\% or \code{cont=c(25,50,75)} which are upper percentages of highest density regions. (c) If \code{display="persp"} then a perspective/wire-frame plot is generated. The default z-axis limits \code{zlim} are the default from the usual \code{persp} command. (d) If \code{display="image"} then an image plot is generated. For 3-dimensional data, the plot is a series of nested 3-d contours. The default contours are \code{cont=c(25,50,75)}. The default opacity \code{alphavec} ranges from 0.1 to 0.5. For \pkg{ks} \eqn{\geq}{>=} 1.12.0, base R graphics becomes the default plotting engine: to create an \pkg{rgl} plot like in previous versions, set \code{display="rgl"}. To specify contours, either one of \code{cont} or \code{abs.cont} is required. \code{cont} specifies upper percentages which correspond to probability contour regions. If \code{abs.cont} is set to particular values, then contours at these levels are drawn. This second option is useful for plotting multiple density estimates with common contour levels. See \code{\link{contourLevels}} for details on computing contour levels. If \code{approx=FALSE}, then the exact KDE is computed. Otherwise it is interpolated from an existing KDE grid, which can dramatically reduce computation time for large data sets. If a colour function is specified in \code{col.fun}, it should have the number of colours as a single argument, e.g. \code{function(n)\{hcl.colors(n, ...)\}}. The transparent background colour is automatically concatenated before this colour function. If \code{col} is specified, it overrides \code{col.fun}. There should be one more colour than the number of contours, i.e. background colour plus one for each contour. } \examples{ data(iris) ## univariate example fhat <- kde(x=iris[,2]) plot(fhat, cont=50, col.cont=4, cont.lwd=2, xlab="Sepal length") ## bivariate example fhat <- kde(x=iris[,2:3]) plot(fhat, display="filled.contour", cont=seq(10,90,by=10), lwd=1, alpha=0.5) plot(fhat, display="persp", border=1, alpha=0.5) ## trivariate example fhat <- kde(x=iris[,2:4]) plot(fhat) if (interactive()) plot(fhat, display="rgl") } \keyword{hplot} ks/man/kde.truncate.Rd0000644000176200001440000000277214336531364014340 0ustar liggesusers\name{kde.truncate} \alias{kde.truncate} \alias{kdde.truncate} \title{Truncated kernel density derivative estimate} \description{ Truncated kernel density derivative estimate for 2-dimensional data. } \usage{ kde.truncate(fhat, boundary) kdde.truncate(fhat, boundary) } \arguments{ \item{fhat}{object of class \code{kde} or \code{kdde}} \item{boundary}{two column matrix delimiting the boundary for truncation} } \value{ A truncated kernel density (derivative) estimate inherits the same object class as the input estimate. } \details{ A simple truncation is performed on the kernel estimator. All the points in the estimation grid which are outside of the regions delimited by \code{boundary} are set to 0, and their probability mass is distributed proportionally to the remaining density (derivative) values. } \seealso{\code{\link{kde}}, \code{\link{kdde}}} \examples{ data(worldbank) wb <- as.matrix(na.omit(worldbank[,c("internet", "ag.value")])) fhat <- kde(x=wb) rectb <- cbind(x=c(0,100,100,0,0), y=c(0,0,100,100,0)) fhat.b <- kde.truncate(fhat, boundary=rectb) plot(fhat, col=1, xlim=c(0,100), ylim=c(0,100)) plot(fhat.b, add=TRUE, col=4) rect(0,0,100,100, lty=2) library(oz) data(grevillea) wa.coast <- ozRegion(section=1) wa.polygon <- cbind(wa.coast$lines[[1]]$x, wa.coast$lines[[1]]$y) fhat1 <- kdde(x=grevillea, deriv.order=1) fhat1 <- kdde.truncate(fhat1, wa.polygon) oz(section=1, xlim=c(113,122), ylim=c(-36,-29)) plot(fhat1, add=TRUE, display="filled.contour") } \keyword{smooth} ks/man/ks-package.Rd0000644000176200001440000003542614550057756013767 0ustar liggesusers\name{ks-package} \alias{ks} \alias{ks-package} \docType{package} \title{ks} \description{ Kernel smoothing for data from 1- to 6-dimensions. } \details{ There are three main types of functions in this package: \itemize{ \item computing kernel estimators - these function names begin with `k' \item computing bandwidth selectors - these begin with `h' (1-d) or `H' (>1-d) \item displaying kernel estimators - these begin with `plot'. } The kernel used throughout is the normal (Gaussian) kernel \eqn{K}{K}. For 1-d data, the bandwidth \eqn{h}{h} is the standard deviation of the normal kernel, whereas for multivariate data, the bandwidth matrix \eqn{\bold{{\rm H}}}{H} is the variance matrix. --For kernel density estimation, \code{\link{kde}} computes \deqn{\hat{f}(\bold{x}) = n^{-1} \sum_{i=1}^n K_{\bold{{\rm H}}} (\bold{x} - \bold{X}_i).}{hat(f)(x) = n^(-1) sum_i K_H (x - X_i).} The bandwidth matrix \eqn{\bold{{\rm H}}}{H} is a matrix of smoothing parameters and its choice is crucial for the performance of kernel estimators. For display, its \code{plot} method calls \code{\link{plot.kde}}. --For kernel density estimation, there are several varieties of bandwidth selectors \itemize{ \item plug-in \code{\link{hpi}} (1-d); \code{\link{Hpi}}, \code{\link{Hpi.diag}} (2- to 6-d) \item least squares (or unbiased) cross validation (LSCV or UCV) \code{\link{hlscv}} (1-d); \code{\link{Hlscv}}, \code{\link{Hlscv.diag}} (2- to 6-d) \item biased cross validation (BCV) \code{\link{Hbcv}}, \code{\link{Hbcv.diag}} (2- to 6-d) \item smoothed cross validation (SCV) \code{\link{hscv}} (1-d); \code{\link{Hscv}}, \code{\link{Hscv.diag}} (2- to 6-d) \item normal scale \code{\link{hns}} (1-d); \code{\link{Hns}} (2- to 6-d). } --For kernel density support estimation, the main function is \code{\link{ksupp}} which is (the convex hull of) \deqn{\{\bold{x}: \hat{f}(\bold{x}) > \tau\}}{\{x: hat(f) > tau\}} for a suitable level \eqn{\tau}{tau}. This is closely related to the \eqn{\tau}{tau}-level set of \eqn{\hat{f}}{hat(f)}. --For truncated kernel density estimation, the main function is \code{\link{kde.truncate}} \deqn{\hat{f} (\bold{x}) \bold{1}\{\bold{x} \in \Omega\} / \int_{\Omega}\hat{f} (\bold{x}) \, d\bold{x}}{hat(f)(x) 1\{x in Omega\}/int hat(f) 1\{x in Omega\}} for a bounded data support \eqn{\Omega}{Omega}. The standard density estimate \eqn{\hat{f}}{hat(f)} is truncated and rescaled to give unit integral over \eqn{\Omega}{Omega}. Its \code{plot} method calls \code{\link{plot.kde}}. --For boundary kernel density estimation where the kernel function is modified explicitly in the boundary region, the main function is \code{\link{kde.boundary}} \deqn{ n^{-1} \sum_{i=1}^n K^*_{\bold{{\rm H}}} (\bold{x} - \bold{X}_i)}{hat(f)(x) = n^(-1) sum_i K*_H (x - X_i)} for a boundary kernel \eqn{K^*}{K*}. Its \code{plot} method calls \code{\link{plot.kde}}. --For variable kernel density estimation where the bandwidth is not a constant matrix, the main functions are \code{\link{kde.balloon}} \deqn{\hat{f}_{\rm ball}(\bold{x}) = n^{-1} \sum_{i=1}^n K_{\bold{{\rm H}}(\bold{x})} (\bold{x} - \bold{X}_i)}{hat(f)_ball(x) = n^(-1) sum_i K_H(x) (x - X_i)} and \code{\link{kde.sp}} \deqn{\hat{f}_{\rm SP}(\bold{x}) = n^{-1} \sum_{i=1}^n K_{\bold{{\rm H}}(\bold{X}_i)} (\bold{x} - \bold{X}_i).}{hat(f)_SP(x) = n^(-1) sum_i K_H(X_i) (x - X_i).} For the balloon estimation \eqn{\hat{f}_{\rm ball}}{hat(f)_ball} the bandwidth varies with the estimation point \eqn{\bold{x}}{x}, whereas for the sample point estimation \eqn{\hat{f}_{\rm SP}}{hat(f)_SP} the bandwidth varies with the data point \eqn{\bold{X}_i, i=1,\dots,n}{X_i, i=1, ..., n}. Their \code{plot} methods call \code{\link{plot.kde}}. The bandwidth selectors for \code{kde.balloon} are based on the normal scale bandwidth \code{Hns(,deriv.order=2)} via the MSE minimal formula, and for \code{kde.SP} on \code{Hns(,deriv.order=4)} via the Abramson formula. --For kernel density derivative estimation, the main function is \code{\link{kdde}} \deqn{{\sf D}^{\otimes r}\hat{f}(\bold{x}) = n^{-1} \sum_{i=1}^n {\sf D}^{\otimes r}K_{\bold{{\rm H}}} (\bold{x} - \bold{X}_i).}{hat(f)^(r)(x) = n^(-1) sum_i D^r K_H (x - X_i).} The bandwidth selectors are a modified subset of those for \code{\link{kde}}, i.e. \code{\link{Hlscv}}, \code{\link{Hns}}, \code{\link{Hpi}}, \code{\link{Hscv}} with \code{deriv.order>0}. Its \code{plot} method is \code{\link{plot.kdde}} for plotting each partial derivative singly. --For kernel summary curvature estimation, the main function is \code{\link{kcurv}} \deqn{\hat{s}(\bold{x})= - \bold{1}\{{\sf D}^2 \hat{f}(\bold{x}) < 0\} \mathrm{abs}(|{\sf D}^2 \hat{f}(\bold{x})|)}{hat(s)(x) = -1\{D^2 hat(f)(x) <0)*abs(det(D^2 hat(f)(x)))\}} where \eqn{{\sf D}^2 \hat{f}(\bold{x})}{D^2 hat(f)(x)} is the kernel Hessian matrix estimate. It has the same structure as a kernel density estimate so its \code{plot} method calls \code{\link{plot.kde}}. --For kernel discriminant analysis, the main function is \code{\link{kda}} which computes density estimates for each the groups in the training data, and the discriminant surface. Its \code{plot} method is \code{\link{plot.kda}}. The wrapper function \code{\link{hkda}}, \code{\link{Hkda}} computes bandwidths for each group in the training data for \code{kde}, e.g. \code{hpi}, \code{Hpi}. --For kernel functional estimation, the main function is \code{kfe} which computes the \eqn{r}{r}-th order integrated density functional \deqn{\hat{{\bold \psi}}_r = n^{-2} \sum_{i=1}^n \sum_{j=1}^n {\sf D}^{\otimes r}K_{\bold{{\rm H}}}(\bold{X}_i-\bold{X}_j).}{hat(psi)_r = n^(-2) sum_i sum_j D^r K_H (X_i - X_j).} The plug-in selectors are \code{\link{hpi.kfe}} (1-d), \code{\link{Hpi.kfe}} (2- to 6-d). Kernel functional estimates are usually not required to computed directly by the user, but only within other functions in the package. --For kernel-based 2-sample testing, the main function is \code{\link{kde.test}} which computes the integrated \eqn{L_2}{L2} distance between the two density estimates as the test statistic, comprising a linear combination of 0-th order kernel functional estimates: \deqn{\hat{T} = \hat{\psi}_{0,1} + \hat{\psi}_{0,2} - (\hat{\psi}_{0,12} + \hat{\psi}_{0,21}),}{hat(T) = hat(psi)_0,1 + hat(psi)_0,2 - (hat(psi)_0,12 + hat(psi)_0,21),} and the corresponding p-value. The \eqn{\psi}{psi} are zero order kernel functional estimates with the subscripts indicating that 1 = sample 1 only, 2 = sample 2 only, and 12, 21 = samples 1 and 2. The bandwidth selectors are \code{\link{hpi.kfe}}, \code{\link{Hpi.kfe}} with \code{deriv.order=0}. --For kernel-based local 2-sample testing, the main function is \code{\link{kde.local.test}} which computes the squared distance between the two density estimates as the test statistic \deqn{\hat{U}(\bold{x}) = [\hat{f}_1(\bold{x}) - \hat{f}_2(\bold{x})]^2}{hat(U)(x) = [hat(f)_1(x) - hat(f)_2(x)]^2} and the corresponding local p-values. The bandwidth selectors are those used with \code{\link{kde}}, e.g. \code{\link{hpi}, \link{Hpi}}. --For kernel cumulative distribution function estimation, the main function is \code{\link{kcde}} \deqn{\hat{F}(\bold{x}) = n^{-1} \sum_{i=1}^n \mathcal{K}_{\bold{{\rm H}}} (\bold{x} - \bold{X}_i)}{hat(F)(x) = n^(-1) sum_i intK_H (x - X_i)} where \eqn{\mathcal{K}}{intK} is the integrated kernel. The bandwidth selectors are \code{\link{hpi.kcde}}, \code{\link{Hpi.kcde}}. Its \code{plot} method is \code{\link{plot.kcde}}. There exist analogous functions for the survival function \eqn{\hat{\bar{F}}}{hat(bar(F))}. --For kernel estimation of a ROC (receiver operating characteristic) curve to compare two samples from \eqn{\hat{F}_1, \hat{F}_2}{hat(F)_1, hat(F)_2}, the main function is \code{\link{kroc}} \deqn{\{\hat{F}_{\hat{Y}_1}(z), \hat{F}_{\hat{Y}_2}(z)\}}{\{hat(F)_hat(Y1))(z), hat(F_hat(Y2))(z)\}} based on the cumulative distribution functions of \eqn{\hat{Y}_j = \hat{\bar{F}}_1(\bold{X}_j), j=1,2}{hat(Yj)=hat(bar(F))_1(X_j), j=1,2}. The bandwidth selectors are those used with \code{\link{kcde}}, e.g. \code{\link{hpi.kcde}, \link{Hpi.kcde}} for \eqn{\hat{F}_{\hat{Y}_j}, \hat{\bar{F}}_1}{hat(F)_hat(Yj), hat(bar(F))_1}. Its \code{plot} method is \code{\link{plot.kroc}}. --For kernel estimation of a copula, the main function is \code{\link{kcopula}} \deqn{\hat{C}(\bold{z}) = \hat{F}(\hat{F}_1^{-1}(z_1), \dots, \hat{F}_d^{-1}(z_d))}{hat(C)(z) = hat(F)(hat(F)_1^(-1)(z_1),..., hat(F)_d^(-1)(z_d))} where \eqn{\hat{F}_j^{-1}(z_j)}{hat(F)_j^(-1)(z_j)} is the \eqn{z_j}{z_j}-th quantile of of the \eqn{j}{j}-th marginal distribution \eqn{\hat{F}_j}{hat(F_j)}. The bandwidth selectors are those used with \code{\link{kcde}} for \eqn{\hat{F}, \hat{F}_j}{hat(F), hat(F)_j}. Its \code{plot} method is \code{\link{plot.kcde}}. %--For kernel estimation of a copula density, the %main function is \code{\link{kcopula.de}} %\deqn{\hat{c}(\bold{z}) = n^{-1} \sum_{i=1}^n %K_{\bold{{\rm H}}} (\bold{z} - \hat{\bold{Z}}_i)}{hat(c)(z) = %hat(f)(z) = n^(-1) sum_i K_H (z - hat(Z)_i)} %where \eqn{\hat{\bold{Z}}_i = (\hat{F}_1(X_{i1}), \dots, % \hat{F}_d(X_{id}))}{hat(Z)_i = (hat(F)_1(X_i1), \dots, hat(F)_d(X_id))}. %The bandwidth selectors are those used with \code{\link{kde}} for %\eqn{\hat{c}}{hat(c)} and \code{\link{kcde}} for \eqn{\hat{F}_j}{hat(F)_j}. %Its \code{plot} method is \code{\link{plot.kde}}. --For kernel mean shift clustering, the main function is \code{\link{kms}}. The mean shift recurrence relation of the candidate point \eqn{{\bold x}}{x} \deqn{{\bold x}_{j+1} = {\bold x}_j + \bold{{\rm H}} {\sf D} \hat{f}({\bold x}_j)/\hat{f}({\bold x}_j),}{x_j+1 = x_j + H D hat(f)(x_j)/hat(f)(x_j),} where \eqn{j\geq 0}{j>=0} and \eqn{{\bold x}_0 = {\bold x}}{x_0 = x}, is iterated until \eqn{{\bold x}}{x} converges to its local mode in the density estimate \eqn{\hat{f}}{hat(f)} by following the density gradient ascent paths. This mode determines the cluster label for \eqn{\bold{x}}{x}. The bandwidth selectors are those used with \code{\link{kdde}(,deriv.order=1)}. --For kernel density ridge estimation, the main function is \code{\link{kdr}}. The kernel density ridge recurrence relation of the candidate point \eqn{{\bold x}}{x} \deqn{{\bold x}_{j+1} = {\bold x}_j + \bold{{\rm U}}_{(d-1)}({\bold x}_j)\bold{{\rm U}}_{(d-1)}({\bold x}_j)^T \bold{{\rm H}} {\sf D} \hat{f}({\bold x}_j)/\hat{f}({\bold x}_j),}{x_j+1 = x_j + U_(d-1)(x_j) U_(d-1)(x_j)^T H D hat(f)(x_j)/hat(f)(x_j),} where \eqn{j\geq 0}{j>=0}, \eqn{{\bold x}_0 = {\bold x}}{x_0 = x} and \eqn{\bold{{\rm U}}_{(d-1)}}{U_(d-1)} is the 1-dimensional projected density gradient, is iterated until \eqn{{\bold x}}{x} converges to the ridge in the density estimate. The bandwidth selectors are those used with \code{\link{kdde}(,deriv.order=2)}. -- For kernel feature significance, the main function \code{\link{kfs}}. The hypothesis test at a point \eqn{\bold{x}}{x} is \eqn{H_0(\bold{x}): \mathsf{H} f(\bold{x}) < 0}{H0(x): H f(x) < 0}, i.e. the density Hessian matrix \eqn{\mathsf{H} f(\bold{x})}{H f(x)} is negative definite. The test statistic is \deqn{W(\bold{x}) = \Vert \mathbf{S}(\bold{x})^{-1/2} \mathrm{vech} \ \mathsf{H} \hat{f} (\bold{x})\Vert ^2}{% W(x) = ||S(x)^(-1/2) vech H hat{f}(x)||^2} where \eqn{{\sf H}\hat{f}}{H hat{f}} is the Hessian estimate, vech is the vector-half operator, and \eqn{\mathbf{S}}{S} is an estimate of the null variance. \eqn{W(\bold{x})}{W(x)} is approximately \eqn{\chi^2}{chi-squared} distributed with \eqn{d(d+1)/2}{d(d+1)/2} degrees of freedom. If \eqn{H_0(\bold{x})}{H0(x)} is rejected, then \eqn{\bold{x}}{x} belongs to a significant modal region. The bandwidth selectors are those used with \code{\link{kdde}(,deriv.order=2)}. Its \code{plot} method is \code{\link{plot.kfs}}. --For deconvolution density estimation, the main function is \code{\link{kdcde}}. A weighted kernel density estimation with the contaminated data \eqn{{\bold W}_1, \dots, {\bold W}_n}{W_1, ..., W_n}, \deqn{\hat{f}_w({\bold x}) = n^{-1} \sum_{i=1}^n \alpha_i K_{\bold{{\rm H}}}({\bold x} - {\bold W}_i),}{hat(f)(x) = n^(-1) sum_i alpha_i K_H (x - W_i),} is utilised, where the weights \eqn{\alpha_1, \dots, \alpha_n}{alpha_1, ..., alpha_n} are chosen via a quadratic optimisation involving the error variance and the regularisation parameter. The bandwidth selectors are those used with \code{\link{kde}}. --Binned kernel estimation is an approximation to the exact kernel estimation and is available for d=1, 2, 3, 4. This makes kernel estimators feasible for large samples. --For an overview of this package with 2-d density estimation, see \code{vignette("kde")}. --For \pkg{ks} \eqn{\geq}{>=} 1.11.1, the \pkg{misc3d} and \pkg{rgl} (3-d plot), \pkg{oz} (Australian map) packages, and for \pkg{ks} \eqn{\geq}{>=} 1.14.2, the \pkg{plot3D} (3-d plot) package, have been moved from Depends to Suggests. This was done to allow \pkg{ks} to be installed on systems where these latter graphical-based packages can't be installed. Furthermore, since the future of OpenGL in R is not certain, \pkg{plot3D} becomes the default for 3D plotting for \pkg{ks} \eqn{\geq}{>=} 1.12.0. RGL plots are still supported though these may be deprecated in the future. } \author{ Tarn Duong for most of the package. M. P. Wand for the binned estimation, univariate plug-in selector and univariate density derivative estimator code. J. E. Chacon for the unconstrained pilot functional estimation and fast implementation of derivative-based estimation code. A. and J. Gramacki for the binned estimation for unconstrained bandwidth matrices. } \references{ Bowman, A. & Azzalini, A. (1997) \emph{Applied Smoothing Techniques for Data Analysis}. Oxford University Press, Oxford. Chacon, J.E. & Duong, T. (2018) \emph{Multivariate Kernel Smoothing and Its Applications}. Chapman & Hall/CRC, Boca Raton. Duong, T. (2004) \emph{Bandwidth Matrices for Multivariate Kernel Density Estimation.} Ph.D. Thesis, University of Western Australia. Scott, D.W. (2015) \emph{Multivariate Density Estimation: Theory, Practice, and Visualization (2nd edn)}. John Wiley & Sons, New York. Silverman, B. (1986) \emph{Density Estimation for Statistics and Data Analysis}. Chapman & Hall/CRC, London. Simonoff, J. S. (1996) \emph{Smoothing Methods in Statistics}. Springer-Verlag, New York. Wand, M.P. & Jones, M.C. (1995) \emph{Kernel Smoothing}. Chapman & Hall/CRC, London. } \keyword{package} \seealso{\pkg{feature}, \pkg{sm}, \pkg{KernSmooth}} ks/DESCRIPTION0000644000176200001440000000267014673325021012405 0ustar liggesusersPackage: ks Version: 1.14.3 Date: 2024-09-20 Title: Kernel Smoothing Authors@R: c(person("Tarn", "Duong", role=c("aut","cre"), email="tarn.duong@gmail.com", comment=c(ORCID="0000-0002-1198-3482")), person("Matt", "Wand", role="ctb", comment=c(ORCID="0000-0003-2555-896X")), person("Jose", "Chacon", role="ctb"), person("Artur", "Gramacki", role="ctb", comment=c(ORCID="0000-0002-1610-9743"))) Maintainer: Tarn Duong Depends: R (>= 2.10.0) Imports: FNN (>= 1.1), kernlab, KernSmooth (>= 2.22), Matrix, mclust, mgcv, multicool, mvtnorm (>= 1.0-0), pracma Suggests: geometry, MASS, misc3d (>= 0.4-0), oz, plot3D, rgl (>= 0.66) Description: Kernel smoothers for univariate and multivariate data, with comprehensive visualisation and bandwidth selection capabilities, including for densities, density derivatives, cumulative distributions, clustering, classification, density ridges, significant modal regions, and two-sample hypothesis tests. Chacon & Duong (2018) . License: GPL-2 | GPL-3 URL: https://www.mvstat.net/mvksa/ NeedsCompilation: yes Packaged: 2024-09-20 13:32:31 UTC; tduong Author: Tarn Duong [aut, cre] (), Matt Wand [ctb] (), Jose Chacon [ctb], Artur Gramacki [ctb] () Repository: CRAN Date/Publication: 2024-09-20 17:00:01 UTC