pcaMethods/COPYING0000644000175400017540000004311013556116437014734 0ustar00biocbuildbiocbuild GNU GENERAL PUBLIC LICENSE Version 2, June 1991 Copyright (C) 1989, 1991 Free Software Foundation, Inc. 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. Preamble The licenses for most software are designed to take away your freedom to share and change it. By contrast, the GNU General Public License is intended to guarantee your freedom to share and change free software--to make sure the software is free for all its users. This General Public License applies to most of the Free Software Foundation's software and to any other program whose authors commit to using it. (Some other Free Software Foundation software is covered by the GNU Library General Public License instead.) You can apply it to your programs, too. When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are designed to make sure that you have the freedom to distribute copies of free software (and charge for this service if you wish), that you receive source code or can get it if you want it, that you can change the software or use pieces of it in new free programs; and that you know you can do these things. To protect your rights, we need to make restrictions that forbid anyone to deny you these rights or to ask you to surrender the rights. These restrictions translate to certain responsibilities for you if you distribute copies of the software, or if you modify it. For example, if you distribute copies of such a program, whether gratis or for a fee, you must give the recipients all the rights that you have. You must make sure that they, too, receive or can get the source code. And you must show them these terms so they know their rights. We protect your rights with two steps: (1) copyright the software, and (2) offer you this license which gives you legal permission to copy, distribute and/or modify the software. Also, for each author's protection and ours, we want to make certain that everyone understands that there is no warranty for this free software. If the software is modified by someone else and passed on, we want its recipients to know that what they have is not the original, so that any problems introduced by others will not reflect on the original authors' reputations. Finally, any free program is threatened constantly by software patents. We wish to avoid the danger that redistributors of a free program will individually obtain patent licenses, in effect making the program proprietary. To prevent this, we have made it clear that any patent must be licensed for everyone's free use or not licensed at all. The precise terms and conditions for copying, distribution and modification follow. GNU GENERAL PUBLIC LICENSE TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION 0. This License applies to any program or other work which contains a notice placed by the copyright holder saying it may be distributed under the terms of this General Public License. The "Program", below, refers to any such program or work, and a "work based on the Program" means either the Program or any derivative work under copyright law: that is to say, a work containing the Program or a portion of it, either verbatim or with modifications and/or translated into another language. (Hereinafter, translation is included without limitation in the term "modification".) Each licensee is addressed as "you". Activities other than copying, distribution and modification are not covered by this License; they are outside its scope. The act of running the Program is not restricted, and the output from the Program is covered only if its contents constitute a work based on the Program (independent of having been made by running the Program). Whether that is true depends on what the Program does. 1. You may copy and distribute verbatim copies of the Program's source code as you receive it, in any medium, provided that you conspicuously and appropriately publish on each copy an appropriate copyright notice and disclaimer of warranty; keep intact all the notices that refer to this License and to the absence of any warranty; and give any other recipients of the Program a copy of this License along with the Program. You may charge a fee for the physical act of transferring a copy, and you may at your option offer warranty protection in exchange for a fee. 2. You may modify your copy or copies of the Program or any portion of it, thus forming a work based on the Program, and copy and distribute such modifications or work under the terms of Section 1 above, provided that you also meet all of these conditions: a) You must cause the modified files to carry prominent notices stating that you changed the files and the date of any change. b) You must cause any work that you distribute or publish, that in whole or in part contains or is derived from the Program or any part thereof, to be licensed as a whole at no charge to all third parties under the terms of this License. c) If the modified program normally reads commands interactively when run, you must cause it, when started running for such interactive use in the most ordinary way, to print or display an announcement including an appropriate copyright notice and a notice that there is no warranty (or else, saying that you provide a warranty) and that users may redistribute the program under these conditions, and telling the user how to view a copy of this License. (Exception: if the Program itself is interactive but does not normally print such an announcement, your work based on the Program is not required to print an announcement.) These requirements apply to the modified work as a whole. If identifiable sections of that work are not derived from the Program, and can be reasonably considered independent and separate works in themselves, then this License, and its terms, do not apply to those sections when you distribute them as separate works. But when you distribute the same sections as part of a whole which is a work based on the Program, the distribution of the whole must be on the terms of this License, whose permissions for other licensees extend to the entire whole, and thus to each and every part regardless of who wrote it. Thus, it is not the intent of this section to claim rights or contest your rights to work written entirely by you; rather, the intent is to exercise the right to control the distribution of derivative or collective works based on the Program. In addition, mere aggregation of another work not based on the Program with the Program (or with a work based on the Program) on a volume of a storage or distribution medium does not bring the other work under the scope of this License. 3. You may copy and distribute the Program (or a work based on it, under Section 2) in object code or executable form under the terms of Sections 1 and 2 above provided that you also do one of the following: a) Accompany it with the complete corresponding machine-readable source code, which must be distributed under the terms of Sections 1 and 2 above on a medium customarily used for software interchange; or, b) Accompany it with a written offer, valid for at least three years, to give any third party, for a charge no more than your cost of physically performing source distribution, a complete machine-readable copy of the corresponding source code, to be distributed under the terms of Sections 1 and 2 above on a medium customarily used for software interchange; or, c) Accompany it with the information you received as to the offer to distribute corresponding source code. (This alternative is allowed only for noncommercial distribution and only if you received the program in object code or executable form with such an offer, in accord with Subsection b above.) The source code for a work means the preferred form of the work for making modifications to it. For an executable work, complete source code means all the source code for all modules it contains, plus any associated interface definition files, plus the scripts used to control compilation and installation of the executable. However, as a special exception, the source code distributed need not include anything that is normally distributed (in either source or binary form) with the major components (compiler, kernel, and so on) of the operating system on which the executable runs, unless that component itself accompanies the executable. If distribution of executable or object code is made by offering access to copy from a designated place, then offering equivalent access to copy the source code from the same place counts as distribution of the source code, even though third parties are not compelled to copy the source along with the object code. 4. You may not copy, modify, sublicense, or distribute the Program except as expressly provided under this License. Any attempt otherwise to copy, modify, sublicense or distribute the Program is void, and will automatically terminate your rights under this License. However, parties who have received copies, or rights, from you under this License will not have their licenses terminated so long as such parties remain in full compliance. 5. You are not required to accept this License, since you have not signed it. However, nothing else grants you permission to modify or distribute the Program or its derivative works. These actions are prohibited by law if you do not accept this License. Therefore, by modifying or distributing the Program (or any work based on the Program), you indicate your acceptance of this License to do so, and all its terms and conditions for copying, distributing or modifying the Program or works based on it. 6. Each time you redistribute the Program (or any work based on the Program), the recipient automatically receives a license from the original licensor to copy, distribute or modify the Program subject to these terms and conditions. You may not impose any further restrictions on the recipients' exercise of the rights granted herein. You are not responsible for enforcing compliance by third parties to this License. 7. If, as a consequence of a court judgment or allegation of patent infringement or for any other reason (not limited to patent issues), conditions are imposed on you (whether by court order, agreement or otherwise) that contradict the conditions of this License, they do not excuse you from the conditions of this License. If you cannot distribute so as to satisfy simultaneously your obligations under this License and any other pertinent obligations, then as a consequence you may not distribute the Program at all. For example, if a patent license would not permit royalty-free redistribution of the Program by all those who receive copies directly or indirectly through you, then the only way you could satisfy both it and this License would be to refrain entirely from distribution of the Program. If any portion of this section is held invalid or unenforceable under any particular circumstance, the balance of the section is intended to apply and the section as a whole is intended to apply in other circumstances. It is not the purpose of this section to induce you to infringe any patents or other property right claims or to contest validity of any such claims; this section has the sole purpose of protecting the integrity of the free software distribution system, which is implemented by public license practices. Many people have made generous contributions to the wide range of software distributed through that system in reliance on consistent application of that system; it is up to the author/donor to decide if he or she is willing to distribute software through any other system and a licensee cannot impose that choice. This section is intended to make thoroughly clear what is believed to be a consequence of the rest of this License. 8. If the distribution and/or use of the Program is restricted in certain countries either by patents or by copyrighted interfaces, the original copyright holder who places the Program under this License may add an explicit geographical distribution limitation excluding those countries, so that distribution is permitted only in or among countries not thus excluded. In such case, this License incorporates the limitation as if written in the body of this License. 9. The Free Software Foundation may publish revised and/or new versions of the General Public License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. Each version is given a distinguishing version number. If the Program specifies a version number of this License which applies to it and "any later version", you have the option of following the terms and conditions either of that version or of any later version published by the Free Software Foundation. If the Program does not specify a version number of this License, you may choose any version ever published by the Free Software Foundation. 10. If you wish to incorporate parts of the Program into other free programs whose distribution conditions are different, write to the author to ask for permission. For software which is copyrighted by the Free Software Foundation, write to the Free Software Foundation; we sometimes make exceptions for this. Our decision will be guided by the two goals of preserving the free status of all derivatives of our free software and of promoting the sharing and reuse of software generally. NO WARRANTY 11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. END OF TERMS AND CONDITIONS How to Apply These Terms to Your New Programs If you develop a new program, and you want it to be of the greatest possible use to the public, the best way to achieve this is to make it free software which everyone can redistribute and change under these terms. To do so, attach the following notices to the program. It is safest to attach them to the start of each source file to most effectively convey the exclusion of warranty; and each file should have at least the "copyright" line and a pointer to where the full notice is found. Copyright (C) This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA Also add information on how to contact you by electronic and paper mail. If the program is interactive, make it output a short notice like this when it starts in an interactive mode: Gnomovision version 69, Copyright (C) year name of author Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'. This is free software, and you are welcome to redistribute it under certain conditions; type `show c' for details. The hypothetical commands `show w' and `show c' should show the appropriate parts of the General Public License. Of course, the commands you use may be called something other than `show w' and `show c'; they could even be mouse-clicks or menu items--whatever suits your program. You should also get your employer (if you work as a programmer) or your school, if any, to sign a "copyright disclaimer" for the program, if necessary. Here is a sample; alter the names: Yoyodyne, Inc., hereby disclaims all copyright interest in the program `Gnomovision' (which makes passes at compilers) written by James Hacker. , 1 April 1989 Ty Coon, President of Vice This General Public License does not permit incorporating your program into proprietary programs. If your program is a subroutine library, you may consider it more useful to permit linking proprietary applications with the library. If this is what you want to do, use the GNU Library General Public License instead of this License. pcaMethods/DESCRIPTION0000644000175400017540000000356613556147214015417 0ustar00biocbuildbiocbuildPackage: pcaMethods Maintainer: Henning Redestig License: GPL (>= 3) Title: A collection of PCA methods LinkingTo: Rcpp LazyLoad: Yes Author: Wolfram Stacklies, Henning Redestig, Kevin Wright SystemRequirements: Rcpp Description: Provides Bayesian PCA, Probabilistic PCA, Nipals PCA, Inverse Non-Linear PCA and the conventional SVD PCA. A cluster based method for missing value estimation is included for comparison. BPCA, PPCA and NipalsPCA may be used to perform PCA on incomplete data as well as for accurate missing value estimation. A set of methods for printing and plotting the results is also provided. All PCA methods make use of the same data structure (pcaRes) to provide a common interface to the PCA results. Initiated at the Max-Planck Institute for Molecular Plant Physiology, Golm, Germany. Version: 1.78.0 URL: https://github.com/hredestig/pcamethods BugReports: https://github.com/hredestig/pcamethods/issues Encoding: UTF-8 Depends: Biobase, methods Imports: BiocGenerics, Rcpp (>= 0.11.3), MASS Suggests: matrixStats, lattice, ggplot2 Collate: 'derrorHierarchic.R' 'errorHierarchic.R' 'AllClasses.R' 'AllGenerics.R' 'BPCA_dostep.R' 'BPCA_initmodel.R' 'bpca.R' 'checkData.R' 'forkNlpcaNet.R' 'kEstimate.R' 'kEstimateFast.R' 'lineSearch.R' 'llsImpute.R' 'methods-ExpressionSet.R' 'methods-nniRes.R' 'methods-pcaRes.R' 'nipalsPca.R' 'nlpca.R' 'optiAlgCgd.R' 'orth.R' 'pca.R' 'pcaMethods-package.R' 'ppca.R' 'prep.R' 'repmat.R' 'robustPca.R' 'sortFeatures.R' 'svdImpute.R' 'vector2matrices.R' 'xval.R' Packaged: 2019-10-29 23:43:08 UTC; biocbuild biocViews: Bayesian RoxygenNote: 6.1.1 git_url: https://git.bioconductor.org/packages/pcaMethods git_branch: RELEASE_3_10 git_last_commit: 0ae2f05 git_last_commit_date: 2019-10-29 Date/Publication: 2019-10-29 NeedsCompilation: yes pcaMethods/NAMESPACE0000644000175400017540000000236213556116437015124 0ustar00biocbuildbiocbuild# Generated by roxygen2: do not edit by hand S3method(biplot,pcaRes) S3method(dim,pcaRes) S3method(fitted,pcaRes) S3method(loadings,pcaRes) S3method(plot,pcaRes) S3method(predict,pcaRes) S3method(residuals,pcaRes) S3method(scores,pcaRes) S3method(summary,pcaRes) export(Q2) export(RnipalsPca) export(asExprSet) export(bpca) export(checkData) export(cvseg) export(kEstimate) export(kEstimateFast) export(listPcaMethods) export(llsImpute) export(nipalsPca) export(nlpca) export(nni) export(pca) export(plotPcs) export(ppca) export(prep) export(robustPca) export(robustSvd) export(showNniRes) export(showPcaRes) export(svdImpute) export(svdPca) exportClasses(nlpcaNet) exportClasses(nniRes) exportClasses(pcaRes) exportMethods(DModX) exportMethods(R2VX) exportMethods(R2cum) exportMethods(center) exportMethods(centered) exportMethods(completeObs) exportMethods(cvstat) exportMethods(leverage) exportMethods(loadings) exportMethods(method) exportMethods(nObs) exportMethods(nP) exportMethods(nPcs) exportMethods(nVar) exportMethods(nmissing) exportMethods(sDev) exportMethods(scaled) exportMethods(scl) exportMethods(scores) exportMethods(slplot) exportMethods(wasna) import(Biobase) import(BiocGenerics) import(methods) importFrom(Rcpp,evalCpp) useDynLib(pcaMethods) pcaMethods/R/0000755000175400017540000000000013556116437014103 5ustar00biocbuildbiocbuildpcaMethods/R/AllClasses.R0000644000175400017540000003074413556116437016264 0ustar00biocbuildbiocbuild##' @include errorHierarchic.R ##' @include derrorHierarchic.R NULL ##' This is a class representation of a non-linear PCA neural ##' network. The \code{nlpcaNet} class is not meant for user-level ##' usage. ##' ##' Creating Objects ##' ##' \code{new("nlpcaNet", net=[the network structure], ##' hierarchic=[hierarchic design], ##' fct=[the functions at each layer], fkt=[the functions used for ##' forward propagation], weightDecay=[incremental decrease of weight ##' changes over iterations (between 0 and 1)], featureSorting=[sort ##' features or not], dataDist=[represents the present values], ##' inverse=[net is inverse mode or not], fCount=[amount of times ##' features were sorted], componentLayer=[which layer is the ##' 'bottleneck' (principal components)], ##' erro=[the used error function], gradient=[the used gradient method], ##' weights=[the present weights], ##' maxIter=[the amount of iterations that was done], scalingFactor=[the ##' scale of the original matrix])} ##' ##' Slots ##' ##' \describe{ ##' \item{net}{"matrix", matrix showing the representation of the ##' neural network, e.g. (2,4,6) for a network with two features, a ##' hidden layer and six output neurons (original variables).} ##' \item{hierarchic}{"list", the hierarchic design of the network, ##' holds 'idx' (), 'var' () and layer (which layer is the principal ##' component layer).} ##' \item{fct}{"character", a vector naming the functions that will be ##' applied on each layer. "linr" is linear (i.e.) standard matrix ##' products and "tanh" means that the arcus tangens is applied on the ##' result of the matrix product (for non-linearity).} ##' \item{fkt}{"character", same as fct but the functions used during ##' back propagation.} ##' \item{weightDecay}{"numeric", the value that is used to ##' incrementally decrease the weight changes to ensure convergence.} ##' \item{featureSorting}{"logical", indicates if features will be ##' sorted or not. This is used to make the NLPCA assume properties ##' closer to those of standard PCA were the first component is more ##' important for reconstructing the data than the second component.} ##' \item{dataDist}{"matrix", a matrix of ones and zeroes indicating ##' which values will add to the errror.} ##' \item{inverse}{"logical", network is inverse mode (currently only ##' inverse is supported) or not. Eg. the case when we have truly ##' missing values and wish to impute them.} ##' \item{fCount}{"integer", Counter for the amount of times features ##' were really sorted.} ##' \item{componentLayer}{"numeric", the index of 'net' that is the ##' component layer.} ##' \item{error}{"function", the used error function. Currently only one ##' is provided \code{errorHierarchic}.} ##' \item{gradient}{"function", the used gradient function. Currently ##' only one is provided \code{derrorHierarchic}} ##' \item{weights}{"list", A list holding managements of the ##' weights. The list has two functions, weights$current() and ##' weights$set() which access a matrix in the local environment of ##' this object.} ##' \item{maxIter}{"integer", the amount of iterations used to train ##' this network.} ##' \item{scalingFactor}{"numeric", training the network is best made ##' with 'small' values so the original data is scaled down to a ##' suitable range by division with this number.}} ##' ##' Methods ##' ##' \describe{ \item{vector2matrices}{Returns the ##' weights in a matrix representation.} } ##' @title Class representation of the NLPCA neural net ##' @docType class ##' @aliases nlpcaNet nlpcaNet-class ##' @seealso \code{\link{nlpca}} ##' @aliases nFit nFit-class ##' @exportClass nlpcaNet ##' @keywords classes ##' @name pcaNet ##' @author Henning Redestig setClass("nlpcaNet", representation(net="matrix", hierarchic="list", fct="character", fkt="character", weightDecay="numeric", featureSorting="logical", dataDist="matrix", inverse="logical", fCount="integer", componentLayer="integer", error="function", gradient="function", weights="list", maxIter="integer", scalingFactor="numeric"), prototype(net=rbind(c(4,6,2,6,4)), hierarchic=list(var=rbind(c(1,1,0.01)), layer=3, idx=rbind(c(1,1,0),c(0,1,1))), fct=c("linr", "tanh", "linr", "tanh", "linr"), fkt=c("tanh", "linr", "tanh", "linr"), weightDecay=0.001, featureSorting=TRUE, inverse=FALSE, dataDist=NULL, fCount=as.integer(0), componentLayer=as.integer(3), error=errorHierarchic, gradient=derrorHierarchic, weights=NULL, maxIter=as.integer(1200), scalingFactor=NULL)) setAs("NULL", "nlpcaNet", function(from, to){ new(to) }) ##' This is a class representation of a PCA result ##' ##' \bold{Creating Objects}\cr ##' \code{new("pcaRes", scores=[the scores], loadings=[the loadings], ##' nPcs=[amount of PCs], R2cum=[cumulative R2], nObs=[amount of ##' observations], nVar=[amount of variables], R2=[R2 for each ##' individual PC], sDev=[stdev for each individual PC], ##' centered=[was data centered], center=[original means], ##' varLimit=[what variance limit was exceeded], method=[method used to ##' calculate PCA], missing=[amount of NAs], ##' completeObs=[estimated complete observations])} ##' ##' \bold{Slots}\cr ##' \describe{ ##' \item{scores}{"matrix", the calculated scores} ##' \item{loadings}{"matrix", the calculated loadings} ##' \item{R2cum}{"numeric", the cumulative R2 values} ##' \item{sDev}{"numeric", the individual standard ##' deviations of the score vectors} ##' \item{R2}{"numeric", the individual R2 values} ##' \item{cvstat}{"numeric", cross-validation statistics} ##' \item{nObs}{"numeric", number of observations} ##' \item{nVar}{"numeric", number of variables} ##' \item{centered}{"logical", data was centered or not} ##' \item{center}{"numeric", the original variable centers} ##' \item{scaled}{"logical", data was scaled or not} ##' \item{scl}{"numeric", the original variable scales} ##' \item{varLimit}{"numeric", the exceeded variance limit} ##' \item{nPcs,nP}{"numeric", the number of calculated PCs} ##' \item{method}{"character", the method used to perform PCA} ##' \item{missing}{"numeric", the total amount of missing values in ##' original data} ##' \item{completeObs}{"matrix", the estimated complete observations} ##' \item{network}{"nlpcaNet", the network used by non-linear PCA} ##' } ##' ##' \bold{Methods (not necessarily exhaustive)}\cr ##' \describe{ ##' \item{print}{Print function} ##' \item{summary}{Extract information about PC relevance} ##' \item{screeplot}{Plot a barplot of standard deviations for PCs} ##' \item{slplot}{Make a side by side score and loadings plot} ##' \item{nPcs}{Get the number of PCs} ##' \item{nObs}{Get the number of observations} ##' \item{cvstat}{Cross-validation statistics} ##' \item{nVar}{Get the number of variables} ##' \item{loadings}{Get the loadings} ##' \item{scores}{Get the scores} ##' \item{dim}{Get the dimensions (number of observations, number of ##' features)} ##' \item{centered}{Get a logical indicating if centering was done as ##' part of the model} ##' \item{center}{Get the averages of the original variables.} ##' \item{completeObs}{Get the imputed data set} ##' \item{method}{Get a string naming the used PCA method} ##' \item{sDev}{Get the standard deviations of the PCs} ##' \item{scaled}{Get a logical indicating if scaling was done as ##' part of the model} ##' \item{scl}{Get the scales of the original variablesb} ##' \item{R2cum}{Get the cumulative R2} ##' } ##' @title Class for representing a PCA result ##' @keywords classes ##' @exportClass pcaRes ##' @docType class ##' @name pcaRes ##' @aliases pcaRes pcaRes-class ##' @author Henning Redestig setClass("pcaRes", representation(completeObs="matrix", scores="matrix", loadings="matrix", R2cum="numeric", R2="numeric", # ditch, get from R2cum cvstat="numeric", # ditch, get from R2cum sDev="numeric", # ditch, get from scores nObs="numeric", # ditch, get from scores nVar="numeric", centered="logical", center="numeric", subset="numeric", scaled="character", scale="numeric", varLimit="numeric", # ditch, useless nPcs="numeric", # ditch, get from scores method="character", missing="matrix", network="nlpcaNet"), prototype(completeObs=NULL, scores=NULL, loadings=NULL, R2cum=NULL, R2=NULL, subset=NULL, cvstat=NULL, sDev=NULL, nObs=NULL, nVar=NULL, centered=NULL, center=NULL, scaled=NULL, scale=NULL, varLimit=NULL, nPcs=NULL, method=NULL, missing=NULL, network=NULL)) setAs("NULL", "pcaRes", function(from, to){ new(to) }) ##' This is a class representation of nearest neighbour imputation ##' (nni) result ##' ##' \bold{Creating Objects}\cr ##' \code{new("nniRes", completeObs=[the estimated complete ##' observations], k=[cluster size], nObs=[amount of observations], ##' nVar=[amount of variables], centered=[was the data centered befor ##' running LLSimpute], center=[original means], method=[method used ##' to perform clustering], missing=[amount of NAs])} ##' ##' \bold{Slots}\cr ##' \describe{ ##' \item{completeObs}{"matrix", the estimated complete observations} ##' \item{nObs}{"numeric", amount of observations} ##' \item{nVar}{"numeric", amount of variables} ##' \item{correlation}{"character", the correlation method used ##' (pearson, kendall or spearman)} ##' \item{centered}{"logical", data was centered or not} ##' \item{center}{"numeric", the original variable centers} ##' \item{k}{"numeric", cluster size} ##' \item{method}{"character", the method used to perform the clustering} ##' \item{missing}{"numeric", the total amount of missing values in ##' original data} ##' } ##' ##' \bold{Methods}\cr ##' \describe{ \item{print}{Print function} } ##' @title Class for representing a nearest neighbour imputation result ##' @docType class ##' @exportClass nniRes ##' @name nniRes ##' @keywords classes ##' @aliases nniRes nniRes-class ##' @author Wolfram Stacklies setClass("nniRes", representation(completeObs="matrix", nObs="numeric", nVar="numeric", centered="logical", center="numeric", k="numeric", method="character", correlation="character", missing="numeric"), prototype(completeObs=NULL, nObs=NULL, nVar=NULL, centered=NULL, center=NULL, k=NULL, method=NULL, correlation=NULL, missing=NULL)) setAs("NULL", "nniRes", function(from, to) { new(to) }) ##' Create an object that holds the weights for nlpcaNet. Holds and ##' sets weights in using an environment object. ##' @param w \code{matrix} -- New weights ##' @return A weightsAccound with \code{set} and \code{current} ##' functions. ##' @author Henning Redestig weightsAccount <- function(w) { list( set = function(newWeights) { if(!inherits(newWeights, "matrix")) stop("The weights must inherit from matrix") w <<- newWeights }, current = function() { w } ) } pcaMethods/R/AllGenerics.R0000644000175400017540000000456513556116437016430 0ustar00biocbuildbiocbuildsetGeneric("vector2matrices", function(object, ...) standardGeneric("vector2matrices")) ##' @exportMethod leverage setGeneric("leverage", function(object, ...) standardGeneric("leverage")) ##' @exportMethod DModX setGeneric("DModX", function(object, dat, newdata=FALSE, type=c("normalized","absolute"), ...) standardGeneric("DModX")) ##' @exportMethod nP setGeneric("nP", function(object, ...) standardGeneric("nP")) ##' @exportMethod cvstat setGeneric("cvstat", function(object, ...) standardGeneric("cvstat")) ##' @exportMethod nPcs setGeneric("nPcs", function(object, ...) standardGeneric("nPcs")) ##' @exportMethod nObs setGeneric("nObs", function(object, ...) standardGeneric("nObs")) ##' @exportMethod nVar setGeneric("nVar", function(object, ...) standardGeneric("nVar")) ##' @exportMethod centered setGeneric("centered", function(object, ...) standardGeneric("centered")) ##' @exportMethod center setGeneric("center", function(object, ...) standardGeneric("center")) ##' @exportMethod completeObs setGeneric("completeObs", function(object, ...) standardGeneric("completeObs")) ##' @exportMethod method setGeneric("method", function(object, ...) standardGeneric("method")) ##' @exportMethod nmissing setGeneric("nmissing", function(object, ...) standardGeneric("nmissing")) ##' @exportMethod wasna setGeneric("wasna", function(object, ...) standardGeneric("wasna")) ##' @exportMethod sDev setGeneric("sDev", function(object, ...) standardGeneric("sDev")) ##' @exportMethod scaled setGeneric("scaled", function(object, ...) standardGeneric("scaled")) ##' @exportMethod scl setGeneric("scl", function(object, ...) standardGeneric("scl")) ##' @exportMethod R2cum setGeneric("R2cum", function(object, ...) standardGeneric("R2cum")) ##' @exportMethod slplot setGeneric("slplot", function(object, pcs=c(1,2), scoresLoadings=c(TRUE, TRUE), sl="def", ll="def", hotelling=0.95, rug=TRUE, sub=NULL,...) standardGeneric("slplot")) ##' @exportMethod scores setGeneric("scores", function(object, ...) standardGeneric("scores")) ##' @exportMethod loadings setGeneric("loadings", function(object, ...) standardGeneric("loadings")) ##' @exportMethod R2VX setGeneric("R2VX", function(object, ...) standardGeneric("R2VX")) ## @exportMethod prep #setGeneric("prep", function(object, ...) standardGeneric("prep")) pcaMethods/R/BPCA_dostep.R0000644000175400017540000000543713556116437016322 0ustar00biocbuildbiocbuild##' The function contains the actual implementation of the BPCA ##' component estimation. It performs one step of the BPCA EM ##' algorithm. It is called 'maxStep' times from within the main loop ##' in BPCAestimate. ##' ##' This function is NOT intended to be run standalone. ##' @title Do BPCA estimation step ##' @param M Data structure containing all needed information. See the ##' source documentation of BPCA_initmodel for details ##' @param y Numeric original data matrix ##' @return Updated version of the data structure ##' @author Wolfram Stacklies BPCA_dostep <- function(M,y) { ## Empty matrix in which the scores are copied M$scores <- matrix(NA, M$rows, M$comps) ## Expectation step for data without missing values Rx <- diag(M$comps) + M$tau * t(M$PA) %*% M$PA + M$SigW Rxinv <- solve(Rx) idx <- M$row_nomiss if (length(idx) == 0) { trS <- 0 T <- 0 } else { dy <- y[idx,, drop=FALSE] - repmat(M$mean, length(idx), 1) x <- M$tau * Rxinv %*% t(M$PA) %*% t(dy) T <- t(dy) %*% t(x) trS <- sum(sum(dy * dy)) ## Assign the scores for complete rows xTranspose <- t(x) for (i in 1:length(idx)) { M$scores[idx[i],] <- xTranspose[i,] } } ## Expectation step for incomplete data if( length(M$row_miss) > 0) { for(n in 1:length(M$row_miss)) { i <- M$row_miss[n] dyo <- y[ i, !M$nans[i,], drop=FALSE] - M$mean[ !M$nans[i,], drop=FALSE] Wm <- M$PA[ M$nans[i,],, drop=FALSE] Wo <- M$PA[ !M$nans[i,],, drop=FALSE] Rxinv <- solve( (Rx - M$tau * t(Wm) %*% Wm)) ex <- M$tau * t(Wo) %*% t(dyo) x <- Rxinv %*% ex dym <- Wm %*% x dy <- y[i,, drop=FALSE] dy[ !M$nans[i,] ] <- t(dyo) dy[ M$nans[i,] ] <- t(dym) M$yest[i,] <- dy + M$mean T <- T + t(dy) %*% t(x) T[ M$nans[i,], ] <- T[ M$nans[i,],, drop=FALSE] + Wm %*% Rxinv trS <- trS + dy %*% t(dy) + sum(M$nans[i,]) / M$tau + sum( diag(Wm %*% Rxinv %*% t(Wm)) ) trS <- trS[1,1] ## Assign the scores for rows containing missing values M$scores[M$row_miss[n],] <- t(x) } } T <- T / M$rows trS <- trS / M$rows ## Maximation step Rxinv <- solve(Rx) Dw <- Rxinv + M$tau * t(T) %*% M$PA %*% Rxinv + diag(M$alpha, nrow = length(M$alpha)) / M$rows Dwinv <- solve(Dw) M$PA <- T %*% Dwinv ## The new estimate of the principal axes (loadings) M$tau <- (M$cols + 2 * M$gtau0 / M$rows) / (trS - sum(diag(t(T) %*% M$PA)) + (M$mean %*% t(M$mean) * M$gmu0 + 2 * M$gtau0 / M$btau0) / M$rows) M$tau <- M$tau[1,1] ## convert to scalar M$SigW <- Dwinv * (M$cols / M$rows) M$alpha <- (2 * M$galpha0 + M$cols) / (M$tau * diag(t(M$PA) %*% M$PA) + diag(M$SigW) + 2 * M$galpha0 / M$balpha0) return(M) } pcaMethods/R/BPCA_initmodel.R0000644000175400017540000000510313556116437016776 0ustar00biocbuildbiocbuild##' Model initialization for Bayesian PCA. This function is NOT ##' inteded to be run separately! ##' ##' The function calculates the initial Eigenvectors by use of SVD ##' from the complete rows. The data structure M is created and ##' initial values are assigned. ##' @title Initialize BPCA model ##' @param y numeric matrix containing missing values. Missing values ##' are denoted as 'NA' ##' @param components Number of components used for estimation ##' @return List containing ##' \item{rows}{Row number of input matrix} ##' \item{cols}{Column number of input matrix} ##' \item{comps}{Number of components to use} ##' \item{yest}{(working variable) current estimate of complete data} ##' \item{row_miss}{(Array) Indizes of rows containing missing values} ##' \item{row_nomiss}{(Array) Indices of complete rows (such with no ##' missing values)} ##' \item{nans}{Matrix of same size as input data. TRUE if \code{input == NA}, ##' false otherwise} ##' \item{mean}{Column wise data mean} ##' \item{PA}{ (d x k) Estimated principal axes (eigenvectors, ##' loadings) The matrix ROWS are the vectors} ##' \item{tau}{Estimated precision of the residual error} ##' \item{scores}{ Estimated scores} ##' Further elements are: galpha0, balpha0, alpha, gmu0, btau0, gtau0, ##' SigW. These are working variables or constants. ##' @author Wolfram Stacklies BPCA_initmodel <- function(y, components) { ## Initialization, write static parameters to the central M <- NULL M$rows <- nrow(y) M$cols <- ncol(y) M$comps <- components ## Column number M$yest <- y ## Original data, NAs are set to 0 later on ## Find rows with missing values, etc... M$nans <- is.na(y) temp <- apply(M$nans, 1, sum) M$row_nomiss <- which(temp == 0) M$row_miss <- which(temp != 0) M$yest[M$nans] <- 0 M$scores <- NULL ## Get the SVD of the complete rows covy <- cov(M$yest) values <- svd(covy, components, components) U <- values[[2]] S <- diag( values[[1]][1:components], nrow = components, ncol = components) V <- values[[3]] ## M$mean: column wise mean of the original data M$mean <- matrix(0, 1, M$cols) for(j in 1:M$cols) { idx <- which(!is.na(y[,j])) M$mean[j] <- mean(y[idx,j]) } M$PA <- U %*% sqrt(S) M$tau <- 1 / ( sum(diag(covy)) - sum(diag(S)) ) ## Constants etc taumax <- 1e10 taumin <- 1e-10 M$tau <- max( min(M$tau, taumax), taumin ) M$galpha0 <- 1e-10 M$balpha0 <- 1 M$alpha <- (2 * M$galpha0 + M$cols) / (M$tau * diag(t(M$PA) %*% M$PA) + 2 * M$galpha0 / M$balpha0) M$gmu0 <- 0.001 M$btau0 <- 1 M$gtau0 <- 1e-10 M$SigW <- diag(components) return(M) } pcaMethods/R/bpca.R0000644000175400017540000001606313556116437015141 0ustar00biocbuildbiocbuild##' Implements a Bayesian PCA missing value estimator. The script ##' is a port of the Matlab version provided by Shigeyuki OBA. See ##' also \url{http://ishiilab.jp/member/oba/tools/BPCAFill.html}. ##' BPCA combines an EM approach for PCA with a Bayesian model. In ##' standard PCA data far from the training set but close to the ##' principal subspace may have the same reconstruction error. BPCA ##' defines a likelihood function such that the likelihood for data ##' far from the training set is much lower, even if they are close to ##' the principal subspace. ##' ##' Scores and loadings obtained with Bayesian PCA slightly differ ##' from those obtained with conventional PCA. This is because BPCA ##' was developed especially for missing value estimation. The ##' algorithm does not force orthogonality between factor loadings, as ##' a result factor loadings are not necessarily orthogonal. However, ##' the BPCA authors found that including an orthogonality criterion ##' made the predictions worse. ##' ##' The authors also state that the difference between real and ##' predicted Eigenvalues becomes larger when the number of ##' observation is smaller, because it reflects the lack of ##' information to accurately determine true factor loadings from the ##' limited and noisy data. As a result, weights of factors to ##' predict missing values are not the same as with conventional PCA, ##' but the missing value estimation is improved. ##' ##' BPCA works iteratively, the complexity is growing with ##' \eqn{O(n^3)}{O(n^3)} because several matrix inversions are ##' required. The size of the matrices to invert depends on the ##' number of components used for re-estimation. ##' ##' Finding the optimal number of components for estimation is not a ##' trivial task; the best choice depends on the internal structure of ##' the data. A method called \code{kEstimate} is provided to ##' estimate the optimal number of components via cross validation. ##' In general few components are sufficient for reasonable estimation ##' accuracy. See also the package documentation for further ##' discussion about on what data PCA-based missing value estimation ##' makes sense. ##' ##' It is not recommended to use this function directely but rather to ##' use the pca() wrapper function. ##' ##' There is a difference with respect the interpretation of rows ##' (observations) and columns (variables) compared to matlab ##' implementation. For estimation of missing values for microarray ##' data, the suggestion in the original bpca is to intepret genes as ##' observations and the samples as variables. In pcaMethods however, ##' genes are interpreted as variables and samples as observations ##' which arguably also is the more natural interpretation. For bpca ##' behavior like in the matlab implementation, simply transpose your ##' input matrix. ##' ##' Details about the probabilistic model underlying BPCA are found in ##' Oba et. al 2003. The algorithm uses an expectation maximation ##' approach together with a Bayesian model to approximate the ##' principal axes (eigenvectors of the covariance matrix in PCA). ##' The estimation is done iteratively, the algorithm terminates if ##' either the maximum number of iterations was reached or if the ##' estimated increase in precision falls below \eqn{1e^{-4}}{1e^-4}. ##' ##' \bold{Complexity:} The relatively high complexity of the method is ##' a result of several matrix inversions required in each step. ##' Considering the case that the maximum number of iteration steps is ##' needed, the approximate complexity is given by the term ##' \deqn{maxSteps \cdot row_{miss} \cdot O(n^3)}{maxSteps * row_miss ##' * O(n^3)} Where \eqn{row_{miss}}{row_miss} is the number of rows ##' containing missing values and \eqn{O(n^3)}{O(n^3)} is the ##' complexity for inverting a matrix of size ##' \eqn{components}{components}. Components is the number of ##' components used for re-estimation. ##' @title Bayesian PCA missing value estimation ##' @param Matrix \code{matrix} -- Pre-processed matrix (centered, ##' scaled) with variables in columns and observations in rows. The ##' data may contain missing values, denoted as \code{NA}. ##' @param nPcs \code{numeric} -- Number of components used for ##' re-estimation. Choosing few components may decrease the ##' estimation precision. ##' @param maxSteps \code{numeric} -- Maximum number of estimation ##' steps. ##' @param verbose \code{boolean} -- BPCA prints the number of steps ##' and the increase in precision if set to TRUE. Default is ##' interactive(). ##' @param threshold convergence threshold ##' @param ... Reserved for future use. Currently no further ##' parameters are used ##' @return Standard PCA result object used by all PCA-based methods ##' of this package. Contains scores, loadings, data mean and ##' more. See \code{\link{pcaRes}} for details. ##' @references Shigeyuki Oba, Masa-aki Sato, Ichiro Takemasa, Morito ##' Monden, Ken-ichi Matsubara and Shin Ishii. A Bayesian missing ##' value estimation method for gene expression profile ##' data. \emph{Bioinformatics, 19(16):2088-2096, Nov 2003}. ##' @seealso \code{\link{ppca}}, \code{\link{svdImpute}}, ##' \code{\link{prcomp}}, \code{\link{nipalsPca}}, ##' \code{\link{pca}}, ##' \code{\link{pcaRes}}. \code{\link{kEstimate}}. ##' @note Requires \code{MASS}. ##' @examples ##' ## Load a sample metabolite dataset with 5\% missig values (metaboliteData)e ##' data(metaboliteData) ##' ## Perform Bayesian PCA with 2 components ##' pc <- pca(t(metaboliteData), method="bpca", nPcs=2) ##' ## Get the estimated principal axes (loadings) ##' loadings <- loadings(pc) ##' ## Get the estimated scores ##' scores <- scores(pc) ##' ## Get the estimated complete observations ##' cObs <- completeObs(pc) ##' ## Now make a scores and loadings plot ##' slplot(pc) ##' \dontshow{stopifnot(sum((fitted(pc) - t(metaboliteData))^2, na.rm=TRUE) < 200)} ##' @keywords multivariate ##' @export ##' @author Wolfram Stacklies bpca <- function(Matrix, nPcs=2, maxSteps=100, verbose=interactive(), threshold=1e-4, ... ) { ## R implementation of a Bayesion PCA missing value estimator. ## After the Matlab script of Shigeyuki OBA (2002 May. 5th) ## See also: http://hawaii.aist-nara.ac.jp/%7Eshige-o/tools/ ## Great thanks to them! M <- BPCA_initmodel(Matrix, nPcs) tauold <- 1000 for( step in 1:maxSteps ) { M <- BPCA_dostep(M, Matrix) if( step %% 10 == 0 ) { tau <- M$tau dtau <- abs(log10(tau) - log10(tauold)) if ( verbose ) { cat("Step Number : ", step, '\n') cat("Increase in precision : ", dtau, '\n') cat("----------", '\n') } if (dtau < threshold) { break } tauold <- tau } } R2cum <- rep(NA, nPcs) TSS <- sum(Matrix^2, na.rm=TRUE) for (i in 1:nPcs) { difference <- Matrix - (M$scores[,1:i, drop=FALSE] %*% t(M$PA[,1:i, drop=FALSE]) ) R2cum[i] <- 1 - (sum(difference^2, na.rm=TRUE) / TSS) } result <- new("pcaRes") result@scores <- M$scores result@loadings <- M$PA result@R2cum <- R2cum result@method <- "bpca" return(result) } pcaMethods/R/checkData.R0000644000175400017540000000537313556116437016105 0ustar00biocbuildbiocbuild##' Check a given data matrix for consistency with the format ##' required for further analysis. ##' The data must be a numeric matrix and not contain: ##' \itemize{ ##' \item Inf values ##' \item NaN values ##' \item Rows or columns that consist of NA only ##' } ##' @title Do some basic checks on a given data matrix ##' @param data \code{matrix} -- Data to check. ##' @param verbose \code{boolean} -- If TRUE, the function prints ##' messages whenever an error in the data set is found. ##' @return \item{isValid}{\code{boolean} -- TRUE if no errors were ##' found, FALSE otherwise. isValid contains a set of attributes, ##' these are: \itemize{ \item isNumeric - TRUE if data is numeric, ##' false otherwise \item isInfinite - TRUE if data contains 'Inf' ##' values, false otherwise \item isNaN - TRUE if data contains 'NaN' ##' values, false otherwise \item isMatrix - TRUE if the data is in ##' matrix format, FALSE otherwise \item naRows - TRUE if data ##' contains rows in which all elements are 'NA', FALSE otherwise ##' \item naCols - TRUE if data contains columns in which all elements ##' are 'NA', FALSE otherwise }} ##' @keywords multivariate ##' @export ##' @author Wolfram Stacklies checkData <- function(data, verbose = FALSE) { isValid <- TRUE isNumeric <- TRUE isInfinite <- FALSE isNaN <- FALSE isMatrix <- TRUE naRows <- FALSE naCols <- FALSE if (!is.numeric(data)) { isNumeric <- FALSE isValid <- FALSE if (verbose) message("Error: Data is not numeric") } if ( sum(is.infinite(data) >= 1) ) { isInfinite <- TRUE isValid <- FALSE if (verbose) message("Error: Data contains 'Inf' values") } if (sum(is.nan(data) >= 1)) { isNaN <- TRUE isValid <- FALSE if (verbose) message("Error: Data contains 'NaN' values. Missing values must be denoted by 'NA'") } if (!is.matrix(data)) { isMatrix <- FALSE isValid <- FALSE if (verbose) message("Error: data is not a matrix. Try to use as.matrix(data)") } ## Check for entire rows that are NA only if (sum(apply(is.na(data), 1, sum) == ncol(data)) >= 1 ) { naRows <- TRUE isValid <- FALSE if (verbose) message("Error: Data contains rows in which all elements are 'NA'. Remove them first") } ## Check for entire columns that are NA only if (sum(apply(is.na(data), 2, sum) == nrow(data)) >= 1 ) { naCols <- TRUE isValid <- FALSE if (verbose) message("Error: Data contains columns in which all elements are 'NA'. Remove them first") } attr(isValid, "isNumeric") <- isNumeric attr(isValid, "isInfinite") <- isInfinite attr(isValid, "isNaN") <- isNaN attr(isValid, "isMatrix") <- isMatrix attr(isValid, "naRows") <- naRows attr(isValid, "naCols") <- naCols return(isValid) } pcaMethods/R/derrorHierarchic.R0000644000175400017540000001251113556116437017505 0ustar00biocbuildbiocbuild##' Later ##' @param nlnet the nlnet ##' @param trainIn training data ##' @param trainOut fitted data ##' @return derror ##' @author Henning Redestig, Matthias Scholz derrorHierarchic <- function(nlnet, trainIn, trainOut) { weights <- nlnet@weights$current() netDim <- dim(nlnet@net) if(nlnet@inverse) { numElements <- nlnet@net[1] * dim(trainOut)[2] trainIn <- matrix(weights[1:numElements], nrow=nlnet@net[1], ncol=dim(trainOut)[2]) wTrainIn <- weights[1:numElements,drop=FALSE] weights <- weights[(numElements + 1):length(weights), ,drop=FALSE] } weightMats <- vector2matrices(weights, nlnet@net) trainDim <- dim(trainIn) subnetNum <- length(nlnet@hierarchic$var) ## ****************************** Epattern <- array(0, dim=c(dim(trainOut), subnetNum)) nOut <- array(0, dim=c(sum(nlnet@net), trainDim[2], subnetNum)) for(subnet in 1:subnetNum) nOut[1:trainDim[1],,subnet] <- eval(parse(text=paste(nlnet@fct[1], "(trainIn)"))) if(nlnet@inverse) for(subnet in 1:subnetNum) nOut[nlnet@hierarchic$idx[,subnet]==0,,subnet] <- 0 ## forward propagation for(subnet in 1:subnetNum) { if(nlnet@hierarchic$var[subnet] != 0) { sBias <- array(1, dim=c(1, trainDim[2])) for(i in 1:(netDim[2] - 1)) { if(i == 1) nBegin <- 1 else nBegin <- sum(nlnet@net[1:(i-1)])+1 sIn <- rbind(sBias, nOut[nBegin:sum(nlnet@net[1:i]),,subnet]) sOut <- eval(parse(text=paste(nlnet@fct[i+1], "(weightMats[[i]] %*% sIn)"))) if(i == (nlnet@hierarchic$layer - 1)) sOut[nlnet@hierarchic$idx[,subnet]==0,] <- 0 nOut[(sum(nlnet@net[1:i])+1):sum(nlnet@net[1:(i+1)]),,subnet] <- sOut } output <- nOut[(sum(nlnet@net[1:(length(nlnet@net)-1)])+1):dim(nOut)[1],,subnet] Epattern[,,subnet] <- output - trainOut } } ## error function Epattern <- Epattern^2 Epattern[is.na(Epattern)] <- 0 #set the missing values to zero if(!is.null(nlnet@dataDist)) for(subnet in 1:subnetNum) Epattern[,,subnet] <- Epattern[,,subnet] * nlnet@dataDist Eitemize <- apply(Epattern, 3, sum) * 0.5 Etotal <- sum(nlnet@hierarchic$var * Eitemize) if(!is.null(nlnet@weightDecay)) Etotal <- Etotal + nlnet@weightDecay * 0.5 * sum(weights^2) if(nlnet@inverse) Etotal <- Etotal + 0.01 * nlnet@weightDecay * 0.5 * sum(wTrainIn^2) ## back propagation nError <- array(0, dim=c(sum(nlnet@net), trainDim[2], subnet)) dWeight <- vector(length=netDim[2] - 1, mode="list") wBp <- vector(length=netDim[2] - 1, mode="list") ## wBp is weights for back propagation for(u in 1:(netDim[2] - 1)) wBp[[u]] <- weightMats[[u]][,2:(nlnet@net[u] + 1)] # cats the weights which belong to bias dw <- array(0, dim=c(length(weights), subnet)) for(subnet in 1:subnetNum) { if(nlnet@hierarchic$var[subnet] != 0) { ## last layer sTmp <- nOut[(dim(nOut)[1]-nlnet@net[length(nlnet@net)]+1):dim(nOut)[1],,subnet] if(nlnet@fct[length(nlnet@fct)] == "tanh") eTmp <- (1 - sTmp^2) * (sTmp - trainOut) #prev trainOut - sTmp (fixed to get rid of sign change) else if(nlnet@fct[length(nlnet@fct)] == "linr") eTmp <- sTmp - trainOut #prev trainOut - sTmp (fixed to get rid of sign change) eTmp[is.na(eTmp)] <- 0 if(!is.null(nlnet@dataDist)) eTmp <- eTmp * nlnet@dataDist nError[(dim(nError)[1]-nlnet@net[length(nlnet@net)]+1):dim(nError)[1],,subnet] <- eTmp ## all other layers for(n in 1:(netDim[2] - 1)){ i <- netDim[2]-n ## the if clause is to avoid 1:0 difference in R ## Matlab (1:0 => Empty matrix), R (1:0 => [1,0]) if(i > 1) sTmp <- nOut[(sum(nlnet@net[1:(i-1)])+1):sum(nlnet@net[1:i]),,subnet] else sTmp <- nOut[1:sum(nlnet@net[1:i]),,subnet] if(i==(nlnet@hierarchic$layer-1)) eTmp[nlnet@hierarchic$idx[,subnet]==0,] <- 0 dWeight[[i]] <- tcrossprod(eTmp, rbind(sBias, sTmp)) #gradient if (nlnet@fct[i] == "tanh") eTmp <- (1 - sTmp^2) * crossprod(wBp[[i]],eTmp) else if (nlnet@fct[i] == "linr") eTmp <- crossprod(wBp[[i]], eTmp) ## the if clause is to avoid 1:0 difference in R if(i > 1) nError[(sum(nlnet@net[1:(i - 1)]) + 1):sum(nlnet@net[1:i]), ,subnet] <- eTmp else nError[1:sum(nlnet@net[1:i]), ,subnet] <- eTmp } dw[,subnet] <- unlist(dWeight) #fixed sign change } } if(nlnet@inverse) { dw <- rbind(array(0, dim=c(numElements, subnetNum)), dw) for(subnet in 1:subnetNum) { eTmp <- array(nError[1:nlnet@net[1],,subnet], dim=c(nlnet@net[1], dim(nError)[2])) eTmp[nlnet@hierarchic$idx[,subnet] == 0,] <- 0 dim(eTmp) <- NULL #a bit unsure if this is correct but seems to work dw[1:numElements,subnet] <- unlist(eTmp) #fixed sign change } ## weights <- rbind(cbind(rep(0, numElements)), cbind(weights)) #old: only weight decay for real weights weights <- rbind(cbind(0.01 * wTrainIn), cbind(weights)) #new } dwTotal <- array(0, dim=dim(weights)) for (subnet in 1:subnetNum) { dwTotal <- dwTotal + nlnet@hierarchic$var[subnet] * dw[, subnet] } if(!is.null(nlnet@weightDecay)) dwTotal <- dwTotal + nlnet@weightDecay * weights return(list(dwTotal=dwTotal, Etotal=Etotal, nError=nError, nOut=nOut)) } pcaMethods/R/errorHierarchic.R0000644000175400017540000000440013556116437017337 0ustar00biocbuildbiocbuild##' Later ##' @param nlnet The nlnet ##' @param trainIn training data ##' @param trainOut fitted data ##' @return error ##' @author Henning Redestig, Matthias Scholz errorHierarchic <- function(nlnet, trainIn, trainOut) { weights <- nlnet@weights$current() if(nlnet@inverse) { numElements <- nlnet@net[1] * dim(trainOut)[2] trainIn <- array(weights[1:numElements], dim=c(nlnet@net[1], dim(trainOut)[2])) wTrainIn <- weights[1:numElements, drop=FALSE] weights <- weights[(numElements + 1):length(weights),,drop=FALSE] } netDim <- dim(nlnet@net) trainDim <- dim(trainOut) weightMats <- vector2matrices(weights, nlnet@net) hierarchicIdx <- nlnet@hierarchic$idx[,nlnet@hierarchic$var != 0, drop=FALSE] hierarchicVar <- nlnet@hierarchic$var[,colSums(nlnet@hierarchic$var) != 0, drop=FALSE] subnetNum <- length(hierarchicVar) out <- array(0, dim=c(trainDim[1], trainDim[2], subnetNum)) sBias <- array(1, dim=c(1, trainDim[2])) sExtract <- eval(parse(text=paste(nlnet@fct[1], "(trainIn)"))) Eitemize <- NULL if(nlnet@hierarchic$layer > 1) { #this should not be executed at all if sequence is 1:0 for(layer in 1:(nlnet@hierarchic$layer - 1)) { sExtract <- rbind(sBias, sExtract) sExtract <- eval(parse(text=paste(nlnet@fct[layer + 1], "(weightMats[[layer]] %*% sExtract)"))) } } for(subnet in 1:subnetNum) { sRecon <- sExtract sRecon[hierarchicIdx[,subnet]==0,] <- 0 for(layer in nlnet@hierarchic$layer:(netDim[2] - 1)) { sRecon <- rbind(sBias, sRecon) sRecon <- eval(parse(text=paste(nlnet@fct[layer+1], "(weightMats[[layer]] %*% sRecon)"))) } out[,,subnet] <- sRecon ## error function eTmp <- (sRecon - trainOut)^2 eTmp[is.na(eTmp)] <- 0 Eitemize[subnet] <- sum(eTmp) * 0.5 if(!is.null(nlnet@dataDist)) Eitemize[subnet] <- 0.5 * sum(nlnet@dataDist * eTmp) else Eitemize[subnet] <- 0.5 * sum(eTmp) } error <- tcrossprod(hierarchicVar, rbind(Eitemize)) if(!is.null(nlnet@weightDecay)) error <- error + nlnet@weightDecay * 0.5 * sum(weights^2) ## smooth (0.01) weight decay also for input values if(nlnet@inverse) error <- error + 0.01 * nlnet@weightDecay * 0.5 * sum(wTrainIn^2) return(list(error=error, out=out)) } pcaMethods/R/forkNlpcaNet.R0000644000175400017540000000131413556116437016613 0ustar00biocbuildbiocbuild##' Complete copy of nlpca net object ##' @param nlnet a nlnet ##' @return A copy of the input nlnet ##' @author Henning Redestig forkNlpcaNet <- function(nlnet) { res <- new("nlpcaNet") res@net <- nlnet@net res@hierarchic <- nlnet@hierarchic res@fct <- nlnet@fct res@fkt <- nlnet@fkt res@weightDecay <- nlnet@weightDecay res@featureSorting <- nlnet@featureSorting res@dataDist <- nlnet@dataDist res@inverse <- nlnet@inverse res@fCount <- nlnet@fCount res@componentLayer <- nlnet@componentLayer res@error <- nlnet@error res@gradient <- nlnet@gradient res@weights <- weightsAccount(nlnet@weights$current()) res@maxIter <- nlnet@maxIter res@scalingFactor <- nlnet@scalingFactor res } pcaMethods/R/kEstimate.R0000644000175400017540000003072013556116437016156 0ustar00biocbuildbiocbuild##' Perform cross validation to estimate the optimal number of ##' components for missing value estimation. Cross validation is ##' done for the complete subset of a variable. ##' ##' The assumption hereby is that variables that are highly correlated ##' in a distinct region (here the non-missing observations) are also ##' correlated in another (here the missing observations). This also ##' implies that the complete subset must be large enough to be ##' representative. For each incomplete variable, the available ##' values are divided into a user defined number of cv-segments. The ##' segments have equal size, but are chosen from a random equal ##' distribution. The non-missing values of the variable are covered ##' completely. PPCA, BPCA, SVDimpute, Nipals PCA, llsImpute an NLPCA ##' may be used for imputation. ##' ##' The whole cross validation is repeated several times so, depending ##' on the parameters, the calculations can take very long time. As ##' error measure the NRMSEP (see Feten et. al, 2005) or the Q2 ##' distance is used. The NRMSEP basically normalises the RMSD ##' between original data and estimate by the variable-wise ##' variance. The reason for this is that a higher variance will ##' generally lead to a higher estimation error. If the number of ##' samples is small, the variable - wise variance may become an ##' unstable criterion and the Q2 distance should be used ##' instead. Also if variance normalisation was applied previously. ##' ##' The method proceeds variable - wise, the NRMSEP / Q2 distance is ##' calculated for each incomplete variable and averaged ##' afterwards. This allows to easily see for wich set of variables ##' missing value imputation makes senes and for wich set no ##' imputation or something like mean-imputation should be used. Use ##' \code{kEstimateFast} or \code{Q2} if you are not interested in ##' variable wise CV performance estimates. ##' ##' Run time may be very high on large data sets. Especially when used ##' with complex methods like BPCA or Nipals PCA. For PPCA, BPCA, ##' Nipals PCA and NLPCA the estimation method is called ##' \eqn{(v_{miss} \cdot segs \cdot nruncv \cdot)}{(v\_miss * segs * ##' nruncv)} times as the error for all numbers of principal ##' components can be calculated at once. For LLSimpute and SVDimpute ##' this is not possible, and the method is called \eqn{(v_{miss} ##' \cdot segs \cdot nruncv \cdot length(evalPcs))}{(v\_miss * segs * ##' nruncv * length(evalPcs))} times. This should still be fast for ##' LLSimpute because the method allows to choose to only do the ##' estimation for one particular variable. This saves a lot of ##' iterations. Here, \eqn{v_{miss}}{v\_miss} is the number of ##' variables showing missing values. ##' ##' As cross validation is done variable-wise, in this function Q2 is ##' defined on single variables, not on the entire data set. This is ##' Q2 is calculated as as \eqn{\frac{\sum(x - ##' xe)^2}{\sum(x^2)}}{sum(x - xe)^2 \ sum(x^2)}, where x is the ##' currently used variable and xe it's estimate. The values are then ##' averaged over all variables. The NRMSEP is already defined ##' variable-wise. For a single variable it is then ##' \eqn{\sqrt(\frac{\sum(x - xe)^2}{(n \cdot var(x))})}{sqrt(sum(x - ##' xe)^2 \ (n * var(x)))}, where x is the variable and xe it's ##' estimate, n is the length of x. The variable wise estimation ##' errors are returned in parameter variableWiseError. ##' @title Estimate best number of Components for missing value ##' estimation ##' @param Matrix \code{matrix} -- numeric matrix containing ##' observations in rows and variables in columns ##' @param method \code{character} -- of the methods found with ##' pcaMethods() The option llsImputeAll calls llsImpute with the ##' allVariables = TRUE parameter. ##' @param evalPcs \code{numeric} -- The principal components to use ##' for cross validation or the number of neighbour variables if used ##' with llsImpute. Should be an array containing integer values, ##' eg. \code{evalPcs = 1:10} or \code{evalPcs = c(2,5,8)}. The NRMSEP ##' or Q2 is calculated for each component. ##' @param segs \code{numeric} -- number of segments for cross validation ##' @param nruncv \code{numeric} -- Times the whole cross validation ##' is repeated ##' @param em \code{character} -- The error measure. This can be nrmsep or q2 ##' @param allVariables \code{boolean} -- If TRUE, the NRMSEP is ##' calculated for all variables, If FALSE, only the incomplete ones ##' are included. You maybe want to do this to compare several methods ##' on a complete data set. ##' @param verbose \code{boolean} -- If TRUE, some output like the ##' variable indexes are printed to the console each iteration. ##' @param ... Further arguments to \code{pca} or \code{nni} ##' @return A list with: ##' \item{bestNPcs}{number of PCs or k for which the minimal average ##' NRMSEP or the maximal Q2 was obtained.} ##' \item{eError}{an array of of size length(evalPcs). Contains the ##' average error of the cross validation runs for each number of ##' components.} ##' \item{variableWiseError}{Matrix of size ##' \code{incomplete_variables} x length(evalPcs). Contains the ##' NRMSEP or Q2 distance for each variable and each number of PCs. ##' This allows to easily see for wich variables imputation makes ##' sense and for which one it should not be done or mean imputation ##' should be used.} ##' \item{evalPcs}{The evaluated numbers of components or number of ##' neighbours (the same as the evalPcs input parameter).} ##' \item{variableIx}{Index of the incomplete variables. This can be ##' used to map the variable wise error to the original data.} ##' @seealso \code{\link{kEstimateFast}, \link{Q2}, \link{pca}, \link{nni}}. ##' @examples ##' ## Load a sample metabolite dataset with 5\% missing values (metaboliteData) ##' data(metaboliteData) ##' # Do cross validation with ppca for component 2:4 ##' esti <- kEstimate(metaboliteData, method = "ppca", evalPcs = 2:4, nruncv=1, em="nrmsep") ##' # Plot the average NRMSEP ##' barplot(drop(esti$eError), xlab = "Components",ylab = "NRMSEP (1 iterations)") ##' # The best result was obtained for this number of PCs: ##' print(esti$bestNPcs) ##' # Now have a look at the variable wise estimation error ##' barplot(drop(esti$variableWiseError[, which(esti$evalPcs == esti$bestNPcs)]), ##' xlab = "Incomplete variable Index", ylab = "NRMSEP") ##' @keywords multivariate ##' @export ##' @author Wolfram Stacklies kEstimate <- function(Matrix, method="ppca", evalPcs=1:3, segs=3, nruncv=5, em="q2", allVariables=FALSE, verbose=interactive(), ...) { fastKE <- FALSE if (method == "ppca" | method == "bpca" | method == "nipals" | method == "nlpca") fastKE <- TRUE method <- match.arg(method, listPcaMethods()) em <- match.arg(em, c("nrmsep", "q2")) maxPcs <- max(evalPcs) lengthPcs <- length(evalPcs) ## If the data is a data frame, convert it into a matrix Matrix <- as.matrix(Matrix, rownames.force=TRUE) if(maxPcs > (ncol(Matrix) - 1)) stop("maxPcs exceeds matrix size, choose a lower value!") ## And now check if everything is right... if( !checkData(Matrix, verbose=interactive()) ) stop("Invalid data format! Use checkData(Matrix, verbose = TRUE) for details.\n") if( (sum(is.na(Matrix)) == 0) && (allVariables == FALSE) ) stop("No missing values. Maybe you want to set allVariables = TRUE. Exiting\n") missing <- apply(is.na(Matrix), 2, sum) > 0 missIx <- which(missing == TRUE) if (allVariables) missIx <- 1:ncol(Matrix) complete <- !missing compIx <- which(complete == TRUE) error <- matrix(0, length(missIx), length(evalPcs)) iteration <- 0 for(nPcs in evalPcs) { ## If the estimated observations are just scores %*% t(loadings) ## we can calculate all we need at once, this saves many ## iterations... if (fastKE) nPcs = maxPcs iteration = iteration + 1 if (verbose && !fastKE) { cat("Doing CV for ", nPcs, " component(s) \n") } else if (verbose && fastKE) {cat("Doing CV ... \n")} for(cviter in 1:nruncv) { pos <- 0 if (verbose) cat("Incomplete variable index: ") for (index in missIx) { pos <- pos + 1 cat(pos, ":", sep="") target <- Matrix[, index, drop = FALSE] compObs <- !is.na(target) missObs <- is.na(target) nObs <- sum(compObs) ## Remove all observations that are missing in the target genes, ## as additional missing values may tamper the results set <- Matrix[compObs,] if (nObs >= (2 * segs)) { segments <- segs } else segments <- ceiling(nObs / 2) ## We assume uniformly distributed missing values when ## choosing the segments tt <- gl(segments, ceiling(nObs / segments))[1:nObs] cvsegs <- split(sample(nObs), tt) set <- Matrix[compObs,] if (fastKE) { nrmsep <- array(0, length(evalPcs)) q2 <- array(0, length(evalPcs)) } else { nrmsep <- 0; q2 <- 0 } for (i in 1:length(cvsegs)) { n <- length(cvsegs[[i]]) # n is the number of created # missing values ## Impute values using the given regression method testSet <- set testSet[cvsegs[[i]], index] <- NA if (method == "llsImpute") { estimate <- llsImpute(testSet, k = nPcs, verbose = FALSE, allVariables = FALSE, center = FALSE, xval = index) } else if (method == "llsImputeAll") { estimate <- llsImpute(testSet, k = nPcs, verbose = FALSE, allVariables = TRUE, center = FALSE, xval = index) } else { estimate <- pca(testSet, nPcs = nPcs, verbose = FALSE, method = method, center = TRUE,...) } if (fastKE) { for (np in evalPcs) { estiFitted <- fitted(estimate, data = NULL, nPcs = np) estimateVec <- estiFitted[, index] original <- target[compObs, ] estimateVec[-cvsegs[[i]]] <- testSet[-cvsegs[[i]], index] ## Error of prediction, error is calculated for removed ## elements only nIx <- which(evalPcs == np) if (em == "nrmsep") { nrmsep[nIx] <- nrmsep[nIx] + sum( (original - estimateVec)^2) } else { q2[nIx] <- q2[nIx] + sum( (original - estimateVec)^2 ) } } } else { estimate <- estimate@completeObs[, index] original <- target[compObs, ] ## Error of prediction, error is calculated for removed ## elements only if (em == "nrmsep") { nrmsep <- nrmsep + sum( (original - estimate)^2) } else { q2 <- q2 + sum( (original - estimate)^2 ) } } } ## iteration over cv segments if (fastKE) { if (em == "nrmsep") { error[pos, ] <- error[pos, ] + nrmsep / (nrow(set) * var(set[,index])) } else error[pos, ] <- error[pos, ] + (1 - (q2 / sum(set[, index]^2))) } else { if (em == "nrmsep") { error[pos, iteration] <- error[pos, iteration] + nrmsep / (nrow(set) * var(set[,index])) } else error[pos, iteration] <- error[pos, iteration] + (1 - (q2 / sum(set[, index]^2))) } } # iteration over variables if (verbose) cat("\n") } #iteration over nruncv ## The error is the sum over the independent cross validation runs error <- error / nruncv if (verbose && !fastKE) cat("The average", em, "for k =", iteration, "is", sum(error[,iteration]) / nrow(error), "\n") ## if nlpca, ppca, bpca, nipals we do not need to iterate over the ## number of components... if (fastKE) break } # iteration over number components if (em == "nrmsep") avgError <- sqrt(apply(error, 2, sum) / nrow(error)) else avgError <- apply(error, 2, sum) / nrow(error) ret <- list() if (em == "nrmsep") ret$bestNPcs <- evalPcs[which(avgError == min(avgError))] else ret$bestNPcs <- evalPcs[which(avgError == max(avgError))] ret$eError <- avgError if(em == "nrmsep") ret$variableWiseError <- sqrt(error) else ret$variableWiseError <- error ret$evalPcs <- evalPcs ret$variableIx <- missIx return(ret) } pcaMethods/R/kEstimateFast.R0000644000175400017540000001215413556116437016775 0ustar00biocbuildbiocbuild##' This is a simple estimator for the optimal number of componets ##' when applying PCA or LLSimpute for missing value estimation. No ##' cross validation is performed, instead the estimation quality is ##' defined as Matrix[!missing] - Estimate[!missing]. This will give a ##' relatively rough estimate, but the number of iterations equals the ##' length of the parameter evalPcs.\cr Does not work with LLSimpute!! ##' As error measure the NRMSEP (see Feten et. al, 2005) or the Q2 ##' distance is used. The NRMSEP basically normalises the RMSD ##' between original data and estimate by the variable-wise ##' variance. The reason for this is that a higher variance will ##' generally lead to a higher estimation error. If the number of ##' samples is small, the gene - wise variance may become an unstable ##' criterion and the Q2 distance should be used instead. Also if ##' variance normalisation was applied previously. ##' @title Estimate best number of Components for missing value estimation ##' @param Matrix \code{matrix} -- numeric matrix containing ##' observations in rows and variables in columns ##' @param method \code{character} -- a valid pca method (see ##' \code{\link{pca}}). ##' @param evalPcs \code{numeric} -- The principal components to use ##' for cross validation or cluster sizes if used with ##' llsImpute. Should be an array containing integer values, ##' eg. evalPcs = 1:10 or evalPcs = C(2,5,8).The NRMSEP is calculated ##' for each component. ##' @param em \code{character} -- The error measure. This can be ##' nrmsep or q2 ##' @param allVariables \code{boolean} -- If TRUE, the NRMSEP is ##' calculated for all variables, If FALSE, only the incomplete ones ##' are included. You maybe want to do this to compare several methods ##' on a complete data set. ##' @param verbose \code{boolean} -- If TRUE, the NRMSEP and the ##' variance are printed to the console each iteration. ##' @param ... Further arguments to \code{pca} ##' @return \item{list}{Returns a list with the elements: ##' \itemize{ ##' \item minNPcs - number of PCs for which the minimal average NRMSEP ##' was obtained ##' \item eError - an array of of size length(evalPcs). Contains the ##' estimation error for each number of ##' components. ##' \item evalPcs - The evaluated numbers of components or ##' cluster sizes (the same as the evalPcs input parameter). }} ##' @seealso \code{\link{kEstimate}}. ##' @export ##' @examples ##' data(metaboliteData) ##' # Estimate best number of PCs with ppca for component 2:4 ##' esti <- kEstimateFast(t(metaboliteData), method = "ppca", evalPcs = 2:4, em="nrmsep") ##' barplot(drop(esti$eError), xlab = "Components",ylab = "NRMSEP (1 iterations)") ##' # The best k value is: ##' print(esti$minNPcs) ##' @keywords multivariate ##' @author Wolfram Stacklies kEstimateFast <- function(Matrix, method = "ppca", evalPcs = 1:3, em = "nrmsep", allVariables = FALSE, verbose = interactive(), ...) { method <- match.arg(method, c("ppca", "bpca", "svdImpute", "nipals", "nlpca")) em <- match.arg(em, c("nrmsep", "q2")) maxPcs <- max(evalPcs) lengthPcs <- length(evalPcs) missing <- is.na(Matrix) error <- array(0, lengthPcs) ## If the data is a data frame, convert it into a matrix Matrix <- as.matrix(Matrix, rownames.force=TRUE) if(maxPcs > (ncol(Matrix) - 1)) stop("maxPcs exceeds matrix size, choose a lower value!") ## And now check if everything is right... if( !checkData(Matrix, verbose=interactive()) ) stop("Invalid data format! Use checkData(Matrix, verbose = TRUE) for details.\n") if( (sum(is.na(Matrix)) == 0) && (allVariables == FALSE) ) stop("No missing values. Maybe you want to set allVariables = TRUE. Exiting\n") iteration = 0 for(nPcs in evalPcs) { iteration = iteration + 1 if (method == "nlpca") { estimate <- fitted(pca(Matrix, nPcs = nPcs, verbose = FALSE, method = method, center = TRUE,...), Matrix, nPcs = nPcs) } else { estimate <- fitted(pca(Matrix, nPcs = nPcs, verbose = FALSE, method = method, center = TRUE,...), nPcs = nPcs) } if (em == "q2") { # The Q2 distance q2 <- 1 - sum((Matrix[!missing] - estimate[!missing])^2) / sum(Matrix[!missing]^2) error[iteration] <- q2 } else { nrmsep <- 0 for(i in 1:ncol(Matrix)) { nrmsep <- nrmsep + ( sum((Matrix[!missing[,i], i] - estimate[!missing[,i], i])^2) / (sum(!missing[,i]) * var(Matrix[,i], na.rm = TRUE)) ) } nrmsep <- nrmsep / sum(apply(missing, 2, sum) > 0) error[iteration] <- nrmsep } if(verbose) cat("The", em, "for", evalPcs[iteration], "components is:", error[iteration], "\n") } ret <- list() if (em == "nrmsep") ret$bestNPcs <- evalPcs[which(error == min(error))] else ret$bestNPcs <- evalPcs[which(error == max(error))] ret$eError <- error ret$evalPcs <- evalPcs return(ret) } pcaMethods/R/lineSearch.R0000644000175400017540000000615213556116437016307 0ustar00biocbuildbiocbuild##' Line search for conjugate gradient ##' @param nlnet The nlnet ##' @param dw .. ##' @param e0 .. ##' @param ttGuess .. ##' @param trainIn Training data ##' @param trainOut Fitted data ##' @param verbose logical, print messages ##' @return ... ##' @author Henning Redestig, Matthias Scholz lineSearch <- function(nlnet, dw, e0, ttGuess, trainIn, trainOut, verbose) { iterGoldenSectionSearch <- 6 alpha <- 0.618034 tt <- rep(0, 4) e <- rep(0, 4) tmpnlnet <- forkNlpcaNet(nlnet) tt[1] <- 0 e[1] <- e0 tt[4] <- ttGuess tmpnlnet@weights$set(nlnet@weights$current() + tt[4] * dw) e[4] <- nlnet@error(tmpnlnet, trainIn, trainOut)$error if(e[4] > e[1]) { #got final interval calculate tt[2] and tt[3] tt[2] <- tt[1] + (1 - alpha) * (tt[4] - tt[1]) tmpnlnet@weights$set(nlnet@weights$current() + tt[2] * dw) e[2] <- nlnet@error(tmpnlnet, trainIn, trainOut)$error tt[3] <- tt[1] + alpha * (tt[4] - tt[1]) tmpnlnet@weights$set(nlnet@weights$current() + tt[3] * dw) e[3] <- nlnet@error(tmpnlnet, trainIn, trainOut)$error } else { #expand, add new tt[4] tt[3] <- tt[4] e[3] <- e[4] tt[4] <- (1 + alpha) * tt[4] tmpnlnet@weights$set(nlnet@weights$current() + tt[4] * dw) e[4] <- nlnet@error(tmpnlnet, trainIn, trainOut)$error if(e[4] > e[3]) { #got final interval, calculate tt[2] tt[2] <- tt[1] + (1 - alpha) * (tt[4] - tt[1]) tmpnlnet@weights$set(nlnet@weights$current() + tt[2] * dw) e[2] <- nlnet@error(tmpnlnet, trainIn, trainOut)$error } else { #expand: add new tt[4] i <- 1 while(e[4] < e[3] && i < 50) { tt[2] <- tt[3] e[2] <- e[3] tt[3] <- tt[4] e[3] <- e[4] tt[4] <- (1 + alpha) * tt[4] tmpnlnet@weights$set(nlnet@weights$current() + tt[4] * dw) e[4] <- nlnet@error(tmpnlnet, trainIn, trainOut)$error i <- i + 1 if(verbose && i == 50) cat("^") } } } ## golden section search for(i in 1:iterGoldenSectionSearch) { if(e[3] > e[2]) { tt[4] <- tt[3] #remove right value tt[4] e[4] <- e[3] tt[3] <- tt[2] e[3] <- e[2] tt[2] <- tt[1] + (1 - alpha) * (tt[4] - tt[1]) #split left interval tmpnlnet@weights$set(nlnet@weights$current() + tt[2] * dw) e[2] <- nlnet@error(tmpnlnet, trainIn, trainOut)$error } else { tt[1] <- tt[2] #remove left t value tt[1] e[1] <- e[2] tt[2] <- tt[3] e[2] <- e[3] tt[3] <- tt[1] + alpha * (tt[4] - tt[1]) #split right interval tmpnlnet@weights$set(nlnet@weights$current() + tt[3] * dw) e[3] <- nlnet@error(tmpnlnet, trainIn, trainOut)$error } } if(e[2] < e[3]) { eBest <- e[2] ttBest <- tt[2] } else { eBest <- e[3] ttBest <- tt[3] } wBest <- nlnet@weights$current() + ttBest * dw return(list(wBest=wBest, eBest=eBest, ttBest=ttBest)) } ##' Linear kernel ##' @param x datum ##' @return Input value ##' @author Henning Redestig, Matthias Scholz linr <- function(x) x pcaMethods/R/llsImpute.R0000644000175400017540000002204113556116437016203 0ustar00biocbuildbiocbuild##' Missing value estimation using local least squares (LLS). First, ##' k variables (for Microarrya data usually the genes) are selected ##' by pearson, spearman or kendall correlation coefficients. Then ##' missing values are imputed by a linear combination of the k ##' selected variables. The optimal combination is found by LLS ##' regression. The method was first described by Kim et al, ##' Bioinformatics, 21(2),2005. ##' ##' Missing values are denoted as \code{NA}\cr It is not recommended ##' to use this function directely but rather to use the nni() wrapper ##' function. The methods provides two ways for missing value ##' estimation, selected by the \code{allVariables} option. The first ##' one is to use only complete variables for the regression. This is ##' preferable when the number of incomplete variables is relatively ##' small. ##' ##' The second way is to consider all variables as candidates for the ##' regression. Hereby missing values are initially replaced by the ##' columns wise mean. The method then iterates, using the current ##' estimate as input for the regression until the change between new ##' and old estimate falls below a threshold (0.001). ##' ##' @title LLSimpute algorithm ##' @param Matrix \code{matrix} -- Data containing the variables ##' (genes) in columns and observations (samples) in rows. The data ##' may contain missing values, denoted as \code{NA}. ##' @param k \code{numeric} -- Cluster size, this is the number of ##' similar genes used for regression. ##' @param center \code{boolean} -- Mean center the data if TRUE ##' @param completeObs \code{boolean} -- Return the estimated complete ##' observations if TRUE. This is the input data with NA values ##' replaced by the estimated values. ##' @param correlation \code{character} -- How to calculate the ##' distance between genes. One out of pearson | kendall | spearman , ##' see also help("cor"). ##' @param allVariables \code{boolean} -- Use only complete genes to ##' do the regression if TRUE, all genes if FALSE. ##' @param maxSteps \code{numeric} -- Maximum number of iteration ##' steps if allGenes = TRUE. ##' @param xval \code{numeric} Use LLSimpute for cross ##' validation. xval is the index of the gene to estimate, all other ##' incomplete genes will be ignored if this parameter is set. We do ##' not consider them in the cross-validation. ##' @param verbose \code{boolean} -- Print step number and relative ##' change if TRUE and allVariables = TRUE ##' @param ... Reserved for parameters used in future version of the ##' algorithm ##' @note Each step the generalized inverse of a \code{miss} x k ##' matrix is calculated. Where \code{miss} is the number of missing ##' values in variable j and \code{k} the number of neighbours. This ##' may be slow for large values of k and / or many missing ##' values. See also help("ginv"). ##' @return \item{nniRes}{Standard nni (nearest neighbour ##' imputation) result object of this package. See ##' \code{\link{nniRes}} for details.} ##' @seealso \code{\link{pca}, \link{nniRes}, \link{nni}}. ##' @examples ##' ## Load a sample metabolite dataset (metaboliteData) with already 5\% of ##' ## data missing ##' data(metaboliteData) ##' ## Perform llsImpute using k = 10 ##' ## Set allVariables TRUE because there are very few complete variables ##' result <- llsImpute(metaboliteData, k = 10, correlation="pearson", allVariables=TRUE) ##' ## Get the estimated complete observations ##' cObs <- completeObs(result) ##' @keywords multivariate ##' @export ##' @references Kim, H. and Golub, G.H. and Park, H. - Missing value ##' estimation for DNA microarray gene expression data: local least ##' squares imputation. \emph{Bioinformatics, 2005; 21(2):187-198.} ##' ##' Troyanskaya O. and Cantor M. and Sherlock G. and Brown P. and ##' Hastie T. and Tibshirani R. and Botstein D. and Altman RB. - ##' Missing value estimation methods for DNA microarrays. ##' \emph{Bioinformatics. 2001 Jun;17(6):520-525.} ##' @author Wolfram Stacklies llsImpute <- function(Matrix, k=10, center=FALSE, completeObs=TRUE, correlation="pearson", allVariables=FALSE, maxSteps=100, xval=NULL, verbose=FALSE, ...) { threshold <- 0.001 correlation <- match.arg(correlation, c("pearson", "kendall", "spearman")) ## If the data is a data frame, convert it into a matrix Matrix <- as.matrix(Matrix, rownames.force=TRUE) ## And now check if everything is right... if ( !checkData(Matrix, verbose = interactive()) ) { stop("Invalid data format! Use checkData(Matrix, verbose = TRUE) for details.\n") } ## Exit if number of neighbours exceeds number of columns if (k > ncol(Matrix)) stop("Cluster size larger than the number of columns, choose a k < ncol(Matrix)!") ## Set allVariables TRUE if k exceeds number of complete genes ## Print warning messages in the first case and when less than 50% of all genes are complete ## and allVariables == FALSE cg <- sum( apply(is.na(Matrix), 2, sum) == 0) if ( (k > cg) && (!allVariables) ) { warning("Cluster size larger than number of complete genes, using allVariables = TRUE") allVariables <- TRUE } else if ( (cg < (ncol(Matrix) / 2)) && (!allVariables) ) { warning("Less than 50% of the genes are complete, consider using allVariables = TRUE") } else if (sum(is.na(Matrix)) == 0) stop("No missing values, no need for missing value imputation :))") ## Find all genes with missing values missing <- apply(is.na(Matrix), 2, sum) > 0 missIx <- which(missing == TRUE) # For cross validation we want to only estimate one variable, the others # are not considered in the cross validation anyway if (!is.null(xval)) missIx = xval obs <- Matrix ## working copy of the data Ye <- Matrix ## Estimated complete observations ## Center the data column wise if (center) { obs <- scale(Matrix, center = TRUE, scale = FALSE) Ye <- obs means <- attr(Ye, "scaled:center") } if (allVariables) { compIx <- 1:ncol(obs) ## Impute the row average rowMeans <- apply(obs, 1, mean, na.rm = TRUE) for (i in 1:nrow(obs)) { obs[i, is.na(Matrix[i,])] <- rowMeans[i] } ## distances between all genes, ignore the diagonal (correlation to itself) distance = abs(cor(obs, obs, method = correlation)) } else { compIx <- which(missing == FALSE) ## missing genes are the rows, complete genes the columns distance = abs(cor(obs[,missIx, drop=FALSE], obs[,compIx, drop=FALSE], use="pairwise.complete.obs", method = correlation)) } change <- Inf step <- 0 while ( (change > threshold) && (step < maxSteps) ) { step <- step + 1 iteration <- 0 ## Do the regression and imputation for (index in missIx) { iteration <- iteration + 1 if (allVariables) { similar <- sort(distance[iteration,], index.return = TRUE, decreasing = TRUE) simIx <- compIx[ similar$ix[similar$ix != iteration][1:k] ] } else { similar <- sort(distance[iteration,], index.return = TRUE, decreasing = TRUE) simIx <- compIx[ similar$ix[1:k] ] } ## ## Do a regression against the k most similar genes ## See Kim et. al 2005 for details ## target <- obs[, index, drop = FALSE] tMiss <- is.na(Matrix[, index, drop = FALSE]) Apart <- obs[!tMiss, simIx, drop = FALSE] Bpart <- obs[tMiss, simIx, drop = FALSE] targetComplete <- target[!tMiss, , drop = FALSE] X <- MASS::ginv(Apart) %*% targetComplete estimate <- Bpart %*% X ## Impute the estimate Ye[tMiss, index] <- estimate } ## We do not want to iterate if allVariables == FALSE if (!allVariables || !is.null(xval)) { break } else { ## relative change in estimation change <- sqrt(sum( (obs - Ye)^2 ) / sum(obs^2)) obs <- Ye if (verbose) { cat("Step number : ", step, '\n') cat("Relative change : ", change, '\n') cat("---------------", '\n') } } } ## Add the original mean if (center) { for(i in 1:ncol(Ye)) { Ye[,i] <- Ye[,i] + means[i] } } ## Build the nniRes object ## result <- new("nniRes") if(completeObs) { Ye[!is.na(Matrix)] <- Matrix[!is.na(Matrix)] result@completeObs <- Ye } result@centered <- center result@center <- attr(scale(Matrix, center = TRUE, scale = FALSE), "scaled:center") result@nObs <- nrow(Matrix) result@nVar <- ncol(Matrix) result@method <- "llsImpute" result@correlation <- correlation result@k <- k result@missing <- sum(is.na(Matrix)) return(result) } pcaMethods/R/methods-ExpressionSet.R0000644000175400017540000000272613556116437020511 0ustar00biocbuildbiocbuild##' This function can be used to conveniently replace the expression ##' matrix in an \code{ExpressionSet} with the completed data from a ##' \code{pcaRes} object. ##' ##' This is not a standard \code{as} function as \code{pcaRes} ##' object alone not can be converted to an \code{ExpressionSet} (the ##' \code{pcaRes} object does not hold any \code{phenoData} for ##' example). ##' @title Convert pcaRes object to an expression set ##' @param object \code{pcaRes} -- The object containing the completed ##' data. ##' @param exprSet \code{ExpressionSet} -- The object passed on to ##' \code{pca} for missing value estimation. ##' @return An object without missing values of class \code{ExpressionSet}. ##' @export ##' @author Wolfram Stacklies \cr CAS-MPG Partner Institute for ##' Computational Biology, Shanghai, China ##' @keywords multivariate asExprSet <- function(object, exprSet) { if(!inherits(exprSet, "ExpressionSet")) stop("Parameter exprSet must be of type ExpressionSet") if(!inherits(object, "pcaRes") & !inherits(object, "nniRes")) stop("Parameter object must be either of type pcaRes or nniRes") if (is.null(completeObs(object))) stop("completeObs(object) is NULL, exiting") if(!all(dim(exprs(exprSet)) == dim(t(completeObs(object))))) stop("Dimensions of exprs(exprSet) and completeObs(object) do not match. Did you really do missing value estimation using this ExpressionSet object?") exprs(exprSet) <- t(completeObs(object)) return(exprSet) } pcaMethods/R/methods-nniRes.R0000644000175400017540000000161113556116437017124 0ustar00biocbuildbiocbuild##' Print a brief description of nniRes model ##' @title Print a nniRes model ##' @param x An \code{nniRes} object ##' @param ... Not used ##' @return Nothing, used for side-effect ##' @export ##' @author Henning Redestig showNniRes <- function(x, ...) { summary(x) cat(dim(x)["nVar"], "\tVariables\n") cat(dim(x)["nObs"],"\tSamples\n") cat(nmissing(x), "\tNAs (", round(100 * nmissing(x) / (nObs(x) * nVar(x)), getOption("str")$digits.d), "%)\n") cat("k was set to", x@k, "\n") if(centered(x)) cat("Data was mean centered before running LLSimpute \n") else cat("Data was NOT mean centered before running LLSimpute \n") if(scaled(x)) cat("Data was scaled before running LLSimpute \n") else cat("Data was NOT scaled before running LLSimpute \n") } setMethod("print", "nniRes", showNniRes) setMethod("show", "nniRes", function(object) showNniRes(object)) pcaMethods/R/methods-pcaRes.R0000644000175400017540000007727513556116437017126 0ustar00biocbuildbiocbuild##' The leverages of PCA model indicate how much influence each ##' observation has on the PCA model. Observations with high leverage ##' has caused the principal components to rotate towards them. It can ##' be used to extract both "unimportant" observations as well as ##' picking potential outliers. ##' ##' Defined as \eqn{Tr(T(T'T)^{-1}T')}{Tr(T(T'T)^(-1)T')} ##' @title Extract leverages of a PCA model ##' @param object a \code{pcaRes} object ##' @return The observation leverages as a numeric vector ##' @references Introduction to Multi- and Megavariate Data Analysis ##' using Projection Methods (PCA and PLS), L. Eriksson, E. Johansson, ##' N. Kettaneh-Wold and S. Wold, Umetrics 1999, p. 466 ##' @examples ##' data(iris) ##' pcIr <- pca(iris[,1:4]) ##' ## versicolor has the lowest leverage ##' with(iris, plot(leverage(pcIr)~Species)) ##' @keywords multivariate ##' @aliases leverage leverage,pcaRes-method ##' @author Henning Redestig setMethod("leverage", "pcaRes", function(object) { diag(scores(object) %*% solve(crossprod(scores(object))) %*% t(scores(object))) }) ##' Distance to the model of X-space. ##' ##' Measures how well described the observations are, i.e. how well ##' they fit in the mode. High DModX indicate a poor fit. Defined as: ##' ##' \eqn{\frac{\sqrt{\frac{SSE_i}{K-A}}}{\sqrt{\frac{SSE}{(N-A-A_0)(K-A)}}}} ##' ##' For observation \eqn{i}, in a model with \eqn{A} components, ##' \eqn{K} variables and \eqn{N} obserations. SSE is the squared sum ##' of the residuals. \eqn{A_0} is 1 if model was centered and 0 ##' otherwise. DModX is claimed to be approximately F-distributed and ##' can therefore be used to check if an observation is significantly ##' far away from the PCA model assuming normally distributed data. ##' ##' Pass original data as an argument if the model was calculated with ##' \code{completeObs=FALSE}. ##' @title DModX ##' @usage DModX(object, dat, newdata=FALSE, type=c("normalized","absolute"), ...) ##' @param object a pcaRes object ##' @param dat the original data, taken from \code{completeObs} if ##' left missing. ##' @param newdata logical indicating if this data was part of the ##' training data or not. If it was, it is adjusted by a near one factor ##' \eqn{v=(N/ (N-A-A0))^-1} ##' @param type if absolute or normalized values should be ##' given. Normalized values are adjusted to the the total RSD of the ##' model. ##' @param ... Not used ##' @return A vector with distances from observations to the PCA model ##' @aliases DModX DModX,pcaRes-method ##' @examples ##' data(iris) ##' pcIr <- pca(iris[,1:4]) ##' with(iris, plot(DModX(pcIr)~Species)) ##' @references Introduction to Multi- and Megavariate Data Analysis ##' using Projection Methods (PCA and PLS), L. Eriksson, E. Johansson, ##' N. Kettaneh-Wold and S. Wold, Umetrics 1999, p. 468 ##' @author Henning Redestig setMethod("DModX", "pcaRes", function(object, dat, newdata=FALSE, type=c("normalized","absolute"), ...) { type <- match.arg(type) if(missing(dat)) { if(!is.null(completeObs(object))) dat <- completeObs(object) else stop("missing data when calculating DModX") } A0 <- as.integer(centered(object)) ny <- ifelse(newdata, 1, sqrt(nObs(object) / (nObs(object) - nP(object) - A0))) E2 <- resid(object, dat)^2 s <- sqrt(rowSums(E2) / (nVar(object) - nP(object))) * ny if(type == "absolute") return(s) s0 <- sqrt(sum(E2) / ((nObs(object) - nP(object) - A0) * (nVar(object) - nP(object)))) s / s0 }) ##' Get number of PCs ##' @param object pcaRes object ##' @param ... not used ##' @return Number of PCs ##' @aliases nP nP,pcaRes-method ##' @usage nP(object, ...) ##' @author Henning Redestig setMethod("nP", "pcaRes", function(object, ...) { if(is.null(object@nPcs) & !is.null(scores(object))) return(ncol(scores(object))) object@nPcs }) ##' Get cross-validation statistics (e.g. \eqn{Q^2}). ##' @param object pcaRes object ##' @param ... not used ##' @return vector CV statistics ##' @aliases cvstat cvstat,pcaRes-method ##' @usage cvstat(object, ...) ##' @author Henning Redestig setMethod("cvstat", "pcaRes", function(object, ...) { object@cvstat }) ##' Get number of PCs. ##' @param object pcaRes object ##' @param ... not used ##' @note Try to use \code{link{nP}} instead since \code{nPcs} tend to ##' clash with argument names. ##' @return Number of PCs ##' @usage nPcs(object, ...) ##' @aliases nPcs nPcs,pcaRes-method ##' @author Henning Redestig setMethod("nPcs", "pcaRes", function(object, ...) { nP(object) }) ##' Get the number of observations used to build the PCA model. ##' @param object pcaRes object ##' @param ... Not used ##' @usage nObs(object, ...) ##' @aliases nObs nObs,pcaRes-method ##' @return Number of observations ##' @author Henning Redestig setMethod("nObs", "pcaRes", function(object, ...) { object@nObs }) ##' Get the number of variables used to build the PCA model. ##' @param object pcaRes object ##' @param ... Not used ##' @usage nVar(object, ...) ##' @aliases nVar nVar,pcaRes-method ##' @return Number of variables ##' @author Henning Redestig setMethod("nVar", "pcaRes", function(object, ...) { object@nVar }) ##' Check centering was part of the model ##' @param object pcaRes object ##' @param ... Not used ##' @usage centered(object, ...) ##' @aliases centered centered,pcaRes-method ##' @return TRUE if model was centered ##' @author Henning Redestig setMethod("centered", "pcaRes", function(object, ...) { if(is.null(object@centered)) return(FALSE) object@centered }) ##' Get the centers of the original variables ##' @param object pcaRes object ##' @param ... Not used ##' @usage center(object, ...) ##' @aliases center center,pcaRes-method ##' @return Vector with the centers ##' @author Henning Redestig setMethod("center", "pcaRes", function(object, ...) { object@center }) setMethod("completeObs", "pcaRes", function(object, ...) { object@completeObs }) ##' Get the original data with missing values replaced with predicted ##' values. ##' @param object object to fetch complete data from ##' @param ... Not used ##' @usage completeObs(object, ...) ##' @aliases completeObs completeObs,nniRes-method ##' completeObs,pcaRes-method ##' @return Completed data (matrix) ##' @author Henning Redestig setMethod("completeObs", "nniRes", function(object, ...) { object@completeObs }) ##' Get the used PCA method ##' @param object pcaRes object ##' @param ... Not used ##' @usage method(object, ...) ##' @aliases method method,pcaRes-method ##' @return The used pca method ##' @author Henning Redestig setMethod("method", "pcaRes", function(object, ...) { object@method }) setMethod("nmissing", "nniRes", function(object, ...) { sum(object@missing) }) ##' Missing values ##' @param object pcaRes object ##' @param ... Not used ##' @usage nmissing(object, ...) ##' @aliases nmissing nmissing,pcaRes-method nmissing,nniRes-method ##' @return Get the number of missing values ##' @author Henning Redestig setMethod("nmissing", "pcaRes", function(object, ...) { sum(object@missing) }) ##' Get a matrix with indicating the elements that were missing in the ##' input data. Convenient for estimating imputation performance. ##' @param object pcaRes object ##' @param ... Not used ##' @usage wasna(object, ...) ##' @aliases wasna wasna,pcaRes-method ##' @return A matrix with logicals ##' @examples ##' data(metaboliteData) ##' data(metaboliteDataComplete) ##' result <- pca(metaboliteData, nPcs=2) ##' plot(completeObs(result)[wasna(result)], metaboliteDataComplete[wasna(result)]) ##' @author Henning Redestig setMethod("wasna", "pcaRes", function(object, ...) { object@missing }) ##' Get the standard deviations of the scores (indicates their ##' relevance) ##' @param object pcaRes object ##' @param ... Not used ##' @usage sDev(object, ...) ##' @aliases sDev sDev,pcaRes-method ##' @return Standard devations of the scores ##' @author Henning Redestig setMethod("sDev", "pcaRes", function(object, ...) { object@sDev }) ##' Check if scaling was part of the PCA model ##' @param object pcaRes object ##' @param ... Not used ##' @usage scaled(object, ...) ##' @aliases scaled scaled,pcaRes-method ##' @return TRUE if scaling was part of the PCA model ##' @author Henning Redestig setMethod("scaled", "pcaRes", function(object, ...) { if(is.null(object@scaled)) return(FALSE) object@scaled != "none" }) ##' Get the scales (e.g. standard deviations) of the original ##' variables ##' @param object pcaRes object ##' @param ... Not used ##' @usage scl(object, ...) ##' @aliases scl scl,pcaRes-method ##' @return Vector with the scales ##' @seealso \code{\link{prep}} ##' @author Henning Redestig setMethod("scl", "pcaRes", function(object, ...) { object@scale }) ##' Cumulative R2 is the total ratio of variance that is being ##' explained by the model ##' @param object a \code{pcaRes} model ##' @param ... Not used ##' @return Get the cumulative R2 ##' @aliases R2cum R2cum,pcaRes-method ##' @author Henning Redestig setMethod("R2cum", "pcaRes", function(object, ...) { object@R2cum }) ##' Get scores from a pcaRes object ##' @param object a pcaRes object ##' @param ... not used ##' @return The scores as a matrix ##' @export ##' @author Henning Redestig ##' @method scores pcaRes scores.pcaRes <- function(object, ...) object@scores ##' Get scores from a pcaRes object ##' @param object a pcaRes object ##' @param ... not used ##' @return The scores as a matrix ##' @seealso \code{\link{scores.pcaRes}} ##' @aliases scores scores,pcaRes-method ##' @author Henning Redestig setMethod("scores", "pcaRes", scores.pcaRes) ##' Get loadings from a pcaRes object ##' @param object a pcaRes object ##' @param ... not used ##' @return The loadings as a matrix ##' @export ##' @author Henning Redestig ##' @method loadings pcaRes loadings.pcaRes <- function(object, ...) object@loadings ##' Get loadings from a pcaRes object ##' @param object a pcaRes object ##' @param ... not used ##' @return The loadings as a matrix ##' @seealso \code{\link{loadings.pcaRes}} ##' @author Henning Redestig ##' @aliases loadings,pcaRes-method setMethod("loadings", "pcaRes", loadings.pcaRes) ##' Crude way to unmask the function with the same name from ##' \code{stats} ##' @param object any object ##' @param ... not used ##' @return The loadings ##' @author Henning Redestig ##' @aliases loadings loadings,ANY-method setMethod("loadings", "ANY", function(object,...) { stats::loadings(object) }) ##' Dimensions of a PCA model ##' @param x a pcaRes object ##' @return Get the dimensions of this PCA model ##' @method dim pcaRes ##' @export ##' @author Henning Redestig dim.pcaRes <- function(x) { res <- c(nObs(x), nVar(x), nP(x)) names(res) <- c("nObs", "nVar", "nPcs") res } ##' Print basic information about pcaRes object ##' @title Print/Show for pcaRes ##' @param x a pcaRes object ##' @param ... not used ##' @return nothing, used for its side effect ##' @name show-methods ##' @export ##' @author Henning Redestig showPcaRes <- function(x, ...) { summary(x) cat(nVar(x), "\tVariables\n") cat(nObs(x),"\tSamples\n") cat(nmissing(x), "\tNAs (", round(100 * nmissing(x) / (nObs(x) * nVar(x)), getOption("str")$digits.d), "%)\n") cat(nP(x), "\tCalculated component(s)\n") if(centered(x)) cat("Data was mean centered before running PCA \n") else cat("Data was NOT mean centered before running PCA \n") if(scaled(x)) cat("Data was scaled before running PCA \n") else cat("Data was NOT scaled before running PCA \n") cat("Scores structure:\n") print(dim(scores(x))) cat("Loadings structure:\n") if(method(x) == "nlpca") { cat("Inverse hierarchical neural network architecture\n") cat(drop(x@network@net), "\n") cat("Functions in layers\n") cat(x@network@fct, "\n") cat("hierarchic layer:", x@network@hierarchic$layer, "\n") cat("hierarchic coefficients:", x@network@hierarchic$var, "\n") cat("scaling factor:", x@network@scalingFactor, "\n") } else{ print(dim(loadings(x))) } } ##' @aliases print,pcaRes-method print,nniRes-method ##' @name show-methods setMethod("print", "pcaRes", showPcaRes) ## @importFrom methods show ##' @aliases show,pcaRes-method show,nniRes-method ##' @param object the object to print information about ##' @name show-methods setMethod("show", "pcaRes", function(object) showPcaRes(object)) ##' Visualize two-components simultaneously ##' ##' This is a method for the generic function 'biplot'. There is ##' considerable confusion over the precise definitions: those of the ##' original paper, Gabriel (1971), are followed here. Gabriel and ##' Odoroff (1990) use the same definitions, but their plots actually ##' correspond to \code{pc.biplot = TRUE}. ##' @title Plot a overlaid scores and loadings plot ##' @param x a pcaRes object ##' @param choices which two pcs to plot ##' @param scale The variables are scaled by ##' \eqn{\lambda^{scale}}{lambda^scale} and the observations are ##' scaled by \eqn{\lambda^{scale}}{lambda ^ (1-scale)} where ##' \code{lambda} are the singular values as computed by ##' \code{princomp}. Normally \eqn{0\le{}scale\le{}1}{0 <= scale <= ##' 1}, and a warning will be issued if the specified 'scale' is ##' outside this range. ##' @param pc.biplot If true, use what Gabriel (1971) refers to as a ##' "principal component biplot", with \eqn{\lambda=1}{lambda = 1} and ##' observations scaled up by sqrt(n) and variables scaled down by ##' sqrt(n). Then the inner products between variables approximate ##' covariances and distances between observations approximate ##' Mahalanobis distance. ##' @param ... optional arguments to be passed to ##' \code{biplot.default}. ##' @return a plot is produced on the current graphics device. ##' @method biplot pcaRes ##' @export ##' @examples ##' data(iris) ##' pcIr <- pca(iris[,1:4]) ##' biplot(pcIr) ##' @seealso \code{prcomp}, \code{pca}, \code{princomp} ##' @author Kevin Wright, Adapted from \code{biplot.prcomp} ##' @keywords multivariate ##' @name biplot-methods biplot.pcaRes <- function(x, choices=1:2, scale=1, pc.biplot=FALSE, ...) { if(length(choices)!=2) stop("length of choices must be 2") scores <- scores(x) n <- nrow(scores) lam <- sDev(x)[choices] * sqrt(n) if(scale < 0 || scale > 1) warning("'scale' is outside [0,1]") if(scale != 0) lam <- lam^scale else lam <- 1 if(pc.biplot) lam <- lam/sqrt(n) biplot(t(t(scores[,choices])/lam), t(t(loadings(x)[, choices]) * lam), , ...) invisible() } ##' @aliases biplot,pcaRes-method ## @importFrom stats biplot ##' @name biplot-methods setMethod("biplot", "pcaRes", biplot.pcaRes) ##' Flexible calculation of R2 goodness of fit. ##' @title R2 goodness of fit ##' @param object a PCA model object ##' @param direction choose between calculating R2 per variable, per ##' observation or for the entire data with 'variables', ##' 'observations' or 'complete'. ##' @param data the data used to fit the model ##' @param pcs the number of PCs to use to calculate R2 ##' @aliases R2VX R2VX,pcaRes-method ##' @examples ##' R2VX(pca(iris)) ##' @return A vector with R2 values ##' @author Henning Redestig setMethod('R2VX', 'pcaRes', function(object, direction=c('variables', 'observations', 'complete'), data=completeObs(object), pcs=nP(object)) { direction <- match.arg(direction) if(is.null(data)) stop('missing input when calculating R2') if(any(is.na(data))) stop('missing values not allowed for calculating R2') dat <- prep(data, scale=scl(object), center=center(object)) xhat <- resid(object, pcs=pcs, data=dat, pre=FALSE, post=FALSE) switch(direction, variables={ 1 - colSums(xhat^2) / colSums(dat^2) }, observations={ 1 - rowSums(xhat^2) / rowSums(dat^2) }, complete={ 1 - sum(xhat^2) / sum(dat^2) }) }) setAs('pcaRes', 'data.frame', function(from) { tt <- scores(from) pp <- loadings(from) if(is.null(rownames(tt))) rownames(tt) <- 1:nrow(tt) if(is.null(rownames(pp))) rownames(pp) <- 1:nrow(pp) dfs <- as.data.frame(tt) dfs$names <- rownames(tt) dfs$type <- 'scores' dfl <- as.data.frame(pp) dfl$names <- rownames(pp) dfl$type <- 'loadings' rownames(dfl) <- rownames(dfs) <- NULL rbind(dfl, dfs) }) ##' Print a brief description of the PCA model ##' @title Summary of PCA model ##' @param object a pcaRes object ##' @param ... Not used ##' @return Nothing, used for side-effect ##' @aliases summary summary.pcaRes summary,pcaRes-method ##' @author Henning Redestig ##' @export ##' @name summary ##' @method summary pcaRes summary.pcaRes <- function(object, ...){ cat(method(object), "calculated PCA\n") cat("Importance of component(s):\n") prop <- vector(length=length(R2cum(object)), mode="numeric") prop[1] <- R2cum(object)[1] if (length(R2cum(object)) > 1) { for (i in 2:length(prop)) { prop[i] <- R2cum(object)[i] - R2cum(object)[i-1] } } r <- rbind(prop, R2cum(object)) rownames(r) <- c("R2", "Cumulative R2") colnames(r) <- paste("PC", 1:nP(object), sep="") print(r, digits=4) invisible(r) } setMethod("summary", "pcaRes", summary.pcaRes) ##' Predict data using PCA model ##' ##' This function extracts the predict values from a pcaRes object for ##' the PCA methods SVD, Nipals, PPCA and BPCA. Newdata is first ##' centered if the PCA model was and then scores (\eqn{T}) and data ##' (\eqn{X}) is 'predicted' according to : ##' \eqn{\hat{T}=X_{new}P}{That=XnewP} ##' \eqn{\hat{X}_{new}=\hat{T}P'}{Xhat=ThatP'}. Missing values are ##' set to zero before matrix multiplication to achieve NIPALS like ##' treatment of missing values. ##' @title Predict values from PCA. ##' @param object \code{pcaRes} the \code{pcaRes} object of interest. ##' @param newdata \code{matrix} new data with same number of columns ##' as the used to compute \code{object}. ##' @param pcs \code{numeric} The number of PC's to consider ##' @param pre pre-process \code{newdata} based on the pre-processing ##' chosen for the PCA model ##' @param post unpre-process the final data (add the center back etc) ##' @param ... Not passed on anywhere, included for S3 consistency. ##' @return A list with the following components: \item{scores}{The ##' predicted scores} \item{x}{The predicted data} ##' @method predict pcaRes ##' @keywords multivariate ##' @examples ##' data(iris) ##' hidden <- sample(nrow(iris), 50) ##' pcIr <- pca(iris[-hidden,1:4]) ##' pcFull <- pca(iris[,1:4]) ##' irisHat <- predict(pcIr, iris[hidden,1:4]) ##' cor(irisHat$scores[,1], scores(pcFull)[hidden,1]) ##' @export ##' @name predict-methods ##' @author Henning Redestig predict.pcaRes <- function(object, newdata, pcs=nP(object), pre=TRUE, post=TRUE, ...) { if(!method(object) %in% listPcaMethods("linear")) stop("predict method not implemented for that type of PCA") if(pre) newdata <- prep(newdata, scl(object), center(object)) ## set na's to zero to achieve NIPALS like prediction newdata[is.na(newdata)] <- 0 tnew <- newdata %*% loadings(object)[,1:pcs,drop=FALSE] xhat <- tcrossprod(tnew, loadings(object)[,1:pcs,drop=FALSE]) if(post) xhat <- prep(xhat, scl(object), center(object), reverse=TRUE) list(scores=tnew, x=xhat) } ## @importFrom stats predict ##' @name predict-methods ##' @aliases predict,pcaRes-method setMethod("predict", "pcaRes", predict.pcaRes) ##' This function extracts the residuals values from a pcaRes object ##' for the PCA methods SVD, Nipals, PPCA and BPCA ##' @title Residuals values from a PCA model. ##' @param object \code{pcaRes} the \code{pcaRes} object of interest. ##' @param data \code{matrix} The data that was used to calculate the ##' PCA model (or a different dataset to e.g. adress its proximity to ##' the model). ##' @param ... Passed on to \code{\link{predict.pcaRes}}. E.g. setting ##' the number of used components. ##' @return A \code{matrix} with the residuals ##' @method residuals pcaRes ##' @keywords multivariate ##' @export ##' @name rediduals-methods ##' @examples ##' data(iris) ##' pcIr <- pca(iris[,1:4]) ##' head(residuals(pcIr, iris[,1:4])) ##' @author Henning Redestig residuals.pcaRes <- function(object, data=completeObs(object), ...) { if(is.null(data)) stop("data missing when calculating residuals") data - predict(object, data, ...)$x } ##' @aliases residuals,pcaRes-method ##' @name rediduals-methods setMethod("residuals", "pcaRes", residuals.pcaRes) ##' @name rediduals-methods ##' @aliases resid,pcaRes-method setMethod("resid", "pcaRes", residuals.pcaRes) ##' Fitted values of a PCA model ##' ##' This function extracts the fitted values from a pcaResobject. For ##' PCA methods like SVD, Nipals, PPCA etc this is basically just the ##' scores multipled by the loadings and adjusted for pre-processing. ##' for non-linear PCA the original data is propagated through the ##' network to obtain the approximated data. ##' @title Extract fitted values from PCA. ##' @param object the \code{pcaRes} object of interest. ##' @param data For standard PCA methods this can safely be left null ##' to get scores x loadings but if set, then the scores are obtained ##' by projecting provided data onto the loadings. If data contains ##' missing values the result will be all NA. Non-linear PCA is an ##' exception, here if data is NULL then data is set to the ##' completeObs and propaged through the network. ##' @param nPcs The number of PC's to consider ##' @param pre pre-process \code{data} based on the pre-processing ##' chosen for the PCA model ##' @param post unpre-process the final data (add the center back etc ##' to get the final estimate) ##' @param ... Not used ##' @return A matrix representing the fitted data ##' @keywords multivariate ##' @method fitted pcaRes ##' @examples ##' pc <- pca(iris[,1:4], nPcs=4, center=TRUE, scale="uv") ##' sum( (fitted(pc) - iris[,1:4])^2 ) ##' @export ##' @name fitted-methods ##' @author Henning Redestig fitted.pcaRes <- function(object, data=NULL, nPcs=nP(object), pre=TRUE, post=TRUE, ...) { if(method(object) %in% listPcaMethods("nonlinear")) { if(is.null(data) & is.null(completeObs(object))) stop("completeObs slot is empty -- provide the training data") if(is.null(data) & !is.null(completeObs(object))) data <- completeObs(object) if(is.null(data)) stop("nlpca requires original data to be provide") if(pre) data <- prep(data, scl(object), center(object)) recData <- errorHierarchic(object@network, t(scores(object)), t(data))$out[,,nPcs] recData <- t(recData / object@network@scalingFactor) } else { if(!is.null(data)) { if(pre) data <- prep(data, scl(object), center(object)) tt <- data %*% loadings(object)[,1:nPcs, drop=FALSE] } if(is.null(data)) tt <- scores(object)[,1:nPcs, drop=FALSE] recData <- tcrossprod(tt, loadings(object)[,1:nPcs, drop=FALSE]) } if(post) recData <- prep(recData, scl(object), center(object), reverse=TRUE) return(recData) } ## @importFrom stats fitted ##' @name fitted-methods ##' @aliases fitted,pcaRes-method setMethod("fitted", "pcaRes", fitted.pcaRes) ##' Plot the computed diagnostics of PCA model to get an idea of their ##' importance. Note though that the standard screeplot shows the ##' standard deviations for the PCs this method shows the R2 values ##' which empirically shows the importance of the P's and is thus ##' applicable for any PCA method rather than just SVD based PCA. ##' ##' If cross-validation was done for the PCA the plot will also show ##' the CV based statistics. A common rule-of-thumb for determining ##' the optimal number of PCs is the PC where the CV diagnostic is at ##' its maximum but not very far from \eqn{R^2}. ##' @title Plot diagnostics (screeplot) ##' @param x \code{pcaRes} The pcaRes object. ##' @param y not used ##' @param main title of the plot ##' @param col Colors of the bars ##' @param ... further arguments to barplot ##' @return None, used for side effect. ##' @seealso \link{screeplot} ##' @examples ##' data(metaboliteData) ##' pc <- pca(t(metaboliteData), nPcs=5, cv="q2", scale="uv") ##' plot(pc) ##' @method plot pcaRes ##' @aliases plot.pcaRes plot,pcaRes-method ##' @export ##' @author Henning Redestig plot.pcaRes <- function(x, y=NULL, main=deparse(substitute(object)), col=gray(c(0.9, 0.5)), ...) { y <- NULL ## the deparse(subsitute(object)) later fails otherwise main <- main if(!is.null(cvstat(x))) { cvs <- cvstat(x) if(length(cvs) != nP(x)) cvs <- c(cvs, rep(NA, nP(x) - length(cvs))) xx <- rbind(R2cum(x), cvs) barplot(xx, beside=TRUE, ylim=c(0,1.1), col=col, main=main, names.arg=paste("PC", 1:nP(x), sep=""), ...) legend(x="topleft", fill=col, legend=c(expression(R^2), expression(Q^2))) } else barplot(R2cum(x), ylim=c(0,1.1), ylab=expression(R^2), main=main, names.arg=paste("PC", 1:nP(x), sep=""), col=col[1], ...) } setMethod("plot", "pcaRes", plot.pcaRes) ##' A common way of visualizing two principal components ##' ##' This method is meant to be used as a quick way to visualize ##' results, if you want a more specific plot you probably want to ##' get the scores, loadings with \code{scores(object)}, ##' \code{loadings(object)} and then design your own plotting method. ##' @title Side by side scores and loadings plot ##' @usage slplot(object, pcs=c(1,2), scoresLoadings=c(TRUE, TRUE), ##' sl="def", ll="def", hotelling=0.95, rug=TRUE, sub=NULL,...) ##' @param object a pcaRes object ##' @param pcs which two pcs to plot ##' @param scoresLoadings Which should be shown scores and or loadings ##' @param sl labels to plot in the scores plot ##' @param ll labels to plot in the loadings plot ##' @param hotelling confidence interval for ellipse in the score plot ##' @param rug logical, rug x axis in score plot or not ##' @param sub Subtitle, defaults to annotate with amount of explained ##' variance. ##' @param ... Further arguments to plot functions. Prefix arguments ##' to \code{par()} with 's' for the scores plot and 'l' for the ##' loadings plot. I.e. cex become scex for setting character ##' expansion in the score plot and lcex for the loadings plot. ##' @return None, used for side effect. ##' @note Uses layout instead of par to provide side-by-side so it ##' works with Sweave (but can not be combined with ##' \code{par(mfrow=..))} ##' @author Henning Redestig ##' @seealso \code{\link{pca}}, \code{\link{biplot}} ##' @aliases slplot slplot,pcaRes-method ##' @examples ##' data(iris) ##' pcIr <- pca(iris[,1:4], scale="uv") ##' slplot(pcIr, sl=NULL, spch=5) ##' slplot(pcIr, sl=NULL, lcex=1.3, scol=as.integer(iris[,5])) ##' @keywords multivariate setMethod("slplot", "pcaRes", function(object, pcs=c(1,2), scoresLoadings=c(TRUE, TRUE), sl=rownames(scores(object)), ll=rownames(loadings(object)), hotelling=0.95, rug=FALSE, sub=NULL,...) { opar <- par(no.readonly=TRUE) cl <- match.call() mainArgs <- c(1,match(c("ll", "sl", "scoresLoadings", "sub"), names(cl), 0)) scoreArgs <- grep("^s", names(cl)[-mainArgs]) loadingArgs <- grep("^l", names(cl)[-mainArgs]) if(!is.null(ll) & length(ll) != nVar(object)) stop("Loading labels do not match the object dimensions") if(!is.null(sl) & length(sl) != nObs(object)) stop("Score labels do not match the object dimensions") if(is.null(sl)) sl <- NA if(is.null(ll)) ll <- NA ## no loadings for non-linear pca if(method(object) %in% listPcaMethods("nonlinear") && scoresLoadings[2]) scoresLoadings[2] <- FALSE if(length(pcs) > 2) plotPcs(object, pcs, scoresLoadings=scoresLoadings,...) else { if(is.null(sub)) sub <- paste(sprintf("%.2f", R2cum(object)[max(pcs)] * 100), "% of the variance explained", sep="") if(sum(scoresLoadings) == 2) layout(matrix(c(1,2), 1, 2, TRUE), respect=matrix(c(1,1), 1, 2)) ## exception plot if one dimensional if (length(pcs) == 1 | nP(object) == 1) { pcs <- 1 ## score plot if(scoresLoadings[1]) { newCall <- call("barplot", height=scores(object)[,pcs], main="Scores", las=3, ylab=paste("PC", pcs), sub=sub, names.arg=sl) tmp <- cl[-mainArgs][scoreArgs] names(tmp) <- gsub("^s", "", names(tmp)) for(i in 1:length(tmp)) { newCall[[length(newCall) + 1]] <- tmp[[i]] names(newCall)[length(newCall)] <- names(tmp)[i] } eval(newCall) } ## loadingplot if(scoresLoadings[2]) { newCall <- call("barplot", height=loadings(object)[,pcs], main="Loadings", las=3, ylab=paste("PC", pcs), names.arg=ll) if(length(loadingArgs) > 0) { tmp <- cl[-mainArgs][loadingArgs] names(tmp) <- gsub("^l", "", names(tmp)) for(i in 1:length(tmp)) { newCall[[length(newCall) + 1]] <- tmp[[i]] names(newCall)[length(newCall)] <- names(tmp)[i] } } eval(newCall) } return(invisible(TRUE)) } ## the score plot if(scoresLoadings[1]) { ## setup plot plotCall <- call("plot", x=scores(object)[,pcs], main="Scores", ylab=paste("PC", pcs[2]), sub=sub, xlab=paste("PC", pcs[1])) if(length(scoreArgs) > 0) { tmp <- cl[-mainArgs][scoreArgs] names(tmp) <- gsub("^s", "", names(tmp)) for(i in 1:length(tmp)) { plotCall[[length(plotCall) + 1]] <- tmp[[i]] names(plotCall)[length(plotCall)] <- names(tmp)[i] } } ## add text if (!is.null(sl) & !all(is.na(sl))) { plotCall[[length(plotCall) + 1]] <- "n" names(plotCall)[length(plotCall)] <- "type" textCall <- call("text", x=scores(object)[,pcs], labels=sl) if(length(scoreArgs) > 0) { tmp <- cl[-mainArgs][scoreArgs] names(tmp) <- gsub("^s", "", names(tmp)) for(i in 1:length(tmp)) { textCall[[length(textCall) + 1]] <- tmp[[i]] names(textCall)[length(textCall)] <- names(tmp)[i] } } } eval(plotCall) if (!is.null(sl) & !all(is.na(sl))) eval(textCall) if(rug) rug(scores(object)[,1]) abline(h=0, v=0) if(!is.null(hotelling)) { A <- length(pcs) el <- simpleEllipse(scores(object)[,pcs[1]], scores(object)[,pcs[2]], alfa=hotelling) lines(el) } } ## the loading plot if(scoresLoadings[2]) { ## setup plot plotCall <- call("plot", x=loadings(object)[,pcs], main="Loadings", ylab=paste("PC", pcs[2]), xlab=paste("PC", pcs[1])) if(length(loadingArgs) > 0) { tmp <- cl[-mainArgs][loadingArgs] names(tmp) <- gsub("^l", "", names(tmp)) for(i in 1:length(tmp)) { plotCall[[length(plotCall) + 1]] <- tmp[[i]] names(plotCall)[length(plotCall)] <- names(tmp)[i] } } ## add text if (!is.null(ll) & !all(is.na(ll))) { plotCall[[length(plotCall) + 1]] <- "n" names(plotCall)[length(plotCall)] <- "type" textCall <- call("text", x=loadings(object)[,pcs], labels=ll) if(length(loadingArgs) > 0) { tmp <- cl[-mainArgs][loadingArgs] names(tmp) <- gsub("^l", "", names(tmp)) for(i in 1:length(tmp)) { textCall[[length(textCall) + 1]] <- tmp[[i]] names(textCall)[length(textCall)] <- names(tmp)[i] } } } eval(plotCall) if (!is.null(ll) & !all(is.na(ll))) eval(textCall) abline(h=0, v=0) } } par(opar) }) pcaMethods/R/nipalsPca.R0000644000175400017540000001513613556116437016146 0ustar00biocbuildbiocbuild##' PCA by non-linear iterative partial least squares ##' ##' Can be used for computing PCA on a numeric matrix using either the ##' NIPALS algorithm which is an iterative approach for estimating the ##' principal components extracting them one at a time. NIPALS can ##' handle a small amount of missing values. It is not recommended to ##' use this function directely but rather to use the pca() wrapper ##' function. There is a C++ implementation given as \code{nipalsPca} ##' which is faster. ##' @title NIPALS PCA implemented in R ##' @param Matrix Pre-processed (centered, scaled) numerical matrix ##' samples in rows and variables as columns. ##' @param nPcs Number of components that should be extracted. ##' @param varLimit Optionally the ratio of variance that should be ##' explained. \code{nPcs} is ignored if varLimit < 1 ##' @param maxSteps Defines how many iterations can be done before ##' algorithm should abort (happens almost exclusively when there were ##' some wrong in the input data). ##' @param threshold The limit condition for judging if the algorithm ##' has converged or not, specifically if a new iteration is done if ##' \eqn{(T_{old} - T)^T(T_{old} - T) > \code{limit}}. ##' @param verbose Show simple progress information. ##' @param ... Only used for passing through arguments. ##' @return A \code{pcaRes} object. ##' @references Wold, H. (1966) Estimation of principal components and ##' related models by iterative least squares. In Multivariate ##' Analysis (Ed., P.R. Krishnaiah), Academic Press, NY, 391-420. ##' @author Henning Redestig ##' @seealso \code{prcomp}, \code{princomp}, \code{pca} ##' @examples ##' data(metaboliteData) ##' mat <- prep(t(metaboliteData)) ##' ## c++ version is faster ##' system.time(pc <- RnipalsPca(mat, method="rnipals", nPcs=2)) ##' system.time(pc <- nipalsPca(mat, nPcs=2)) ##' ## better use pca() ##' pc <- pca(t(metaboliteData), method="rnipals", nPcs=2) ##' \dontshow{stopifnot(sum((fitted(pc) - t(metaboliteData))^2, na.rm=TRUE) < 200)} ##' @keywords multivariate ##' @export RnipalsPca <- function(Matrix, nPcs=2, varLimit=1, maxSteps=5000, threshold=1e-6, verbose=interactive(), ...) { nVar <- ncol(Matrix) ##Find a good? starting column -- better way? startingColumn <- 1 ## sum(c(NA, NA), na.rm=TRUE) is 0, but we want NA sum.na <- function(x){ ifelse(all(is.na(x)), NA, sum(x, na.rm=TRUE))} TotalSS <- sum(Matrix*Matrix, na.rm=TRUE) ph <- rep(0, nVar) R2cum <- rep(NA, nPcs) scores <- NULL loadings <- NULL anotherPc <- TRUE l <- 1 while(anotherPc) { count <- 0 #number of iterations done th <- Matrix[,startingColumn] #first column is starting vector for th continue <- TRUE if(verbose) cat(paste("Calculating PC", l, ": ", sep="")) while(continue) { count <- count+1 ph <- rep(0, nVar) ##Calculate loadings through LS regression ##Note: Matrix*th is column-wise multiplication tsize <- sum(th * th, na.rm=TRUE) ph <- apply(Matrix * (th / tsize), 2, sum.na) ##normalize ph based on the available values. psize <- sum(ph*ph, na.rm=TRUE) ph <- ph / sqrt(psize) ##Calculate scores through LS regression ##Trick: To get row-wise multiplication, use t(Matrix)*ph, then ##be sure to use apply(,2,) and NOT apply(,1,)! th.old <- th th <- apply(t(Matrix) * ph, 2, sum.na) ##Round up by calculating if convergence condition is met and ##checking if it seems to be an neverending loop. if (count > maxSteps) { stop("Too many iterations, quitting") } if (t(na.omit(th.old - th)) %*% (na.omit(th.old - th)) <= threshold) { continue = FALSE } if (verbose)cat("*") } if (verbose) cat(" Done\n") Matrix <- Matrix - (th %*% t(ph)) scores <- cbind(scores, th) loadings <- cbind(loadings, ph) ##cumulative proportion of variance R2cum[l] <- 1 - (sum(Matrix*Matrix,na.rm=TRUE) / TotalSS) l <- l + 1 if((!abs(varLimit - 1) < 1e-4 & R2cum[l - 1] >= varLimit) | l > nPcs) { anotherPc <- FALSE nPcs <- l - 1 } } r <- new("pcaRes") r@scores <- scores r@loadings <- loadings r@R2cum <- R2cum r@varLimit <- varLimit r@method <- "rnipals" return(r) } ##' PCA by non-linear iterative partial least squares ##' ##' Can be used for computing PCA on a numeric matrix using either the ##' NIPALS algorithm which is an iterative approach for estimating the ##' principal components extracting them one at a time. NIPALS can ##' handle a small amount of missing values. It is not recommended to ##' use this function directely but rather to use the pca() wrapper ##' function. ##' @title NIPALS PCA ##' @param Matrix Pre-processed (centered, scaled) numerical matrix ##' samples in rows and variables as columns. ##' @param nPcs Number of components that should be extracted. ##' @param varLimit Optionally the ratio of variance that should be ##' explained. \code{nPcs} is ignored if varLimit < 1 ##' @param maxSteps Defines how many iterations can be done before ##' algorithm should abort (happens almost exclusively when there were ##' some wrong in the input data). ##' @param threshold The limit condition for judging if the algorithm ##' has converged or not, specifically if a new iteration is done if ##' \eqn{(T_{old} - T)^T(T_{old} - T) > \code{limit}}. ##' @param ... Only used for passing through arguments. ##' @return A \code{pcaRes} object. ##' @references Wold, H. (1966) Estimation of principal components and ##' related models by iterative least squares. In Multivariate ##' Analysis (Ed., P.R. Krishnaiah), Academic Press, NY, 391-420. ##' @author Henning Redestig ##' @seealso \code{prcomp}, \code{princomp}, \code{pca} ##' @examples ##' data(metaboliteData) ##' mat <- prep(t(metaboliteData)) ##' pc <- nipalsPca(mat, nPcs=2) ##' ## better use pca() ##' pc <- pca(t(metaboliteData), method="nipals", nPcs=2) ##' \dontshow{stopifnot(sum((fitted(pc) - t(metaboliteData))^2, na.rm=TRUE) < 200)} ##' @keywords multivariate ##' @export nipalsPca <- function(Matrix, nPcs=2, varLimit=1, maxSteps=5000, threshold=1e-6, ...) { nipRes <- .Call("pcaMethods_Nipals", Matrix, params=list(nPcs=nPcs, varLimit=varLimit, threshold=threshold, maxSteps=maxSteps), PACKAGE="pcaMethods") r <- new("pcaRes") r@scores <- nipRes$scores r@loadings <- nipRes$loadings r@R2cum <- nipRes$R2cum r@varLimit <- varLimit r@method <- "nipals" return(r) } pcaMethods/R/nlpca.R0000644000175400017540000001664413556116437015336 0ustar00biocbuildbiocbuild##' Neural network based non-linear PCA ##' ##' Artificial Neural Network (MLP) for performing non-linear ##' PCA. Non-linear PCA is conceptually similar to classical PCA but ##' theoretically quite different. Instead of simply decomposing our ##' matrix (X) to scores (T) loadings (P) and an error (E) we train a ##' neural network (our loadings) to find a curve through the ##' multidimensional space of X that describes a much variance as ##' possible. Classical ways of interpreting PCA results are thus not ##' applicable to NLPCA since the loadings are hidden in the network. ##' However, the scores of components that lead to low ##' cross-validation errors can still be interpreted via the score ##' plot. Unfortunately this method depend on slow iterations which ##' currently are implemented in R only making this method extremely ##' slow. Furthermore, the algorithm does not by itself decide when it ##' has converged but simply does 'maxSteps' iterations. ##' @title Non-linear PCA ##' @param Matrix \code{matrix} --- Preprocessed data with the ##' variables in columns and observations in rows. The data may ##' contain missing values, denoted as \code{NA} ##' @param nPcs \code{numeric} -- Number of components to ##' estimate. The preciseness of the missing value estimation depends ##' on thenumber of components, which should resemble the internal ##' structure of the data. ##' @param maxSteps \code{numeric} -- Number of estimation ##' steps. Default is based on a generous rule of thumb. ##' @param unitsPerLayer The network units, example: c(2,4,6) for two ##' input units 2feature units (principal components), one hidden ##' layer fornon-linearity and three output units (original amount ##' ofvariables). ##' @param functionsPerLayer The function to apply at each layer ##' eg. c("linr", "tanh", "linr") ##' @param weightDecay Value between 0 and 1. ##' @param weights Starting weights for the network. Defaults to ##' uniform random values but can be set specifically to make ##' algorithm deterministic. ##' @param verbose \code{boolean} -- nlpca prints the number of steps ##' and warning messages if set to TRUE. Default is interactive(). ##' @param ... Reserved for future use. Not passed on anywhere. ##' @return Standard PCA result object used by all PCA-basedmethods of ##' this package. Contains scores, loadings, data meanand more. See ##' \code{\link{pcaRes}} for details. ##' @author Based on a matlab script by Matthias Scholz and ported to ##' R by Henning Redestig ##' @references Matthias Scholz, Fatma Kaplan, Charles L Guy, Joachim ##' Kopkaand Joachim Selbig. Non-linear PCA: a missing ##' data approach. \emph{Bioinformatics, 21(20):3887-3895, Oct 2005} ##' @examples ##' ## Data set with three variables where data points constitute a helix ##' data(helix) ##' helixNA <- helix ##' ## not a single complete observation ##' helixNA <- t(apply(helix, 1, function(x) { x[sample(1:3, 1)] <- NA; x})) ##' ## 50 steps is not enough, for good estimation use 1000 ##' helixNlPca <- pca(helixNA, nPcs=1, method="nlpca", maxSteps=50) ##' fittedData <- fitted(helixNlPca, helixNA) ##' plot(fittedData[which(is.na(helixNA))], helix[which(is.na(helixNA))]) ##' ## compared to solution by Nipals PCA which cannot extract non-linear patterns ##' helixNipPca <- pca(helixNA, nPcs=2) ##' fittedData <- fitted(helixNipPca) ##' plot(fittedData[which(is.na(helixNA))], helix[which(is.na(helixNA))]) ##' @export nlpca <- function(Matrix, nPcs=2, maxSteps=2 * prod(dim(Matrix)), unitsPerLayer=NULL, functionsPerLayer=NULL, weightDecay=0.001, weights=NULL, verbose=interactive(),...) { ## do some basic checks object <- Matrix trainIn <- NULL trainOut <- t(object) stds <- apply(trainOut, 2, sd, na.rm=TRUE) scalingFactor <- 0.1 / max(stds) trainOut <- trainOut * scalingFactor ## now setup the initial nlpcaNet object numNaN <- sum(is.na(object)) ## always inverse in this version, bottleneck is not fully implemented inverse <- TRUE ## DATADIST (nlnet@dataDist) is given by weightOut dataDist <- apply(!is.na(trainOut), 2, as.integer) #0 for NA, 1 for everything else if(!inverse) dataDist <- NULL ## setup the network architecture if(is.null(unitsPerLayer)) { ld <- dim(trainOut)[1] lh <- nPcs if(nPcs < 10) lh <- 2 + 2 * nPcs unitsPerLayer <- c(ld, lh, nPcs, lh, ld) if(inverse) unitsPerLayer <- c(nPcs, lh, ld) } featureLayer <- ceiling(length(unitsPerLayer) / 2) if(inverse) featureLayer <- 1 if(is.null(functionsPerLayer)) { functionsPerLayer <- rep("tanh", length(unitsPerLayer)) functionsPerLayer[1] <- "linr" functionsPerLayer[featureLayer] <- "linr" functionsPerLayer[length(unitsPerLayer)] <- "linr" } hierarchic <- list(layer=featureLayer, var=rbind(c(rep(1, nPcs), 0.01)), idx=getHierarchicIdx(unitsPerLayer[featureLayer])) ## set up the weights wNum <- sum(sapply(2:length(unitsPerLayer), function(i) (1 + unitsPerLayer[i - 1]) * unitsPerLayer[i])) if(!is.null(weights) && length(weights) != wNum) { warning("Weight vector not expected length (", wNum, "), using random weights", sep="") weights <- NULL } if(is.null(weights)) weights <- cbind(0.2 * (runif(wNum, 0, 1) - 0.1)) if(inverse) { numPattern <- dim(trainOut)[2] tmpTrainIn <- cbind(rnorm(unitsPerLayer[1] * numPattern,0,1) * 0.1) weights <- rbind(tmpTrainIn, weights) } if(nPcs == 1) featureSorting <- FALSE if(nPcs > 1) featureSorting <- TRUE nlnet <- new("nlpcaNet") nlnet@net <- rbind(unitsPerLayer) nlnet@hierarchic <- hierarchic nlnet@fct <- functionsPerLayer nlnet@fkt <- functionsPerLayer[2:length(functionsPerLayer)] nlnet@weightDecay <- weightDecay nlnet@featureSorting <- featureSorting nlnet@dataDist <- dataDist nlnet@inverse <- inverse nlnet@fCount <- as.integer(0) nlnet@componentLayer <- as.integer(featureLayer) nlnet@error <- errorHierarchic nlnet@gradient <- derrorHierarchic nlnet@maxIter <- as.integer(maxSteps) nlnet@weights <- weightsAccount(weights) nlnet@scalingFactor <- scalingFactor ## ****************************** if(verbose) cat("Training network with", nlnet@maxIter, "iterations...\n!:\tSquare error is NA -- accuracy in line-search might be too small\n:\tComponents were sorted at iteration n\n^:\tToo many iterations while expanding\n") newnet <- optiAlgCgd(nlnet, trainIn, trainOut, verbose) if(verbose) cat("\nDone\n") if(inverse) { nObs <- unitsPerLayer[1] * dim(trainOut)[2] we <- newnet@weights$current() scores <- t(matrix(we[1:nObs], nrow=unitsPerLayer[1], dim(trainOut)[2])) newnet@weights$set(we[(nObs + 1):length(we),,drop=FALSE]) } ## for further applications newnet must not be inverse anymore newnet@inverse <- FALSE res <- new("pcaRes") res@scores <- scores res@loadings <- matrix() res@network <- newnet res@method <- "nlpca" R2cum <- rep(NA, nPcs) TSS <- sum(Matrix^2, na.rm=TRUE) for(i in 1:nPcs) R2cum[i] <- 1 - sum((Matrix - fitted(res, Matrix, nPcs=i))^2, na.rm=TRUE) / TSS res@R2cum <- R2cum res } ##' Index in hiearchy ##' @param hierarchicNum A number ##' @return ... ##' @author Henning Redestig, Matthias Scholz getHierarchicIdx <- function(hierarchicNum) { res <- matrix(1, ncol=hierarchicNum, nrow=hierarchicNum) res[lower.tri(res)] <- 0 cbind(res, c(0, rep(1, hierarchicNum - 1))) } pcaMethods/R/optiAlgCgd.R0000644000175400017540000000414313556116437016245 0ustar00biocbuildbiocbuild##' Conjugate gradient optimization ##' @param nlnet The nlnet ##' @param trainIn Training data ##' @param trainOut fitted data ##' @param verbose logical, print messages ##' @return ... ##' @author Henning Redestig, Matthias Scholz optiAlgCgd <- function(nlnet, trainIn, trainOut, verbose=FALSE) { tmpnet <- forkNlpcaNet(nlnet) derr <- tmpnet@gradient(tmpnet, trainIn, trainOut) dw <- derr$dwTotal e <- derr$Etotal dv <- -dw if(tmpnet@featureSorting) eSortLast <- e eHist <- rep(0, tmpnet@maxIter) ttLast <- rep(0.0001, 6) for(i in 1:tmpnet@maxIter) { if(verbose) { if(i %% 10 == 0) cat("*") if(i %% 100 == 0) cat(" [", i, "]\n") } eHist[i] <- e eLast <- e # line search in direction dv (downhill) ttGuess <- max(min(ttLast), 0.00001) linSe <- lineSearch(tmpnet, dv, e, ttGuess, trainIn, trainOut, verbose) tmpnet@weights$set(cbind(linSe$wBest)) e <- linSe$eBest tt <- linSe$ttBest ttLast <- c(ttLast[2:length(ttLast)], tt) #shift and add new tt gradRes <- tmpnet@gradient(tmpnet, trainIn, trainOut) dwNew <- gradRes$dwTotal e <- gradRes$Etotal ## define new search direction dv (conjugate direction) ## b1=dw_new'*dw_new; # Fletcher-Reeves b1 <- crossprod(dwNew, (dwNew - dw))#Polak-Ribiere b2 <- crossprod(dw) beta <- b1 / b2 dv <- -dwNew + dv %*% beta dw <- dwNew if(e > eLast) { dv <- -dwNew if(verbose) cat("!", sep="") } if(is.na(e)) stop("Square error is NA (critical) - accuracy in line-search might be too small") if(tmpnet@featureSorting) if(e / eSortLast < 0.90 || i == tmpnet@maxIter || i == tmpnet@maxIter - 1 || i == tmpnet@maxIter - 2) { eSortLast <- e if(verbose) cat("<", i, ">", sep="") ## somewhat secret method, sortFeatures calls ## nlnet@weights$set(x) so the weights are updated here ## 'behind the scenes' sortFeatures(tmpnet, trainIn, trainOut) } } tmpnet } pcaMethods/R/orth.R0000644000175400017540000000354013556116437015204 0ustar00biocbuildbiocbuild##' ONB = orth(mat) is an orthonormal basis for the range of matrix ##' mat. That is, ONB' * ONB = I, the columns of ONB span the same ##' space as the columns of mat, and the number of columns of ONB is ##' the rank of mat. ##' @title Calculate an orthonormal basis ##' @param mat matrix to calculate orthonormal base ##' @param skipInac do not include components with precision below ##' .Machine$double.eps if TRUE ##' @return orthonormal basis for the range of matrix ##' @author Wolfram Stacklies orth <- function(mat, skipInac = FALSE) { if(nrow(mat) > ncol(mat)) { leftSVs <- ncol(mat) } else { leftSVs <- nrow(mat) } result <- svd(mat, nu = leftSVs, nv = ncol(mat)) U <- result[[2]] S <- result[[1]] V <- result[[3]] m <- nrow(mat) n <- ncol(mat) if(m > 1) { s <- diag(S, nrow = length(S)) } else if(m == 1) { s <- S[1] } else { s <- 0 } tol <- max(m,n) * max(s) * .Machine$double.eps r <- sum(s > tol) if ( r < ncol(U) ) { if (skipInac) { warning("Precision for components ", r + 1 , " - ", ncol(U), " is below .Machine$double.eps. \n", "Results for those components are likely to be inaccurate!!\n", "These component(s) are not included in the returned solution!!\n") } else { warning("Precision for components ", r + 1 , " - ", ncol(U), " is below .Machine$double.eps. \n", "Results for those components are likely to be inaccurate!!\n") } } if (skipInac) { ONB <- U[, 1:r, drop=FALSE] ## Assing correct row and colnames rownames(ONB) <- labels(mat[, 1:r, drop=FALSE])[[1]]; colnames(ONB) <- labels(mat[, 1:r, drop=FALSE])[[2]]; } else { ONB<-U ## Assing correct row and colnames rownames(ONB) <- labels(mat)[[1]]; colnames(ONB) <- labels(mat)[[2]]; } return(ONB) } pcaMethods/R/pca.R0000644000175400017540000003477013556116437015004 0ustar00biocbuildbiocbuild##' Vector with current valid PCA methods ##' @title List PCA methods ##' @param which the type of methods to get. E.g. only get the PCA ##' methods based on the classical model where the fitted data is a ##' direct multiplication of scores and loadings. ##' @return A character vector with the current methods for doing PCA ##' @export ##' @author Henning Redestig listPcaMethods <- function(which=c("all", "linear", "nonlinear")) { switch(match.arg(which), all={ return(c("svd", "nipals", "rnipals", "bpca", "ppca", "svdImpute", "robustPca", "nlpca", "llsImpute", "llsImputeAll")) }, linear={ return(c("svd", "nipals", "rnipals", "bpca", "ppca", "svdImpute", "robustPca")) }, nonlinear={ return("nlpca") }) } ##' Perform PCA on a numeric matrix for visualisation, information ##' extraction and missing value imputation. ##' ##' This method is wrapper function for the following set of pca ##' methods: ##' ##' \describe{\item{svd:}{Uses classical \code{prcomp}. See ##' documentation for \code{\link{svdPca}}.} ##' ##' \item{nipals:}{An iterative method capable of handling small ##' amounts of missing values. See documentation for ##' \code{\link{nipalsPca}}.} ##' ##' \item{rnipals:}{Same as nipals but implemented in R.} ##' ##' \item{bpca:}{An iterative method using a Bayesian model to handle ##' missing values. See documentation for \code{\link{bpca}}.} ##' ##' \item{ppca:}{An iterative method using a probabilistic model to ##' handle missing values. See documentation for \code{\link{ppca}}.} ##' ##' \item{svdImpute:}{Uses expectation maximation to perform SVD PCA ##' on incomplete data. See documentation for ##' \code{\link{svdImpute}}.}} ##' ##' Scaling and centering is part of the PCA model and handled by ##' \code{\link{prep}}. ##' @title Perform principal component analysis ##' @param object Numerical matrix with (or an object coercible to ##' such) with samples in rows and variables as columns. Also takes ##' \code{ExpressionSet} in which case the transposed expression ##' matrix is used. Can also be a data frame in which case all ##' numberic variables are used to fit the PCA. ##' @param method One of the methods reported by ##' \code{listPcaMethods()}. Can be left missing in which case the ##' \code{svd} PCA is chosen for data wihout missing values and ##' \code{nipalsPca} for data with missing values ##' @param nPcs Number of principal components to calculate. ##' @param scale Scaling, see \code{\link{prep}}. ##' @param center Centering, see \code{\link{prep}}. ##' @param completeObs Sets the \code{completeObs} slot on the ##' resulting \code{pcaRes} object containing the original data with ##' but with all NAs replaced with the estimates. ##' @param subset A subset of variables to use for calculating the ##' model. Can be column names or indices. ##' @param cv character naming a the type of cross-validation ##' to be performed. ##' @param ... Arguments to \code{\link{prep}}, the chosen pca ##' method and \code{\link{Q2}}. ##' @return A \code{pcaRes} object. ##' @references ##' Wold, H. (1966) Estimation of principal components and ##' related models by iterative least squares. In Multivariate ##' Analysis (Ed., P.R. Krishnaiah), Academic Press, NY, 391-420. ##' ##' Shigeyuki Oba, Masa-aki Sato, Ichiro Takemasa, Morito Monden, ##' Ken-ichi Matsubara and Shin Ishii. A Bayesian missing value ##' estimation method for gene expression profile ##' data. \emph{Bioinformatics, 19(16):2088-2096, Nov 2003}. ##' ##' Troyanskaya O. and Cantor M. and Sherlock G. and Brown P. and ##' Hastie T. and Tibshirani R. and Botstein D. and Altman RB. - ##' Missing value estimation methods for DNA microarrays. ##' \emph{Bioinformatics. 2001 Jun;17(6):520-5}. ##' @seealso \code{\link{prcomp}}, \code{\link{princomp}}, ##' \code{\link{nipalsPca}}, \code{\link{svdPca}} ##' @examples ##' data(iris) ##' ## Usually some kind of scaling is appropriate ##' pcIr <- pca(iris, method="svd", nPcs=2) ##' pcIr <- pca(iris, method="nipals", nPcs=3, cv="q2") ##' ## Get a short summary on the calculated model ##' summary(pcIr) ##' plot(pcIr) ##' ## Scores and loadings plot ##' slplot(pcIr, sl=as.character(iris[,5])) ##' ##' ## use an expressionset and ggplot ##' data(sample.ExpressionSet) ##' pc <- pca(sample.ExpressionSet) ##' df <- merge(scores(pc), pData(sample.ExpressionSet), by=0) ##' library(ggplot2) ##' ggplot(df, aes(PC1, PC2, shape=sex, color=type)) + ##' geom_point() + ##' xlab(paste("PC1", pc@R2[1] * 100, "% of the variance")) + ##' ylab(paste("PC2", pc@R2[2] * 100, "% of the variance")) ##' @export ##' @keywords multivariate ##' @author Wolfram Stacklies, Henning Redestig pca <- function(object, method, nPcs=2, scale=c("none", "pareto", "vector", "uv"), center=TRUE, completeObs=TRUE, subset=NULL, cv=c("none","q2"), ...) { if(inherits(object, 'data.frame')) { num <- vapply(object, is.numeric, logical(1)) if(sum(num) < 2) stop('no numeric data in supplied data.frame') Matrix <- as.matrix(object[,num]) } else if(inherits(object, "ExpressionSet")) { Matrix <- t(exprs(object)) } else Matrix <- as.matrix(object, rownames.force=TRUE) if(!is.null(subset)) Matrix <- Matrix[,subset] cv <- match.arg(cv) scale <- match.arg(scale) if (nPcs > ncol(Matrix)) { warning("more components than matrix columns requested") nPcs <- min(dim(Matrix)) } if (nPcs > nrow(Matrix)) { warning("more components than matrix rows requested") nPcs <- min(dim(Matrix)) } if (!checkData(Matrix, verbose=interactive())) stop("Invalid data format.", "Run checkData(data, verbose=TRUE) for details") missing <- is.na(Matrix) if(missing(method)) { if(any(missing)) method <- 'nipals' else method <- 'svd' } if(any(missing) & method == 'svd') { warning('data has missing values using nipals instead of user requested svd') method <- 'nipals' } method <- match.arg(method, choices=listPcaMethods()) prepres <- prep(Matrix, scale=scale, center=center, simple=FALSE, ...) switch(method, svd={ res <- svdPca(prepres$data, nPcs=nPcs,...) }, nipals={ res <- nipalsPca(prepres$data, nPcs=nPcs, ...) }, rnipals={ res <- RnipalsPca(prepres$data, nPcs=nPcs, ...) }, bpca={ res <- bpca(prepres$data, nPcs=nPcs, ...) }, ppca={ res <- ppca(prepres$data, nPcs=nPcs, ...) }, svdImpute={ res <- svdImpute(prepres$data, nPcs=nPcs, ...) }, robustPca={ res <- robustPca(prepres$data, nPcs=nPcs, ...) }, nlpca={ res <- nlpca(prepres$data, nPcs=nPcs, ...) }) nPcs <- ncol(res@scores) if(is.null(scores(res)) | is.null(loadings(res)) | is.null(R2cum(res)) | is.null(method(res))) stop(paste("bad result from pca method", method)) colnames(res@scores) <- paste("PC", 1:nPcs, sep="") rownames(res@scores) <- rownames(Matrix) if(all(dim(loadings(res)) == c(ncol(Matrix), nPcs))) { colnames(res@loadings) <- paste("PC", 1:nPcs, sep="") rownames(res@loadings) <- colnames(Matrix) } if(!is.null(subset)) res@subset <- subset res@missing <- missing res@nPcs <- nPcs res@nObs <- nrow(Matrix) res@nVar <- ncol(Matrix) res@sDev <- apply(scores(res), 2, sd) res@center <- prepres$center res@centered <- center res@scale <- prepres$scale res@scaled <- scale res@R2 <- res@R2cum[1] if(length(res@R2cum) > 1) res@R2 <- c(res@R2, diff(res@R2cum)) if (completeObs) { cObs <- Matrix if(method %in% listPcaMethods("nonlinear")) cObs[missing] <- fitted(res, Matrix, pre=TRUE, post=TRUE)[missing] else cObs[missing] <- fitted(res, post=TRUE)[missing] res@completeObs <- cObs } if(cv == "q2") res@cvstat <- Q2(res, Matrix, nruncv=1, ...) return(res) } ##' Wrapper function for imputation methods based on nearest neighbour ##' clustering. Currently llsImpute only. ##' ##' This method is wrapper function to llsImpute, See documentation ##' for \code{link{llsImpute}}. ##' @title Nearest neighbour imputation ##' @param object Numerical matrix with (or an object coercible to ##' such) with samples in rows and variables as columns. Also takes ##' \code{ExpressionSet} in which case the transposed expression ##' matrix is used. ##' @param method For convenience one can pass a large matrix but only ##' use the variable specified as subset. Can be colnames or indices. ##' @param subset Currently "llsImpute" only. ##' @param ... Further arguments to the chosen method. ##' @return A \code{clusterRes} object. Or a list containing a ##' clusterRes object as first and an ExpressionSet object as second ##' entry if the input was of type ExpressionSet. ##' @export ##' @seealso \code{\link{llsImpute}}, \code{\link{pca}} ##' @keywords multivariate ##' @examples ##' data(metaboliteData) ##' llsRes <- nni(metaboliteData, k=6, method="llsImpute", allGenes=TRUE) ##' @author Wolfram Stacklies nni <- function(object, method=c("llsImpute"), subset=numeric(), ...) { isExprSet <- FALSE if(inherits(object, "ExpressionSet")) { set <- object isExprSet <- TRUE object <- t(exprs(object)) } method <- match.arg(method) if ( !checkData(as.matrix(object), verbose=interactive()) ) stop("Invalid data format, exiting...\n", "Run checkData(data, verbose=TRUE) for details\n") missing <- sum(is.na(object)) if(length(subset) > 0) object <- object[,subset] res <- llsImpute(object, ...) return(res) } ##' A function that can be used to visualise many PCs plotted against ##' each other ##' ##' Uses \code{\link{pairs}} to provide side-by-side plots. Note that ##' this function only plots scores or loadings but not both in the ##' same plot. ##' @title Plot many side by side scores XOR loadings plots ##' @param object \code{pcaRes} a pcaRes object ##' @param pcs \code{numeric} which pcs to plot ##' @param type \code{character} Either "scores" or "loadings" for ##' scores or loadings plot respectively ##' @param sl \code{character} Text labels to plot instead of a point, ##' if NULL points are plotted instead of text ##' @param hotelling \code{numeric} Significance level for the ##' confidence ellipse. NULL means that no ellipse is drawn. ##' @param ... Further arguments to \code{\link{pairs}} on which this ##' function is based. ##' @return None, used for side effect. ##' @seealso \code{prcomp}, \code{pca}, \code{princomp}, \code{slplot} ##' @export ##' @examples ##' data(iris) ##' pcIr <- pca(iris[,1:4], nPcs=3, method="svd") ##' plotPcs(pcIr, col=as.integer(iris[,4]) + 1) ##' @keywords multivariate ##' @author Henning Redestig plotPcs <- function(object, pcs=1:nP(object), type=c("scores", "loadings"), sl=NULL, hotelling=0.95, ...) { type <- match.arg(type) panel <- function(x,y, ...) { abline(h=0, v=0, col="black") if(!is.null(hotelling)) { A <- length(pcs) el <- simpleEllipse(x, y, alfa=hotelling) lines(el) } if(is.null(sl)) points(x, y, ...) else text(x, y, labels=sl,...) } switch(type, scores={ labels <- paste("PC", pcs, "\n", "R^2 =", round(object@R2[pcs], 2)) pairs(scores(object)[,pcs], labels=labels, panel=panel, upper.panel=NULL,...) }, loadings={ if(method(object) == "nlpca") stop("Loadings plot not applicable for non-linear PCA") labels <- paste("PC", pcs, "\n", "R^2 =", round(object@R2[pcs], 2)) pairs(loadings(object)[,pcs], labels=labels, panel=panel, upper.panel=NULL, ...) }) } ##' A wrapper function for \code{prcomp} to deliver the result as a ##' \code{pcaRes} method. Supplied for compatibility with the rest ##' of the pcaMethods package. It is not recommended to use this ##' function directely but rather to use the \code{pca()} wrapper ##' function. ##' @title Perform principal component analysis using singular value ##' decomposition ##' @param Matrix Pre-processed (centered and possibly scaled) ##' numerical matrix samples in rows and variables as columns. No ##' missing values allowed. ##' @param nPcs Number of components that should be extracted. ##' @param varLimit Optionally the ratio of variance that should be ##' explained. \code{nPcs} is ignored if varLimit < 1 ##' @param verbose Verbose complaints to matrix structure ##' @param ... Only used for passing through arguments. ##' @return A \code{pcaRes} object. ##' @seealso \code{prcomp}, \code{princomp}, \code{pca} ##' @examples ##' data(metaboliteDataComplete) ##' mat <- prep(t(metaboliteDataComplete)) ##' pc <- svdPca(mat, nPcs=2) ##' ## better use pca() ##' pc <- pca(t(metaboliteDataComplete), method="svd", nPcs=2) ##' \dontshow{stopifnot(sum((fitted(pc) - t(metaboliteDataComplete))^2, na.rm=TRUE) < 200)} ##' @export ##' @keywords multivariate ##' @author Henning Redestig svdPca <- function(Matrix, nPcs=2, varLimit=1, verbose=interactive(), ...) { pcs <- prcomp(Matrix, center=FALSE, scale.=FALSE) imp <- summary(pcs)$importance if(varLimit < 1) nPcs <- sum(imp[3,] < varLimit) + 1 res <- new("pcaRes") res@scores <- cbind(pcs$x[,1:nPcs]) res@loadings <- cbind(pcs$rotation[,1:nPcs]) res@R2cum <- imp[3,1:nPcs] res@varLimit <- varLimit res@method <- "svd" return(res) } ##' Get a confidence ellipse for uncorrelated bivariate data ##' ##' As described in 'Introduction to multi and megavariate data analysis ##' using PCA and ##' PLS' by Eriksson et al. This produces very similar ellipse as ##' compared to the ellipse function the ellipse package except that ##' this function assumes that and y are uncorrelated (which they of ##' are if they are scores or loadings from a PCA). ##' @title Hotelling's T^2 Ellipse ##' @param x first variable ##' @param y second variable ##' @param alfa confidence level of the circle ##' @param len Number of points in the circle ##' @seealso ellipse ##' @author Henning Redestig ##' @return A matrix with X and Y coordinates for the circle simpleEllipse <- function(x, y, alfa=0.95, len=200) { N <- length(x) A <- 2 mypi <- seq(0, 2 * pi, length=len) r1 <- sqrt(var(x) * qf(alfa, 2, N - 2) * (2*(N^2 - 1)/(N * (N - 2)))) r2 <- sqrt(var(y) * qf(alfa, 2, N - 2) * (2*(N^2 - 1)/(N * (N - 2)))) cbind(r1 * cos(mypi) + mean(x), r2 * sin(mypi) + mean(y)) } # .onLoad <- function(libname, pkgname) { # require("methods") # } pcaMethods/R/pcaMethods-package.R0000644000175400017540000000772013556116437017714 0ustar00biocbuildbiocbuild##' Simulated data set looking like a helix ##' ##' ##' A matrix containing 1000 observations (rows) and three variables ##' (columns). ##' @title A helix structured toy data set ##' @name helix ##' @aliases helix ##' @usage data(helix) ##' @docType data ##' @references Matthias Scholz, Fatma Kaplan, Charles L. Guy, Joachim ##' Kopka and Joachim Selbig. - Non-linear PCA: a missing data ##' approach. \emph{Bioinformatics 2005 21(20):3887-3895} ##' @keywords datasets ##' @author Henning Redestig NULL ##' A complete subset from a larger metabolite data set. This is the ##' original, complete data set and can be used to compare estimation ##' results created with the also provided incomplete data (called ##' metaboliteData). The data was created during an in house ##' Arabidopsis coldstress experiment. ##' ##' A matrix containing 154 observations (rows) and 52 metabolites ##' (columns). ##' @name metaboliteDataComplete ##' @docType data ##' @aliases metaboliteDataComplete ##' @title A complete metabolite data set from an Arabidopsis ##' coldstress experiment ##' @keywords datasets ##' @seealso \code{\link{metaboliteData}} ##' @references Matthias Scholz, Fatma Kaplan, Charles L. Guy, Joachim ##' Kopka and Joachim Selbig. - Non-linear PCA: a missing data ##' approach.\emph{Bioinformatics 2005 21(20):3887-3895} ##' @author Wolfram Stacklies NULL ##' A incomplete subset from a larger metabolite data set. This is the ##' original, complete data set and can be used to compare estimation ##' results created with the also provided incomplete data (called ##' metaboliteData). ##' ##' A matrix containing 154 observations (rows) and 52 metabolites ##' (columns). The data contains 5\% of artificially created uniformly ##' distributed misssing values. The data was created during an in ##' house Arabidopsis coldstress experiment. ##' @name metaboliteData ##' @docType data ##' @aliases metaboliteData ##' @title A incomplete metabolite data set from an Arabidopsis ##' coldstress experiment ##' @keywords datasets ##' @seealso \code{\link{metaboliteDataComplete}} ##' @references Matthias Scholz, Fatma Kaplan, Charles L. Guy, Joachim ##' Kopka and Joachim Selbig. - Non-linear PCA: a missing data ##' approach.\emph{Bioinformatics 2005 21(20):3887-3895} ##' @author Wolfram Stacklies NULL ##' Principal Component Analysis in R ##' ##' \tabular{ll}{ ##' Package: \tab pcaMethods \cr ##' Type: \tab Package \cr ##' Developed since: \tab 2006 \cr ##' License: \tab GPL (>=3) \cr ##' LazyLoad: \tab yes \cr ##' } ##' ##' Provides Bayesian PCA, Probabilistic PCA, Nipals PCA, Inverse ##' Non-Linear PCA and the conventional SVD PCA. A cluster based ##' method for missing value estimation is included for comparison. ##' BPCA, PPCA and NipalsPCA may be used to perform PCA on incomplete ##' data as well as for accurate missing value estimation. A set of ##' methods for printing and plotting the results is also provided. ##' All PCA methods make use of the same data structure (pcaRes) to ##' provide a unique interface to the PCA results. Developed at the ##' Max-Planck Institute for Molecular Plant Physiology, Golm, ##' Germany, RIKEN Plant Science Center Yokohama, Japan, and CAS-MPG ##' Partner Institute for Computational Biology (PICB) Shanghai, ##' P.R. China ##' ##' @name pcaMethods ##' @aliases pcaMethods ##' @docType package ##' @importFrom Rcpp evalCpp ##' @import Biobase ##' @import BiocGenerics ##' @import methods ##' @title pcaMethods ##' @useDynLib pcaMethods ##' @author Wolfram Stacklies, Henning Redestig NULL ##' \describe{ ##' \item{plotR2}{Lack of relevance for this plot and the fact that it ##' can not show cross-validation based diagnostics in the same plot ##' makes it redundant with the introduction of a dedicated ##' \code{plot} function for \code{pcaRes}. The new plot only shows ##' R2cum but the result is pretty much the same.}} ##' @name pcaMethods-deprecated ##' @aliases pcaMethods-deprecated ##' @title Deprecated methods for pcaMethods ##' @author Henning Redestig NULL pcaMethods/R/ppca.R0000644000175400017540000001463413556116437015161 0ustar00biocbuildbiocbuild##' Implementation of probabilistic PCA (PPCA). PPCA allows to perform ##' PCA on incomplete data and may be used for missing value ##' estimation. This script was implemented after the Matlab version ##' provided by Jakob Verbeek ( see ##' \url{http://lear.inrialpes.fr/~verbeek/}) and the draft \emph{``EM ##' Algorithms for PCA and Sensible PCA''} written by Sam Roweis. ##' ##' Probabilistic PCA combines an EM approach for PCA with a ##' probabilistic model. The EM approach is based on the assumption ##' that the latent variables as well as the noise are normal ##' distributed. ##' ##' In standard PCA data which is far from the training set but close ##' to the principal subspace may have the same reconstruction error. ##' PPCA defines a likelihood function such that the likelihood for ##' data far from the training set is much lower, even if they are ##' close to the principal subspace. This allows to improve the ##' estimation accuracy. ##' ##' A method called \code{kEstimate} is provided to estimate the ##' optimal number of components via cross validation. In general few ##' components are sufficient for reasonable estimation accuracy. See ##' also the package documentation for further discussion on what kind ##' of data PCA-based missing value estimation is advisable. ##' ##' \bold{Complexity:}\cr Runtime is linear in the number of data, ##' number of data dimensions and number of principal components. ##' ##' \bold{Convergence:} The threshold indicating convergence was ##' changed from 1e-3 in 1.2.x to 1e-5 in the current version leading ##' to more stable results. For reproducability you can set the seed ##' (parameter seed) of the random number generator. If used for ##' missing value estimation, results may be checked by simply running ##' the algorithm several times with changing seed, if the estimated ##' values show little variance the algorithm converged well. ##' @title Probabilistic PCA ##' @param Matrix \code{matrix} -- Data containing the variables in ##' columns and observations in rows. The data may contain missing ##' values, denoted as \code{NA}. ##' @param nPcs \code{numeric} -- Number of components to ##' estimate. The preciseness of the missing value estimation depends ##' on the number of components, which should resemble the internal ##' structure of the data. ##' @param seed \code{numeric} Set the seed for the random number ##' generator. PPCA creates fills the initial loading matrix with ##' random numbers chosen from a normal distribution. Thus results may ##' vary slightly. Set the seed for exact reproduction of your ##' results. ##' @param threshold Convergence threshold. ##' @param maxIterations the maximum number of allowed iterations ##' @param ... Reserved for future use. Currently no further ##' parameters are used. ##' @note Requires \code{MASS}. It is not recommended to use this ##' function directely but rather to use the pca() wrapper function. ##' @return Standard PCA result object used by all PCA-based methods ##' of this package. Contains scores, loadings, data mean and ##' more. See \code{\link{pcaRes}} for details. ##' @seealso \code{\link{bpca}, \link{svdImpute}, \link{prcomp}, ##' \link{nipalsPca}, \link{pca}, \link{pcaRes}}. ##' @examples ##' ## Load a sample metabolite dataset with 5\% missing values (metaboliteData) ##' data(metaboliteData) ##' ## Perform probabilistic PCA using the 3 largest components ##' result <- pca(t(metaboliteData), method="ppca", nPcs=3, seed=123) ##' ## Get the estimated complete observations ##' cObs <- completeObs(result) ##' ## Plot the scores ##' plotPcs(result, type = "scores") ##' \dontshow{ ##' stopifnot(sum((fitted(result) - t(metaboliteData))^2, na.rm=TRUE) < 200) ##' } ##' @keywords multivariate ##' @author Wolfram Stacklies ##' @export ppca <- function(Matrix, nPcs=2, seed=NA, threshold=1e-5, maxIterations=1000, ...) { ## Set the seed to the user defined value. This affects the generation ## of random values for the initial setup of the loading matrix if (!is.na(seed)) set.seed(seed) N <- nrow(Matrix) D <- ncol(Matrix) Obs <- !is.na(Matrix) hidden <- which(is.na(Matrix)) missing <- length(hidden) if(missing) { Matrix[hidden] <- 0 } ## ------- Initialization r <- sample(N) C <- t(Matrix[r[1:nPcs], ,drop = FALSE]) ## Random matrix with the same dimnames as Matrix C <- matrix(rnorm(C), nrow(C), ncol(C), dimnames = labels(C) ) CtC <- t(C) %*% C ## inv(C'C) C' X is the solution to the EM problem X <- Matrix %*% C %*% solve(CtC) recon <- X %*% t(C) recon[hidden] <- 0 ss <- sum(sum((recon - Matrix)^2)) / (N * D - missing) count <- 1 old <- Inf ## ------ EM iterations while (count > 0) { ## E-step, (co)variances Sx <- solve(diag(nPcs) + CtC/ss) ss_old <- ss if(missing) { proj <- X %*% t(C) Matrix[hidden] <- proj[hidden] } ## E step: expected values X <- Matrix %*% C %*% Sx / ss ## M-step SumXtX <- t(X) %*% X ## Replace the right matrix division from matlab C <- (t(Matrix) %*% X) %*% solve( (SumXtX + N * Sx) ) CtC <- t(C) %*% C ss <- ( sum(sum( (C %*% t(X) - t(Matrix))^2 )) + N * sum(sum(CtC %*% Sx)) + missing * ss_old ) / (N * D) objective <- N * (D * log(ss) + sum(diag(Sx)) - log(det(Sx)) ) + sum(diag(SumXtX)) - missing * log(ss_old) rel_ch <- abs( 1 - objective / old ) old <- objective count <- count + 1 if( rel_ch < threshold & count > 5 ) { count <- 0 } else if (count > maxIterations) { count <- 0 warning("stopped after max iterations, but rel_ch was > threshold") } } ## End EM iteration C <- orth(C) evs <- eigen( cov(Matrix %*% C) ) vals <- evs[[1]] vecs <- evs[[2]] C <- C %*% vecs X <- Matrix %*% C ## Paramters in original Matlab implementation were: ## C (D by d) - C has the approximate loadings (eigenvectors of ## the covariance matrix) ## as columns. ## X - The approximate scores ## Matrix (N by D) - Expected complete observations. ## M (D by 1) - Column wise data mean ## ss (scalar) - isotropic variance outside subspace R2cum <- rep(NA, nPcs) TSS <- sum(Matrix^2, na.rm=TRUE) for (i in 1:ncol(C)) { difference <- Matrix - (X[,1:i, drop=FALSE] %*% t(C[,1:i, drop=FALSE])) R2cum[i] <- 1 - (sum(difference^2, na.rm=TRUE) / TSS) } res <- new("pcaRes") res@scores <- X res@loadings <- C res@R2cum <- R2cum res@method <- "ppca" return(res) } pcaMethods/R/prep.R0000644000175400017540000000732213556116437015200 0ustar00biocbuildbiocbuild##' Scaling and centering a matrix. ##' ##' Does basically the same as \code{\link{scale}} but adds some ##' alternative scaling options and functionality for treating ##' pre-processing as part of a model. ##' @title Pre-process a matrix for PCA ##' @param object Numerical matrix (or an object coercible to such) ##' with samples in rows and variables as columns. Also takes ##' \code{ExpressionSet} in which case the transposed expression ##' matrix is used. ##' @param scale One of "UV" (unit variance \eqn{a=a/\sigma_{a}}) ##' "vector" (vector normalisation \eqn{b=b/||b||}), "pareto" (sqrt ##' UV) or "none" to indicate which scaling should be used to scale ##' the matrix with \eqn{a} variables and \eqn{b} samples. Can also be ##' a vector of scales which should be used to scale the ##' matrix. \code{NULL} value is interpreted as \code{"none"}. ##' @param center Either a logical which indicates if the matrix ##' should be mean centred or not, or a vector with averages which ##' should be suntracted from the matrix. \code{NULL} value is ##' interpreted as \code{FALSE} ##' @param eps Minimum variance, variable with lower variance are not ##' scaled and warning is issued instead. ##' @param simple Logical indicating if only the data should be ##' returned or a list with the pre-processing statistics as well. ##' @param reverse Logical indicating if matrix should be ##' 'post-processed' instead by multiplying each column with its scale ##' and adding the center. In this case, center and scale should be ##' vectors with the statistics (no warning is issued if not, instead ##' output becomes the same as input). ##' @param ... Only used for passing through arguments. ##' @return A pre-processed matrix or a list with ##' \item{center}{a vector with the estimated centers} ##' \item{scale}{a vector with the estimated scales} ##' \item{data}{the pre (or post) processed data} ##' @examples ##' object <- matrix(rnorm(50), nrow=10) ##' res <- prep(object, scale="uv", center=TRUE, simple=FALSE) ##' obj <- prep(object, scale=res$scale, center=res$center) ##' ## same as original ##' sum((object - prep(obj, scale=res$scale, center=res$center, rev=TRUE))^2) ##' @export ##' @author Henning Redestig prep <- function(object, scale=c("none", "pareto", "vector", "uv"), center=TRUE, eps=1e-12, simple=TRUE, reverse=FALSE, ...) { if(inherits(object, "ExpressionSet")) obj <- t(exprs(object)) else obj <- as.matrix(object) if(is.null(center)) center <- FALSE if(is.null(scale)) scale <- "none" if(is.logical(center[1])) { if(center[1]) center <- colMeans(obj, na.rm=TRUE) else center <- rep(0, ncol(obj)) } if(length(center) != ncol(obj)) stop("center do not match matrix dimensions") if(!reverse) obj <- sweep(obj, 2, center, "-") if(is.character(scale[1])) { scale <- match.arg(scale) if(scale == "uv") scale <- apply(obj, 2, sd, na.rm=TRUE) else if(scale == "none") scale <- rep(1, ncol(obj)) else if(scale == "pareto") scale <- sqrt(apply(obj, 2, sd, na.rm=TRUE)) else if(scale == "vector") scale <- apply(obj, 2, function(x) sqrt(sum(x^2, na.rm=TRUE))) } if(length(scale) != ncol(obj)) stop("scale vector do not match matrix dimensions") if (any(scale < eps)) warning(paste("Variance is below eps for", sum(scale < eps), "variables. Not scaling them.")) scale[scale < eps] <- 1 if(!reverse) obj <- sweep(obj, 2, scale, "/") if(reverse) { obj <- sweep(obj, 2, scale, "*") obj <- sweep(obj, 2, center, "+") } if(inherits(object, "ExpressionSet")) exprs(object) <- t(obj) else object <- obj if (simple) object else list(data=object, center=center, scale=scale) } pcaMethods/R/repmat.R0000644000175400017540000000166713556116437015530 0ustar00biocbuildbiocbuild##' Creates a large matrix B consisting of an M-by-N tiling of copies ##' of A ##' @title Replicate and tile an array. ##' @param mat numeric matrix ##' @param M number of copies in vertical direction ##' @param N number of copies in horizontal direction ##' @return Matrix consiting of M-by-N tiling copies of input matrix ##' @author Wolfram Stacklies repmat <- function(mat, M, N) { ## Check if all input parameters are correct if( !all(M > 0, N > 0) ) { stop("M and N must be > 0") } ## Convert array to matrix ma <- mat if(!is.matrix(mat)) { ma <- matrix(mat, nrow=1) } rows <- nrow(ma) cols <- ncol(ma) replicate <- matrix(0, rows * M, cols * N) for (i in 1:M) { for(j in 1:N) { start_row <- (i - 1) * rows + 1 end_row <- i * rows start_col <- (j - 1) * cols + 1 end_col <- j * cols replicate[start_row:end_row, start_col:end_col] <- ma } } return(replicate) } pcaMethods/R/robustPca.R0000644000175400017540000002205113556116437016170 0ustar00biocbuildbiocbuild##' This is a PCA implementation robust to outliers in a data set. It ##' can also handle missing values, it is however NOT intended to be ##' used for missing value estimation. As it is based on robustSVD we ##' will get an accurate estimation for the loadings also for ##' incomplete data or for data with outliers. The returned scores ##' are, however, affected by the outliers as they are calculated ##' inputData X loadings. This also implies that you should look at ##' the returned R2/R2cum values with caution. If the data show ##' missing values, scores are caluclated by just setting all NA - ##' values to zero. This is not expected to produce accurate results. ##' Please have also a look at the manual page for \code{robustSvd}. ##' Thus this method should mainly be seen as an attempt to integrate ##' \code{robustSvd()} into the framework of this package. Use one of ##' the other methods coming with this package (like PPCA or BPCA) if ##' you want to do missing value estimation. It is not recommended to ##' use this function directely but rather to use the pca() wrapper ##' function. ##' ##' The method is very similar to the standard \code{prcomp()} ##' function. The main difference is that \code{robustSvd()} is used ##' instead of the conventional \code{svd()} method. ##' @title PCA implementation based on robustSvd ##' @param Matrix \code{matrix} -- Data containing the variables in ##' columns and observations in rows. The data may contain missing ##' values, denoted as \code{NA}. ##' @param nPcs \code{numeric} -- Number of components to ##' estimate. The preciseness of the missing value estimation depends ##' on the number of components, which should resemble the internal ##' structure of the data. ##' @param verbose \code{boolean} Print some output to the command ##' line if TRUE ##' @param ... Reserved for future use. Currently no further ##' parameters are used ##' @return Standard PCA result object used by all PCA-based methods ##' of this package. Contains scores, loadings, data mean and ##' more. See \code{\link{pcaRes}} for details. are used. ##' @seealso \code{\link{robustSvd}, \link{svd}, \link{prcomp}, ##' \link{pcaRes}}. ##' @examples ##' ## Load a complete sample metabolite data set and mean center the data ##' data(metaboliteDataComplete) ##' mdc <- scale(metaboliteDataComplete, center=TRUE, scale=FALSE) ##' ## Now create 5\% of outliers. ##' cond <- runif(length(mdc)) < 0.05; ##' mdcOut <- mdc ##' mdcOut[cond] <- 10 ##' ## Now we do a conventional PCA and robustPca on the original and the data ##' ## with outliers. ##' ## We use center=FALSE here because the large artificial outliers would ##' ## affect the means and not allow to objectively compare the results. ##' resSvd <- pca(mdc, method="svd", nPcs=10, center=FALSE) ##' resSvdOut <- pca(mdcOut, method="svd", nPcs=10, center=FALSE) ##' resRobPca <- pca(mdcOut, method="robustPca", nPcs=10, center=FALSE) ##' ## Now we plot the results for the original data against those with outliers ##' ## We can see that robustPca is hardly effected by the outliers. ##' plot(loadings(resSvd)[,1], loadings(resSvdOut)[,1]) ##' plot(loadings(resSvd)[,1], loadings(resRobPca)[,1]) ##' @keywords multivariate ##' @export ##' @author Wolfram Stacklies robustPca <- function(Matrix, nPcs=2, verbose=interactive(), ... ) { nas <- is.na(Matrix) if (sum(nas) != 0) warning("Data is incomplete, it is not recommended to use robustPca for missing value estimation") svdSol <- robustSvd(Matrix) ## Sort the eigenvalues and eigenvectors loadings <- svdSol$v[, 1:nPcs, drop=FALSE] sDev <- svdSol$d[1:nPcs] / sqrt(max(1, nrow(Matrix) - 1)) ## We estimate the scores by just setting all NA values to 0 This is ## a bad approximation, I know... Use ppca / bpca or other missing ## value estimation methods included in this package compMat <- Matrix compMat[is.na(compMat)] <- 0 scores <- compMat %*% loadings ## Calculate R2cum (on the complete observations only) R2cum <- rep(NA, nPcs) TSS <- sum(Matrix^2, na.rm=TRUE) for (i in 1:nPcs) { difference <- Matrix - (scores[,1:i, drop=FALSE] %*% t(loadings[,1:i, drop=FALSE])) R2cum[i] <- 1 - (sum(difference^2) / TSS) } result <- new("pcaRes") result@loadings <- loadings result@scores <- scores result@R2cum <- R2cum result@method <- "robustPca" return(result) } ##' A robust approximation to the singular value decomposition of a ##' rectangular matrix is computed using an alternating L1 norm ##' (instead of the more usual least squares L2 norm). As the SVD is ##' a least-squares procedure, it is highly susceptible to outliers ##' and in the extreme case, an individual cell (if sufficiently ##' outlying) can draw even the leading principal component toward ##' itself. ##' ##' See Hawkins et al (2001) for details on the robust SVD algorithm. ##' Briefly, the idea is to sequentially estimate the left and right ##' eigenvectors using an L1 (absolute value) norm minimization. ##' ##' Note that the robust SVD is able to accomodate missing values in ##' the matrix \code{x}, unlike the usual \code{svd} function. ##' ##' Also note that the eigenvectors returned by the robust SVD ##' algorithm are NOT (in general) orthogonal and the eigenvalues need ##' not be descending in order. ##' @title Alternating L1 Singular Value Decomposition ##' @param x A matrix whose SVD decomposition is to be ##' computed. Missing values are allowed. ##' @return The robust SVD of the matrix is x=u d v'. \item{d}{A ##' vector containing the singular values of \code{x}.} \item{u}{A ##' matrix whose columns are the left singular vectors of \code{x}.} ##' \item{v}{A matrix whose columns are the right singular vectors of ##' \code{x}.} ##' @note Two differences from the usual SVD may be noted. One relates ##' to orthogonality. In the conventional SVD, all the eigenvectors ##' are orthogonal even if not explicitly imposed. Those returned by ##' the AL1 algorithm (used here) are (in general) not orthogonal. ##' Another difference is that, in the L2 analysis of the conventional ##' SVD, the successive eigen triples (eigenvalue, left eigenvector, ##' right eigenvector) are found in descending order of ##' eigenvalue. This is not necessarily the case with the AL1 ##' algorithm. Hawkins et al (2001) note that a larger eigen value ##' may follow a smaller one. ##' @references Hawkins, Douglas M, Li Liu, and S Stanley Young (2001) ##' Robust Singular Value Decomposition, National Institute of ##' Statistical Sciences, Technical Report Number ##' 122. \url{http://www.niss.org/technicalreports/tr122.pdf} ##' @author Kevin Wright, modifications by Wolfram Stacklies ##' @seealso \code{\link{svd}}, \code{\link[ade4:nipals]{nipals}} for ##' an alternating L2 norm method that also accommodates missing data. ##' @examples ##' ## Load a complete sample metabolite data set and mean center the data ##' data(metaboliteDataComplete) ##' mdc <- prep(metaboliteDataComplete, center=TRUE, scale="none") ##' ## Now create 5% of outliers. ##' cond <- runif(length(mdc)) < 0.05; ##' mdcOut <- mdc ##' mdcOut[cond] <- 10 ##' ## Now we do a conventional SVD and a robustSvd on both, the original and the ##' ## data with outliers. ##' resSvd <- svd(mdc) ##' resSvdOut <- svd(mdcOut) ##' resRobSvd <- robustSvd(mdc) ##' resRobSvdOut <- robustSvd(mdcOut) ##' ## Now we plot the results for the original data against those with outliers ##' ## We can see that robustSvd is hardly affected by the outliers. ##' plot(resSvd$v[,1], resSvdOut$v[,1]) ##' plot(resRobSvd$v[,1], resRobSvdOut$v[,1]) ##' @keywords algebra ##' @export robustSvd <- function(x) { ## We need the weightedMedian function provided by the aroma.light ## package. However we do not want to make the whole package dependant ## on aroma.light if (!requireNamespace("matrixStats", quietly=TRUE)) stop("package matrixStats required but not available") L1RegCoef <- function(x,a){ keep <- (a!=0) & (!is.na(x)) a <- a[keep] return(matrixStats::weightedMedian(x[keep]/a, abs(a), na.rm=TRUE, interpolate=FALSE)) } L1Eigen <- function(x,a,b){ x <- as.vector(x) # Convert from matrix to vector ab <- as.vector(outer(a,b)) keep <- (ab!=0) & (!is.na(x)) ab <- ab[keep] return(matrixStats::weightedMedian(x[keep]/ab, abs(ab), na.rm=TRUE, interpolate=FALSE)) } ## Initialize outputs svdu <- matrix(NA,nrow=nrow(x),ncol=ncol(x)) svdv <- matrix(NA,nrow=ncol(x),ncol=ncol(x)) svdd <- rep(NA,ncol(x)) for(k in 1:ncol(x)) { ak <- apply(abs(x),1,median,na.rm=TRUE) converged <- FALSE while(!converged) { akprev <- ak c <- apply(x,2,L1RegCoef,ak) bk <- c/sqrt(sum(c^2)) d <- apply(x,1,L1RegCoef,bk) ak <- d/sqrt(sum(d^2)) if(sum((ak-akprev)^2)< 1e-10) converged <- TRUE } eigenk <- L1Eigen(x,ak,bk) ## Deflate the x matrix x <- x - eigenk * ak %*% t(bk) ## Store eigen triple for output svdu[,k] <- ak svdv[,k] <- bk svdd[k] <- eigenk } ## Create the result object ret <- list() ret$d <- svdd ret$u <- svdu ret$v <- svdv return(ret) } pcaMethods/R/sortFeatures.R0000644000175400017540000000502113556116437016712 0ustar00biocbuildbiocbuild##' Sort the features of NLPCA object ##' @param nlnet The nlnet ##' @param trainIn Training data in ##' @param trainOut Training data after it passed through the net ##' @return ... ##' @author Henning Redestig sortFeatures <- function(nlnet, trainIn, trainOut) { weightsAll <- nlnet@weights$current() weights <- weightsAll if(nlnet@inverse) { numElements <- nlnet@net[1] * dim(trainOut)[2] trainIn <- array(unlist(weightsAll), dim=c(nlnet@net[1], dim(trainOut)[2])) weights <- weightsAll[(numElements + 1):length(weightsAll),,drop=FALSE] } netDim <- dim(nlnet@net) trainDim <- dim(trainIn) bneckNum <- nlnet@net[nlnet@componentLayer] weightMats <- vector2matrices(weights, nlnet@net) bneckNum <- nlnet@net[nlnet@componentLayer] ## ****************************** nOut <- array(0, dim=c(sum(nlnet@net), trainDim[2], 2)) for(subnet in 1:2) nOut[1:trainDim[1],,subnet] <- trainIn ## forward propagation for(n in 0:(bneckNum - 2)) { E <- c(0,0) for(choice in 1:2) { sBias <- rep(1, trainDim[2]) for(i in 1:(netDim[2] - 1)) { if(i == 1) nBegin <- 1 else nBegin <- sum(nlnet@net[1:(i - 1)]) + 1 sIn <- rbind(sBias, nOut[nBegin:sum(nlnet@net[1:i]),, choice]) sOut <- eval(parse(text=paste(nlnet@fkt[i], "(weightMats[[i]] %*% sIn)"))) if(i == nlnet@componentLayer - 1) { idx <- rep(0, bneckNum) idx[1:(n + choice)] <- 1 if(choice == 2) idx[n+choice-1] <- 0 sOut[idx == 0,] <- 0 } nOut[(sum(nlnet@net[1:i]) + 1):sum(nlnet@net[1:(i+1)]),,choice] <- sOut } output <- nOut[(sum(nlnet@net[1:(dim(nlnet@net)[2]-1)])+1):dim(nOut)[1], ,choice] Epattern <- (output - trainOut)^2 Epattern[is.na(Epattern)] <- 0 if(!is.null(nlnet@dataDist)) Epattern <- Epattern * nlnet@dataDist E <- mean(Epattern) E[choice] <- E } if(E[1]>E[2]) { #change features changeIdx <- 1:bneckNum changeIdx[(n+1):(n+2)] <- c(n+2, n+1) weightMats[[nlnet@componentLayer - 1]] <- weightMats[[nlnet@componentLayer - 1]][changeIdx,] weightMats[[nlnet@componentLayer]] <- weightMats[[nlnet@componentLayer]][,c(1,changeIdx+1)] switching <- c(n+1, n+2) nlnet@fCount <- as.integer(nlnet@fCount + 1) } } weights <- cbind(unlist(weightMats)) if(nlnet@inverse) nlnet@weights$set(rbind(matrix(trainIn, nrow=numElements, ncol=1), weights)) } pcaMethods/R/svdImpute.R0000644000175400017540000001132713556116437016212 0ustar00biocbuildbiocbuild##' This implements the SVDimpute algorithm as proposed by Troyanskaya ##' et al, 2001. The idea behind the algorithm is to estimate the ##' missing values as a linear combination of the \code{k} most ##' significant eigengenes. ##' ##' Missing values are denoted as \code{NA}. It is not recommended ##' to use this function directely but rather to use the pca() wrapper ##' function. ##' ##' As SVD can only be performed on complete matrices, all missing ##' values are initially replaced by 0 (what is in fact the mean on ##' centred data). The algorithm works iteratively until the change ##' in the estimated solution falls below a certain threshold. Each ##' step the eigengenes of the current estimate are calculated and ##' used to determine a new estimate. Eigengenes denote the loadings ##' if pca is performed considering variable (for Microarray data ##' genes) as observations. ##' ##' An optimal linear combination is found by regressing the ##' incomplete variable against the \code{k} most significant ##' eigengenes. If the value at position \code{j} is missing, the ##' \eqn{j^th}{j^th} value of the eigengenes is not used when ##' determining the regression coefficients. ##' @title SVDimpute algorithm ##' @param Matrix \code{matrix} -- Pre-processed (centered, scaled) ##' data with variables in columns and observations in rows. The data ##' may contain missing values, denoted as \code{NA}. ##' @param nPcs \code{numeric} -- Number of components to ##' estimate. The preciseness of the missing value estimation depends ##' on the number of components, which should resemble the internal ##' structure of the data. ##' @param threshold The iteration stops if the change in the matrix ##' falls below this threshold. ##' @param maxSteps Maximum number of iteration steps. ##' @param verbose Print some output if TRUE. ##' @param ... Reserved for parameters used in future version of the ##' algorithm ##' @note Each iteration, standard PCA (\code{prcomp}) needs to be ##' done for each incomplete variable to get the eigengenes. This is ##' usually fast for small data sets, but complexity may rise if the ##' data sets become very large. ##' @return Standard PCA result object used by all PCA-based methods ##' of this package. Contains scores, loadings, data mean and ##' more. See \code{\link{pcaRes}} for details. ##' @examples ##' ## Load a sample metabolite dataset with 5\% missing values ##' data(metaboliteData) ##' ## Perform svdImpute using the 3 largest components ##' result <- pca(metaboliteData, method="svdImpute", nPcs=3, center = TRUE) ##' ## Get the estimated complete observations ##' cObs <- completeObs(result) ##' ## Now plot the scores ##' plotPcs(result, type = "scores") ##' @keywords multivariate ##' @references Troyanskaya O. and Cantor M. and Sherlock G. and Brown ##' P. and Hastie T. and Tibshirani R. and Botstein D. and Altman ##' RB. - Missing value estimation methods for DNA ##' microarrays. \emph{Bioinformatics. 2001 Jun;17(6):520-5.} ##' @author Wolfram Stacklies ##' @export svdImpute <- function(Matrix, nPcs=2, threshold=0.01, maxSteps=100, verbose=interactive(), ...) { missing <- is.na(Matrix) temp <- apply(missing, 2, sum) missIx <- which(temp != 0) ## Initially set estimates to 0 Matrix[missing] <- 0 ## Now do the regression count <- 0 error <- Inf while ( (error > threshold) && (count < maxSteps) ) { res <- prcomp(t(Matrix), center = FALSE, scale = FALSE, retx = TRUE) loadings <- res$rotation[,1:nPcs, drop = FALSE] sDev <- res$sdev ## Estimate missing values as a linear combination of the eigenvectors ## The optimal solution is found by regression against the k eigengenes for (index in missIx) { target <- Matrix[!missing[,index],index, drop = FALSE] Apart <- loadings[!missing[,index], , drop = FALSE] Bpart <- loadings[missing[,index], , drop = FALSE] X <- MASS::ginv(Apart) %*% target estimate <- Bpart %*% X Matrix[missing[,index], index] <- estimate } count <- count + 1 if (count > 5) { error <- sqrt(sum( (MatrixOld - Matrix)^2 ) / sum(MatrixOld^2)) if (verbose) { cat("change in estimate: ", error, "\n") } } MatrixOld <- Matrix } tmp <- prcomp(Matrix, center = FALSE, scale = FALSE, retx = TRUE) loadings <- cbind(tmp$rotation[,1:nPcs]) scores <- cbind(tmp$x[,1:nPcs]) ## Calculate R2cum R2cum <- rep(NA, nPcs) TSS <- sum(Matrix^2, na.rm=TRUE) for (i in 1:nPcs) { difference <- Matrix - (scores[,1:i, drop=FALSE] %*% t(loadings[,1:i, drop=FALSE])) R2cum[i] <- 1 - (sum(difference^2) / TSS) } result <- new("pcaRes") result@scores <- scores result@loadings <- loadings result@R2cum <- R2cum result@method <- "svdImpute" return(result) } pcaMethods/R/vector2matrices.R0000644000175400017540000000252313556116437017344 0ustar00biocbuildbiocbuild##' Tranform the vectors of weights to matrix structure ##' @param object an nlpcaNet ##' @return weights in matrix structure ##' @author Henning Redestig ##' @aliases vector2matrices,nlpcaNet-method setMethod("vector2matrices", "nlpcaNet", function(object) { netDim <- dim(object@net) posBegin <- 1 posEnd <- 0 result <- list() for(i in 1:(netDim[2] - 1)) { wSize <- c(object@net[i + 1], object@net[i] + 1) posEnd <- posEnd + prod(wSize) result[[i]] <- matrix(object@weights$current()[posBegin:posEnd], wSize[1], wSize[2]) posBegin <- posEnd + 1 } if(posEnd < length(object@weights$current())) stop("weight vector has too many elements\n") result }) ##' Tranform the vectors of weights to matrix structure ##' @param object an nlpcaNet ##' @param net the neural network ##' @return weights in matrix structure ##' @author Henning Redestig ##' @aliases vector2matrices,matrix-method setMethod("vector2matrices", "matrix", function(object, net) { netDim <- dim(net) posBegin <- 1 posEnd <- 0 result <- list() for(i in 1:(netDim[2] - 1)) { wSize <- c(net[i + 1], net[i] + 1) posEnd <- posEnd + prod(wSize) result[[i]] <- matrix(object[posBegin:posEnd], wSize[1], wSize[2]) posBegin <- posEnd + 1 } if(posEnd < length(object)) stop("weight vector has too many elements\n") result }) pcaMethods/R/xval.R0000644000175400017540000002715113556116437015206 0ustar00biocbuildbiocbuild##' Internal cross-validation can be used for estimating the level of ##' structure in a data set and to optimise the choice of number of ##' principal components. ##' ##' This method calculates \eqn{Q^2} for a PCA model. This is the ##' cross-validated version of \eqn{R^2} and can be interpreted as the ##' ratio of variance that can be predicted independently by the PCA ##' model. Poor (low) \eqn{Q^2} indicates that the PCA model only ##' describes noise and that the model is unrelated to the true data ##' structure. The definition of \eqn{Q^2} is: \deqn{Q^2=1 - ##' \frac{\sum_{i}^{k}\sum_{j}^{n}(x - ##' \hat{x})^2}{\sum_{i}^{k}\sum_{j}^{n}x^2}}{Q^2=1 - sum_i^k ##' sum_j^n (x - \hat{x})^2 / \sum_i^k \sum_j^n(x^2)} for the matrix ##' \eqn{x} which has \eqn{n} rows and \eqn{k} columns. For a given ##' number of PC's x is estimated as \eqn{\hat{x}=TP'} (T are scores ##' and P are loadings). Although this defines the leave-one-out ##' cross-validation this is not what is performed if fold is less ##' than the number of rows and/or columns. In 'impute' type CV, ##' diagonal rows of elements in the matrix are deleted and the ##' re-estimated. In 'krzanowski' type CV, rows are sequentially left ##' out to build fold PCA models which give the loadings. Then, ##' columns are sequentially left out to build fold models for ##' scores. By combining scores and loadings from different models, we ##' can estimate completely left out values. The two types may seem ##' similar but can give very different results, krzanowski typically ##' yields more stable and reliable result for estimating data ##' structure whereas impute is better for evaluating missing value ##' imputation performance. Note that since Krzanowski CV operates on ##' a reduced matrix, it is not possible estimate Q2 for all ##' components and the result vector may therefore be shorter than ##' \code{nPcs(object)}. ##' @title Cross-validation for PCA ##' @param object A \code{pcaRes} object (result from previous PCA ##' analysis.) ##' @param originalData The matrix (or ExpressionSet) that used to ##' obtain the pcaRes object. ##' @param fold The number of groups to divide the data in. ##' @param nruncv The number of times to repeat the whole ##' cross-validation ##' @param type krzanowski or imputation type cross-validation ##' @param verbose \code{boolean} If TRUE Q2 outputs a primitive ##' progress bar. ##' @param variables indices of the variables to use during ##' cross-validation calculation. Other variables are kept as they are ##' and do not contribute to the total sum-of-squares. ##' @param ... Further arguments passed to the \code{\link{pca}} function called ##' within Q2. ##' @return A matrix or vector with \eqn{Q^2} estimates. ##' @export ##' @references Krzanowski, WJ. Cross-validation in principal ##' component analysis. Biometrics. 1987(43):3,575-584 ##' @examples ##' data(iris) ##' x <- iris[,1:4] ##' pcIr <- pca(x, nPcs=3) ##' q2 <- Q2(pcIr, x) ##' barplot(q2, main="Krzanowski CV", xlab="Number of PCs", ylab=expression(Q^2)) ##' ## q2 for a single variable ##' Q2(pcIr, x, variables=2) ##' pcIr <- pca(x, nPcs=3, method="nipals") ##' q2 <- Q2(pcIr, x, type="impute") ##' barplot(q2, main="Imputation CV", xlab="Number of PCs", ylab=expression(Q^2)) ##' @author Henning Redestig, Ondrej Mikula ##' @keywords multivariate Q2 <- function (object, originalData=completeObs(object), fold=5, nruncv=1, type=c("krzanowski", "impute"), verbose=interactive(), variables=1:nVar(object), ...) { type <- match.arg(type) if (inherits(originalData, "ExpressionSet")) { set <- originalData originalData <- t(exprs(originalData)) } if (is.null(originalData)) stop("missing data when estimating Q2") originalData <- as.matrix(originalData) originalData <- prep(originalData, scale=scl(object), center=center(object)) nR <- nObs(object) nC <- nVar(object) if (nR != nrow(originalData) | nC != ncol(originalData)) stop("data and model dimensions do not match") if (fold > max(nR, nC)) stop("fold must be equal or less to max dimension of original data") if (method(object) %in% c("svd") & type != "krzanowski") stop("Chosen PCA method must use krzanowski type cv") if (method(object) %in% c("llsImpute") & type != "impute") stop("Chosen PCA method must use impute type cv") if (is.logical(variables)) variables <- which(variables) ssx <- sum(originalData[, variables]^2, na.rm=TRUE) for (nr in 1:nruncv) { if (type == "impute") { nP <- nPcs(object) press <- rep(0, nP) q2 <- matrix(NA, nP, ncol=nruncv) seg <- list() nDiag <- max(nR, nC) diagPerFold <- floor(nDiag/fold) suppressWarnings(diags <- matrix(1:nDiag, nrow=diagPerFold, ncol=fold, byrow=TRUE)) if (diagPerFold == 0 || diagPerFold > (nDiag/2)) stop("Matrix could not be safely divided into ", fold, " segments. Choose a different fold or provide the desired segments") if (nDiag%%fold > 0) warning("Validation incomplete: ", (nDiag%%fold) * min(dim(originalData)), " values were left out of from cross validation, Q2 estimate will be biased.") for (i in 1:ncol(diags)) seg[[i]] <- which(is.na(deletediagonals(originalData, diags[, i]))) if (verbose) { message("Doing ", length(seg), " fold ", "cross validation") pb <- txtProgressBar(0, length(seg), style=3, width=20) } j <- 0 for (i in seg) { j <- j + 1 if (verbose) setTxtProgressBar(pb, j) test <- originalData test[i] <- NA test <- tempFixNas(test) if (method(object) != "llsImpute") { pc <- pca(test, nPcs=nP, method=method(object), verbose=FALSE, center=centered(object), scale=object@scaled, ...) } for (np in 1:nP) { if (method(object) == "llsImpute") { fittedData <- completeObs(llsImpute(test, k=np, allVariables=TRUE, center=FALSE)) } else { if (method(object) == "nlpca") fittedData <- fitted(pc, data=test, nPcs=np) else fittedData <- fitted(pc, data=NULL, nPcs=np) } ii <- i[ceiling(i / nR) %in% variables] press[np] <- press[np] + sum((originalData[ii] - fittedData[ii])^2, na.rm=TRUE) } } } if (type == "krzanowski") { rseg <- split(sample(1:nR), rep(1:fold, ceiling(nR/fold))[1:nR]) cseg <- split(sample(1:nC), rep(1:fold, ceiling(nC/fold))[1:nC]) nP <- min(nR - max(sapply(rseg, length)), nC - max(sapply(cseg, length)), nPcs(object)) q2 <- matrix(NA, nP, ncol=nruncv) press <- rep(0, nP) foldC <- length(cseg) foldR <- length(rseg) tcv <- array(0, dim=c(foldC, nR, nP)) pcv <- array(0, dim=c(foldR, nC, nP)) for (f in 1:foldC) { test <- tempFixNas(originalData[, -cseg[[f]]]) tcv[f, , ] <- scores(pca(test, nPcs=nP, method=method(object), verbose=FALSE, center=centered(object), scale=object@scaled, ...)) for (p in 1:nP) { if (cor(tcv[f, , p], scores(object)[, p]) < 0) tcv[f, , p] <- tcv[f, , p] * -1 } } for (f in 1:foldR) { test <- tempFixNas(originalData[-rseg[[f]], ]) pcv[f, , ] <- loadings(pca(test, nPcs=nP, method=method(object), verbose=FALSE, center=centered(object), scale=object@scaled, ...)) for (p in 1:nP) { if (cor(pcv[f, , p], loadings(object)[, p]) < 0) pcv[f, , p] <- pcv[f, , p] * -1 } } press <- rep(0, nP) for (p in 1:nP) for (fr in 1:foldR) for (fc in 1:foldC) press[p] <- press[p] + sum(( originalData[rseg[[fr]], cseg[[fc]]] - (tcv[fc, , ][, 1:p, drop=FALSE] %*% t(pcv[fr, , ][,1:p, drop=FALSE])) [rseg[[fr]], intersect(cseg[[fc]], variables)])^2, na.rm=TRUE) } q2[, nr] <- 1 - press/ssx } if (verbose) message("\n") rownames(q2) <- paste("PC", 1:nrow(q2)) drop(q2) } ##' Simply replace completely missing rows or cols with zeroes. ##' @title Temporary fix for missing values ##' @param mat a matrix ##' @return The original matrix with completely missing rows/cols ##' filled with zeroes. ##' @author Henning Redestig tempFixNas <- function(mat) { badRows <- apply(mat, 1, function(x) all(is.na(x))) badCols <- apply(mat, 2, function(x) all(is.na(x))) mat[ badRows,] <- 0 mat[,badCols ] <- 0 mat } ##' Replace a diagonal of elements of a matrix with NA ##' ##' Used for creating artifical missing values in matrices without ##' causing any full row or column to be completely missing ##' @title Delete diagonals ##' @param x The matrix ##' @param diagonals The diagonal to be replaced, i.e. the first, ##' second and so on when looking at the fat version of the matrix ##' (transposed or not) counting from the bottom. ##' Can be a vector to delete more than one diagonal. ##' @return The original matrix with some values missing ##' @author Henning Redestig deletediagonals <- function(x, diagonals=1) { wastransposed <- FALSE if (dim(x)[1] > dim(x)[2]) { # matrix must be lying down x <- t(x) wastransposed <- TRUE } nr <- nrow(x) nc <- ncol(x) if (!all(diagonals <= nc)) { stop(paste("Order of diagonal number", max(diagonals), "is out of bound")) } indexmatrix <- matrix(1 : (nr * nc), ncol=nc, nrow=nr) finalmatrix <- matrix(ncol=(nr - 1 + nc), nrow=nr) finalmatrix[,1 : (nr - 1)] <- indexmatrix[,rev((nc : 1)[1 : (nr - 1)])] finalmatrix[,nr : (nr - 1 + nc)] <- indexmatrix dia <- 1 + 0:(nr - 1) * (nr + 1) finalIndices <- NULL for (i in 1:length(diagonals)) { indicestodelete <- finalmatrix[dia + (diagonals[i] - 1) * nr] x[indicestodelete] <- NA finalIndices <- c(finalIndices, indicestodelete) } if (wastransposed) x <- t(x) return(x) } ##' Get cross-validation segments that have (as far as possible) the ##' same ratio of all classes (if classes are present) ##' @title Get CV segments ##' @param x a factor, character or numeric vector that describes ##' class membership of a set of items, or, a numeric vector ##' indicating unique indices of items, or, a numeric of length 1 that ##' describes the number of items to segment (without any classes) ##' @param fold the desired number of segments ##' @param seed randomization seed for reproducibility ##' @return a list where each element is a set of indices that defines ##' the CV segment. ##' @examples ##' seg <- cvseg(iris$Species, 10) ##' sapply(seg, function(s) table(iris$Species[s])) ##' cvseg(20, 10) ##' @seealso the \code{cvsegments} function in the \code{pls} package ##' @export ##' @author Henning Redestig cvseg <- function(x, fold=7, seed=NULL) { if(any(table(x) > 1)) { if(any(table(x) < fold)) { fold <- min(table(x)) } if(fold < 2) stop("too few observations in the smallest class") res <- sapply(unique(x), function(z) { if(!is.null(seed)) set.seed(seed) tmp <- sample(which(x == z)) seg <- matrix(c(tmp, rep(NA, ifelse(length(tmp) %% fold ==0, 0, fold - (length(tmp) %% fold)))), nrow=fold) },simplify=FALSE) res <- do.call("cbind", res) } else { if(length(x) == 1) x <- 1:x res <- matrix(sample(c(x, rep(NA, ifelse(length(x) %% fold ==0, 0, fold - (length(x) %% fold))))), nrow=fold) } res <- res[!apply(is.na(res), 1, all),,drop=FALSE] res lapply(as.data.frame(t(res)), function(x) c(na.omit(x))) } pcaMethods/README.md0000644000175400017540000000277313556116437015172 0ustar00biocbuildbiocbuild# pcaMethods R package for performing [principal component analysis PCA](https://en.wikipedia.org/wiki/Principal_component_analysis) with applications to missing value imputation. Provides a single interface to performing PCA using - **SVD:** a fast method which is also the standard method in R but which is not applicable for data with missing values. - **NIPALS:** an iterative fast method which is applicable also to data with missing values. - **PPCA:** Probabilistic PCA which is applicable also on data with missing values. Missing value estimation is typically better than NIPALS but also slower to compute and uses more memory. A port to R of the [implementation by Jakob Verbeek](http://lear.inrialpes.fr/~verbeek/software.php). - **BPCA:** Bayesian PCA which performs very well in the presence of missing values but is slower than PPCA. A port of the [matlab implementation by Shigeyuki Oba](http://ishiilab.jp/member/oba/tools/BPCAFill.html). - **NLPCA:** Non-linear PCA which can find curves in data and in presence of such can perform accurate missing value estimation. [Matlab port of the implementation by Mathias Scholz](http://www.nlpca.org/). [pcaMethods is a Bioconductor package](http://www.bioconductor.org/packages/release/bioc/html/pcaMethods.html) and you can install it by ```R if (!requireNamespace("BiocManager", quietly=TRUE)) install.packages("BiocManager") BiocManager::install("pcaMethods") ``` ## Documentation ```R browseVignettes("pcaMethods") ? ``` pcaMethods/build/0000755000175400017540000000000013556147214014776 5ustar00biocbuildbiocbuildpcaMethods/build/vignette.rds0000644000175400017540000000042213556147214017333 0ustar00biocbuildbiocbuild‹…RËNÃ0tA[è…c¾ _UB¨BUÕ«»ÔRbG¶CÄoæñÒ¤ 8x홵ggmocBˆO‚È#~Ë`!‚1ï ÉæûR#äÛ†57é«l\âVÕ¶\¹i•Ó·{Åzv,õ¸ê¥’÷N+eU[j…’.ÿQK“FØ}‚ÚXëIZ­XŸvÿi°b»ÏÜÐ`ÇŽ¥f½::>wŠÒÃ>/N"½ùeþ'ú<§}%iÉKFŽ —¢à¸a-ì7^²¥[zèî:ã— /ïæ™4JšhÕ¤Xì®û ŸÚ¶=\:Ê jÐ’1ƒÇJwÎ:~ꯟíLpcaMethods/data/0000755000175400017540000000000013556116437014613 5ustar00biocbuildbiocbuildpcaMethods/data/helix.RData0000644000175400017540000004656613556116437016662 0ustar00biocbuildbiocbuild‹]œxTUÓÇÓ‘"ÒŽ l_Tæ*JG¤AD@é½Jé)Ò›ôÞçÒ;„@:i›ÝlOH„@Þ{Šfž×çy?öƒdg÷Þs~gæ?ÿ¹]¾î¥{§×;AAÁÚÿ´—!AÚÿ (SBû3tèàáÃ&kÿ^:  ä\Ww÷œS ZÁ¬Õw Zä=ƒ½{®h]ç]Þ{ où7øÿœc˹_ñ¯~ "ƒÒô¸ëj²}öÜWx¡ÿñò—jgâÉsà çlÆË£óϦ x‹'óï,,‰Û÷ï²ñPk\ñÅ_k¿ª‚§²ŠJ{õ3^ >±Ý³—x´w©w"¿ØˆwÔ¹åFþ¹ï©5®Ç“}ÂwÙû5^?5– ˜…ê{å®L4ïÅ­ù~¸y žÍ}§Á$ï·xáó-n<ú=^ß÷Ú>|±öó‹æ?ýÐTOëËîj™}Õy'oÀûx½l›7?…wó§¾PmÆës;÷:S¿ï ìÒhtÅkëί¿™6ï•ßרiw^‹s´øª9w¬Òøhð›Eñgó0ê£AÇz]ÆÇ%S’£<ÀÇ ¹ž­ƒßùbôާïࣻ«š7¼9£½§ƒ|¼7<6+ûã½§ul¨í9Þií~»qZ>ùiCؽí Ó«Ì£r0úf™» ûoÁGí¶•ªõFZâíü¹Ÿã“ [§ÍÃèÈ!Kº¦cœ²ýß2Ò0ÆðdÛÀŒ||ìý¶^ô¬1fÏü+õfmÆù½â«õªû}p_Œ]ãµ~Z#ãÊú¯v¨²é˜ï®„1A¥ÚžëoħG’æ¼™†Q®U´ýCŒÿ£ÛóIßÏÆ']Óž|´ «˜Ò\ÕZbü7½˜ÙñÆ<0?t臘Ô~Ó©y‹]ûUì•Ý’1¶àÉݮߕÀäÍ7ë^MvàÓ„Ú‘¶_á³N‡|Tç&,÷æ¹gÝľœÞÅ'Ú§Ytý2&¬v¼¸ˆIß.?Ròn&&Ô<¼Ýßg >ûì§>ÕD`úÅŸëOÿæ>¾¸åÚÙa˜2oà+Õù˜ú×—?|“rÓn è1Ûƒ)ONõùkV[LÑVÇì‡iâ>ø£q¦¦¼ÿòäBL¯Ýµì‡¥Œ˜¿³pg#f´œµà»˜úÛ+öI1½ä’×…ƒ¿Äô–ïš:y8f¨.OAË·˜1Îy|ÈÖW˜¡-†A.ÂŒ¸_~ ¬û¦Éû”öàpÝj•?Ãô¬÷^mý f^L}qZM̨ðlp÷럢­J©ìYïLÇŒÚçÝûÏìÖÞÇŽ1íÄôêö «23Jžò ™íC•_Ó mQý“Bµß;qúãÎëÑVp€­DÌLd7¦/¦ Q{× ÁŒ!•V¿€6W³ÏÒ>íöß¾¨v9ÀŽöùMz/|VÓÿž¹doååè`»¤^ ¦³»øñ.Ì|±¾s½“.Û×ÍÀôZÿ´{½2Ú²™áZÜQ«Qm3Ц/ñå­ZÃÑi:óêh³ŸÑ±£Jû„­»0‹­ò/Æ M®ÛÐ÷µKö÷¿ó%núíò¾ÛòæîúlY´ßmѳë7fNªöç{Ó¶ {ÆþNw¾:ŒÛ„æ-Ô&è8þÅá¥ùÚûÊuë¿jJyÇû˜9òɤÑS2ÑycX茊sÑÅ>U§­˜5àË…æö7ÑYÑu¸ÌG?¡«^Öäè7¹˜u£BíÝÏó0«Ì6óó%ÝЉ›~Üvòzxßûü¸ ³–ž9§m1tv2O8S>ìÿkøf½I¾¸¢’½Yì”D—¶©G7žw«Ž™°¬6ºÛ ;½¿zµ¿ümeº7~|u`yt?)Ív:o-b;Ý•öŽóüGÌ’ûÖ•õmúªTz‡L ßÒû9fÝØ@Cºf1¢¡#añ1¥Ž£KòÒÉñØݽ¦|7szö7´´®=¿¦ÏȲºÐ“?­¤µ2ºÏLYä;®ý{¾¶£<èk<úþ­ èûñ£ÔùëÚõýr|‹u-Лöi¿Ô¾{Ð|:« zg š°½÷·Ìͨ{}|[@ψ6ùCÑÛÆ±§zæFÌ®Ìo º#ª}ª÷å OÛœ}®EßÒN 9è{½wÎ/{]èŸðâM­¢OÑ/¹êÿõ»#9ÓŽ£èØ#5¯mB¶Ù->#úµMøËoèÉ;¸n÷Ör蛲ûdÉ÷h÷£cTö¾Btïhòõʆ Ñ_{h‰èµ•Ð_±WZåÁkï[=ó¯èÑ{áÒÀ‡o ïƒ6J꘿ѧQíDj ôë7•¾ôl'f/ù­A’“1»¿¦{Vbö®²›R&ä£Ï^Ô«y׳è;½¿nìÓÍè½S‰‘Rûu=‘]V£?(9àÑéŠèedz-}=b½³³§F­6¡_Û5ÏOµB¯<ï|óîÄL»™Þ?Su'ûç §p¹Î¼è(ú¯ÞöÞ}ýú%½æÛeÕÉÁlÉCi¢ èÿùá—9åžbvNˆ=„Þ²Zëº\D/£ÒXŸ7¸Ub¿ŸÐÿÑöM“þú½9¸ÑWjófC¯ûè¯þù&îžè«Ìlô<ÅȆ^§qõWO¢mVê‘…[Ñ'¹çöõKõßÐkpýUqDú%/½ç -£­¯í†Ÿ`ô³¥èmµ`êÒŸbÑWîàìƒÎ…è=uhFá¹èÕ¨rc}yë7gWl‚ž¿ÙAkC_D301ZÜJ64½„~y¾{6/ølÇ×_ ÷«pvr ·ÙàÇFíE¿v(~³Xû}Á_ïÕ¿ß Èꄞ“Ç¢£ƒÚ£'„-ÔDô¶ëظñ›“èIϼvýOm¥öÝ}'ø:z#<¶öm¢Ñs¯nxÍß¡§eÅ×|Úú)É/ztµW¾†=è ,蕯­·o[o¨zÝcNLþªéô$<[Þº©‚à`@¿] ´û–¼zËÁ'èìÛÅþɶЭÁiœz=üëˆ^ޱéè½ø2w\H:Ûû7ô9<=ìÉEè\^½EL}zÞë|௷5Ð}fEÌ®³§Ž_¹ê§ªè˜tÀ›’«½ïŽÐ®õ*¡ó«o´¡}N¹Þ]'¤õÏCWÓ_W )Òîg»%:¬G翜ëÿCSmIcÖõ–¿.ߨñäïKo°¢c~•Ñ=>E—•ä1zÇ¿¿3ꘊÎ+ßø¾z>7•5„8ÑŶúÊtì^}èMÌŠc ôcV핆Í팮/î~÷tüž0ñøùyèHz’Ò!t:¥~ùçªëKÖAûߎ¡ú—Œ…Ÿi© Úo_‹_ìCûìªÕW~‹™S¤ÜŽ{Œ™Ýnÿî5 3Ù·×-ƬР;F6Á,Ž÷˜)÷™ý©áÓ„“[ÐÞ?›4è8½j|ç*è¨Ì.híÜ>—dú r8XÖÝzÚµ¤nB™Ž˜õó+}¿:[ÐQïËã NÅ¢cɰ&#ç¬ÅŒKì†BG[ø#ÑÞöçi?Tõ¡£æhFT̘þÍÝ µŸct¿mü¯»a†\ǶÎ-#]£1#¸š3§ó´iÙÞÒWÐÞjó¸¯<çÑöIž¶ÂzcFÄÕç)æw1óæO‹æ|Œ<=ª¡->ùÑé÷ù}¦h‡sç–ˆéZv«¥˜ÆÀ‹é χ¸1Cæ-©òNc«yÐL]Pöh\ãY˜–Øæ—¨Ä£˜Æo3¦¼{¨ÅðZ|¶|Gcª<¯ÓVßb;S%ÿRã.7wEKLš¸PËDÞ†ý&Ï®ZSY68y6¦k›}¢ÑŽ©Ú*É83Ód‘p“SÒøAÉ,+ÜÉüXí„i‹t¹mËmÀ”_ËÞx”Z:¬ÿ!ìÅ|vmäa˜ÒùÏímÖ\äÉ=Y¦‰žóZfåÀ'³ÝÑûˤ`R=üc˜ûLbØ~ñm,pŒmÊ`Œûå˜ð ?ßÃÄÊŸ5zô/|ʪƒçññЫ^8b<û±â06žwôf+ŸZ+k+£!ÆH<íáÔ2â9uèu£C Çc̪eŒ¬Se‹Ùß¼‡OµCÛªº1‘e==WcBÓ5Õæ|p£:6×¶.Þ¯Îâ#m—WßsoiIY•R9øHæéwg?Õ*ž7ø°ãú6³ðVÅ?ïY‹žàc-ëøâÚY‚<øHÞטé£.Öÿ¢/>l²~\Ñêxÿ·ò»ûjbÔüÞírý±xGî›ë2O¿;†]ñÆ`–°4Å3;։܊*«Jš~€w_}áUxmço,ÃkZÕ©¥xýŸç–1Öâ5–ÝüXÏ_â •Øxrr5¼üûoõx6“tŸàÅΟ޸àèŠ;ÒV‡®éÄ«!Á¡þ¼"ëÄÃÇcSæ[ð°všø6Ã}1ý½Që ž ­ä½y÷5«}úœ;Wí¹Øæn;níSÒVfÅ82tbjòÛ°öæý…‹?Ý ë–NØâXúwêCJ¾ý Ï}ؾFZËž8=óEËKáüÓž/O÷™€m6]>ì:üÍËã°OæÑN}<©g§±¸ïU¨õõÎΰ&¯cuW áoèRX¯ÇØó½-.£Ô<¸T6áκ•{àöF8UuDËHë08üûÓ¯ó÷€+öv²Ã‰Ï4è5γãhÁY¸!ø WÙic=7D~ gÇvÕÙ½îÞÿaÃÕÅpmSÇ}-œp¥`ŽcÆÇÓAÛDŒ(p…}ÜáÏàÖ1v€ÂíØ¹ìĆÈZö+Üû°Ö'“𭀫Úé”SwDÞ7îü³¨¯í<¨8‚e²pãÅšVãfÇBO“‚á Û-mw'j@¤üûûbÿCô™]¾ÿª<âiR{x¼ìö÷E!=á¾Ýùƒ–ÁŽéÊ%êqx"ÎaˆaeÊN<Ý•Þ5ÿA$¤à'ñãO¶ÕÖJ_ˆ~þÑäJ-„Ø¿>þyEÏG%ö?<Ü€Ø1ÉÃ~u ž Ý‹¼’&Žíû…ž2U¡Â3ˆÎºw:ºâ×7†fpõ]vbBÂR¶€CÜ4^¨A¬%þóZ•~ƒ¸Áûö†¬ ‰{YA» âÙ©VÆ1Zu7ýÐHíVêæ²¿…$¡@RÕìDgó¿îCò ‘Qv|&$É÷Ü€-k]°|>$l`ÕJˆ›YúqlïÛò]ºóroBÊW؃–õ« i,Ûþr!¤jEñ({eÈù(¤Zúy–Þ’D¾I~º½ßRH9:ZÛ¡U 󃱺s†´ØƒÿüÝ}#dˆ|R—óƒRŸtd‘!-§³V™ì´ðv¬’„ Á H:¤Íý|@Ðß!=⯠ u[A:©üX ‹çÌܶõ!¤f?ötù­ ¤ý6=,uãJ°=«÷ëÎÕ·!½–úuƒŽo ]K.ëwé6ýÑðñC@fÈWóž=héIL¨j¶÷ÎJÔ…Ìf¼ð‡t‘Ç‚m‘ïØþ´?ÁƲ¾ê|ÙׂL¦’|ÿ Ød‰æÈÈÉX¶hL2Ø7¾ºÑ¨j_ÈZ7ädö¥<°s,}™Z5ñ*á#°}sêêÄ’/ S~nGZËël‹À!×uæÈ9kþêÑ ì⼇Ìž¸€£p°ro²,ëþÜQ—CÖvVHWÇ^žxhŸ‹'Ä`—û6KÔÝà`ßÂ:œl[ÿó=Y42{ñƒì¢.‚,‘߀Ëûudù}õòÝ«æZ1à”ë×ác• œå˜Ñ\ÐNŠæÅ囑àêóû«“ÖƒGèK%ôÈzsBÛqµÀ͇oÀåË^Ü,ìƒnv:Ùö8µ,·ðeip²»ò]¸ÄyN.G}YÝ}&¹âÀÝyS(ÀÉÔ¯ÏKoáÀ/´ n¶ío€Kœ»à~Î.Äeð·|Ûwå#í:½—1eå…à¾ã?cVp—~Ã2@ðÈuîyÈØžàdrήà”ýeÍïUðü9¨ÿø†[ÀÍVGõnàÎh¾gã¹DðÉï噹{bMd ܼ|h Þkó~r÷}p³múËð‰ú ¼BG÷¡ ÊÒe·Á+×…ë2O<Á›RûýzoÇ‚—¥ñ—߀ËþÏ„Ÿš6›ßiðÊïçYÃ1ððcQ{?F™Á=ÀÏÔÌÞ'ÀWÚñè½ÎÁϲà´¿gx”ö{¿÷Û>v>øµ‹òÙ}xãË598ü$øu<ÏÔÔÖ‡ÞŸÞ†<_­Ù}ÂwŸÐGÀ·Ýßgô²dð.¾<ìÊgSÀ'òPðÊõàæò#€w!æÀ_í«´À+ÎSðhUDÅË—€};­”ï&HÖ/›¿?Ǻvý,¿—ÀMà¶%콌©àÿ¤øªð |Í/¬ý“ü5¯mÜ•Þ üò¼ð²ÝQÖ^‘G‚Ë!£ÁÏeæàûùóñ¯»m—ÓÆA¶Ü/ÞÏÊLjµ`øOÏ2Œ¸þüBGßYuøÃX¢[üwت~ùû¾ú0¼qÝÇ´ßt¤.>ŽO3xµÝÏŽƒ•¢ú'_è àzøyš¦]gyß}ëÊ?®ºqøÔ'y_Ÿ¯ ~VUý9ò|–º ø¿±° ¾#?=zM~VÍ—^¾M&‚ÿ ;` à°ÿ§+/-à¿Çþ»¯}~&̯ïw×YÅ Ù¬¼]”^VæU¼^m‘4é½ü}yá ž½ ØÑà¯7ksÛÎs!gK€Oä½àg²Èk;øØvªT¾¼&Wÿ¢Ý碖_ni >¦¦6^yŸ=‹ÿÔ2ëãà™Óø’vôƒ[[,†ËûÀÏd Œåàõ8øêñ‚ \,}øexù±Õ¼¼|ú¼lU*4îh«²ÛIÈîÙâÓÔ2àºÕ×Í[}«hÁÃ>MàwàÝR•(Ú÷á:#ødžäyË pkŠO·ƒ›©ñƒß‚§Õ;»;4™ É/w—€Ñ_Á­%¯m”TpœñÓ†°ûà–ëÍóGc¶‚À½—olðVb‚cðj§ìÌxp›x" ®—ö¼0|³9ÜZUñqÆ`ðñô¾78™¬ñ4œ2OñJN»DŸ¼ÃFöÚÑDã©ëZD¬{¸D=¦ý)Îßfp~øwà™Ÿ3ÀÅT·nwÀíÙÍ2ðj»Èµœ=ß?{wh[°kÕÿK.‚ëjí†/€¬ê¯W/Ø™:Óë'°u\®Aàhú8lûÑÑ`_ìÝ^âÈdÇuV>Ø®œe?öÓk?ùèÛs`û½Æ™Î®ß!Eæõé·Ú÷»7(Ò2[e÷¨\ÒµÍó]›Š`‹Y¡ އt¹®lr¥_;²e}’–ŸXY'Òå:Ìð4Xb>¶Ë™‡ï¥A:o_4‡4ÞNøl2ÏI•y{òšk­µÒz´ :ñÞH]Poÿ²ƒ eUc̪^ƒ«àÙÆòô³þ I×úž¿rö[HVeÂý’§µ<ï{¶Ó •oß~*¯{ºÐÁ´ïÇ0ð¤Éýš,óæ´ó:.W”†x¡3C ìN.œÚRXº½`'<ã²G;HfYß;»á™ßóÅï›–Bb!KHjC’–å·»Sqr¿Æ‹~$0yyT$²«Òm"<åÇç—ÿ=®þ„'ZÓ2ë'H’çO© ó¶f”…x¡³C< ³d5Äup™T¥6Ä Ýb°78ÉÕg`›™÷ V«¶þIVàá Ö`L‡(Öõ¬}¢yzðÄj‹¥W™á +&v?ƒ'³n² bò;‡Ät~ 1WO¼ÐRdˆj°íÙYËFx,ú/ð„ËÁC!JËz?8Z ¢˜ŠTw4<âÛ²<ÖVñ¾½¡#Ïÿ»U}möpÁ=¡¯ÀÃmüƒCT˜5&cÇ\¸%úSp[ô}àÖŠORÞ­:î} ®°¯_æ\f*wï¿!úóö%²–>€+—¹€7µäpZ¢î¯oWúË’ƒà*oÕ‡»Z£¥Êp£çëaÚ?ÁÕ›;Ö׬önŠúneÂéh¸-ô#¸/×óÕd8ÿôX0œªcdä‚“ç-¢v7Uô£áJVœú‘ áz-_„¾ÇË÷žR©rœn±pï¢uá„õ¹¦å*ªö_<ý± ì©=#hQõBXYoЗ_~ Ge ‡kŸÁ™-ç~¹3Å»~õöXô9 lkvgƽ׳ìâÇí°£‰»GÌÃ`uÁÑ’»ëÆÕÚ·ˆyµ w©e©r³»óÄý\&/~¶º>|Žþº‡2›½Ü†î™Ý[šv½¼½ññ}ï¿ú¶ì¯dkÉdS:zcêÛjõ†Ù_þ÷}³µt}EÑŸèÙ´mñ“'£oö•ôE•Ñç½ûºÂÉñèùúµjóêmú˜œ°é.Êü}üöD¯Ô¯¼<=Ó~þ]†ô±öæú™ªÚn z¥þdzêÀeèarßkÐÇåÛÐ{˜7rÐ;† ÞUпñ‹ð[íûkï'úg¾fËY$ô ~¡ûV©qÿÜn¾¼¡>Öûizx]}L¿v=?óÿг„ ÙÐÍ–MLºEŽÞÛ¬q6=¬ }/GW-3ÓÞW^/¯Ô{½\ŽŒ@öá'Liˆžé\Gß'¼q^–éî K+–[dž¡Wè ècryÌJôp[À—è©¿ÛlºŸƒ¾Ø‘lÅ£×r¤Ý©°3è*Á„Ûèæe`gô±tæbúÊŽõ½c3¡[êƒ~‘'k߇çèïÎ/zJò†úÙ_‡å£—Uѽ—7Ñý[— ÞAwäy¦l¢›©ÞÝ£ëGÞ¨@7k_9¡ýyþk[©Б;ŽeÐèaex¥½èÔŠÙûcÖ1&ôvAÏÊ®S þLE'_Fu0KƒX͸5è²²Fƒ]¼MSÚ·×v"ºyzÚ]¬½¶ø: x¡„Î¥ü`Ä,¶‹?)YÌ…Ñ0óáǯú}»í­ç}ôÞÃÙ˜Ån˜µ¯‚ùï}¿¡=©‰–™OB§àfõZfZÑ m²/éz8eúÕ­ÿéöÎO&i+(ã†k'ð.t±toÊ´Ëþ@ë*‡•A;/óÇ`V‡§m믴Ï]Áœ èýu?ÆCûX³Ñ.}!Î}\DÇ0&œ5@»Üw™v.p`Sý#£CË^Ú®™Lþ¸aGû§DÝ<³mZ–w©#fæõ>9ýw´}ÚÆ±§ÚD=‰¶x…öf/·jHE›¶ ¤ÜAÛeÓGcºÔïmrŸ¦3ùÚö7¦¿á Lç¦1÷C$LexHsa†80“·%J`š¨Û0UÔChë7hÁà5‡1}L©ck³Ò0“—³?`ú¼àÄ´¹™t˜*ûãéoYâ» Óe?1 †ä&<[ŽiM/idXŠiƒ/m=÷a{Lã6—/0ÙÞsÕ²R[0MûöÚ%ÇÔ©Ö9÷`ªÈ1íý—'¤ÃDvÜ÷`jC¾@1e30åß÷ g`ø “e*Žƒ&S¸ô>ceÝ”U˜°§/Û˜$õáÄŽ…éÞ=„ lU•oŒ‰3Þê¶þü9ÆU¸òýåñû†öw1NòáYÁå¨Í7ëa¼–%û=-ð©ã›>:ƒO¥(Aò8^ú‰¥_#^öažðmXSê2eqÆj§:œÅ¡ àcÙo"ûpѲ+ò*ŒaêeiÄ[²Ïþ˜Ë¨“1Ž—‘V¼?‰„›ñ+{>ÿ£Ú®~ç“S³ž VÉÛêu0ºýG`´ôçD4ÌþµìM¼?‚`É®rÿãÿéí‘\>ëƒ÷YW§w2F¶?z¸ùûñ.·×4ÃÛkwzw^G¼)ÎG¼ÇÚƒµ_ãÃYVgRo¼'t,¼ÉÛd5ð¾Ð‘ð*³5äœÇ›²w‰·Çá5.#uÆk²/~fvãŒE½Nà-?3îTûZ6OW‹oܧf2^õ.^õâñ.LÉÄ[¼Ü™Œg¶»Û}g\8øWl<ÏÚé Và¶Z­¶í x‚X{λç^r[§lé‚«ì­•³á=<øîÃ{ß/½Ž§fï;swYlÜU% o«.¸ÞÁï¸üTùéÉaãø¤—#?k KyšRÿY=´~Óc`µ¬‡wžÚªâgóq¿ñÉ´íæ8>}Ûë!λ°iä‘ÓW¿ëǺ˜ÐbôopbûD¦ÈÃÉu5‡vïꃳK"N÷ß=Nï)Ô¹lÂÙ6“†~0ì!ŒHìÖ°cÀO0îá"ϧÃëÀÆ~ÁeºŸÙ»S{M(UõO¸ üWpIìO¸Âèøáß0»þ®“¬Õádhøî‹#ÃÉ*ëÀÁVU4;·4˜ïºŽ_Òˆ95 d¿®K½êžVåôJ« 7e}ýXôwážÈ#à»ý¯×@$?žÃàá‹™/¾Ì…k¼Ý7TŽëoá¡ÌçjÙ]½µë fî«?ª‘â<…(© =fmÅËk z{‰ý;V†H©C?<ÍR¥Ußï­ø1ª5© O¸}§)<þ/ˆjTµïôØ pëÃJ{n@ oßt‚;Í™¨%<(hº' 1—ðã2î_5å}`ŤîÝ Ée»ˆ–õV,Ëþ¿Šƒhá#„¨u „ŸB+ó;-†'"_ƒ'e¸` 1â„-éúõ숕õa²à3Ä'–˜uì«pˆ—:gÜßÿSkvˆãiù8ˆ=;é÷±@‚¬Óž-g‚þ8HÎæ†Häv¯i,óÛ¤àðœÏo%C·)Œ‚TÞV{ž1[cËuÆÐvʪœM!)ÿ¥VAç@²8w!…ÛšóƒÃOVCªð9Aês&°ü ©'7'Ö˜2Òǘv¼ÞiRwIÞø0K¬ƒd‘/AŠà$Ëú?•• Î[’¼—)œ|€­¨òwúÍ!SêØ¢ÏC¸P\î €LýØê]ï'AÆûÜ`鲎I‘ú’á`ÇzHûªZÈWóRÀ¶„z ý/ ]øcÀöÖ×ÿV)->OW¾ÛéµÌª-dÊú<3¦ó“%Ú@ª–Ív‰Òê\^n| ™“+ÁÊÌÖ`ýaí}"X‡ÌÝ×wØÙ1pz-¸ØÕot¬]íº™ÂöµÃßÕ­ØË¹}ð»2Ù2€U)|V`gånÇ5`?’3íØÛë`cå\€¬üL‘;oç~ vÙÿ°1y»æ|p•ÞÇ:Šà~+p0whéY`¿ŸÄ”È\Ò蛎v'8…ÿ œÂN.? s Fp²O&ŒØ.©Ç9’+ܾõspØþn¼ïÑ ð°ö}cW·E쎃«+´Ë‚KôÁûÎK;Kê±nfûêèç ^`€ƒ·Ýꀋ57‚û¯YÌqÎë•§ÖYØÜZ1d9Òþ?×!úËàŠä#pî˜[HÇÀ+üŸà^Þ·AÞ•åàýfðÈõçf¶®¯~7/#K‚KrË5éðOaÖ'àÔNKËüÞàçx¯°7º îD&X¸ÃR™BžQ,X^žÀ#ü³à­Ì6ÎprÃ%øØ±úÑ}ð Ÿ-¸¹L£?OwÇ€aqõ"ðÊÏï—ûÏÍå†ÿéÿÿéq˙ҼýÛYt'€›c!â?}ÝóäÇI×ú^ϧc™¢^y²Yzp)¾ø¥®šýœÀÍí#WÔàùøú4ÛyñÜ;àgí…ƒÀ'×›OêþûF­âh~᳟ԹÜÌ:µ.xð‚|Ÿ^á³Ow)–X»?d粆Ðûàý}u)xúžüÖ÷ø¥Nãgéò¼Sà?Ëçà™XÓ»´Lx¥Þä‘Üòóôi&dsùúGðÁFð3F«ñ-õv_w¶òÁ/ü)àå²F}ð·™ywUóFàåí9+ä°U×·!øGqãøt'ûg¿y~©3úã¹ñ|L&;|¼ÂG>áK—äpvUžˆþ·³ÿ½ÿöäºó3ùê]ø7²ÆñCð3Y|Î~ð¾—z}NðË~‰_ê…^–v˜ƒGöý¢]¾Çß,‰z§|¼]Û|ÂǾm¹G½uŒÝøÖ©_Þ×ç«<ßÙµxá>ŽçÚ:f·yøyþ%xžñ|"ïŸ<}F^x‚Ÿµ›j?ðwƒŸ·é>ßH~¡À+êIð _¸n°ÄVãÒZô3)¼?p! ¼Â'žïîÑ=û\L.8ò7xx[¾"¸eØ#ûênáKëj–¨áÿ/Çvdðÿ´8k߯ÿ\<šê}|n¸dÆ#|ÜàéËa%û&nVæ.?®&\@·ô¸…?|͘á¾*¸kÅ|žô$ÜjïZÁ¿ß;¶³£Á-ûšÎò¬C›¤O|á€SÔs%÷S®3'—G¿§› „àfÕ]§Òàà6÷R%|Åàâmÿ6à,|À&À)¹à>Hpðq à:8M¬¡ŸYòû;3ƒMupËþCø|Á.û~ümƒÁ)|·àdêFë«`gòJƒTpH_F—¡ÇB&;V-ƒLÖ^/L»ð¡‚ƒÛÏêAfö¥ÜMy£À!ó7ûœîGŒ;ƒCø)ÁÁÚ¢«Ÿƒ-—BtZ¾Áì uÁ&Ïw{û6 íñ$ÉÔÀÏò!“ÙuAúKÖ@j£º±Ê2+pcØ…n6~ÛHó#y™ßp° +Ø»°B82…ñÒåz³ =Ò¤Ni~HÈäédoHmΞ(H“¾Œtá[…´ŽÜ` é-x ©’÷iL¥¡‡4nC®©\3Cb¹Þ[žÞ(„tÙM“ý®¤è_ëá¿@:;RBŠ\ILÝÑ„TÖ®Ù~2¸­®•}¡šðìô¯Ÿùü9¤”bƒ;A:›&ÑÒßgF¾ùÄl1íîBJ[nƒÔsm?_˜û¤ó²¤)$s[HHbß"x#ıl®K$1»B^,ijéšþí %hÀÇ+f_…xéãHámîŸá™ôW$ìaƒ!™ÿ¸¢ùrœɬË3>âëG²Lb…♽7þ!<•z|ÒÌð`€X.w ‚hù9“åúNdåÇQ²ûoÃ9ˆ{t¯Lé_Þ@dÊó»-zvƒH&W×[™ÝèÕ-ô=ˆœWºbåüW+û½QÌ•úHÿRÜbv ÷ƒG븀¤O"Rø×àž¬?Î ¿%ÜŸ2|b{xôËêœÿœ€¨Ü¶åÖ/x ×d>{¹#8ïÂ=Q§Â]™·E‚ívõ÷3àºð±Ám¹oUჅGòzÜõÖNw¾:ôÈ ×Ä< \å¶ÖFpFæ×·e鞘¯Ó?MÛ¿±íf¸rš%Èá·ö;B‡€ã”è¡>ÑâI?ÌU1ï7™ »_*œš°ÐûË ¸xªê‘]€ý[Y½;þkIòV½N™Ú»‡lº ²å— «xÙ\öG^êÒú.(½¹ó‰·SñЂMU»Uîó=žC×v–ÀãŸ>³­.“:î? î´˜Q¾3®¨ØíÈ÷Óßâvvµ§%Á|Óˆ©=’õx<³S7&íj’°<ùÓ<ܺ)cÏ ['\]cƆQÛqI®cü­¢³¨æ°–‚WV2aiž¬kly蟩xFúá÷¾Êíøu1¸ Ÿý6˲÷þÞ!aªm î]øÛ†AåòP­t;¢Ú§xU;„ú½·Õ±¼@ÄE_ýz²ªà-=?ÀPç.îZ:¬k@/¼9™2xYê©{ó aö‡x›Û£ÄË¢ÿgº—y¼$~$ÞéÁ'x—ɘá[ð®è âާx——W#ñæ¡÷Yåw¤îy—Û«`ÔÙ»CÛ\:…ë0£ŠÑ¢nÀHQà•Û‡Ùd FVd‰ñ|*}¼ïsáˆ|Šz £¤þXêè°ÄÔ‰Qrþã±È§1u‘›ïÅ'R§Žbîw}ÆídÞøXœƒø˜ËUïã-«ùÞY ٸϜhŒf.ºÉŸ`¬V¼³²+>fÝåð$Œ“sbñïn`)Œæc7ƒð©80^ú¹g0îb}­2¯)¢¾Á˜Fó,¿¿ƒ‰ÜÎbÄxÞ~lƒ‰Gxƒ˜ ½y7Læ2H'Lf] 1¹7n`üž1·GçŸÃ8é»Mú >“zyŠð‘aŠôѦqÛÆxŒå2N>û‡eŠçÕ^>›úíÔÓ˜.çDRÚ¥`—·{a·ùMǦBv¯Ži]XC`¦Ký9G[6`F}.@ ½L^•Û‡Ó0U®‡´QÇ0¡#¦Éõ™ÆíwSÑ&tL•óxiÜÖÜ 3E_Ó.ÌdfÊù¿T9wbc*rS¦‹sÓx;f:f^â $L_\yg…Üþh‹]öò31]Î9ÚøñÔNû9᯵ñ´ ?f,:rkQkÚ:ó‚mrfhYr÷#&̾9Ìäem{̾PÌäáJ¢]ûÔ%fC»Ô-íƒ s.-¾Œ6©K;¥Û>õ½ëÎoÀ,©ß:E>ŒYr>ÐÉd—Òk1«z&kÿ^"kéýãèÐ6]3]ÐÙ,];¹Þ g¼D'O¿‡¡Éak‹0‹Ïd¿E—ÔݳD~‹®ïyƒÚiþ:j<º¥¾én=Š9 ÐÉí-Ññ|I7Ó™tÊ>”‹ÛV[¢Kø¨0‹·uF`ÃèÃuè>t ß ºüÎ*ôp9³ºÄ¹‹Y‡Ž[¹¢‹.á×@7K¿FD7¯ëŒîfSï}ûr :™MÜåEï*~  Kö?Üryùêt‰zÝükŒC·èï¢[~¿ÈOÐÉìzñè‘ó•>9?âu::™=åë•ÿÍÿ¹E_ƒgžÚ0oú¥_ßǪÛ÷"ÑÏËé‰èåÇS5ísŠy&k´|Š>^–ÖCÏfœJCß 6`„y¼ýk¶ºùÓ!ôð±È©èËOgz˜ `)^Ù—‘~%ô48;òòLô²¶ü“%èþ=ôÊy°lÞö˜†~^Æõ@ï].iŸ×…èç6¢ºèû®ÄÇMGðù WømÐ+çB³YÕÞã úD}¼üéŽÒŸˆÞ^.ÿ‚Þ§¼ C//¦£oÏú(ë—ÿæõü"ÄlÙדu8Jú˜8³ùe™Œ>ÉÏç¸þý~Þ†ª†~ÙGv úd?Ç/çþ×ʺ°¶¸]e½¾Ÿ˜a§Z¢¯gÕÑ?V#„n¡F]M(äu°!t5âÛÈÔásU#xÛ»ž!üfjDG>¨Fˆ|ZXøëÙ%¿5P#Ä:R#ÄçW#ðýr¡‚Ôˆó¼WÃ7p_ˆ!t75BÇ„íÊjøÖ௦†Ÿæõ§!t íç’þ|ÑøW5Bèõj¸ÐGÕðÕò÷yy¯¨5¹¡B z‡žÂ}vjøs®ë©ábÞ\¾X5\èæ /5LøÕðº¼VƳBív‘¢v…'j˜ðÑ©ab>Q ¾5ÔÁ÷ÆÝj˜˜³TÃ:íg‘ÕP1W«†MarúojØ f¯¯ª†ŠsJ :‰šÃç*ÕÐeÚ»kUã¿qBE>üßÏ…Ua™ÓÕСòïÙéÝõ¥*êb5Dìk5T^§Pá'TCDݪ†šÄú •÷;Tè jè׺;=÷ü¡†°±”O©!̶õŽ*úxj(»=¾PCÅÜŠ*æþÕþ65ÕPqN¨¡¢nWCY{£ÊV5dï¨!K¹/B :‘,ò5ä[^À©!"¯SCÊñ>·ÌdøÛóÕáwUCJ/b“jˆà¨ÌÇßÚ«ÁíµU0Î¥†¼ßJÛ‰jðDfÿk«GsC§Ìþ:!O îÏ j°ð]©ÁÅç 'ÖCp*[øûÔà?x?D :ÌÓÎÕà£ÜÏ¡ ΩA§Dü å|nS .¸©Á#¸ÏL jÌûý®$¯WSéfhi´\/AÝØ‚UÕ ù¹ƒD¿¥ø÷÷rÝN Êå†i50/D5h÷y«©<1Tƒø˜@I5Hô«ÕÀ ¼ï«rÜkïSyƒöß_jÜ·A¡â¾й 5P«Æ´ŠY <ÁD5p1×[ÕÀl¼c´8Ÿç¨7ؘÛÏjàÎ=5P̨AMׄŸ[ s¯j ðÕªL&Ù2O HáF5°++¨Â'£~/8ð–0j·3”Sdü¶M¿Ý¥ˆºU ý5@ÌÕ¨|Ÿ`‘ÐMÔ~™Û`‘˜SZhÛÏv‹Ä¼— tX|+yÐVþ9ûHÔÖ%,£ôùtÃ÷Кmpíÿ|Æ"9ÇY$ÏÍ·BoÅ7²ÿ_$ý6Eb½ã9ß]$ç ß }‹ø×Â7ò¹oäœî¡oãF±þÀ7B·ÄB™¿¾‘ói¯Ùm¼ … åœ|¡œ~#çßHŸL!·Ó¶ÅBÑŸÃBÑwÀBÁ ,äXjˆ…‡yƒ ëpa Eÿ xúú Šû‹…òº¼–yÃë†|câkÁ|­*¿Øc°P~97] üøøZœ?X ¿Oô]¼æÛ ¾–óï/…þНäóE^ 8¿:ðq¤ÑXÀlSíÀí·ûÄ9ùJœ XÀ.oô[|ÉÑ:øêž ã+éKyÅíbð•ð×ã+±ð…|žÂË |?áKá§Æ—‚CXp†_ÎÙ·ºdÊ3Ì—þ–W[¹q _ÊëÿR£›óøPÌ—¿ŸÏ¨f>‰/żæÏä¾’s”/Øñ×ì¾:æË|$_pó™M§º_ˆF¾Ì˜/æ]ð…Ìwþý3OúûòاíìÆ<±¾1OÞß\é;{!ö+æÉç—ä‰ó_Èõž'óÍ—×÷9+wÖßÂ\áOÇçlw~à×>¯xþDî~^`®ô¡åÈçwäŠ}Ž9òûæˆóŸ³¯±q7>—ï—-ó»n# Äf{½» ŸË}÷\ø^1[Ö£ÙÒ'÷œ—·e0û5¿±˜-ïkö%Ál¡·cŽÌWŸËçbäˆs³Ùc%²÷c6O‹>ÁléËýAÌýÌæcÆ?£OÎ{Ÿ§˜K%þ²/~¡£_>Ÿ"[®¿ÐSѧUU @Ïsž£_a@½^ásFé‹G_r¡Vi¡ì+¡[Ì} WÎ-{„¾^.·¡sâøGƒ sÑ+ü«è‘y¼/ŸÈ ×ö¼ô̼yèÝÍpè‰z‡9Ð)tB´3›az!zfSúÍíëè•Ï‹q_2úD¿N«WD=•É ñÐÅä¾»Ðññ´…ïöé‹RÇ,Ñ_G¯ÔEœâÜÀ,Ž£ÆèfíÌñÐÅÛ<3ÐÁÜM– ‡¹ÅÞ‚©?dqùïKÌü–ÆÐÁºQÜè>¾,©8y¾!fñ1Ý è”ëÀö 2Ñ–óMc&ÿ8hgn‹’KÑ!ý¢6©ÏØN±Ö´Ã éÌ>£L1oŠ™Lv3Ù¸°{6Ú„éÒ¯˜¡oÓØq¡"Ú¤Ÿ3C>O(íOþ0Cr"ËPSµ÷¯:dËý˜.Ÿÿ“šÍ<1UøÀ1]œ¯˜ÊÇ“yÓ›³ ÓgTœóxÞ$L•sÇ)ì]¦\ÀTù¬déÏM‘çRj+>pißžÝ61é ¦ˆ<“­l`y ¦ÓäsH’XÙ‰Ïnp#¦Êçr$ñÇŒDé¿ÇD¹Oãå|t¼ðg`¼|þV¢è‡`JõGÛkìèÉrÎ>Qê˜1q¬Yãåó¨¢ä9;” žó’çÎb¬<—b¹ŒÔ£EŽQ˜0³ŸÈ9þhÑÇÃhù<„Gl¹ï#|*uÒ¨÷xƒ£¥îU2ï邲Ç0’© _yñªÈãñ—=»ãSùܰ{ìéRðç…;FŠy<|$ŸkpCø0R>¿æ–0LáU1‡Š7¥ÏôÖ¡®ƒÓza¤˜?Åbo°t MÞaŸ®Ì6<݇ ¿xWú ¯Šþ+^>¼ÈÓ/¯æL<= ²ÎèZ𜘃ÅÓýŒƒ?ÿë<ÞsAx•Ù]w®Ás|,d .üvë¤àá›ðøR?f<ÚÇÖ“8÷#tžbÝsyêB<>ÙÛ`Hýžx¦vÏ‹r þ¹'`Ѱ©^øëÅ}ÀÄøÀ¡'ïÁ–G85úØx\ä\WrÔVX[µÔ]µ_W˜²êÖõ€NGaß²3kÃmÇ­ß¼-[nâkÜÜh×õüá8—Éjâ±¹U¢Ê÷ûN×½ºL¹Íprl¯RM—»àTѣħa¥àOgÚÁÑUÍn´ß×N–»c ïfOl‚Ë«ÙôBì³öà´'þšê0ûëzp©Û‰¥Í3çÃ%‘×Áu9Oq ( .l(}IÿαŸ¹úY×ÕuK 9ßyQÎy œKF9¿z¶ã\e¥ú\ûð›Ð•=–`f*<àãÉ­à›–î;nIܽŒ2§šî«W–2!õ¯O¥ìá…Ækâ”)ð@úÈbvò|мžlήø©ÿ<óp_ÎgÅ>àºÏÚ8†ƒð˜Ï›!FèQÅܸ£wB´Ð—!F<÷žVbΛ ²;b!^<bŒ¿ )ýqyH4VXd¡÷‚ô'Ã3^®@ÚÎnÉ¿íz ©Ü&<Òe?fpCšœÍs™iÒ?”!ò,È”ó¨ÜÞßÒäÜ}{àË)ÏGt9'”ÁQa+Øä\l†ôßeqyy؆ʹ!SøhÁÀO`¯ÀkùxÞÄœ];Á.ïc¦ôAföÊV.8¹Ýa2Øy¸Z`—}ÊLÙ¶ËyÛL¶¬‚çƒSø¯Á!ž‡Y‚¯àþ<§œÿrþÌ©ç‚í»~ïíZ>ìB—‡,é7sŠº\Ò·æ•ýH÷ß|À¼òó»k”a“à’>9×&và†ƒGÔ—àaÓÁcnƒ[èrà–>·,ñ\p˹}/,Jx½¼qûŸŸÅ/êðìç |ÿ·*|P¼|yLúo¾ØÇTû‘½ÁÏvkÙÍ Ÿë^ñ¼ðkIrØÒ³à—ß#[¾v9.ü@¶x~ø¶r}üLNm´üZõª¥ˆà—s 9׸ž~¡£A¶8—![οf‹~d‹¹È–ónþ¶µŠ¨dó²­*äH?OŽôãæÈÏ‘ó!kŒŒ©ïBŽÐàyÿ€#ê2ÈÏ·„\Ñw„çr^9ç)/\à¹ü>9¢î„\þ5 ð\ÎÉåJl.“Ooì…\¹žrÄ 3<ú%äJnçŠùÈ•×å9/;Â!OÎGæŽçú äJÿW®ô×äJ?]®|^F^0/¨!?nékÈ}È“sÕy¼ j/qáž‹üòäçÉõ4¼óL/tÈõ1¼àã·ý!Oú_H_j¾ôS½zäK_T¾ÈÏ _öã_ðr¥:¼àå[?x!ôkx!êÈ“þÝ|éw|)ç}óåúx!ç1_ŠçµÁK¡@¾x¼”þ–Ò§—/êHx%tVxÕŽïàK Oᥘû—¼¼ÿ ò….ÒõJú¯ ä\ø+ù|ŠWr®û•ô ?ˆçiÂ+1W¯DŸ^Êç¼ú ¼–~ª×rN±@è¤P æ PÎuˆçíA¡ô Ê÷+Ï9‚¹o^kYïØ#µàõO|@ ¾â åùùZ®÷BQ?@!_¶#á5s+U~oäû¿æËz¼•óó…r¾‘>¿7BŸ·lÛ•L…·¢_¯E^Ëõ\˜Í'(”ëäô¯ Þò4ú;x#ÏÏB‘WC¡\ÿoKs]Þ†q}Þ6çƒaP$9ôFÌWÁy.¾‘×÷­xî” ¹]$×E‘¨Çá­ô5ùWÿúZ<éã+’×»Hî§"¾}g@‘ø¼J€Â0‰óB ×­HúÛ‹D¿Џœ3S ó}J€à‚Ї¯O%àg¾n”€íü”€¼ÎWÅÜ› æz”q]•€‡\¯S˜ü“üX srJ@>Ts«(ø¦–æ %àµXÏE‚J@$¯ãµ?¸—)ÚîÓJ{% ¯k%𾑕À‘ÜW§Šy)%@ô” ÑçS×ò9<%ðΤ(ôi%PèMJàu®¿(B_V¤þ¬ ÝP þ %PpJ ý?%¨-׋•@±¯• ¡£)A—¸¾¡ }K zÌõ>%¨Ï” Íü P…ޤåç†$ê>%踞ý§‰çÑ)A—£ñ¯?] þšƒM þN\—`ÞNQ” Ë|nI žÇç°•à?ùúS‚W”à™|žD }%%8œsK º“”Ë,%xŸ{S‚ÿá!%¸!×c•àg¼¿¡‰sO ©"?ÏT¾¾•nìXh¦„Tc_¬¶²‡ï%XèYJ°xî„"úÛJˆðí*!òþ…lãƒvJ0o³uWBÊr]T -Íõ/%ä-¼"öŸ"×eÈÎI%ä"”"ž_§„~(Á6îgÕ~Ÿ-,í}ZóX YØJè®7*¡¢ï§„ØDƒ/Tô •P17­„ÄòçÖ)!…`J¨Ð¿•Pq¾)¡¢ß „•d‰E]%TôE”ÐR\ðVBÉ÷óñºç¿ýÎ.Û°åý¿Ô’±’ìYžîúG¿í†«‡ýEÇ7}oíÌk’_2Þ÷ô(—Ük÷¼99à«|õCƒ=+3Ëš¿üwÙ3{·—ýýç¾ûµöå~×”÷6O~½—ÿ ½ëX“ùÁ/Öíô†w ÞÙüÜq&9»Ï}ן™ôÙ¹ ÿÓuïçv_yÓ—ýñÃ7¾ã·/»ÐZ¿ô’7go¡ßÉ}œÍý¨­¹qøª‹¿âËÞÿÖ§··Òº“_ùZ‡ýhãÜó÷m}úµoù·ÿæñ îøçùÚŸN?ïæ®ßÛCߨ¿êèëÚ¼ýß?üåÃ7^lË/ù§WÜyáÇý·á¾[]rÚË–ú¬ê÷üÏSo{™¯O?~»­øÏ?è¯+>mϬÿÙU?÷½.¹óþÄ%'Šþø•÷ÄçuÏO_úO;ßÿ6{ò‘“ïýV‹?zׇ:~|Ã[þó=ÿ¼|ß8»/ü®ýéɘïó†/^ûõÅ7{Û'~\ý§þ‹þ߯½å+g–v Äç1=ëá-–÷þÌ׆õi+ï8ùÜ?¼ð¯ñøÛúªëÿ˼ô“_øÙ¢þÖ ¾öÒ“~l¦?Ñ~Ðçömmúò=¿öY¬Ûî+®{Í W7YM¸nÛ–ìÂÿ^kå/ÿÐmÏ´½Ü&ÉþOýþ ¾â‘JV¼ÏONsËc^ðó×ÝúȽŸ·•ÿ9nòÞÖé]éÛW[SX¾ðÈøÖÙ“÷Zç[Òo™p|ëa}®[¾oìȬû¼è¯(8pèŒu§Ûòs¦õÿ½Âmó^ûD·ÏJo¤­ûÇ:~3ñÓe †¼™ý·6]v·÷Õç¿6=a/èúÝÒïüí&]“lÓ«OÏøø\ás³3¯ÏÖeÍ;o~¹¯ä>ìÏß³ï]sª®ÃŠÚnsÖdض¬Ê.YŸ›]E#—[qÁ[“aõS?úÍËßþ#{,k<¦7^lϼ950¶õGé²?a‡~ö©ÍO·Éëì±OÊ*>¯?°~›ìÊ—Ï÷EoüÕ{/øÛOÙMw÷||ùEß×s²5ÉíþÜn›Å:½ú‡éBöUá~ØÚ‡RÃc¿ûÁVsÅ—þÁæ…uzŽ=cÿÚ½ér‹yòØöœ²uɶxÿÏmÛ±ä…ñó•ÙÝš5!~÷‡ó´™<‡Vîÿ’äö¯~ØþŒ}½‹ûÌÀ7möø‡‡¿¸Íв«ø™½öÝä²?¹ÙÏ:mM–Ù£ö‹°^ã~›µ÷ëWýfâglñÿ\fYj7¿ïßž:Xà¿á÷×ûn¿ëÞbÿ¿Ëú©~õ»³þ¿|^ÖJ]r¢ØõyÙ£Ç~|C²õ|ÑcÉ?Ûª¤Ú„ýà?áy”,,¼6»eü_°“+ŸL7œÝ¾è_>ÝÖ”üÜØÓ^ú÷‰~‡Mb­L¬Çw¾›³Ù›ö¯«JìAöeuø¾?ì®ÍÃ>mc?‡Çü*[]õ_ ~Yônùó`Mç=wÅç^ù‹«Ï·UÁú†ìÝyÇѽ¶¤Úï>òÐOíÉ=åÁŸ³yÁþ[S¸/öpòô§1°À<ù™/µÚ†°µÞmvɸìÿÊëØµÉjù×Õ¶rÃOÆßRÚjs“Ç諬²ü®ä¶üm©£2ù©•¯OoŒìµcßmñ³?ºï…GþIÏÅW횺ìÃÅUÖ=6q¨Ï[&» ìµ™¬Ó¹ì«šp<ÿÇéH¬ÐGž·ÕצÝgýþêçKo±½¥/ÉžÁT¿w_²Òž¶jöS!ñÉî7¬»}ÿš›mvr˜žUÞ•ÝGž\íE7¿¬ù¡7¿7îûeø¿_|ìöŸ¶u°•ç¥ÿxé’§³óo´}mâŦ.³Š~ß›ð³àÇæ„øÂWã—îụ̀Ɋ·2ü|&Ø!ÿÉ ‰áþ£•à—VgÒ dmÿ0æ{cÿXj…écû{+HÍŵv_Öˆ½ôÞM6?øGŸ›}8ÙÌ qŠ-~Oú€l6÷yûyAv5¾ë'OÚïB|as²7ï+g–ÙïC\æó“}£Þu^“6üÄ×ñüe'a­Ê^TÖ3ºöË’›ÒÊ7§Ëè"×ï$aâP»ßü“ýÿÿ-ö㬰Ï|Éñ_ýeÓ?>¿·5Äqÿ-Âÿ.MË öd°«¶ffê`}Uº=/õ<ׇêÓ€ ®›g2©c÷‡ˆs “§}룦õÃýóöÙ/°“îó“ìkíƒù& d‡?”ºÕ/Øo³Þ²bíq_kòo´7u°¶8 ýû‡¯·ƒðk°C³²7õC·åâ =Ç_`æ¥îúlë÷ÿöçWµ‰t¼ùìÿGBœm‹·ù²—½ßò‘ôÊ-„™_ŒÇy‚û•«ì)ìë[y>‹~Ÿ.Pßt¸úÙ7¬›i SŸËÜä±ÍyÂîH¶×7gñ€ö­¸2]v_° .¢uôp6xÊ.q›ü”¯'., þÉ*N3[÷Ú42²gç·ÃºS< ø×þ3œ§¯b_oÇnÿ2\Ÿ×§Ëì ÿyXßöxêžoóZò–®ŠO5›J˜âèÒW¤ŽÁ—¦G»ÈÖâ÷ðÿÎïšüé†Ô@}Û’(4»d¼,‰òÆt{ùÈ’ø/Yb÷•ÈÞh}Í öØ*³‹ãò#—ÙÁŸÛ–÷¤Hôë"NÞìœ+ûëæâ¥íiØù¯ŠƒìnÖãÓØÃg^ѽ«ñ×u¾ˆã±¾âý›ƒßM\¼=uwy1Ïÿb¿$ß[òda^cëÈW¿çWÉòÂ×§Øçñ¼ óñÖY®ý?' ÷®ô5ÄÃÊó¶g­Ñ¸ªë|}øòò2â…¸ÿ%î_™˜õkF¢íHÝÑt_žX•ë^cºOaÏ}MrcûÂ/ûfìÊýY+õŠîÝŠß¼9õ¾ûäß}ýÉäÄÛí©ØÉ—üÿWkìAüÞ,üŒìó}Øù2ò òOûbØ_ž¦ãõK|vºŒ^)»èÕÄË’trýr[–æ&×úx»™Â %Oø—Âó‹÷çñl–³&óC[†=.×»]ßû*~€xÇ‹‰›*V¦5Æ÷qõ!N°eÁžÇõ½1ä!ö0~ùQüŸÖx†éùÎN+_FF¦ã»Öùæ÷Ø ü^CsbøÇÚÃ?ñ%ä)˳Þí÷ý/ñ9|oc’ÍüøFßÎó_pƒègÖŽKx¼OºúOŠßlUbæ¿pÐæs=åÁ¾y ~þ©‡E;µéeßÞ”5õö}ò€Š4Ìúª—¥·õ¾1yÜÛá™GƸ£"Ù…·]i÷s˜&¾qÁÛ“'㵉¹uÚKðk›‚?ó€­MÂó‘Ù¾*økÿÞu‰#m–¿0å!SsxU¼žºp?my‚ð˜~ççÄï+C~ì›Èç–rK«ûÎëHÂìç~ç?ÃŽ¶†¸Ê7I?ò‚$Jå/¿(¿ëÿï÷²îô¼—NH>øi'NóÕ‰y.ú£k½pݶ(Ľ.ük5÷±ä“›³Ö•ÇõPøÁtzIðƒ®üqmš6~Õ'mþ»ôFÇ8}hßö0Ú7å¿öÄW§n𫞙—&èÂi\ñÊ÷‰/¼)y`ïó‡ÉÁEì#ÜÙqây[›DcïóÆÔ¾Ý篫9¬£ï•…xEvÃ7†|•þ‚<ÄWñ<N\;çÒ'^-ûá›8§aƒ=’>æ7Ù†å%ߌ8ÊêÔœ½Õ‡|×Ã6¼ÎW&g=q]<¿âïê¯(8íàÖ_).ñ’ךâÓ¹)<òF[’º§÷5évx™ìbü]ð)Wº€øàfp‹Ò`WíIâMâ2÷€‹ÉFQx yª&0ÉÝæ±‡|=Ÿ¼W¸…oM~ý{+Ž÷Õ›.D»'Ýïv]vÆ BÜä³ÁOWcd_ñcÖ“]ä_¹¿Á„{­NžÊ›þÛÊC¾êÂ1¶§é×Glë™Ôñùæ€#ñZÎ?_—6&†û¯Ã|Â? 'U¾Ú• 6²Ãwßš:tÅkÞÇý"Žô?a:wõƒ{|g}ꄯMÍæWmqrÚw÷x÷ç1ìÜ—‰kÀ=í§øÁÖ˶V~»Íw¨| Ù0¿±Š«’Dt¡ßpÿ]ÀÃ|3Ï|ʵß+oõ‘‡}Q#6uáï<¯¸ïºÓÛÿY߆ß$Þ‹vu¸~ÉïK è¸ü xÏ"Þ® 8|Ö!Þ“¿úsêÖ_ãó“Óß¼ø¼/Wf37ºüߟñ›Äå¾,à_v'ëhq¸Ž˜Ï“gåpët{Üi›ŸË~sŽÿ}³"1cÿùg¯Î`?VÜùådA=j‹ƒÝ´p€‡Y?ì[ÿ}ð#þõ4Ÿ>5úáOà¯àñÞ›nÿiþãsÁ}ïæºµn¾ÉýÛ’:Ô9?ž t_ÿ°oOÜI埕¦é“…{ùìÊߨúàßãýþqͱ‹k*ÓÄ/ú[ì»/g_/Å`·ü7¡nb§Ó€Â–‡€Õ—€¿(¾œŸ.Ÿâq!~>Xp*_“ýôñ_=çkÈß±[¾…õBœï ÒÛý[þ•ÔàúЬó¸¨o›ýœ~Ñh<×?}) x±W¤ýC¾"ÙηÿÌWœÈî!¯;”&pZGîi!Äç;fËk1óZŸö‘Íy]zÃl÷Gù—âšôn}{s<êHV”À»o~À7&Ëá’“þ öŽ}×øˆ7ÿc+sã¯õ¾êVõD²WÙƒàâÕéǾa›S÷ÿ!+âýòl•]‰Žëå»ÓÄ=‡'¤…'ùO«<š^`<òzWœŽ_sáÛ¡.bª—)Ÿ%/RýºÒåu¥).Çÿ+OyÉCà!娕ˆ#X?þ#pY'_“–/>êþë+ópå_"þYE\ô§ðÚ„û€‡ÄúûFq¤ÏqQ뻩¸ìðÂ_ƒƒkß(¯^²ˆýõË’õK[,\zå}ä=O…ýoøjão ëÚÖâÇ•W­Æ®6±^šÀM¿Ãóx‚x˜ú¦ÝüF\_Ã.>B\¾/äŶ$uŸ3}u°'ñü×QßÀÎvÛ…·ÕÏJ GŒsÖÀ΋B¼«ëº{tœd…©»>ÏÀY|Ãd÷çß¿! Lí‰?Øõà[ˆ[— ÌW‘O—€®Hàù=ôÅøÙ œ/ö[ëÓçà_…sV†xÔ—¿&Yñ¾ Ø;ßü˜«žI}È7§áû›my­pËGƒ2Ù¡©ýlÄ–&îï¥îî¥)œñïÑ®Éë÷©‹ø6>·‘x’ú™oʳ›Ý™Ô±Ë_ gö}!N‹ŸSnCºÝ>é Be›©+=M„¬Œý ^çÅÁ¯y!ñ…êeµØå]ªËmHÓõññøu Lóôv[’„ã{+Ï£"ø[/}[šàø6ê+ÅÄÝ›“pùŸW8uÕƒcœWòOQæ1¯²ÊÔ¼ÛŸÁÿ Øö$Zü¯…Öž¤í×GµÙØñ ´îOáÍwZ¸uŸ”Çü^€ê²µiØ÷&Õ%mKš_lß RŸz…Wo­Åo$^’Ý{:¬oßú¶ìdW¬­KÝà ¶ƒz²êQŠ›U¿|.jS~ܘ–ÇÿÁ¶Ÿ&¤¶8ØÜ~¥NUK½¿jÄýž¤[å}^Çï”$n¤s³55½A¶*äçŽìë¾i{È«áA4‡ë9Ì öÄŠ’]ôƒ¿³àÌÜGp6/ ‰£}Šõz¸3½k¹&’ Ü˨÷˜êö+¯O7P\g© Á+;ÁuTï…Ïâ«°_ !^·%ü>xK¼oaÇÚá/¨Þ(>Â2Öâüg°¿U@,î—MÊ+BÜæ¿ öÜB¾nlÛ³AÙÛ÷ýÙÄÓ©LÓ‡/Zú©Im˦ôû#ða~ >©xHõ‹ê¬³šôÙy¶†xdUˆâõ Ÿ¬N/ãUÊK »kEü<¯ûÌûžM½­™xbCK²@_oIxö£û½~uS\Ý®7Þ‡µàÚ™°oýÏ!°Æ4-¹ÛV¤pÒ-^åªûGÄü¶¹# ˜´¾¼Œø¯€ºøŠÀ+ŠyºòΞ4ì©Á3Ø^×þ"\­ƒ¼Pv¡‰û-¿³úê4ѵx+Bì—²~N¼¹õ¹>Yͯ}Ò;‰ËT?mHÃÄïxGX_ñù¨^´>¥|ßþD>ߥøÿ¿#Ü—˜7ÀÛð ®Ø ¼:Q8qWöÍìÒ¶§Søç VÀû•Ô¡fQSÜRü©mÅήM¾vÑ#_0ù;}¾05ßoŠë§…:[÷KŸ[Éq΄øÝ B\¯?ÃmÀ ï-"Ok`½’ÿjŸV¾…ðã}Ö}oj`­6àVð?“Ý-ø` DX_r÷>ø€5SGÌÀS³Ÿ¯¥§ùÛO<ÚOoý7Ò|yO.íŒqô¡´lñukùpzáF¾e!ܾ;òàoØö4½¿ùž—Ãßûü${þÄ ¯L–ÙµUÖ”l÷¹‹rü@ü ö—x ¡¾b C vÄJ‰›W“‡?Ÿó¼;ÀÈC#O+õ^7ï°ûá]´Âïôäé¾½?Æß{7¤‰“ýLû&ì?/Kîò;ŽYoà9|$á^“ÂnŸˆñ$8¯ü•x.>'…¾å«°Ç5ä¥ÁžYAÈ›½:à¶*à3ösÕãáËa­)àÂ~þEyCIvŒsáC]!Þ´öOg¨;¸¾=IÜ®7êÄÏZì©WQ·”ý^ðgßl!y¾pÈ äŸ[À[‡”„õnkÂúsòƒIPWݼ)äŶ5Ô„cØBÖi&]îW[òî²dÛ^°ËÿÈõ<W¼ð#/QñuSB›{h¿ö±³Þãy®IÝÛ Ëàÿ‘÷ó»¶6MO¾gÛƒãÔ÷¾ÎûŠWŽO’ˆn…ŸYâñ¦m¸b98XëdSØøªD>ô ââH_ þrÀË…¿k]ÚÆp?]õ‘¥Yã‘ÝR—*"ß,xerâïŒ8ÖšôöžoàÓ¶¾ u/÷ÓKØïà Ѿ—$Ëï3ÆûÏÜw%náø¯ÿÙÓÂ7Bü?/|K|öŸ²‹‰ÄÇÔ}( ~7_|uá>Ô‡mÛ%É‚Ý(~£ËŽ=œÇ—qSØ÷N¼e›Bâ%ÔÉcsõâTÕéG’—³ÿ]oŸ­›EœO:þÞÏ8ÿúwÅ8)€iS¦Šûû4¸iEàoY{P›†“TW¶âÀ‡6ê`NÂ6–±?çbG„©>·”õ³Þøøn‘çJ]Åž„¼™zK-qëÊé)ñÚ—²tÝâ娞IÞgmðgTçþ@¾ïkAº¬?ù>ÂÅæÂ³W]ž£¯ ühÏV÷P?+'ÏXöÃ×¶'éÓ¨ÂN–rþZ߬ q†Ù­àXòËŸcgâ~>+¾ÛötùO0Åçâ©.ƯG.å¼V†üÝU¿«7jLàÊ]MFýÂu¿V§ðØ­±ÿœßꈷòñÕ&ì2ý%G¨£îZü¬ËïnëN ¿/I~æ+Ïzi¨ãƺè,ø­âÿµ…85W_ ^vðÿÕȰûøÞº4í¸ÝzÈU‡ÌÌIO·ÕÛ)6“º­7Ç<‚ü6òÞTÿW_“plÅk¨+×{ ñúÂp_ŃSœ¯'ò¢Rsù9WýŒçî‹þêê'( õÄI<üå:xŒ-Ôã×7 ¤Ä™È;oH¶¶ò€w¸ø›­Ø·ò€ÃyERyO¹—…ü-Ú•ç£ú‚§Õ$gsg‹µ†óëvsàÊλð½|~ú7¼\Iýƒéí¼ÍJCžå 5òõÅ£÷Rü'üÞô»ÚßÔ#Ä_óBìf9u{ð‡/೉Wd·Þ“«ÿ€þ©>IÿÖ±ìZ†8r{G8øzð§o€‡Ê~Ãëõõ©ýbŒÿXGq]‰/O?‹¿7ÿRÈõ¯‚W2/ì‡È+Ú’Ò¶?ì[‚?ŽçOÿo&¿,"/R=R8_u[êp½~Œ«Ÿ {Zü¨o ø¥·‡:ƒí·DvÏÓåvwÄE2iºù:+¢žµŒúÕö”fs‘¯…çO}Ê3àý¡.þ^\‡ô˹î/vÖ oϨ‹xKà YS°ó¾0É\c-à-Äꇪ€‡ÙøíVJ} >¼‰¿¤zvÑ•TQW­GžÒ!/$<Ö~”ÇÝŠ½Ϩ<êqââ‡_åâ9Tw«ÿ ^«ÑOf%Ô…V„}ïð<‹ÓÛüN[ OöYu«àS÷E_yuâw›ê+cσ mû^º!­ü\8%¸”m ~ÃÚá“«®Qž»ídß6RßÇ~Xæ ‰¡Z8¯ þÑ·¯ ‡·óyðæ»“,ókmÝW§…*_OýLqSSàG¨/ÙÊÓåñ?ÖÂúR=¡,yÜSŠ×ÿ$ùéÜÑýqVš„Ÿ.·äÓê(Ia×8üK[CdSš~3Ö [áŸ?siúú\¼MÓ=iC›?Ãþ]>·uò«x>Êsï%ÚÖ‹«o¬>¦ì'üýØÎõöÅt_‹¨›lj z®ú\ÔYÁ:¨ .\@][y£x‹Õìõ#Ò7æ…ð«YoÊÃý;Ü7xêwtâ8[Î× ©*OW‹þÃI~3à°[Àùšéó7ˆý)ÚÔ›­œ~Í"âÏ&âØJü}5×UâYW}¶ð΄ ö±X[žCŒw÷7ïχW¾2Üo/ñV.ÎNh¯è±¨³ü¿¢xsyTM¨óǾ-úŸãúŽ"Þ¨òMÅs%ð^?¨¿§·çkN[®¯þ¦ðWñÄUoU_ŽìḊÀÊxžŠ´>À]|0îƒmƒG¯þ·¹á>;q•Q³"ø¾yu×È«Z(ž;vmaàÚRx|ð£c¾¨u/¤ò6ùuê•ÞH_¸ô ÄŸXø|QètÕ×Ó·G]6Þ_ü­oŠG¾Ž~WøXñó ðˆUßFýo.Q°Þ ý9ñ{sNj…ðÙá¥;ûÍ‹àA<@Ýö7ð?ÙñúÏ­ÈÃOÑ]0pžx\ðýIò¼Uç¨Áÿ(NÜú,ÖÙÂó4üÀXת'îÛú¥ÐxÚÔ§ŽÞ„êòôË8|_õuymàeÛ.ðÓ ý/ëéS/lK›KG ¼ýx½7²¯ï øj¼núËâçÁÕc?@-:'Ô âçàø3aG¾V9¸êððcÀbÿ’úé×>CìÛR=…ü8òÁþH=Eù¼ø» iºúOñzÖÂÇjæzéÿôáùzsàý¶)ä³¾-…Ƹú3éû‰ù¨ì"qAìGÁ¯Ûòká(â¥ÊÞ>xÒ¾ƒ7Sß Ÿ‰×GýjÒ)lýÖÏ«¬›¸¦<ðÉŒ~kÇ?·½3u¨`Š—ZùM'uÙåàm=§SÄÚˆàƒÚBðòä#Gè›ïÿÝG~ û­ûU†Ýžù¡­u+_ê"VIŸð1ê´Ãä¥[‰ÛÅÿŠýîØ©²$ úÙbKÏêŽS‘¢>8åïmØÁ-èt‚–œ)Þgñ"ék³nâTå‘â]ü„¼½ \–¾Ùgí_ú¥¬ ðÝ­–}ÛN}]}£ëR3û帴náÚø ¥§èsñêGŸ ^#¾ýDV A:=ðaè׉×]ÿ| üíäC…ô[ÔPo¯$Ž¥^jÛȧOJül›|/õøNÂÁ"o_ýåmÄ#íØ9úyÄ3õØWáÎEÜ—=Á^Æë¨ö8ׯêN¯®†w[òZïÄ®Kg>Ÿ5+Î'¾‘®¼S[O©ƒü@}( áÑâ_UQ`.nÅT_“n|+w® q®Í'N¨¥»2ukSrÏ+ôÇ{_à¡™êóØßêoT=»Ü¼½?‚Ϩ¯¨:؇Èרœ>ÎOy9þSø6u^ŽŽ¾RÄ WçñX?±îF¿‡£³2é‚—…ûÖôɬ'Ô}­œ°œº|«ÿÐBž§zJ8'úOQ_ë`è#¶Jp«5¡ÏÍf±N…—Ãç½'ðãÔ‡û± Cÿ›!Ž:ÀýVüJ?†•œÇ†áï £¿"ÿ|yŽâ¦µäòÂÅß)ÀLhg*fý4‡~ aÒx#>1õŠ÷пW‰üÕ[zòøgŠï比„†~[.^Œîü5ÿ-<ºõ'ßÀÐ׈ü_ôXÄ|:ኊËÄ3Q}ˆ}ù¸½?ÎåIøñJt„«>"{›ÏgÙD>"˜‚—¥@‹K{^¶úÄÇ«J~þÃ%±?¦>yúe¥ôÑ—„õèÒË .ž¿ö_7Ÿ“¾”î-ý~káì'Þâ§Üú…¸9Äy¶”<‰~bë þÆÔ¸™üo+uxé?»pò­Wç+áQo€Ï›¡_¥|1o¥žº4ð òyv^½;ôYÇ|^ÔQSݸ;Ô­›>‘^ü®âÌè®Æ©Ÿ¾ à»®>ás›¹ßeôËÂÓÐ~sì Üœ(¼ÒæÃ?WžL]Uü¨I\êX‰\Ôï–ÚΠsa ÄôyÆþ´ÖcS|Ñýµàt­!ð®¯> p¶•äK¨Ï ÏìL5Ûx,ù¼z¹ ~ûaúÚ‚g ?½¯â> Ã7fÿn¥>ÏßÔ¯-û[FÛÈsï=‡x^Øs¯äxíè0Àó”þ›©e|Ÿ&xÆâAKçD8íFòˆUðË×P* ~VÌ÷èDZbøFè‚EÝ€Ía;õ©.~µ|¦[upñÒÀ÷c]G:côwGéa¬ èâ}*Î+û%^ôUà·G=Áðž*úEJàoOèÏ•¿T¿ŽôóÄß³^úÂÙ—^DBõŸMaý{õ õ£…ïIË»àkPséUñû»À/v£¥º¿úФ¢û¢¿ ØÆ}¦ÒJÉ¿¥VD]¾Vºw¡nb­—¥ú t:¢?hGe}ú6úó‰×#ÎÝ•>îŸ8¯£î•p;êÀÒ[´Íàžð"b<³µ&ù§Ötå×· ÅÅ­!ÏÂïxk¬7Ág™ôÒkC½¡ü®{IÜzm%¼ûê“¥äåô“µ€Ó ïWÓ ÿNþJñžêKÕðè{©3ö†üZ~ÉÙVJ¼|ºÝ°Ÿs}@äyy‚ ãÇdoÄo¿|™ê±ðé ÿÕÁW"²™:[ÿÞf-â\9êì}—|Bõ…nòõ#/R]ª™¾õ³im…o!> p3é\­Ih …Û£šøuªo®¤¯ƈÇWsòPé÷¨oȩ׊·-}½¨? :=}¼V Þ] /¯ ܨ†tDÔïYK†úˆjàã«MúvÒ+ê"O”_RŸ³øÂêó/ý+[GßQ1õk'~Q}¢ýHœhE§è‹©oÌ ºÁN‹«ß»<\öl 8LïË^ÒŸkÕàA•i:ññº¼–ûR >4öH¸ê/ÕŸM]³œ:Žô‹ôý*âXéh4*>d]ƒ§Dþ|k…רIü‚N²°^Å÷è ýqÖ†®vä~+uõG>¬ú ~É}o\5—Wûâ+À-ToÜžö¾G"ÿœÄѽ›ôWµ)b½Ä3½Ô'Ðs±RêÊ;Ñ‹k†w·!à‚ÖÍ}ÜAKñ`+y_ß/`Ý)^V½b¼¨7…mßb¡O*êÊ ƒ!žªÀDSúŠÂ½3ð†¨%þU-‚® ­JÊúluÀ‹r¼ÞgG½ròå³¾z$-$XSh¶ÒÀ{ñգ󰈇´ÂÇÜ:ê2gÈç¤Ç±‰úôýZ|0tb|&õ°þ}ËAˆ~tÌë„ÏÃ7vê,‘oŒNQ_‹qP]ЇŽû»ŒþÕ‹6é˜ÇjfЋ©$oG¿Ûf÷—à/J>ã™Ñýx¾=äí.ýtô¢ny3ýz½ø‰ è'‚WÆßë"½!¯ñÂÄyÊ*ð¯Q7‰þxë^ÁsÝD¼ÞÈ>àï^øeVŽ>‰ð³Jê[›é'(ÆN´ã7T·Dg0ê#„ëñê§õàÛÒËiÅo¢#y¤âã¶±Þáá{5Ç­‚Ó¾ >ûµàS+п’þÃΠk½ôñ¨½X8wà?Æ>-õ%•ÇK'¹œ:cõté] wu¶w×V‡ø&×oôg•/Lz)û½z@7ø^ûTxªú~;¨O´‡8'þn7}²ÛKÁ9๱î­|Tϵ†}s˜¾ÀüÈ|¯…äݪ«K?ó@ÀÁì°tìÀóTÇj ¼~Â¥BüáGêo܈ýÕy6Ç´€_ãWáUÆý \§ú¨ò4ô±¬|_ë |iéY[3‡ï&ž¤x> ô‹—ÄÁC¥¯ìÔ½½xåŠ#/^qø‚mVÿ+û@8g=õÍ%x˜º¢êg™€Fý˜ àÍÒwÎ\8GoÐa× ËÊñ«ètX;º{{ñËÅI˜µú« ü''}…¡ÎŸ«xŒèäö:ïð~¼¼‡¾L—~s?ñÒ,æ%¨Ÿ·žy;ÑM|8ô7xy¨yÍhÝãx\ú7"¿˜¾)ø*V‡î€x˵èÞ¯ ~ÁTïØ_²ˆ>3ñ¥/'þÀ†”~ò!+"Ÿ]ð$/WÙ•nÏ ¬'”VßS}¸»‚>m¬«eBŸ³—‚wƒïZEð§‘GÚ‚Žq;}DšcÒÔ.ý¿Çû.j/Àã>BêõƒÃìÏNð¤ìpë|Wès´vòäè·w¼)ç(}ÉûÐïj&>F×J}™6Bœ»—úÎÑÔ üUÔŽ-¼ê8üªSÔÇTO†× ®9BÞ;Bž\]DwÖè'òaúvŠð;‚=Ž×Ñ~#ÿ³‹¸\,÷¹ ßc³6ôZkˆ“múÃVìýlñû¨ß`Eýñb<)—/®3úa¬]·µÔiêÁ4Ä©;5ckèË. ®­÷VÿRA°›®ë¾.}à ù«ú쎯¡ë"•Ä»ÍÄ›tù—ƒÓ©:XÔ1Ç.ùòÊRt¥±Þ²ðù‡ÉÃvÒ?¦tÅ냽qOºH®(´íüî)øŸµào ºÁé{åŸB_un]Sw¯£.pˆzË0ü€íÄÔ§m'öÞ´U†ý¬¼ZzÃ&¾õUk¢þØ„î©ú‘;YÕ¡ÎvN]C:ÏÃó?ê"¶Ÿ:Âö§ ûÀ¼ëÆÏ¶cßZàÂÎì&Ÿ©"<î×ÎÜZôvTï×\¡ôGR7;IýR:câ7‚§.sŠu[À¾‚§ÑƒŸ8E]h„õQI²gÈgÅ×VŸ^ ñ˜pº¦Ð÷óO¸Ž*öôr4Fzõú8\¼TéKïB¶ð‘›Éëè`Ýø2âÀB?›ÁŸROëÃ6«Öþë"ŽiÁ¯)Þ·bž˜ Ø1ê êçpÝׅרjìÔpþ⇓ìg鲈¯"þàñ€cÛItëÄ»ÆÞôý´ø|FÀÏá Y)ñ8<ÈýN"~Z%ö=«¡®ÏºµÎðxrý:ô¯‰Õ¯ý]+Ãþ€ ¸ú…·a×á9IßË*ÈË/Ð'©úw½´VìMeàÿºù1ŸÎçOÓT€n¿ô ‹©wÐgû©K¯Þ¶³?;X›É«Úƒ‰õòºœ¤,¹·NaÔe.üîðܢߗÎS øùÚ<}Vê`uð"Ñ×´â%ú¾¥ƒæáú¢=‘NœøÄÌ»‰}VÔï¤ÿk«Ñë_¨}š6ú¾¥—¤y,[Òmp“æõ–„:D|½/à]9ž6vý¯á~÷¥´£÷Ú>êð{¨G*åw£.ÚêeõèyVRGTݾܾRø ×ßLœ"ü~'ý' ØñJÐëò*t:Ÿc½Á¾ bº¸è¨ÞF\ÓÂqZÉ¿÷sü]Ü_õ Å¿ˆ?ÆEéÅ‹?gÀç¸o'YÏ໑´øö8yö)tºá»Û0ëO:'óú¥«èO-#ÿGïÄr]»B=ßÚX/»¨#Pç·ìúêùâËÓ'j-ÔÇ·¦åÔVÿ‘mH ïnûá7€“ǹqõăà]škWF¼Çüº¨_X VsVzÐÛÒ<Ó2Ö5:ÜÞúy-ÎÙÏ™¸ß·S÷€Wå¥Aï˘[cÛ‰73œïfúfºáëyJ÷ùŽ5ƒ¿k]k>hy€úì—ÐÏ)½XÍ ßH}øÅè»*îéÃ? ?câãÕ©ŽÞò‚ø= Ô±ËB}6®âÌh¿úÐ3jF_b…pλ^ÅÈWv€—70Ç£þU'ù¹æÑ/íõð53àò“šo¹ »,]rÍKcž”tW}˳iâ›ãÁ¯“îÏÃëñÓðkí(ø1u¶ÕIøtK™õƒt¡/~ë¼ÎºÚËýÙö­ÜIx'zTÖî*=ú+½%äÒú®k—ëS£>N}ÈÚÁ¥¤ëC>ïÐ5‡_;éÂW~ûðôôtºY—M£û~rù4:¤ô÷ÁÓéà<œzѼôÔ#Á £žýGv\®ŸzÂPx®¼_8«æÈ Ño|"ðæb?Ý üÜàáGÁuGÈS—Ï ŸQ? ó#ÿL<ø:ÒK7Í&/Õ\.뇬$nÑ+õ5ªï­~?ñ.¨çDý‰è"/D_zl^ø/Æ\ͯózì¥ôÑ5ïp!|‹•àÒÔaï8ø}ÔÙ€ïáŸÛà‰+¯Yôt"ow€zˆú|«ŸÜ®:ÜoÍõ2Í_k /¡^=õ-é#yaðþ‰õßC?\/ó}4/Jýbê[P¿[7ùÐaÖOzðe\ú1e¡'ÚKærY'úHÒgûu^µLq(: MaŽ•À[T?p©ò>pÌê0Ψëúbüu+ÏCó³5/à ú6‹à{Õ‚Ÿ¡‡ê›áµµ2¿S󈱯g<‘§k$¿ÕŠ{7u(õ9£SiêÍÒ]ìß b÷4ïA:iíÄ7š£¶žˆòEé¯)OS¼ þ‹&òZéû6<ÅŠÙ‡è0ªŸßÑE˜tá×®{€úÝ û¬'øÅœnõôޝx¾4@¼#ôiG}ø£¡žc­ä­åßM@NgÜäýćȟ‡Cþ¢y¾šë§.ÏëñÜpþäm;‰_±ŽFèû‚¨ú°k.f%øC†:ˆâ˜vòÞBê¿ÂÚ‰k4—¹‡û¬ùpè3Æy¤ì÷}ÔO°¿.=éq÷’¿©Az³å!ÉéÙÇ\þÀüÜbžŸ®SºáÂÿ5GIó~´Ÿ—Q—U¿ì6^S'sÍ“ŽÐbòOõ©IçI«&¿•½Ÿu6òü±Ã«OÂ)h{sÞªà£h>ˆæ Á£³Jp´¹ôQuÂ÷=ÎRŽÞ“ôˤŸ%´BtDéϰâ 7lõÔo”_çõMÛú¦[‰TïOY¸“úE[Ð)“Ž‚ê›˜ÐHÝ>¢¯C/QzrâOH—\ýûðãÄK\I]§†ý%Þè)øªï֠ù;Ø¿x]ÚŸÒg øÖö‡ J¿ž ukÛ ž®¹»òÏ7S_ÏñÍ‚Ïñ¤ÀW™·b»ˆw¥ãßüDŽgÞøD\8¬‹øyñ–˜[ëQÃ蜀¯¶Œ|óùýNüâqâ©Ö‘ú†K‰ ˜ÛkÂiÅ?h‚÷ÓÎ~ßF<>ƒµ\À™‹aÝè1U® ¹âé*öy?8ŸêJ+À;5—¶¼®õ-=é’—Á§!°6ÍáÄ¿h®ëJÖ;¼?×¼ ÍB?*ê0jÞœúõÐÑ}ÅÅà…+©Û ž^n!¾V+80}EÒ#‰¼pô¨œ:°k¾Ò¦4l¹ÀÄ+“N®æ¥H¿QøãVô?¤kž¯G{€x«+ðå¨÷þ™«£Ð‹žIü»ðæòé#©ÿ ˜õRæ@Zx‚ôa ™S^B]ÙJ¸ŽmÔ+ÐR¬›¾gñ7¤iÜ.ýFñ‚ÐK_)ÖgÀƒÑ!ˆ×£ú¾ò“êêõàè†ÆÏ«Ï¹uÐG_a/ë=òÈÃh…ÿ(#ñÁ5ŸD}«À©Å—ªE§©|žªíçV_¶tÉÔÿ½8à“. ‘ÿ^Ö“pÊê Ûàl…ŸÖˆn¡þÛËœ…ì²ôqóúôÐ;u鳊·Ôƒ=<ˆ>Ø‘Àr鼨o·œÔêÞBÿºÑFÿºÕ¡CÍÜW¬P:§Ì Ù‹VŸ…òzõÙw‚Ÿ«ŽH¾$= õA:ºj±¢µ(Ä/q}5SÖüMaަ•áŸÑ»°Ãì£ ðàÔ¿Ððµ\™ôð9Oø´QG®>m+úÑ…ÌmRÿ}ü&âOËÏeð[ªûo¤¦Œ9#»á½eXßèoÙüL¸vsBÕ/*}üàöýà|ûÁÉ÷ƒ ‘¯[GÀ ãõòÚ2œÏúØšÓtsÎÁÝÑñÚì`Ô%<Å}Þâæh/E—Þ¡Ág<Š.ÊHžý9 ÏgøÑ‰¼º_zéåÁÏÅ㣬\x2븃ºC q´êl§à«xý‚­£×…kîI9üÞCð>7ƒ3¡ãêìêA캤Qgr-ýªÌsŠx4<ûØ] O ý4ÕC·‡mMsêNë‚_T¿§KW²{QIžY„ýÒ<Ÿãàžô³ØNöQ³ú‰à}E¬ Öçzô]ŠÀ¥•WÕ1ǼÀÕyœ¼Gsb5¯¬e´žíƒGz„¸>/]óúÐïó¯O³_öÁ7ê¦^»1ð“4÷Õ¤÷©yÀ5îÒÓoqWÔËT¼VLýº\®ݦrðÏ>øO«èØòvÍð¹Ôÿóôñ·’ïÐáÂó¤s/ÞíIxE…ô> ν‡~œApå—âß·‚Зrü¼ð é [7y“ôêЪGG¨ƒü€>ü˜Gî÷j¤n†°]àÕäMÌk…—®¹‰“Îg~A?ûrý¬¼ýÅsìR'øy#×ÓoGXG9žuµ½ðÒØ7Û~‘~Ðjè£(ul†_¶3MW>m]äßâÍÑ,}·\ÜE\ÐO\!ü˜9Ï>'ü;~~§xéÒâ¼|IÐaò•Ì­ßCÝ|\p Ïg5ü¸âuá½Ägà9½Až‹æá2?ÌóñÈ>ž7};VD?uS¨OÆÏÕ`gÑ»ñïJ7?¨xFuèòKédgÀ …Ã5ÀëYÇ÷é‹óûˆ;5/±œ8„û#¾tÔ],~ëÞð0¼¢x:š/0ÿMzµêgV>§>õCèÁÇy˜:‘ê²êÿëa?H':¯•¹qªçäò|pänú«›ù^!ø›~k ñzœ¯!|÷yPÈ+sz'äÁÒ+ê ÿ@g·½špÔ~êÿ;° âUŽPéÅÎg˜ó\Àþ­Æžî‡Owˆ|é$8êløÎÊ·KáÙ#!Ž•ž÷)òÿ]¯ç¸q#<ƒ ò«Zæ¢cõ¿í€Ñ‰Ž‹p¨-ø×SìßæÕí¯h'Êï¯m âù«?Xúëñ›ð‚©û½ø½üëFøwØ¿ÈKª¥®\CŸ¶òcòFéAǾÁ;Tÿ}åš«íè$y1óºá™€' GÔ¼Õq„st`oÚÑ“¡ÏÅ*XÏóÂ<1§~èÒ‹Ùø^|ìo×Äü–¹‹¾ý¦!ø[«ÁÍÅ·, :29>-ý’ûGÏë´uA?FõMûsꎔ‹ï°“ôaåê™ô¥©ßWkG£ŸúO3û1Ã:ë7‚c èŽ#ßhbž æµ6`ߥ·±ût˜<ä$}“ô¡û#Ø3õ¡cÈ<ÕÑNÒŸ¶ ÞÜ øw#àk™Ñú9±®¦|¶¾†êc舠§ÏÅgÜŸ}¬—–Pï‹¿O»µ¡ß¿¾–êÞÅá9Çd&ñ¹xÍê×?£þê…5ôIî‡gØB.½6æ¾EÜù(ó¤¨gtÓ§(}PÍuí ù—«ŸSsP¤ç£ú=s·¬;ÑJ|}OÞܧ qj9óB…»¢w6~1HÝl?ø[öMú†»¨‡Ã'ØMŸ[<øp«.øêï"¾ÛÏsoç¿§ ‰<×ÃÔQO៥C¾Mº2ôsÓÇ8‚. ~?Ö£»‰ÃŽ3ì ø„tkà•bGÕ÷ÒH~Ñ ÿ¸»±xCscNb_Ë>œ0¬zL¼Ö-Ô3¤ƒF?Ž v8Îcþn..à}æz¸æ‘W’o/ zsñóê o¢I¼þxß4Wþ¼«ÿ]zçeàܕଛÈÃZÈÇÅ*ñZ®ÞpÇÜz£¯ÜZóD՟隣WÏzdÿúÆ'ˆïáïe˜û Ã!úºÑeðµè¢›-]«½¬Ë£ôÕ”ÐOÁü!_Á|%pWÿ³øOÍôÝö1¿£Hó”Ññ)fŽŽê„Uè!‰Vµ>󬘹ÈêŸT?¼ôW´TfþlÔaUý¯1à^†~B-}dM!>9`é9üé}uÓ2pÒ¨÷Î<(«ïíg=d½ ?h¾ÒòÙnêÂíÔœ ÷Ù™cmëÉÏ+¸Înð´Npä5àäýðWÈÏڨ˃£øö€ßO:½‘Aìþú/{àUÉ_í¦Ïä(qÀ.âqæ]X?øbuÕòŽðlúÙgµ!óÌÚè¿9Ξ£ÄŸ#Ì£v¡üyœf„ºzYv ]®nê­G©û!ncŽžÕ…ûèÒ½há>·ó¹ð_éßÌå¯ÜñâûÂ÷ãûÌ£Šs©cïðwñásÄï/“~C }8màEÔãç¥çÙK_!qDÔ'+…¤9Ëè•çøñÌw`Þ©sÝ^„Žtª¥×VJ]Z8æ¼éA×}+k—/Þ†xâ—Qÿ硹3Ìß0ôœ¼#àoÞ€žÔ }qêÇQ^F"þÞ^ú›‡Îÿ¾¼Mü3ôº¼)àu9œ Þà® ç¡y½1¾ÓœCéRÕP@ÝÛCHÄÕWU„n:¼×t#½ÿ¿„¾ˆrê,Ü_ï¤Ï¶šyêWbΠæ(õïÄeØ™Içã(‡‡‡ç_kNz5yyýæ½a_ÚAæ†/¿Œþ€ÔŸzÏáõJM}a¾ÿÏyë¹XõªêöëÑ£Ýï§ܪ:©xášü,<{õ à_vaÅÇÛAÿ÷x†;¸^øÌ¶¼°#Ä;9<>Å>â[ú¬\óZÁÿ‡É ‡é78â@—n/¼G{ <ÿ(ñå)âjéïŸÂîdýg½Ÿ·!} ôk½\»¿ß>oËè¼Ls/mcÞÜcÍËì#.ëËç]6/ †­×rtO Ñ/ZN‰®ˆïÇzÆ÷ÕÇ‹^ó"¾îéóÑ…gÞ /D_…8ͨ‡Ö.g~V#}|Â#×·Ñÿ$^Ø êsè™ôâ{ÐM¬€Ç+}ÙbúiT§?ë§oj_ˆKrø õÍpzæ[«úhéÑœÔ*t)2GÉÕcà™ÕÑ6@~Þ•WßAÔéC×>ú5ã熘ÔD>ÓCú9øûÝàBòǰ=è {{+ý¡æÀ¯pñË À¿ž&¯¡.+ÿ©>ý:敜€ÿw†øGõ…góüñiòÁÓÔ¡:ˆûÁ«ãs8¡9¯ð”öoî£ÎÒŠ=.€6Ì}ÿh?}á{á—6GæÇùûÕßúð¤ãõ†¨ ® :¯ž¡o$#;L¼ÙŠŸD÷Ìú‰ëZ¹eèè*žm#Ÿ_²¾æm5¡ëPp ¯$ŽþWNTpÆ\œ %žP}Öð>,N®z¿ø»EàuÌ›Q®¹ó+CÝÉ[éWh§.q, á×Ä‹èQGœ÷(xr+qâæ8Ç9VÕØÓ~â¡^ôA#‡»á·™oèÅêûƒÏA½&êï27ڼd«oÒ¼¦“ÿùnÔ÷:ÐëC§0Îó"ïj…ÇÖþ"¾Œîû–À;qésØ!}?ì ß _­Þ\¸¹ú4çPyå ôYŽã?O°OáóÆù&Ø¡úËÛ±s !¾ö8çóiÄK§¦z,úQÏ£ƒúY:wðŠL¼ÅAæõbÝÖùƒãg4ÿ\ðYúnÕßÓÿý ¸S¼¦}y8ËÁ k` àè4Y vøYÎçq׳ðš4oPø‚ôÄNža_Ò¿mgÐ7ÙËþ;¾ãiìÔžoö y¼ðž5ÄëUà—°üyŸë¹þ°nâqŸ©ï¬\  ÞLÞ€ôôÅç©W®ãz÷–Á/Üú¢>ÁqæzhžL_àÛ{?:Ô à-àŠ;àûƒô?£_*ÁÉÒK„ŸÐððx}š#¢þÀ޹æeôÒ÷~6‡ïmb>ýRúß…³;ºXûð{äqÝè7•çù¥EèUV箄7¶‘8@}èùŠPoúâà°þ+úô¤C^ >Åœ0ל/ê–~Lzzè1—ÐÚà[mƒç*=ížO98ûIöE3ù°æ¯®J¿Ÿy7¹üý3ô·ÇÏŸÏ„9êó³}è‰I÷¦Ü¢.ÔÛ¼ ìÃ/<£Iùe-¸øAêà.9\{õSêÈû¨ë’]#Þ’ŒêíÝðpÿ‚ßï†7Ö@Þ{€¼§‡~ ½Ô³ÿ¾UýuÏÀók ûÈN’gŸ@_ä4}õê{CçÓV†ëÊå³Ä ΀sö’猄x2‡g·|^ü{?Ésï ö0OuñWž/n%ð-¡´|‰9Ø>;¯î{Œz¾øeÚï…ø·!ò¹õø_ñava_ÔgØD^'|¢_ºõØ Õ)Jèã픾7uHô¨lóu[BýÒ›ÁkÑvé²i yJîþ³—vSoÉÀïh¤?@/»×@}¸\s3N³ß›ÁûÊÑ€§©~~ñœ¥û+½ÄÓ màp ƒç ±5/Yýû¨£– “Ù  ÿ^õÍ3°Fp8énõaÇÀÇ¥sÝÊ:êû<‡û`/û‰k#¯ô°ëàíw“ga?jþGñØ ‹lÝð ö„øÿœ:ËRøt»¨¯-§Ÿg<àu±ê—kÐ+žÝÖ€¬æøºàš‡|†<»¼k;wŠøvÇè~ÏñîÀ!ô}xQ~Óè>Xéìڡǿׅ¸<Çëà9ƒ;j\ÔŸ <«Qö~Êfâäú°Žâï÷çVƒÓTàw‹©;ˆ¬9Õè$Ãwpx7ŽN˜×ÀG׌·7œ$N?gø‡p8é–tß9Ä<ê~ÇŸ….…æ ¢«ãº^ôYè‡R?»t%_ÖN¼To§‘u-ÝBé}.‚¿T†¾«tø4ç ]öÞ¾1›9_ê«åz]ý‡eÔÁZÑÝÐ\bô½…ýTòœ.ï½8ÿ±¾PtÖ¬Z: àÏ2oövg¸°Bs¸X×{à“I×Oóq¤#']6ÕWk©kí?žO=õ…Cð›Ô7ÅzpæØÅ¸²‡8ù`žNv~¾ØÞóyp}ÔÁ±vcÏ÷kRˆ¿Ó‚ýd®í¤æ/äÏÃô…ž·"\@¸8öÿ¸‡øù³‚>Œ…9®ÞœÇ§ƒ‡¦º¹£ë¯Cs3þ‚N¤t¶€ÃôÒßÖÃ~”nnIˆl>œaìÀ0¯O€_TÒç"]˽ðxUG*¯GÖ7†>F“x†¾æ½ðò÷€_ÕäéNÖñ÷.øæ[˜O¦>‡~ìâüÉ~ðÛAô^Zñ—Ý!ÌñÉ‹Ä釽\\ý?gØçÃð؇èŸÇ¾ˆgYÃàT#øëû‰›¥Óó}(G¥£>ÀŸÉ·ÔÐËþ>‰NÐ pSôú­…¾ø•á¹[ü´Nâ²nx˜=ðK¤çV~}ˆ:[!vjür/ó·P§\Kž!üq{òmâ(ú:#4Äþ* q«/¦‰Í9q¯tuÅ7D/ƈC¬‘õJÿ{Ä—á5z9ö osd¼-ðwœ¾©¨·ªüò$ú“èÄý¥¹Yµðú–Ó?ñ ú™ B¼èÌ_¡¯Î Á5)³xÇ Þ¿'ðÛsñ5þ[}¦Êwà­ž¤¾[ðk‡ŸéÒg\ð|­ƒÜúfì÷”^¸TÔõ*g^|=“þýÁêÿuƒúXš©£ÀuÖ7SGCÌ5‡úvæeÓåðä}#ø^ý$ôyœ9œ^¿Ks¢žª¢¾F?¥tºâõ¯$~SE~ù xñu×£ììŽîCޝò§øº‚º¹ì´úЫÀú¹oûèãj O<^ÙA}¼+ünä«â¿£ïp>âáHaŸæž/¼Œ3ÌýíÁ®k®z?úÅ]Ø›½Ô³ÏP<Ìõƒ³ 7è¡_ñ~ö8uáÓôeˆ·´Ž}CÝÏŽ7ÐG#>Òø_qôIüá0þŽº°µÂ㿾ýÚêš¹üý3éÔˆ73ÿ¬úH?q…ÖýÀ³C,þÞ0xÿ\ú'kÉ4÷eœÿå{¨_ô¡ÇÂ|9ë@ßm-û|?~…9ö$yY }[-ôT3'–:ºÿ9ìKg®›÷†ûîÕè63ß×JyšC(Ôís}1ðb×g·ãß…”RÓ<_Å¥ôqºâ˜ZæØî£O||h?8 ü¸¸.j‰w¶¢ï±…ãàjêÛpv“ž÷ï°ëÍAOÌzé7ÛÍzÐÍk£^!]Ñ­ðÑ´FâPõg=ŸñRp©&ñ"Ð'ÛÎÔˆZ)uvüaœ? y‰ŠÃw`W¤ë®yƒêï®d+÷=ìIç„þî s¤s<@]y øÄ†´¬1Þ‚ñ.ú•Fd7À‡ˆG*ˆ÷é×ÍÙEü)}\V¬¾ið{õ£W±ÿûáÃîã~ªßøxV#}ý;©/ƒëj®«¿íOh%#³m?¸aý=ô'¨!Úé¶òüöÁ3èGÜN4ÂzÞO|ÂÜj—NH7ñÊqúœŽcçGÈÓ™[oô;E½ñzO££ ¾ýùºô’NP?>C]±$é°çì6øÉnúòûð¯¥Ô‡Õ¯)ÝIú•¬ýŒ½àô%Á^KÏ=ǃÏòþ}¹Øo<ìn扗©^Gü!Sõ‘h^ú|—3ØÕ£ð):Á ©gYkèÛ›t~ `Åß{ܯ˜ü“>âÈ>a»ßú¬í ü˜Ýä}#àa'Ä?#ßm'­`þù qi{ð[ñ<Á_n?YK]÷|tpl>þ0zQ]à~ãÇÄÿ:ž¡ÕºØåØí“Ô/©C¯•Svv7xæN¢.ÜÆ¼¯:Åñàk=¿û×n^Çü8§O`¸D?ó^¤wÞÄyTIW\ýrÔ ô|ö¢#´ƒW£ã#Tóïçû{¨ŸhNÈ~xL'ÙŸì+pÿø|Ñ=tôã|™}Ôå‡CÞqN~¯¾_é2USw”ŽšpMé4|-~_uC]çRÖç¸ýgÑü ›xµKòô£ëàÓTÑ—Øp}¯"Nmb}=¿ØÃç4‡„þAï ¿¨ ùBüýgñÏ3G`¿t†ü¾9ôáy)vï÷¡’¸º“ºO5yŒæÅï$^jCü ë–>wk库vƒ‡÷‘¿·°¾¹žnÖë<ç!òÎcøõnxÁ>Nº€óì'Oìf~ÅøÒ>p¡PçštÞ…ïsb¬)|.ê&ï þaÒùy}‰GVÿLxi<îãÔGx>Í迟ÝM>4Â<ùêyyó0ÈûýùÅ>üÅAöíAò½"xäÿ¶;¥>©aô‡¨;u²Ï4×}õîøÒUØ îZÆ\‰SÔ“ªÂy¸xo½<ߎ¼új=}´ÇBÝ,× < <¡ÜDüzé¥W‰‡ÏÜöP†ÿš×“oÐ7`à¼eÔyÊÉ›N«ßy/v]xQ;q¥úžÔßHßb.~B·˜y)VLüVòÅ?š}¡yEè÷¢‹÷$ëjzØIâp%õ kýIç&ƒNäÇåü÷±„ù”5ðUg ¿€Ÿ)}¸zæúÕ ›õÎÉjÈcÐwòG¨›®!nW}èOàuáï>@¾ZŒ¦ô+Á‹›±‡ô×`_z˜7ÕŒþÉóØé>t(Úá‹ üÈÑ…4ârËð»#áõ¤ó{C?ƒx§éh õfÏ^W®N®3¾.Eߤ-ØÇø÷6æÉ‡8€~Cös~æY¼‘&Ö¿ô¹÷`¿_Á':J¼ÙMœ×ð˜Içc7ú©·îb_&_ÚúÐuž“ÎqJŽ7G¼ü øÙžà'7}IöÍsÔMÊÏÑ¥O+^é1ìû)òzø§é“ì$Þ<Å:8E¼<¿ý8ôÓý88Ñ^ø‚™u ì[§)ÎÙŽ@UЯ¶!öéñVWx^®uØnsŠ<`ûJxÌ2xÒ'¹Ÿ5èË^îá¸]ÄC½è§•ó<™ÿÝH½ƒþ¾XoßÍyUæñq¥ßÈœL‡/ì-Ø)éhHïEû };u9u™=Ô»¥‡-=ÁfúP¤©ü«;§ç¼Šº9~Þë¨ 0?ÅNâ×2ìCñâzÐoXI¾¥¼ÿ8xc }6sÑ«BïÛV1gQuªøñCðÊö²®ÐŸõ|þ]s¸>_ NÒ„_ÔœÎñ'Ò«&ïo¤?Vucôü½ž:K^_šk¾“xmêgP½¾œp|z~Ç5Ç«iô^ëc=ŠO;Œ~ô4¯ºþÉoNºà‹áýSðµFàÕÂ[Œó¸÷ƒ;UÓ×ÛIße¥ô èê¢þ)=…ýäuÒ¯h¡0̺jƒ¼WóÄàé7î¡}ˆõ;D}f˜¾®ÝćýáþåøÄÑ;À·³n÷ƒ;êc“Î{˜¯4DßSú èZ:ƒ/à·˜?cἫ¸?êhÃÿGgýxxëð4º4;à‹ ‡¡ï¢ ^I[ȯ".{Œ~)tíì vý`ˆc}M˜ as©o$o®uÐ\ÞM߉î¸_Ž_‚N_qs;øa‘ú¼ð+õô­+ŸìÅþÂ/µ]Ü'ô캉âÄyæ¡îÓ¿€'ÝFý¦žù4ÌW^ãê7óá/ÖãKžãÃÐÏ'þVüL=NømÄ¿ä÷çœÍ[ØWuôIÇ#Cß#º ^Ëþ÷Y^ïúùû½ÔAöRo?ìZ´ šK >TtçÎÉ'°î{á{–°_4ou=}­ÝØÅJâðvìíNt^ÖÂã¯DsÒw€#ïßÚÂ:Õ:WŸ„柩à&p¿rêÇx1»èCOœ9➉8î/àß{Áñ•'gŸ‰WX¾Ûúgì@°›“.„çu é$zÈð´]º ƒàOÌ9ÈÕ7CþæšFý<âôá×$Ojc}bŽT+qF/û´»Y/»õ=D_àýnÃðÂw±Ñ ™tödè#Í=ò¦½è­G9nwš<½ <²<¹'䟓Îõ˜ø{σû® ¼3rtDí(ýn§ƒnaÎ^ì§:A=÷YöK üøvñÙØ§»©¿ïâþìÇ~ì'~~¿^|;ì›æ Q¢žÓE¼üxo/ó!OPÏb™uP·®?LU¾BõžpÉ6êÙMyvåó2ø7øOæðÜv“Ò/eØûuöÂÀ#wæ4ºúE¿¥¼*úˆãqã…ÑÇùŠ¿Oÿ±øÅ1aN‚wRO>prô#½!Ôß¼|c„¾BÍoe>«õRÇ¥?ÓŽ ‡~>RsÀ×¢î7ëÊ™KkûÁ1ÚˆÇæé l ýƒš£ãMà™[‰S6¡[Nàšû^žC·CÏd\²œõrPóàÝ \/ââ}æž;ö2æôUXyðk®~¡*êˆÏÃï¡E¼äßy+üöâ twíYôþ¾\ë€%ý³úÆvÑ_°óU¢Þ·úAö“wiÎÝüÄaüoGž_j“>58Þ<ò‡ö€OÅÏž†6ÀñÎ\'ÛG]§ ?Þ Þ\ïCs{ák[Yà¯ZxqøÓjxÜÌ5öò&ÍsjFgC¼ÿÇÑé…Ÿ‹ÏЙ–ND/ÇU½¶ ÝÍ5€Oêðzcý–ú¡+î©|0o ®§¼âÄ7Ò4ËŃØ_Œ9|Ò1±Ó¡=Æg›©·®GG>…k^)qˆuË?1@}¾ÂCqâ÷«ß›9Á¾VsnàMˆ—"æ=y1߯D'a6ù½ô°CýkÍAqpQ×\eé6ˆ/ÝL|ŒÓ‰_žÇÎ÷À›„lÐ}éÄ—bÈíYê·/Ð|Šû}†þpÍ?Šý€:¼æ%´‚Ç•‡ýïâ'·/6RÙÏþÞ¿V¼¤!>×E_8Uäsˆ‡¥8åûˆ:Óá Së§ ñ÷¤óá÷Hç^s£º‡~º#à6/Po8N\Ó_@ó<öQ÷|>ðÚsõúPJˆÅg’þã±ÑóÄrx18ðμúìáàÂuà{”ò¥c»Ý?ñœ?æ+À©¤Çt€9 àȹ8‰çG=ÔÃîöÀ>}š‡=Ý…^J Ç=Nmúàá¶—:NKÀ§sø*ý"Ãôª/ox\=õ“¥þùÒ« º˜1+Ë;Oæêúú°£Žm ÷Yüínæþ°ß—I?îyê-½àËúþQúx7€ÏjŽx·õýÈI'œÀNž¤OJu'Õ°ß*±›­ì§jp ôš¬•|° ¿q¼à0v-œî›â}è‡ê#ïÈ _ß·ݦ!úÕ†8Îæ7í îu¾IçSO„7ÜÃ>:¬>ú †î>鼯€³§þÙFÿÍãäÁƒàš/€×ˆ¯÷pÿ2®WõŸFð•ãô½ÀC`ÿÛiø0»BÝ.gÁZ°[íô»oç¾õQÿÛKܲ»|(ä³¾~s+ì} àÆG°CG˜_ÞM_˜ž[/ç}Rü}ž;s¤/h‡Èkkˆ_47R:†;àçHO~½ ÑO‚Eœ—šA7®;,¾œê©áûïâ5?0¯Ó6P'“®ŸîÃqü³øg;ˆ[5OXû»»€Ž x ^I?'öÐ7£Ã×ögÑ›ˆoa7VÓ'ºœ²5Ä]¹º$÷aÞZ zVàO9 t{Ôg¸Ÿ¾;ôrù!sßÚÉ“ðÇqn|¼·ð wüÙJ”9™Þx ®¹öÒ‰j¡î×®Zø²ñøK˜‡ZêŽþ[|ÿ 8#}fÞÎs¨…ÏÛ̺yºU} ¥ð”9Í.å~4`ÔÏø¼åÌ-:ï¼—\=‘óÏ;ï5/Ëþû›¦Þ’ý×+³ÿKþüxöãó>ò7ÙÜ:ù–)·gÿûõ|î5|6yÿ­×Ý5múíS.¾eÊÌ›§ßy×ä[¦Þ:åâKÆ_{õ5ï᯾ö®üùÒ³þü¶™3&ßzû˜k¦Þ:ù¶©·N½ñâÉ7N½éâKÆõ‘K®½yÆ”ézoÌØ÷3mò3§'?•~î}‡áóúü賸èš;n¼ñ9ÎE×Ü<õÛSoyÑï½ášÉÓ¦Þx×´ýâ[¿8õ†ÿûâßüù»fLÿæ´;fN¾åEᵟ¿cæŒ)·ßxÎßøù›§ß~ÛÍÓgį}ó>û¹·Nysî¦^yñU3¾™½¼ì¾ÿáÏÿ0þõ¬ÏÑÏ:ÙÏÝ8sòMSnœ|ëôxÔ±g½ÿö[îš>æ“·N¿}êÌéÓÆÜ–žàä™ÙŸžxÖ‡^ö‡.¾dÂÙß¿zÊ´©7L=÷Ž]~öº¸zò´™ÿ÷G^{õäi/z[^ý™1_ʾׇ>ÿ™1Ù5˜=£ü5ø®ìaÉį\yñ¨¿ýéì;×L™1ú£úÃÙŸ{ÍgÆ|~ÆôsOãÙ¿ß<åÖ»¦Mž6ùÖsÞ}ýgÆ\½â©ÓÏ}+ûƒŸ™rÇ/ö•OÞ>}ÚÿòÖ'¦ß’½çž^ö¢?äY«gÔ߯>é^¾—޾³¹ëö™SÎ9HökWÝ~Ûä3_ôk¯çÝÉß<çxÙ_Œ+ô’ËÎzãÙ7Fݷ䤯z±;ùªO»ëÆsÎèÕÉ_§ÌHÖäÙ¾(üùEOó³W~ã3reô’<ûìÞ|4{»ÇL8k_ŒþøÙ;ám|üÿøÅ7~|r0vñÔÎþ׆w§Þš\ÏgßÚÝqËä/nN?6ãŽgþÿŸäßéóÿ×YþÇŒ»fÞ<ã1º¯ ïKpö—~tÊÍwÝ4cúäÛoœ>ã}7ëj¦ÌÈ>ÖKÞûžìú8ûÄV鈙š5ú·ü/à-7Ný¿}Ê›“\ucö„ÿ—¥yÔ™“sëjÔ[žrëÝÿ‹e¼jÆä²ãÿôƹ*û×é7Ü1ó®¼'tåÅ/úÞÙG÷¸1Ÿž23ø’/¾ ϾΗ^ÿÙ«¾6î«g¿ºtÔ«ñ£^]¡W~oÜè——Ž~9~ôËËF¿œ0úåÄÑ//ýrôqÇþðøÑ?ú×}`ô˱£_޾„ËF_Âe£/á²Ñ—pÙèK¸lôY]6ú¬.}VFŸÕ„Ñg5aôO]1úÝ+FŸó£ÏùŠÑç|Åès¾bô9_‘w Ñç|E<ç¿I^ŽýÀò^Í{=.ïõ¥y¯Çç½¾,ïõ„¼×ó^_ž÷:ïüÆæßØ¼ßŸÿ:ïxòŽ7!ïxòŽ71ïxóîÇļû1qôýwÙèß7áy¯Çæ½—÷:ï÷&ŒÏ{}YÞë y¯'æ½¾<ïuÞùå]︼ë—w½ãò¯wbÞùMÌ;¿‰yç71ïü&æßļó»<ïü.Ï;¿ËGÿÒÜz{ÓõWŒåÅ×Θróä4O ¤Þ~óõWŒÍ¾}Ž3Õûo»þò c³ï'Þú¶l“Zøàõ‘w†|tê7ƒ¿¾ý®iߌ‘ƨ{ãõ¯¸üÊ‹£Wý;o¿~â¥ãÓ˜æEâ&}è-×O—=X6Žžr÷Ôð#cÇžý·^?ñÙc$yIÎÿ]6ú’'ŒÿÀ•Ÿ“wäÎrBzK^<|CöÝqÙoŠoõÕk¯¿lBö«c?‘ÿoÊÆ¥cƽïÒ17e#Ú)ï»xì˜KfN™1sLâà¦%Ž~æÍwM»}ê´»¦½gÌÄ1·ÍÈÞÞicÞwñ¥c.ùܘðæô;³Ÿ:ë6_6îŠ+/æ‡/3ùÆ)3GûÐä“If:žo\xÝnæ?ÿ*ûŸÿ:öìãÎ~qéÙ/ÆŸýâ²³_Lˆ?<6÷ÃcÏþá±gÿðسxìÙ?<öì›ûáôÅÄx”ñ¹£Œ?û(ãÏ>Êø³2þ죌Ïå%×§{iòßñ׫q£^]:êÕøQ¯.õj¨WãÑÆÏmÜø³6nüÙG7þì£öÑÆ?ûhÉ« ñ÷Ç_žûýìŸõûÉ«q£^]:êÕøQ¯.õj¨W¹«¹bBîhÙÿ>ëhÉ«q£^]:êÕøQ¯.õj¨WÉÑžOþ÷ÿ>[ýo pcaMethods/data/metaboliteDataComplete.RData0000644000175400017540000006350213556116437022146 0ustar00biocbuildbiocbuild‹Œx]W•ý‡a&ô: †0$NB`f¶ 5¡…aÊ€“bHbpHB !ô@z#ÅvÜ«lÙ–ûVïå==uÙ²zu‘lÉ’í”ÿ»÷üÖÅïÏ÷Í€¬§÷î»÷œ}ö^{­µ?ÿ‘/]ø¢/½èŒ3ΘuƬ³ÎÌþ_ö¾`VöÿyƯ8;û߯½aÞ¢¹W/¸~þ¢y™»hî‡ÜðÝëç-š—ýƒ—qÆ[?n_ºoÑîoÿä¤o_˜snñŸVxÉ=×eÿW‰­œ÷ÌoÛvLض÷ÿò?9>fK>ýÁš‹W}ɽ¤{äÍ>ß?Õü¢ÿøî2¿'ÿÂ׊þðŸ|宽䶗ßãׯû‡ Þ1t«ýüïvŸùÔïµ¶k §.ݲÔþ×GÞ·÷¥‡üá§?ûñ׎þ½=õ‡cW~ûµÿè}ø=Ëo»åzÿ]ÛçÞPû÷_·'öþá·ßë°¼]ï=øÅÆ”g¯òDû?ûw¯¾²~lõÓ¾é%›êÚdú»-‹®ŠÞÉôûøSλËî|}ô/óèÕ9¿Yl _~söÊÞé¥ÿø™ï½eƒß:ïÝ™¢¹¯ö²Å_ÀZsS?úÅÖýž~ÇÐ- ÏŒûSÑÕ}æûþÔœO/[Uðßþ«»>sàƒ×~Ù=|Í[{öYÖôż>{ ýîãîGMõ5c—ŸûU_ÿÞ7>¹÷ûÖ½Ë×[í'Û?7rç?n·mñŸ}Û¿óκù_>ê[n_pÆuí¿·¾YwùÑW5{Ë~èˇ¯9×6œ÷ößrÖÇü·Ü÷Úè²×¯óÅUïú¿'Þt¶o‹_~“müï?ýÇß—ÊžÚö_K*óïðç¼*ºóþØy…üê»üÎ+þú ßïwÞþ·ïï›ìñ‡Ž¿ûÛþð­ÿÑúÓ«ólýÿ²¡ÿB»;¼¯ÝûÖè,´5ñãø®å}þ ßXsm‘7ü§U÷~à_íæð~¶öá²îŸÏÙn57|õäºÖÁäy,ÏÏW4Ö­©yâ¶ù¢Ç^¶÷À2ïäzîØ]À ¯÷ø m[ö[­}ò‹¶õ_>ú¢W¼ÏrZ>’zõêµewŽ.˜ÿ/{|óù/ûâ ¬,ûnwô¾ÃïÝóï—¼rößùöÅwŸßùÒ7YÙÝîz¸¬'ù~Eá¹Ùº7Å ÊR/¼kûβWÚŠÏÇhwgŸò¾Róú3Òy¯ûR·,~gô í7¬›ªì*¾ûÁÝžùíÙ³:Ψ·m~üÅëÇßj…OýÝUïx]¡íïÌîv_}l÷Ï}KXŸ–sóñgþé¹óõý6¼çߟ´½}ÓWþǼäŸûùêû^bùÑ]ûè"¬3~¡/ëÙùò¿öŬÛÎ˾ôŠ«¯ÌXuøÞ¶'Ú…ÿ»Åʲ»-»ålîì ò÷¯ñ½=Zñ¾"ºÌ]xþ¯ºñ¡»>k9ÿ}áܽ©ÍÛã__i™°.|Õ‘ÙMKæöYÛâo©ðùÖÅúܺ¡ÿ‚©Åw{á?_–?rè¤uÆÛò3¦õÿƒ‚=Ë_ùX§¯o¤m ûÇZsé§JWŽzûoK¼ìn4xe|Ážßþ»uß{I®?/ƒ9¶ãå'~ì¿.ó¥QØYÞc[³‹æm×½Ès¸Ãáù{öwWMÖ¶ZaóM¯Ï† Û“ BÙ%ë˲«±pêýV”ÿÆhgXÝü|ëýoù‰=’]­ êϵ§^Ûý“xٽġ_‡}j+âmò*{ä“ã£Ù²ÕÑ·þð³ööÁo£]ù¢¾úµ¿z÷¬—|Ò®½­ëcÎù¡ž“åF·û3l1ëôÊÇ Ù7…ûa[ˆý>ìÛÌ÷,zsü¶<¬SÅ3ÛÔÜ>þ«§whÿÚ]ñrƒyôØz'mk´-Þ{‡í9½ ÀÊ®ÿ݇¸ý…V‘Ý­ÙâWs¸N[Äshâþ¯nÿæíOÄ×[¹O! |Ë–Üü‡zðó{¬0»Šß–ê³ïG_û;ý÷¬Ó¦h™=\n¿ëÕJâíþŸ¾¸ï—ÿæÒ+lÍÿ]lÙjѪâ`¾ÿ†÷Ï ñÝ~Ö½­"þŸõSõòwf#üÿøòlT8o¢ÈõzÅ£G~zu´õ|õ#ѵM?Š7´­ ûÁÆý+^Uð…ì–ñ%Næ<o8»iõ¿¾/»%,½Ý'¼ä£üV›Ã:ˉ¢Ç÷¾ï…sËÖeoÚ¿m*¶ûÙ—Uáïý‰wm9ñiû9<æ—ÙæÊÿYùËÂwêü1ñ×tÝË6~楿¸òLÛÎQÏËÞ·í³µñƒ´Ç¯ùõ¼û7~Æ–‡øo™p_ìÁèéÏ{Ö–\~îWòßxi²ï6DÇÃ[k«Wl}qö¥¶iõ©ì;}Äv†ukë³Aèãƒ+Í£·ùb“å…ý›¬÷%Åf¯øW^Ë>¨‰VË¿m¶œ¼ŸÍ¾¡¤É–EÑ7YEÙ­Ñ'؆7Å•éœÊyu|c¯ønkNýäîçz»ž‹o꘿þCE•ÖyAt >k©ïEw·-b.c_U‡Ïó?ð9­ìÇÍ_ˆ7º/>úÃÍÏ–Ü`}%/È^Á|¿‹øZÅ~* ?9ðš­7 ç^gKB~àíÙÝxäñÍ^xÝÙ ¼þÝɾ_Ïù÷‹Þt{sû„åœÿÇKÖ>™˜ÿ }ì[¢Slþz+ù‘qî{†sö>α¥!¿ðÍœKœ·VÊ9Ÿ qÈö\¸ÿhÅœK›Sñ²æ:ÿü±Ä âÇö–‡‹/ØÝÙ ö»vØŠp>ú²ìÃɦdöHÈSlÍ»âdK¸Ï+ÙÏ+³«ñ?{Ü~ò [š½y_=¹Þ~ò2_½ì›uþëèøÎû™oåù+N>ÄþÚ”ýRٓѵ_Ö^'P¾3^Fç¸Þ7?JG[üžp>ÙO9ÿ¿­|‘}¶6ŠrÿüTòw»CÞh…Ñ1—›Ãœ¿ëâÄò{<ÄUË]°¾)ÞžùJžëuqB¬›§RñÁîDOûƇMë‡ûçiöÙ/ˆ“)îóãìkíƒ÷G dŸ?«Ÿ³ßfOËò-—%ûZëÿö‡»ãÖÖDÑñC_±ûC€ð«ˆCäͶ2û°³+ÙõA|X×ÿg»ðèw>›sÔî'ÿÑç­`ÿ+^­‰ŽÍ³]ñ~ׇãon!Íü¼o‰Ð:ŒûÒ•Ëí âëy>«/Pßq¸êÔk¶.²•Cñm{ããà¶,zlK³›£íõ­%F> }c?/ »;Ä×y¢uô`6yÊ.q[Î)ßF^XÎ'«x0NÌlë+ãÌÈ>Jžñ$ù y“†ëõØ_[¸oZ·[Ãz°Ì;^—½ÂW%çÎCäC§Ëo°Â÷ð!Ÿõû¹9ái½$ëè¦ÿÛÞ(ÈõUxnØ÷¦õκS> ü×þ›¼{ûz/qû—áûy]¼Ì.ó;Âú¶GC¾í5Ô-íåŸúZ¶”0åÑ%/Ž_Ú9¶…sóßy_Óyš¨ïX”…f—Œ—FYÞù^I=²6äþKÖÅýÄ}ÕŠ7Z_KC<¶Šìâxÿ‘‹í–pžÛ®wʼnHr®ßKž¼;Ä9W>ö+ÖÍ äK{ã´óß’<è6Öã“Äç^ÜÙQÿëZ_Íç±¾’û·”sw yñÞø¸Õ‹xœ/öK꽵ߋæU¶•zåÑ;Ý!/xuü€}9Ï« zìo\ìÚÿKãtK>¬:oo6]Xù%ß'>‡<ŸºŒ|ÁëÃùí“÷çDaýª©$޶ÆÇÑßE•/½Âtÿž ž=üŠèƬñU¡^öÄ•{²QêÅ”¿yC|úöë|÷mÇ£ o±'Býa¤>ü]ôÿ¿Vm÷sî-æœQ|¾›8_J]AýiŸûËãr¼n­/‰—ÑK½Š|xuTNnÛ`ëã¹Öµ>%nÆðBñcþEÖ)×ãÆáÿǦºiI¼\osýÝ×8Èw¼ˆ¼©<'Þ¨I>p7ߣ.ä ¶>Äód}ouˆ=ȹü0çŸÖx†éù.‹+_OF¦Ïw­ó!ﱜ{é†(ð_`üÄ×R§lÈžn¿x/åï¶GÕÌO¯ñ½<ÿõ7HΙ-Æ Ü'ãmó}ϋ˄·+³MQ˜ÿÜA[Á÷) ñÍ«9矠S|ÚqöwvdC½ý: öÔs븎uQÔ}[ÚZ£múÌïüçÄѦWyÞ‘x0ð#ϲô—þBù‹êÛd½pþû]¬;=ïu—D/ü”“§ùæ(<þѵ^øÞªC]ø×fîcñ'vf+¬«þö{¹Î‚Ä+ЋÃ9誷Äeã7]uÒÎ7Ç7:É؇öÓø¦ºà÷!žøæøüš§–ǺpW¾òCòË•¯‹Ø{üAêFpû0÷_qœ|Þ¶DÙÀw{}|œ¾Å—s/ŽÓ͹Þ@Þ¢|¯4ä+о=Ô®zðÔà ¾‰ç±êÒ-K/zì劾3‚sÒyöP|û^gyΉþ2ÁQ6Çáì¶&Ô»¶á—<'ºêK·&÷¿è‡ŸüçËò&qpwȯ”—x1y­òÓe1<òZ[Oÿé¹ñv8;‰‹¬wŸrå¡+É®·( qÕ'ß$/s¸˜ÎÁGÞBêLòc·åìÇ5ÔëeäÔ½Â-|wôî?xTy¼o~I¼íÎx;¼Óõ}gòÉ›–€Ÿn&.(¾rŽYWv‘õž´ ÷Ú=•×ý¯•…zÕ…cì˯Ûînøï 8²‘¯ù.ÎQã‚Û”ÔGû¯åü* ø„+Nªzµ|üÀñ®|Í{¸_ä‘ɾkŽ«Ýéû£ì蓾%›_³5ÑeßÖåiîÏ#Ĺ/“×€{Ú휃iÖËžV~‹-w¨¸/Ú0¿±òË£Bt•ßpÿ]ÀÃ|'Ï|ʵß+¢ÓêÃúêF¬?êÂßy^ɾëŒoÿ§}ç¦ò=å÷ËÁ58—üž7Ä€ŽëœïñÅäû”g€ÏÚCä{:¯þë¯ðÑå¿g¹+î+_.,ªÈV ¯uâÜ$/÷õà_·°ŽÖ„ï‘Ôóª³Ö”½"ÊÄ,/Þ·Øö†g²¹Ô̾Ù…±ÿþ“ƒ× g°Ÿ*ïür´ ¶5!nZ18Àƒ¬ö­ÿ>œ#~ý8_0?9G„? ‚¿ ÿõþãËÀ}oã{kÝ|‹û·+>PWºêáqYö!ß'G7…{ù’ê¼±mœï+ÈGï%¯ù1q1·".ü’ó–øîØ×ëˆÄ-ÿMè›Xþ‰8¡° !aõµà/Ê/WÄËçj­ ˆüEø`IÀ©<÷ö}Æs©ß‰[¾‹õBžï+ãÛý#ÛðÕ8àúÆìáqNÏû#8ýjò߇èÓ|–øRðb/ûøÆh;ßôó¤t'ë~Ùh\Ài¹×Ç_â˜mˆ¢Å¢/ø²°lé«âfû¸?ª¿”×Äwë;;]ç#}$+ŒàÝ×ßçÛ£åpÞq¿ŸxÇ>KÖøˆ7„óǶGQæš_ë÷êXåcÑBÜd÷ƒ‹WÅ/û¦íŒÿÿ°B~_–­¢²+ÑÁq½ì@\¸[úßãÅ*ãÆ“ÎO«8A¯aßP×'ý Î5¾Ýú"¦~™êYê"õ/¬=^^4ååœÿªS“ºäð2âÊ}ä¬ÿ ¸¬SÇçÆí‹x^À¿’¾E¸‹‡u’Ô_ Eþ³‰¼èÞð³ ÷¢þûFy¤¯yÑMÖw!ç¥ò²§À  ®}£ºNxÉjö×ωÔ/Mç4ûÎï¦îy"ìÛÁWÛwy&¬kÛÂ9®ºj3q5ÃzÉ€›~çñù0ýM»šzoÇ[¢ øŠ}¸øyy¨‹mm||.òÍ!ž$׿•þqƈÛ.¼­nq8’<'7-^òÍäùë{).êß âãú S_“zÃ÷”çß“o<{,äöðŽ]ä­P曨§‹Á7Fð|ï} çl×KüÖuøRÎWáœ!uÎ_âï 瘫ŸIÈwÆéûëmu­ð¦‡Cœ2Å¡íqýt‚¬‹Ž¿þA¸»«£¸¦8¬÷§/â{xÝvòIúg¾ƒ¸Y¬ü ì:¯„3{ÈÓüúêÃåÅÛí®i;é+ïY¼7.„¬”ý ^çEœkäê—o’ºK}¹¼¸\ŸmÕÄéÚ¦yr¯­Òñ‘ÃÞÄó(ç­—¼).p|ý•"òîQºü/>úÁIžWêQæó_fqxÿŽ?Åù.`{£lñVYKT¶%_8ª-!Žˆg ýpO o¾ÍšÁåèû$ùH1uÌŸà¨/[§}¯S_ÒvÅeð¹öMêõñéWxùÖÎáíäKŠ{O†õmœ§¾'û&Ùk[ããájÛG?Yý(åÍê_>µ©>®Ûãÿd{ÎŒ R[CÜQ^S@Ÿª†~çªÕ“÷;ýûZÞ§8:FÚvZÃ×âd›B}nàȾ5à›ÖK]U¢<\Ïa)ñ¤0ÚE?z³€3×sÁÙ¼0ŽöIÖëaúWÅ@²{ýSß>ç+ñJÖ™pªðÊ6põ{á³ø&âW:äë¶–÷ÞBžmÇZà/¨ß(>ÂzÖòü§ˆ÷“¿•@Ìœ¾ÙÕ!oó_QwþŒzÜØöf“²·ôÿÉÄÓ©ˆË‡ÏÛ.ø»vÄ78ɳ~ >©|Hý‹ªìa5çÓË-—|d}òПHðɪøk¼Lu©w­÷ƒGãµW¼÷Žlém äyÑ}µ•GéÙOîñ4üú&Éúë$O½\v ¸v*ì[ÿSȬ>.Kn³1œtƒW¹ú¾äI}ÛÐ'LZ_^Jþ—O_|cà%uºêÎ.ú÷…Ô«ûÀëÁ_„«µR*.d¸ß:w6_ºÖï`#ùÍE¬ƒ;È7÷°>·E«ù•{y™ú§é8Müž·Š/¢~Ѷ˜~ðC»—z¾]ù9çÿ>ò+Õ ð6<ï¬yÙÔÉ[è5ƒ·g™]Úöd ÿ¼Æòù}}¨Åôã”·TRÏí&În‰þìœ^£^0wz}A¾_—¬ŸFúl…Ü/½.‡Ï9òwË×ÌЯH±@ÓœGÂ{ ÉGÒ¬×ê_íÓªÀ7°~¼Ç:ռÀ þgŠ»ùˆë‰îÞî³úˆ)Î9xjv;ùZa|™ï²aòÑzxzÛ¾ÿÇ7Dð亶$_:·-¾aŠ¿¸QoYH·oKøð7lo\Þ_'>ùÙáaqžûЍzþøçDËì •–‰¶û²ÕVK_D8£ö—x ­¡¿b«B>cêÃïŽüßZcès%õò“á|´êUõ-Γ~×–¨Ì_R•ðÿ¢(ý«û:lŸçœ{{¨wÀem9xpòâ¤^^…óÙÍÊk\}0xL–‰—åm<|Îã ê¡TÜVûJr‰§&|/œ@|Ÿ²‹è€x†sbW„õ^/†W Ü&Ð5ÎKê”}Â:ø(Ä+!oÞLªüyuU<œVðõéÅÓŠ“ëöy 5Áïôèé¾e É¿ûòâÂÉ~®}öŸ—FóÖcÖxcI8„WǰÛÇ“|œWç•x.¾4†¾í›ˆÇÕÔ%!žY>¸mUÀlSÀgìõãáË-paÿ.ç‹ê†â8í¸Ð…µ‡|ÓZÙm¡ï`à>V@ÝDŸ8Á³¶…xê•ô-¿·üÙÄ7[E/2úsx«òâ°Þ-7¬?§>˜3‹>¸úæ™PÛîÐGŽa«X§©x¹_iµð0Á‹,Ã÷­ùŠ—±ßZ9gUÏ©ÿ%Ü@8m†~ãþý¸5ü~ :L¬eâtûs¶9~ÌWx-ýLñ†ã·Ÿm-ý¡áÓz»/¬ÛŸŸGuwi´mguøù>OÀïüÈÅKT~‰hs k'uêòÜøx[h)οÕÔý¼¯m‰Ë“ØÞúøàNißà÷ÊsfÇ I‚Yòñ¦m)¸b8X9ëdGØøê$|è…ä 䑾Rüeúgâ3nS4ÜOWd]6xdq‚KRoæ¿4ºð·%8Vn|{Ï4ðiÛ _¾¿…ûéÅìwð_§ûç{q´ü®xØvÜZô‡mÏyÑ‚Ý.~£+Ž=8ƒ/!#x‰“oÙŽ‡x1}FÕ±ê*OUŸNyDui}ß øløÛbò ñ¤;_ÜÎõ×…¼+É“R˜6õi*¹¿O‚›–þ–¥ˆ5qº1G}e+ |hSŒ>…í/,e.#ŽÿRnëyy|ùÿîÛÁÃ…ÑϲŠÐOKâˆò“gBAc…ïdqö8ë%_Áž‚ï#¼°˜s½”z\ØÀ+ÊØï¥à¡žòòÕŸ«CÝ/ÉàfXWõob+à7‰?´•~K´M˜›ÄÑܸ]ñï^ OÛD>7ôÆø€°ø‘;£ãá·›øÖôA}D%øƒø®Ê#¶…¾Šê-ßÞ½'ÊÖÎʱFÖe!ý™íðöÀoÄwKx®ŠWÃGÞI¿¥†ý“³ &^û:ö¾·x9êgªîk†?#djitðtZ¼éNø™ÂaSñ±õú¤Ž ¾MxoêÿK×$[ùG.}òz/&__î«xpʳ}øK‹ŠÃåg\ý3ž»¯ ø«KOPú‰sf=ÎË­ðéÇooÊŒ‰3 ï<¥m?:leïpñ7›ˆoe‡óò¨ ò®2/¥^SÜÉ <õ‡¼<­:ºš[­‰uZK¿wgà*λð>þ~ô^®$ýÅP|;¿k%¡ÎòUðQÅ×Ü+ÉÿÀïMï«ýM?"á‹7ËèÛƒ8|_B¾¢¸µ:ðž\úôSÿ9Ñÿh+®¥È#÷†óÆ·?}Ͼˆ7žeBœ÷UQõøÓjk¤Ni$/ªf:ðÛ­„þ”ÖƒøKê·]u@%}ÕJpäåñÇüÔ ¨CÅG5ú”‰Þo7ñJ<£&ð¨GÉÈÿ~•‹çPNÞ-ý¼VCOfÅô…6’_ßÇó,ŠoóÛ,Ÿâ³úV÷Á?¦î«¿ áÕ‰WÞmêä ÆžÚóƒxCZø¹pJp)ÛÎ kO®s!Ÿç¾Ÿ}[OŸøa©×DrÒT×èÿqëh¼a­½J7õG&ô‹’>žÇ°ÛWm'뭗󶂸v7<“:¾ô@àf)xxà¿IÿTù–úHÂgÄkÙê_ßê)§ßiÔM^p>«#+§sÖ«x5›ÉC cùÙöú™âék¿mñ?áE±>mGæWû®€ûºâX:ü»•£§ÔúXK=·Œú© ]|•Ÿ»#‰ë:÷Ä;~–O],~ÛvøxÛ/Äu Àͯá­ûCèXÄW_ìâ£qá|ô]ÔëÂaÀí|9¼ùΨÊüúß[ç•q£Ê·Ñ?SÞ” üé’­,^ÿg¬/õJ£Ç=ïßúFþ8õ)}!k ú+‰Ò‹O•ÙJêiñp‹cäÿÒré“ìˆËÁo%}Â&øçO]ÿG:¯ƒG“¹3´ùSì_=÷x*ÔWIÞ¬:÷.ò¡=a½¸tcMð1?áï':p}?â‹é¾Ò7É'@ý\é\¤¿,g”ÓΧ¯­ºQ¼Å*ö‡ôˆèƼ~5ëMu¸û¯BzG'³­àÊôU§«…þpÎYß 8ì.p¾êláÒ§h?Òo¶2ôš…äŸòØ Îû*¾WqÈg]ýÙ‚["‚ÚG“þÛ†ð’ü¸¼?M¾¿^yN¸ß^D¾ÕC=_Ñ*^Üe÷Ñgù-çŠòÍ^ÖCuèó'º-éŸ×…¾G‚£ˆ7ªzSû¾ÞëôïñíùºKÇ&žö:ø›Â_ÅW¿Uºʼn¤ÿ)åy*ÐúwtñÁtöÀ£—þmY¸ÏN^eôǬ>—ú®y¬ñªV‰çN\[x„¶Ž¾üè¤^ÔºW¼Vݦs~¥×£ —_øâ Ÿ/ ú@W?qº=ú²¾7,\·¾ƒ8(ùVô® ‹þN±úã{èÿ-ƒG*½Tñ´—sBxˆpñÍj*bàËï#ŽQxCàÿ›úÅrt~^ +ï-fè:ÄWn®þ©ðHé/×ÁW/o|"à­¾¾ð.ô´Úûè#jÆðã×õœÝà%òÿx¼ñöï*òõ^ðÕŠè2/éIôeÚ§àꉠŸú > ÄŸ ë8ák•¨_?Fx€xvÒÓç>C¢ÛR?…ú8áƒý‘~ŠêyñwÓq¹úv'O°-ð±ø¾è?}{x¾Þø€†^Âv„zÖ÷€¯KŸ‰î'©GÉ = çºåQ_ G/Uñö¾À“ö}ÔÉZ¯ÄÑ?ãß¡âúÛ^úÆÏÀÑ!&ø|'õ·t]¿C§"¾ö¨§Òñ¶ºÙŠÀ£À·ÕpÎißx¬Vï½>ð2½p Ⱦ—óEúô€ÊW\zÁmôUëÐ)ÔË£¿ç;ÑcKRÃuЗþ/Ý©øq¾7Sg(Or‡VnŸpÛ‹¾I<¶=ìë*úPêK‘·JÿdÛðÿC{àK¸tvè¨à§Š”謫ÑŸ~Lø­ô›ÂI ¨r¨Wѹ%ü â«åÆúû­ˆøÜ@Cõ ú*£5gV [ÿ—µ‚wƒƒu’×àcèQ¬…ó¹ùmñj¦x¡uQß´q¾loë: ÖLþÔV—¯¤9‚n¾ ü·Ÿú@ñ[÷«”¸;J<ÝúV¾5ôE¬‚¼éçÍuénòvñ¿½;qª4*ƒ~¾Æâ«ºy2áƒH§ú½™8¸ ?6pÂpWÕ#âEê¾w’§ªŽïâgÔípYérˆÏÚžôR–øîVþm¡¿.ÝèÖ8Ì~9ÙZ·âæÁo( xø艬ƒ|*ºàà×Ixœè ,þÎ^ê¡ôÕôÛ+Ècé—ÚΛúÀ“?Û–ÃßI8XÂÛ—¾¼™|¤…8‡žG×ç+‰¯Â ¹/½!^Z9¾Qåð©„k…¾S‚WWÁ»­ u­·×峟ϔç“ßÈWÞ©­‚§ÔJ} Ê*úªœ¯êÇ$ú@ñÚˆUWÅß ÁkCžk+ÈjèWäÄÇÚjêw†þ u¢éæÜUž¹s`3ñHzúü€ïºt.Âçvr¿KÑËÂÓÐ~sâ å_(¼ÔÔÿTL_Uü¨9³f…>Vkdõ»u¶?ø\Xš|g¢Okd=fBCK÷×öšB>àù|¿ø€ÂÙr¨;×r®ÏìŠ5ë‡ÇÒG§:Cxäøí‡ÑµÏ~ x_É}ƒo6ÆþÝMž¿I¯­ø[ N³ç&Þ¹üÄ%ž{Ÿ×‚<ÏÄÿM:”…ð}2ðŒÅƒ–ωp‚íÔ›à—çÒʧVÄߡDZ"øFø‚%¾;ÃþvñÓÄë@Ïl¹à3àØêƒ‹—¾Ÿôuä3†¾;ññ‘Æ–À•N4ѳ‘î£Îõ=hqñ_«©³8w ¼5é7Ág™óÂ/„~C õ]ñRykù‹øh9ðîÓô'K¨cÛГ5‚Ó ï‚WÓÿNç•ò=õ—ªàÑwÓgìõµÎ%gÿ%üžƒôíFØÏÂ{vPW¨NÅwkŒsLñFüFñË׫ Þ†.\üW_Iø“ ôÙBú—oâ\9ñÙãLú Ô«…©/Õ€n@ø°ÖÑnøâ 7“ÏUnDS(Ø›ø¡‰_§þfºZ|<¾ŠÏ¡•tCN¿V¼mùë%þʯÐñZxw ¼¼Jp£Bxò‘Þ³}†tDÕàÒÒ¯ÉßN~EíÔ‰:—¤s_X:_ñ"…wmEwTDÿÚÉ_ÔŸH³É­0ð} ý¥Á7ØÑ·¸ôÞÍÔáŠg»ÀaÒü^ñ}®UUÄåÄÄëòî‹t:â Wý¥ôÙô5ËèãÈ¿H_I+z凬kð”„ÿÁšà5¶‘¿È'9Ÿõ*¾G{ÐÇY3q¸œ>Ý*úëâ·Ò÷—<áÃJoðKî‹xûÂUÁ?­ Äßn¡~ãÞècï~(ៀ“8¾wsþ®&ø#w“ÏtÓŸÀÏÅJèCªNlÃ/®Þ]^À­“û¸>–òÁ&ê¾fþ>Ÿu§|YýŠ!ê¢î¶}ƒ Tâ+ƒ†xª6B<WþŠÂ½SÜßQúGÉ•Gë<ojý‚Í3ð¢úg'~åÔªg}óTÜH°L[Ià½8º2S_NxH|,ñÇÁ­_æõœü8vп“¿Ÿò1ðBˇ†o@’ŸÉ'@úÖ¿ïšoøÑI]'|¾±ÓgIøÆøýµ$ª þÐÉþ.Eÿ¢~ÑÎà#Ô±ÚŸ)üb*¨Ûåß½¼¿˜ó¢8à3žâó[Ø{CÝîòOÇ ñ-o@¯×Í9‘‡"x¥Ëï TÒéuDÇÑ‘'¬œó5ñMBÿoÝËy®;È×ëÙü»ç~™•áO"ü¬‚þÖNôEĉÎ õ-ñLü‘òÃ÷ñjú§uàÛòËiâÜÄG0ᑊÛÌz‡‡ïU|n%ü˜ðé¶€Omä¾Éÿaðǵnt<Ò£ çüÇD§%]R1y¼|’Ëè3æÓO—ß~—‰ÏvxmuwzügU/Ìy!û½‡~@'ø^%ûTxªt#­Ô--ä9âot¢PÜ^Î1Ïuoåà£z®Õì›Ãè9Gá{÷“¾ºü3GЯ–xžúX™Àë·QÎÁ£è Ä/~$}ãv⯮3MÓ~]Ĺª9)øÀÂuZèªNÃË ¨wñùµ®€+ËÏÚà‰âƒiÊе¹ú@©ç=M=Q‰=ø´øn†îÒÄ“Ï'>X¼ø#*e§ïíøÅ«~Pž˜ð╇€/ØNé_ÙÂ9ëèwHoó ¸úg©€&þ1yàÍòwPÜéçè>¬¶–χeeœ«øtX ¾{}œËEQšµù9«ü'§Otô9j0á1ÊÇ^³íÄçÞ7‚÷ Ëtù7/-f^‚ô¼ũØoâƒAßàe¡äÕ3}‰—ðÐ>~¼ùï×â; Þr ü‰-á\0õ;úáKÂóQþrâäÅô“ÿ°BêÙMOò|p•p®PZ1|Dép;‚?mÒWKÛ•€wƒïZy8Oi#õT :"Í1IƒµÈ€ú]ú4pQ{÷úT#ô³?ÛÈ/ò‰Ã=¬óŽÀ›¶êäø°ð]êÑeE—Ü/Bù1¾VÒeÚùLý£ñ1ðwIŸ\8¶ðªqøU“ôÇÔOƒ× ®9EÝ;E\G\ÄwÖÐùºBÎ}ÄãNÞ·üFçOy‡p±êÎæàßc³½aSÈ“mˆú}7qLz¶"tò#ôï°ÿñbqI_Dü˜ò;üÀ餤–ø˜—<º²ßCùGôÁ[>/Éö£SÒÏ>9Ìõ4±ßñ qñ¶ÆÇõ'”ß[u\yD;üéâp>$~GòWU?¯žºn+"Ï‘/Uuø½5€ƒp> »´màƒè§6 ãWKtÀÒ!ÈŸ[~jÒ¡+¤/á©H.+°½¼ï$üÏð7áàôÝ:ŸØ¿Õô'šé»×Ò8D¿e ~À^ò úÓ¶ŸøoÚ*Â~V]-¿a߉þªeè?fð=•¹õQEŸ­™üByŸ|.ž…ç$ôEl˜> ñ'C|`^Žurζßá"Πž©$œ÷kaîG ~;ê÷k®ÐúÈ!úfÇé_ÊgL¼ñzðÁqú2“¬Û|öÍ(<.ΉIòÓ)ÖGùXQÐ… 7óqᥜۭܗýèð©´"öß>úïò}D¯ª~R)¸„|¸Zð ¯´›z¨œBþþƒð2ÕßÙEߤ„û¦¾Pq}²µÄe×O­Šü+pÞÄg²!–G}ÀJÉÇÕƒ¿â¥¡ã%ôÕÈç]8ŠêÇ*x9âÝ¥ŸÃZÙß)¾¯ð0ê#» üG> êÇ o•?M5óÀÊñmG¿cCœð‚“9+ðDé«Y|”!êÜÆ°ž}¸t¶H9<Ëž€C&þ>Ùð‘=E=+¾¶tzäcÂé2A–ìçŸñ=*ÙwòËÑ|ùÕ{Ðq¸x©ò9–ß…x®ò–\zF\,gÞU}ŒúóÇ©/ŠˆÓÒE“Ç:º'Û ÏtqI>éxîÃ<·ƒøôà+%ý˜5ÏuŸ«ù½üè5G?7«xŠ5rßȇ¥cÛ¹º.>@s4Ö“>ôlv„ó|ˆ~Zça=¼jí¿vò˜FÎ5åÛCàVÌ3p;FTzî×}Cxm&Nç$8Î~V__|ñÇŽmÇá;‰w?F¼éÆ?m’õ3~_ÈJÈÇááY†sl?8r+ú‹Ž§\ü]ô#Ûñ?{мj€s£‘¼¤ˆ}/>(ñÈFèoî&¾Ðï³­ðʇ¸_ÂIÄO« Þá§“œ›¬[kã¼ÑyŽ—ð±2ð:ðßµR⸀K×(¼µ¸ÏIþ^VN]®|q¤úß)øMÄ›ŠÀÿ5|ó“zZü)颊ÐåãÛ/ÿÂ"úè¬=uqàÕÛ^ög+ëc'uUKð!±n~Nƒ“”F·ñ¦ù&¾ ò…?ÀsK…úÉåóT ~®|[¾nMôÁjáEâ¯iäKè¾åƒæåáû%ñD>qâ3ï&ÑYÑ¿“ÿ¯mƯ-<¢Bò¨føqòKÒ<–]ñ6¸Ö ¨ŸŽržÓ‡ ßkýè}óÁ ñ×–?£Ws¿{bÚÑ»­Ÿ>|/ø½êWÞ7ñEË£n©Ãϳ‚>¢úö…àöÂ_øþ ä)Âï÷£?IÿÅ+Á¯Ë+ñé|†õ~„ø2DÔ£W?¼™¼¦‘Ïi¢þæó;¸¿Ò å|Œ9Šò‹ÎFÀç¹oÇYÏà» /h/ùí8uö$>ÝðÝmŒõ'Ÿ€ãø—L¯DŸZJý߉å{uð|›Y/ôôÜ‹‰ëûèç‹//h#ýñÝq;õ_Ôwµ¼¸ñî6 ¿œ<™WG0Þ¥¹v¥ä{̯Kü KàÑjÎJ~[šgZʺƇÛÓAÏk)pÎ hš|W:ûµè9å«yAâI‡_„¿«òžÎügL|¼Zõ!ðÑÛG]п'M»•t—Ê39‡¼?£ü%6 ‡àºàUüócxyš9âá¶QŸk^zi¯ƒ¯™Ð9©ù–ø°%¾äš—Æ<)ù®ú®Sqá«þ¢5¯“ïÏÃë8§á×ÚQðÿ"ò±ÍQútC© €´ã/E—ôyuÕÇýéÇ_´ÜIx'~T–w•Ÿ úJo u…üÁ×\p¸Rðï4ýqúCÖ.%_êyÏÃ×~휳^øí#àéƒøét².3ð’ñ ´fpöf|HSø1öÃÓiå:œ~Ñ øô~ú‘à†‰Ÿú#; .7@?a4<WÝ/œUsdFÑOÞ\¢§‚Ÿ»<ü(¸îuªæð Ÿ‘ž…y„ ÿL<ø:òK7ͦ.Õ\.뇬 oÑ+é¥{ëAï'ÞýœÄb?¾ ñÇ—›×þ‹1×%™_WG¼”?ºæ®‚o‘.-ßjéÊÁïŸ ø.ü¹ž¸êšÍÁO'áíÒ‘ÎWs3àvUá~k®—iþZ]B^=ý-ù#yA8‡}ë¿ =\7¾š%½˜t Ò»uRfýäãWQñÀåSŠO–â%s¹¬ $ù³É^sK•‡â’ s ¬Þ¢ôÀ%ªûÀ1«Â8£¯ëk8¯›xšŸ­y¡Û,„ïU~†ªï„×ÖÄüNÍ#&¾&8ã„ü ЭèÜj‡½“>”tÎøTZŠ~³|{8O†ˆ{š÷ Ÿ´òÍQÛODõ¢ü×T§)_þ"Ã9#ßð”"ö!>ŒÒó;¾sÎúFÀuGÀ׆Øg]Óqñ{­•ÏW>ßÀy4H¼/è´ø£¡ŸcMÜDzïÇ àÏ>“à&ÇТ~ õ‹æøf¾¿úR£¿ˆo°MÏ­ç¡nÛOþzŒu4…î ~ úà)ú ÊcZ¨{ èÿ oh!¯Ñ\æ.î³æÃáϘÌ#­g¿÷Ó?!þº|ôäÇÝMý&ý‚üf˘P-sÑþd1ós‹x~úžò þ¯9Jš÷£ý¼ž¾¬ô²{ø™>™k~œ|„ÖPJ§&Ÿ_|$­ŠúVñFütùlp;}G‡W)ž„ÓÐö|æ¼UÂGÑ|Í‚GgàhË8wÛÀÑŽ€³”á÷$ÿ2ùgÉ­]ŠôEø ‹©úZ:TõrÑM7‘¨ß-ž²p'éEñ)“‚ú;˜POß>¢oÅ/Q~râOÈ—\úýzøqâ%æÐשf‰7:I¿Vû¨ÎÄ?ù¾jÊDü2ò[ë#’9<ù:ïO×Ü]õ€§È›åÏÀü=>ZE|¬×œ%ðˆò]ùø§9çä/%¼ñYâö|"ÍGjÇ¿!E](©®Gü»&pƒü zÈ—à9Úyšæ³—°O¤[)#>D^˜u1î"Þsë’~Ô>gðÕð/µêûýœ‹ãÜ·)Ö‘tÃ%äÌí5á´âdàý´°ß÷ÏãÏ`ÍpæbX'~L•ÔÛòQV>]É>çS_i#x§æÒv‚×°¾å£'Ÿ€êòBø4äÖ¬9œœ/ÒÝç°Þáý¹æmhnþQ‰£æÍI¯‡~¢+V6‡¾ýùt.¸…øZMàÀzîø‘$¼pü¨œ>°k¾ÒŽ8m™eâ•É'WóRäß(üq7¾ò5W=Þïy„¸SÏørâ÷þéâó¥ÐÅÊGŸgÞœf}ËIúƒ"ÖKE˜i•à ò‡)`þM-x }e+æ{ì¡_ô±ò)“îÖó‚áªË¿Q¼ 4~éëá1%ýð`|lœþúûªOºè«×È7T<^éœ[X=è »YÇø‘'<Œ&øò9\óI¤ÞN-¾T >MàðTí8·tÙò%“þ{MÀ-æÌ¢‘ëI8e#}…=ðvÃO«'N7ÒÿífÎB#qYñwˆ:Fû ¿S—?«xK]ÄÃøƒáÜ–Ï‹t»eÔ1cø;MàÓ+_³òÕvæJ#_„¡>­tÉøds1­…:®^msƒäk¬ýÖòCñ}­“ó¿qÃïðÏY|N™ÒG–ÎBu½töm3úˆÔKò’ÒñUKtipÔBòõÇèk~ˆüÈK9Ÿñ»°Ãì£L¾9Ç8÷úÀ7ëññÚâ`âK8É}æùÇêa|yáZ!|Æ£ø¢L±ÿ—3á8<Ÿð£ â×Iê ~éeœs•äûÇЃ• Of·Òwh$VŸmþ¸ôÀ½è›Xªë5÷¤ Üñ¼ÏàLø¸zЏzø/iâ3¹½*óœ<ž}¢‹®"Ž㟦~èÞp`[†çLÝgNßik8¥÷tù 6/*¨3 ‰_šç3î)ÿÁýì£éÙâ}É«œõ¹ —BpiÕUµøÂR¸ô—ãÔ=š«ye¼/¼`ë‡Gz„¼>ŸLóúðïMæ_+Ÿ(b¿ôÃ7ê¤_»=ð“4÷Õä÷©yÀÕîòÓo yWâ—©|­ˆþu=¸\ ¾MeàŸ=äO›Ðì u»æø2úÿõລàÉ»©wÐA¸ð<ùÜ‹w{^QºÀ'©ŸzÑã «¨¾ÿ¾ |xý|ü(¼ð ù ['u“üjñªÃG¨•ú~RG÷ª§oFœ°p‹*êÆ óZá¥knâœ3éÿ °/‡ØÏª;å¿Ø†nø øøy=ßgÞþ>ÖQ |°)új}àqiöÍð¯jt”Ä¿Ãàplj_ZßøT'úçcàSè¨À}ì8z¯ð ô®'З§Á1«ð¥ícðËöÇåʧ¬ú[¼9ôÀòw³I®[yÁy…ðcæ<ûRîóAl›9ÿŠÉ×…Ot“Ÿ ÇHSgð<’y¸Ìsá‘øº=Æ)x­ÌS?'ñÉH#w¢;jÿ1ø›ÞŸkBÇ¡ùÂw'©ƒxÞƒÔM½ÔÁò+j¥þÄg·¿šFpÔúÿûxÎâUNÑé&Χ˜óœÏþ­"žç;D½tu |gÕÛ%ðlQO‘ÇÊÏ{’ú¿¿Žãè‘'Áëá”S_Õ0õ먋ümŸp|\T/ïâ|dÿæ3¯®¼¢œHþ¯òOS§Žòü¥–Å6ÎÍAÎúã~ç^#çëvøwÄ¿„—TC_¹¶êcêFùA'ºÁ›ÕÿºrÍÕv|’¼ˆùðLóÁ“…#j^ú8Â9Z‰7-øÉ s±rÖ3y´Ó?tùÅì üGÏ^ö’ܤ¾eî¢çàß4 k3¸¹ø–¥øÈh½w¡—æºÒÔ“[ƒLÂwøsj™Ÿ½™þ`?qR:,p./E—&½¯xžðÓ?ytóÉ}’î·šú´‘ýW„OÛ|>þOÉ|žtà^ÄøüAä­ù*šï7Ò炯kÔ]Êå¯8ï\>/Ò§•€K0Fy9ÒÃäáÒaKWÐ /L¸kúfÍ!Óœ Í©©¥~èä¼o`_«ïÙÞ•âÜjÇ_m¸a }ÀIúqÔÿŠgûÙè`¬Œú? ìÇëlÜ>Ž¥ñ?F½‘až æµ¦‰ïòÛè#>¦9Žnº?D<“íó@¦à¨v}Z¼¹ øwSàk)ú)šs¨¾šêÙNù:ÓkÃG¬€þ8½MtúY/ôû4ÿIúöfüû{ák©ï]žsRGˆÇ&ßCñqÓø?j.º:¯ª›z£JsìÀ95§UóX+裨Nƒ'èॠFþ³ò—aNºm‚¡ùî£à¹š#%~µþN~“ÅÔÓòõ‚×ëµÌ–'¾¾Dd*ñ¹xÍÒ럔~…~a5:Éax†äáòkÓÜ7áÎG™'=H?£¢üA5×µKø2}*ÍAIËÏFý{ænYq¢‰üZϧYŠ<µŒy¡Â] ð;›¿¢o6 þÖE|“¿aýÐ1øйeàSR‡µÃÏh•¾‹ün˜ç^Î-~O%><×ÃôQ'9ŸåC¾G¾2è9ÇÐ1NáK"üLýèNò°qæä<’Ïc5ü â¨t/õÔmðÛˆ«È747æ8ñµôCqÃzèÇ4ÃkÝE?C>hÒã q8™?¦ù»âôó{æz¸æ‘WPo¯Ãoe¢ oBÇ$^rß4Wþ¼Kÿ.¿sxŒ^κƒ:¬‘z\ü¨ü¯¹ø1àŽ‰/¢Î pëdž(úL×½:Ö#û×·‡:A||×¼ËsÄc8D¾„/ƒoÁ‡´¾´|­úX—GÑÕ£§`þod^ ¸‡Kÿ,þS:€æwjž2>>EÌÑQŸ°Þ€øhµàQÊ÷˜w`EÌE–~RzxéεÔfþlâêþ_}À½ÿ„td™ŸÈ§Ñ«À5ô?‹ß#¿¯Nt!CørÊïyPV Þ;Àz>ÈzÐ|¥\êÙNúÂí¤N…ûì̱¶mœ«å|ÏNð´6pä\pòø«iê³fúòà(¾7à÷sÎÀodˆ¸?‚þ² ü\çÕt&GÉ:ÈÇ™waà‹íôUÓÔðlØg5¡ŽOæ™5£¿gÏQòÏ)æÑ'qŸóBõÄ(8Íu~Y6‰/W'ýÖ£ôýN’·1GÏjÃ}tù?(^4rŸ[xÝ~ø¯ÒoÊŸb’û"^|¿sŸyTÉ\ªÃÄû#ü»ø¿âsÈ'÷x™üŠÑá4ƒ©?XÎ/?Ïnt…ä‰?Y ü ÍYƯ\sm#ó˜wê|o/ÄÇF>Õòk+¡/-s^ü «ƒŸ5ÂËoC<ñË4ï¼;™;Ãü ÃÏÉ[þæiü¤†ÐÅI£ºL} Í-éCß<Î!àê6ñÏðëò xøõâ vàç¡y½òUÔœCùRUÓÀÝ[‚>$Á¥«*Ä7^‡k¾‘^Çù¿]D}ͳhCg[Å<õWbΠæ(u…óÞóÉˈ3s΄7&Q'ý»6ž?8¦iNzuy)zónpƃÌ-: _~=ú€¬Ë®Àßsx½òC“.Ì·rþsÝz.ÖE¿ª•¾ý6p„^êßzžg }RñÂ5ø<{é9_:ˆÇâãíCÿ}žá>¾/|fë/l%ßJ|Š~ò[tV®ù@M¬³1êÂ1ú ÇCèòí…÷hÌ=±£ä—“äÕòߟ$îgý³ÞOÂÛ¿þµ^®]Á¹ßŠ?o#uºHͽ4òˆDg¯y™=äemÔ¡â]jݤ©—†BÌ÷b=&óX WÔ\;æÿ$ýVñ+ùŸðüáÝiô®òáÍçüd^†çpŸá=¸|Ä_…é¥è™ƒ‘øRJ/qˆ8 OÉ5×° E~š­óZóèÊ©ñY^o›Ã÷wô‰.^wšûYÂz¾náòÉí!_ã¼W£9ªÒAì næ¡{íÛHW$8X~}Z¯eøžà_$´œ_'ßOú?”Ž¿æ-$ø¸§¯Àžyƒ¾ ò4OÓ­%/ÜÀú¬GÇ'Ô¿Ó ºkt´‰,ó ½Hº?øôkÿ]æ¦X>d%x“æ5ü„ï6H¯¿>| “y¾£Ô]MðØZÀ_Ä—Ñ}ßx'..p;¤¿ûÂ÷ÃW€7×n.’檮œÀŸeœós‚} Ÿ7™o2HÊG_ÞBœK‡øâÉœ®§žó[>5=ôcñ¯Hü•×nãû²nÊè«*?“î¬ \ #¼ùé‹ÏS®\Ë÷QÞ[ ¿p7sòäO0Î\Í“éo?€¿o¼±üAyüa¿ŸûU„•|çÊ/~B#xøsÏ4GDúÀÖ¹æet£{ƒ×ÓDßaóéסÎîøbõsî RÇuâßT&ò›ÕøUVÁ§® _Ø®9÷àÑšwº1ô[qpXÿ8¤|ÈkÀ§˜æšóEßÒÉO¿ æZ3|«=ð\å§ÝÊó)g?ξh ÿçqUþýšwS nÓ‡ÿÎIú]ãœóãàó©0ç@:?ëÇOL¾7àµ¡ßæùìÃJpåzðŒp$Õ—5àâéw wѹ)¾îíô‘û9o)®‘oÉFýöNx¸Osîw¿JS÷ŽP÷t¡꣟ý4ûVý|Ö=ϯ9ì#;N=¿È tõÒ½áói9|¯qøË|/á'Á9»©s¦È'Çñùi#Ïn ÷Qü{?Îsï!öÒí§ ¾¢ü¬FåË ~ >ð%æ`û’}ßcÔ1â—i¿p¾RÏmãü†¹Ù&a†¼\øÄ€|ë‰êSÓ×k“¿7}Hü¨l=óuCÿÒÀkñvù²iŠêò:›`=¬$/í¤ß’‚ßQ>@´§àŸVqNá•ðä•'§Á·¤¿Ý§þ³úÕMôñÐë[#õ¼â^šþp\s3N°ßÀûÊð„§)=¿xÎòý•_â‰h[®¶1ðœQö¯æ%KOÑÏó*Æ'³óþ½ú)šg`õàpòÝê!މg#Ÿë&ÖQ'û\ø±øÁ䵇À‘q®t±káíwRgd?jþG†|l$ø"['up/ùÿ¸[3ý@Í#ï ¿¶¾ø xÀ}±)ú—¹øÈÏŽ¹b6N}0þ§yÈ'©³»Á»¦ˆs“ä·ûøÍèâ…‡OÀ³é‡ÐßÃÃH|àµ~FÁ3ñÙµCÔã½<¿Zô:â5ˆ‡:î¨9p‰/øÌ˜x`àYõŠ¿ðSv’'ױބ¯ ·UÓ”sîÑw/Xs:«ˆÇðÞãæÕðÑ5Ôù<ùvI?sŒý®9<½ÔSàŠèš‰S;Â~МT¯$.ˆ×Î<âd¾8óa¼‰¼á8y‚ø9{8¯„ÃÉ·¤þÌ!ø‡êû Ç_Œ/…æ â«”äuÝø³¤ÑCIÏ.üRs¹[È—ÊáíÔ³®å[(¿ÏÕð—Já·Ê‡OsÎðewÍoŸb s¾¤«åûºô‡¥ôÁšðÝÐ\bü½‘ýTŒŽjŠz­Þ-~;ÉüÇ4ºP|Ö¬J> ৘7{ˆ¸3D^X®9\¬ë^ødòõÓ|ùÈÉ—MýÕâR¿øéð|Ô÷:¿Iº)Öƒ3Ç.É+…«¤_w¾êÅ&ð¨.ðžg¨ƒ{Ð}´‘c ž÷ƒkª0Î÷l$~j®Á~t4OS?¡ =I?A¸€pqÅó£ÔKSà¸ÒŸL¢ÇïA7þq ¾õ˜5†ß'u,¼„Ä_¹‹¾,}äïUD¾{˜>ójlˆó©W|uü8WÒ·ÜOÝ= þ£ù«øû¹ô¿ÃÄeæ¶zyÅ~öU õH#<‚Ôêßà‡Ô›š[›AG²Œº`1óH¶|ÝËð‘.ʳ7hݹð±máœut6WóQÚÉ·ã7%¢ê@Íù,ƒ·ÁQ̓qæk. 7 „8ŸT<6Âÿà—ÙçÎ$úd|Þ„æ´ËoÍ…oõ¿3ÔËGxîâ—2ÈwásÞÀógÞ”3ÏÞë˜wìáÜÔ|(+ Ÿ¥8·ª|ûü{àâç/þ0^æ¸:þ9~|Rxhê›;¾ÞòáHæfx2ÌG´nê"ñDàcï—þç$û| Rš}ˆÿy¢‹8Åú§šâ¼¾‡¼Y>=¡C9*müþD½%@7ûû8>Aà¦òëoDŸƒŸ{~ZyY'<Ì.ø%òs+¿>DŸ­€85~ÙÇ|Á]ô)·Pg#ž|‡< ]g‚²¿ CÞãÒÅts^âCcÕœ;»È{ä«+¾!~1Fbõ¬Wôï ¾ ¯Ñˈ_øŒx:Ì‘ñæÀßqtS‰ßªêËãøOÊ/€~ºknV ¼¾ è'à§z~Ƚœù+èê¼\“6‹gøüAúý½ðÛ—kÞ.ç·t¦ªwá­§¿[ðk‡ŸéògÌ x~² 7S~Þî)¿*p©Ä׫ŒyðõLþO胥ÿOúÅèXè£ÀuÖ7ÐGÃÌ5‡ú&æe£‡rxò¾|/…þ<Ž”—ÂïÒœ¨gÁ*鯡§”O—gèwæ¿IGÑù|¼x”ºà(ûFqG÷aû™Gý$\¿œ¾¹â´tè•àÜ·~t\iêÄ#à•­ôÇÛÃû&|•ÎïÃø;ŒÐߎÄ>/°nx'Ñu×5W}ÿâvâMýì“ôóý‹ÀÙ„t±~&9gÇé Ÿ`ÝŠ·´•}CßÏŽq~N¡£i ü¯‡<ú8çáçúÂMðxįof¿vÒ×ìâ:{ñ?“Ox3£ä³ôGÈ+´þÓè—€óaŸŽ÷/#Õ°Ž4÷eœóË{ÁM{ðca¾œµâï¶…}>̹¢9S—U£ÛjDwPÅœXúèþ§°/¹nÞî»Wá_˜?f¾¯•ð<4‡þ}{ù<»…ó]¸A }1ÍóU^ŠŽÓ•ÇÔ0Ƕøyþ0ç*ü¸d]ÔïìÆß£…qp5év—SÏÉÏûwÄõò˜nôfXšC ùaÍô+ä+º~ þ‚VO*}Öð/—ʈ?Ù08“úG%ôÙ9¼Vó•‡ï#®È×]ó¥ï¯d÷?ì9gN=vйò9¤¯,ßÞ¼¸­1ÛÒ!Žx;z¥)Å ðÅQò‘rò}éu{áLj„ŽËФ›¿—½’ý?¶Ÿû)½ñ!ð¬ztýûé/ƒëj®«â|Žh%SKlܰ }úé’8#ÿÏ&ž_?<ƒ6pÄ^p¢)Öó0ù s«]>!ä+ãèœÆ‰óSÔéÌ­7ôN‰_„x½'зˆo?E½.¿¤ úÇ'é+VC⃔ø°7)ï?9€.¿‡óµ„þ°ôšòD¯døgôÓ3ÏN~îZ_£ðYÄŸÅ_.Ñ»yâÒ%v“ÈçT:Í‹ÇïÐjè‹©Þ”üÜ àÙáO/? /'~ÖÂëóJp{å“òc~«…ÂÔ¥o#î—ÂwªÇl|Lé‡å?’ õ út¾ =¯ædˆ·â]£³‘îÁáí€Ãùø÷áÿrŒ¾‡æVà¯è¿€·îÌ¥³èáüÎpží#ÿï% ¿ZþïyaNx2G9âg8/Àuä;í!_HômÒ¯¥Ákˆg{áSlƒ÷ ?kù6êy§yÎMð¿Ê©G*˜ãQ‡Ï+u½Mq^Vâß$ßÀ^êÕøA ßrΛ¤þÒù9~Dÿ:Ñ/dÿKÑNYK°¼¨¾Ÿæ€„Ÿ|ç&è°^ïBçK#Ÿ—&/ì€Ï0 ÏHýíq^ßÏ|(¼~Î,ê×Ä—›û=Jþ1}üí$qõ(|Š6pBÍÙm º½9g–C'õ ¸_õ':â„7<aЏŸ:k; ?æußxØ„øgÔ»-Ô¯åÌ?"/máÜ:ˆàx~ øÉúº#ðEðÁ±QîÓ~Qíà~rމÿ5ɾAÇjíì‡2âöqúõô!„×ʇ©8{¼Gº…Cš—Ǽ¯Zåñàk]|~šû× n^Ëü8§/²\b€y/ò;Ïp•ò—^¼Iϧ¾ã†óžðCn<Á¿>nQÎz®áÜn½+ð¼ŠsG8¨æß=Êß÷Ò?Ñù4 é8û³•}%Ü_< |?Àd¾L?}Ô1êŽúm;¨ë¥û•/S}Gñò„kÊ? |z+éê{®c}Ž‚[ÈG_çÀÏÁÄ«•~ þ»]bkÀõ½’<5Ãúz–ó¢—×i úAo¥¾¨¦^¨'9Źñ,s¦8—NRß7ž—÷F¸äÕmô}ª¨c4/~?ùR3þãY·Ò¹7ñ½Sv‚‡÷P¿7²¾êù>¬×QxΣÔÇ8×;áU †ø8g×9@ØÉüŠQð¥~pÑÐçšsÆYï sb,^—ø&ïçÜ3Ñ%Np=OÛ¨†&¼´ üœþïϧÿ·à³¨‡¦˜'?E?¯:¨‰<‚ºßQ_ôs^dߤÞ+„w õ8Lœ’Nj ½QúNmì3ÍuïÕœ)øòUØîZÊ\‰IúI•á:\¼·nžo+8£æ†Õ¡£=F߬¾HZxxB¸‰øõòK¯Ÿ¹í¡ ÿu¯£Þ@7`õ༥ôyʨ›&ɇ¥wî#® /j!¯”îIúFéå7_ƒo1óR¬ˆü­2ԋ‹ìûBó ñîÆïqÖÕý°ãä!­àJÒ kýÉç&…Oä8ÏAÍ™ä>3Ÿ²¾êáð3åWÇ\¿j|s¿sê‡jêü4ßÊsÉÛÕº¼n;üÝû¨W‹ðÁ”?pxqñ0žx5ñ¥‹yS øŒ¶Á'h†ÿ>1‚C†ø9®Á"ë7’aýËŸ»—ø}„úr>ÑQòÍNò¼€ÇÌ9“¸1@¿µƒ}}˜z©üj4àsÎ$O9FüU¾,]o8眱IöÍ3ôMÊÏÑ…7‹WzŒø>I=P¿ã:É6òÍIÖÁ$ë³ <¿ ÿ8üÓ}œ¨¾àA|ÅF¾uôKɼ‘a|*ñ…eŸŽ’oµ‡çåZ‡­à6“Ô{ØWÂcÖÓ>Îý¬ÆŸXñ²—Ïm'êÆ?­Œç9Äüïzúèû’~û®K~Å­Ì”#s2¾°7§Ä7‘ß‹öþ46u}™^úÝòÖŸ`:ùGªþ*$Îé9o¢oÎ9ïµô˜ŸbÇ9×RìCñâºðoÈ¡ÞRÝ?ÞØˆÎf:mü¾m÷A}ªÎñCðÊúXWøÏºøwòEhßÏ·€“d85§ó~ÎùÇUQ÷×£Uß?¯£Ï‚.Mþð®ùNâµIÏ ~}8áâ1ïãšã•_q ¼¤‡õ(>íþòмê4ü‡‰oΙõùðûIøZSðjá-&ó¸‡ÁªÐõ¶Á“¨¿ú vúŸòS¦®“E#}€1ÖU3¼à>̓§#ܸ‹û6Êú¥?3FŸãùá@¸sΤßO½|û0ë¶¼üHèe_æ+¢{ªÄÿQëÂé9Î-ÍŸyšë®äþHÑÌù7ŽÏú$xxšux_š}ðÅ„ÃMÁŸo‡WÒê«—=†^ _;;H\?òXÏ s!lý­ƒÔÍåôAÇÀF©ãÛÀCås¤¾Ã>}ùäÍ-à‡…Òyq®Ô¡[W=ÙMü…_jÜ'ùÿ‡ï ™gNß¡“ç‚'ÝLUùÊÂk Áëà}X'ÿ~}>úíLjkÍèi5—@:TùÎq¿ÝâÖ}7ü bîŸæ­nC×ÚI\¬ o!ÞîÇçe <>ñJ4'}8ryö.Ö©Ö¹tš&Áµà~eôSìÇtâ‰3GÜSâ#‘Ç=MýØޝ:aœ}&^aønówFBÜœs<¯Ip¤ãø!ÃÓvù‚ ?iÎA#<îòP¿¹æÑ?OpúƒðkR'5³¾1'¥‰<£›}ZKÜ,†—ÝÂúE8ŠÞm ^xëŸ9³ˆ'ÃèH»À­ŽP7õá·rå(8Ü êôvðÈöwW¨?çœ ^>NãYpß-øn?ˆO¹|D¢w;oa1øêyÁpª ú¹§Ø/º¯-â³±OÐïàþ ?†É„ßoߎø¦9@£ôcFéç´“/?ÞÛÍ|È úYÌ#³Vú–ÂõǨ£˜sîîG¸d3ýì q¥˜úçóRœoðŸÌá¹ >”^ªœx¿5àÏ^xäΜF—^ôÛª›ñ©’ŽxõÜ8çÁÃð#z¸^ñ÷Ñ‹_œä!ÌIð6ú ò§Nޤ§CÿÍ«À7¦Ðj~+óY­›>.úL;‚ú øHiæ"€¯%¾ß¬+g.­ ƒcÿì(ÿO|ßø¨š£ãðÌÝä);ð­'pÍ}¯ã9à/bøÛø¸dëå æ;Á»¸^ÄUäûÌ=wâeRw¢«°²p®¹ôB•ôŸ…ÿÞ.!^ò u}üöbò |wí~ O³/'ÁÍOÑ?MM÷5²!p‡bxÉèÏöiî4þ‡itúé¤nn&ÞhâûéœI‘÷Ëßy:‚Qx:£ôAŽ¢“é÷¼È9³V_þaž“|…Ž_ûÁ‡ŽÐwxŸ IòÀv|)ñ/¶°>æœÉûj>ð3ä½Ìa”A=qé(yÐ|ârpŒ)òŽýà½Ç‰/§ÈÃÑáZ+¼(ùŸu¡ë@_0 Šù*‰ß·ô ÃÔ]šs7 ?qŒóWxìÁ‡šåO Ž·œú¡{øÁQâkÎvæ:Y¿pÎñnðærxÿ£àšÛ _ÛJպċã<­‚ßÌ\c/¦nÒ<§|6ÄûŸ^ñï5mŸiùDtó¹ê×fðíÑ\ø¤¯×gøÊºòžrø iúzª+&èC‹''D|1æðÉÇÄN C¯}ßI¿uzNø®y¥ä!Ö©ó ¿0é|…‡®æ{ˆß/½7s‚]ójŸâœ/E>Ì{ò"þ¾Ÿ„%Ô÷òÃ~½(þךƒâࢮ¹Êòm_º}|’§“¿‹Nh’ë?…¥˜üQ|&ù?Jǧ:Ró`O7ì'/œä{.\¾G ÏQ>¶½øþ‰ç4Î1ßN%?¦æ*GO;Âó£êw»Àñ'ˆOðA­ƒ<·šÏGG[Ž¿x¸õI>½þŽô"cè?¥ËÛWGÿdÏYóæ„Kçá« ùžåÔyô·’Gï Þ._ØcøÔÝ,Ÿ\ü¾ÙÿÖ Î)ü4¯Wû]y™ü㞥ßÒ ¾¬¿?ŠŽ7|VsÀ»­;øGΙE0Aœ<ŽNJ}'õFØoÄÍ&öS¸~MÖD=˜áÜ8^p˜¸ÖŒžá(}Óvp¨òÝþõðq{ðm%/åsŽÀßÙGÞ1ú|sΤŸ<o¸‹}tX:t£wŸsF}ÀÙÇé6£¿y”ðTà/z†8sŒ¸±è^pʦw%¾z‡¸SðÖñ³Jø×è*¥·aÖü†©·÷2÷­…:‰ó8™Ÿ†÷¶>á€?ûíôA™“é™Ðs͵—OT#}¿pÕ|xÜ:÷Ö2µ"ô-\þouè‚3¢3óžC |ÞÖÍsô­zÐ5”ÀSâ:4¸„û‘&HÏø¼ÕÌ-:ãŒÌʦÓgžqÆ+ÎÎþ÷Y×ο!û_/Íþ_ôÏfÿoöŒ—üCö%7νaÞMÙÿýj^÷ ^ýþ_ºõú7Í;÷†y‹®[pË­so˜ã¼sÏ›ý…+¯z¯xùn]È?_tÚ?¿iѹ7ÞtþUóoœûÝù7οæÜ¹×Ì¿öÜó.<í%ç}ẅóèwç_ðžÙç_?÷šE ¢·Š_÷žsÛðús¦½~úUœsÕÍ×\óW>眫®›ÿù7<ï߽檹×Ï¿æÖëŸ÷ßøùùWÿí/ÿúÏÞºpÁ·®¿yÑÜž÷^ùÙ›-œwÓ5ñ‡¯ýìu núîu &vúÍûôgÞ8Ñu¾©<÷ò…ßÊ~½ì?üðCŸýqò¯§½nöó¾nÚÅ~æšEs¯wÍÜ$ŸzÁi¿Ë ·.8ÿ7.¸iþ¢ןÿÝøç.ʾõ¥§½èÕ§¿èÜó.9ýﯜwýü«çÿå{ÿéëâʹ×/úÛ/yå•s¯ÞÛòò+ÎÿböWÉúÐë¯8?»³W4s ¾#û‹°d’?ùà¹ÓþíôWgsÕ¼…Ó_ª8ýu¯¸âüÏ.\ð——ñÚì¿_7ïÆ[¯Ÿ{ýÜÿâ·¯¾âü+³ßxþ‚¿üUö ¯˜wó5Ï÷'Ÿ¸iÁõåW_pCö6þååe¿ôÇ‚'üóó^æ?g¿ù57/üsD™¾$O¿ºwF/ÍÞîó/9m_Lùé;áM¼üo¼ãk?67»äÒNƒW†ßο1ú>—~k?zó s>8ý蛯Yôÿ‘oÖëÿÖUþ×Â[]·ð¯ÝW†ß†Hpú/.úȼën½vႹ7]³`áÕúÛìQ3oaö±ž÷îwe×Ç鿉b•>ñÃó³Aÿ†¿ÞpÍü¿}¦¼>zÁå×d/ø¯,Í«ç-šûçu5íWšwãm%2^¾pîÕÙ‡ñ7Ï‚KfŸyö_\}ó¢[g<¡žû¼¿;ýÓßyáùŸš·(œ% ŸAžþ=_ø•O_þõ ¿vúOMûiö´Ÿ.ÓOÿÝ…Ó¼hú³§ÿxñô/™þã¥Ó|ÿô§îìé/ž=ýų§¿øâ÷Mÿñ‚é?Nÿ Oÿ Oÿ Oÿ Oÿ O¿ª‹§_ÕÅÓ¯ê’éWuÉô«ºdú[]6ý·—M¿æË¦_óeÓ¯ù²é×|Ùôk¾lÆM¿æË’kþ‡èÇ Þ÷¾?_0ãç gü|ÑŒŸgÏøùâ?_2ãçKgüüþ?ϸ¾ f\ß3Þÿ’™?Ïø¼Kf|Þ%3>ï’ŸwéŒÏ»tÆý¸tÆý¸túý¸ðâéïwá%ï›ñó3~¾pÆÏ3Þï’Ù3~¾xÆÏ—ÌøùÒ?¿ÆÏ3®oÆ÷½pÆ÷½pÆ÷½pæ÷½tÆõ]:ãú.q}—θ¾Kg\ߥ3®ïý3®ïý3®ïýÓ?ÿ¢?¯·×}å² gðÜ/,œwÝܸN ‰¤~ýú¯\vÁÿk»ìYb0<ôã´Ð–‚é”B®X:Ý]BÖݺtx*I ††ÄCÓ_ª³ã;‰d0èAÒûÚ7X:NŸ Ó1ÿPÍ

Në-_cú?ü0Ç’§¡äµYózh×iÓPb·U˜ÏEšªZç± –úæÌÞ4ÝUÙŒ÷èú·DdÁ}Jöˆ÷’<ÿœþdOå¢8¹wä·ôý‘œ_o8‹Ü­öÛ±õ£rž[á÷ÿ不֬x£­§˜IWï:\}·9´û¦=´Ï&˜íŽ·5Óšɻ’ß?\%ŽÙá|Q…­ùüª;=Cce¼™Ò±ãbYnŽá‡/ %X $ÁIðI²0Ha …A ƒ†,ÜCH.”]Hºt!éBÒ…²Ëåp»ŽqRYE¤È)òŠBrCÊnHÒ Iº!I7$é†$Ý"ù¤O³¬Ï±Ð„Ь"RäyEùkæ>»q,Ü"¡"«ˆ9E^Qtû‹Ï?zNP«# pcaMethods/inst/0000755000175400017540000000000013556147214014654 5ustar00biocbuildbiocbuildpcaMethods/inst/CITATION0000644000175400017540000000164613556116437016023 0ustar00biocbuildbiocbuildcitHeader("The pcaMethods package implement algorithms found in several different publication. Refer to function documentation for reference to the original articles.") citEntry(entry="article", title = "pcaMethods -- a Bioconductor package providing PCA methods for incomplete data", author = "Wolfram Stacklies and Henning Redestig and Matthias Scholz and Dirk Walther and Joachim Selbig", journal = "Bioinformatics", year = 2007, pages ="1164--1167", volume = "23", textVersion = paste("Stacklies, W., Redestig, H., Scholz, M., Walther, D. and Selbig, J. ", "pcaMethods -- a Bioconductor package providing PCA methods for incomplete", "data. Bioinformatics, 2007, 23, 1164-1167") ) citFooter("This free open-source software implements academic research by the authors and co-workers. If you use it, please support the project by citing the appropriate journal articles.") pcaMethods/inst/doc/0000755000175400017540000000000013556147214015421 5ustar00biocbuildbiocbuildpcaMethods/inst/doc/missingValues.R0000644000175400017540000000332013556147040020370 0ustar00biocbuildbiocbuild### R code from vignette source 'missingValues.Rnw' ### Encoding: UTF-8 ################################################### ### code chunk number 1: missingValues.Rnw:43-44 ################################################### library(pcaMethods) ################################################### ### code chunk number 2: missingValues.Rnw:46-49 ################################################### data(metaboliteData) mD <- metaboliteData sum(is.na(mD)) ################################################### ### code chunk number 3: missingValues.Rnw:52-54 ################################################### pc <- pca(mD, nPcs=3, method="ppca") imputed <- completeObs(pc) ################################################### ### code chunk number 4: missingValues.Rnw:58-61 ################################################### data(metaboliteDataComplete) mdComp <- metaboliteDataComplete sum((mdComp[is.na(mD)] - imputed[is.na(mD)])^2) / sum(mdComp[is.na(mD)]^2) ################################################### ### code chunk number 5: missingValues.Rnw:64-66 ################################################### imputedNipals <- completeObs(pca(mD, nPcs=3, method="nipals")) sum((mdComp[is.na(mD)] - imputedNipals[is.na(mD)])^2) / sum(mdComp[is.na(mD)]^2) ################################################### ### code chunk number 6: missingValues.Rnw:71-80 ################################################### library(Biobase) data(sample.ExpressionSet) exSet <- sample.ExpressionSet exSetNa <- exSet exprs(exSetNa)[sample(13000, 200)] <- NA lost <- is.na(exprs(exSetNa)) pc <- pca(exSetNa, nPcs=2, method="ppca") impExSet <- asExprSet(pc, exSetNa) sum((exprs(exSet)[lost] - exprs(impExSet)[lost])^2) / sum(exprs(exSet)[lost]^2) pcaMethods/inst/doc/missingValues.Rnw0000644000175400017540000000565413556116437020757 0ustar00biocbuildbiocbuild\documentclass[a4paper]{article} %\VignetteIndexEntry{Missing value imputation} \usepackage{hyperref} \title{Imputing missing values using the pcaMethods package} \author{Wolfram Stacklies and Henning Redestig\\ CAS-MPG Partner Institute for Computational Biology (PICB)\\ Shanghai, P.R. China \\ and\\ Max Planck Institute for Molecular Plant Physiology\\ Potsdam, Germany\\ \url{http://bioinformatics.mpimp-golm.mpg.de/} } \date{\today} \begin{document} \setkeys{Gin}{width=1.0\textwidth} @ \maketitle \section{Missing value imputation} One application for missing value robust principal component analysis is that it effectively can be used to impute the missing values and thus obtain an estimated complete data set. The pcaMethods package was partly written with this application in mind. PCA is a way of creating a model of a matrix, $X$, by defining two parameter matrices, the scores, $T$, and the loadings, $P$, which together have less values than the original matrix but when multiplied with each other well reconstruct the original matrix. I.e.: $$X=1\times{}\bar{x} + TP' + E$$ where $E$ is the error matrix and $1\times{}\bar{x}$ denotes the original variable averages. Now if $X$ contains missing values but we still are able to get complete estimates of $P$ and $T$ than we can use: $$\hat{X}=1\times{}\bar{x} + TP'$$ as an estimate for $x_{i,j}$ if $x_{i,j}$ is missing. This is can be done as the following example illustrates. First we attach the metabolite data set with missing values. <>= library(pcaMethods) @ <<>>= data(metaboliteData) mD <- metaboliteData sum(is.na(mD)) @ Now we get the estimated data set by using PPCA and three principal components. <<>>= pc <- pca(mD, nPcs=3, method="ppca") imputed <- completeObs(pc) @ If we compare with the original values we see that the error is rather low. <<>>= data(metaboliteDataComplete) mdComp <- metaboliteDataComplete sum((mdComp[is.na(mD)] - imputed[is.na(mD)])^2) / sum(mdComp[is.na(mD)]^2) @ When using a different PCA algorithm, we get different performance. <<>>= imputedNipals <- completeObs(pca(mD, nPcs=3, method="nipals")) sum((mdComp[is.na(mD)] - imputedNipals[is.na(mD)])^2) / sum(mdComp[is.na(mD)]^2) @ If the data we are interested in was gene expression set of class 'ExpressionSet' we could simply do <<>>= library(Biobase) data(sample.ExpressionSet) exSet <- sample.ExpressionSet exSetNa <- exSet exprs(exSetNa)[sample(13000, 200)] <- NA lost <- is.na(exprs(exSetNa)) pc <- pca(exSetNa, nPcs=2, method="ppca") impExSet <- asExprSet(pc, exSetNa) sum((exprs(exSet)[lost] - exprs(impExSet)[lost])^2) / sum(exprs(exSet)[lost]^2) @ Different results will be obtained with different PCA algorithms. Which one to use depends on the general structure of the data set and the imputation performance can be estimated by cross-validation. Please see the 'introduction' vignette on further details on how to use the cross-validation capabilities of this package. \end{document} pcaMethods/inst/doc/missingValues.pdf0000644000175400017540000034752213556147040020757 0ustar00biocbuildbiocbuild%PDF-1.5 %ÐÔÅØ 10 0 obj << /Length 2122 /Filter /FlateDecode >> stream xÚ½YoÛFú=¿BM8I·§ézoÔÀ.Ф-Ò’v%QäÊþ÷û]C-ÚiQlDÍñÍw_3ïo^½ùhÝDÊèÌMnî&Ú:U–ù$Ï2ðSO~I.§:Ùìàs¿ãj ßÅtfÓ4Ù¬09¬àÓ­êä·if“ æk<Ð oÜa¸lÝn:3Y2Ç£WÍq -­Ô„âËÍ?€¿L¥…žÌŒV^—̲†gæS“'ÿ%â " 'Þ|Ô~¢8Ë 8+ •— (Ué #ø×´DRÀ0üîöˆn埧̡ /=C¡ Š$+‘¨æ3o¶²‹ Ÿš:èã(» ’j–yå 0¥­ò®`¦.Þ!ñêâúG&t=5eRí…'ÄÜìõå6Æz ¶"rtôÅÛóø»Ï.‚«È¶SÖÃVÖŒý}XGÉ[Ôí#o|N³f×—ïáfš­”z¥ˆ#FBa–‚–t&ˆû5£»F;(˜šâ?­],#EV„Þeª(ò!úÞ¤Pmrår=Ôè=)¥Ú²YK0+iæR°ôš|ªFT€èü*h¤™ß˨ Jà?¢`þxíZT»x$ÎË\eà±Ù®ÑñZaäP³{vJû‘ÝB”%BììùN=ߪÂë!jTîÑîÞÂç þn‘³6(]줖9E#ˆ\h%{»ÙtæÐ¡E¦M±˜²Që†HŒeáØhf¬²…cÆ~BË ¹öR@Þ9º%ÿº£ûi/òÚ‰vʺœåÕâº('3—*› ^=im\rÅ™ åÙbxÒVnÄë!,ÐàEFJØìezåÂ679RK)%•iÙ³¿E>M* ³]Hó*ø ¡ý;r(ÙTXâ¹àUȦYR­%T ':iÖä¹ÛG¡2ÇÁ®ZóÖœB~Gªl· »eR‰Ç@ýÀtºaʉ˜G/iÙ|,íJx(MÒ|Nµ¥˜bþ†Ù¬%iÈœýÆ6ãòÉÒÔ<9röG UWB$ÊþRƬ86¦¤[8ŠËEq¡Áj2uÀz¸û´J¶f5ÒJà•ÌòK ¢nȶA‚ sp¨&Ê"°SWìO„–„>rÂ˼In–= ¬`…]Ë–ëžå]%9 å]4ŒöDµB„¦ êÆšì “Ó>Ô bM¤=¡,¬˜K%‹¯±ÝìvQA|âÛÖ&AVŒ´­IP*Ò¢Çüxñä×3¥GMáÊ MUáç‘×Û;þGí›.AIðà‘Š6½ÎÀg%Ñ;•Y$¸9—:Å4– T`Tœƒˆnªº³Èá½–%´t6 ½# Y|Hó´äJ:[öÄÛEÓñ°çå%ù ç ^¡Š É‚&g9‚»_¨Ì{>eƒ<‹…Åit¿XI2…e÷Ä. {Ûµ]sZ6%x€ÒÑZl¾’&m½’Ʊf §¹K>ÕTAÄK¬¢€HñøZØÚS†¦8=¦ý=WÜ‚¾á˜fQ]O¶–øîE}àl††ºÄÊßàçíˆé¡+cQÓ^¡D–ÆÒÚ¹ÈwÌe¢¡‹ÁWtçÀŸÓÔŒ KÁ¸ycòÎf©ÊmGñaÌ)}Úyå7c±„£>” íòÁ¹µäʶ$v ÂMË)­Eé95¸õ¤ ¦è«$8Ç-ŠV´I¬oË€§eè ›±¬½™Í^@Šž¶»v¡ìë3Œ ù>´Î°p°ØUfÿŒmASºøŠm5¤`ûgmëUžw"×\5‡U3"ÞXìÃ2ç–°WÝ®CûV&QR¢$¿hJRO‘©¬Æß?[?a`\w#BÙÜȾPœ£í,5oå^z£ßpqÞÓ ´%çÝ‚ßJÃñI’,B¢†I‡jñ¾aÀê6\œx.­ž[4=¾9ßQ×’úŽá€^¡¥éê³·3p™6íSX[¼PxЛ|äUw¿|ŠÇ+›/TMÌ8¦³Í ’X²3%Ñy%Ý÷ÂoùNèTi±S‚êëÄ'~ók£JôWÈò/Ì¥Ù_’KíïÍ¥åRûl.A £rãèÂYâ!Ô*[›4± #¹42Zi—=‘S§™Ê=R9®NßþgD«F“虘÷ çÿ‹ž¨ãÉóAÈTpJ§ng7ᵆ^üŠ~˜‹kÃà–zñ†'u»ͤ’SqÃÓ»ð¦Ào%paB`ô?TQº t’|äÝ1TœJÒ8·(6K> ²}¸J§y“È]‘ú –ŽËP-Aü‘{L×´W,l»Ž^æÄêÀKÅóCsdÏ‹n`ýÅépˆE¶ç·[ÅfÍÒᣄ.¡c—gÎïá‚ilw[» ‘ÿG~ÂK¥çœËÖU”—òXU‹St/~é°¥Jux5‘Á¾‰0VÂH]#C>ÿp~\õ§BmCGhø7z9íVÿ8±ç©ÄŸ6½> stream xÚåXÝ›F¿¿ÂÊËa)p°Æœ‘ÒJIî"µéI9)×TÂfíCÅpœ4ÿ}çkaÁƾ$ŽT©ëýðìÌììüffysqõnîOb/ŽT4¹_Oz _M®ýÈ T<¹O'ίSw®fÎÓ4pV<|C×^ œÚŸþ܇n íÚK¦(`x'T5´_ Íº¿‘ZCk =B+¡¥BöÚ“4#ä ¦ŸîŸø7¼x>ïiš‰¸m'¬5±j¿†:·”ÁþhKÑ[Ngt1:\½[XD…`†Tè·õÔÍ”óeª®!ŒEæÙ$•–¿³ÖDbl!;Ê ÆH²ÁŸ"Éy×çé…Á_lU<£kT±rÖt3­ÌŠWÓöÃE&¯°nÏLb;~Áá¹xŠÃփ̲ÿZ‰ ”uíB'†Ê$G´×GòW =?–Ùà{I¦-mÖ–ή¯ø¬PÙ¿IàFÓcuD±7‹æ?ðT(Ô ÿ»¸½¿øvܲ^ endstream endobj 40 0 obj << /Length1 1579 /Length2 8358 /Length3 0 /Length 9390 /Filter /FlateDecode >> stream xÚ·T”í6 HIK§ C#!ÝÝ éÀ 00ÌÀ0´t‰HK‹´tI‡”ˆtI‹H§t*õ¡ï{ÎyÏùÿµ¾oÍZ3ϵ{ïûÚ÷³†I×€WŒ°(#à(^>~ €‚–¼©€ €Ÿ_ˆŸ_ÍŠ‚Aþ%Çc3† Ý ¸Ä?,êN¦BÝj!àuw@@ *! &ÁÏäçÿ—!)Py@Á->€:qÃcS@¸x#¡ö¨»<ÿzpÚrÄÅÅxþ¸äœ!H¨-С ÎwmA0€Â AyÿWN)ÊEôôôä9»ñ!öÒ\<O(Ê qƒ = `Àï–Ú gÈß­ñá±  n) v(O¸À ¶¸Û‹; Aî² Ô4:.ø_Æšðþ@€Oàßáþöþ ÿã ²µE8»€àÞP¸=À ƒt”5ùP^(þm‚¹!îüA ( dsgð§t@YNºëðïþÜl‘P”Ÿö»Gàï0wcV‚ƒÎÎ8Ê ïw}ŠP$ÄönîÞÀ¿× Žð„ûþ ÙAá`»ßm€Ý]€Fp¨«;DMño›;Þdö@„ŸŸ_LTq@¼l€¿z»@þ(~‹ïzðóuA¸ìîÚ€øAí w?x¾n …t‡øùþSñßO@†Ú¢6{(ï?ÑïÄ»¿ðÝù#¡^sþ;ú øþýdyÇ00óþùŸ#*éÈ©ªpÿÝò¿•òò/€/¯ˆ€WPD $áøýw]ôï:øÿã«·CÄÿ*÷nNÿ*Ùãopþ½ \€ÿŽ¥¸c.Àù¢[ð‹ðÛÞ} ü?ÓýËÿËGù¿ý+Rv‡Áþè9ÿ2øÿèAÎP˜÷ßwÌuGÝmânàÿkjùkuå0ðÿêÔP »]ƒÛÃþ=F¨›2Ô Ö…¢lþ¢Ë_r£ß‹ƒÂ!º7èï«À+ÀÏÿ?º»í²uº»>Üî8ùG¹[žÿN©·E€o™ ˆ(„D‚¼ñøï¨$("ð¸[G0Äë‹@>8uç¸kÎ`‡@âý>Q11Pë·è¿C #;Æ!ÿ€B ý? ý| ÂþÅ@çÿ@~þxño(|g{wKþC-ºüÞerûQÿ€wE»ÿÞÕáñþ× lݑȻëâ•ï¦ø/üçn‚@¼ ¶x³S[ÉPÇ÷¡­Urtž¼kÃO¾°­™¤rñúÎ"?¸ÿ"ÂIâª|üy&—Ô×E2¿¢Äy*;Çxí»ÓT‹Ñœ¨×rùüÊ:^l­of”²g$G®æÃ}z^CÙõç׮σœî5¡·«³e»º?&ÒÍ%»ðìVñªùTòu0|jMo½RTÿªdœ7Úè•EPá[ŽMæ$536Š—÷éñÄéÙÒ¬‘[Fõxn<¿ÑB¾f‹‚1?'}¾• ºuаҘQ3Ü;%c÷•ßLV§šö-z«õÎŽÁjªAú1ÈU¸š©}Vªd`di€zË:H¡„xe,<ÖVo™M¾o fÕFMš]#V8~-`‡ç¦×¶qMéU>ñ"Óã„ÚÕŒHû˜{zÊ)ƒÝ¶±™<-:C“ÛóIÜœ£6á6@6~FÐ&eÞé ò"Ümû8ôQã¾5?1|º+m8ŠFßÇêp`Ê’Ä~¶ûÖ¤·§‚ôÖI®8z CÃq°ú»z»ÑíGí·;¶!2sVRë?e 0§] ×ÜCgÀ%汘tG õN8æ€ÆåKAƒÜ*ó×;Ÿ2åmHe˴Ѓ¡ïÅ*‹@ üäÈ\ÚÆû“Âs]‰áÚ±âñø[àJypÂó1µ’ѽ0U+ÏZÄ¥#EÅ}ÌR!“ļŽH¾ÁÑãvu>JÑéŠ ©TC­¾ÌuÒ¡HjNš2íàÏo“ –CX2¤Ì;Ïrûê¸Sn<+oWð®jYãV?Ž.[,™hó)p[ìk¯¯Ógèjz@„5!'s³&që,Ö ×m8/Áo‹ñ¶3ð.Zšj·Y|0ôñ™¸;ƒKLÏ™îro\•¶;rŠKó‰»CqÌdðx¯âðlNs§îD”U¡uC##VGÇÞ*oѰѣ¡¯…Ê6üÏVT <©B|Öô$0&ûpZ‰¦–†ø^QØz5—PæyP§”Ðr¦“J™I½IÑø:‘^ƒþ…Ûçêʇ~t®ŽäӲėÓ•ä}¶¹è”ò„Å9bjÆ“A×XSkÙ&·Mú–„uoÄ»±ÌÌ2œŸÛR!B#ªã‚ú‚Õ,œÃµÑÂýái¯·×v βiø-h®í4 2˜3S˜I­A:Ý·þ¸s߆vc2Ê;ˆÑïI¢‡rösÌ­3¥tÚU‡^ lMRBS?_Ó=Féÿ¥ßÃiA\Ú-¶soÛ£ï+‹ÂïÈYg_+šñV&wV¤M˜Ìi$»ÔUìå6ÖçøºŸLéìÔPÄ—1ý ¢÷e°‹ ç‡l`½Úšoÿ„Ãó¤™{ ÄX™Êv]è“PÕÍçÇNŸð~XcíbÃZ(6äà¸qßAékuÁ gÚŸ 'Ú%r ÍÓX‡²À£P[Š©4Çk…K-ÁÕgƒ"È ÍØc‰<êÕšüÞ_‹åmCþ‹7¶Ù¤ÓZÏî-F’1Ù‰[KënWúá*矕Kcͬ§’%Õ§â¬á '¶2<ö`ár«Sïbý¤qz@C %Í£ê_aUÚ‹n$]éÿ€tÁŸ Sœ*Y%‹—Ü:”{d";Eð¢5Î9ºþs8S YTüãä“´YéÞà‚ÌHÌÐãe]3tçÍ8lgˆ¢îÉXzÑ Ñfºø×?ÅßyÌûEçé sàˆ ûÚ’ô¢aS;*uõË„bhL’ŒèÞ®ÜK륇„ãêµIr;XØ÷×=ÎÈÑɹ7!f º$\ÂUvSÜ:.¦Ö/Pþ‘dóX}3V)‡l…ÏÔŒ‡—ºrd¯=lÚT A•üBƒë̉ œøœ'éS\°2|¹ºg®”-ìKÖíâ£4ù¹EÄ{¿%¨d­†µ^—èìX±;âhm Ü0wÍh(ÄåËœRàw§7Ê=gøñ½dí+ÜØ}O^fnáÖ…éñ¦üÔ',YG&­J‘•(Lë>^õ¯?ãÛ0 çEä›cŠ"’ãÞGÅvK{iïç.¸¬ÄºÝ{—Ù¤¢8«Sà }œõ}òe'»(“¶mN]ò@ÛÂo˜‘^æR”º¿Ë‰†,fÊ9Hœ[›ðLq¼ÓüÃ`ÓAÔÓWŽØVêÉB…‡_Æ ±.šBÁwÜñG/Ÿ»ÉM§ü°Óèá>)` *.L4€ÿÌÿ´âÏ`ýàQ¸%ÃÛ: èÅOõ!è›~×Tgu€F ÀîCö ïG?щüÎ"¡*Æ¢ž tª%Vç÷áR$>fztVš“üÚÛú®égL_‰õ¹kê´?± òß»²þ¦~ÿZ!XMFJä° ßÀÇ­°v×Ï]óY²ÖL`I³Ø[¶/ø÷öÎ7ZŽVÿ=öÖDùÅå?ÏöxÆõ°‹O1œž^(gv=Iƒ5‰†0W9ЋnÌDÚ¨Z]YˆˆD$±¬ÌzP˜‡=©å¢Ø ÔX5(°¢¯õ‡7¥•kÑš^òâaÍDÞ{tŸðFl)”DÍŸ ©ø¢Ólp¤ÏåÄÂ;Xz­¤1bdz§8²y31ç•/Ú<µt ã*ÝÔSôèšku†øâÜ¿ ú`¸ÐÖp"XaL_¡íö}<æòˆ8 M>dÎáÕÑWVúiž¥cé¬õÚcÄ–FʳjæH%Ñq6 ±TpXž¤ÈhHŒC½š³~¿…ásÆØQ¹­€®Ú·¸¥s>X¶ÞºcŸ2›cV‘n-žÀ^‹`_EµIà‚·³FChSÏg;æy ^w¥äåY²X.P¬³üÉ‚wCŸ-ðåú"<™^]¾ž^¾Ž~+*Ç¿ÕåÐA®¯,^•e8EàǙ˞zÀú=,.@bàÄ¡öüö4VëÖ±c§(‹`†î[œ2•qèíÜ •å^†`l™g‹;êýžÁú>â‹€eñŒR´WÙߤžž›Ã{SêËO>/`Äìà’Yl[š y¼Íú«…‚;ú„tÒ‰ŸáÐnÁ.4'9öUÖ~iqp-®'Rxáé«vKèœTîuÑí fGS¬S·{×ZHÚ™m¥ÊÚ™“šWÌ›L?ÙQsg0¡ÉEÆôcðÈ‘v¬x‰%Š²Ðµß‚˜Ž_>y±ÂÒKvyß—Ôç^@e“ÛLü®Ò^3ÏPÕ=·(¶xf³´2Éz†|ÒÆ5$|G{.#• “7"ÙÞ™«àË)^¡–¯_ÑW¤Õˆƒ˜©™ÈVؼد¦:%=IÿñŠ`&Ð]ÙúMÕ|¿"}Ut¯ÙÀCv<ºUï¹1îÍr N¿ÅèØ'asëÿîŸaqò3òâ‹%gÌþžÄÚH°9_fŒl~Þ7,{mE%kê-®ì{‰žXÙ¯On õÇ›³m*ÍãÆ/íZNÔ´æÌ·í ª¡K=kÚ‹á•Ù;µ”ä=Þ¡ÄVñ3`}zNp¾„ ˜®wrçØ÷Ì ¹®LñBSÞN ]pùX+ø¼)±ÛOíÑB-•¼Âˆ´·LºŽu:V«p é•ÄrRµÒF¶±e¨@ý¸ñÁOžæ „2?Ø&DÛüƒæv0kÁNXÅj°Ì)ÓwããÄ<nZÝÍ‚%/D½aƆ6iÀ6hR†¢:ëjСïÜ›t87àå¯ë:·Œ·=ï©27x@¨"o(zõáØ ò¦‰­öHYˆæÊž¤ «`!Š=WËoækQ¯P–¢S¯Ã³h‹,¤;Xúû¡ä9´IŽÎºm/’¡l=–fëtª1uÁ–ûËcšb’K0=ékZÂ*+Ë.¡èÏ1ß›}Ü«ÒåŠz eS£& øí°=<6/,sûÚ ’Ý0GQÿRü Šyæ„+'˜Âû}oâæy6úF®Q6x½É¤ò ¹Išwy@ùĆÿ±™ë‡Ýâ9F¨ZnXë»,…ÃŒG^!%Ÿ,/K6’¹N¡Yèè(a³Œdc´ KçµÈ ã/hñs:—‰A§­\öhЛ&¼#6^ u£žls»*CS± m˜meT®æ žs±ýÇZ¹Â#Ûô÷}6Å%Vk–oœ‘§º'1,„Ò Ù¸6^G-+Öç-ï4?6vgá{—†—$†TŒˆéË}ívôýJÁý¸½Ógþ3+Èš-¦Ê{Â-×Ycq§-¯ìtó »íZ_¸­8xR‰öD­_M|f†úöû€‘céÛƒŽ5~ã7»ê>øcná/'ëW_.M9RÝ/Z¨ñ÷ÉÐôu•³@Ów`劒ðÑ6P×üæMïhõ¶ûõç†:ô¸ú‡\ㆣþ£f÷ä(~éòcÅc«ëRúsJû‹/³,Xðè\¶ºeÆ Ájãúa*²A÷Äážöx0H›dæ}1[´Öò”¶À¼½‡ÌF °hXêY<åÓÍqöÉu§c0CÑòÙ6›o·kùMÑè¨å#ï£õšµÓCö}ßcO=sn³—ê ¹®xÜMK²øû+ÎpžïŽZp%;5gü"}†J#ÌòÚ§D•»Ò/´Ã¨Šòn35H^JÐYvçË3•|¢Ò– T§x¹¿Á1-ÇwæÔá+¥ÛÛ•"³Æ ´É]a³Õ¯ _‚·2a> Y9JÏ''ü’q^›ØÝÚ0s;] ž€‘}ï$_ÛŠ½Ð]íùéÏPîÛ)þâ:f| |ô80•ÎAyi—h…‘穳ÒôeÃèÁÑÎåØR=ŒÙÄcæÌa^üähV®ñÓNÍ|}ƒ£?ž,0ò…Rì>øÌQov¿¶sÈÄf&ôo ˆ³YÙÔ0™(Ý1Êö ?B?Ëê×Î Ï=ãQŒBö ÷70`ÒõWåK¶+yáXÜ@˜|ö˜?ù=Ÿ°sÞõlü6m¬7&d&¶-ÈØQ²­°HªW2]Æ<à@t VÉwe(zCý6r­ÇŠ?G5³ÜžúF*Ö²ã_„(ú‚Š-%Xá$àþ,Ž¡! %‚d—Ä5З‡góÖ ¨Ð¦ÀDDWýW %%¢yaæöwRæÛ}e~Xbbl3ÍiN@Aµz§HŽòBÏ{6RƒòI¤gÏ?K>]ŽÇ¯h‡u(ûΔðPœ½Y/Ç#ãà‘íÐ癹ºÊV=£Œ¤&WˆªÚã!¥™{~dD³ùÞ\ðs±ïEãàÚºï‰u ýË;rÃ2[Í;8OSç•ÊVÂâÞ°×j0ññ£¶¸Ê âHìN''ŽÅ’°c`cðt-Ï5S©7øÏûÛ/r<©YZú%žŒ¶Š.äL «þ¬*˜pBú«…¢)Å8ŸL¦{”GVÚðt™§{•-õ’< “mH]DÒkR5tpübxV '¦ëoÊm½žÚ¦Ï¢²*gܲ|©¸&l«eº?Ÿü(Èý;ÖWŸ›UªªàÔ*ºRäÓ±Ýwïµ_VeÕ³ÑØã¢õÒ™²; ½›ÑiÛ8þµ8a2,Xm5G}‡Nä£R¹{€jg—^_{Û}°6¸ež•¾ä³vF+/ª²wÝ»#ŸôðªlºWG~Ódªä‡Výf7,lbBÚb½™ƒé“‰^cCE¡A ÊR5q´BiWf÷n¾-k³Q.èa[yS ‰úè›Á®TþJ"×é €8r‹î0^¾¨ºAíyA^–ÿä\F¥F1ÔÉ«»çÓé\×Àÿ"ôk¶Æßy× O?·  aÙ¢Z½ã–<9weV¸VV‡iyÓiI•+ÅTOáÀ5øø<ÃÎ\•¢6.ô5¨ßü{„OÑ' »p?äéuaK2Ï&x,Ââ¸]º[91Ó±ÙD«ÔõMeÒœdœŒÈ®¶Ð[6PÌÔ<›€$ g LjÂ!Ú¢.Wt©rªÜÇgõW`ˆaª]öË&\ ÄÔÒ¨Á—k0“8˦ië$ŒGø‘V(ëå,HD¼¢LlÐu0À¡^ŸóhºN7ʃ֠?.¥žÛ¯µ´:} 6ø±µ,1Ä8ìQº¾Tç÷£åzæ} ¾{;þÃi¸ìå—á³u†÷_£iÒi‹×¨ø?á åM‡·pÖâÝâµßëÿr‹/qhé]gX΃gÃ×ð—šðJ]v÷i³êÙR=…ÿdM6·~ÜUÞ|ãidÞ©ÞØÝÑkK»Y°«Ä­ó”øöe€ˆÝrS«Y2ãQƒhüéÕ²•V׸÷¥ ¸éT“¾P”é=ÛC—Åxžô¸K'åCMEîçný`ÑŠ÷Þ86¥ä8¦ÂX*ïYº÷ÄÆ6„&y‘”$$ñÂÌpÂa¹dõãYR:ì¤%F I>ôXEIôyD°Ù»QÉYË9—l‚ÖØ*ö-+›xy-µÃzɧöëãMÄ]ì?Z:kf³_—Õî¢eߎk i½…bIîú¼ÁP‰·[þIó-Ê3¥É5Þžä³ðþèÁd^JÄÞègÞÜ5¢M¡ïKÍã{¼l^Reã3.ÔÏHÌL5ÃS²ìCd’>úG±o‘öæ>†¯¶JŸÙñ/6N„92öPL>=<@‰¦í ƒ$ÓévFIF²´ £¡Àä1÷uZp;‘>æØs䱨›>%ëžx5A8¹Õþ˾N\ïÃ/ëÈŒdÑ驱îpUŸ|V‡‘4¯ëu*qûEÏ ùù´%J"zˆO“w!_¯à)f#©þ ª…äem÷àT¶€yëäÆEw墩w^:Úù¡ZÏt‘=»[¼Pí;»rYð½k>Ÿ0«œü´Kô“°w÷¼[,¡·8d¾ô¨FYéböqóúE%³¬ÑUÑ… ï†cVïlÔÇ1Ëá@vÿøÃ0r²¤z6PÆã>Àv=4¶ãäm¸™ÃCñW¾2¯qW¸ÌÚSÊÇ  ˜ì1Ü“Nã®26®Ç)8?n[zr<É Xúð¾eÌèžY•7¡Ä_°–ÞÇ=û.4Ž›ç=Åá•ÝH:.b"ÚäòÍKâÛ¿umV*m•t<–]—Ò]–[ëà =1_¸áš+¬¸ çB¨Üÿ„¹n£"k%µÞ]í¶|£ë{Á$C„“Ϫœo¦áÆÆ¢Omñ#õX…æ´-Ðy|ÙtÉxMz鈖3ë ¯ñŠsOG‹GÓPG‘vFì͸™j1@ztO#4õaäøÈ V¦­lˆûC’y?bnµ~Ö†·%¸ôÍá=­QœrA‹2øû¶çƒenáƒ%h} IMPÀPmQ‘ Žvl>Ù³mþšÎà9Yßø Œ6é<MçÈš}ŠLˆÕN>El÷Gïéºõ^æxfìâ’ªFù°]óoR*Ê9½~|º\¾ÁdÄ"TX]BhŽ¿}Ú“AÀx$,JÂãI}ú‰ÚŸØ÷%c`³Û^ÇÈXák…7(ÄwOÂ"™Ú QVƒå?Z¹Ï®J!²,œDüXNïS‰!©Œ&°=·‘ýn­ÿ%ùòIõ–âµ>ωËÕî¶Ô5;6Œ•áÃ+&¯êêU*þ)Éò’µù‹6æg¡”«Y®£ÆéçEñê%×§%tèZé¯Ke§( '£ÆbîgœRoË›~z‡Õˆ^A}8 ît̰ÊäìŸ{¾¼£S¼Lן?ï0ŠòÄàýé1‰¶Â«óm”NAM µŽõ}M¾Ë‰€ˆe¤îµ›÷ª9Ï‘´‚Ë>ÝáÁ¹)ñ¶öÇrê¡ÃFm(½{ªÈ*Å©Î`”&Óc"7êoç` K5†ÜꃎO]隦Ò7/2óY6¹@Õ¤M{À™†×IÜ|…änÈinhöZMŠ`ȵkÛ¬éÙS¿öt¥ E“äk Ì*§SÌ­ærÙÞ™Ê4±Ç¡z)§½Ý+’ îÇûüÊâ¬çµ㉋–šGó•‰5Ç[/@ï’n8¤–¶·P²ªçž¡èähæŒRj¹¾o]²™›,Ÿ…Þý;6z«M€PZlz7‚ÂVRxB9u¾’‡DyUÿèõO#2ÄhF‹(Áê;+­ ì¦ú•ÌLÃöй"°g$6¸ßÂdz0c1æ.&L%笪¾u”ùÚ H=î}òÛg’ͪsö7Ç¢²7–¡ã·mäEõñ8æÏLÑp÷Yà.{.Á¼eÛ´®û[£Åº‚ ‰®Pû¦QåÄw¾Àâô2ŒmöCÃDZçgg.Òšöò`çM|WǰKì>Çy° $£ãaÚ}dC6ûX_w[ÇFô³™G‘¦ iɆêÀ²îtqí*KâņÀE*Q(ø…np‚#\ íW‹ùt÷¬!óR¼˜Ï¡å Ðûð%ÿ¯ýj4ITÔäÖ¯-åkS’MÑÄáΉŒ C£Ø3`›°tôDlñΟnίå­ÎÁùX±Ë>`š|°@ç$êgó(ž«zY1WÁ+Ú€R5ꄜڰFê>âÊü«ya/3LØzQz‘¼K°`)ª©*àôLÏÖÖÖ^±:4=ò€ïÌ‹ž€ëIWlÆW®©ßX>^ºZ.#ëé×L9u(¶²’ttlŽÒ¡|:CÓ<)Úö>¿M4(Hà Sfz­HI±4ý¦XrÊNr¹ý³TxP/ENöö@Ŭ4nð»AȪ®\®ØmÒHÖåû›Õt!y&íoßôÄÏh|?sÖ[0!Óól=ÍÙ±<ºí°:ÆÞ[oúZ3IfÞP¹J…³¼·‡·§aWîEM—ò8_)ïÏÏKy2¹Ò›Ì\L)e‹‘;_dSêÄ| ×øTóžg9ضõ²f£ÝÜ"æK–ëÞñØ%ÄÚµfA@|Ù~VéGQ >Kµm)ÖãH0yÉ’j„{‹ñó—Ev|Y¼W i#i•x# *ùs¨r«›«çjËH Tï•êA”yL;ʲFÙ¼c|T gg§ºÅÜØ™bKŠýÝ–$¤?ö'ß\*Õ…~™ÒÏQR|»­uËR=äÅ¡wùŠÚŒtíp¹Ä/m>¥uè™zûU øÊ3ÜE3ÚŽ² ÿ(ÞWW«Úë5!ÿ¦îS±Ûè¼3‹\›ŒÌ=ÏôâøÐ܈X3ú9ù("^Mö‡ÎÒÁå‹—Ú0‚›bÆŠå› å=Zâ7ïc¶zªø_$VcfIÐjyäßGy½Ê—òMï/·Å$Z´õ. º%GLÙ¬.La_¬,‚âñŸ¦õß§œ]’0c‚á`~ߣ2ã¾qÆ2 ¬,ÆÐÑK¦ŸÒå±ÑO8Þÿý´ø]ÒÝKQ‘V7ó„߯Úϱ&Óö …ðܯ¢¾Ö±æå—¢†[ÞŠÁGTâçªbxo7Æ£™Ñ¤•ßÊå"t¯OÓìùI¡¬q6ϤwôWBέíZaVQ©¯Ö:Ë7»ây|Åÿf­ê$LQMø.øµ¦ûПÞÂÑùóŽù:©Í'ï/ vy§â þ{[i„/Ô³{åÎù8Ö;j¦ ’˜ä›?™ UìÕñ¤USüÀðuµªŒ¥ éí”øÛ™<ö7ÉòøöH_O¹st ùv#­àcÜ’kL Wgà«yvìŽ×_§*Kð\FLÑsµ××ÅÝ?©êë ž…ÈcàPR¼¡Jd×Y®ìŸ!o\?jG‘f=4x˜²¹üþ f)WÑk#«®%cÎØSã„cîóþ¨HÓ„bâ?ÀŒ]‘ñÕùMÁÔxôaÁbðaÂv8NkíyÏQ<4î_kõ¬ÒgT®ß“'èÔ_Fm5Þz4.p¶Y{»ø3w5V«gôN +ÓñA 'Ö »k½ÅåO.·©Èýœüµ}_]3OêÂuC7ïÃUÃQù¤viíD¢l*‡¢—z±/„^¹qlFÞÇ¥¬ùÊV[:v¸rPq`ß^Qî+)ß~¯HUê¯ Q¶"0¯šËä°ô”(ƒxMx*ˆJ”:¬vöÿä6¡K›ÜÍÁåè –%ñ2 ‘ßnÑ í&`J($qMéü’-5òæ“pЧ Í=ó\ÞnŒ‘„¥JvÔ¾&rsb»Cøã¨š„VÖcyC§Ê¨HVhk¿{…&¦6UùMËØå ±›Mmÿ&þs¤«¸65 ñpçàÂòâÿŸ, endstream endobj 42 0 obj << /Length1 1450 /Length2 7213 /Length3 0 /Length 8198 /Filter /FlateDecode >> stream xÚwT“[Ó.‚ Ui"ˆ€tBï½÷Þ¤† ”$$‘Ò;JïMºté J¤*¤HQ@zçGç;ÿùî]ëÞõ®•wÏÌ3³gö<³×æ‡:ú\²¶pˆ†æâåŠä55Uy ŸäÃgf6€¢!ëñ™ Hû_y$„¾Ñ)€Ð7@M8  öÔÀËàã|@ èß@8R  rƒÚ4¹jp…Ï,Gx"¡öè›}þ^XÁl^QQaÎßîY Áš ´ÄåfG0È C!hÏ…`•p@£b<<îîîÜ 7i/ÅÆ p‡¢zé±ü* rü)Ÿ`àEýeЇۡÝAHàFá C`¨—§0[p³;@_U €Àþküàü9/7ïÂýñþ ûí ƒá.Ì ³ØA!m% n´š‚Ùþ‚œQðê ²¹üNP’Õ€n*üS Œ„"Ð(nÔùW<¿Âܳ"ÌVîâ¡Qø¿òS€"!à›s÷äùÓ\'Üæý·d…ÙÚý*Ãö)‚Çu} QUøƒ¹Qáÿ£³‡ ‚@a~AÄñ;ðüÚÀÀùmäý¥¾©Á×GìnÊ€øBí 7/|oÈ @#ŸB|½ÿ·áß>//À Fl öPþ?ÑoÔ»¿ä›þ#¡3à ýxÀ_ÏV7 ³…Ü=ÿÿn1¡Žœ±¾ÇŸ’ÿc”“ƒ{¼¹øù\|‚@/P@ |³ðýwôOÿËWfü ö+ß›ƒú;g·?$`ý3!l€Ó‚ßP`ý‡éæ@A øæ‡÷ÿ›ï¿]þo4ÿåÿÉôÿÎH驳óo;ë_€ÿÃr:{þAÜP÷)úf 4á7Ãûo¨1ä¯ÙÕ„ØBŸºü·U ºY˜ý ¥¹x¸é¡(%¨ÄVŠ;üE›¿ô†¿Î ƒèÀQÐ_WÌø_¶›);Ý\#¨nþe¡nFý»‘¿dÈÍPý;Enûkúø… $ä‰ÓüIàÍ{3¦¶ßìðpÃàèÀM;;8ÿW£…D<Š¿T¿% €GçIÀcð$à1ùÄËwõø-þ+ðS$ò&áßDºIöoù÷Õx@ÀøS“p°x°cMpëIµ,;×êöNê‰ ×P%.º[qÔj)^?+sZ½Tiª‹WÉÒ±]KÎõ${þó¾÷J-C§ÀƒÒº=ƒMìäõÑ­ñïcZÆÉ{M/“åèÅŠ‘Ý:tQÄxÒ¶öÝúÌ÷|×t~Ipg!/QV iÕjë¨.ÉÔ  5úþUÝlÓg°*x@ù8ç«Ë‹ åhóˆ­tFu‰¶Süx÷*âÅÞþ¾Y’.ºwQjn«»L¡hEñ>Z;UF³yw/rŠÝ¶ÒI½m”Éd¶l»ËØ’Oj´öVö¶,²&v;X'ÚEó~÷:¥#AÌ’Ë¥"só×"mg²Ã°{¥ŽQFiP•÷ó_éœÔÂBÔ=§Ç;G/žª§¨š ÷³rjÆK^•¯|÷­g~ãfŠÙò¸:™dE‘ƒ"“mÓ×.(¯‡É9köˆ5dܘÏL$Õ¡Õ@ªhV4¸Åß„„ŸŸÈ šZix—§þ[ïBÓ#o/«¼ý¶69–iת(X i´×bí¬OzÁ˳^žuà´×Ñò^î.KG>UK½.ðBß ¯VÁ‹ŸIžGÃÅ Ù¯6Û#ƒc(¹["+f–»ÂéË1cm1J´rËÎ(>öÀ[1øñÝJ¢Ï¶¼.·0ÓZøˆ:/ìRûâF£íöB½yݘRSœ×?¥ZR´ýÕ\dƒ—zü “Íìä}Þmy-+¤¢$K!Ч«Ãš‰+Ú2÷|¹ÈfXŸ’ëömÄ¿ˆn•]Yª•~cH ö3_RÄ×è¦Ý¤%t>i…ßMã,àe´±¿ÓÛ_l§³ŠÈTÕ²czù$Ç%C¹EÙž÷¦eÈ…ê?_@ñ¿xá”ËimŸ¾ë5ÐŒ "°’ÃYÞ$Ã&=ï:ãÝá™&»uï»q} _?ž©pžYÕæ!Õo÷Y¥öÌŒÁÂ@ñècº^_¤Ç9ÁÊ Å& =•xÎYdÅËd6Qs“‘¥òÐèx6À¶áaƒ¿¸­@^¾Aá½)A1Îå37òG`õ쳓â^Ã÷ï˜DW‘z9èŽ-¸wK\{l„VycTÏš‡Ô‰YB¤!Ÿ&Žã$Ã7íš$vÕ/ž1959Í3[¸0Y§‰.suå®>pi <£ž=M5Öáõþ@FMËx>‘­4ímcðv°Õ'ôëÜ ïÑ-–ðŽü~“,Ñåþ¢Å³ãíú)é²é•:ƒÜ/}ðm=÷ô±%1ü0ú Hó>Ï^DË:í;J ÏóMw}ƒsÃÐkÓ&yA-ŒÂþ{®èàE»3ßÒµ‘D¶êP‘Ûúzc&IžH%2Qu|ŽIÚ–­KA-ÄÌ„/ ³ZB½¥ÆcxrË«^É»ÀVCÌå¾/3l¿y;PHE’FÞ±tW(Ú•èØÞíÂÚAmMG,eÚë¥q„P~qÚš¹îfY:Èù9}ïÁ!!Öc…:xcÍ ÿ@êâÃ…|D²€–¬éÃÀ™gÿÔ#Sšá-¨È¿Æ?KP—ï ¿eV#MŸ„©òˆ7ïÉ5ØH•ªKˆÕ ©Éx±õ÷š+ÕÚ å³CAKGû×’;t‘ioܼçÊy[Ê=~ÞdS⫊iLuÏm7Ò·d¾þ+J\LŒ6äY뽜tõlÐâµ½w7 šº²œV´¨Ž*ÆpNÓjg¯eÙ3¬íGÃtK_×à±}¨Ö%2d–Gج”?é”Ç…l°(™±Fáç]„œÆžIQÆ 1"¤¶¶XT†Ü!µÓÆ5žÿ„¹\’Sòô8wãÄ_ÉR%xHøa´ä´~â ^Àk® S£'¥6ßL²ÍO)¤ÇjükštwhĶ5œ¼ì3L©>ÊXh†9Ÿ¨ÐöÁF¯<3º †==רÙÛ³™‘ð~†! @•EÚÊWsÀœV%Ku¢ò„!¯L}(7(¦xïˆê­Ía%¯D¦¢6Ã/<ê˜XΟí„<\séð<ñkÜ}_ª¨ðÔ²3ŸÍØ$9%T]˜¾ö„{|jp¬ÜNñȯÑ›ŠeîYÏ{ Ý’m´¨Ö@EŠÃÁÐGã É5£ðš´K}íOšòÙN•}ÎÌŽJS“”‚Œ¹õç«Jâ×ÒÊmVìvs®'£©è³·Ô:ñUØh‘ºG±õtº‚GùÏ (m4|y¤†H¿SKêºçi_ S^„?¢KúÌ™ghE¡¡#WßÛÿ@K¾rM_óhº3A4-•t—¤gø8|•põ'?ûgÇGÆY ›"ëQL{c>{Ù•"ú#›®¨7¦5Ä­žáÛê&«©25´$wšö±~¡¤pÑs]';ÎW3£Œ÷³›$ÆåL.ªýúÀŠ÷,ÌÏ@ðfë­>'¶4ï é \‚ýIÙñîOgÖÀe[ÍÃè7¹ X;†±¤Ûß×ñ:¨Ê\dFÙ)Îc ÈEcg›1 µí¹¦“'ßEß«VÙ@¬QÞ©àÐ3âš{\¾'æ£Vt4ÂöÎâØç 2xd¤q9*èi¬îãå?º¿œ²ë3Ý€¾Sà(g»b$Yé5Å?Ýa0©æ×ºfVÓV5Áo}é–{…€!ƒ 5ŒK3ޏ¿‘™r¡×ÓR7ª+jø¼²&~Æb„áKè ¡yë»æ¬ßؤ¯Ebä´`¨nQŒÐY±e=IÛ·‚+ÞÕG‚¿”_’‰ „w ÷zh(|l—MlxvÙ­UD÷I õ#Ó®ÇcO59p–[3=™9)y"êœj úzƜǭƢßW4{Å…U~ÙøýðîzPI€×ÛäÕÈJì{›‹£L…r¥˜<‘±ßÄšä(“óŸYü•i²G‡xŸ -PÖøñI>é- ¬¯)rŠY7våj.¹HœZß4g÷$;.œ¿Q§¯t‰š“™·*}¼ö¹ãIËX).'“Ïe\Á «"¦#á7°[¹A€c2› {)â"ë5uø-ó¶m›–8ãJš%ÎH|Ib¡õ(å~•¼rÑ]ÿ­O9ÕÏ»|Äêw¶…ˆ;dŽÇ{­­WÚÅx0`<©}á}eXÄÌ~&Dcþe¿Anm)a:N¢sU'š åXáßѳáÞ‰ဖÌC+ ÔJdz/Æo »vQë¦-YŽ!¾yøæïÑ*ò›wvúOéÓxð–Voúñ¼¡®eÂ4¬r‰—U4m›Ã—>ï¹½SÀÕaÎ'Àçùá¬ÅÓÏäqS9¶?Ò6ÑN=iºB0X~ƒŠïÕ1)–Æ+©Ä¸`°-<˜/àÇ.G/è­B3œ™€ÌN/åÓÄüE7¾Y q½ñuòÓ‘ùö£ü=Ñ€²d—e™Ã—M6§ªŒf(žˆÍX܉ ÊòÁ‰‡ò5w\(î½-JíãÐô'QÛŽUËJä³­Ÿˆ8™@;AÔz3Õg(»Äºü‚/æ.þ`'^øk޵g¸aƒk ·ýß)n†ÙìDqÊWºîw²„^šm©ægôŒ§ÞF¾5Ÿixøáúë¹ní¿Ô‚´}€þ><^Y•i7!?tçõ«Éfcì#Dèø¾N^ÐPŸl.1ë”…æ Õg’{xÄäG³±–#ÚI¯ç†?yJÒ”Xz *FPy1]€Ìœô#/ÞÊîb:#Q!oí±Õ>&\9Eµ6*+:Œ³¬Kß ä“-VÒX'e-ˈF=L…›Ëh$€ R,ôÐßïyµW=¦fwl(Ê¥8B½’c`óu®]p¥ª±ýâ-_b›TÙwuWþÎÂÅVèbºŠŒØO>#ÆŸ™Ãùƒ’kŠq2U-‰Ü;|–‹ÈÚB±V:©rÈú´Îà\í¸xÕñ—ÞÏ/ƒ}]Œ37¸~Ök.Éy¡¬Q6¾±ÙÑS—¨&r«T+fJö²Rx{ð=íxÅ÷e)ÎÛQóæÊÖ6·Ž@ô"éL¬ó{}–†üdU®éuã‡;ËûòØŒe×níQM‡ŽW$1Ò©9g³²tck‚Ü%(±6‹^q¾Ê·5ÒÎíu5`rÒ•+ ÂデÜü™M.4ïó/<Ô<½q6¡I ³—Nç<«Þ 8#‰ðŒ¡\²à{°–¿Š(3“D †Z?lõ4jþK:ƒ]¾gëûoÑ>,Ïn&E~– y˜Üà·¢€7öHDœðÙ˽ºë}ÔmºËÍjI{ùݼ¸º€9éÓõ­Ï^wÄc¬,Z‘~7ÿUª"xÐûëBò¯„•íŸ7§ž~ »w*')Êtxçgï%S{£PÜáYŸòÞáà <H¹øúÀ]ÿQãnî[¿cå+ bêÒFÁ¼ ïJCaŒ´:Ò¨Ô•0n°©ª,±‚xñ;@ë–»§Å¤yæ,ŠÏÕeûJ:›ûK¡zÃÙ‘¸÷á¡A³@ qhGÎûÅZy;ãÄ‹8mN[‹]¶‹ïT%{ƒŠÃ -â-ƒø_¢–mÕÀó|sT•ˆ Ãn§]mv›áU§àOăŸ^z6ê8-bÅÞ|‘Ô÷GД} h öi5ÑÏ5zÛͺá5?²éâíööy§š=gðg:˜C!íõ u<•8Úi|ž+#kS€Ž¹¸¸[©ät'ŠðdîG õ-,YíÖÊPÔ»×̬³çˆ!…F>ìT.D$úŽ_u„ÍòÿCɧŠûº‰)¸¬[ ⵎÕ]ýg Sª«§|(rº¼†…sÙ:ûPå„Zú½‘à¾ñ‹–ÏHRîйeá ùç–òr¹KQëÕä’sÈØëä=òLg× V>‹¹ö4-M–úz"°–î»tñ(‚7ùø'eLêÞ]òû$Ö:~ú× ’ês×ÒŠ§kê4,µûùHs&><&³ÏÑ<“þ6Œ­¯¦†/Ë]ƒ°ëÙ1­s‘€WDz †"=½ÝðXvŠö”H±ðÈÙ©ØF¦Ë±?(® ËŽà󭢞j𥳮¾ñEÝÈ èQ™õRtvBQ¬Í½?+{„BLJÓYÞ)©z¸Q¬Ök½û!Ñ„MÏcg>~#«ËOp„{Ú¡än‹‚ Õfΰ\ï÷®™B­p9´i' PÌõ6EOŸƒ§ú\Ý•’¸„éuΧ ¿NŸ½²Æ¼q²dûeCÕŽl—Ífko~ÝùŠs9³³×0ä‡ú–  #¯% nÞ0··&­øè§lô8‰qsÉéÑ\ÎA¼Œ~î™ÛõŠùmr_ÎÄQ7Ci 6ÿä‹‚ŒË‚í¤7»ñå_ç7”?õ’¼KÀÆhL7`ZWÕÁZÃát¸òQf·¾(5 ¤éþ>ùù£þˆÏyêMÕWú4B´F˜ÈÈk`J¸ÏðØf›6=Û‚ë§Ub¥$4^²vf¾¤GMaK7vÔ4Ñ—"«µnÉëâKŽ”ªQmÑÕPYßAÊ—æKž–÷8%=¬¢ß•b¿yѱðRH7ÌQnzYD:D:øÛªö›Œ8Z³çž ‰Síw\Œ‹¬ÅÛD®“}ÜÙHT­ïUtBA Pê-·Z–‚‘¿OV†eÍSæíНc<88!˜Épž£èj‚¤ýCŽA vᆕa«]‰J^7”¶sÁ­Ú§ Á[ÔY²Êó×A¶NÆáUWO½Q‰^ú%É®L·ðÛÊÃä¾Õ¦ÍÛ^à´ñƒªa.6(æO£ÉÖrôäecû¨W·®C’<¯ïµ³ð*ø*Zülå¹P¬ÏLË¢Ùä“o*_ˆgsºÜTÃ'ë~_=ÿø-€¶"./_Qh…ÎÙ&#pu¢ÙoEmažæ÷8Õ­Šî\Z×ßq1äSvÁRâ³w©>\Õ` âÛÁu·Ir¶½üúHH03¯i4ΨC…Ç +”9žt äÈE¿^ˆlß”lÂeg 0.ñ{û(ut±§Ý9é½Ãý†šî·MjçEïrióXc®mVÙ)$}ÐÑ4œàÑ i7йq§ºî#J¨Å£¼—w4ñø%;à %ª ½®’ΙÚS’g^pÁ ¾ÂÙ²ZÛr,bÅg çý,ä_pzŠ Nži4sKëæ¸&–§õ­å 1…]!éÅ6…º¦q~´m`¬mJ;º+À­íÁäïñäçó“&ªÜò{uer«m`íÀIøW‘qÐîàT¬Ǹ•cïvŠzmkûŤÁm÷o¯P®aåmBÈéÍ=1cÓȮִ8À²}iA½†­TtD¦ú "™sþ óm}-1‘—ùíoo-àJϳ­ŒëÙ0œOãýS×¥"…ˆ–»zýØRñ´ãñË^õš¸ ~äaXòqJ+a…P1ñPΫ•µ)4-&Ð÷¾Ë¦jÂ9 F 6˜»öVoqéå²J²á–|ïã wh]ßiñQå|úÎZW'ÿx2Íúç«]ÝÞ† Ö‹;Ç ‚”Iµ–ÅAóÁäš Uðp:\AevkézJš%D‹þWƒöËÏ7vÓ ˜0¬+Œ6ÁÞ{¶åÍq?Ô»¶HÍ4ð=›\ñêä:˜:eS…÷ m<4ÞZÄŠ=S¼ßУڼ#tÂy®jðŽ),æjÈ3c*l;á–J2Ú¼vmvȶ tf™Yx©Ô9'ÖüžX»`ƒéQ_$îC±Œ(Œf6Ì;ä Êå M„˾^F+”3iÚ÷¶¤ìgba“À> aÖï-‚J‘VvZîÛn‚tÙ¨‰Ã絺¡Àö“õIŸ9e¤ Ü\ÓðÒ±ëWòT•T÷n'¿ìñwî'Þé1÷£Òã¸ÏäÖ OV(L›¶]À½ò 8ÝÌ]1„~SЂ!×Ùw*½r+N‚¦NÌXŠºYÂ:®×Æ/\¹+£úOèÚç>†=è*ÜêuN4òà?4È*%ïÏJ%Ï&Õ0ÒØÜ_ºâÈš$ë*MÂI/*}‘‰èRžÁÿÜòßïò©³pžk#Õ¬mù4®6çSA¥º7Nžæ|ïίù%’ËæM„M³ ˈ¾«˜}âZJi”ÿ&˫ݔo}ºI”ø‚1ÂõÛñ{>s åñK ;“™Töø[`©0áp‚Ž˵Չª§xíè|EŸÁÇ}pˆ‘¢cëÈ“‰.ˆE;Ÿ˜Jil9ÅXâKsóu*ÑÕ·E¥FÞ?í†Þ_›Ï¬›ÇW—i f·ò-l3ð×6$ªžRI5ž\¤õË÷<±'GŸ ÆÔÛMÜ:Œ(öÕÛ×5°ó²Ä=mý¶î×^ˆéñÎElÓ+<ƒOe¥ò[<®I?‡j‡v=¥-Ød‚U~læ^_ÆrØŸõH &¿·ÙO—ârD4êì·4ü¤Ž^Xîón™Jw‘1Ä]®T |~þ øÒGq“5àý›þa.ì¸?ÆÎé;á…Ö‘ §œ–R¢Êœñ¡—ÍÒH£Çm2$TJ`Ág¼² dDDDgÒ àïGe·Tq,^e¶¬´=ú”§`Æy/VF‹2¸˜d±Õà·þÜb`<3A!ìc?ÇLˆcS¹Ã×¢T@~vº}ï Ķ$?_ô¥»‚¬A9ÚwÓ8–Úø ®ÑãÛ(¡E9jáe_øÝWñJ¾èá‰å_™/=Cˆ² <Ò9b¾?SgºÆ ß„ðXV}OúØÈ5š ·µ “M("ï­ŽßTï‘5Š3O+;ÚÍŠú SdÊx:n2é+ngšÒ· ¢5,Ø[âämš«= w1 ºÅÁÄe÷¸öhd+p•Ev woM71ËÆ>ÚÀo! LjrÔøo‹VGà$‡“}¢#yüì£É©ƒ¯›iÈ«s†ábÉÕ¥«KÚÊ램[tž|ú]bŒ†Y`OÐC›Ÿß)¶]KÀó*#`­ ÌASHÖ̦zÖIÒÞ» ‘šo‡·Ïfpš,î¾`]­øºt1ƒg¤½õ¬³q~¼²·¾%QHVÛ÷\÷±Tkòv(Òb ô0•«GPòý´Ö%CKêI­¦-Ii|r««–RŸdбá2U½aXéc÷î\AÚó3ÔÀÕOwûÀPžÕ~^ÍL`åþ“IH9pƒ0û¹'~¬Ó¦çNØ}ÃHn® ‡#uæÙϨº-æéÇ}V4aqæ¬:éW˜êótd8²?¹”V;—­Ð]O,®ûg®»²í»Ñ[wú3¸n§K´N-*ÒãôÜ3È€ËѼî î}l±]hÛ§ƒçê#ДŸ¢²Çnòd"¿È&ÅâÈlŒÅ¢½NwçÁ>ÜãÁÇÜ:qÂ$rã1®‚JOþÈ)®Ÿ½Ÿ¿¡~ð…WNÒð»´$-ÒÚÖº¤f1ô‘AÞJi˜*—áµ_+·æqÕ2wH&r¬Ï¨7Éþ\‰ô“­{B5³.Gùó±œ]‰òÀÈÄ{'èZªiDâÛšùWÛ;Ɔ°;ÎæñÇÄ.Ò™4ª…/j ‹¾]mJþõìü endstream endobj 44 0 obj << /Length1 1418 /Length2 6406 /Length3 0 /Length 7380 /Filter /FlateDecode >> stream xÚtT”íÖ6)ˆ")!C 3tIƒt§RÃ0À3ÄÐ ‚4H*!%]’ÒRÒ!C(ÒR’"Í7ê{Î{Þóÿk}ßšµžyöµãÞûÞ×õ™tôyålÖ0e$Å æI45UE $@ÀQN°¿`" ÌÍŽDHüG€‚ ‚Â`Š&N‰¨y8À‚°ˆXT€@âÿ DºI!žp€&@ ‰€¹.>np;{昽8 œ°¸¸(Ïït€œ3Ì … š”=Ìs"âÐGBá0”Ï?JpHÙ£P.üü^^^|gw>¤›4'À ޲èÁÜanž0À¯ZgØŸÉøˆ€{¸û\i‹ò‚¸ÁÀ …!Ü1˜s8@_U íCü ÖøÀøën`>ð¿Ëý•ý«ñ;…"] 8Â` w‚´•5øPÞ(aó+âäŽÄäC‡ÿÞ/ÿ#ycS-eî?ÿÛ'/ôøñ‚¼ , ˆb^þYEÿ« Ðß©ª[$ úÓ-æšþÕ±ç_àøKœ€ÓBbX püMr30Šy€ÿÏTÿòÿcø¯*ÿÉÿ»!e'§ßnŽßþÿÇ q†;ùü€!­ #M$Fˆÿ5†ý­&Ìîáüß^U#9„†Ì¼`!>Ðî® ÷†ÙèÀQPû?”ùƒþ’šÓAºÃ}[0Y Ðù0ú‚:b¾î^þqAÜ1bCý^ã/†‘Ó?ûPB@‘6¿t' ,€¸¹A|ˆ0«ÇXÂ?0F 60ïßÌðó!(L 3sÀéFôkÍÂâ~Œx!¿àß$ à‡ÿ‡)àwømþã`¨‡›¦³ß|Átõ/û·úa0o”hf • q¨ i9©’£óâ]‘ÂßM=1àÉ· Dõ([.&êgf̪—*Ïtƒ•-Úµä]O²¾Lú­¼e¬õ:æeTþfÇh?u}Œ=‘ä÷“žyŠ´+Ç8Yþ¾D‘[–Cô÷7elìzô¤þì5]è$/6ŠbeQµÐ­¶÷UÅTôâFÛKz¨&ë>ƒUá#öì%ç¸gË1f‘)Z¯˜Õ¥ÚN‰½*ï|íè›'ëfèˆRó ^Ýc C‘)IöÑ+ TÍç’\dyn½ºëgý\^…óñ8gRqüjzñ^`xå€óÚó5<ÉΧ[¥rŸCjDH]2¬t‡>¬~[}4¯vÄx=@ƒt£%®BpO6ëþvþJ ˆ¾Ò©»Œ`¸Dÿ@ÿ –8zusšÑ×P –éÍô@%ÑêPC/vÁ6™Èü.ÛŽ “͋ǥO¢»ì’>WŸD¥Øm=º2°¡–© ࢉ”¿püTÜ”~úQa:Ä!—«©‰nλaùœ¢‰_\\?S¯ñ–Ó¾m^5ï‘y0ÉÍ7— åJ†=t—Ý–©ŸØ~fÌÜâ:søIQ~š@ü:ò‰U3‡c‚»ì7ÂWñ¦*õßJê"Úü6J¤ß»O ‹ó³Åß~ù9š¢ËþØ™osk´¤.’½üî*®k–åÍ ò ÀÐ*‚ø¾±ån«$χ| ©nKÕÄÒ¨û•açY qx¸OÉ·QF½| öOÝäÛ”R ‘‰’ÛI”­÷Iße¦kõú©žcAF©¾ë™Ý¢6©P殦ð'^p€å£uõØoCmÞ”š¥=«[ðSlt÷`çåæÞ¼ýC˶U+÷ …#?8oPœp ÷‚ÓÚØ—ÙýâÑ¡xeŸh<ÛO]‚zJZ6œ«µQOú“ ~Ë$¿szÁÙ§%;ƒŸ"ÃÈøØ»¬‘báÃÄÏvÚ¯Í8¥öïFù— (âI-Wuy±Î¹fêd»C:762˲JÅevôÚþß}Ü>3èN­|’٥Ž?ÈJ¼¾'»lv7ÖÎp=B¯$tˆõPœfMàÕ,ö“ܧÙãZ˜j2–¨!¶—°=ý ·¡•ÖíyÉ~¼Ô}W?ôʇӧE±A;üÉôw)ñâC(^ªŸ_n‰uû l¢™g?æ\ÝäNï}9+QÞ˜û˜Bë‚,ÿßYRjA“vÊ’°½KŒ´Ï J“0öªíŸg¾~SCÊéš3®Û²ñü¢d«îU¬)2Zò-ZY‡BD¥F“à¹óÃG&×⺞86®Wë¢ K„ú2²ŒÍ™s­“ÀQP­‹(ÕK~ ï÷,ú¬si’·MélƒÇõ-͇“Îq5ŠNpkM‡ûß3µ“jðµ‘(-Â,s²úÜ ‡ªÌ³,Â.Na¬$-+á,Æ]Tá[—GÓ=¹F\ä´¼TìÍãO/#4»lFÙª¼l‘ÏVê½µôN«×έûtòi”´„O‹ùZeQHõ†µ„‘ë¥îÝóãpåS×3ùÑ~AÉâöýeSçò‹Â„ÞÈäÔ‹ˆn X½÷VûäzöaHÎcFºV­Î[óÎT׿¥g¦n›C.Êúï+m‹nãüÌà )ª,¶ •h`ݧ§¦é$#ÂU t¥¢´ød…5‰oTëmÉ;Œ·âbÖ3[pý<ö,tlÏV¤oÒ±ô“RõsãÅÙ-œ¢fðˆ÷öâ‘¶kÒ›çI&S•Ÿ{³ø:õGó..Íž–W,MË¢(Jˆ×û…¶º?m§šJ p8M„Ïág¨ ¯ôr/GÚÉî;ÛÕŒWUánD脘°‚1gnLJpAÐÙAÄð´HšgÀE¡€|„4 NOñøÐ¡ëàþ“e©òrïa!«9d×HŽSeûK=´ÖéLê{ßêºT.¹©:€¼…Ia)ïŽ2ÇÝH¢KrŒ¹lð%¨ŒÉòÅ5ZE‡PðWÜôÖ-·+š¹ââêò4ÄY‰fs 6áËJÖX5ý´’x-ûÜÅ‹Jâ63½¯pRŽûLx"ùYÉã÷iZoŠ+ téäVß þÞ(ZÔÜÿu¢¡Ú…dU6g±,u}‹{"¶(6²ìIû!ɲ…ñ;êÔæ–TS]ë‚AWÕ6V®j1ÑK·ÆÁ&Ú*šP£|ÙÆÍ gÔæ6@¶Ê^/ˆÞ¥UcÔÑNÖÎ’DRf~Ø4¶v½æ&ñþŒãcH5ã.-G± Í‡c]ß)Ïd£IÙÕª®>›ÃT«7uj–{‡÷X|‹*]’!´»W—ˆÔ«)8BèP)µh‹;åI–¤ß“Öy0PøphBHô…,RPõQš!9îä‰ÌöÃyjOÜYŽ÷ôÖnáOkÎj-ž“G‰Ÿ4‚5·9#X®¾Æ•ßí\™8Øä.sv_“(] Á½24¤Æmí ¥çkJŒòŸ“³hFP…b #Þ& v»‡¨@L?Äyƒ6¤ðFÕóÇC{¶ÙÔ±䃖 —U%³±+-ÔÁ vºµfôË $ñ¥º&$¢H%g‰àbQSš©¯Ïáìò?[EPïr¢?Y}m¸p䒱з~Ј“H•»TðŒ®àÝ•#§Xî]ÆO½ã·×k½,ðïiÅIŒ¯™>‰Ú°ØåIï„Í®fƒß>Ù=¬8Ò«ú$“±=ééêq‹Â_®aºÉ·¹WkM±ŠB˜’΂¢øüxXµ)çáëWë³°~®گϬ’+§SGi2?À¯£‚% ŽZç|Ep£„Û_¬Å”—Ž5Hâ8gOë‘öíIì¼KúQsxÐ:’‰V sþHg“>¤¢C8Qï }VhôóRwy?XŽ] ·çév»ƒÊh™[2ãÜN.îôLÛã«8ߎísØñâ ÎÐ ÎTsE:/p$’Äßh5íN0´ï—&â~_ v_Ù&‰fØö6ªŒah]âAß7;%Íè¬jkÝD£ *6ÑÇâ®sdºHÕï—uFÉxqšæî¸S@cùÅâÓKÿjÛþ%å8·ÒS¤zFž G‡x{7±‡ü»Kx_¸ ½î\îâ–‘>Ï$”a'X¶k÷ES9Õ'…Ÿìmò¼^¥yËÒXøªI®^A¹<-ÙÔ9+ÔâÖ¶l y!Ú}ª9~WšÎ–Éœ=­Fözü“‡—U4Í“ºT±ã> q-Æî3TËÄ]UKl‘5AªtdÊÃOÕvÚ`ƒ‡¤ßž³ñÖÀX^:©knvÏSí${:±³‹ì§ßŠ~+v’M¬²Ð&ŒMÏ3¯‚OG ÂL g½‡VV#ø}îÙšïmJŸ`»Z¬Z%Õtb‘­=a›€ku(x…×±q0Á±¶.›šc|Ò¹á;¡ç çuâQcAi T¹æìâ$) Îg½ƒýåå3Š}íŒ1mÊÜMÙ„òöèºwt+>d ~àåž²!‚ä®gœQÊêAOÔøNÆd„|IfÁm/–Yqïñö );ÒëT!Sbx9e÷¢N+ééÂ\t>ëºÕ)tÂߤõ¡Í]åŸ0²<£ˆ¨¸ÌzaÖ c`lÀÞ,Lòaó wV!Û73&µËÑô™ò=>½»cnþ7ä>A­8SÈ Z†Â;äcÊ’RÝ‹º;åÛ±»–³ŽÕ¤Ù¾2GÊïߤч¾šŽ:ºw PX1hÕYÑ0¹Å~bC=¬ÉéøaMÈ©sLäåw"óGmá7¦^A[Ï©“ n[Ú¿_°¯ÆÞîxÞÞ2ŸÀPÞ¯PWí8ú ã±9²ƒZwSëR¶oÇíVyšF ÅÙЂÇo)*„Ð&½bnÃæ"ë`.%jœdܼpÇÛ s½67¬xÝ—k¬6ðÌÆ£Ù?\m î´2-)×N~L³­“oœ² íž/âàÄz[™Æ¬ûü+Ðú£,M+w¢…NƒˆÜØŠciwÀW¼‰‡xøÏfU¸Ô ÌVz³–Ùu(’þžæ‡XGô…„[~š bŒŠ\5z<'+@Þ²« 2½©/S–£¹:¾=ºCŽ« x®7&49Pb€6‹uªÍŸÄ5fNêqÙÏqʃQ¢,æ+[¹¥w½µ^Ê)ê8 ÂuëäªâÈ fÄ& {¯F‰Kåi”¦šBW-¡øz±wÒ<µî~pRþ’nÈÈÞ¸âã¿6 „•wL>#HwÁPmªŒ|mÕiîâNEùØrÔ¿ê;§HÜ&‚žÿéMêö×pÞ›ùÕ4o)òºh÷KÅ~l ý6¦NݧËfx;¥¯êÿí¦IIŸü—y—Ýa´¾Ê©>ÀGë1»§vZ³™ï ¨Ïø4”x…­ê4Ç,›me§GS–íä$-ª9ñ—ôÉîŠõµ µÀH’áŸ=‡g+Ú¼óÜp6§¨ûpT+ºfä*ô&MÕODÀÚ~Ô׳ǣ+¬> ½Ù³%ÜÑœ‰øÞ® õ+AÝ—ó§[Q呯QéIñ…-ÌwøQ!üšAIåSw½ÞÍþ8%‚õ3ˆrÛàÑ3>ÉýaÍ!2Ú!3µqÈ=#2Ÿk5š´À›ÌqaÙkQÃ(aå7,Sx-ßnxN0MÀ–ÓŒ}cuþèxâêæŒ¸J·«¡ÀÁ)ï8Zø´ÅfO±þVþ=…Õ÷z'žêkJgp£/Lgë§MµÕ’®/Iï“®¡â/´¢Æ¾I‡);Oª;w íL][BVwÖãbõ;è“köŒ‡)¤l'×áÚ-C;Ò&ê±»Ì=Ò╬‘™ Ÿ÷Ê%—o¤¬â‰”„‰•«:°éiäœÒÉüƒ¦Ï¤J=jÛÔgévïm3?¸Àãæ¤¾|×l&ãÛûIm•qoí…‚oâ“‘K7h£&NÞÌ vâú±ö“6?9/Nƒ¯Nò ± ‡íöŽÔ³= uŽ¿9jPósŽ}/æãƒáÎâ²Ww2©,r—ˆÏ×¾*/ÙÕì7ú™œ:ÕÁ¦<ù½\zÂü~ìlY¾ x—ž*$XǨÜQÓ%È1Ç=L¢Se)“ÉÎ{Ù“«µù0 Dþ\'R—K‹uzúæ¢cv†˜¬S5}›7¾;ýéΡÎÇ!f|s‡ÔtöKµ&Ÿ—ëø,4S‰N­æ*&dã¾píóßÒº¾‹¶uôAiTl[»›T“æ€B-óU(÷û¸cïÇÓÜN+n íRjDñ?»ïêäUv-IuàÁËËÆ§r¢cÖErñÔ¿gzó¦à{VÆÜM²Û—*¯b›u%òú³(nM}oPü^ZR—è!à~pÁq„ŽØ 1ùy‡¸¬¶ZpÍkfv*¬ê†LÄÀÙ}5;ÓevO{Ì@îšwý^Ñ3ƒJ22'ub~qé\ŸC»¢Lˆk2 )R^¬Ñ‚ƒyÔ:J"´Ïôjß>o(hy*$“8 '¡9 i{ °Ê߸á€m$¸œSoê¼ÏâÁÒ¥a5ß ]$W*}“ b–s¼Ö¢Þ©/Ѧڣiv„j»žÄ0G‹Ñ±qªÎn&Ê&¾yS@äÌñ ø±kC ë£û« Kþ„™¢-™Yf¤¦ÐúÖ|IÙ˜Z@åÈ$z-;*9’ÙZ„C+ö:Àƈ ~L·H ä —Qäñð­®F%"mC¡|’ùcߣnÑCNù£+]€}ëNů¯çnry0ÓhÙŸŠ¤ò¬\ÒERˆô|)Еv°=Ò#+úAkÐñ]c}§§h–¤–|ÂЄ±§È!‹£c­¤,‰º^¸± ê æ»—`…|ë§ŒÇy|¶çaG^ÿ*.Iž}‰úÌš)„S*9$œI¶Kp–?ÓBDÊ+|&EPéâëöþåº%M[Ÿ=öS¯a0¾Ô¢Üü÷Þòþ9±w–F¢Io¨`\›,Gû?uö\Ê­yU’,c‚¼kÛ®(Ä‚#öxXÒÍV ¡ÁÂDá«óXxIÖ+‚»8fvão4›!ÀŠÏò~KYÎùÒF@Ü—âÂ_¾ºâdÎüXX-)I¹zLžä/xUªÂçð´ví'Øïû…j€ì8Ö……>‰* ×™ÓÆ3ÕH&³p©U俹±.ï=^+ëÃ#4‚ÅÂÅÕ¢c*p³t)3­¢ïéȱϒÚ@ºìè„õäqyò]“$?EðŠrOF½ÂóP=èw•;Ý<îð%#íjƒ2œ¹_è§°å‡sh×SªCÓŸ’¼k£Él)…ç¢e¨œŠN ßèîŽx9ÒJÓàæwjº»¬;¸]îÚfB1MT³›îuôqØÂxË«ûb¯TWÑœ&½ë‰¬ØŽÌWÆ/ 3)…;^©3c +\ÁSR=hݘœ”yö¤*Mhtm½–[^šHÜ:Ì~rzú²ÿë`ãÌÕÇmDÎ$"ºÍ¶ÆvÈ”M{·Åè¶++Á=ÒæFû{{bSSE_úžaQ5ÆMs¿Í<],Øò¹}¨}ÑûÔˆgä Øz¼ ¼3ƒk.wŽ¿x+Là˜j¢,Þ@޼<¯ edàôöýl<„¯ŒF2®tØíGÏ"Þëkn,µC¾ŸœŠõÐ÷0;¦Ù)hLÄ}BŸ3&½Ž»?êDÁ’±™K N•$ÿTÓ¡Õ$è}à¦àxùlUßñBé^s •Á–{ñãºzòáÙì1z­ʅɘßâ·Á»EÄ·´uCF ÷² ú©6,ŵ§ÈÇæ;Ì{[‰uA] ø[¡»\;Tu> â­ö€ï©[;7®JÍy0™-Ë_S,”$v4MôEé{±° Ù*èÉWŸ¥ãn‹ÔÙ¯œtÚ>‹‘"îgôÓ±CÀŠ÷x >šNø¶±Óç©ébáBùXJ/½‹²+ê¶óJžJ ð&p›ù›l°Æwµ‘êxç bã4¿—b§œé•¬†2ä¾{ÓöÓ&-Ä~ŸÞb%˜ëðî865gÊeedéß ua~Ú&¨rƒà–¾q“®®GvÕe¸õ-=W,÷çcüM€Íþ¼k°)©QƒÒòŽÚâ±Ðá5rð©÷ÄtB<éÅPR{¹ÎªåÅ ü8î˜×£÷n56&utŒÊ´&®M²ÝÜ')õ 9†×¹¦UÆ®d€8E&½L.),<3;›ï ±Œm³0ÅÁÙOw›U†ë»÷õ#Ä”`Í­ŒûѶ%iãtLn¹±+œ`Rþµ¾ÜË!á’|ÕÇ  ´™îƒPSVBï¹&ò”MjÛÉ€iq¦¥¹_Íú]Bß³­’<ô£Í5oŶæƒ;Ò§¼UÇLk8?¼Q™­ æX©¢L”|çiÓÒ(j˜ô~˜iƒÖ¿_ÈJóÙÝêíL|>0®ï¨\>ù5™ÄÂB‰ —ŒêÁñ£æFÊy–T\¶’CU߉³¤Œƒ»Ú5ë÷” i],{/Ž+3xÝhäÍ‹67tk”_J:qeÑ[§Õ4ˑՑ®ë ÃÈ]Åvv!‚öM)|Ãä€@ÐØšTI›r õÈûñ°Ú¨àaƒ±+}>+Vé‰ßy~‡ÉÜr‚b,ÞåaÄܱ±3¼"G”÷è~ã-tŸ¤ü£À gGV@Dôú¤Ç†ÏÐÔ‚³»ä˜ìO5T¶®&˜¿J£óƒ‚Šp/þ¹ÅHÞ6±øÌ›æÄqÔZc{¤–ÏÞkùï)¦õ дñ|ÀGcæAù§ªX¥§µ>D¢©]š…–·âZýúD}Ÿe¹ð[l&âœçW¶= ȹgnƒ»ì2Ó®W£<“Oó F¦»²cN(bí_"E IN¬¤w–ҋ'¹î„¾g–¾Àºõü‡9«Q'þâÐÓ䬚´î¼œbfŸNÊ÷+YGµð@…°LñH5TªOǪþU¼[À6XÜQ†~îpþ„ôúü¬Ü‘=vMÇ£½›{žpïëÏ^àgd'þøCsS endstream endobj 46 0 obj << /Length1 2109 /Length2 16826 /Length3 0 /Length 18106 /Filter /FlateDecode >> stream xÚŒ÷P]ÛÒ Cpw ²qwwwww6|ã\ Ü-Hpw×àîîÜyœs%ç~ÿ_õ^Qk´Œž£g÷Z‰²ƒˆ¹ƒ)PÒÁÞ……‘™ ¦ Ê `ffcdff…§ P·v±þÛ O¡ 9[;Øóþ#@ 4qy·‰›¸¼Ç)8Ød]m,lN^.^ff+33Ï@¼q7ks€#@ÖÁè O!æàè ²¶´ry/óŸGµ €…‡‡‹þït€ˆdmfbP0q±Ú½W43±¨9˜Y]<ÿ‡‚šßÊÅÅ‘—‰ÉÝÝÑÄΙÑd)HCp·v±¨ 7 9à/ÁE;à¿”1ÂSÔ­¬ÿeWs°pq7ï[k3 ½ó{†«½9x/P“‘(9íÿ,ÿ¯zÀ¿{`adù/Ý¿³ÿ"²¶ÿ;ÙÄÌÌÁÎÑÄÞÓÚÞ`am (IÊ3ºx¸ÐLìÍÿ 4±uvxÏ7q3±¶51}øûä&I€É»ÀËs6Y;º83:[Ûþ%‘é/š÷.KØ›‹9ØÙí]œáÿ:Ÿ¸5höÞvO¦ݬ½ƒ»½÷¿…µ½¹Å_"Ì]™4ì­\2âÿy7Áÿ±Y]ÌÌÌ\ÞŽŽ‹w@k àûxog7 Àä ôñþ§ã< ÀÜÚÌ` ´´¶‡ÿÃþnZü ¿_>ÈÚ Çü>{,æ¿~þûdð>^æö¶žÂÿ¾_&5Em 1º)þ¯OTÔÁàÍÀÎ ``å`°ü5d\ï>ÿK£lbýïcü#WÆÞÂÀó¯Ó¾·é?'vû÷Pÿ{9hÿË¥èð>µ@õŸ!×gæ`6{ÿÅòÿyÔÿNùÿ7á±ü¿ ùÿ=¤«­íßnê¿ýÿ?n;k[ϼ­«Ëû(8¼¯ýÿ Õþki€æÖ®vÿ×+ãbò¾"ö–¶ÿm£µ³¤µÐ\ÙÚÅÌê_Óò/»Æ_[fkmTvp¶þëµ`x¿šÿã{_-3›÷W‡óûHþí¾oÎÿ–”°7s0ÿkÅX98& ‰'üû%¿#€7Ëû.š=þb£½ƒË{ à]žÀÂÿ×rr˜Dþ2ý q˜Äþ n“øÄ`’ø/âb0IþAl&™?èEázgQüƒÞY”þ‹¸ßY”ÿ wµ?ˆÀ¤þ½sjýñ¼#“?轂éô^Áìzg1³™¹ÚYؾwâ?fŽ÷3Û÷®ÿÇÂþWàûûäí_×ÁdþÈ`þiÇ_ÈÉõ}`ÿ°˜,þÀ÷‹À¿œÖÿ~kùø^ÞêÏaÞoÅÊÓÑ hÿˆwÛ?óß[`óø.èŸGyo‚Ý!ë{m;3Ð?ä²¼¿"˜þ ßOëð§þ{úû'ëîwÇ?nÎwô¾»ºÃþ.ÇÑÖÕù¿¶w 'W‡÷Mú{aþ0½ ý¾«rþ|gþü>ôL.îÿp¿‹vý|íöø^Òý Ïöø|Wèù7üŸ2sÞ¿!¿âÞ·ë?øïè4ƒ_œs0ã þTÜz_)òÑawL`šbW+™†Á{Ôæúˆ ó¦"=pt+ò}¨ue[‚úFx‰øÅû¸©æKs¼JËÓçg£oª“»-ð ØýãùÇ"5}„p êÂ{Ÿ_œ>kØ@4wÈRd;¹r#+çbÜ»÷JyÔô•,„ÎíªìUpÊ!<—L1DiDêü˜¡È1͘Å%…va „¥E¿ð@™¹¹FÏ#–ýFïsÅVà­»Áý0ëµúSÕ¹O—â}d’Ò[ô QgÞ»¨`¥Á£‰¿€8 ‰>i…•ñ€5­ÂZõ«}CO•ÛÈb;ËNöw@0ÕÇ~ŒøªâL#©VE‹ÑW —*6 áA»…èjë­f—ŧ$‚%ž©š7À ’ŽÕ±c—÷S­ÓêÐp?Ã]hRóýPÃðv—Ö¡>!KwZ +½ˆ•eÞ4Â$¨—ÑZ@LTú%·ïÜÏP&!üÞ¿EñY¶ò™™u¹¡zÀGÎ8²?ÿŽö)«»PlߘA›·uG7z âT|3íHº¸Èdzb“0N¥ÎI“è ÏY‹¡Rá«ÖÍ•0]l”a3Š{ËÇmq,ü /`—‘êÙ/ŠÐôüôÀ“Çx¦Y¦ÑË‘§B²sò•aô .`P4(¤—çÖü`ë¶´TÔëÐAóá¢Ì=Lt(zS+8'¼]Õ?«ÓY)EJäë‡Â»íÙ8S²SuQ7»†-uÇíÑÎÂgwúq ª‚E…>¹Cþ/Ÿ Ä zð*ý…±ÃoüY«Û”î©ÁÓQW£)Ò÷–!Ȍچ4¿â‹~TS¿Ï%óÚ뙟Ý&ù5¹Mîm[¦ÿƒ½·ÀÇ™aû’d~Û?õj]€:2âÕ"¢qMÃ]¤v‰ *m¦¸~28)JÓÆÆŒ)8v" È/¦J_“/ Ã|›‹Ïü´å•!üHÓ(Pî¬_½†yÃR* ß©N¾..¦q7±÷Ìù=kÊC*GÅ~ ôcjÞŽuÅ\B&Éü$¼<+N•^f*Úõv I¤Eˆv ¬ÙÚWX°mþY*ÌÁê˜ÄߪÊl9$Á ¼UhTh¯"Ȳ(ŒP‹½Õb²c`².%`”=Þ\бœUðdš=dzsÒ³㯾2#KÎ[º“Jr$ĸÿV #ÿ\ñà Â?¶?–„€0iì[ÏŸAÄÂ\ü6å¿åÐî¼¶áa!}mÈêb-;x½ìж®®²2MhÏXèŠ+ñÁh$Jï{û·‡2?;Ë,á•€Ëø½4Ø gJS£HΕåí)UC Ê:»VÚ¥¬zrÚçÖ—hîÕ׌ˆBÝéÜ_¿¹y]‰ÝD/ÍèN¦9·ò¼¦â«Ïì+>ó:a饂ìƒB–óÏ‘7oÎuÜ„9ýûÀP–ÓP*)žÀDB†ú”ŽãŠîÑM=ŽÔ°œ·I˜˜®¨Tf<¡ç25r„Uh«ó´fR9§Œ29¹íÓáãÏF†xœ ½NÜãàèÐ×û|X<¹‘yJV ËÓ ¯/w²#x)ô®¾ ?Y{q½Ö\ãR²¿ÌpÒ¦¶Ð"õu#•N¾ñtŒ­ õæÉ%`›Ñàþ¿ÛÔ¿°Ìr)£*ÞŽï½~²‡V?vo‹ßk%Ï6WÔkLË[Rº.Ðbõ >àv"‘7Ÿ*ßþèpì»Ã¢É9(* \uýxJÂ×¥“ãàõ@~ú0q7àOÑ[Ö19„å¾¹¶ÉjàPÒ³(7z’±ú…âz#jw& Ã[Ç‚ »h[<èÉyع´ ”D* Æ•·çP»t§,f8ž®¬È”òº/â9¤ŠÛ¡yi•y¶R|¬óÞQB '³@È»™ÍÚaAY×vR’*œÙõ 7h͆–áæ—Ù­˜R0/&CÑ>§ÃôæÄ¯Ç3YÜôdضQ(K°æðiŠÐ¨‰Ê3íñ—`Ÿq`”ˆeŠä­ßeØò•{ÿêD &k^І²—„Ž#…ý – Ìí­ºà‡ÖÕ«ŽC½c¤uñÁTäÕO‚ÆÜt’Úú+Á·6$Ï…ƒ7NÿÒì*ö;—LbXa%–㦠0ÉßÄ õlžDd)Ûf( óR}º/«zv?ßš½Æ#%ý áÐmKJI»eË š¢1Z¢>ûš4±+jƒËÖxô9çk?ÎUѱ± ÚZ$£ JÚ”3Úä!'ý=î! qôÆ´5ÃÐp®®U²+Ô•ßnÜ‚²àÃôqޝKøA ?«øÇÆxWìß±Á‘Æt¶Þø°w@•»¹X#a N6|éue·I/ïÕ’¸€26ìMM¾…œÖѰid$ôr%F¥Ì kúVnîÒ9¢ “žÐË,} aRÉG9É{õÓ§äoíšѶԶ–å]K¦SeQ«$%³‰³¿ô7_LØUXp*G|¥<[Nš?F qΆ6§¹ÎæÑiöÈ^ߺ`aÊòöóyçä3½ð ¢Ñ.J@¹%AnŠâí_¤¸xÈeä­_)ãE,?nÚÛrz‚‘…‹‘Tª»ä—ƒÛ=VÍWŠÁv ÷[›²Í¥Ü¯»×Ù¯¥ZÓZ𨠢ìÌe¼âa¥2YÃY/N¹Â™fâw‹Ásôz‰HÌf±Ã{2^êe{4m­º°ˆ¸–•d :ŠŒ3ê±ç©-°šº'Й\ŠóÏø¸Vâ#Õ«}¤)$D¾;¥|t\Žúùîxè:[Rð†«å£+ñXèYF+M‰Ruw›‡þÔD"ˆ¶ “g,Æ+Ü®ŒW®tgÛ‹¤W‚|ô~^Cµ“ð-ëÙƒÌÉÎhsƒxèZ,’ãg3IëËW»oQušž7uKh“¤d^¾¸I?Ý.»2C¿¥Ü.ô1‚DyÂɱìuÄÊÄX{]JÌ…ÉÁ&l¿Ì|•«ÏžæËë<š‰· |¤†FXˆT¼Š(çÛØ,0ýÙýt W†ú¼NÛÁß_˜ó}»#£¹³‡¿N*ILÂ8Dó0×j£m¨—K%~L²ÀA¹­wk-ȼ$¶t’­õN™gžçD7Ê™ç 8•ñvÒ hIëW1¡àdÐyúf ¡XÚ"1ÝPk5„’•ݽØÖ a|†Uë£Ú}5¤ÿD‰Çô-o—T©0 cxT«‡VÝ`•©Ðس$R¥ .l(A_å{÷GIÚ’kuìßQ†Š€[s£<ã üP5 õÅf9®˜ X:Õ”¹=PAÚó ÎMÆ©ŽЧµj±àëÝ‚ÎåÇkø1Ip ‰Ü“«×È‹i¹ ÎYéHªê:Ap]á1™Å м·ÄôN½"®^Êöçl«1û Ä*¼Cª™e“îzõžßTÚ ¿™õ&{#ûï²ò9ذԉÂðÕûÙ¿&RáËê ½å$ëÖ—/ mÒL׺í|±Oå+ò `—Mò5rÀ›ú¼ŸÃÐݶÎå×Wš÷rÙã¯Y’Éý0@v"ogÎß1»û¯.íu×!hSZŠï†o.Ò0M…½¿yÎ!À’ˆ‰õÆ1ƒeÔòÓ­?9'ÚcƱGØ0u ±tVN6'g˜Ô¿­›I£”çwì€DoD±Å@üÜä›VË}ðÈ8Ô›XOg16àåYäD¼iT-êU>ü%+ôšUÈm¦úÃÑUëÍü—mp“<n¼Ù|;øÄå/¸–î‹$0ð°÷\»’p:žj¾¿C²>¤”=Ø:ˆÎþŒ‡#NÌp#a~8hˆLD\]<+–X7Ñ ´¸Wö`÷—ÉÆú|£õ\Ë·,E4‡‰ªƒAÊ&zͱxõ¼w¿)­O:]ª}Ï’8–rjÆ¡xÏSâ(ÈÊ%Û"úrUçÌqQoI}]ì¥Bì"¨W¾Å=/åÜ×ÀK¬JÎcÀ‰IöhçYNc]‡á6tø©ûV«è«|¯ª¨!¡ª¨‰mDZGQö¼No5‡Â Èõ½³eÌóŒè–2”º~EOLŸòC6úsH|֌ΎÅ)µüKvæT´›È¡¥×죌¸Èä—µÂù Î2.êü^°Ç*%¢]ŸØ¯tÃ"µ+=o»ˆ(B1È#Ù'~M‚2ZÃ64 Y¶ÄŒ«óÍÄîá‹2&ªœ8-’ý²vÁ ì­ïZ!®øÖ/˜p`Û¡âø-¿4×oš¦3´>ã¡®à’)ŠF Ѱé†wÇR¶ê“ð¦GiM5³Ô0ßX3Ö²ŠÓ‡èâuª¦‹LT¾LþX§ö ;®.¶òb,ñÚ½ŽQü$x§oó½NSB;_'¼XÄÔè: ò”Yz|‰€CÔXGØJe–†ƒ¿ÛRÿž®#‡%j©ì2Ù9*rÈqk½#>Óô,/¬™DH'–mè6MW°»Fiøtôõî÷\Acj&I~%ê|?9{ ™—ÑÙ-šÎ“£80+¨hãrÅcç"êÜ̉ՔõÙÖÖWH¹rI¥ÁÜ×-²>¡¸0EÎ ¼nDl]Ào°‡W±”Ù‡=‰ ú8BÓíY”XòX‘>õŸ+</N5IÉÝòq”Bç»·@îÈ€œÎEŠØ,D×uЬ]úw¯îô\¶MžŸL¡ë€P Åonyc¢–̉`T Vñ".hˆÇ %±ô˜4[Í&%ùCÍV[ãx|ýè4”ç )CøÖú •CþG‰L§YÏg+¢Nòo^â[ãݱî-:Qtm/?uyÇÎ÷{ÇóûÒY:.ëlýÀ]ï…?à†üyBŒ5·vP›Úo¸ÄùñŒ)ÓsòÉØä"½p)B¯b,Õxª}D"3¯Û¿ êÄ„è#bŽ×žõ¥¼I0ºÏ§áÊý!á§¼ˆ]Ÿ>?”½s­z(˜”]û/ÎÑÉÌö´ÈÞȶBºÅ^W_¿£lÜç>%vdŸ¤Ì{‚8—3×Qt'æž­v 9—5»ýjé7ԧÚ{QÞ,R4µ¡[nm_ Ît …‹Zü¡×ê5yeyd9j²ëZ+™=™Â߃¡ýè¹ÚXòsâÃ#më‡AËÛ7v!mVƒîXö"éÔÑÜ´ùËTî 'y Cuz²Æé•/â]A\Ü—à MþUŠ´}læ‡æ¢2.ÒCoñPU~&V—…¾ëçÝBÿ'íë8‹¼âôIˆ­·ñ—':ŒÅBm‰“‰¨ƒŒG›$=¤´²Ëj\ÈK:D°P矖‘)§EdéJ÷Â:šZ`Ñ—ÐnsvÉ9ݪÕðÕuÐÂÉ–qËå“ÞØÇ~œí ¥SÝžÜõ<øIâ˜kJÉÏÑÓíc;cý”zÍß0dj,vªíº·ŠìCùùð KôÓ+X¼ØE‘툞š–u ° ,ø‰†_ËÊxG¡5Lðm>wèÍw¸qª).;û9©¹YÅgò‡(b§_}PdðÁ÷åCr^\‚²Tét«Üí4ÙØ7õìdÉ»Gý}ü—;ÙxB|¾¸ÎÏGªqýÞ@$i×Î7•¾—‰dª ¶AS>²Æhb¶+bµÞ’O¦h§Ë.²ß`˜‡s”ÞÎÒ AškÝx•MЭ‚4[çY]uþ¶í¨zû9’|¨ôŠ”ETÖÅúUŠ*]2—Ò3£×ÄBê¼§Â ’ç8ûŠo0jÚ]«ÌUü?iZ¬6´æ ÚØ-š$Ŭb–jÔ?ЪAM2…Ø»û¼Ð#„é*6J`Mˆ_WÜ!è D©B뼺gn/ØZ§ôáë\N__u69Œcp$}Ò<º¢‚ 42?¥ÒŽ«ø,%Œ¢1‡Ê)ƒ Oœ¼¿r+øÌ;¤iôôaHp¶‡V†Ù{ºNŸÔ{ öœMcÉ«ÿ÷ëŽdºÄÈb~ÈòzÑ0Dl7Jà³FAú£;Mù…%FÄ/,§Þ óOdM+XaV0Né­ÎH MÂqðrR\1Û¤eŸU&¾|Ø+yÜè^z9 ôÉ땟­UÈ(àõ·(™µkýÑ-Âv3w¡†ÀÚ+€÷mN¯Æ¼a!þY6]¥›¨ŒàÛìd ×ÿùå3ô/—öO,“ýi"û ×¬òÎÉòW¼Y·nz9âÏàp—µ!ãß5×bcÓᓟÓT‚kõVÖø5º´y —–…Ìèn)6ªàN[VDÕ8xBjÀ£J2I{æzÓk,Ô[Ww M‹µú`¡—©«iÉ‘îÐ?ïh°""ã·C 1I@e=:^ü¡”8 $>s±ÆLa>ð Ÿàœ^ú¿l/QF^"ÒåØñiGäðæWµ Ôv•¿Ž–‹”HQõb÷®<âÔLLHœ³.LH+suáTêìh 1TK¡K(|›7H ßÊ"G3øÍJÁ¯Ò´·å»ae¸I´}8rRé—7@¶4àm^èè>Lfýh}DìËŠÕð¶÷úåéfp% l´}Š^§hUQwT)PTÕûq§äÿ¢¥ÄYÂhÞ¿Y†/L²?Ÿt bˆ_޳/‡P¾"ÖÀGá²22Òµ§FרՅ¤×ˆ‘@:F´“%3f†'ö•z@à‹.!LTLš’lÅê¼Þ1w¸°ê_ðkfS±ßÚ…ñß¡A¶ÒãD‚«ï˳Îúüúz<¬uàÂëýéÌ×DgGÈòeðF'J©p JšÀpƬaÐÎ*!àCÀ9Ÿ„r?®TÙf[öþØO¥ÎÇp³gÍO°å€ár)dMªå^–¢kCÈÌãU¤’4(Ÿ÷Q2üAùãbbã*ú#á¶IhÀpQ†Ù U€^ GI®„ôy0Ÿh}J™§ƒ[%w…kü­Ñe5íiõ¡“¥ ¬78X÷÷䈩Éë¦ZXáìaNRåxHO Bùµ£Wt¼Ç”Ò`Bò÷9Ðt #p'à ¯ñ‹wË-¦w²{mNkr…­À¢oF:w­FB™Õ™ÄÑ*Fcù:³í8ÕpoŸì0[&—ó—pnmóê_Ìëö¦×]Ó8šÝÇS¼*EˆU¼uéù¹/â±ó…w©|¸0Häëí—øÛ]¸BÃ-LŠÀªÚ7m¢D‡“è½q¿¸ïçxgçŽZ¦6ÈÀVå¿Ö k~ôNŠÉÛÏ\ÀâÙS `áÀà%î׊/ˆÊÕI<>êÒ5˜D %ÞhÀnŸÂG.És)ÓcÍC¡ «ˆ÷ê4]‰ u6Kð°ê%¤¡Ané©1ˆÛ¹dÍœ•c¼,3rŽÏ;++VÝTm¼ Ì ùjÇätÕbŠÃñyv4:FLÅyŒ[7GOÙ9Â0ÿUfA9û´5Áe‚oHèÔxèúZhrÕ5Jlîã=ƒÅ‘P¾Vn­¾¡ñ5Œñ)·Oò¸o†Ë8*¾l!LÒÀÄ:¡]XYï›õgý®òÚ¿H™Ãà0;“zz$éÈÔ'"§²¤aü½½‹5@û aèì{Ül¬çàC²iP¿-8q@ˆ23MS!®¨ÐYp=z¬;²‹°Ã¬Ô"T{ôññú˜fFΈ]V¿“œNl´Í+®7]YüUw“6)ì-C_³d§¥D -£)íÀØvjÊ.Øå÷Kht•ŸÐaˆ ç°…U!—.guZëò¹J“fyÌö}ëóT‡”2 F ëcðÎ~É· ~@¥0n,÷Z›×îi¥½OoÛ£»(nñ³ùs6E¨ þe¾œRâv¼èX>›Ú–XXä!Œ*„àPÝÛ#%öP†þ­ì…丫XöæE1º–j-i!Þä'h“q£#<Æj±éŸœÝ¯Œ`yÜš ö¨¦¢]·?ÂãP´&TŠ·*(‰§ A¶!Þ³¬uHš²®ª´¨ "ÓÇ—,˳ýn±ÚsÞ‚®4je;&ÅâÝÔÕ0@„ü$굎ó|ç±Ä½«”Ä𠧉õüdtM`«F#O‰ñSnÔ¯«æƒ‚²c‚ã¢júa¦2¬×PmÎlx¾=„}wòÀ8(l{%Cï]gÎì¥8u¬‡Ûzƒ¹Ø' †ÍÙ|ÈÆZ=† 3Û.p‚ìÏòhè˜Å«ã ~?¼¶²4ì¡.böNù«D¾¸k‡·ŒìõJÿÔ'¶& :Ð×buLûg¾Û6eDZâì!*Ò²ámÞ÷ÍòÕá§f[WåBAü `4´R³¬ª¦¾'®bAY¾Ÿ¶ìö‰Õ&{fQ¯[œÞ¹%/Ëþ“þ~'ë&œõ8{WŠ\E6qåñ>H–^#’ÓhPqã", øü´û–ìYùe-¾ˆj×S_cŸ¢lÇèT$DÖX¨N bæêÊô†/4Z«È/l«ÉDÛíF[ßÁÙ(Lƒ®FÙh„ÛL1ýùqf£¡×åË+aÿHÓÚ³äå¦H(-²4Q»[›M±—-¬šiGëôœÔ%c$ Z~«•ƒª%=¢²šhÎWB·¨–¡.ºQýš`@ÜT\Ÿ†Ö "! À;E~®š'¹RF'}:S³vmάãÎ`ya˜±Ný Ü { üÛ|T¯âÎÙRdEõ‹íªí=wR‚ºZ /ì ‚ºÙÚ]ä» v\’`½VÃuè‚;>‰Þv4µšOŒb€ò~Õ2 ‚H4/$“®ÚœíñeíqÉ'(,Ý‹§‹aŒi,,®_Í ˜~ÓÈ—%ÚK¬Lõ-s¢Å÷«ŸÛÛ#àZN/uKXõ ø·¡éo6Öñ‹zxƒœNŒ‡‰º+Žä õ¬ï¿ž_d@MÙU‘[fÙ½¾I")ú©ò÷ûr‚ˆŽÒgÐêR~Kûú=|Gƃ•˜:h¦Å'bËŸë?f¿q6"[ž¬¡2Þ^Xþ ˱ø¥Un_j,Î3èbêUl­}ð|'‘Š„/Ÿ^Çe—Ï›ÑŲ@õòøèá¡â“yÆÄÝî¤UÄý/ÃâY¿vg›oÂóÐ-'НZppü.ÀlëX»-×¶5?P(<ïËÙ•õèbÀŽòÉ—9{¾Nz‹GÿºÔ’R izéLÍÍ_Q¿öl.3¢”= Q|æ3ª²œŽöð±á.[0;W—)ø5A¹yy‚¢ðT¿“¢†(ƒ¦U,0¯FãûFGŒŸkÒ#‰%Æÿ;Ãs“öÀ7a;@Ð}ÉÖia §I–‡˜òøè7~ãEÓS<}CæYŒ×©Ü´Ëp°½maøÂ\šì,¹Ý\˜Bô” Ø¡ˆÖbÅ™™°·Ðeü0;ºžLà§«Ñ.ÝzÛø46‹·©XÏ,k¿¹ ìÙ}&i_A[8ñ˜hƒ«åjTd©„dJàY ]ð'KYˆ[ZÜãç­æ1ߪåÁB±«•kÜ"‚ׅ˶°õ@OÂr¡™_ší`öž¾cSäA)Œ+GN½M€R¨Ìâ¤ï‚‚îù0„ey^ªëûº ýe’!§);Ê"ÚË0+5Û4Oot×]ð¬©ÇÏBuÍrø§ÃË]Ãk”»)ÓM\P!{.˜þÎPb0DÒüM¼ëKýíl1‡~¾F(sKÈÇÐmö4­ÅYd"džGtи¦šnìR~5®ðë8BuÐ b,@n®´†…/ž¹ðLB_êšqHØä°˜\óêh6åÓAe9­ˆhÕPŽ–BB^Ý“T%«9ͤ,˜uK0Ç߉Êðùé£17)¸|±©Ç‰î hr êÏ(7üNÙmÿ™„ï‚^(Èn羚¹“VêÛ±ö}naA5ªÔì¾|ÊpøYuº›ËÉ}¦@ÍþmZW£þy'µK_ÎŽ~Ÿ‡5˜ITy±t#LÝËo¼·Y Î-”jí¶púâÀ pB‹HÂÒt•ÿ3Ûv[ :'–ªI[w§¿û/€:à EZ Ç0êr SîA‚|^ù²¡¥AÇ%D´“ìÅ?x2 v4RÎ<Œˆ’øE‚‰É35¦ú…ýxˆ´²0šÞ# ƒÁCBÍ{<$àb˜és\WBd­zkHåVֈƣ†¤fôìiŸDü0Õùu—tL0N=QûGGy©í ï51­½Ÿ*÷1¨gKÉ´ØÖõçÉ€ÅL¹T™9ЦjÏFÈÕÀY“…ÜÓÚAô‡&Ëä09,6ØN<¯»_6‡¦X‚#ŠMÖ–!Òiiéã­X«síÂRî$Í壂_ÿR¤!$ΨÀj­ïªwstt™x\_««;óiŠD[©ì'Ç©‘è»â¥Î½End·2VlQܦsÂ%©Ïª7òƒqy™k·[b%BD•Ùžó±+¢>º®æ…&¨9ì3>N >—àqÙym¸.lÐÜgH®÷ïy¢=«¯$ÁœâL¹WäÊQÙµ|B%¸ŠßÍ]‚öHy&Éç|ùëg«ˆP›8ÅUоâ ª@95«·s¹£zR¶)è౿ Nmð=]£—4ˆãÑëPÂIøÔ(/k H7%¼ÖU†ß$÷•20H6.YÓHôN¡öèí/,ÉJ`fʤ&•e-’»¬£†B6`¡f¥u)ik˜ËÒ¢­Å T+J–@ÉÃõ`ÚÓîXrô2ÁæI6ßUÑšþ^/gÀ¯ŒB*פ]ÝÆÉLq ÑµÿýžãKV6¸h º«çúÀZÙ·PŽDÙ'ŸNƒ&ŽùvòŸ·†»<ž?Kéõ¦˜â¸¾>öÊ+“ß ûT‡›€Ÿ‘MFrrV§g6ÌÃQ­o¢ßqT5°Õ’©l7~øµVýÁ¢  èXŸ52ËL;‹w¢' ¿„Ue6ÚE²C0!©kŽ»x$1o¾B1ÖARIçô™/_j°»žk#%?ó–·üþý¡÷†¿¦7Ø‘¡];ü #gYÐe&­º¡xŠ@DnD‰‹ÃÈ9rÖ»D¿ãî$TRO÷æ(n~Ýk;.+.jÝs#ï5¸xʇÄËT·º3GÜEœm²xåš‚k=f E«ØTÁ‹Jq+Õ¹‘ÓiM†àI‡um†L›GíOü™Ùx˜¼@d~(D#u9¼èÎ8ûék™8ÆV j P¾Z"r¾êþžësÃb{iœÏ=L,xu¶_˜}à‘–rKÁüP?q›éÇ,ÝòϪL‡${Ïž“S£ñûc–*$Κ†z³¬ôõ¡uÖZVÞÊ”?ás%1©¸t ÒךÒ:ìäøÎ«Ï—?U@é•>Ť6œ·1àRæ=º×Ï­#`˜&Ë…p`w…c‡ª7=ž­@üH’ÂJ±Î•¦&¢Ù Yˆ‡Hä$@Ãl{àr‡àJ0æ0Ž! ~§Ý·Â¿ÅùÔ[ËqƒV¤ݸ9)âþ“å…^H´dñãJgö§!¤å:é åÍ›ªF4}äx#/ºŠ”AO5å.ÊÓëT°ôò䤡.A•ƒ´ùÖsR©ì› šZµ5G-cZSÙà`™u¤’›Ѽ“œ2ÆÃ¾J ¾bw€MðvB±Û«lÍ/Èeßx|Nèɤ˜¹=†–’e#ú $ÐË x¼ª,•ãWÐ%VZ÷Lªd¦'›üZþ‘݇Í­ñÛLp|/ÊÒEzGè>D³ÄƒÕF_)4.%‡8¨˜ñù¬Ø×ügQ?±YŪ@ÇÉ*Á¦{‹-¶]êsôß:Ê•Ïð゙Á¡ M[¢Sž=ÈýÅ©÷ѱÐÔ‹IÂÂÏ`HuE½­ÁÇ~D$>`K±oØjÖ[ĵwÄØÁ^bëLsfåyYo[` l‰~Þ£% Âñ’üú%@bŸ¬äsÓ¢ÿ9 £X4hÀX/"oŒË9º{*Þwq A›O Ê6 Ì¢·õó:#‹’w>D2EþÁž·®4lkë1µÁ•²þåb÷³àPôwä³u˜†¶?'ÚÉôd%ÈÁÛÙ¢_6Ø‘k?a†ùÉÆ7¼>öHOªêàò88²3dä¹l |ÿG‚’eÁz[è-ÃC‡èXØíuc{›kßµ‘üǶ|ßbI„uÖx:´X]üºÚá¾M]¶t’vÖâ›1ž¤…´æâlS‘:¹™.–cù/ß’*lÕ„æãiêjy·ÆD:7S]žN0â!­zŠeuªRUävEqù>U¾Õ¥;FÔ8=œV»QP¨HÖ%è[0Bœµ,f|~Á4­ðdÖÙr5cIð ÂëDxø~+H†‘÷£ÂÌ1³‡!ZOrUýg¡ÁƒãœÕnàgR÷‹M[( úP¥·Ú@TOXJ× B<ei¤×­É‚‘§‰Qcqþ]µ:d+MyÊvfæïÄm ìËÙdꦌZñ̻Зköv½™û†•kÂïe—ºYûÕù9CÏœE’n–úÕºÇ +6µƒrÞ§LäÇî~B$¶¨Ú¤Üƒi ÌF1È)F©¯v3ðà©¥6™d¨Èiï_Ïx¯Œ¾ì–!Nø8˜!â ãÕLc%Åz*,4Ùqü¾i',|=ñEr¼¼%£ý¬å!l¡ÁËñzhÁ¶ÀBnë®þ]'ÑÒ(† yíögd”xKá}ôd=bMbÇB^)1ZÁ´||½l[¹?ç"G/¾0;yÑã ë¢ê^ô1²Ö…„Yu݃m¬¯¸$¯j£9‚3#ünˆïæ$ú6ti›õ¸sÕ^ÿ,#ü¸pY©×ø`ÚS· …ÍBÓ$§­b[É þ”xÔƒ—ŒRl25’k[hü¢S2M½e&ýY]è˱Tg[†'¡b渜JÀGœó69iŽ¥c£‘µ!ñ ¶ÉÖðŸñŒ×2¶÷?A·Þ~åÏbÀêVI¨ËÛÖ‘êÆô›¥¤ÇŸ¢c¹_±‘B†FIqÀ ×yÁº ÷ÂDhGÔØØA q%™’‹†¬é]µÏ͸ÅQoOyo·ã&Ô5AÇϽQ‰Ÿ²_bia¨dËÁ€¡‹W-%bÁ˜PÀöì퀙oш_–ý&ágQ„iEØ©#ÚA¡£ ¶cü}±ƒ£»ÜÑþëþwŒX»Aw“6¾Y¹oÈ,–¯êœéØU·çL…xTëe‘×Ç©ê§Íý™ÊS:§ ûêÍåÛG¿'æÎ“aÀ¡x/y`Û[¢,ðg’×’¯ÎF*#½ wR]‡>÷ý­¦`Îâvpá78Ý3ãñ+s¼pör·•†è»çI7%ΘOýî –d*M_Q…Ïý¢Í;ߊu±£ö§›Fê¡ËQ_ŒÑæà½xrA„éBŽFw_Ĺ!¢)L>(Ì÷–Å/+ Å‹ŽyÀò±ç…q‚éRWm†¤Æ­Þºi,‹{Ç E0H³óŒZÄ Ö#\BÁ¦‹öäc´Œ-v´§7$9"–v¾ÉÍF^›˜=ÓÚpsSò4=¿ÒƒM¡+>ättå¡ï̈˜`ö´6Ö¥ÉwPO‡i?´~UÂF™ôÔÒ´Ù>M8_.ÎEñ.ˆ¯w´AâÄÉH~«×°í(6-e·¹Ð•ýJñšv|¿6g!²h(d˜ŽS|麞ƒt%4eLÛ ÷„I „Vº Tñþž´ žüx­ƒo¥+AÇJê>YUo_tË‚;Aò܆=¤›JÇCÀ’®åðd¨X¯ó¨¹›Íà[Œ—±½¨¥¿ ]Ô÷ÂîSž\ëÆRe¤¸)Á¿K,^¦ˆNõAÞ ¢ Q`Y0# £êÖ^Áõ¾f]£¹æƒ«›Iv¥qJŠüThe«Æð3øoC$häÀ¦*µÂ"çm¹7j¥Ïy#É="Þ?á "uÑ*ÛÖ¤Eµ°ËÚ³@úM0ÝQÚÉxE¿5GáE!|ý­µÞØÑõc?Ð(zÝ*Iª3 YX‰ÙP»Ÿî:ùLΑЪwô’– B=¿54vñÃU‰FAvÅTJ–+z݇·Ïa'¥«-\PÑ“@ê©heñFÒZ}›/&H’Uăðåè:¿·|„ûíªI(Q8ëqçi¼XB'í"(G¾4±ùѯÀ[zL ‘âwÒ!$`/{éËãå*0*è"ìn·9ŠJŒm~ÑSË):'Qh÷CV³'—qú>;ú †Oe¯*£¡ÆÉ.XháˆqJ‚â½t)á›YSAæ¾—)’éÑ2sÚÀ¡÷Áaòiô{­Ô0bà— ã’T˜à]êQ M%TWÙ¾¤U„ ö ´Þˆ‚§ÄsùçQϧZ$¼lÅ„’¾¶Ç¨xe/ RÚV:O(£[¼ª¾`vìî#ë½µ©°þ€žÃ/ÈÕ¡ª-œæfkbs¨mt¢1ᯃƒÃ&º½ñLë†I§ä¦«tºÕõ*Éý=Zãæ¦€mnsÅh”©ùDaJ`"`,þ3ÅÐæ¡ ¶›ßÛãä3ûF>)êö‘¾·÷zü§s,çë, }äCÚ83‡Š¸$¦¿ëUmœÄùmiãáÀDÞz”¤°ª“üêFÍëKœÖ¼¬¸l ¾Ç!.Ü2m8HµÄ+ÜÜú:giŸˆé_÷:¾„ŒýÊÓ=ñk¶‰›œe`m:(c EŒ-ÁFrÞ<^S³'ÀÞNTýÌôß—(öõF[cpüBèˆ7¢Óo¡ÈËL¨Ð™)cU’äJ»Çw$z´µ/ägí®Edj`ʶ1b3Ã2_xœIMÀ}eÅEèèõƒíšY ¢ ƒFÇIúíUTT+A¡3Ð=–µ›|:B”ÙK„÷ó·ÊG„iÍå)¹rã¹g¨ì;‚šìxµÑÎ \èƒ.yVºÆAýáã7,uWNÇz=ö:öNFR3ṫ<]–)~œ”a?%âä¹ûŒ.|3>kC¬[á-x¼TÜ‚R4†üŽ#û+áþ‹Ü3=hÂUË{»'Í¢…rAIDʼäž5µ”XœºþÂ4‹Ý—wç‹pG=lg$Z¸ß‹_Xî¾¥ÑdþFÌ-†=4’¢°ÓÉüXü¼ávx91üôC–_M46ª +JûS<¤æ§pƒWœC/)É‘ª‡PÌ'{þªÐîÑìLÇdDñ˜o­øÇÈ»é¡kGÏ£| »íˆºþ¶A¡ 8BkÚv @’Õô¯­.Ò;’{öªJ¿®g0ÍÕ˜6\1šÑ·ë̺ðêoÝ$ü È2>Ti€"È-ðÝüŠ wMxø@]µ¡ —cÕå)ÙŠ‰€ˆµÀAÌãf¼À©…þ±†ÕËtñ8Ÿ”ùHÇU`]“•ÙæÐfÇʸºMôçÛ .ˆƒ†8<cÁ]}ㄨҕ„uO›ÿÝH8¯XKU‚a޵{ûñ‘u NjÙriû#Ô‡ÒvÎF’y›+Ú4ð s…R­»¾L…ßT‡·°%SB„MÖlô=¸¤÷âÛmá …ü«‚ÞGÁåÌðcűÔ¥RŠ®;ö2ÅKžßYÖi,Káûõ3kQè£Ñ²Ìäí¨7ï@C|½°âÃÛt­mM¿`_ì’¬ áxËEœ–U¯;laewdþò6¯kŸØ¬žœ˜2ÊöãžÙ zk WfmÜç[¤3 Íï°š{wRެj(@ïC«eKh¾íØNÞu N35GÀÌaÔMH×ÄÖr–—³¥£‰=Í=ûh±½A2vIéNN\ÿ>‘ð£âÓŠ'Ìç˜m¸;˜v© ÀZæ+‚>cÃ\Êø¥ßòbEø:¢¢¾ÿ':Åú/ʼn>C½[¨eD'þ¼óŠ»8g½8qk•˜çÌìO½º¨)%ÆiûØéÒ1%ÿEÆD÷"RZÅç™Fh±‚\=í¦¸Òî Ä‘JîBNŸáu’Ãs¹B- 0 'K±+ùÌ’×´|L# 8#ä½±ûÙϹö†¶ê5C.ôæË­ìÅ9Hòƒ6'±\ `•^à–7÷e|jv=HëpG1‚&æœReHi÷>ôö|¾Ç…¸û|Yï0­jNÔ›èû†ÉúÎÑ!±¥Ø)½Ì•ز‹·*××Tj¦}æÍ¬òôê,3E&žÂàk-¼÷A7ü䜓ò²2—LˆV¨wÝކ߰’Û ¾Yäʈð‰þƒÔnÏ[! Áœ°Í4’’¹~QyŒfÍèxk£Ì®]ÆÙ ±6©÷¨ž4ßì‹¥‡d„ÔFï»Õ£c ™É.žÈBø'©Š&·¤,n¿Ä*ÜØÁ¤HøÔé‰XW&`Ìf—!JŠ­IyÚÛyõçÄÙlý¾v÷µ™o|hg «-Ãñ[²üY/л¹èáw¼í›Bi/ê‚ÄRù R=OÏÑá?§9U.ÕcI¼5Ayìö$Wü’.HôÂù6fla×[¡ [èòše%oÚô~T|°ðŒ!ìh±û;H«çI -0 h™Òs·m01?#Ù”RµyZãÖ7XS>à£f(‡fþ —@¶D·šö,Ãw¼äYÍÅ•#—ìÏ\`@4~ð@½Â­·^ï€þäkhZ¸€2YÈhp¸mÖ®Ÿ=@'«Ïë»H—÷}Õ¬ÒKÛ¿ÿ„æ„mÌ—Ö§¹5hÛ”%Èé5¼'²–º>úí¹›doº¨’N±+Ñ`#F<„’‰êÅ[†ÜäGVáÓm7²˜ž˜Œg±øôǪ…¬_ùB%º‰{#¨ì‚ù”L8~¦}¤Pög\#½]Søå+ɃÓ=¤UΈˆÓŸ²:~Å ®÷¤ÆOG…¬"Ñ!Ã[5—ÓÛUÑO]ƒq'[>yý¦Ïý¬å_žÂÖ¸'ËÕ†#ØÄ£ii fBâ©ùFGÏ{dÆÒ gœ_*§Á³r æ5 ›s¶ÚLÑÑBTlú^s;Gî·(¿æ'^=@‚ì'™äù Ù äÖ?+ê‘ߣnÛ×5ž¶Ø[)_Ö|3@Žj¥s˧ïÆ7RSÛ_f{A­)ƒŸó¤Û®Zk+úÌm-çÜŠnßRËcº·mð ™s šD¹ü³œt_`v¤nÿû@Ä:1Fv ¸ÎwFhíDj,—­*³RŽ…àÂÝПàðÀaô—v3åwüÃC6v²3üý˜¯IžÙ.lŽÓÁ¿ûaë;rMóŽùCNb«ö uÏìC>MtÜ À%™¢ˆuÞçx.ŽÅ¸D´µ à\±É*ñÛîÁ<ÑóHŽ›LLŠBUV.Ef> stream xÚ¶PÚ¶-ŠÜÝšàîînÁݵq‚»{pw'xp‡àîÁ î‚;<öÙçžìsÿ¯z¯¨‚ÓÖ˜s¹h R%U3{ ¤=È™…‘™ &¯Â `ffcdff…§ P³r¶þÛ O¡tt²²ñþ#@Ìhìün7v~“·d]l,lN^.^ff+33ÏÿÚ;òÄ]­ÌòŒY{Ð žBÌÞÁÃÑÊÂÒùý˜ÿù 6¥°ððpÑÿ+ bt´25ä-vï'šÛTíM­€ÎÿU‚šßÒÙÙ—‰ÉÍÍÑØÎ‰ÑÞÑB†àfål P:]f€¿(Ûÿ fiåô·]ÕÞÜÙÍØx7ØZ™ANï. 3 #àýp€ªÌ'€¢ôwð§¿èÿž €…‘å?åþýW!+п’MMííŒAV €¹•- (ù‰ÑÙÝ™` 2û+ÐØÖÉþ=ߨըÊÖØä=à_Ì’"Êã÷ÿÝž“©£•ƒ³£“•í_-2ýUæ}Ê 31{;; ÈÙ þ/~âVŽ@Ó÷±{0ý}³6 {7׿¹ÈÌü¯&Ì\˜ÔAVŸ]€2âÿy7Áÿ±YÌÌÌ\<ìàgÐÝÔ’é¯òjÀ9Yþ2¿wàãå`ï0oèce|ÿïådì 8;º}¼þéøoÏÂ0³2u˜-¬@𪿛æã÷Ëw´rè2¿kÀü×Ï>é¿ËËÌdëñ'ü_÷ˤ&¢®,¯N÷wÇÿñ‰ŠÚ»¼ØØ ¬,nn3Àç¿«([ý›óŸT¹=€ço²ïSú®ÿ¾êï à¿k)Ø¿‹ þ£q=ffÓ÷_,ÿÏJÿWÊÿŸÀÿªòÓøÿ&$ébkû/7õ¿üÿ·±•­Ç¿Þ5ëâü®yû÷-ýïPMàß;+4³r±ûß^gã÷=YØþgŒVN’Vî@3%+gSË¿Åò·]ý¯%³µ•ì¬þzU ,ÌÌÿË÷¾Y¦6ï/‡Ó»"ÿå¾/Î)2µ7ûkÃX98ÆŽŽÆðÌïBbåàx±¼¯¢Ðý_01‚ìßSïíùÌíáÿºQN“È_¦¿'€Iôâ0‰ýq±˜¤þ V“ôÄ`’ùƒÞóäÿ “â73€Iéz¯¢ò½WQýƒÞ«hþñ¼#ã?ˆÀdò½Ÿ`úÄÎþŽÞß…?ÑÍ•Éìð½à?à; óÀwÿ€ïå,ÿ—¥‡ƒåû“ø'âÝfõøÎÔæðªí?à;W»?åÙŸRï© ÷kÿ‡ÿªýŸÓß“íÿËýNÝáû½˜Ãûc²š;ÿ±²üÛú÷&þÇü~ÕïûgÿÁ°¼÷êøøÞ˜Ó?à{ÆŸüwA29»ÙÿÃýNÞåd}÷»ÿ¾³ðøsøûP<Žgÿ—žM]ß©:ÿëÅyûÿàýûݦðË ö¦|ÁÖuÁ÷5"n {“³{ši4 ^ËŽ.ȰÉ4ÕYŽ·"É#½¨«;Ô7Â+$/^'­ °am‰ÊíOÞφñ*3{íðKÓØƒSE'"õDp„ jÂûÞ/Ÿ½5l [Á»d)ò>»p#+`Ü»õK¹×”ÿ]ØSÞ¯æ”Cx.ÿÁ­¥P2G‘o’=ûÆ™è-úow”¹›ÛYôÜ©7Ùx:xŸÓh¶b/MÖ˜‡yϵJ5V§nlpƒ¿8ÞÁ,W/©®ÊãÎB+‚“†yh†ÛšpÁ–šÞâ_GÛ’ y¥-¸Ò¬ê½šø~u®_ÈÀÌÏXn¡*Š‚¨`›„¡“£÷é|6?†ƒÀ€§¡ó•·D®Úû—$|`ð›W_`%Ì¿|A‚~/çésŽ`§5³AªûIJòÀ+•VßÑ™ßQHD‡âž‡ÎˆÒs ÔN¶à[íÒªÃ1K€ªF ¸Oìï×îõrã+3fËþE/??~;˜yV·³ñÜ”ÇP¬ ŽA_ÈÀ”ás0ø‡ B©@Kÿ(ß©³«9~6Õ\¤³® Ù>°œOÄŸi¿ÜtQ„ìð»Q±ä9\(ˆ»6Eå f1Jõ7¨‚:ÜuÎ%`{Îf>1f«~I”¿&¡ ýÉ?ÒõËõEiÖÒŸŒŸE†LÑpÏi­¨½XòÎÂ¥¦oç ÛdÓú'3Lèý[bBþï'¤\;¥ wªÇXJVN/#óK”œ»©gTØh+¨‹ "ùŠ)½vØÊ+ä âZ!ŒÒ8î?2µ€¿L ¬¾ŽÊ!ö¤£¥r­ö )§Qbg²CôëôÏѤ¾ÑáGØ)÷Saðpn-ÉpÁsïQ‰gAÌ|$Ñb.¼mÉx ;ž2[Ó=JÁü$4”`QbìÑt.òµnÆ`·¹Ü&ö†;ƒƒñÈ*™.(ê»,´{5"oOáþêÔ¹¶ÊÁ”ÅÐoõô‘ó‘­ú#–'i¹>U˜£Ð“­9ÉÆ½wÜüí uĵTÏ`—0t;"È&Ó×óx¶9P¨x+Ø·¾#›U/’K“3™V¦ÛÓ^‰™A¤Ñ'ðíqËqƒåô»E%Wý®åƒrˆ¹£­#.*…a3³¬Ž#yÆYæMÄv¼|‹tÕ‚DWÚ_Œ°‚O¥GìL}ÓÿTˆÔök´ürå «†âáOF̵~®DÜÇ&·©D?¡VÈb±y)&˜\Û6™9e7²|•¦Hð ÝI´WHËÑR„Ý_‹6—½3æúˆÃÕ‹ÞL'¾m"”ÚÎùm%rþ°Ìà„Ö…c]ÞhœÎ#&ÑÙtU¿ÎضíPò¦ ÑjA¶ûݶX‘M„Ž.@Ùa P‰}-Óç”lHT8^—íã·Œïí¥XÍÙß3š*šËe-s‡€æ?œÆšŠÿRoŒñ\Bø–&ÌÅŽ ü†uÌN@ØÁ(þ 0Ab”óÃ’Ê¡òÇ{²ë„–OhЗq4©64q"ÞÔ@ƒ¦ØÆüÙÜ8ZÊVö ‡%Çm÷èç- $q^¬%J¦ð…ªáx̉˜­̸ÏJÃgá¶™ àÛ:œpmûëWؼfïH7ùHÇDSLž %,êÝ~îvØj¾är®!i¦ÄÂföT•iÜvŽïU³þ0 {êX‹ô’„Þ´«AQÁ•‘Æ${ȸA ªˆÓº¤”UÐq„{ä~Þówv‚jS65[Ž]Bj´…åA‘Í.“šFBø(ÊÜD×nRlDþ¯|fÅ0-«¶£öpø­¬/ˆ‰ê[W7.ÝœÛib _(ð´Õn¨T®£†ð­€åÚ/°É—ƒ|>½ÆÒýNäU|md7º:˜ÍÙÃÃao‰i¨…GEü2>L^¡m<_#'ïgÁŒ-£œ„wſΓúäêÍÆ ;'µ¼+ìýY+4Jpk$ÖA§Š¢†šXëÚ5ytŸË‡¥ܶB¤Znå›?8 g1Ú‡”ýÄTºŒÇÍZÛøkCN¬+ª¼“i¶U#Z5Ä&êoeú9}(œ*K º=ÔU§"²¤lN¼Ð@Z‘‚ãÕÂqDh×òa §…™(»Óéò FÆŸ‚ˆa½á³—Y‚Ê]a”­)CîßöNÆ[=AiÈ]á¡6b¥5æŸ5–“_‚8?GØÃÒ¦9Ð1ÞZXÛêä9[:ÄN ±ñ²;/_,{PH¦“ºñß~&9G>«i—d”ŸEöÂ-ÛâËa=š$|ríÇÐV˜¤“MÜ©Æà¤Ì#?Í|ŠãGð@?&s@¹›ªá][×0ªí&#€Ý„áRV …qïÍåzÖdKÙ‡ótÏ}4À©zðWA-ÂpóÖ!5@ùî&a¼h^ûã×H³•Ý[V4;;“É7…·S*üô[H¹Ÿ½l{h—GµŠ–S²"Céõ‚%è<®¹³q´°T2vÎ-³ßxÍÈÎS “>Í쫸çõ‘†<›‘vª €'eú…g4Ò^äÁtíòìçá¨~7ðYêÃO0!¯yÔ¸ -7pNa÷±äƯK+„;öKWÉÓ+6ÃrRmãmM¡ÙÔ“„f¿…*(㮹ƒÁ*ªå|ÐØch§n&•mlÖù×YJª_í¹¸*¼ÈØä_ö€…Èq6Å›(þ©ÃÐqYzï¨ê³E`I5±ãAI ê¡ÅDµEâw@¥ÙqWJ·{ޜȋéÉ@ÚhÛT~Á³¶Ùu0-oν}´™QxAˆü£’IpT(¨(>ì¨Ú*AŸ¬±8bÐÖ‚rdXšZ…r˜bÂa†1C«(LJT«®fùˆ·fëÚÅÓN`Hþ0‡7Nî –™1xB„J7NÝ"7ß±¢ô¤ñ%AÑ!tÈÙT̨m>o›93¦¿Îÿ7G¥¶õÕQB8‹l "~5FÒ¬À(HHÂ!&fËø3ZA¤^õÃ1͘!û2Ñw9íüjºF³AÓu/œƒèÌ:ƒÎôÃ/˜x4•V·@?€"I}Õl.í+·(Ç+íþ`pªû&¨S˜¤Gâ B\éM°-HŽ)sóóÔ ´ÜmŒ ].ˆuŸ_`C¶ùÍM¨Ç¡o«ám"±ïe˹1›èg7©!  c¶ß̱ΫøÙ 1a‡>ñ‡ð¥>ÝRS|g¤ØOÏ֚Գʩú·.¹‹(BzÊù€•#’=üÁª¸ÕÝé–O¢5Sèd²»uà¿»¯^û.Zˆpp˜±&˜`“Z!²èhñr\:ŸÄ?&_›ï˰” ÍY;÷Qpó2QGÉäHåç®À€6rFAá…¥!ñÑåו9Þî³Üs–¬ÆÂüªÛyºö<ûd¾“B}lrDJ)ti)8ˆD¤°Ã¨;é›ðYăG3|â=ˆºcª\ù6î€a¼”ò…wgšÅýa(érx×(F£;ù‚RB"7@¨ÉyáQìgêP:WƒÑÐ29Tu•¼ëì1Ñ1¶w}¥SÔ$â7N-Ò~ 2ÙT˜n¢Únpˆ:Øl9&Õ~Ξä®Ä0ä4Éžy‡­½"¼–p©‚8Ì]>N¦Ë¾óºJ—ÜL\íXgõñGŸuóf`¢•·"㛉*D›Ó€(¬gÜÖBÆpic«Åoˆ¢…gLÎRÔÍŸ&°¦Ži$­—?ã5ñYýD>n€´ Æ/XÁÛÙC@ñrD=Šb Ø "e[<8gšîãÔÐî ýo+9~Ä5=+ÎÜ8¨> ¹× 21â£ÁI i]Å$]¤q¸#ëÐþõ}À›$1»ºã—%Ô0X‡íT¬øá¾ö:²c3 ›o' B·GªfQjbyi,€}ƒùiø†T޼Ù_ÒÈ”š®i €´Ç’¸Ä‘ëÏ]OÀ„ض£Ñ¶S]r½\ÒÚw˜Jk éƒOB&„¨D°ÓâÏó|o×çký;)+ფ+¡Æ €$Bî÷#*¬•vÁí`̵N`d.F4QDø‚÷>:%lVçòº¹#ÊV“¿r‡/¨™ð‰#B}ïŒI7¢“W71Ð_QÅX‚ó,¸'Õ?…Pì öXm­4VMo­y„­²òÍ…«á3Òiû½µ-|hd[.éêo<Ç„àüh+¸#T#—~nÒ:rüzS¨ÒuõôgÅQ>_ŸŸ¥(-h÷f‚Á‘ÃøF‘ÇSH,Ú¾ÃBƒÛª?Û»¿DæÒ¡ÕT¦ðè'/UwáoÆ’åT‘J½mXsµ[ß­Ç¡¸yèkÏ­ÿ©_) $æü¼jȺ:4górƒ‰Ay2pĬ•âZL€.–I¹êdV6ïaj‹¿…‹|"ÁÛ8ÏÌ’ªn{³Æ$JNóE/İüÙxo¤¬ãÂîh|m’3Æ—å©D¾”üT#lÛ7ðyÔ5û$êÒBí–6<@ ˜—’¥E9Ÿû®Íìy(æ>hßUl¥å/'¬ œu‚™vg^Ý¿»öĺ^SÔkµà˜Ï5JAX€á93Á^ÌõåŒ{¬ù,Ø¿6l¸€ÃÆwž÷µ3[ÜÍácŸõEÔcܤ‘ªŠ^“6r-ã7…Mœ±Úx›=Œ ßui²#jnä)®z€+Ê“ 1‚ßa"³¼9¾×R“$\btª›4ÈÁG¼–ù öñäÚãfר¼(È¿4K„MHâ6†Q5W|z/ ÚtJÝ[ñ²”®Uˆ¼ÁqF`9,"î(î¥á€Ãh<¹~s‹iî·#‡>ÏèÒoAå,…E)+Ø Ê»¬x‹ÔP¯ÐYC”átæ9é Aê±ÍÎÙbOïLµñ† ùjsÚÂ`Î^,ØqÓƒo+ü©Y*® nßh÷äQ–bîRX5 ¾"Z… íÊŠš‡äáeŸ©|…€µòÚ•&k`·D Ï#°ðí€çþΰ³xÒ9)|Oísš‰û±­–£† Ÿš¬TUhtî0žoäåõJ•™îŽqŽE—‘?†Ãÿ‹ßy1G!c¹g|ÃVFš(oÍ)»OàÜ ˆR3¾=…sëWÄ";Y^F½XÓUöI7½¹?Ú‰ÖÔâFxÉ×ÎÆÜµTŒ{mt#8 ‰GÍAG8Ö-h|²$8ÏŠË¡¹‚µ)9PùSE=W]‘C7R3¶™ðŠ~Ï›¥ÐŠ®]ˆƒÙ©{a*3=”Û¾½gýefóê™mQiªSðúß18šåXQ¤ZœsÀ ­ö+&?§{‰ìŒ3êŸ><,gÌÝ»NËiáý0w.Ðû×85NC…RHázE8b—²`y4”ÝvòP;yã]°zP"]=Óã—\b•y…ŸJЋeE[™onÑn2U| íÞ8ù‚bªà\•¶i0¤ªŸwö”p)PêÌõ)büb[÷vpmÍ;Dl`·óÑGrC¶I|š„à §šëvÞ Ë®41)Ji É&ýk2¢žèòqÈGâX÷­ ÓËD³=]¡ÎõÅŽ†ôMéýf¢Ú‚ iƒF¿íÉ0·v„Þå/êM ¿ÂØ’ü¼Š¹²kÓ€/£/VÓí¸‚)\^0>§pŒYù¾Ü<)@—ÇxälØøƒžÊÿk¯ÍÊÕtIŽ#í,Ú»fˆÃ"y¡ºCTSéL hÁ 7*ÿóJ^z¼±‘â¸÷ç³4èüõ­I¶#|ÈÙ¾6ò­c¦+ïò*Í FQ›Õˆe|œûòÞ¹}¾—þwÚ—rÝZÛKjûŽå7˜¹ üt[ ”S·Äæ–ž…`:ÿ(?‡k³•z-[ÌXq-Ì‚)„VbVy`Ò^-pqñ~ÃJå§–;w«ÈaªH=YO+&Ï#Ubv % ³2½7ˆÀzêY‡÷\)±Ä±ºiùZ_ͻƅ9ÝsSS>Rõ@ &.ò D¾†ôÃ&ó!óµ?úp.šþ0ºþžD¤pA2{ɦÉDïfç°ÎUfín¢2_}ý*\¼vîøêÉæÞ)¥ÀR°)Ã($>×8ºL©†ÚךBé[hËðˆ |£”¦mò§€!ˆôì°ÝÆß_o¡ÖâË({"-]nÀ(;‰EÀCÉQ££ç(cxLõ”?JQ\GØ.x¹i Ý:…éîÁLþ æE•›J÷1qן s¥%¨\c[:2=áf¯1šEj%÷NÚ=r[¥ãH­nâGŸSCŸƒjLEX}]Þg^ø@Hw )Œ¾Ê¬—§X-dX›‡b›êã>ݦ™Ñ³ÑÇ%@Ü M|– «H¦à%aAÝÚ!žŠçŒ€9ZÄQ ͽ…µÈZ"•{˜‰¨’õN±ÁúYR”¾|×ù=»%YÒøZZþoµLªuýòã<>ZDº?UضEXé‚ÌN>n»‚IrÂIÜ$èž3¢äOÎê((ŠíôXâbEz“ ÿ–;äøñ…‘ƒZnëkV64ˆŽ½Ýc^¸k:;fZtþ—®6IKGÒ \%ƒÑ3ZTrDRüÅDëà×X‘ÔìBò°€ÐEZþ:¿~Rjw"Jr—o¾kQ ɺ!;î\7 ;˜s©-wàÖµÿªä óÉŒB3¬4ñäòs| ŽÕЉZ‹ziE=¯Õs_ެv¹a>l ^l¤Âjê¯uè ‘Ÿq¼½â¢ÓÁ´þWq¬T>¿y“Qß<¾ÂŸíuÒP룼aQ×¢f{±=Á“òG…E†§÷ÜÎS³ Œ g;“….Ê[0²X7÷гAAjö*åÃK#û9Ÿc_ÃO%Ö¬>W3¯ŒBš½f‹|ªl„ŠÃ3Äut¢§Æ úì"ø ˜!à„œ¸!ëO&Ÿ¾‡bC×wsÆ–÷¸–ØÉß »T¡T!yiDvlņ¬5ޝ\eïè+ŸZÚ]ZE/×õ<:ãAzsæG‹Ò¿?©¾HN}"@-[6¦g]d¹X‹UàôEÐÌöéõ-eÖNÄʼnGQöΧ$9é…êæ¡·¬}Ç2 „Øw $Ìdê&Ø…ì°Œ†Š1±4g\òßö àhòì!€·˜Ç*0· Ð¢§LGVƒ&ÈKñï*è•9ƒõ]² >òæñÔŸõW6ñ}–ÏN]Kn†ºÿ>VËàNf6-ÚBMÚB‘P"™•&•C^ÀÒ·˜xS§ïGQã´t°¯º\¬žLaMéŒûYo­;ÀÝ V­ÉW =’a‰NãM¡>|ÑkÊ‘PÈ3˜íîÔÒ%ï„(QW˜Ix‚ c¿Â…‹IÌßÜÏùÒ´òËåkº}Ap…P{ù¦Ã‹çåN“Y×#{›ë¦ï]z` i² ´}Îy=#]¹›ìÀèšnÙ8LΣü1U¤kÑÚ~ì²=ÒeYrœ>É#Ôª¾œ_Ã&57#.ø$’ŸŠµ”$nÒ/vOä;i^\ÿeujy&[Gç–{ôI<¦Ä·  ¡{«ésþ¤ £r8,‘êK@1µ)Ÿ¼MU‹÷viãùD1Ñ*·{—6‡H–Ø!ò ¬;¶ûGy°u­/÷5ô2—h¿]Ò¨2¤ÁmÓYÊØ2–ø£¢ñU|ø†–)*¡FFßÌyßÝA3H˜Ú?í^|kj"oMŒbÞ;Ÿr¼ÛŠÇ¾TƵ¦†ºÐ>¹æ/ôD¯ÛòÔWLßùuW´×µ…ús?pi¬#«±rykÛYÉ< <γíÀ(·w†¡±œŸ|K®êrlfž´íY©x|¨Î\˜‰tÍ!Y°ÓAüS2 cçƒzµÑyÄêOÓÇÁ<ÜB­)²ëqX‘Dî|š|Ï(*bÏ G’ÓeeÞ*£YǼöIW¦9Öí¹š9 §v”¬tMv·ÅÜ<+ï±°p‡HÁ×Âtñ(‘oœŠ’ŒŒ6Åòãø£…Ÿÿ½Ž’àÒ‰7ÝüØË\>,O•Ù/E¹¿‹}2ã»@VžûÛº¯à⪒ç6¼©î¥¡g„Ê$¯æÓ,â~ÃöÑn/Š¢SbÆ­Ðîøràí…wñ¯êŽ‰Ãˆ)‡T¿HghÒŠAªrâÏï9DK¯LSÞ1¿àLøßåPVèÃÁËEX*Žg±Ä[Ø• ß@qƒAmoS ±Åˆž«ÅhÊ 5’•Cò¶Çà-¤R/æ–‡ÖûÃ?J¦ù¶@<éxÜtÄ$tØúS6ÍÇ M5ï)¸ÚŽÕ5º¢=Ð{pV`Ì‚»­+~I¨^Ï:g@×^‰ô¼ãÞ.f8DÞ4Þä×Lt†l›$]‘ 5FïÔ}»šðF‚=,jäD)âi3ÓÀ؆îV©­K–雳¤œkìó+5FA›¾P1”XÛ³8-Þˆ¸f ­Êð-?®G?Ç oÎûâx»ÂCĖܪþ•1 ¹çëæïOréØ\ÿÊuàýµ–¤vçoúAxÑ€EîD½ˆÎÒZèÀm>ªá>׆Ž%»¨bkoJ!lì°Y É$"°¾àUßVFL|+\¾G¶­—ÜN¦/(šcŸàØTFm>æôX|HOo±îTÙÐÍ_ðž u½²Š‚¾¸,•~š¬ŽfWóç}8Ô5Ãøj 3;qÏ멟vû7ë3 _ÞÎräMÝ÷Y:cÚ"ÇPŽ Ó$w|ü¦ûׇ*’-A.Ík¼…xâ:»Žü3'²>äUž„HÔ•öŠëû¨.*E˜úÿÇÉS”køûŸªøæŠNëc±×q¶K®ÝYœaGì¾WÄÏ+´ëC¨<QþüññðJî»$‹ýƤ„ÈŠ¾Œ†sò«‘}×¥GGzžYœãwnÎБôˆjSL­©†ÜЭxQ^DyH,- ¶ÜÝ0œÖÖÃhd=|?Êò/Dªµ:h%I|%£øÏJpqÒ¢k3÷¡³‹”l;vÁrð¿Aò­TÕÔ…õŸÂ$­©—ªÊ÷<é€HÕóN­{)n!6zŽÏ¥ë» ñŠ8½‘»ºGáÁ‰u*“_j…±´ÍìNÆGœzä·P “¡Á4•¾ÚÚ¨0:q­¬;û-—tíž@¼„ò}*Ê—(‡^67mZ‰Еòªü°˜->ލNKi `þq R«8üüÚUR”EÁ!¾óM'N}t}³øso!…îJŸdh9l—nÇÿJ¾ÕµšO1ÛÀ™L)'¼ ÓV¼­kÍ>EGÌ`O †ÜŒ ËGânÑ­aĉ:ü’)*ë2!>Õ9–N¦©¼— 0§Óæý†¡r7™®|“1 Ã„J9›:Ï×û-‘)°)F´ô'ò 4,'íƒ<˜{©óGO‚(¿ž¢U%ˆ­ò¤ó'L´si‹—ýk„˜\Z1ž¥/)×¼UÁÊÌ87 x—²¹zؾýºûé¹ÎÝx«.F ¥^ìwQœF‹öÓ˜@œ^†ðÉHŠ~áÄ@Þ *[Jáho¯9ãÞ”öç"«®rX]º`”ÝÎ÷-[üˆ?¡ÞÈÄSgæáÙb[©¤¬¡@¨ÒÄ?ž(§ð¢”å›ÜØüW›Çò`üºVpêŽQ›V¹‹²±Of£ÁÚXÐ’{OBû· İbõ?^³gvuõnEyÖ1‹À Ú 4Ô‰iù< ç‹m+2&pÁî6=´ÎƒSQ>¥Ÿ éK'4FCB¼VdxÜÃÿ’jÊlW‘g.§õóa: ¯üÙ«Ì0€E[âóç…nÀÅ%Õù§ŸÄÓ8¦><ÈM¦J`âp4îW:4Ì{)¶u2y[¿K,>…¹ÕšÏu¦x×@8Fˆ¢ãÝ;¸ÄÛ8Ü¢÷/´ÂwZ•Ç"t¨µêIø/¢D¾’¡ On Ôìe‚{c[îµäÔK­ößFXV-¹òSeó]V 2Û›$i–P7":oWÜëP»ª¸Yõ¬ÊãE5ÕG!5椾pNwú鯓C`gS£Qà"®Ú…ž¼ìó¤Ÿ>5ú¤ÂëÅGS~ôbûÝ´h&ËýÂ%T»³(¤hµ“¢Œ•ÕÕªŒ›wÍLªJ}ZlÝ/+÷>,\èžÎÙa,øÆX¸p_ˆïëûìÙàñLñ£µ™§sý²RÁR‘ó+RßÓÍSl%Ãó÷cŠC_ùn2¡\J™‰»8Q1@—²s›×°¢`ær Éêîa¬ºî‘ª¤Î#<Ûª„c•fÀŠÕ„«pò¨ Z!4bÎ}Æý….å_îÅ8•M'Xq…ÑšÒ§»VË?eCÌu)‚ý2œÌ#[FÛ™tšäjƒ>(^ #UÅî»IDÉ`hÃ<”ÁØ6Xú$rðÀËE1o½zO¬ä¡æfÍÖ×uׂ¾z­2´DH|[q‡U…8 I9áÌ‘IÛßX¶˜ðR7¼ßw—䌾‹9ÅâÁV„V\ ²,4ù¡ÙÓÀ‹«†ƒÏ$¯±B‘S)!6gÈ_G·Æ?èlò ùC»æÂÒ7¤º¸p€FIiÑ}jd’J–£ H2Úªƒ¡aÏ’(›ùHÄh¤ß䥸ÎÊŒÇvÍ300»¬­”ÚÂÙħ„¿}KÀ®‚óóíaOð~ÌÊÑÖô¢"ž7„Nã[¥÷<ê€4x‹{ÂPÖ~ ­­ŽÊkÚM?<ê°{»±€ÐÈ–.tšŸ°á,u²'–qªwOÚ†k\úx%,î{Ohô\ØfÒmv"Çq1Ú–»¶L´±êL=&ð16g£6-òó`U> F: WÅ‹AbÞÌ«}Æ'žEËpϳ¢UâWª¸Ôœ m×ÊZŒ·ã’£”y&!@ô땤U§ðC¸8ÓþöSŒ^Q/žÀ -—_ÂXpW5uúðcÄgMu\ëˆyœ1©±°º»Þ’­úá ²Ñ°xÙ¾ö– béÅhÔ8†vòó ìd>!mQuB1 .Ü`°O…>)_êP@+Ÿ)蛕ZÓ\'FzÕn; GëŽ}°¨ |ÔI£#Rþ¾Ix9….ü<¾a3gñlVÈ Hð#¦=QÈñ!ä™ùªþf$i–ôºU¼&œª˜ëG­ÔÌ«qÉw ÿcDgS¨•Ïs[‡.~1κ¢ª|äß¿’ûò¶ûÇíìqÙl7bß0Šlö\tß4]ÍÇ7eއpaí²ƒùŸ˜a!·¡â»fí&ø½v â‹¥zTL´ÄµjáçNpx’ı±dÂZOêVO°§†ƒZùAÇß/àö< $9ð›kO©Æóš6÷Ø|\ì0µ.½ô›"~§&²¤ÖE2Bí…Ý: lŸ2¦žàPÔ˜¢~+l@\v²›óZØÂ+$!ÍÔeR˜rýêæG^ øÅãXÑÊCø]^‹ÍžDFÁ£Pm×Ó"äy´FÔcÜ3ßMáOÞ’Õ-5#•êr6ÖŠl\!«‚ÞúÈ’_)I®>×eÿ4‹ŽÓ˜XWš(c§iD„uýU0¥{ླQ\f¡…,ACªÕSm&+1.JOªèÝXeϵ‘!–™·ïMå R Æ¦¡çø)ü»Hº+,^*˜îˆ˜ú-¢ÏQ¡» ]öC!ÝWxC˜ù}îIZuÉ+ÍFÅ @¦m]ã+Iïë~oG]–üxæ¨ÿ‰gäFÒ£ù¦x\DÌÑ>ý\K›ºKSö§Æ¶ç©Âó‰IêØ™«ÛbwgªÞ:”€Ÿ6Ý—÷“*ÚôUaìP¥¬ Ë!¦}«H æÉ‚Û0‹“•ç}Ã9PõnfµöëÒ-ŠAçå«JÚ=<¦9â"m …ûHêXÔº.OÚJ &Þú†ñª)äˆ5wé†%%ü £+¦KO4Yô£=DŽŒÇE?G— ¤öј#J^Ïêc6&s³"ˆÑlyW£èxêù7„{‰&“Â|³¢]s(?ØÞGÔ…æ™ÆÃª½Œ‚Œ’È0‚Àƒ/¶§1BCOë *•Ú²A?^iÅB½Û.ðop"ªÙ—w´ ݘ²ÈÎå‹Zm½à½tg½œ¤Út³ãÃO-Ik2§×|ÄØC—Y6¦§¿Ï¸ƒo(Œƒ¤Æ””'°õ—Œ‡If<œT¸½6­v°x.uôŸ'£õ§w>‘®ÐýJ®†øÈ·íú ó»u’þè{Îf]npJ^€=•~Ð;eõæÙòJËŒ®@J­Œ:>Ê´º\I[ Ÿ°™7š“ âðÎ(•Df"vç‰+^Æ3Ç(ä·ì[KçÆÂÐÐL!™†Ê7ÝcZ *›ìCzŒý~ 0Þ“H” •&2¡ºtïë/”8ꃲÀ¶Èa˜åÝxJë—³ê`ÖCõÕ©¸òÏ—Ù“}ìç-zdD(¤WADÕÝ.¹¿9 F”ÆL0¿ß%v:&K2«WãˆYž5p“‘4ÔøÄ…Qåk˜Õ²m„äÃj:ר¹¨%nÒ;R®y¨Ñlf|‡ôÚ¬nÖCʬ?>ÜÉhP¨$íg¸¹iÀ›~v¶ìà7>ZLë¥KZƒ~µhÉ[‡¿¨d%4ß‘9Þ¿^æy$*ê’nùåâ|ň5y’Áxä¦ô¬ƒrJò@viô_Zˆí´–A¶JÎZbaýÎnáÚvÖ.Ï\ãs²Ç3Y¿{Íç”nCï™PYa|a}Œ5~Ò^h2V6¥K‹Œ4Ü–X™d½ªÌº: Ý2 õÅ¢æùWMÌŒÁ©)š–½{€üyÅTßã STì³}£*á y>œó¯L…ï¸F¶†‚G›Ø¼§9ÛSUÊâ³ÜúÑžMôxq!¬q-Š¡s¤R¹tŤd¸¦èÕç¢u( «·0¥O•kQÕUõ: ¡ê¶N^†ºý>YDÝ J»Qy3ŒŠ€ù2‡Ý»3"ŽdyÌxÑɰ­¤U{ÜÓÐôyùœ!Ôg-^×{x‰1JV8uÉyLL èñ¯ÞŠó[VòD~üÃȱºbê/ëÔ#Ú2÷¢úOË^ϜӾhðe—ávQßÐí™»?ÏJs%Àe%ÞP£§»^$ †U:OÀFa>&’RÛø‰1šØ[„Kü{fP~È…ø”{söXµÊÎØùàXíxFLҚէ禋¹øq'^ŠêНÉNøºÞNM³ÊÐRºK‹·Ë³;RÝW&‘ž[ãqñv\ŒõA®üÛ%i¾ËïôêÅ2²6ß½ âªTûœ¢_2UWÓv?|Ü#Æ'ÚPVx˜u!2~º­ S†’\coøA±hgCÍÊÉ_4XˆMÌÂ2µú/­ ¿\ž˜MvàXZf\gÁ¬œÈ­Khòáʨ°H"¨¿J†qœrp犗÷<•¯1vm=Æ28ù©QÚx8R/‹ÝÒËA«Ð7X8µËÈe-w fcñÝ0¡.ya2t“ä(×îdñ&)Súuø¦qtwZ~aÐÍ|>àI`_½’}°Nps<šqA±þëeÖªò˜þèê⾡ü5Íäz¸Ë$Ç”ÉëœZ²•ûo#Z ÿz ÓCŽÀDp¦E|ÄpÈžý7Òº¯c––d™ê¯¾C88öŽR¡«#IWV‘xn]È׿«Óc\Œþ–MrMã“ÆÛÅW_)“.OÏrY»à¡,˱±bTˆ°ƒÚm¢›+ÑIõ~»/•ЦÈ7ÑYR‚ý|þအ÷{ÛAÎáw=œ„Ï<ä69|½ž! F{¶I #—ÇÑœ±KßÓ?Gc©ŠBO£Âs4›B…Éâ¶£5â—†&Á 1Ò4XÛ¹ô¹ã`ÙÀèÌAg˜t¥›žÚeK̵{¥­¡PUÀ唲S¼ž‹—´Ÿ†AÜ}«¾¯ÿÄ!e0mcßÌBís°ëà1¿¡þML|Ày€Õ­”®d9Åí%¹ø§šçRLšzá6©Å_ØÜ•&R«ÏB’Lœ†¤Fbäw½)®×~£L¡r1ÖZ…lö8aFUs'ã™´ôô¹¾¤*ý`·…ôf›±ãpìÝÉ­‹%WWç^|ïldrœ¨zÚH(£¼b:ô ,Å]v гóOW\K’ M,„â šÃÉ)fã¾j.¢s®:¤Q&´Jê?øˆø_Ôe{1äüxXð<Ðî_#U¥ýHþÌN"«FÛã4ÓÒÛúåá+›²Fþ¹ºõÆnü*¿‹ßõ<Ì<´JDè6¶ê,L ¢LùÄe§³x]@CÛÿG7£T endstream endobj 50 0 obj << /Length1 1629 /Length2 9340 /Length3 0 /Length 10387 /Filter /FlateDecode >> stream xÚ¶P\[. Á]ƒKc îwww—éÆÝ]ƒwwîÁÝ „àN°Gî½3“™ÿ¯z¯NUŸó-_kkWÓP¨¨3‰Z@Ì€R° 3+@\QÀÊÊÁÌÊÊŽLC£r±þ#F¦Ñ:9ƒ `¾? Ä€¦./2 S—;E çj`ã°½ãcãæce°³²òþËâÄ0uY™r0ЙFâàé²²vyIó¯O­9€——›ñ/w€¨=Ð dn (šºXí_2š›ÚÔ!æ  ‹ç… °vqqàcaqwwg6µwf†8Y Ñ1ÜA.Ö5 3ÐÉ høÝ0@ÉÔøwgÌÈ4 kóßruˆ¥‹»©ð"°™ÁÎ/®`  à%9@]V ìÿm¬ð·#àŸÙؘÙþîïß@࿜MÍÍ!ö¦`OØ ` ²”¥˜]<\¦`‹ß†¦vÎS7S©Ù‹Á_•›¤DU¦/ þÓž³¹ÈÁÅ™Ùd÷»E–ßa^¦, ¶‡ØÛÁ.ÎÈ¿ë“9Í_ÆîÉò÷ÉÚ‚!î`ï€%laù» WM0ÈÑ(+ñÉ‹ù?2+  €‹•••›— t=Ì­Y~‡×ðtþ¥dû-~éÀ×Ûâ°|iè ²¾¼½MÝ€'W ¯÷ŸŠÿFÈll ¹ À h#ÿ'ú‹hù7~9|'@Ÿõ…{lÖßÏ¿¿ _èeÛyþÇü¯óe‘—‘Ugø»ãëÄÄ o&;+€—“ ÀÍÉ ðýï(*¦ ª`ý«,Øàý»Ø—)ý«`·ΟöŸÝ üw,%È iÚÿpÜ€•‹Õüå‡íÿ™é¹üÿüw”ÿÇÿ· )W;»¿Ô´éÿ?jS{ç?/œuuyá¿"äe Àÿkª ü{g WûÿÕʺ˜¾ì(ØÊîßc9K<€* së¿Éò·\ó÷’ÙÀ@ˆ3è÷­`bceýÝËf™Û¾ÜÎ/ŒüK|YœÿN) 6‡XüÞ0v®wS''SOdÖ"±sq¼Ù^VÑèñ‡,Ì`ˆË‹ à¥=_€%Ä ù÷‰rsXd‹þFÜÅ#ÞdúÄ `1ÿ7ú]+‹Å Àü¾„µúrX¬ÿ€\Ðð%íÀb÷|ÉkÿÈö’ü|É ù²Xþ€/‰œÿ€ï,.À—¼®À—¼nÁÿ𰹫“ÓËó×¼Œÿ_ø¯ ôš#//@ÌùClêBÚokD‰Ý™v'givµÓ蘼—:\¡#¤ÐUg}sºMîÅ\Ý–¤½Y!ôþù¹!¼5IµíÞçÁ8Amz· yi op²è§hý) “†ÈwŸGG­@[˜ÏÐ]r4yŽ®<è*8·îýÒõe_ÇÂvU¿W¿“Gy(›aŠÕŒ1,™£É7Ëž' „wa"E¤Ç>õÀ˜»ºžÅÎ|&—K`@ö=Œå(öÖÛ`7ïµV¡ÁîÜMHM¨G@ s…=6ýÆ[lÿƒþ¢wéÇhÌæQö‚G¹lÃÃHä%9 Z£Ý€>o› ù‰Ãë¨7À°ŽÑï7¾ðˆ§^Ríâ ˆÑßµCu¥=z[íUI ÉÃXÌæ¯0õ®»|nç]ä¦K2ð@fhc¬ëŸ£Þ¿vJÅE¦åÓÒË)‹÷etõP_*êé<µ*ˆtêãiÓÅÒE?ø¦2+œ"½R‡ ÷U%»c-³ç{ªI´1ây{!v,?º}ÀmÌ‚”†âÖý¤Vq©¢Ì«MªÅiO=éŽdó³ ÍÔ_+[GÞllüC4¼„+ïÊGQïˆ[¡Û0µÐ£;ܽo¶µç®8÷¯é=¶¬gÚ¿q†7’ʺp@IÈ‘àY8ÕåÐÂßwËø2]õ(KÆüú™Z‹Ð¬3T”êkb|øvð¥U»3&øŒQ£~žÜì ¢É:óˆ†%<ñ»_X´¿HdÕx¡DèÊ ŸCHºÕc’nr­L'VÔŽ)“Ní¥²éüIf7ë1ˆ*zE¡p<ª%A¤Ì’V‡ç3 Ç{™#‘*Î{ü0öVµÕíYƒÇ+’êqéɗݾÔVñ"–Ÿ•ÀKè™[¦ Ÿu{³Ø!5P4íF=Ô²[¾µCyšU,¸¯Úo¿[í‘u¨¾8OF·øžG z}2$¯„Y›·»H0Dy,NJðŽ’º‰ºÃt[fFãÅÅþò€í'Ò‚mŽ™n8n(Gô`ô}Gƒ 'rÞÙ’;èt‘L­ä*ƒO©'ÜôJÒÂò$<:ªÀ™ {^bs’ÉÛÓ(Eµ7yÛžƒ<À=^]Ò•¯c?Ù‡ˆ¯Ûhõõ•”—±¬{ÞÈ’e€Èý÷ÿÊÔ8íj©Ã0ÙšÉËâ»wXÓÑÎî“É2ØÊô¼Š#3”¯jóÝÄpb®ª6…è,K>€ }-¦ã"ÅVê‰@Oà\ŽäþäWQ¥uß×;¡þŸ ŒŠ Û°½yRMi&+ŒåkŒ—9DMá %ø³ ·bû#ð…l ÞkÖû^åCŒã¾¬¹zyP ,u¢oåðåúwÆ›b±äWyhæÈ<ñq‘ …#š!™r1ĵíÑÉÉá¶äp0P”8FDŸ”$×¶ Â(«eÒ›°ÿ).Ã&B,¥,„‚r¡ñÖ°¶ùx†õÃèÊJ#¯Úê–ñbOt¼´hÇõ1:FBañp[J©{‰¶®Fïú`ðvÀ± ŹÏÜ„õ=ãôã/†Žµ'W¹¿9M1–uÆñ:· £²ÇlxóÕjš†kDËJ§|ŸªjŠïÛ YÍb¨8?üR´€Ä…”žOƒ3­¶k‡ÕIbŃ@·ä{pÙêàü¡vx¸E»#^Ó\LÅ þ†«gÈ—Ì ÿѤ‘F²Ð˜RÙÒãê^)[~Ô5R®…™Š}Ÿø#ìÂÍeß—€]ú‚Á`í,VùQ‰0ôr”ú•é©Ñ‹Uùà“ÔÛ[xîž9¦üOÝ0OÊ–™btµ0šÓâ–’<Ìõh²Ú¬Þë[_¤ç>‡&ï#5âˆÕÞD‡Šá¨:PY·­Ác¡tmÉ ¿êR¼¼^öÑ”ˆZO/Ÿ¦?¢X>í™Qñ°ïÌ©q";x½Í3É¢3¢Aøš»C-DŒÅÕýV‡=ðÆO{D+ºÐltc-:8oèéf‘kð—êï²u’Ío‹ÊvÊ㨲ÕÜi]ÁŒÎÔ6OŸü GÇv`bxŸ©ÞÔ4ÊA_Aq˜8¨Hãâ*¯·ò‚Ꮄ’–©Œ˜§ |¿<{¿ÒÊÌ”/#°Aˆ8NˆÉÓ&!·¾¬g¶Tï/=ç¯ü¦í&(tx@wp[pkðaQôÍ¡…øUËHÊ~5R°ð8„<ðׯ[ø˜’ Û1æë“û]‰´¬¥kœ":‚ÓæçmÏkó¤;lÄÄGj¢GÖï´ê¢™JúMyXñC²›øZ饀“6Áwï\bÞq" ¿›*8à‰ì¥÷¬LȰQ5¶C/,È0¦ÊrõMƒå]åµÞüÄ«š†ór‡ÿ@„­5"Òä¦s@ÐèýfRði܇KÓía£éˆº:}\M¯0Ú.}*ÈÿˆtәǮ °î›è(Ezº• €©»ÂÞyæŸýDHàöU3\…;èh?õî Å“y·çæœ%‘Vá<|Ë[ 2L`PÁ2Í&Y®v‚C¶7«²Å<µHå{>úå8ÖrÀ•Ò¤æ Þ¥lUbÛòƒ0\Ä»Ök]OáìÒÑåNŲ\B7¦í£jY`ñ 1L`sª°pƒj¬]¹´>eö¸8?u·~)|™ Ü!m£êL16râ WSغFùE·åä]Ëe†ù˜D8¯l¬âlÁ>‹x ¶²05E&§±ºÖÑ›ñi-Ç®ųObÊ-¢•‡‚º­»À’}&z÷õŠž™'?+”§ìíU µ–XŒÁ9.Ìè›`ãëù…Î-ö¡±ÇàÓìwõÙ™;W¥¯°¼Ç‚XC¯Ú² ßÔ~HÖ ñ«˜˜éÖµX%T’In@rCKs¢¬ÔÄ¥ÉÇf IîÙíÕè剖ˆgÚk²I7å* œEM}’Z}T>ø2üÖ\˜Ô¦IÞ íÎÜ]ƒÜËf\'¢ dE/G~üºyÝ jm?M³­ÔlÑk$‡ý‚ JÆleoMg’™@¶\ùSvZ~{³E û"± Ï«Ø?û=ã¸Øi¡ÿÑ¡Þ.p_§¦Ô“÷Ž* zzß>ž¸®BRh\W’'^èËeª¦…ò+¶ºþ ’­ûÁ•jÛùÁE¾¯f(õnNꈎŽìþCŽÈr¦GMH MÚð€»í8é^ÆëOß°“K“Ía«*›à/aE„úçS‘Õ^gø ñvÄ…‰«·)v×¶^íñ†¬ ïÃÁÉÉà¦âM I+krÚXT¶×[ð Kì¨l‚zšË9¶øRUšR‡”ò¹So^±[•— TªÕqÕÚ9~ j%ᅧ9 žŸ[œÛ|å ™Ù$ÛzBäû¼jþi)-tÊõ  ÂÁX8§ÇÜM'€iÜÏ&´žG^üEàcÄ#>é¨á™¦Oú¥Xe0摺d°û· Á¯pÔ6ÏëšÇoY=Ÿv$àúòíbámLò IñÐu<4šÇÚnìz`¡Ÿõ„@×x#´J‚´ãÐîYÞP1l³—¦u?»ª€*ü¸Râ©}zî8î·>x ~»êˆËŸpÓK’T†o¶Ú„–%øaóãBgš}«ŠÅ‘·É¶a£îHLká^K/¬jV.Ñ`úöb«U-ŸõR)«t^\¸ñó†’j ­$it£…”€À bÂæþ•O; ·†ëˆ¥™vi”…ùvÉó@`TÉÑרC4•æä‘¹›ê5iñ¼~<»ÂdnÕÏŒv¬ŠZáÈ‘*£døX˜´‹i¿p)ÖX Q·ºƒU.Æà,%>r„+ TàV=¾¾ÂÕD †kä€4¼W„BM…ÊþI 8óÑ¥l„Eú) }ÁHu9¥ ¶Áò±$Ù&y­ñ$ä׊C_fœÇ·-rõX{wŒ ›£Òí7þŒó˜Öä~!ÏæC½[ŒãfÓ'TŒqÛ`QÅ^¤Ú¥ÀØU[¬PT^…Ôá‡ï¢ïs›EF¯åÜ®Ýõ3ÔÆóŸß«{± îP»¤JÆ–Eáên¢ÛIC þ¼-HOõÛ²Ž¥ŒÞK¬ëÒª’ÖŸ×Ôä¿*Sydž´ý¹0ŠÐmoc¤™3¼ÓÒUžlíļÊÀ1“-09ɧ (ؾÍÊþ€M¼=®~}QÆõ¶oâ4ÐÛ@ž¡•Á± N›µ­?C´2¹´q.?ž°Þ!¦QPVu‰G$TÓE¹Nû nl€%É.Úp%±‚Wá4PKV¤*¤ñÛ¢ú1øê§òéQP¡Ø§>=¢Ö÷F_.»õQL¾ÈÄV®!'ÛÜQÁdd¥ƒ‰ß⨇$QgàïsèÌøMgß^ÄÔñt‚Gí£q£[›é‡ŸlP!Û¯’]­Z'À™=¤wÙªÿdÎÜñ’…zfüÐfdêäÕTd‚þìñ]^ù»’9°øIÿ·n<ãÞäA’zÁµÆG{üöÜ Ìº‹D¶¨E`e…YUð¬on2¸M*,NÖùñ»Ð‘kE-ZÆ <ª½ "Ž·¦äö-»ÈÎÅZŒ@ :NÄÖ.\2d€%ïqž¿tÁâqS}Fnõ2ðÍÙŽ+PÖßi8—Y˜ÿ¬M×ËÙ$ G¸¿6 3á¦H˜¨ù‰ÑœŽu'‚!w@âõ`ä·KÂü ¿×Õéç ñá] “GUwËR‡Re_[îʨhåbEÞÂøNõ ‹‘Q˜<œÑvZ>*³@ì?ºE^`ØD6Ôù¦€?rŠgíWŽ+›^s„ÌFŠgVO {îF “ªS¡O7#¨ëòÅú~ÉØLjÅåÓÖÿ*°“ 'gèÕ²å{gâÎhIîÉã×ÍþòŠEìÚ„¿."³ùÖÄ~³ÔMØÔ!}çúK‰$J=¨–_ Hßš'F­sïwG…¬h†£"z|-EÀÃ|»ó†Í\8Œ*Ÿ"jUL¾öòipkÙÿq1œoNšü«AÂ1yë'üƒb!NٜꡜæDF•‘Fæl#jŸåùp]—úª¤ã-Û±Àª 9Ó.Y´¬Žéz6r<'{fd4È#ÿ_¦“ˆ®²— Š®¨‰ö W_U;P¶´KÔ¾¥Þôž}N"ÑÁä(LÒí–÷•ÎDÅ匋Gót]ì/¬V—·#wþ—›?†}ùzÝݯIk‹K‘õ²kÞ›k=@ÐІ8Ö›ˆ¼ÔO|1a%¼! ?›¼Ó·£EÛáÕíµë¿§rQF?:Iº-ïFPý¢%Y±>ËÏê×è¶ÓÀ”ÓijåMï •‹ôKýÅ)qæ8ô‰!ÍÓ¶Ð’ph…>èÈ/§Ú‘Bõõ‰‹#OJ[·ëʕܚçøAuâ€1ýp&ˆùÓñÈ´¯lV4*RðçCvÐ|>˜GºŠÇS®gü׵Êuë kaÕ1èZsŽ;#1¸ÓNqn=ÜpY&žú³Ý £DÍv2Ê`Ž8.0s³ýÑ»ñÂX ‚; òfù"üÔ*z`dƒf+Dµæ•T)˜2ÂÑÉ”>÷«_ÖVÐ[i§uêhÞ†mÉ›ãØ&¼CM¢—ä "ÄåTÌ!‡ˆ¦­Yæ%Ð]°ù§¬e‹4©< xÙid‰Š”G0êr8ŽçöÉÚ5iY"TÓK Oz¡^*¸¢eˆ°a¹»ƒ†M Þ^¡º£°3:Ìå³wÝÏc&í<ªI ø{ýE²X#äùñˆš‡05ÃK ‘#g¹É®qñþë.òNXÎâ¯OI®ùw’s®W²ÓdK +¯ÉåÞOÂŽ&ËQÂÚðFCz‹@\h‰— c._gX”ñ¦t±¡ž3¾³{–jr"®xô2Øøþú[™B[ƒgøÝˆQ¤,é´'e ,Þﯩys %±…²ñÁ¤V‰| Š9îY¼;íÕêµ^’PŠ‘<ÒåF¿zŽ|:¶¡bå&øÓÇ-¥EÒ^Å~!íhU{7Ìö’Yí_TÁAÒàñLìQ5û“9ë7Ãߥ  ÆRTçhÔ12ëO`h„PN»¿|p>yŽjáÏe˜1GÚwÃÜžÙÌîÅDµvó<Àܾ—ì Ç{”!–JÅ5æº6Eg®72’éu}ªÍèl ¥8j)@áÜà¿Áþ´8€Gšº$!»‚<ñLçð­À«3WË‹Ño\”´"^Ó¾H|-…6¡Izò3@ˆ«¾ë=‹®)+Ó𫪒Pºì{Wè_ASáßo=mýOn¿œ¿É•/›; tRGá쯎8Ùr5´&hT«Â1îX(E<º‡­$x{gKêÓa0@nÉwÈò­a)ÍáªbzÁ;iLtI1ƒqŸP;y¶›’ Jyñš–8[òº7X:jÚXÊŠ{Ƕ±9ÂQ¸ŠÐ°âˆY ql\ÿLؽ‘¬ˆN§øÀ~ S1®³Ô”¼/â1_kÞ¨[»ºgqþåLJ0ÉÑv{p°Ø¾0¦]è»È¤ÿ0à罋9p8FˆÈþƉ–4mywYƒ„ºõAý¤ˆ>>?M&p™d<€=;jÇ;*‘-ª\Il˜úvå˜ÿg­'kÄwNv[ê~ÇJ—6áPF—ÚkþԋΟHϧH$ß|%ˆ½6;é9ƒmÖ^ù|S³<ÆJ[ÍN™ŸQ·}+¬ÉLæVן W"ÈË-!>ƒIr{],åðÃ5Úù‰,Ü2„–Of¶P\=µ+¯õ=*Õ Tèw™òOIDO¨3Ä +¹j—~‚#H¹‡·ÜX§„dÚ¸e µENQ[‚ó>E÷{­÷‰¤¥š¤oTb9) æFm~¯‹«/gû¹Ö—4¸áé÷ÞVLjC:êfó©˜Ûlÿa}­}éž]ÝO|WÞé~°XDc‹Ù¸yä§ÌÆ žþv%Ùê’ÕTö¦£`ÕïU0lY…Šg€‘X+Ï4À)); Š>°]$Ú/…X5öœ|–æó¥j=“‚Ø01Âêƒ@?wðmöþ@S¬’~Jõ¤*¶y}  VlŸr" ›×ÀcjvëzoŸò$èu#{k²;wñl~PO´¢ãíK‰‡Ò»¼Ù ׇ•}h㪱˜}ru½¢h@Ç|üa°—±è±‚H8¯£Á ô$ô9NM³üÔ^͘—'Ù³¬Wûëyʵ9Èa½Üaç²ãx9Tú°ËG5¼¯ï±Ÿ­"v ¬d¨ºÞ«¬cÐáLióJÝÛÓwΧ:âð*î_°Ó+ò ø#Øt´²¶#Ê_ÜlŒêP¼Éý™h\7¥Õ‚”ÁĆ)îÕk ×aöκ¢Éù@öN¨#f¢×È9ë)ŒÔv"wCCü­ ,éýg“i¦ÒJnñÎcæl221ý4¿¯:jçËs4ös)+8¹@Õ¸ph Ý.,‚·uÌ'¥++`¨&µ“oÔ:z¼(‚¥ùd௯Ç¿„ÐNi¦ÉvékY[œ}JîløQ¯–ˆ %’:e鋆0$ˆ|º…mPpEroç:é'©c߉Åãéé“ Ï–´›Rå0F“ƒ>êNhyàÊÖIœò·þQó=¾ËÁYŠåp/ßo—¤¨G°!ïT÷­CC|Å0‹™˜K±e—°¬Á<¾‰ƒ ÞoHöcqÝ*?³ö]uXÿ}ׂ}7HÛùuAƒ!^Š–þ¡ö®X†}ÓI Ϲ€6.‚K„«rÛ¨»sz%Q¶?P® £.V:©—?߲啌î ÛÏÛ{ˆû¶m¶¼ ùÃiÒÐXRì'²Î%á·c£>búßk>Ó!+¢fíU€ šjà/Û$³Ð%vØ«SÁaWƒ™û¦Œ§ú„îÇñ{ EØ„ ¯:&XjÔd_ jÊ$bžiêu¢ÛSRÝ.Wí¿Sí·ÃžR*€-Zßk'M¸;0ðæA?5>³?‘S bdÁyr]Àê<ÈpRÐS“m:oÐpÞ))ŸñˆÓÜ×a‘N?‰Ål?¼tÑ,)…u–ª·|¨N{ŠÃ#t© … °X3@\‘O~\ì AãÔ«- š“1Yß_c‚]ÄvßàûWâc)Ø—{×ì¸QT RþTxòÒ( Ÿ@qyµÿ"©· NA ¡æèâ@'Õ ¡ö½F« Ïïp0 =BØŠ¤¢Ì nZ„B[0~3ëEÀ #7`ª ë±4øU ½À9¶¥ëtW4SÚþ®aožŒ«U4¦Àˆâ-a7ˆÇÅä$Ég†¡„qZ;1ßÔ¡ %e̲ÂÃÛTÀ‹'p(|Sk»eÞ}IOìú€£-à%-êÊPïÁü%¹-Ÿý{ 9[„ËÝ×ðX­È†ë”_°‹YƨR ½Êõa¿ûIdX%v+Ì#è#© 7’Ù(^c@ÜññéS9O=ãVsj‡o´èbÚZ3êzd_ 8i#˜7=uœÞ«‘Át-§ºÇêÀò¶Æšnv¦Ÿ÷Ú5áñ¹º°C£ö«Ü•°4Xæ¸#˜O |#Ä'xïÃ')QžñaýÙÔädùŒöÜÞü£°Jm°šã¢>äçªFS‚¢JK9Еÿ˜û†%£1bë†pZ?ZlM‘¡¹;2¥WŠa¸öLN óP>Bn\†bÛ>”k+eò®8/ÄrÛnà¬â&‰^JædënF±ÙlDc)…n/yŸ4hªðÈ÷±\!ËÄ>}_Mº ÎI#õ䊚J0B2ÛºlÆœÓ∇(f`k ½„å!æq¸˜3»?ñùÜæ€SVèº&¾­7ÞË+Ð63"AyíÝ®\8E´¦^ÇÞYX=BkJUàõ>–{“öUë¸û¾â§*Ÿ¹ôœ[œ£*b)fG'XU÷«|{ÕA†>¶fçÂ}£/‘.ÉoëöÊi™ã¼–B`,Q÷ðRJ̺Ú&™^7åû •G¾û Mãhç=¸FQhë!jÜô£ü² ½[r©BcÁ?-©áA/ÞA–pmCè¦þ¦¹–€lX…%TÙ6tEä9Ki.FTM$CµT0ÒPÜdu$9øH““ù5>ãLÌNm¾>*zžˆ#¾‰8@‚Ð8ê^¯b)R„™›áÉx/{gÜä3Z¯í£"1¡³±ém:e–ø¡çAÒF"pîn‹½Þh€€ò…ƒÅ¿¼÷G:~+7o\™•¥Ó,Ê|‘6ò×ËêOoŒ²ÎXñ[ÆÈy×)R€Nî:#ÏD ‡S~¬×(­GIHžž5ÅŸlb…±¥Wý½»j;‰KGuA»¾0NU ËÞ¤7s%Lr*š+:wÊî=À kU®“Y/ôÝãšØ-çw•™Ÿê¨Â2š²F_<"ß9I%9ך°BKØÅ'=²O¯SãÖ²jŠ„ÃÝ;wÎ{…Ö¬3X×öW)½c˜ÅìUNsº¤<&˧Q4éîÜ1œàšdÅv¥|T3öðEvË2K÷Ñ'P`—F„¿–P:Sm’‡¨MTtñ³­6ø˜@=Rôrø–*%^ʯ¤ó°:Æ©W[˜kð¾&av%ÐQ²_q׈©”y=mñÞ‰–±Ë‰s^í1ÇáïgŽ^|@z4¬4ÆU"ÓÓ‚] ’`fxóÓbÎ}cÚ¹¥]QgŒ°'íVZý¨æ•¦FŸÍ¥E>’Î9+v½$[me;ã¬ì .R@xzoP}1NupAÒM6ÓWÁ!b \Ÿvháfx‹wHÄ€Q>o0t<›~¸¬Á\{m£J¯Èò,‰ÖÂxßl °%˜º4§ˆrfe;#¤ƒ¿µ·[ÉîhÔψ>Êû«°ïP#\/Nm!šÓŽýzÌó€µynˆ3žXà –½Î²(ö3Qô¼Œ@b•ZþIQ £f˜£ÞØ zùýÙó]HþJñ)žÔÌgý@™fÊ5VAï‚zͰ¤§ZÞTÿfZ½}¹‰|ÄÞ²ú­2ŸíËЛµŒ íöHÛŽI¡>­ÙÔ¬1ôw OkÓHBrÑÇ5è¸Ú‚þg]¶³º›a¢L 5åHµ:Å€_¶)É{AÖTh1I¹g-…õ=ýf`ÐZR©‚Ê|³ûà”’PÁÎáÃûßÞºgD©õÑù‰–òJ2 Ìx(æDn®ã¢T†¾iù5a2<{`Iú¶*åÎ1 ºûG?lÝ*— ôçS d t¡¾ —‰yÚYGiy,¢ Ûö¥êô•þ÷q”4“{rÙ]é <ËxØ”õjj×*…Ú¶6_Ï܇·O¾€˜â=ôüù²z£Äóði,â‚Ñ»ÒyÔ}¤M ¸*9ñs¨s]¤BV"ñuÑI,­ ­mܸš» ×=•ÁËÇN½ÕüSتÓè|ê,M¥5ƒ˜¡U…'è(“~ãZÆbÑŽÒ§:¥QÔµ©«Î3kò|Dy/öJ•´B¿ÅFMóÇ´mqç>QµŸêô?ØÕýÀy Ô}uYEµ{´ü„p:F"ž8JùP¯b'Ëb–w¦$î-†g;ŽÖ®>yJ8”± FÞ¤oå+wéà FUQfƒ9%::sôö“Í oß±3s¢)QÓ­iËT 2ÃùÀ>U¨L›Æ¿£µcÑD$QáH€Ü'êÕ–ù˜]XÂ1ïÆåœë“¤¯·”¤ç’í‡ß2¸É¶X¿å¿UJ³1xÖžlçr#Ñæ¿S#å«£„>( )j Þ]\Tñ5|°ã«Yó’‚E vÎ_e‘œ½±Ñ•®ažÔ1l^}C‰¨J œ`N#šó*Öîÿ”Î;† endstream endobj 52 0 obj << /Length1 2117 /Length2 12185 /Length3 0 /Length 13473 /Filter /FlateDecode >> stream xÚvuX”k×/%-ÒÒ äÐ)ÝÝÄ0t3tww§´ÒÒ!HHH7‚4ÒÝyƽßo»ßïœ?Î5×5óüV¯uÿÖý …Š:³¨™ƒ)XÊÁÂÌÆäˆ+ª+hh°@  ÈŽJC£a±ÿÑ Òh]¬ìùÿe#î 6@e&¨©¢ƒ=@ÎÕÀÆ`ãægãáì@ ßÿ:8ó$Lܬ̊,9{° *¸ƒ£§³•…%šéô 76>>¦¿Ü¢v`g+‰=@Ñb ¶ƒf™ØÔ@V`ˆç… ´„@ùYYÝÝÝYLì\Xœ-„Þ0Ü­ –5° ØÙ løÝ6@ÉÄüOs,¨4 K+—¿Uêæwg0*°µí] N®öf`g4?@]V ì¶ÿÛXáo&ÀÆ`caû'ܼ²²ÿËÙr°s4±÷´²·˜[Ù‚ÊR ,ÀÄÞì·¡‰­‹ÔßÄÍÄÊÖÄjðWñ&)QU€ ´Çÿtèr¶r„¸°¸XÙþî’õwè %íÍÄììÀöÔßõIX9ƒAÐÉ{²þsÄ6öîöÞ°¹•½™ùïVÌ\Y5í­œ\Á²ÿ±‚ŠPÿÈ,Àä…,Ø öY²þN¢áéþKÉö[ íÃ×ÛÑÁ`mìke†þ z»˜¸gW°¯÷¿ÿPÙØfV Àlaeú':T 6ÿCYàlåÐBIÈþþüódå™™ƒ½­çó¿šUZERGB•ñŸ¦ÿQ‹‰9x¼™Ùfv7€›àûßTL¬þSÈ¿\eíÍ|× ÔÿÔìö"ÐÿgOÞþ;–’”À`ý¾¹€ èÛÿ7ëÿrù‘ýw”ÿ¾ÿ\mmÿ² ÿ“ÿËÂÄÎÊÖó?6P »B ë è] ûÿmª þ{‹ÁfV®vÿ[+ 1®…¨½”ÚÌ|,œÜ‹­\¤¬<Àf*Vå_4ù[¬ù{íl­ìÁ*.V¿¯3ø¿tÐ]Ù@¯(;ÿR¡«ôßY%íAf¿wŽ‹`âìl≠„’Š‹ àÍ]N3°Ç_|°²Ø;@ .h‡¾sgÔßGËÍ`ý-úqXÅþ «øÄ `•øƒø¬’ÿ ¨¥âµTúƒ –Êÿ ^ €Uåâ°ªÿƒø QLþ N(rYY¬œA®väÐè¦9›€lÀÐKÝòGÎñüo*þ£€–úqB“@ï-»?i«Ù¿ 4øOÛ¿‘“+”¯ÿHج¿_PÖüq‚†µü“:cKOGKèÍúÇ*³ú„6õ¯˜ÐlŸú=´ê?#`ƒî9ë¿‚AïV‡?é ¶ÐwпÔÐÿ¨¡¾ŽÐ—€ýÍ‹“í?Òÿž”ĬŽÐuuø3hoN®°™©í¿r@¥ÿ´½?åsBÖÅÖÄÅò_ÐÀÒ@/Vˆ¥3øOÑP¾³BÜþåáúB¯5VAhÏ?ù Óô;ÿíý_ëru†öùëfƒîÒÿà¿ÞW`°„º0ë±® i¿ù$Jìμ5ÊÑíyõìJΉk »Á÷U‰À¦Œ˜³à’75WBeØR¨Z^6,X2–Ç݆O¬ƒ 'N&’耥^P‰ý™~üØC”Cþ~Wܳ_昢âJ—FH|[=ÌÏŸôFï׎“2"1YEQŠòÎì>ü» #e[?U~ÄéÁN  uPx:šÜ—Ɖ[:9=´Æƒkœªlã™Ù^±—\þÓaÒ¨ºSbÝ2þΑ/ìw–ò"°]L%Qùt¿Î™J¯à“ü¨¨ß:ó¨â·E\¿¦”JHÍް·^ÊÆ*:º M3„ðÏ¢gX£˜»î—XÊÅyåñ”ípOÉпǞüb5ùݘj á½R´ ~X,u¸ø¼vԘŮL¨hKÓáÕ³mâ–wâ;£Ï„(º3)ß «¨rZ¿°ÁLh/œá•%º§ŒåMô‹H}Ç鳆wä×fqý$¥4ôyÛÙZLoaòavc` ªð£÷Fº‡c¸qq.ÄäÄIÝQ‡AÚ“å➉ö㳄¦ÂÊ$\+ ØÇÛ²u‹Sbî¾|ÑC¨ëÃQ¸Êvõßš:—Æ_h]ä¨`ÜVËî9¼MÅui^²NM°Ò¨³˜MÛº|S ÎèìÌâÄ>óPU ¿‰¸, cÆÞZõ²Àqåà:n§´à$rHÒL"vî ]Õûo¥W#~ð†ç"@©n:x€a•îKå×Ó£Ó¸S5WdÇÈ÷FÌI/APÎÎáq´*H(wF¦oÅ@´i%"{£È‚axÓ]¯¦ûnŠÖzÅ£ûÚ0í~ZIüÏm£½GT_ÌåCoÐ\Okã+Jbáò3#*$/Cc? •»'È à–úk‚Û‹E4x&Qæ£Y-ll¤6A‡[)aXÜè¤ Öª¸*(*Éþlu¶#ç[ÑÃ$dìð$eÖ«sˆC†èõPfûQúB\¬3$b2¦^õ[g ‹Uü´LõÀãíFëÃeÏçJ-í$9ÂOåÝ\È¥1.TÎoY ` Ì;ì¾[Fc8ëL™-ÏÙ?¼ús^WÿYÂÇÄpJN,§bâxð6}±™Î…F‡ Ñ|79`ÿJ5älâ ¶”È¡)è Ëé/›55¢<$¯ønöÌ=6¿C±݋ͺ¨ÏJª¯RX)°*âÎwÌdÙ^ÅíÄX¸ëÁ-¨ˆª¶¢‰f iq˜wìfV}K¦º‘â‰ý°b"K-¬)îù6üUw#™^ HZMû‚ò]áBD¤+6ª=ƒW˜¯cÀdôŽ?;ÎuúùžVnú?¯e(/µe~ð ÌÌ¡f.Zä|ƒ<<¡{½Ïâ%Ê@4ƒ WçŸÐ7ïÒÙ‰¿UÁÖÒŒc!sÓÐïÚ>.]•U)<]¡/ŠU± ^|ü¸¾l)'ÅyP÷Y ÁÉ}?]õ+WI¬‰?{îÞƒl‚ìé¾aË‹ññšOQ¬Ðé¶à âêȰ—xÚã£á'W®¯>ÝaÍ׊ŸWþ÷µ@ôó¯G>¶Ž$IžJA?Òüº³¼°R;Ü$0ž†AÖRä ¬Ãu·K™ÈRúúAt;NïèÙ°iLKGŒ$Òj÷Ä)ÝDT±/³<ŒJ¯–2è†d)€ÆüÛH§´ HjXªžâ•2[î©+”¬¯ Ê5ôwA0í˜f‰ 󠸀p²÷^,y…²"²ï&ê:Ú>Bˆ8k¥;Ø]ÈÛÖƒžÄ_Uò%ïís²~¨:Ú‹vSóM¼WKt‹J\–ìè6CPÁ nÇ_ÔÔŽa9Å-H:äg²t±6 %lº}€ëóÁwe¢!2°ÏT “¢û¥Í™ÀÕ¦¢Ã#¸€yOvG”£9²·Ëò£ÆÏ›¦jA÷ÃYþ@(ê—Ó½ÄÏô‹Ùn³¤wÏ06qv"ŒÐâœpå15ÌåmmäžäZã«L¬Ãd…ÉæË‘õp9Q»TóyÂ1ÚÜÙSï>Æ*ö‚™‚ËìGCc 0˜ Ô) w’Ê1¿ Ž ±p|¦ƒ;'þ‰¸öêXħåËxeH Ñ˜¨ILh*ÓìŒUÿbìX#¼Í÷Ñ þà±úŸFªuíÁÍXžRõ¹RÎM;½)Òs¾5â;p}fåºÀÌ+¤¥Þïå§åæo;„?6 Õ!Lfó Knâ Ù”nrC+–«¾yÚry»¤Åêc·=ôHMfIá¾S•b›x!Ó'ñ¿¼1íy7ÇhÌ<ÎÙ}˜àɤºß¬·âË, y”ýÒDëâdiðÑã¢Ö7¶â{$¶eÁC[y1éý'WfæKªÆ-Úñð›Ø¿qæÜKT;»àÖÍ= Æ$#ΑŠøouL–óÄÃ’&øÏÑú–ïŒ œèL+Š¿¤µH¥\,'lŽq&`É!IS¼\t.Û¬PÉ4-Sc¿ ØÏ­ènÎeDá$Ï[(ˆšYÖÇb c%°¾®P±ÅX´~79¯ªR„ ¥çà 'Z³ópð js$Âe-¶†~}8xysi6õVÒ¨"SŒ32ej”;P¯À ãL¬k “ºÀx®8¾U–ß'&ÑH¨}ãb:ÇÂõSš<¨Mwj—û&ó üõx'kZ“Þ,%†"å9[ œË·/ºžp" iYAƒO¦W ¤-Qhè½uÁ&¡ò?øËñ¥ö–H’`7ù¬ETb ¢M³Þ p7¬”†M¤ä¤juåyÂÁ:“És¤êa~_>»Ê !6ÆP6Kšý ƒ!±e@O~2Š»ùW‹¶û–b©&3½ÐwZæ›D[… Yxò>Q4nêSˆÀG_È7Ö•%/¬£X7JtÇzG]w®`Ђ·ÉO†m³vžÓUŽzFÀKíg¶Ò¯˜›á~°T¿p‹ÖBƃL?",Kf^;} 7±¢„}ÉÂgÉ…)_Ví‡Üh¦¿.À´Ôýt¦S>zÂW%ï0!}R˹ßÝñ^úÒú@œ4ù†»HÕT’I%ŒÃÇ1‡‰sCX£ò.¬~xsÿ™x0fÞ{Y“ô²ÚYþ+@sÀ² ÕÏã1yäëX(YcŽÈSþEŠÝKµÌÀœìqO‚  ÷8ž•hì–Lâ,B…V¯ÉÏèCèüOÄ£ÌnÙ-´Þä€Ò€D4õ×çÁî™<í®›µ{ÂC¡ù´²K›²+ùÞ4SŒ?@'¨»¹ð¼IORR®ØÈÄÇA¸(4‡œ‚bV#ålÞôRC{oØÖ¹€n…2Ém!A'/÷Œã±^Ë®ïŠÏn„ßð€‹çŸºß}9KV°¼ÑÑiz‘çåý¼y¿_ô¡J!ûûèÈaŒ–Åó©JkÍK„·zýíh®˜h—%áÚyš3z—÷öˆûTyJ~ X¾ú¥$ͧŽ?Ç{oD-_ë‚ç›’KÁ¹Â ñ¿îÖª©­ka…t›Ë°v ð-ÇN‘ÐßÂÞ?Û½Žòç®_œO–›ëýµ¨>H=•JI³Mž¦ršRIœ1ò¨w\˜~ZÆ> Õ±ê{¤ì9«Îïï¼*[ý ,ó|ŽÎŒ3@ËqÍo½‘nòˆ kx¸UT 3&Ë×r ”š7Ÿ­NcÄ4ñª]cºê¶cçäèïëj'ÈÀ›õ%×ú!3ó)ôºqy±Ok…Òæ|Š=²|ÞT”tNÛäÐzãjµ8¶SÓ„ËѽÒQ1Ã¥{Ô»•ÄÁ&yÑÛFV‘D²j©_2‚‡ÒÃνÜü3eTaöNtÃrQÁlÀn¶v:µþÁ=cõ '*¢kâb0󡾚ÑS#žÇlÞyÝ/ýÖÄóû¼u0"ÛÌìÒ ¬¿‡ÀsÒŠR‹ö¾Q@QûI 8Ÿúà)"^ô’ÃmrHPã)ÿõôgÛ§Qñ²ñÐ. C'€Ê­jø`/Û¥Õ[òvž>žcb¹^MÛØ^{‹pÑõñ•x”ë% åþù⪠ýj¹äGsô¹Ë•T]ðnIÜ ëê–ŸZ\Z³l‹êÁé¡ÏœÝyû¸ØY‡×ôI+_à—h½Šæˆa¶17ò½ tNùœÎÃUûðÜ÷¨œDçH ræ=Gðìjžf) ö»µXÅ^0‚[…ˆ¹L¼ä%êy©… ypa>:?×ט;Í(ÿÂÆ“L0xáÉ­ŒëÒ™t°º0ämàt2=šä,άWë–Ç»ƒqšÌH¬xç&{c'|.jëšY¾’TAVÛëìD¯¨Ú 3° Ø|›Ç°™œC g`–SÒýˆš”iƒP`Œ°Eµltö­§å¹†€df.8ƒYb~beJYúûÛLM*„ùQ„ˆCl3¾æúxÁcˆÞZÞ˨¦/º1»„uÛOf!ÛS9R·)™†÷ŸJ„EWX a¶T¾iÏè‘|òL°?að²»ÂÓE6•ðå&j‡ø-誚) qu³eýä{6D[°FŽ€ø¥†IgF¾w)¡}¶Íñ°ÎTÝÒÉÕx¨ÔR ‡09OÑ\¾Ÿ¼²GdÿibòDrköo·N´cŸXŠ,Úœ–üsÉ% y ¹FƤãLIw*d b‡ñí³GThØj9þùs-ñÚ¥ë˜@Õ3ñ='¥|À _ܛʬʯ»‹;^â€O60N”ñ\{oáù²ByXô¦D˜¡Y¦¨^á÷¨ä”Ì’ÉŒÖb²WD\åßEºHûwG‹Á{ŸùöÖÄ®_31 ä»óƒ=LB¬ÕETm›ÏÛ))‘÷È3Æ=ÙÕkv7?޵±¼Ó ,Üy*2{½¶Æ¤d¥jGß1Œ%± q. ÎhýÜŽ|¥i«‘*Ná>ï€O S·õi·Ûëgo[Zd\:Y¸BÕS‚E<³«ÉGcÇ+HÜ„ ¯ƒÌázlàt:pd3eôdÛ3I¡pàï@'hÀŒy”Ø‹Àzxj¸v^ãâOÏZn&‰ïÛæí¤%7j_GÛs‰ŸýãP;à[wžÚsþ>ODUq´SñÛàQÖî}ø¼ KëïÒ“«QD^b<äùèR¢UUˆ…îp×%[‹öî«:ÀÇe÷T)8zf3Õ+v¦ÜøéB3 ?„B¹íáúžFÒE°P®6qcºõ0–—ù’רãCòúÓ3¾9Uœ ³Õ­ÐYŒ¿Š¿a‚5ËAÑKõná˜_§áYR o·Ú¿MѸ½—WàKªAÞÿP!ÁŒfû”³áýÉ Æ°]=(ÈèŒ1Že¤<€âé̹—‹ õ›¨×`IÊÃ+¡™å—çꛨ¼I‘pÇŒõªIœ#„¦µÞpñê³§L¢¨.ýãwZ(è÷eâו‹gãððJ·?$¶ùL4¿]K-‘(-V?8îÙqW§dhÞE4×´Çd'pî{¦,òÛ PÇù¹Â¶o£;Œ×îw²á |T±j—fÕt­å>°ø2®-«&cõ2.§õrõ­j3x –í3ò”O5eº Ȭ\F±´Ò!ƒaˆàŸÍ–2F~ò¦œ-òåíBsö¨›1¢ó¯üÄ7hŒ-n”}}óõÛb¼Ú1$Œð€Ñ\œˆ}äÛÀ·UÑ?"~D%ÐG¯¡w¡Ê`ÔÞksSIjÔ8t#๔­RMF•_['9¿#m؉‘Ó9ŽÕßH5Ðì^BrÊv‹xÑJU\dV•b¯G|ÊyîÏO4ùÜ1 SSq÷#zÓ«´Ÿª,¬=Ô¨ÎöÖ3ÛŸ˜™r£•[Ynðɬ+Ø¥"©Üië ðÑÄWõ0† V†å“zg8 ÑF¿øºuòJ„_>Â÷ò›¯ü­S<wX7µNbÜ8Ê|g8Ó$÷d› êдØ;ÀR?œrïïb7Ct [9Šñ‹ªÄcæFçL0]b|éå";ª¼óhqd—c„6€Ú‹Á­‘V0;ÐùKùÅõ­~ð“ˆ6¨+žéhä–p\eAk޶eá«opÁ#ˆèH¢,œçÝppPH€VWe¼Ò‰\HbÔãÑ£rͱ”xrœ™€:˯æò-bˆÐü>ÍFŒ0\ ½.ñ³K‰‚Ô˜]â…ã&Óƒ?uIƒ!]Á{‰J²Ãp$f¯A’h‚J6sc¾áz.Ajä×ß'·vmá(gºþ˜ GÞìç‹D.^v®©ÊhËʼnÏï\’•#g&i¡3ƒQ7´9NÑ'i)_T¸_Âä¶G=¶v†½Œ‹˜“Žˆk:.eN\WÜ]êÈå¿>°cebo“'@nÌ!k×,)G&ù¤_ oq:ïY©&ûÅxñ®„©µÅ¡G.:ë¤þZ|?ÉéÔþ¦ñp뜿?Âd°êÚPÍtOwòm¤c®4ì‚J¹¿šÉ—oDÞü¿xˆÈöìÃa1Ç3Áë*%žV„nÌ zHH𪣢F o.œC!É H7ÄÖ—E?¯¨îˆ5R;¬+PÈÁ §+Äex‡.:(dÓ×Ó¥T¼p]Îc³!ulTÞ~{´¹öù”˜Ës¹Ax´<×()_•‰¥äÇ;-˜îÖP Í‹ó´7I/mncîmÞÙ=!r4®ðÊŠ«'ü«{NXnFù ác•”7ºÇŸP]¾•ôà¶Þ¡-Ñ.6ô_`ÅF&ÉH€@|ÛeæKÊõ…Âç'-¤EwÓzÅp G(6R;:qÎWº0e…íOÒÞºïn χ ¢ƒ÷ѽ­NÜMá1ÿ LWe¨¦¨eÓþêáª%Y‰³©°yŒ¸#wØ[„=ZGC‚ɨ¥ bI‘×Ó@$ÇõÎ6ö¸Qž5iò[ÎJ N÷MêOiÒòxÎZʺòh&[…–â­;Ôø['ÙoVÆÄ6¬½IwµUt©õAÉe~ÙÌáwߊ懳3“,¼p³„ ÞÊ—ÄŸ©q›™oäÂÂðà±íÕŠªàöNìö&5W"6”¶ÍåÿS¿êš©|¥ïi.FF1sÌOh£8Ìïb`Çáu§qö5/[w_?Êžk®?â%Ò&5g]%{ýÄ·k?URŒŒ‡Õ¸·µtªë‘ -ßh²9Áu¥:lXÜÖOR>ê*šUÉoØêB²à'Ý©Gž#·-t®útÛ)ÑžÐl™m•? Îío†™oGß›n%wì<oÕ5î‹‹Í…åJ¤- ¨O¡´4À)Šz¢ ’„ ÉÔ/õDð)’¯¹»](&°û—¶ì‘½ôêdÙ6Îŧ—lºDV8Bãut¹Å$'%’[+¡Ü÷R‰Ëìs­*ê+Éw% |úd'%}‡At¥ï‡ŽÏK—EyE¦€"0šZ$ ä-rFíd“Y6·k Í$ †iœ´ò¾ÿrðÂNMÙÔçï_±MkÈP^ Ey¸IwY—Ëq :[1 ÈŠŸŠ8£l(1Á‘ð¢îbæ¼ã‹výùzˆé:ÓœkñÄf’•SBÁO9ß§õÛ€ZoªñÎ9»›9ßò7sçÛ•¾ ”ˆ„¯23’Z¿õe FƒÓÀßi#ç˜"MœLK9ßüQەܸŸÒ~%öZ”å%P)¢,*D Dë‹4¢IéŒ :¥æÊ°ˆi>ÙaöC€SUryGÃyR™»ãಶWÕ=×Óg0Ù¥èM >XßC˜,¤hpÍÃpˆvÛ€Ùr€© %¹ÃV3Ë}-dÞÎH¼_¶ÊIÛ™)›Ä|˜ý"Ad þ°Nã¡è/3¸‘gé´Ga²ò©ð üØEªt)==+£W˜RC1‰y‚ðQR2,„˜s˜6è£`6o«o‡TÂP¿÷SÖ0¸ýWØ0Ò¤ö{Áe­ì;~Óë©­´2χ+¯:ÁÁƒ* pÀŶØÄ6N8aKœ0‚"òÞâ)Êk¾>ìC Iuø‚†IJŒÛt8kì8ŒŽ¸°4 ËRܨ*¡#3[þëkåªO%žWäarìš;Ì÷õ&mäŒï…ˆu¶@á†)ïNO\CÓßãÍW‡¼Î xaÌLUò8lFº{µžsØáÁjKcíŠâÄS=¨ãjtyšõµŸÏÌÓH-?–•‘±’Ûa¼ñhëº,f2 NžC1ÎLöy6ûL?JÜÚTY<<Ð-q“ȶã(:rZœà©~J™³ƒ Ç@MP¸‹«Ô¹g{5ïù:²¶|ÓŃãiv^É|£>Ð*³G[åáb"ô¹·Óbèk¡YkgdEG6¡z"Lüðª­x¢ä`Jmžç\{:P1¥Úïpu£Àq4ëš'd$k¯í€£í¹Ý§G£®oiyòý»<9KAsZ!$æzQÑéu“CsÉ+ÁËðöÒ½U´,’£¼”üòäc¯ òRþ‰¡ÓŒÑω­3jøì¹=Úý3C?,H9 ]½-àqµ«GX^âí’Sp,—)R­Í)ËgÓg>%õ0HsÜûáØ.E6|ÖÏòs»£RNTwˆ"›™ •î!q2ï9Dý¤¶5m±g¾|‡T׿…'%22ÏãNˆz²cùÎT ]®#/§æÒâz Tò$J{w Â"tžÓOêÉaúV`ò®­¦w¸Ì:{E‘ªìM0$»àªÉ.žìŒN"_ó?‡ VÙù|}ù•5í"p Ïö –…gÅ}à‹%ÿŠZ”BüÛï1“~ñ:˜w†ÐIä:œ}ÙT!ÝL†CEÊŠí+AiåE‚ÿñ2 †­ôÇæø•Å.lýƒH§ˆs3åŠc¯jB`[F­ÿ›à6|×ѽW¿dì—™óiîÑk͙ŷôÝbeÎ;"÷ j÷l¤è^ï/HŽ· Å_”hóöT~X%g «iGÊñµ±ÌÛ+ªÕol (àÆì™nùN!ãÚJwö^"tT“D-ÖïJˆÁW×§žÀ“m¹ZÜ%ï#/&÷‘TÕ®^¤í2LÛbÉÉžmå8©èëÑ´7û?´Ê–Ô܈ª">Ésر½)ª½w=Ü…`…K#¥õ¤Çäýâiƒ ÜAtwœHðFHº|•Ï‹ñnÁ‘»®[øç[ßWåp!]’sBï•ÊÔÇc_.Ø?žût»º™Bª›LŸ•!—¨5ì¦E¥ë¡”ºñ»3×¾øP}7õPpIžôé½³¿áã¶Ž¸RUy»ÐX@J¨ ¶¾1,ň/FD$FìRM÷$ìÞhö^æfãA"cç3™TŽ­­i‘Rån)ÛD"£St1 BÇŒ“æçÊú‚‡ê£A®I"`c ‹\¯r 7í÷ô3"Ë2ý’<y¸‚Eú•áEeD\°ƒ…ÌÞ²)sµ³Zê'ð®ýÅ¢Úª*óÇD!kÁKƒ¬…Ôœl¼Üæ4Äm.·sR Ìâ~ó<7ÃÜÊ•"²ó‹¹Ñò<7ô»œ} l~›ròÛÍ}l5]U÷÷cˆLÞÖ‘~ƒÌêjæ£gY{þ¼£ !mó¶_~Î JÚ`<Ø|Í’¥Æ®ât«&É—¨¹Â½ŸŸnÊ_z0˜¤ \o†IC'ÿAÐǹ"”ß,òš¾¶ûþRÓ¾VäÁ¾h¿ÀË5ÕÊO"8ⳇâE>Z5°'"εDUÁ¥³¡‰U&W˜6Z—k‰b‘>¤ #+’ÄFõM«ˆ'.ã”L¦|ƒÖ ©ºñ}mY+DÓYÆ*ß*ߣ:å\ƒ¤‹yi–À>‰§Š´žg˜H±åÃåýZùçõ&iÅð²æ'ƒÀN†…\!ꃹ%ü#&oj@Xq-nÉhø ¥‰÷*ö-‹ ÏØð7O¦'X4‘«”]~Û†BÙù]bÞ+ÆoŸ¬Xˆ+u„Ž´\îµæîØ3˜N\W©Q«Üèõºö!®£-–$pwVˆ/¾6 ¬Džæ²L°õ”¹ˆ¼T«dH9vIˆO‘öÂ]i›9®‘6{%h*õ2R®$;Òâ:eaq8cPù=ÙQ”evóË[hÞïæìÈ`ÞWBXæÅƒ&vÎ3M©¥rM\\W õm"H'ú³«¡ÇƒÿI½Ÿ²>Ç®Ž9Uã.ÕV  žC!bjéG ¬~æ„•„È÷§{E%y`{~*Šù·=Ûìôœ_Òe>¯j/¦x¼¢ ¬ôÌÉp‰w¯Q|¦ËO>"tH÷L' mm`”^~eõÀ˜}G‘¡®O­H>ˆCi»Ç¡^¥ónNïÄ*bß¶§úÔP:Kæž|ÚoûÑ+Ö‹ëï„æ¼,-ÜP÷6ÃÞøž,Áó’ÆYÞ­Äp,’Íß9Ý{;'HûÇTæò #¢Ñ¤n+QF×·ˆú»£¸ÉgN¬þäv AªŒ°è²"åÒ¡Úæ Ú³ˆ–¾·óàÆ‰»‹JgÂÏÁ8sFƒ"ŒHíTŸÙoã7>³2WØ á!X´ N¢Î[þ†£-ó¬¹û«»@Kr$kâj[‹nÁ YõÃï±mC–üãx*~Ø£‰ËxZÔGØþº5¨Ž,*©Hn-ëW®Ï j/êç‚ëiV¹žÖ‡šîZžšÒ7ü7.ÒËP‹B+òJ.­,îÿ&/}¿úÚ]ãÖ:œ=׺L¿êØG{QŠDWÔ'dÙÐge;%¢à‚Ó£Ò @$z`È¢Iy5G«Úf¯µö‡Id"jq Œ7Q¨0VÛ§¨‡Ë1ÌêpÝ HÂë!(ÒNIikDÅŽŠ.ßOÈždÌ 0ñd)&5ñP©x{)¶á~äÄy.ó5F(¥. ݰY]ïmïĶ_ žaÒ­SÜÛ¨ž()ge“§¬='rBµÃÓWÃÂç" ìÓT†+`ÓÛ`eålÕ¡´ÿÀ64®B=ýîÍÞ´¥# $ÆI¢q’¦™òˆÓ±ñšÞ±Ã'[¥zHgdzú¯Ðü®sP?Ï@5îýbS¢o·ºÅu~/¯ÒÏN×>v69ð&fÒ{9»„k°á2H3Yú8CïnPoðÜ«"OFœð,ñú †zÒ@”¦Ñ7¢ ‘Y†¬ÄLâä…'¨Ã_ŠsÔÇñ€ëº)ÁŽäœÞÈydiçÉIØ‹—3½ü>´Õ°­ª:näΉ#Všgž9z é|bic9™š²lû矂x‡ ’§.£—´8‚ß…öò ¥¬²´~óÌíWçyS?5b“kÈ‚øíhCÙ2 rh¼¤¾sB.\£djfÓ›D bXV[°Ô'%¢šüÑf¡G\l4ðU\sWŸµuÁnîÒ«š—\_(Hþ540RÔn‹˜øHu4Ÿ/®¿[æ”}ZîÛE/ß§/#ÏTçe¾ uϳRÉY»³˜;!ìùN›¥ßBV6Â÷çô¦²NÒiáÂí÷é“”Úhò?í)׿Ê’^⳿˜NÅ>|d´¿üÑÖ`Mò,ŠöÙ›[h–ƒ½ä$štø8RÇi_ábiî5”1ËHL}õÂÔ0x C–û… "ÕH£eÝ4bº\K|ýæý˜ÄŸ€"½«¿Ùîì~Ãù‹”¢ªBƒÙÞ2„øq¿ëiOû{ø4†X_>p‚‡kQ£µ€y²B‰Yiù±`Œ½l7`éÉŒ-Ö.¤È±±½>@1æœôH(¾k°&ärÎM½¶OkµlHÁ|˜JÙÊ'›Ÿ’¨X‹±…Û!ÚØÐÕ5©O¾òÝ¥Z`'‚ \Vg£êe/dê , » iÏ‚¤‘w_ÚÄŸ?luPËÁc2©!îþÅ-Œ^Â~DTŸ£ú.…?X"\Ѷµ2Px4€CÏÂ!y¿Šõ×dÃRÀºÆR+O»ßµúéä¦Sàg ýîO½ÃfÒëXc³n¥ª?b^7+ÔÜ^u̘hKí¢­#¦ÄÔôÃòNpðàž¤`ªd>°Üíp Ê\È+)‡¨ëèŒÄTyº º­¼üª*àSËÀAø–PoF ô£ç¥ÙZäjê±n ò'é,B¸” Ò3‡ì.^£ÌÀ²öç—Lpì°Cæs!Wúu±“Þz¡N5—=øeª<)ÕÁ)ól Ïv{¯©fý¢&¢s]½Í©ûü™Y·äQQЖ_næŠr™"4¦"»ùEµ }̰|Çò4tF´âåhæyY|îdcÍ£”_7{N\á½SJ½^¢ÞÕnÏP‡Jâjý‹¤æ19ÔçIгsI®òÈ뮟ßËåн%MÎT8‡T»Â¢¥}Î% ¡ó Ô€«zdCh'^¤TE¥–|#Q8Û¬nà×ñ ÖdWó]/l«&ŸÓ.6%?¶ 5¼mØEc^¸šçŽqeÛ•Kû\ªÀFši Ce”¢4|Ò`^žóᚢtda`fOý€òò›Ä@×FðÍØÌ‹ˆPk/Q8 ¦…¤ÊY×YÀ ²ÇšÑ°ï{?4‹%£W¯µLqú“„ ø•ûŽhå;¦Ùü™ù©-gUn‘¥Û×ÕUŒöñ¾Ó´<¢Õ@¶ËÅ”MiŠM‚aä˜*òÝ/”-§xvIVÏH˜HõÝ Ù7HKÔqcÌG‚Ön]]ìïÐVP_©¹õeÒ×6^ã ²¼û2Ìr‹ªÝ”ŒÆÄJÓg\Ý4ÓÃïoÎŽ4‰Z{x{å´ôBÜËfE uûb®”&[@gØ/søõu Q€’µÖÒè‰ëA‰ÓÂ:žÇ=˜óFrS`Ü>‰¬àø±Ö¯„á?ª=&w̯é/áOûP5csÍ’†QÆ!†Íð^=ÄùÛ×îSkqh¯D,_˜?`÷è_`£ ̶7ÎIÝ+Sº/q'†¯¥ÊŸƒÆtw2Ï/IÓ1A{_ÙPw#Xø0(}¶2lj§u˜Z.KOg¸VZ£%8ðPù"Ô mEjW5÷ð]=}qK&/ ^W[²êT-xíœqtZæ 7lï SÅ1‡ ôê¥] ²TGñ7_Ú-ø®ô3صB[ÊÃÏ_Å-‰²üôEµµIàË©ïK„—bðÔi1kù_J°ˆ´£ž§$p`‘Œé°Rqü®²â:jƒ 3Pß]«>$® ‘ÖòؼSÌL'†Z”FX¾Ýý•¥g„zºf5ÍÛþcºücÝ«ÇØ§4~ØŸã4*KËäy[‹ýØ’-;žqºD¾ZQ‘;½+²îR66Ô¯_³NAºbN®‹¢œôLfÜ/£hÔê¨80´—·bai­b8×éîÒ0½¤kx­œ¢gèæp«5X>P•SW¦¿·å¥cÅ‚Kjèmâµß|š cµÝ s[»¬Œ'"†ÍðÐѬªjƦgتM(õgYñ…üI+ËÏyé§Ñò²sNÚ˜Oœöf$kô ÌH¥=唽`ÈËzc0 Ò¤L’I¥Ò[>§¦ )îŠÈüüžicÊô©)bíl®¸ÿ¾ÙÀÛ ãú^ô¢½ùƒ†Ë/“ü:ÙPq0}ÖýZ÷˜w—ÞšmªR¦F΃Há$¯·«oOE—zŸä ¾†ÔΗ ý…ŽLåsºk±Èa¬¬.íú¯‡wø.¸pYsQФÿ‹¦.^ endstream endobj 54 0 obj << /Length1 1382 /Length2 6009 /Length3 0 /Length 6959 /Filter /FlateDecode >> stream xÚxTTíÚ6RÒÝnΤ¥[ºA$†a€a™¡ARJºA:$¤QR@ZZAB) ô}}Ï9ïùÿµ¾oÍZ{ösß×]Ïs]{fmN6cA%¤=T‰@ Šd]cK‰ @¢Dœœ&04ú·ˆÓ ê‚!2ÿPñ€‚Ñ›*ê"À]O8 "ˆHȈHÊ€@€($ý7é!¨‚½`€®p‰€¢ˆ8Uî¾0'g4¦Îß·„‘––ø(¹A=`0У¡n˜Š00FB`P´ï?RðÜqF£Ýe„…½½½…Àn(!¤‡“<¯à C;FPÔà êüлAÿŒ&DÄ ˜8ÃP9Œ‘Žho°Àà0„x" ¦:`¬¥è»Cuþ6ùWº?Ñ¿Á¿ƒÁÒÍŒð…!œG è«ë¡}Ðá𠆣˜x°Ûc¿[êJ†3áŸùP˜;%„‚ÁÍ(ü+ f›Õ*H77("úÕŸ*Ì Á컯ðŸÃuE ½þ¯aÇ_c8xº ›"`<¡Zª0Ñ¿mNP4 ’– ¨ÄYøW_wèoço3f†@w¤;àˆs„b¾ˆüQ`/(€öð„úÿ§ãŸ+"ÀAöP'‚èßÙ1f¨ã_kÌù{À|+†~"è×ç_wÖ†9 pßñ°ª™Š®‘ÿŸ‘ÿåTVFúþ‚¢Ò€ ´‘$%ÅÀæ1Ãþôñ±ZG$ ýW»˜}ú»e¯?àù#^àŸ¹ôæBžý>HÁ\DþÏtÿòÿcù¯,ÿ+Ñÿ»#uO8ü·Ÿç/Àÿã»Áà¾æz¢1*ÐEb´€øo¨9ô/éêB`žnÿíÕBƒ1jPB8a-(r[tû/; ¥ó:ÀÐç¿Xó—Ýô—Þà0Ô‰‚ýzÂ`¢@ ÿòaDqÅPÑÂ,"îò<¼ã¼N‰É[ðã(ÞÊjWt¢e”8šk³~¶ÆÔe;‡ú‘ ƒùŠˆÉÙ&¾ýó1Ÿû¯rC–Ó5 ÐXjýëʇ‚…Cߎ_9AYŒNã’ÑëI†,dvV¸VÜ)æ³—${Ä6Ÿ}*¾êBmÞ“ÞÜ€ÒSr¶t&tœR~×.öcK6ŽUô°å}òxJ«LO-Óñ+ìǰø3åŽX†o¾ëVñYÃïß¿ÎèÏ;*Òî­•¨‰ b ö»Ë–?’<È@“”{÷ ‹KYüT¥"-‘¡Ä-‚kÇNrùûtÑ@{Ò ˜g““žJÈ ý‡oš=â³rù=ËG̶9&ß„Iè[’£ê28”9k°N‡3jÅླྀ´¬^C×ÒOÝidYù™¥öbs,þÃ01›OôçvqÙI"¢Öirp©ÁO•% à¬î#Žæ‹—wÒ*FN*-=6—UÏ‘[B_Ÿöê,n¦ N¶Z^w ÷D"9öYh¦B-:³wûv’^h‰Ÿ¾¥0ãS“@×kÏ!»myò¦ƒŒ òq Ú»4dï)JÏO°'˜‰‘šPÌ™èlbeêZÊS ÔC;ÜÝž íH’óÓ s½›,FÙj©G¥{Y­ ÓØ'ß¹R×3á7ÕÒÞÒ¡À6ryƪžoÿ¥´fŸ~þ}“þ«Å[ô´¹lV÷›/™ƒžŠ<·*Ñ|oʨ¦o˜>ÉõXÿúÕ0,öQ½Ž“Öu,åƒ[Í„_<…NÖ4W*ï©Y“m`[4[žøA¢Ò6Lf†lÈXòjËÿœ¬ÐÓzS65·^nä¼@øÀzYU×}öªoÍiÙt“F{×½ŒTÒæKps U‰H§mòiGrÉ:])iãìá›JZç6Ø• YÞ ÙT¥ZCš~®<3±0Ÿï‰¬1C!Ÿ‡M ´ýµÇ*FQŒˆ÷G-k; ÷„?{½J·*lý>Ïڢé_¨¸›¥‘BygÙ€;þîÓ‘¥ ¯¤ŽFvgÓ2]…“†]…Õw¾ÙpI/Ò9:¹%TàºYŠb±^²¸Í¡XZ´ÉÇ)Ƕ€Ñ×°42¬UŸ7¼‘¤$9ýiŠ“ôa¸qñE²‘ßSªãcÇmÿÈ  Uçw'w6Šü³²Ö”ÓFvf/^,ä£D¹UŽ}—×êlŸM?S´†°J¥Ô#%p‡1|º¦uyÛU_ŽnïG•â¬)\þ.ðÞ©öx™î“+öÛ,>RI‚8V‹lxŸÏ.^ŒÍoMßG¶‡ÚÌq¼àŠýx|dM!OKØxjå %œ¸¶fÛÍfò±/LrŒ½Z½¿0ѰJÒiÌ^(›°•véiƒeM¤$~”%¡,G±Tä…XÊ2Y'¨ÉJ¨ï`4yVAe-Œ7*è590¤X0Îÿ˜9FzíªsG -­7N˜ñ¯$Ñ«Å:pæD}«Ü>̰ZçVõŠiC7>V›-nŒ †uƒ+²­O¶fˆ±÷ŽìÑ…ð”Lgr‰¤ŸúQ^=¶exFÛo¤ƒ=6C÷3ÂW‰Lg­g¯dKî¥oulxͳJ»ûRÄÓ6i&ûºË2ªÍ­). {"¹2‡FáÖÕïsìý4TØ9C ïK¼Ù¶êÓG%âFJŒ ¶8>½h•ùÙù–w¨3^íÖüV‰wuïÇn&ÁÖfâ×¢XÂLýç—øÀÈ…n¦ØùwtÄn#jï]2öéJ¿ˆÒ û$w¼~â´m\>TLæ ð¾Þ‚—Š'Ì2ìQß™«žy“åË=õù;[Ê éÛ® K½ø+F{<‹3ã6¢óœä l¥Ë«ÏnáíX±cdÛð·0äôëÏ· ?¶ÑÔ„úæQ¤$,zݤ<à˜X\ÀÚ—ªÖ‰û|'Yw`wN¸æ”…v=ÿçRš±` Ò¸í¥·¿É®!H\ú™d¿ ߺI¤P‰.ñel*ÀfÏF«ÙÌ—•òj¸d¨¶#“9µß²GížÜw vîÊÛ#º@)O7}oZ}ăŸð»†æ¿)Þ{á­Ñª×Xn ’}[©í¹7Ç0Þ3¾h9V\¬&ˆjx½Ê0ߢ´‘Úúª ïªÓ¼*´Â2 ñAé±<2ˆæk|…V$”:–vay».FÒ³žkŠðJ”'ÀØã&zB9@,ó?B«zÍ`Ý””Ñ~Ç›RýÝ%ò?]|M²ãBz›ÌÜ?e¬2<({ã¤2Ìtœt0ÿ²-ûš´ßõö&Q*ËAÂ}è‡mIý™×Ë,|{Ò±ã3Z ‘.µ¦{Ú§T>.…mB„x£çüö"u³Ï¿÷jåÀUu ïህœ5"'GÜw&,;Wð ‚ŸE„ÏIc $CbùÔÄoe8D~)FƧû[«TÊsüÀQ'ôsb›ˆâ{ŠlÚ¾îóVPï{ÈÞhÉh òH*¡_ÛÛÐé}œ{.Ÿˆ i¨ÐlLTÞXÏn=±ŽYôÓ×?/ÏÝH¨áà1kNêÏBì¥v‰í¶Ï _À˜ƒ7Ü_žödzã㣒±Å}…,Âp˜§FŸñ~ùÄ\dR¸‡Uzê ¢]PäZUõ&Æå}PUäGW©ñŒeõæƒè »à sÃmN¿ðiú»Ï[ç-Š‹zÄÐZ¾ ¾®§B¸Ó·¢û­Rn³„©ÞR{šìî^«ÎWUþµ~9ÓëÀCÚa !‰¸QL1…ã(ÆWËôÙBækצü`G¶ ß貋#öhM¯t28ÙE–ŸÅŸ¦IÈ;±¨”ìÔ[§Í·b¡Ýƒ½dÖÌí·pïøË"/!bÿtüFÃŒIG*EoVâ’ÎØ éÝ݃£mUXù•Ué¡ êN޻Ѷ/Ô˦‘9Xé·ð¾‰›…ï:NšÆì<0 <»œ û›ù?Ó#®`Ћ·…Ê–þRƸñ1…ˆ…ßÉ])X Nà•\€K•¨8…ÊP¨dÎ×âäƒôÚ¦Œyfò ¡maõr&¦¡ÖPQ:Ú(wµ-[ˆÙJìÌÈÐMjüº1ëæ~´7Ù¨XW¿–òÓöpGi*H¬lÍf»[™Õìä!Ç¡´ZQm¢‹Ñck j+Ïßûz­-™¸ØY‰ú¾ÓtwÃí<嬭œÏJåU”«žŽÝ\y¹öY`XÍ¢â 1[ÝtfDçžú¹r¸ºrXßÞ­¥Á=„8¶^Ô’·«àš§+Ë;Íp¿ 7¬}’bø)+×~µœFØýøZÓñf“”]ê“R¸¶?f”(’»»T†oñ¹iÇûsùÉÞ™éÎÇ q;Z·‚I¾§mÜÍ}ú…›öêE°É5ÄúÓL¬ë‘0B…¨oþȸãIkí^ÚêZѹï-žõZ;ÎEÿJ&ò2¶CÖ\ajŧõÄqS¹Š¿zÆãñÚÊzrÑI_›9Îð)ÇÇs9jÆ¡s¹ ;Šb^rk±JÆtÒæ,N>Ý@Mæ^Ƭí’|Y‡§«™×‡<Ü;êD¨]® Ô)É»£ P,è´= î³5•ûém5¼§x?rmofSñÖÏð´^ä”4ÝÐÕm#jj'y0àÅÕ\;|¦²òQL«YçªÁÛÄÀÓ?‰ÕÑ6^wnÿ˜_0¯¦¢ÐôqF½úWÓsöví±ˆþªŽ)Ë.¨$•¨Š'í”æÀÄ”­Ó(ëäŠ/ýŠ’ÇRK01n÷ìë¾¬Ý Íæ'¥ M™ÆgOV-5YìERšÖÚ5ý[¼Ãl,'³Hˆžp¨—¿M×T]82cTp?Â÷Ìh¥6XæT—kíPŽÄ;]8-ÔО7pnx¨×hÕ‚«5l»¬BphBë·<åH¸Ixg:=5ö2äšÝm̯²šÈì]Õ÷vF„xd¡ø‹A÷˃)ÖôD]‹ÀÇHÏ»’byAä;G^«;Ý'AÚKæþ¹ºÓËb s qºGÖßdÌd©)¼n&>"ϼn±&{=\ùf2at™U‚Áª–¬B›Iq tãŠö3Ω©õƳ°d¹I‹˜7®’Å_Æ{ë¢å}åj#øÖCìUpºÚt¢û™”<ê áÜÍçý¥i—Ç×Í 1ë„`pI¢Ñdاž~²ÚN·'ãXe_xò²o7È^¼3NKïñê9¼üªò°A¡ûnX»p²:»„©PH~û„ÿ¶át8ˆ}xºjRn7ä j{y/Ìy×zu®’í÷¥„þ{¡ÀÇùÁèÝ‘ qŠÃ"³µ‘ŠþjŠ’Õ5//Y$cLQ¡ÙÏf+|7?÷¥à›Īˆ_jx8­™âõ‘‹¿áHˆÌˆå ¨KhêŽ!x˜ö^£rc^*?c‡¤°O×ëã º,l} ù2^„ø²—c=fÝoˆýÆœþ‘þ|c”©PìÓøšCB#«.º‰L¤Þ—[Ÿ~p‡nöuˆÔPÍ“¾” ¼Ivøè¾nPd–GfšP¯pú(bê^ÖmVí(§HÜŠþ­âžŠM2*ÅØ\“êT`†|ÊÎ9 `g¨ø¡èõú‘§jDnJ4ià1WðUŸ”´ÐÕlYû¦=£—2„?H­›_xH¼4Š .}bÑj ?p:!ÔRß”QèA§Î'â›êAH]DUܬJ>TéKʇ¸š_¸–UtbXÔr‡=ßg)¥Í%£é¹cZö|NoIJé಄š##”'κÅÄø%}Ím‘æ@#ͯ&Y¨·Gå+åßo}ëÖš'Épcà y~û\2éµ£öœWN)Ê1y±cý ûñ'²Ê6† ú)bߺù”oˆ5„y"Úq8^¬h îü,°Õǃþ°{Kt3“ãåŠ]h  4¡p_ÉGÝhp¾Âü«Á¡©º¦k¬äŸ7p¬cJí» ¯¶ÛwªlwDïLž¦ðã’F3÷TÖYçÖG1Ü-üQ¡¯XÌ©›¤º¼M‚.mš÷ ‹˜¹ÏØ*l¿>Ù{¤k¦N3 ÿ9ÆïªõQÔ\ZÞ1=°â@>qÞ×|"†Žl®úÙVèRQÚÀ^Öøéd?q–¨`pÈŠfZ'ƹ;Þ’’¦w).éJ³ò#åg¾fò},jŸ±€T-Êg¯D3ÿŽº6ÅúFäüõÑ=ç²­÷$&èa¹ ,O: šß£L Å»KÏý¢îlX®|ã·ÞǼ0¼R&šã\_|Å`#Ü»×v"Z æ½,=1Ânóqò›ÓƒÒ zGµYwv¹ÃŽ9¡W ÔÞŒÊÈnÈz ¡þ·/»v”ÃeîÉ™þ-+åZt*éÚW*8ÂÞuynÔøUï†ßr<3üÎ:©¢sRtœÔBë“éMø¡«”å‘zËrÎu.’‡ìáj)ò£Ͳv¸dýÄ^ß9Ïo.ÖšXíAfñ¬B¡Ø1즌.q+Uxá¾1‡h/ŸOÃ$ZÏ{¾¶ÔÏMØS*ŸÚóo¾Ò°ž‡×ú|à–ð>Z—°x®BÞ­ªÍªÏð*Ò1‡ÇIÑÛÚ±Ô,k¤Çj'èán¬ˆú·mV2%1¯jÂÁ5ý¯©ZæfÑk^ÄD~MZ–/F@ ï©íoæ\³ÊñóOžáÀ…ŠT/¹Îˆg8)}¶ãé¤èŸ(wÝ×\¾û0jCþ"©vÓWñSÿÏVÓÕ ‚߯z5e!þé‡w|•í%+àl\> m‡×+"ߤ:â.uznìÜÇêÅþÄžtÆí¹?@Ã&¿$öÔG×ÛñÔfæY*=ïL– ÿ ý„Ö:ýÙ Qr£òžF2Ÿ×[»7£ û N¼•&’Ö:®Œ¯ÚósØç*á9ö‡§~²­õ™Z¾3÷Õæ®¿Ù‰v÷(wÕ%~¡ˆ§ØåzùD,-Þ;6>ûÔÇɲǻ%ìn4x¾¼ð²’^´É“(S š,~Îß·P‡*»wÄÍH@]ßZö5óæ}”ásÆ“Ã:c8eKÿ)ÈÀè‡ïÅ€E¾¼öSˆJÛï¾uaLøæåú­Ú¾À:ÔLtŠBN—IÅK-r2hö¡XfŠÑtAt²E~H$jîu ;Y·¡yÛ“0ý«¯“I·ùôý¾· AZŽ] ©U–\—+ïæçÇ©>ïÈíJR¦ZÒg5žž¿>ÜPG k­¨Ç¤ ç*\¾‰ÉZ%±£„õø6Ÿ›´5“Ò‰?G›äLX§?°o¹­¬ÕNמxTy9¼×óA÷mð8 ¨.çåP÷Ø–÷lyú~a¦ZÍ´ãß^0W@â:ÈW=cnæÊÇz]¥”§ÕkçJQ~‚ðõöÉaØëïïî‹+|; §Â©óÍû‘yA|nLyáÞ·íäÅA8üEþWzè§:Ê£¼Ói1¹Ê®ñËÂ=.8ªW{£ß¨¿FÕo‚È|RÙ³*b"+ª6 ç¿'J¸èÌ2Dÿ”DcÓDÜ& œ¨Ûé¹^|2^/9kΤ‡\ Hþ¾d:ú2³PJx¦³_¢]k^·l¤v&Ÿ\¶{ø2N<ñ ¿f†»lέ/âÅRsªÅs®ÍOS'—Ùd=µ/¼x²Hp#ý,ðUõ‰hN—4.öjkÌê±€hô µ†ï&Úhk‰+¾Ù9ÑVî§a¢|r˜±£¾G_Cyà °–T©P ´G¢8uŠ ³Aq×ß ²"ŠÇ-Í-’S“Yù&¯n{CÀS°OÛŒVw5*×ã h,¶²éÖüà¾Â1Eµhn½Íÿ ‚p endstream endobj 56 0 obj << /Length1 1372 /Length2 5926 /Length3 0 /Length 6871 /Filter /FlateDecode >> stream xÚtT”ïö.Ý(Ý1( ! C‹„tHƒ´à0 0Ä Cw7Ò%R* ”„t7HÒ)Ý!¡ÂŒsÎÿwî]ëÞ5k}ó½{?{¿ûyßçù8Ø´õød­–P%ÅâÈkè‹„ø‰88ôa(èß0‡éCÀ%þ@ £Ð10 Ó@Àj®$*“ <ø”(€Ý`V ~€u!âG8y"a6¶(ô6ÿzpA¸ Äîÿ.È:B‘0Уl¡Žè!`€ƒ¢<ÿÑ‚KÒ…r’ÝÝÝùÁŽ.ü¤4÷}€; e Ð…º@‘nP+À/ÂM°#ô3~"€¾-ÌåO\ar#¡tÀÂ]Юp+(€Þ §ªÐr‚Âÿ€Õÿîþž Äúw»¿Õ¿Áà¿‹ÁÂÑ ÷„ÁmÖ0(@KIåºí~Á.t=Ø s[¢¿'”du`4Á¿ô\ H˜Ê…ßæð‹"ðWô)+­äŽŽP8Ê…è×| 0$‚>vOàŸ›µ‡#ÜáÞÖ0¸•õ/V®NÀ'p˜³+TUá/"úOÌŠˆ<@Pˆ-ðW{}O'èï$èWÍÀ×Û á°F“€ú¬¡è?"o°€BºB}½ÿwâŸ+"`ƒ –Pœè?ÝÑa¨õŸ5úò‘0€©Z{ €À¯ß¿ßž¢åe…€;xþþû~šzjŠZê¼ÿ;''‡ðxóD|D@ ú!&.ðýgm0ìïÿ©U…[#þL‹>¦MìöW\ÍÁ øg/MZµP×Dn& "A?@ÿßRÿ]òSø¯.ÿ/‘ÿ÷@J®¿Ó\¿óÿGìsðü @‹Ö…6€møC ¡L«µ‚¹:þwVFAnƒ3H˜_@øOæ¢ó€ZiÃPÛ?’ùòËj08TáûõmAW üWí/ˆ=úûá‚ÖåïmŸî«‡ ¬~ùLPDF"ÁžDh9 Šˆ¼AhCZA=~+ä‡#P蚣/À$úu­h©04}t˜è­!®H$Úb¿€Þ÷_ëß~†B= ¢éIäaˆ]eHãE¹,£;ß×!¼ÅåæÈxã®çÔKo[uü å1g¹gVet©Ú_Þ„Nú0òì_ {˜µe.¤¨ä¢0»Våùòú.OÚl ,V›±ÔÈUòƒæÛÏLqLï%NþkÚx·^ð³ÙeÃäÁÀPôŒ-…+›Ó¼ƒmÌÆ¾Œê76#‹£@š¦¹ê«>ù‰ ‹;—kŒ¡¿ô\5KïÍêîJ:~õ¸ã½hY%=•—[Î@b=uB–Z­ §œÈ7ù7Éñwè Ã8°+'@LLìqôŽ€b&»YXf"Rë¶ÚHuCãè  g¯$Û/O¼—RC·‘úÉZ`,p¯¬Œ ²åYܟмÞ|©\ýuq*Ùm8`lÌž¥×sH7°×Î%Ÿ»`í|¨rKSë*ji£-=Þ©A 6ü”UçˆÂ¨ÉŒÞ¬z%‰Žè<4¦à³íɺ™<ýs :ù;GFëSÊG;˜“ròøL;†™¸ïïa(n˜.kX\©hG tˆ¾á…P.[Î8å,šXÖ(~"oñÏÆ=0GÌ ¯Uo ÷¿$‰Q.×[˜3»6Î|:Vbdx¥žå‡â‡=d„|”Žã_ß“³lÖÁªð^øú6@³’Xª“¿ú-Ž?Eÿ<0ß½î˜1ByT’ánÍVÐt<]øKC±W“—óÒ(Ë}U¼äš ™|¯í÷Ëׄ˜–i|ß5UžSÐ&ÑôP€‹eÈåy £Éæý1~L¾†aUO&ß¿¾·]îšvœy#S̵“ïIA™PWµ8~ dwÚ ÍvéšCŽD°êÉ_"8ë¨Ç»‡ÊåLXºãð¢–ÔW Ïns«ôU]œ9ƒ“ÓdZ±ÛN²ÆÅÙÕaU¾*cL'o¯½Çå…”@ù1„?Â’û™S]Ö(]ÞÂ-9|7`ä&5Ù=ˆâIƒÔv²õ¬,ãU7_¿²½¿Xš>â™ÐÍqÛòÕe½ûòÅ'KH€ ?‘,êH$ÿëš³ã]- 5;¼ùÄìðNșɧaÿš™Sva•Fm·|¾CS.úM°XÏ‚ˆ54E}"v¶¢AÐÂÃffþZý³Ï‚¡˜d·)”·ØyÏñ2Žn"¨ë3á;fC¢¼* Àò±vË›Yë~Y­¹„g²Œ8‚^ã$ms¢/çB<|5ç*zrvfS n¦æaG¡üÇw°.¯!8n:ö…»é7™“3ÀäìBWMgÅ>±©Kg‹M$7o‰ŠCÈßpŦ½?‚ìéÎÂEl¿d˜Ð[{yËD=îg$¦š¦¿b0¾¬QC£œûÉZ^„áĬï󟀬ZÎ?†f'ªQÉ‘»ß®û(ó“ÌWá˜ÎéOÒk"eš’\´2ôvU¬VÑ_êícúí¸ô®hê¬,}0{†—”ŸV¥""Ü9#¼ÉK ˆõ.ŸœNu‚© ýø™ÓOó.J3t‰CŠÊÓµ„¹rI{¶ž|tįÿÌsKŒÀù1°ómìCâCœr®>ª0q! LQII`Hgz»1kÎåÏ&‰JZœªø™Jâ.nX׃úlÎÖ\¹Q‰>w"ýüú(ø=†nÜ7—øŠ"bêKv+–c/Í}Va2Ö£[/ØÊ@ãk‹Æ÷ hXèe&VnÕˆ|PèGt Ãÿ~ãÂå ¤忆ïQkd-ŸYRar¢B†€@³d•âÓQŽÙ‹­ü ¦¾’ç«b?IgbðxÔÄü \rs¤/ËûbcÞû,ôgÑ~û¢p;²o/÷Û.êœñf-°çíÎ,AèVgËà.Û¤ÑQÅ{’[,ÉÛ펜ÜûqjÍdCzî4ÂlBÃÕ»{¥ËC™J–rª†qGªBÃR ê·ÕR¾:ߎ{RfÒ2oš©ö¹Iç1=aLzªøû•Ъñ|úG±å)½Ô߯gfmð}zË•*žé/µ?H¦|ž"˜–=–Ù0Û~Xv7Šu%»oËòëY°qà#Â’¾8Â;gbœ¦{ |îöÝ_jÏžÝKQÃǕ¿'ÿbù;޶âžËílNé»ëƒF¢eŠƒlä *§"[œrÞ ~&Ð×ÇD†ŠÝ$Ú&¯õUcFe#ˆqT¨¸Ã½=ï®UfÐÍÝÚªyö®!‘)süõ›±Í—Ai›‡ãqúó4”e.¹8lÎLÙIà»+ßÁT0²©öøëdZl"¨së]æI6Áëyá ½ô¯wJ)=ô4Ž‹”¬óOƒÚx#¦o½¨³bà@fjÐÉÍØêx}jÆz(v¶¾‡a4L¿‹ýÁR9²'*Œ¸ò8;teº€s¯ :ù$D„Ä­$Fò*˜aÖA]Þ"%zhå_b7S ÏOŽ •×M£gVOü ú'H0…¬$;ÈÞ…+äÃhQfaå…7~Á-À7¹Q ƒEqû·r‹2üÁ\¹ø‚3š"Ò §êV´®šb¡8~=Ï}„¯¥tŸ¸Ù17U?§®ð#hMíý²9I$¾É’Q<Â*Zá`ãÜÍ}Ÿš,d-ž;épµeüCC7“oë¾À¥K83uÆ=ãoÊî£;®ûx<±7–k,UF¸ïÜˤd¥š=”ËŸî$«Ž'åÑvÙ8ß&lK©m.ñ´ßxiô¬§;5ʈ¹±²€Û;ÕÙÒÕê‘yƒo#J0F³ÜÊm<,-ú‰mLjtí©A¿ã7|\ËRO™ ÔíøÍÒ’v§à„O %N'éÖåaRЬN¬"sò†-Ÿn‘é¿GTŸð8“ 7¨s&ú¤‹ ~²cº2¿*ãb²PŸšï˜L¤Ü_«.öŒw=oLÜ~ò­pB^I¡ðkø­ˆ¦O„¦aŽúä.K´ _UǤ{‹ô†Ö³;÷º"î8]JåÇR®®Õ{z±½ÏÃ$‘î`ÚDسׇʙñí<¡ˆ£Åós%—M³L®¬_f=pl¥Ÿcâ†;}¤OÚ²”JÌÌVù4ê,­øVo¿íÜþ$õFp­p$o¯®¥Ó¹®—"â}´òë tåf2\IÙú³)ÕMÇ?ÃÊÕ WfÁtHÝ;̲yï®×Ÿ’Ò”šý€Ï7BÉÙ8{7)"vj"9>™Á~œD |ËôG²¯Çß.›Î Œ˜ÜdP3N…Ë`Ë<çñäįŽYÆQ}dÇcòz}ô’°¯çûFWAJ#*mmP]ÔÉ mbºÉ!µo HXS`=ýò*CŠ©’AõµoO œY¨ãEÞÎÆä!ã½fÖ —ÕÙß`Äö™M‰k`|[óT•Æh9h¶dãs4ÉBá‹. ÝÅÈ|c¬jM‘¾•pÑÝq‹Ýæ"ëdìÒp¤)½“ž hn‡wxp”ÐÀå}I@gAëyÙ&°Ëê<Á;0™‹‰0©] :[r±Ç+¢ÆôUŠLFÌ$’X$8atWçzÖÑÚn1½îÔ$)ËI¢\%í°ŸO›g)U»aªØ"&ÙT·aZÒÒú©±m&XÕ˜MO¸bÃ2À+™èÛ)³jçSDðp%KË‹•©µÚü…mW÷JktzaÂЗvÑ4ñ(%? jñoWÞ§g®¶_<8Æ„V|ZðYäæ×ÃOñ–eamvuã:â¦]WQ75Cb$5Q6;úÞc"1N~ÏíQÊT³î”Ç‚Z%B„Çô…›Öt‰Fô>hœ-.~ÇKPÔ*À, {]E¾£=‹›T­‚iùðžóçó÷Ç)÷&HDÚe5®ïv,~ÂzÿIS϶}9-¿2±àÝà̧a2Éøµj¦&`õbä­Æg–€Æ—F?É*UG+ µœ2 EZóÀ¯IÃäv8º¿\ lNií [òTšÝݾ%=ÔN1ÛÙí8غ'1èkÔ¬Ô¶›S3g§÷›÷œ;B±T^¶¦amb‚C¿Jµu>]YíGµl®-‘‡5{Üý³“n8O¡Iù~þd÷únšÖÿî¡áBdiÖux•äõ€" a—„º”+;áºêâ™@srEà-yÖî¬5¯™ž–Í’I±‘cÿ˜™·U]ÓgõÕ,1?Îgº¸cy$|®ž¿3¦ *]âz9y‘ƽ°g[Ó^H¥o;eEsC•è·b\6¾ñcÓ4|•ïy..­œ©†+L·£:ÌGH#¦"ÈÖ1€Óž´´ ¤³÷ô+`å’Æ¯á@óÆgC5L !ovx{ÛT=¯¾¬&%F6Çpi,²cÇɆ© ëog¾½ÃÉÏÌlìýÍf×ô~T~æ.’³…aåVqèGuÄŽj)c£‚FÖ)^ÃÝŒ¯Žsá ¦äÿ¸mÉ †œ^€MŸ:£Ôµ‚cbR˜¶ÓßIIð“1ŸsjÀ±P-!‘Sðà×bQ±ÍÑ?øÏo±pcÐxYd†Ò}o“~ÝT%ñÃ3Q¶ UýÀ0¯tãá¤s–|¿¦¡ùúhãò2ƒêx€Ñưj•òq¢$Ë5ÙQÓ«›í'eZHæÍ¯|Q|¯PD9nŽç†Øº2bÚ:Á_…¦ÐÅ&Ô¸ *hšf™ðÂ¥ ®š0Çá’i[ù I¯£t;øñ>n øAoýlqÍ7›àxŠìõ;¢]r¿õ„³{¹T>KµÉš"w(üHz‚ôÔ‰ŽUèy¦/óVã¾Iü€7–ÝRíÀ";´ž‚Í(×@$ÝÀ yá3è¡› ;êHJq,ûÕÙüùChÈì·Ù¾~Œ å¡#ûV6l7²Í“ªÓ-”šÚ]&•혻cƒàç—åökÙ®.§© Yˆ,¢© {ù¢íeŸªCÛAï¼èùÛÛ¯¢ŸÉ_å^·kØ[ë< '«·ôUç”.Âõ9ô6Çl¬ ´,Çõ}ÁMØ5ÐR‘ˆ®5±SËk‹"i³uOH 3Ú¡Rš¼,f%ÅL‡õ:¼hß›‘rYÃp‘ôÍñmêäö„ã=×+.Ü@‰]ºæ'Ô¥/Jöº¸ø¶Ã·¬ñ[‚ˆšz¶\Þ÷DÝÎ1œäµZd6™C]m£[ÚáÕ„×LŒ$¹,r¥Ym–ìin]%ÀÄÞ쉰ˆPe"åÖ:“Mf%=««¶Â“)¤“¯m‡þû7xîÚVUiÛ­7{¯=c&nQ?u· ´¥=­–; AU€ªY²oû¸a‡íöw¿×=!~Šý¬ŒAwÈ5h/Ô$žphŒÁMcÙ½•Š zŒÓD´ÜzT#ûɧ K™wA_.z+[wýp*˜THkòHvU®Î¡}Ô{ÝÅáåðè\ °:lß?Æ`‘]壸‰× Ó,UÒÉ øL™Iôb?å0–ÖÜ)µU‹<§$¾ËS™GµÄ!¥QÆèˆÖjZ<”÷K¬V%(¸5Þ‹WÂ}QL®…$$)®"±Ì„QŠÒSÔª‚ âóUú¤Ö_&OÂÕ‰Ïå“îŽÚf¹ËÜî*±JGŽ|ñ¹¨öÎÆèÕÿ*’§ƒ˜lÞkÞÉ–Ñ.Þp„F| É™åç)ýúÔ¡²»¥ï¸L<ïF8ƒšêZrX k‡O¹x´ó‘AËQãð@éá™y’óh%Dnr\]]=¿ã‰ÜjÛ´†N/ç·—Z޳÷qSæ”ê4‰Ïf³è2D=8éU°’7Ë$ªVêÚVñzCU°¿D0”¿ŠîfQÁ¡­¦C^:Ì”ôœ­Çd3ò’”A€¡ËkNÖ‚áBè~‚q»È\önžÕÄF…Ùš¹tD¢£1Aït‚ T‹këúgo+fŸ‚=å{$ÉB¤‚{ˆ(Æó¬5–¶æO9ì!K]Ð-¶zA¥¯pNýùë")’RüåÔG/$¦lr éË׸Sù…Ô>‹î„Uˆj±bFÕÐ.ƼTqè–Ô\ÆàÓ2I¾6ûØÆ6{ráoV[AŸÓ6‹µªÝÀ™²~GÇ«ûy/ 4eùÊ`R¥Pø¤ÉH³[¸èKŽdu §ºÒc›¾b6äëp¡f$ /©S´¦ó6”Û"£Þ@U%†³³ŸÈék+¡"Ùa¥ìý®ï¼pp4iË®F¿¯Þv±¬îß$öA:+ZU§E#Xvž.ž^üòR® endstream endobj 58 0 obj << /Length1 1540 /Length2 8620 /Length3 0 /Length 9647 /Filter /FlateDecode >> stream xÚ´Tn6N Ò]R£FI·€4Ò5`ƒŒNFn‘îînF’FJi„ù¾ïÿÎ÷³í¹îëÎçºfz-nYk¸%XsåæãŠäÕuuù€ P€äÇbfÖ…º:€ÿƱ˜õÀ(&ö yäú€=¹>Õá0€Š›€OÀ÷TŒOX ð¢á1À3;Ô ÎPÃÀ.XÌòp'/ÔÆÖõ!ÏßlVì>QQa®?ܲŽ`Ô ¨ƒ\mÁŽ­@¸ìêõ_!Ø$l]]Äxy=<. w0ÀáöóùOß°øøÖP+W€%Ø Ãú7ú †üy~¸Ô` |øûóÏ?Ó…YÃa^ÿÒÿ¸b^] U%yοZþÇ('÷øp ¸ùB€§¢O~ÿE ý«ŠÿðT†AàÑ?‹}˜Òß»ÿ¥¶¿Öƒðß±4àºØþ•¹ PhõðÅ÷ÿ,ö?\þÿ4þ;ÊÿUæÿ[‘¢›ƒÃv¶? ÿ;ÈêàõãA·n®; ØØÿRõÁ.®:Øêæø¿VeWÐÃ.ÈÂlþ$ÔEê ¶Ö‚ºZÙþ¡?á—¿÷Ì kÁ] ¿_7ø?¶‡å²²x=\$ù‡ ü°;ÿQf·þ½düBO ä…|P¿À‡ïa­ÁžˆÀ˃»>¸ºó@à¬ßW*ÊàµD€¬ìÁ/ Äõ·ñO\àüÏ+üÛð…üßÈ}þ?ÿwCü xaíüŠxáÿy~êô ¸õ¿>@¯«ü_Š€×Œøø¯X¹!ïÀ*}˜Ïßç?0Øl…µ0 ·¶«n¹¬”¥òàÞ“œbÞÔOfçöY@´º]ãa¼g¯HZAœË¾ì"XÚP`;“Y¤ûå³ßX‹Úÿ¢ùÆ÷ÖE”5~O§ljåw%ðÑÇh•?újÆ{¹T—ߥƒ’‰Òˆ‚õŒhd’ÅGn'Q…|Χ0?rEr.È΃´2ÒKç;ÜNÿ]ÌXM™ˆè€XUaòk®á 4èÆÁLßÔ±pâ¨vÚÇ›£à„xx(*åj6ä÷˜m „b^„!8­ÕÝc04¼|¶ÖÐÎ ÷ÏT÷¿W{CÖ1±ÖI—êå–Ÿý uö#QG§þ¹%Òá›g—VÄTåÓm7~à4å¹å©ÓÚW¢p膇…‘÷>Ÿ€8/õ“0ó`ÜY±  f…h³%¦I'»ñ å¢âË/<Î'[w‚ßsFŒ’ETo¼=<¨ë¼Æó•…,‰êéì¬â¾‘–¢à¥„sÅ gåÍ«V‰]2ʨ|±.—™lV<@Uݾè4/ι±&d´á¬¾Þ—¬ß‘ÚG›6ʽj¨OÒ§×8# .dÈ»¸!aTÝ—·“—ÚLF¥0úÑxxŒ%;J»Kpgæâ7òz">…­“=¤>Lø,™Lîvpè´B±KÚF;Eµ¨Æôíh¬„ΔVPCÇ=~LéGɳâuãêîøÓJFÿ䣩N£y¯!ÏŠÈêЪþ@}–gÆqS»Hi,Du›M3â ¢ÔÍo´ü®üÚâVÚ»'Ag³v<æ¿Æ¢²‡ŸÓÕ-ö¢Ìö.û*¡É¨Mj§ZœIˆUTõW& ù(,C¦+,»¤~£©…Î#½ù|ˆbP•)îWýî2“dË®×ÐàLaέ„®åƒP³Ó—X9å–üµœpÙmQ¶°Ó‹Ö(A òó”éÁc7ã\ÃÖîÄÔÌÍ þc׉σC‰‹"kÓ¥m¦„7X¦¯Ýq,Р0`Y÷}´gh½Áy“öV,€!(û¤ˆ=ºÈ÷åäQ_Ñ~$„ýª„øY]žë%Á´uRWA¶diZºÐè–WœTŸòí„;™÷¡2÷IdëÄ„ì”È“aú¼HÎaO^.M(îV]Ä6ò%¸*Œ}®_ΕÁÌ ç£ìþP­h¦-Èñ>ã´+ß™,Ã,þ&…ƒÏÆãb›¿#G7?;«A ]¦dݤRØû È{o’z—ým=ÏjÊñÀÏ CÖ¦â‹=¡0E«W’0»6þõÖôY+’X>œ¨ÛË‘Ë,eÓÐôðN¯†j±è‘o¸Ùœ¢¥½ˆøÍ½“ü›BˆÍs»SÎô„N“–Õ§(f¦zÏ£äNgKœÆ*òìþdx’j½†|¬cËöB±,@n¦>Õ÷Ø íugÒ¡Q~:öþÂ=ëûزÓöR?ûžÌ»j±6}Ö-.ÔP¤;¹^¥bJnĻݜ÷3d†Ð4ü`9.û ÔÂm‚Wc% in'Í †«×EGR„“f´Hµlï˦XÑS®i°UYñU×§pY¥²ÓVµ4¸ÄÔw|Âu”øµñdžšd­JsϘ‘D±|Ù9{¥HÆãl›ÜýèÂ*ëôÆ!T\è1² QþÒi´À¢¬ìçzªºhú7s›Ï–ôëI8#-.=CºŸÊžu Z Šé;Ó›¾‚öåx4Ç[Üïä¡aD²nvY¬Úš¡wáèêÆ(—Àž¿;|û”KØ:úî4ú¢;Û‚hù|v#09à|nÀZHë¶‚Ç&‰VbÈ׸u^ò6ûeûdi!ïæõ#],f –jŸ`6 Õ^1ù訪hn»çó^nb÷u…ÛiAOs!ڰ͘셦ÛðÔ®Ÿ™ÖëB¬–ÖËÎ\‰ÏÏyšñ•;"½k˜° £¥ÓGÎn£´õýU³ê (m?;mnv§­yQQx0àßý9zZ‰ÖšÚð‡ÚÁ$N—ÃuX|‡Õ~püU¤¶Ä#Ù'.ê4ÒBœÌºa?J_¢Ót"$¢B³Ï‹l7‹|íîYêÈ·ó²v±ôzh?^Û Iîõ±¤æ@i<~„ û’Rd„¸ËWDh½7Ѫ¾FíZ(¾^WÁФï(¹óîËéïþ ÀÃ.òA6ÝÕ‚3i ¾·°—𸼠7àµ7‰Å½ïÝ«2Gù¨õöKìX4 vWej[縹»0Z j¶óa@SU@…Ôv°iÿKÌ^>º˜UaüžQM×OƾQÇB êF§¾ù»›yŒ•÷úÉÔÎÏÔz[[¢}–¾[T“ÄX{$PùÎã*©Ü{óÖ§ï–'•·ÓgD0œ÷XBxŧBÍé+a÷…;31ÑÚŽýÐ’Àëàƒ]xÞ­<ÌOPªÞè1PéÓŒ ï|it/ªPœ5;­T¾æ'(º(µj§tnå Ê[óßÅó\OŒÑy㛥›+•î08áqó„(Y*Üòxè‹ß¼Isu`’”S³·W­óÁ…<¶±¹åBvØõ-ai¡ øL§>âø0Ítþ–¦Ýqé“ÔðË«ÝW£ÝŸ0JXëQHBa ø)c|‘©Ûùú_ûJî0 îõ–pŠ’5ÜÐ9Eïñ… #¼”ðÙ&»îVVp q»:Âʼ‡Öñàš]æ ¾¥VFwÔÞry[¯eêdé–Û®Ve@Àž‰›T–Ò±ÛR c¤ï‘ï1 ;ަϒ¿ÝÕÍ·‘«u@#^ˆ}ùKqO¥BšÈØÉ´ {ñÑÓ˜mä|göXǻ“—("ã3ñÞ9Û_ÛÔB¤k®QâÉoÅhU#ÄËÏ®sôï™G ‚ÏИæCÅÈ-E6jͬSåÆhgU´ÙŸsër|¤Ošô2D²`U¯ûņŠ]ö¸âû:q‰–,c$.@—BÉ€& ¾Hº3(`y§Ý֣пmZȯ°Òƒô çDÚ–]„e„”WDy}žOOÞ «;v%¨W—è@5"{·ò­0=e"­÷» Ô—Xá³Û„® ?ÝþÅÔ&XF®¬­l©mr¹å躽Ø7Š-ŸJžøÑÔ)Ô²mè yâÛ+¢ù¾²c[ßE¼fü1¤Œ>*Þ eJNÔ:XwÒH¦Ü~÷þvç}“2*WÀtZF|àù†¥\ÖøØÀ휈ïÒ¥ñ̯ËG&âMÏq°`}kèVüa+žtcè$Ù|ôÒb«¤k„‰›d®Œ´Ý«dÒÌ“ïm‚ÃuÚÁÆÞ©YàQ¤7~€¡[ö¤ëg8¨vÜ×o÷h?u?óvË‹•±|Ñ‹ õ¬µê[¶@|!bAtlSãâ)ì¡ÅPÆ,6šK>ÄåÓÙÓ±~ô‰bÓéð{^WÙÙdŜأp²Ãö÷„¦Ø—ÚD¦ŠÊåºöžÇiv6–™Ïã­ó°‰au®•¤SóDN $1ç99·ÿ”"{U*e¶ëíɵècönÔÊ‘¶þ^^R×x–3ðµÔõy¼î’ü¯ñ9À{Mñ çõ@S‘+&¯†z¹0¸7ÀGÒñ©”‰ xÙëÇqöšùŠ#èºøÚÕÚ%—,MYÃÛêSsjÁM’ÿè«ù®†L0`?‚fÚº„š9|ÃÔÒÅ›µ}@ls¤/[[œ"%HÁ‰õ %_ÈmER…V÷ôº,þy)²H¡áZ=¬xbÜç¶Š;¿åòúq¦ëýg?­Ûƒ×ÒÜÏš÷"sSÚÔ_ÖxÈM|†,Ü–ïIèèê ~2ï6 t†Åܾî2™AÀ‚B4Tèêöî"ÆOS~e+–Ù—Vc×<:mÓ´kOj}R¾²¸ÅY‹¡~23áR¤>£—Á^|™œø£s£XåÄPïd8M¤ö ƒ¯óº5£Š™4±zí€41CošeÁÈ‹N<ꣶFˆUýÄT¢¶µ[ÂÆÁ?ê()NȰBu\2PgóÄtßK&,/ÀChæ¹ËjÐÍÙißïézSĈ½õ²Û–Í·W ¿r"f üXÞO$lÊ\U´›k*¯4 S¦óúÄš¡y{q[#F)NYÙkY>öï DšÇæTuu_æzSm(O•Ñãê¾nˆÐ.fšâúHTxÀ›½#‡”w°È¢Œ.ÒY“Ò«Žéñò"ý@0i~±‡+}šjh±ÒÈž75äŒrþ©—ÇQŸáßDÕá½Ô$ñÕ+ðå–ü„LH’ᇪ¬g¦MÃ+Žn!cOoü4w)xt-ëŒgýS ÍÂõrk)³Ñk©µwbhÈòª'êöŽ×ððp‹à!*IÑú,í¤¾úFlWÁ{ÏÕ0e¶BÇ„­5»?l ;ëG„€âN)+–ÊÈ3Å\È™1Üi9Ô%FÐi“½vçÖíD5”M%÷ EÙÎ,ãs›Ä‡_}S£µœ&í䪰ф±ŠÞ—•Ó“?Q˜a§¹{ÃGÝIzÅJùdÔpئ.œ*=¾bȳ Ðý,“Èð‚*ÁÃiÌ.4~ÆÖ€ï¢Fœõ¬pâQ­Y¾3¾Ä8ÛÑW×BJØÁ`óXИ69¦u§C‰.ƒ-ôi¥¹;BOó YnÈcäXzâvVP±íëÞÏæœJÉï}S/ù¥OmäI^–0:aù*c¨]õ†qµM]È&ikÐahÙC%.ôu‘7ãóbqW÷-Â\Ç4ÃÞiõ¥:ŒÄœëúp˜u±Gi™ä7³5²qª¥’œ(ï¿ß©×&ÔVá.~½z–Ò@ß”!Zçâgd(¯¬ÛŠ“:¶Ý+„HjPöšÒ…À¤=Ï ÚœõV’ZP”{—’/q"4w©ÓŽŽ©_ÄÊ*ÔözLÓŸû–éÄwUV…ÖÏÚ1Ú Ž/‡ij 6·+;šûð‹¶ñΪ©®R6…§Á:OñȮ·ŽÁy‡íôÈå um’ÑÐÑ«üÈtstAÉcaAaY%ã(Ò´i›¬üæé»cƒ|sN×¾~vÙY<Ÿ•ÂË£ «×s•Ùq%…ÞZ/†‹’tæ´7­9è¶4¾“ÏÇXÛ †‘ x´º™÷ RPW(>š·;vK¢òzqò³Æ òs=ë„US-·y]zØÙ®ø$…k(z?uªš3óSŽ,Qÿ$Í"w½¹è)Ëó‰yÑYGÄ#ßnaD5Ñ¥+f±ÐSÄÝ'‰¯l%6Jf]=µ5þžZ»èfªë_§)É;W ¯ ß<Žg¸lH`"‘†Ô Ó&-¦®ø6¾3þß Û¸Mg]çÔ}ö>óÏò¼Ul{7‹>xOâ†5r”œÝxl‹„%ªáJHIóîhP³ÔjËeô›J€–•‘ï§N’GLòÆ“´Ðð9ðñkìÏ H7Õ«m}âO|eã¼× Ïõ1æu¨Ÿ%«¬dî²ûß“0§X·jJwŸôŠ“fi{„®4ö+” }fÈÔßô Ÿ)W~z…’`€Âjb¾£zõÓÅ?’Òç…,»Ç•=W ƒÕhR¨€[¤6‘Šnš³Pœ¥/­ÿþõ;‰]þ›aJKY¿7Y)O>ªLk>ݾá%`E±£Ÿ>iE©m¯+¤mB&=±‰TFf:Sâz?¢náå§´—ŸN>$ÍnõnéÑ׉©÷ÂkªØô"rC­ÑȃÙ%Á¡ä9ï§jçèiWS¿(í?ÚN³lOÈWW¥LùXÂUÓ”·`“/ƒá T364§ï:g¢|ã´Ûœi¤ KX¯‹ð·u®gáʳz&=_L¡³º6QÆæ+Fuæv!Ôåî„ò* Y5D<ðÕØ18ÄÜ#±Â‡‚Q¦ŠŠ^­Åƒò=ÉÄÇ q¢-jq.ªÆúÀQ§GGE –.«Pg)®ž'F°!šîÂ4­®HðÄÛÇ*›¾yøM>i @’é@ûªþ>«în©ôúQE+꺊˜•]±@~„Kã]˜€¼|¬¸&Á;ƒÄ+²YiQÉöéPл)Š’ëkõáæðOç,~j[0·ŒüW¿dé¬ròäš¿±Ü+dr=/‰åÛãùä78Û?Õi5y[áó­DB"Íýàb©a…lÄ{ög¨Ä5ö(ÒIâÀće$ÑúWŠº@†tË ×O)CÉL^»ų9zSýDŠ®ÙÏÓR–7€¨™„g¯è´òäÞÕësÐÉö¥…kfîé`îkvf@1äMAc³Mè›:ÁUBcï/¸ý„ß=žî#[àîLã oP–/ñáh ‘ O^1ê­ù†QN‰ÐŒcSzŒý–LôtƒŽ¤%&Wº!´bIøì—ÆE5-«‡Z..ò&QµÃû]$}÷/õ1Kˆ{ZcÞ5 ×LÈsŒòדŸZ”Œƒ o¾÷²Äiux6Ô²(Q¤¦ÜïçÅ@%Ùð‡“²w¼Å‚Œý©L{×R³h“¦µ9ÛÊîå^Z?n­#§ŒiQƒ\ ›¼'=¾‰OÚ"Y}M¬Ø5ÀÑ9 òɈÛÉ0,U´'uiZ;‰@£'I•„1pø­íðm6å>,vÙ,Å¥iP,Ë,Iݧ[T!¾·žd’eDdrYnÃUbQŸq¿Fßå ¦À;¿.‚À-Óq Ù_“žôù8ß! ÇÌwŸSzú/+=º¨qÂö»‡k¸æÛ_ÌnDÖêw—F#g=Š-yFB[ŽGNå…ÇÔR3cãD„Æ"•Däy /D"ÊlSÈã *h_ÒéS±ô¡;¬l±äq˜¼Ì"‡ƒ µa ÓÑ–—0nÖâ(fNˆíã,­xTÑŽ/(’7m¤ q]ü» †½ÙðÞ9>\œR¦$±«³²š"Ëž×ôJ5¯MrêÉt îNÝÒ߯Þå„ð™¹çö¹žCÊÆA`—Ü*uïKÇjµQR$åÞÔy¢à«$l×ÍÖ´ï—9àzªh;Û^8Zšâ²¸Ïl.#‚B"ç¸vwp£±obL(Òôò—àGÖm•õ´ìåü±¶5›ÐGôZµj“#,Ù¾;¼:þÚ*±Eœ»û/Dâ‰QÕ©·ò| ‰;Ò¸¯H?ˆG%†ˆÿJòdNi×!”f`L áqáŒUö·U¿ix7WîE<•Ú˜dxfƒ—àÓ²†=¦ŽÞ–µÎÙ—Èû¶z3ºbÍ ´ü²þ}¾¹€Kœô/ˆè—œM‘½~ iKùîxiâ2¢@÷E÷Ì뀅è}oI&æ,[Îå\¡Uyi“j/qzQ’ wW‘D¦|,ïÀ(;ƒ.˜X3’O³²–ílô‹–ØPI2Êéñ‡û’g&g­³Jòçs’M»CrUeTí-™RZZL寇^ (×V¤ÔSµ«©âŸßöÞä zÀó!uµÑßü:øèŸ÷×Ýf}ˈF¼ðBÛy>«úɳV±³5îpNfþ £ßÓ¤dg¨R<¢Ô率TÕ%ÿöÔ'kîmÄà{m•f5Ì( :ž¯ÉŠ&ñ?ý‹ëæhjë–Ëîú2º1"?¸f§)+¤ÐýŒÂÝ endstream endobj 60 0 obj << /Length1 1680 /Length2 3950 /Length3 0 /Length 4992 /Filter /FlateDecode >> stream xÚw 4”mØH˜’²¼ÙªGK–™±eß÷%{$ɘyÆLÆÌ˜†,!¥R$EÉRʲ“¥‰$K!ƒ¬5–7J}–7ïûÿŸó}gÎ1ó»®ßµÝ÷ïºÏ—°sT0À’½@S2‰®€RDjF6NN(%‰TVD"•`p¸NÛaðà •F “470Œ¨ šÙŒÑtˆhC&–þD¥  Ô4Q5‘H@ ‰ÔøM$S5ct Ø(–dHƒÁÈ” *ÁO‡êüþ È`d”†ÆAùá€/H%`Ð$ÀMǃ¾PE š8’1ô¯2Úx:¢‰@0 E´/M‘LõÖ••:pi 5Äë#‡Ð¾à¯ÑapÀ O ýt8’qtš HÀ€$âO‚Tª8ZX¶ô“lý“ ü:¥ˆú'ݯèõDÒ`4Cö¥ IA’7€#AÀÖÔZ‘H—Ð$ì:M¤‘¡xtš@D{A„­£S{ Møk>†J ÐiŠ4q}FÄzè˜MHX#²¯/H¢Ó`ëý¨ :÷ įËõ!‘¤àßG aqëc`ý)gÁÏ´0þÅL°?6o¨"‘HueeôÀ@ ±^À)ˆþp¢ÖÍÐ ¡Á2ÀAc€¡}Á‚iè SýÁÐàŽ# ` :àzH°?Ù!3ˆû‰¡û§£HH~(¹þùç×1HaX2‰ô‡þãŠFGœŒí ä~üÓÐ+ %e@UIPÓPBÿÅMøÕòO¤ G4~6 Òï†~)@æ×zÈÿÎuˆ édþÈÜ©ŠÄ@Pÿg±ÿùÿi|=Ëÿ*óÿvdêO$þðËü$ü?~´/ô‹éÖŸí€ ÚÒ©.àÏŵ±ßÿz-èhh HÞÄ’@3%‚X;ƒÿ¡Ÿfçõ=#H ™FXYù´\èõ A’üá¡ÝùwE†Œ]_2%U5M¥¢ƒ`HHIJªª@0 ÚF,øCÄB‘D¦C!4](€#SaëWªq@ ×M?‘:€ðúƒ4æ¤ ù0d"ÔèoËz¿ìˆà¨ p 2€ðÞUþ¨¢ ¡ ܘ²6@¨>q„šóýQP+b¡C7@¨ÊŸbj‚´GÞÐ; j‡ºBµi C¢Ñ4ü”ƒþþë:0þT*ô&ýØè®~ã ‚Ø›^2F+êDqTÕçB1†³CI…·™ua‘ÛeÌD$`Â>ÒH{eA93¦Ï ®gï㓃I‹éu­1“<¤Øeû0yº¦Ã‹µŠÛúE‹[N-î¨97Ï!a½³t5m´ð†ë±èNá¿è«ø“~n'²n¹?4ò/`·íÊxû/Øì>¼“ŒD<]YÔÂñp/Ü祟,=>w­÷[—ná~I+”ýuáLîz”ßÖðטsŽo5&ù4Ý–¥o<¾Û½½ óNg¢Æ†5þhmñîaêÒ ÓŸ»WÔú–ÆÑÿP±<©XjY¸u\ÄbI€ÖÚHVy(Z.‚Ĉpr„mÊèš“EÜ0ÿ*õÚ¤ÄTä[ôS¬—À-ùäŒZ~Í×–V-3w‚0†ÏLe¢œ‚ÙØ¾Rm¡‚•s¶×+Í”¦Ñô÷½-Lƒ¦¤Só×D¯´Õ¤Uâ¤ÞiZ«f5ÏV&›–=™j¸ ¿æ@\‹¬«Ð9ð›ÊD !†ø"µ%¤kÓšïÏ÷Ïj™´=ȤD¼v¸•hQ7Ymp7ÿt±tûRP«K¢å¾bÚÛƒèR‘[új CûiÆÌ$> Oâ|è†ÑUÞjñª^} ›±Ž¹Ûã3ïWù÷ÞÚ\6uaSqvý}ŽZˆ¾i‡elqí–ô0ú#)–í-¾­^róÍÛíqo2¥c➊ŸL˜¹õ¤òh¿šÈ‡Í.ìgè'ï;Ñ’<5Æ<Ô[V–¥ŠC‰îNr.Î ':«i„|asèHñT¨;¯®°bÝJ >«0)CÖ0R—zpÝ9 VèyÂTžV5›òѲüý$“*êWÀeÔߛ֔Ós75üÙÙoñ¡Ê<ó3{Ä–b"©Ò‰fƒ’&ÈJßÓCV’Òˆê)Vˆéç“ä´=A(yWRcpgÛ¢™ÊbµˆÖ=—]²Y¢Ì[n0É-i·Q:1©`Ú‘›³2ËsD݆‰¡ÙÕ;t&ÏùXËQ‚ôuN"n‹´Ð_ùl÷êÁàLÁÄé.Ê-ñ—³ª_L¾ \ž ÄëX‰?k=Ö]š…{óíç3ëM“u>H†Â¸ý¯dÕ\êíOynl®©Rƒ£Í¶Újë‚Çu{H‡œ éÖû»ä;œÒÚßû|æÐjø`'{pÌÀÜRðµòu_ø‹§E¢GûëCF,˜\¶1…Ek‰^ù·{Gf<²75#Z6¹é\:FÊ­OXû"ŒýÒæølî•‚Ä=J••Œ-ªðñèçm!çÌÖZˆ÷ñÔÎ »ô¶ìzÿÎSª‹í»$ ŸhpYO6ñÈ,â|€òæÇ®X~kàe°ø¤"“]kOLk~ü1žÜ›,J }À YõÑ}ãs<Ø Þ²(jXÏðëš“;+4_,ص¥&ä°ýÓhÆÔu]]¥‰Œô¹x‰ ’W›Lün4Uáu}÷ŸQW8O'å®)ˆ½½;t‡ÇêJø~Ô„Ž¾àüël÷s3öTó<ã»2hî,rµÃéáÍtÙœÝ ’6å\Õi‹QQYyZ¶[f<'=2ÍÒ¢Ø7 繜ÞqæÎ.K#¶3õàœôÃ(!j è‘eIÞœõ¬Òòöüû©»¾lš}V;no/H"ݨʗ¶[©éØ’®ç¡S¼¤:Ï(±(Ʀ —¦—;Çó½©ÌtãÍ—]婪¾;ŽQº©ú1~_d\m÷C WZ¨d×¥vï®Gîf=o‚…¾Ó^²8•*ÑS¿tÑX4HwF,¡ïšâb†å©¿§­¾žÑ77W· ¾ÈýWÇþƒÑ[ö[½Ù~ñQ^PØ\ØõÈ—¥)°216-‡’É/ÇûNwpU—ßa >Ò3﮳>¼—FbV$ÊTfÛWéM ^h†YG×åÙ˜kŠ?ûªi{LSÁ³-úH>´V'íp°é^‹)<f\0ìs‰nÌSn°UšçŒñ¾Ý66ž;÷ôÆ­X©F?uÌÐiÜ#ä¼Ô»YœÍæIÞú~ë¾´^¬#µ¹ÀÂò嬓øŠ5ò±uH[¨Q6>“Ïo,ÐÅLa& o°yæÞyCÞ¡èTöÊ0>Ž­sÌKý¾…{$š»Ï–’£E£HÁQŽåÖß»ÈS‘[Y‘:‚þ®Ü6©=ÜóekæßOÄrÕ8[BL/É\ÙA“Çrˆ]l+ülºÏöˆsbÙò¦Ø)>7Ê˶Óûì£ÅŸ® _ÆÍßy¯–‹|<ßå²ä±”Sêûzí«ì¼jg|ðà²/éÅQʋ⽮ÛÙu݆ä¬ÅJ¯ºË<[½M}Z"Oʱó¼?”WŒfîýu†?QbßJ>á1rÏTø..SĬRmÒðÀþhíhOYåÞçÎ'"ŠQ.왋ÂÔát-3ƒ‰%Ž¥ó™9]£á±¥Obn×íJÆMé»õ^ém_U\*³vmšUðˆ2 ã?‹T~oÔçJˆÉÖÎoúfxÁè E3l!.Õ¶"y­l9Ñ*þ•j澿ï¼â©X.ËG¤2ÜbÍÕ%-9ç٨‹¬iç€PSc¹QšñÄ[Cåû;Txò2L žá€¶è!Õ{µHYÂeJ ï²¢j4ºMpŠ$§tT .¶PÕ©y¨ßñ©æšö yëëäÀ'©dÿßêááR÷x*ÉÛô¹¶{ÿ¹qv}ÀÐÒ<†?Z‰þ(9æPW9ß\ëÞ—àxGæã+÷/33ž+)Ü'äÄÏrà|¿èêœ[°ăQ’ªÏ»f“Ôí0¡Œïsú-Aǃü ·#åV•™nF™CµSE³6Ñç~+§ŸdÐïˆqp”6 mYÏê¼Âr…/®ß›h';¬;íS¤û,ä«…4‚fèÝ ÃÞ¸²7Uc…d‰~å~¹*L~WPƒçpRNß”¹žÎR§:|®ó~Ât^ïLJ{Žõ¶z4”9^ ~ºÙD<âYJ’)òjϰ,Ρps= x+lÙs° ´óû‰ã wß–õs?d›öf y©YÒBGküß¼³¢z?÷°òÕ½ª€î7ÉçVoñÊ Kž fQZmwœ~Ó-±+ýìçö»‘±åvE"Ïxšn¤séù.Ö¾Ñr˜!Væ³Ì'…_É_bE¦¶pÎ÷|$Ì(•æ[9Â>ôå`ÌåɨΔi8²oÀGTCÆ“ïð¹ªó£^óå¢s%—x‹¶‡2"¯­lâöYæñ*ªIIÉro[¾}‚*I”ó7 a¹Ÿ«;ftÊNtW4XÁ²Ô‡ËŸ¤;ƒÂ‹÷(*U¢§6÷\i×Ýf•|„~­€xZk“BeîôÚ».Ô4JÎ(+ÆŸûœ)WE˜Ý’f‹_‰øZ› Š™k‹·WzÊä¸ (èâ¸',|[‹–gµ_ ÿ’ À½¤ýw¯¾0Ã×#ãòüz—íG=Us´Ç(Óf—ÆžæqÓï“7oÛ»f‚ñÞ LïS{k~ ³ìñs[„ÈÜŠ¾|AŽaÑ W%Îá%ˆ§±ýC“/õÚvÇÐ॔Ï{©"‘vÑ)Õã°Õ59gã=÷¨¥ Œ'‹E—é†Å¦ï”{bÃò¢2ܧô?¢îwU]éÔV›ág5ý Üè|1q©ï#{B¼µz²QõBvdò)ÍöØ{KµîTØ #•ç߯Jï¬q_k wb¬Ö<Ö3kÐ98}r³®ÚxîùºÔ}‰¥ÑTmé-Øå¢y¹üøÛ’;J·uÊïˆgeÝ ü¼¤ùæcgâp·íåä´|ƒ†Üú¶ùÈ!‡¥ÅŒy!³â×+W3¥k)G);†Ü%ljÉÁjú‚êÏG”='ã—¿4]GZ(n^Áž—=9¢{(KªkeæõWÎ.aäÇù箑yÎßù-*-¦Õ‹r„ƒÌL³®šg:²ª„u”Øýè‘Äñç¹Õ°esTÎv§÷+›“‡MJVÞìM8t$ªÿLµî¨ègb£ ½W­`¼…ÙØWº91AI»Ò=nB„Ñðy:‚“®Xu/D£¥™Çº±‰ß§Êþî ÷œoó—"÷²Dêcè#˜ëC[ÃMDÊsžÔ ð»í“òîÏ “ÅÇBJê'Dez™¯š¹˜Ñ¨xðük ¤¤ —r•ýûðHSÔ51T®-ƒ{—¹oµ«ÂiÒîô¼Lxp[òmê0j.êðƒ¬WçøÌ“\\@c%±5ì.˜!í—Ò_ò9ï|ßè{.yÈ4‰]],Iî~ä×Ñ<á·WÄÓž½XU˜hnÌdΰžW´ª†Žo÷Ͷ·9+c.xÏŽ½ˆmõ—M‰œ¼½U=6ËÏ+n”Sч d\^°ïöVĸæGfwVepYì)ãÞ¥T~Åum0;ùï¯ÚÕ/ll.vݰ¾™½’ÙË­òภœÜ„çY_±Ç²òaNœšT™ÎU¡†­m³»œþѼģ6¨Ô÷E–¸‘$ ¯sïœÓh<'ݬ²¶™Ïûpà C½Šî-®æ™]Ÿ>ƒ©ÿ¡ˆÚë endstream endobj 66 0 obj << /Author()/Title()/Subject()/Creator(LaTeX with hyperref package)/Producer(pdfTeX-1.40.18)/Keywords() /CreationDate (D:20191029194120-04'00') /ModDate (D:20191029194120-04'00') /Trapped /False /PTEX.Fullbanner (This is pdfTeX, Version 3.14159265-2.6-1.40.18 (TeX Live 2017/Debian) kpathsea version 6.2.3) >> endobj 2 0 obj << /Type /ObjStm /N 51 /First 382 /Length 2689 /Filter /FlateDecode >> stream xÚíZ[s¹~÷¯Ð#©SX£»t*µU¹È.á–ÂRy0ÎÌÁöx=cί?_Kc{<ŽY ¼PuªbY#µúòu«[ÊX°Œ¦2f™qÌ1ï™g2ÓL¦œbB2-,SL›À3™dÒ2ëLOzæœfÒ0/°$0"#‘ùÀ”`BkÅ”dÂ)|ƒ“Ó`©éóB è-ØÓSK æ=H<ÆA$ÉÆ_° ¬´† m˜¶4èÈð •$tBרb̰¢ÖV˜Œl²Êb20k¬Ç ³ž”²ÌeÊÂ) Œ`Ni<€òÌ`!Œ²Uœ‡²"0¬‚òÌgÐS*æe:  ÀWjæ5 •Ì› zã }¬ð6ƒNxvRõ,ÈV½ý}ÆÏ\^”Œ³U>¬‹rÒ{ì·ßzΊª*&·ìÓ`4ÏY1žÎëÍïu¾spã+ÆOŠúšVîï÷øÅ—iÎø‹ÁmÞãGå¤Î'uça¿Ê«r>æ|Îò›bpX~fïˆÂÓ—ÎÂÑ÷áº63¬'ó"õÁdR‚Û;Ä É½î­ÉŒ³=~XÎnòYä'®ù~Êð€À¸&ùÚ½Nö  6R÷á-³¾Ïà,õÕùü} Žüi1ùÈö÷#~1âçüòÕ)}ÜÕõôßœ¿/Êbò¡œÑ°ê§ìám9£{Û¿Éù´l]AvõöO¹‚`¸Ø¨~@Mæ£Ñõ6RH…ígA~ ©ö¢oáñ5Òx„‘OhK¨Æ}/¡›>¢QÕôElê{ Ò†\P7}MÁÙôc€6ýŒ‚;õ±!ÈM_RðRñ³rxžÃ%ˆ˜ãÆ/òÏu×­P’®JÒ|G,­ “v;üýYa×6¨…Ë·Úö.¦Œ_÷sýË[ð ¬Tè#}ÙG­ñ¦Ó·"ë $ù@­U}ä}úš¥iDù>* ‚ ß§qÓG©Í,­EÁX¨¶€Ž*'‘dCÞzš. ˜FéSÙ`fr J“‡@ê£2eø"þˬ#"šÖ¨—,Ú­Šx7_:1$å+ l‚õAÓˆFq5®/b ¡–g*"±!};BHBm|;çúž¾Uä¨V9O–z%H:„í&¢¡ Uëɤà"†™"¡!Ê4¤ºÍ5‘N²Æ“…Ö…bŠM#^zÒ7ò¢YWâ/ø¿âŸ×l’»Øôçá›·ÏN¢Mî~“Dc‘pâ'YäÅV‹†åx<€çþ³f‘ÞÅ¢ógW—G°èÕ6'Á듲ð#&-£Ñ†­ð#~ GðSæ3þ.;‡ÓÞ Hßó!³á|üa”†í£rÒ pƒÎÿšFüÿ€¿ñ|Çï¾Lïò àùC{<ΰbáÓ|V”7|:šWü¯yYã*ñY ýúï’ŸÿÍ?ó/kðÚ]à½8¸|yváݲّïãfÞÿtÍWÐ=¾ãÿÏ"¶¯€îÛ%”îƒÈ€m ýgi”¨S/A× ÚFèñÿæ³r B¿ „¿=9}ž"tËžS*EhШC:|3„f „Fµ!ÌÚžÆ|9Œ0H+tÊûs¦Ù)g>~ñèêø%l=zq±mCJÚŠnêÚ–¹B¬™KKs¥”¾iC6[rµ|P ‹"íE„ÎûÙ`ø1¯c04ýÃíy;ËØgíèñQ^Uk¥çk±7êÍûQ ²j4¨î(Ôîfyþ•€3;%ùã׈µ×ä„·[]Ÿ¥4Oo7œYù[øûҼζ:a<ÕÅt´–†ÌNYþÙùïž?&m«[Ñ™ƒr®€?âía5ã|͜ݲê³?]Áœí›D§=bÔí‘o©Y[Ã?R£}BF-ç³&R„§P¦x݈ÔRãÑÛ‹ã— ¹ý<¢âkØ`~‡íÕeQE¨,ßSEVY²U%š ÜFÀnI˜ËWT,Bôá Êãû‰3õ`ôb3¾+9)fUMúÆÈ|:h„ð=þ¦¸©ï*zGº³ðî9·#\u…Û5á2[ ×» ïH;²MW¶Y“Ù•l±»ìÎѱ#ÛueS„µì+ÙfwÙsUGvØpx¶M¶ß]vç@².Ûl›SÛ‚-ì.{ó€Ð¿nJo3ý;6Z·4v„oÄ›lÉ–+ÉjwÉ Ö¼lÚ·$ëæò;b­Sm:²7ƒÍ¶dµ”-¿ÃßÝ ¿.Ûnf6³l­ î6…Ó[æŠ^3ÏéqÿQÜT‹×êé qçµôóy=B«‰¬ù‰ \ô3±\WÔTÕšŸ2°Fßæµ´]s•4ˆp\ž×ƒY½G¿ˆ!]L¡(ýPCˆÖ³Üc2½‘oÿ–C%µùÓb\Ôv-ºÕïò döžˆ< –·½Óò÷ Ú7z[½0í6?+or~Yå+¸žOóIú-[V˜ÿlæ† endstream endobj 67 0 obj << /Type /XRef /Index [0 68] /Size 68 /W [1 3 1] /Root 65 0 R /Info 66 0 R /ID [ ] /Length 194 /Filter /FlateDecode >> stream xÚ%Ð9N‚Eá{@DPp@TP•‚ °HÜ€-±1n‚° j* íM܉ ‰µzžÍ—S¼ÜâIÒO"%²ß'¢Rp†V`Ò–òq·ëp—Є\@®áΡ·Ð‚,d¬ô4F7`r°y(À6ì@J° {Vî3ö­ÎwÔõøU¶^Q‡Öü+ªb-›QGv¥ulúQ'öóÿJÕõ¢NíÙ0ê Úpбߞþžø1Ö/Gt endstream endobj startxref 118165 %%EOF pcaMethods/inst/doc/outliers.R0000644000175400017540000000343013556147050017410 0ustar00biocbuildbiocbuild### R code from vignette source 'outliers.Rnw' ### Encoding: UTF-8 ################################################### ### code chunk number 1: outliers.Rnw:59-60 ################################################### library(pcaMethods) ################################################### ### code chunk number 2: outliers.Rnw:62-67 ################################################### data(metaboliteDataComplete) mdc <- scale(metaboliteDataComplete, center=TRUE, scale=FALSE) cond <- runif(length(mdc)) < 0.05 mdcOut <- mdc mdcOut[cond] <- 10 ################################################### ### code chunk number 3: outliers.Rnw:70-73 ################################################### resSvd <- pca(mdc, method="svd", nPcs=5, center=FALSE) resSvdOut <- pca(mdcOut, method="svd", nPcs=5, center=FALSE) resRobSvd <- pca(mdcOut, method="robustPca", nPcs=5, center=FALSE) ################################################### ### code chunk number 4: outliers.Rnw:77-80 ################################################### mdcNa <- mdc mdcNa[cond] <- NA resPPCA <- pca(mdcNa, method="ppca", nPcs=5, center=FALSE) ################################################### ### code chunk number 5: outliers.Rnw:88-97 ################################################### par(mfrow=c(2,2)) plot(loadings(resSvd)[,1], loadings(resSvdOut)[,1], xlab="Loading 1 SVD", ylab="Loading 1 SVD with outliers") plot(loadings(resSvd)[,1], loadings(resRobSvd)[,1], xlab="Loading 1 SVD", ylab="Loading 1 robustSVD with outliers") plot(loadings(resSvd)[,1], loadings(resPPCA)[,1], xlab="Loading 1 SVD", ylab="Loading 1 PPCA with outliers=NA") plot(loadings(resRobSvd)[,1], loadings(resPPCA)[,1], xlab="Loading 1 robust SVD with outliers", ylab="Loading 1 svdImpute with outliers=NA") pcaMethods/inst/doc/outliers.Rnw0000644000175400017540000001203113556116437017757 0ustar00biocbuildbiocbuild\documentclass[a4paper]{article} %\VignetteIndexEntry{Data with outliers} \usepackage{hyperref} \title{Handling of data containing outliers} \author{Wolfram Stacklies and Henning Redestig\\ CAS-MPG Partner Institute for Computational Biology (PICB)\\ Shanghai, P.R. China \\ and\\ Max Planck Institute for Molecular Plant Physiology\\ Potsdam, Germany\\ \url{http://bioinformatics.mpimp-golm.mpg.de/} } \date{\today} \begin{document} \setkeys{Gin}{width=1.0\textwidth} @ \maketitle \section{PCA robust to outliers} Away from often showing missing values, Microarray or Metabolite data are often corrupted with extreme values (outliers). Standard SVD is highly susceptible to outliers. In the extreme case, an individual data point, if sufficiently outlying, can draw even the leading principal component toward itself. This problem can be addressed by using a robust analysis method. Hereto we provide \texttt{robustSvd}, a singular value decomposition robust to outliers. \texttt{robustPca} is a PCA implementation that resembles the original \texttt{R} \texttt{prcomp} method, with the difference that it uses \texttt{robustSvd} instead of the standard \texttt{svd} function.\\ Robust SVD and its application to microarray data were proposed in \cite{hawkins01} and \cite{liu03}. The algorithm is based on the idea to use a sequential estimation of the eigenvalues and left and right eigenvectors that ignores missing values and is resistant to outliers. \\ The \texttt{robustSvd} script included here was contributed by Kevin Wright. Thanks a lot to him! \section{Outliers and missing value imputation} The problem of outliers is similar to the missing data problem in the sense that extreme values provide no or wrong information. They are generally artifacts of the experiment and provide no information about the underlying biological processes. \\ Most of the PCA methods coming with the package were not designed to be robust to outliers in the sense that they will converge to the standard PCA solution on a complete data set. Yet, an applicable solution is to remove obvious outliers from the data first (by setting them NA) and to then estimate the PCA solution on the incomplete data. This is likely to produce accurate results if the number of missing data does not exceed a certain amount, less than 10\% should be a good number. The following example illustrates the effect of outliers and the use of robust methods. First, we attach the complete metabolite data set and create 5\% outliers. We mean center the data before we create outliers because these large artificial outliers will strongly shift the original means. This would not allow for objective comparison between the differnt results obtained, e.g. when doing scatterplots. <>= library(pcaMethods) @ <<>>= data(metaboliteDataComplete) mdc <- scale(metaboliteDataComplete, center=TRUE, scale=FALSE) cond <- runif(length(mdc)) < 0.05 mdcOut <- mdc mdcOut[cond] <- 10 @ Then we calculate a PCA solution using standard SVD and robust SVD. <>= resSvd <- pca(mdc, method="svd", nPcs=5, center=FALSE) resSvdOut <- pca(mdcOut, method="svd", nPcs=5, center=FALSE) resRobSvd <- pca(mdcOut, method="robustPca", nPcs=5, center=FALSE) @ Now we use \texttt{PPCA} to estimate the PCA solution, but set the outliers NA before. <>= mdcNa <- mdc mdcNa[cond] <- NA resPPCA <- pca(mdcNa, method="ppca", nPcs=5, center=FALSE) @ To check the robustness to outliers we can just do a scatterplot comparing the results to the optimal PCA solution for the complete data set (which is \texttt{resSvd}). In Figure \ref{fig:svdPlot} we plot the estimated and original loadings against each other. \begin{figure}[!ht] \centering <>= par(mfrow=c(2,2)) plot(loadings(resSvd)[,1], loadings(resSvdOut)[,1], xlab="Loading 1 SVD", ylab="Loading 1 SVD with outliers") plot(loadings(resSvd)[,1], loadings(resRobSvd)[,1], xlab="Loading 1 SVD", ylab="Loading 1 robustSVD with outliers") plot(loadings(resSvd)[,1], loadings(resPPCA)[,1], xlab="Loading 1 SVD", ylab="Loading 1 PPCA with outliers=NA") plot(loadings(resRobSvd)[,1], loadings(resPPCA)[,1], xlab="Loading 1 robust SVD with outliers", ylab="Loading 1 svdImpute with outliers=NA") @ \caption{Figures show (from left to right): \newline Original PCA solution vs. solution on data with outliers; \newline Original PCA solution vs. robust PCA solution on data with outliers; \newline Original PCA solution vs. PPCA solution on data where outliers=NA; \newline Robust PCA solution vs. PPCA solution on data with outliers / outliers=NA. \label{fig:svdPlot} } \end{figure} \begin{thebibliography}{2006} \bibitem{hawkins01} Hawkins, D.M., Liu, L. and Young, S.S. {\sl Robust Singular Value Decomposition.} National Institute of Statistical Sciences, 2001, Tech Report 122. \bibitem{liu03} Liu, L., Hawkins, D.M., Ghosh, S. and Young, S.S. {\sl Robust singular value decomposition analysis of microarray data.} PNAS, 2003;100:13167--13172. \end{thebibliography} \end{document} pcaMethods/inst/doc/outliers.pdf0000644000175400017540000034102613556147050017766 0ustar00biocbuildbiocbuild%PDF-1.5 %ÐÔÅØ 16 0 obj << /Length 2488 /Filter /FlateDecode >> stream xÚ¥YKoãȾϯP(`Äa7ß¹yœìÎ$pÖ˜1v±ØÌ¦Ú’0©ˆ¢½þ÷©g³eÓØ{°ØÏêz~UÝþx÷îÃiº0elMž-î iã,¯E‘Ŷ*wëÅoѧfi¢þÖð·ßIg³\¥‰‰zh>`3¡éæÜp§íi-£3nßá–™­#ü÷8çàç4,¿Ýýóæ^×yn‰©:®L¶XÙ*®jË<ý²¬˜î}8áÉ jóè+RÄ>üµxþwY…§»Ñ$íùä:Y„‰/…BÖ†³ÌnÅÅ*¯ãÚæ‹•Iã:«˜©ë+<|7·?òA·K[EÍIxBÊîĤ?w!ճꂎ£­(Þ‰Û×Ðìðs Ö¡gÜÜ é†$%êu%ï7ðóÌÿIòz·Ÿ¯?Âz†ä1I+â”, ³²¤í í÷LîíCÿË¿4v½ Ùù,˲¸ÔÖdR¨±EœærÍ .ú]ŽBQšŽÍZYI3Ÿ…ʤɗjDˆÎoT#®¥Õ¨’õvÛ HˆQ Ðä€~/ªÝ<çUçà°ª»EÇë…‘aÍîé•ö#»‰,RgÇÏ’ÐñWe—õ • rÏHöø7øù€÷ÈY¯J…:©¥E.Ð@äB;™;®–« Zd:ø›%uí舙˜,ëd]Ù4NËŒû -ƒÇõ÷ËhNÝÂPý~j£û™Zäµ “ÅiV°¼BÜ”Õb•Ù¸ªj¦k–+clÝb\2NtÌ8†i„æ&áÊkP©²…-ð¤X7q•ˆN¯žÐ^ þ@˜dYE^sâ £~ÿ@.ÕqÏÝö¸üI±mÃ3Tî00’euô¸Ìsö»nmdeÝìșᤞ ‚`,`Ý¿7¸‰"ž5KÖډサZºáóˆš#ïIÙl,iÿ ÑÑ!—€ÓzòxÔ©5O=é)[î;ŠEXsL<8žñFX˜é5)mˆ5v/²T»ÑÓœäxþùïÜ&­r“À… h/¨ƒH”¼¡uG‚àPtIc ÷{Qì¯ÁæÄjà8ÄcQˆÖ°ª [Çm÷»ì }d´mÇ¡žÃ7²o§âí… ìŒà4½öÖNŽläO ¹g[Ü2Œ¤HS´ªZ¿˜3g}/ãs˜ëxŽœ#³)³æ}ñ Ç,ðˆ}>‰Öye¤‚ªÍZÉnxôxÒVG9e¿ï"k𳉦ž8®y“¦JR1¥~°O™ÖÑæœaNî#ÅØ½‚?V eÉòbCðŠ;ÍZÅwàÑ@‹¦©jTNÚ„ŸJ`I}ñ,s]Ñ=œÈ@²Á×ìl9L}¢äà<¤¡HóÔ&ö«99½o9×:À<¼ÜÀZy’v/˜43_ÉÌ’”/13+âºÊ”zA.›‚ëL€7REèÏs†<À=^KU+Õ JÞkfåb¦ã½l¯‘S;¨>€F?žÃª¥â9™!¡X(qÿ@ê[ÁáæµÔ+[qiËKE‹3ÃAk,‘¸¾â!J°Ç½GI©'4‹¸’ìß Ï›$:9õìš›ÜÕMùh3’ß©%ŃŠgdOë¸(½ì_ø£8w+ª`3¼6v—™÷–9_}Ïú¤!¹å2ÎVYœaµþz·âiÊ8Еªí¨üniÖFç-kŠ:;‰#hRL¹aFRcð¶Rÿ ßΓ8M¬ ê.5:¿`î_£ëšŽ³×tÀ% Ô9y©G oóbª¸.½0˜¯ÇNK+ñD Ñv ×Îù傲$‹¾bņdYj{,Å.„)zÆQCª½tUXîSeÎENKjæºÅ-°î¢ÅOš4NŽ0~ 𓻯yåN2 ¬ù kÄoÊózëp‰ý&º” áA³á Qs¦Ø”€ºoß)¤ÂœHí´ –8W5Ѱ»@ù¯âû%=ˆ>§Þ²›Šs>8!ÿ â|FdGïSePU*LXKê¢áσ8Mˆª¡íác«9×rŠçˆñGüq\ÑŸ&údY,Ä:v2/pXÀ(ó •|4UÁ¯Ùz3Ë@s 7ᜠC2„ãg_ªÀæáø£¯öhþ—e•ñîÍV‰Q-QDÖE“|÷¶X•IœUÕ%¤7¬nº ¿¶Â–Rà_æ.y`ƒÔh +·»ŸÆÉh'­’8øéRwP㇞…Ë"å—ÑñŠ­ãHïNg½wx TI¡¸>úJÎŒ0J¿s>F3þžR0ŸÄµ¯ˆ’©†‘»Gpd7þ,¸¼àŠ«fÇM-ô|Ê«ƒ &uæ°åò2#|U¯ï®8Z +ŒB$i¦ a+=»ÀÐÓIÓÖ†'^¼S¹Á°ðNcà™)4'¡ÎèG*'žöáµ7Þ  ­óðÆ6ØI[Ð&ñún!%œóAÕ„SØ~Kzu= /‘žHÉ5~4xÁ S1Aœ„á³d•ñST¿Ñ7®ço>T´µŠ¡ƒ{ˆøÒÐT¦’Ò ðL£[®qSPŽú“”„saïŸDLVˆ¯ÈyËÃ{ÂF¹b8irMA©ÂÑÚ§Ê;œºÓ)'gþ*÷FéïõBÏ&º/«g/<£U 4ÃÒºÎSv ×Ë{i‚½–æ[ r#̵h«<éœûù3ÏúêßJp5šeFä¡ßéA«9[XÿmøÓâM\ÀNÞ•p8¬ålÁµ®Ól‰ä×%8£“÷ [ôÜañZxTrôöØè› Èqç˜ôĦ˜›t-:ôšX‚û0þúñ­Òa¬-ƃ>ðaG½ÛtáRO,™À"ú ÎËã–¦Y¢:ö|¾ünº„ÖØÿ÷=~Ñ&…=¿±¥ã)çß¶)ÝkŽåöùÿ»@È÷¢àÅa_—ôÓVO­MàèV_^†i³4ÉøßÕ¯÷Ï"[ϳ‚ z™Û‰¦¥Çù“$û^ÎÓAò&†Í·KgBòQîÕ ôî(Â9[3•+Î^Ä€ô™q7ðH'5Ò ÊÃÔ»ܽûƒœÃë endstream endobj 32 0 obj << /Length 1780 /Filter /FlateDecode >> stream xÚÕXmÓFþίˆ*9ÒÅx×ë·–V¢PZ*¸¢»”ª¢|ð9¾\ ±O¶sõÏwÞÖ^'>z´©¯wggfgžÏì÷Ë{ž¦³ÌÏbÏ–—3f~bâYľÒÙl¹š½ö–Wså•óE¨2ï†õþ¶õ\'Þ{m*ø[órù'ŒóÝ5 ›6òºÝÃ_ÛÁ_ƒd8([$I=_ y ã?â³èx½¾äµYtÂoS «–ójuÈ-õöí ˆå‚{ê ø£Ånþfùól¡ŒŸOågQÄGß•"/à¼+æÏil¼§¨*Ð ÍÉ|)ØA IB>#þòb®SïŠg­z8.P•Á`+‹+½ìü‚d“Ù72'4+óKËX0:e5ˆhÊÞâ<}E§ÆçÌ»!R9œØ„™÷Û< É”:ðv¸Rè¥@Ê ÏÝÙm¼Ð;Æ+«AŽïÊÃSÅ^Iˆj„ÆÚŽ˜ª›n÷ÿÀ7æ½Û-Zˆ&psެˑ b6AÞˆ„ `Ð ÑùLLâ¯å•÷=Êù½ÇyMñ±ý Ó¨ ’^г`®Ç hÄËkŠ++—ÀWâŒ81•>¼Qb:Ro+®‡mU-"ò­,HØNøž!ᦌ"ï-Ú°°ö¸AKT5à:·Ú¶uÅ âƒiÙ›Œ£¸\‰‰C”Ñ øÅ„ýÞÚÃtm¯•DЬfu®/É0`5ê{ÊYÕ ª¦-멃·…Ygu¹&Suí°åÁÓ(tÒ$2€7- ¾céгñyaœ‹ä‹x’¥*dòŒâüÂäʪªAU•Ž$­„E'Oá¾s8æ¢Hm±Ú'å=9ØþXèv¢áÖáShãN´Ù‰â½fÞC/x©uλ¸}NeOXr!¯•³„Îø~KøÁïWøý0l™RÉ1Ñ?‚ßsøË–0I!JöYXE®YP‘½,Û\ fÙ:z¯t8V[‰ž¤ƒ<ˆíC~ øŒ¦Rܑߔ÷‹h„ñlj«ì@v¸¬”÷Zæ[(ïU¶— HY ³te.ö—}H‡zøP„llÊo%Œþsd9S¼|L.¤qKßÑ}'¦¯…!}/$G0!·Jç͊鯞÷J&.&¸ºqÈîž;q{+»aàúÂ͈8¼1•YÌ÷EÈÌݶáY_ §VÈ£»EÔ¿‘騃,ú„Ó»Àû¿ZâL”¼ãàó[¢qïE™Î1CþE,t”ÜP9•Úeb¹eTÜ{.íh¿ÉQè§a|hÿKÑ™“Á‘4‚g2KÜ‘=HTiméa×§[¸ ¯Äü[j”öÎκÂZE…½‰;Ö¾-epÀm\jÚʤåÅS3*¢‘f*å,’ÄO2#Ö¼ý#p*½„ÊÔùœ \nIþ*0£œvjÝq‡ð»¬iÂO «ØCâú `ûÒðïR'–Ø‚!.¡-0®¸D‡jûϺ(‚Wnuûƻ➛q$ÔÄ.úHkt}¸¡\jûâÐ{Ûñb®I…SÎt©«iGÁ}oßCHu>nÄã!åT& xŒýVØcd2{¤9ØR_¢xËô6d‘¢µw£â©.±/2BÔ2„IiFm;Ls››óKß’ÃXIɦ¿0™4qÇÙë ¦?˜G¨ •jmÓ˜ÔˆÐEaà=#Ç©ÁÜsƒM3 NY|¥Æ¨ÒNJÜõ úŠ)¹(ŠÕT'‹ÓÄ5_mlÛ2Ç|ÍM]k‹P¼^+*ä\ãëŠ^­†É6Ô9¡‰µ1ÔCÖ;£TIäeU”ƒå§ªÏ4ê“™‚ìe²Ìû)§Ñ;{ùÕbb¸UÁÂû…Ï©çžã©ööüŒìí Ž§ˆ¦n`-d羸Ø&sã*—ÁGÍ„ÖÁgî=1<ïÃÉöÏyÃ+¯PUÊû’gž¦Ë ¾$õÁøc;_ÓW¦n7£Ï™l9£ý H¬n§ù@L¹áYÕC§sª ž8¦ :|žwV›¹pPX.ç…ÍQUaS]¿žuÂ~‚di(9&jÀ÷¬¤œN×iÍpõ'j2ŽÒÚ“D¾Iõ8#´Àáùf/âM.–  lZrp‚O!ÿ‘.évHfÎñ*#K9 PÎR„æv¤è(õuO#¶}î&t^ æHÒƒ›Ð›yØÑðÝ] p¦ÁÁ4yE[𤕻*œF7kcè ¼)lõ™7|kfû "ø¾phZ~”h·´;}t~ÂÛá7¯Að5>B'ÉÙdßèl@ˆ³U˜úY˜¡ÓSÃ×îý°¼÷7f¥U‘ endstream endobj 40 0 obj << /Length 376 /Filter /FlateDecode >> stream xÚÕT=OÃ0Ýó+<:CÜœ§MC)T‚ Ê …¦RÛˆ¸¡Ÿ;Û j%dÙ¹ó{÷|/Ž,Æ d"±dr EÒö¾ >¡³cb…_ç7z·[`×eðˆ£¢F3:½ÊƒÞdÀ2‘¥2eùâûà4Z+–ÏÙ3Ÿ„À×8—u…kFJ)Ã0Jõ4„JnV˜”¡ì󃣿Ä:Æ­ΪÜ:Ò¦ ½Kö¶Àr«FrE`5àc¾æwè&™Ö®»’¤Š%ÕìpÎ6Nr:9ACÚ\ê½—.wùĘPá̘S,Ù²ç$NàÌå"Q¾ryYûœTÖ…oÌ\üuÛUùF,cO÷72ÿÓÔtŠñ¸ëÁw²îz|Þ1@ß^Ñ~ªÈªÛöÄ\âr?:iâ©û6!æm'çúG–ퟚW©­ø]Uóì\Vt®xí5Xô¢Cz áœKa]Bš •júH!3oVÜäÁ­ƒ ‘ endstream endobj 29 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/outliers-005.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 42 0 R /BBox [0 0 576 576] /Resources << /ProcSet [ /PDF /Text ] /Font << /F1 43 0 R/F2 44 0 R>> /ExtGState << >>/ColorSpace << /sRGB 45 0 R >>>> /Length 2748 /Filter /FlateDecode >> stream xœÅ›Is\·…÷ý+ÞÒ^èó°ÈÂVâTT‰ËŽXÉÂåEb)‰T’‹Rü÷s€74Õ$Õ§©Ò¤`óŽíôd²ÓËé×ÃÓ¯S¨³1“e®irÖÏÕL6—Ù„éíóéïÓ/‡¯®ÿúÇo¦ÇO%Í´ÿùôñwø¿9N¿~üi2Ó³ƒžà¿—ËÓ_ß\¾úÖ¢½«ñÇÛ)Í΢¤éÿÈígÀ¯d¦«×Ó¯¾œ®^"÷êíáW÷Ö.l;ä8¡vªsEmçfãϯÃòm™]ÚNœçPÜì…ÚÙ̶LÁÇ9¡í/ìyg獵$*w¦+ÜÍ^±uµ´+ª`Ó ÛÆÓn‰˜µ… ~¶¼_am”êÞÏ)5Ÿ( ÆÊ9ÛÆ’6óFÎ5ß"Šgõ`yW ÏsÁ¦õe¶‚µqp 0ó¾4ƒ+,\¥O¢î:Çƒ`m\4mêRsǥΜ8¥KÊÁÈYnͱâØqÔ|­Ú¦… Gk-SBõHçÄÃ5*§6-¢¤Ëb ܽð^Fˆõºµ‡Ã5Ò¯¦§7 ,I7×Òò-–ź­ÀšÞJÀP'»+±¤×Îà63»^,é­ÄQ¿vý\Kp"ÛDü|}b"®F¬ø­óÜÆ¾ýèéÔ¬‹Ç9 ýJäÿåd¯Ëtgu‹[ ×ÑqýxF} ­”ãúöœökóàëŸÑ¾ƒWÒq}·¶?"q„g¸:_À<Àý }öoæ{Æ­#¹ËÇ!‚ùZó{rËg-­ù#¹ËOð¹ò–ß“»üÞ¡5ëß~ü»A?ÚÍýxŒ±óð*aNNà`—ìòÓ+p÷èXúåÛ:‚!¯.9†éËI]Òë‡b¯FñɵôÓÆWúÂÇüà×À+*‹ˆž¢7~ ©ÿüæÏ^üòoTzú·ßŸ19øgÆòä´˜á£M¿½x÷ŸéÍûw¯^<{½|–cô>Nˆä±GÛ¯xð­ÆJT!ÀÀ×¾ùªP=’]cáàÆ+q}ô–Lƒ N9ïÑÙæ.ã®RâúhjìqËŠpóc»òÔÄηêE:qA6iœ冸^Èñ0vJD¾¯ñ}ÑJØ?††ƒ8É\øÒ«k(Œ/¾iêp¿[.À¤=vçå§Ù Ÿ–S'í:ø6ƵΟ|t¸˜&ùXh<ÆI7K¬éäÄŠ{ ´¦·¸oö@iMo%ª9Jkz-qܳ]O(ùP/"J7Ó%D‰÷ñ%D‰WÛ%D):QZâr^œD6#M{n[僶S¥º,Ñ(ú%º–é} Ø ·ûÆHïKà ûoôô®ÄèÙZb×Óa;tì1É,ÿ¶ÍpÏXü¶Ýp÷èÇxçv¸çç³¹s?ÜóDÖ%Þº!À£9\#½­eYn¤o ®“të!WÀ¯‡\þ)§¸9d?î>õöÍ?¿œ—}ñþú݇Èë§éêɼ:ìÃVû|²)ç@y^í²)ǶéR“Mñ}Ù+Þt“M9o.M‡:Q6Õz®´ÝdSŽçK—MYÄßÒ›t—MQO!ÉwºlÊ qvÙ”¥MÑeSº"´²)[œý Ù”-Y"]]6ek¤ÍPeS–V‘ÏtÙ”‘‹¯Ê¦lб²)›4Ü2dS”?%¥z—MñÉ.(S×eS&2("Ž.›âÖUÐòMYµ=ÛeS¬^”ÓÞeS|2RëM6ew‡lÊ"ˆSbÇ!›²ä]ºlªV‰ªÕ÷Ž¢Áè¢)¾RÅXtÑ”µI"C4U“$ëš©ªº!™²¶U2•ôŠ4SÙkJ·.˜*šÀqè¥àë*¤bÈ¥x³)mwµz®°ø!–µh½PwÁªØó.•JM¶£*¥p«] ”¢7§ë¤P[¹Ó†L û\q&†J >°òÊ=DRðÞ•ÇΡ‘Â+§j?DªGS«ò¨'7Ô’ÝSô©ŽäQKöHîò»jÍïÉ-H¡ÖÖ{r—¿ïÎÚ»-ÿ¡dQõ2UT½LU/ÓDÕË$QõvETŽ|9YQ´†R¿ûîëR¬ÏÄ¥øÒXS¾ËT2ÅÖs–ÑÔPb©lŠ ®äd8åaL•éUhÉËxŠ2$›eWöZåS>ºu* j©ae@Åê¼D@Õ$X: ¢‹˜DT”‘ÁN¨€ŠÕí‹€Šrd@E5ù–¨¨à‚Ϩª!‚RUèzQPQ½f’ ¨â(x* â_ä…$*j¡ˆi4@E5JŒ* jê?'*þ¹zÐU¨H)ªP›g/*ª°pNT@Õ´$UT¬ -* QŒ ¨"}©ª* €p1‹€Šµ ¨ÚÛV¥gNTT áZåOðÜE@©ð5* Š®r—‰€ŠÚ§bgbˆÎD@ÅYsFTì9Oè§T” ¸´2 è©ÇF¨nääÆˆ¨°‚Á_òGr—C¦5¿'wùôÌ76’[þQ‡Öþ= £ò¤®^†TZÁp«*øÀÈB…T.'+Cªˆº“jDÜ †Éú €=U#%U£ÀU-%Vµ”Ø`ÕRb¡UK‰ W÷k-q&°êê¢ ˆUÿÀȪàf5¾p´ê_¸…Z˜Ubµ¨‘\9Õ²,7ÒÇàjpªOˆ®H^øêº «Û4Q'þpE}¼ô*­ÑâÖÈõÿžýéõß¿{~;Ý:üìÎf endstream endobj 47 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 56 0 obj << /Length1 1724 /Length2 10030 /Length3 0 /Length 11125 /Filter /FlateDecode >> stream xÚ·P\Ù. !¸†àwwww Þ4ÝÐ8»»C H'@p·—àÜÁ=A™™;3÷þÕ{uªúœoùÚë[»ªé©5uØ¥¬œ,!òNŽnì\œÂ5iC.n''''7=½.ÌÍò9½>î srþ—… r{”É‚Ü ÕœÊîö.¿0—€0''€›“Sè?†Npa€,ÈfPã(;9B\1èeœœ½á0k·Ç<ÿù0™\BBl¸¤ päP¹Ù@3‚Aö'0 âæý_!˜DmÜÜœ…@OOOƒ+‡ÜZœ™ à s³hC\!pˆàwËuä¯Ö80èº60×?:NP7OxØÃÀG×GwG+𘠣¤ Ðp†8þi¬ú§à¯Ãpqpýî/ïß`Ž8ƒÀ`'g£7ÌÑ…ÙCòªn^nl£ÕoC½«Ó£?ȳY>üQ: /¥=vøW®`8ÌÙÍ•Ãfÿ»Gàï0Ç,çh%ãäàqtsÅø]Ÿ, ?ž»7ð¯áÚ9:y:úþAaŽVÐßmX¹;õa.î%Ù¿lEÿȬ!n>NNN~!ÄñÛ'Ðõv†ü¡äú-~ìÁß×ÙÉ}lâƒB_¾® À îñ÷ý·â¿À vXB¬aŽÿDC âÇùÃa^cÎGúq8?™>2ÌÊÉÑÞûó?F T‘–ÖÒPeý«å¿•ÒÒN^_v>;7€‹‹G ÀÇ ðÿï8š Ø_upþã«äuýYîã9ý§d¿8Àôׂ0þ;–ºÓ#s!¦ˆnÂÉÇ ~üáú¦û.ÿ,ÿåÿJôÿ­HÞÝÞþ=ÓŸÿ=Èfïý—Å#sÝÝ·@Íéqÿ×ÔòçêJ;Ù[ý¯NÉ ô¸ RŽÖö#ÌUæ±Ò„¹mþ¤ËŸr½ß‹fs„h:¹Â~_-v.NÎÿÑ=nØîñúp}ää*ÈãòüwJ9G°“Õï-ãæã€àp7ç#•¸ùø¾\ëhñúƒÅ ‡£“Û£ à±9Ô Žñ{¢ü| ÔoÑŸH”ù €#ANPóÄ jÿ„ý@ÿ AÐòôü7úÝ#Ðê_ „ü >†…þ ò€Öÿ‚ÅÂþóØÿ >&rør=&rü|Läô7ä}´}¼—ÿ¥~Lìü/È Âÿ»þ ò€nÃÇÑÝ<þ¥~< ÷ÁÇ2=þ€ÿ59°;þxý±[cýþ㲄@¼ `Œù'°H˜muXëu•¹'ûÖ¨Ø$ý–A3»ï<¼Íý'js囸¥TÚ@7Þâ†Ó…äÕï~S-jdsªVË/¿[ódíñ­Œ¹1¢Þ¯ö¥jz(Ñ)Øu%·ýî\üôƒíš;”és]Üq4óž_{~Qðªé)ý61³¥µ]ɯ‚y[:Á¯g\4EÿÞòí4 Š;% þ±îÔÅå$~Î×*ådV ÿñ<¾F«Ü 7Ó>KåºÜ®¤t¤F$”HøÃã ¾Ò»éÊij¾Å…‰Ê±…PJ³™ñ6\ ï'êŽyÑÒ¡¯kC$ß̓eJq7Æ#ÁZëôÒßíurÚIðskŠ&î¸Ìó1\µÚwîˆ|b?N½‡JôÚ¹hrGf~N±vÁ$Ò¾¹›>Ë?Gš×Û#0ÁX›ò(™<ÇmùÞ͸Ëä…}ÐþDàúÄÿëWG´wÆçº/ÄuÇ(hAƒ\6Ô9"(ï ú{+ðìÄÉKâÇŸ¨ØZQîðÁyø¬þn*±`&º}#‰õduÖEwË=lΪÔ2že+¬ÜåøtHåW -·N^•ñ«ýž·Ò–øÃåjˆ!NaÕ•ŠЇôèœl`ÜÖÒ%m~[7Y3Pç)`ÒŸ09RßÞ¦Èmënäy†ƒ<%%qjoŽçÚU¢¡ÙpUŠÙžà Õñ.^›é°\}6òÙBÈÒ9¡÷Rs½?©JÝ>ì*ænS’02Ñ/;:ÿ¾¹Ks*֬ȟ¤¡‘ ¹³s c“½xÔGeäÛ žò€Ë ÅOâPŸ--á'Ó¨-T83k#¯Æ Á^Í¥Dù$¯Kɘ²BðEDcøç²øT¾MeÕ N²úÜ>Sh{ωÈÜYž#<[I0ÎC$’Æ.y/ ¤?|‡<³•kðД­mŠ]÷Fè ²‘Q6¥ƒ˜Ø),òSRð@ˆ’‰C„:BD€c櫽­¬Ë\RNÒ;¨*V6üéÜÌÓ´Ö`/h K# Ù;q‘Dض©‚_wA?…bß À¯Ce3úîH‹Cj÷2™à‡5>Ï{úe~¢#ë–Îùøy*õ åâ™PMœ6@LÀèÔ¿äÍb’)ë¾$œDÕ8s š޽ee£s•·€nþ•¾¬{ezWEæ”Á‚Jºs]%Öa^cýû]÷«‚ãý.Âärê›`ŠP6ƒï‡Ðä¸ï‹=¨lbͬC¥úòÄàmžžª!Ò>A»L›æÈ(ö-„;RŽhI+ ¬­º†KõÎ>ì©áã÷EÆ™t#9VÂc00áL¦íÌ/5îM‹a>x…jâ™p>Éf͇þŸ«ÛGVïÁ¹ø³jèÅ«ÑÏ©¡Bæâš{•þhòJ.?Š#ÏmgŒƒ¬©Ùú0Úýãßãh²£ ñèûÚÓ´âígö*êê×±¨E¨ž z»°®m—`O(צJí#£ o{\ÙácÓ oB(5ñ˜y« 3¬ΆæQn¾tãDm#:Q›\™eÏŒ6ª“_¼šRÏ_yÚXv*ÞðÀäüÃBêŒq ~ø\¥i^ÓQNÞ"Ëåòú>ÿâì£ 4¸ùŽÃŽÞ\ 2WBÞ®Ku°I-…@O[ƒvŒ]²ŠÐ8ÞÎÈpºSèå]Qþhc(ÝúÆ&5ŽîÉNÃÊ[C·\ >Ñô!CŒ¶"©Âbð¤U.ºÒí©ù»rÛ«>LKÊÑüÈü3#«Äã›Èô¤êØÄ/â^ê%yËΉ®H…o›Tƒæ5 lÄÏr&âÏ'÷s‹ß’u",(‹+³š8`ŒRQ(“þâ'ì¶#}ž0ã,Dƒ¢Ž°è2n;ìÚð{úJá‚EÅ–+<ü³‡q5d VÐ&Xbü\¥f_ÿ€ªô²žä¢þ_”ªãxó¡g#€6>ÐüK„)å»:Ø15áòìÍ ËP†ƒ2@¥mËòfѹAÄñ¿Œ†)èó{‚‰×èª#±äúù­,qËê¤öCtjA·.v}Ð0øj÷äPúà[oYµÚg:BŠsÇZñ­ÓoáÐs…"³[>¾È4ÚyBãð—_k™ wˆT6u Ì(j›2?ª‘þbdÈa÷ZÑÝ¿^JÃ4sàÏyíìøBR„vVÎÛ¨¥âË%`U×ÈT¢k]ecaärQ²Ûðí8 ƒhõëX5B–fì*udØÑ¡*ô *?,’ÀÌG~å„7:l¡Î]Ä^—Þs3½ð 1i#ÎF._~¥ Ì„ó~ØßÝ'RûYœN4=+ïØácL0= ¬5êÔyáܘ0;[_¯”è×L… àŽª¦2²#‘÷`|>+]¼âQð/ÐE$ÝaÌZxŸèØIÛo&þ$q"w†1—=»ñiÐrТüu»§š†nR¥¬«rÆysíÏ®P_Tô_ÁmºËí çÜú¤Éê®+ ¿NqÛé¥ClâN¿ÑQ̲­‰çlG’9}Wy½@£øök%În. ±Œ{T¯XoD€QùSýQ e_öøéG0—¦ÒRÒÚ‡}®Ö¢í€<½mN±f-×a‹%÷@Å'ƒ e5V•†°¦Þ>(Í"' ÿK¥È¯Ët< @×G±!'ömú ˜íUÇt eéz IÇmÄ~)ÎïÝ6ÚòBU9º3X¾ILy Çt+áIÂCç6dW‰j¶;¨Å9Xsä?è“äŸë‡=,D)¬÷{4„ HXè¯î +Cœµ{‡ëp¯×…²Ëâr‡–D_^;ö¿®ÿxÞ·ü$a`rµ}m&Tpîg- mL ÚŽ“òº {Ÿž—3„Y_R‹¥éI.W'Ttñ•N½%lA4ï®ø!ŠÆÖ¹ÅÐéN N6·'WYûãí´ê-Í.õ( k65¢žTtò×Á'Xl<ø^é'ü´ä ê³±¨ Úþç¿î0ÒÜ ˆc-ó©+ 5‹´EÍ+“ ˵iºK¸X;óˆD°ìJ(A4Ûé¹´ÞÞþ¼@G._ɉtý†¶ ³¦@DC¢Oý|ƒÒë—õvÑ~¹K†]šXzÄ8v€WkD|eë’¢ñQEÖ¦<©YÇCr"ö½Þsg›öÎ*ëáI÷ÐQ½ëàJ@¶ ä&úzÒ”)áhCxëÄÝüAbüùâ¢oxîÖ†BÎÌ;4ÉjáÞ÷öå?{\AÛ‚»óí Íú1Жsåg­ïÛ÷ †•E-šØ%/LEØÏö‰Püuè^€ü°®Öœ°fëíÜ<³ñî*_{#Èïe„-;®å~AДúÅ_‰e¹–XZ櫸·D–†ù"r+ïkü[áõ´O–r;¹ú¦a\õúÇ7–l¡4‘*NòœV–I ²æ¤Ã9ËP^³ár»&ßÏ#‹h™u÷˦ìå†9K²´!pð´á§œÛa›+WVüѼÀ˜Ÿ½æu®Ùïzuª‰ßîR²ÜнaˆŸNƇá÷Môµ§ò<¤·Öx9˱ yjþsߊûyrdíúm,âMràîVâ+'"W°&)róöÃhB 'yóñL°Ý…zöTì5•£EÞOYÂifo™)›t´Øü7ã¾÷Gh·e;ëõ&òz? (1;‚ì1è½­k§ðÂmù ðe0§iRRÌììeª¼q›é)Çèë ¦åßhÔð2½?L휹´”,PÁ”òƒé sd†t³Y¼BK{L•î¤3ŸSÁrÝx²Óõ¦L¶¢ƒõ'’4~¥U\´2[#Àî›0NéÙŸ(ëõæC«t ¦€dáàH¢Ø`4½§QlW6G‚jy¼_÷(Ð}v…„7†kÖlïàÙšç ´Øâ)¹h–A^§-æW-…CªŸ¿ä`z—E”¦†VåÃ×–úB÷µdNüˬ;Ñúg/æ6à5ß©+‘x[îrÆ“.Zâ šùN Aj“¬fŒl8‡üæÀo¾6s#GÀèñ¬½aÛÿ‰ûåHÌq׈˜éú͘µ[bôâ嚟lU_)í :æXaueÕ%o [³w_^õ¥PÖ!&õR¼`žÐ S1B’"ü©É‰œ‚¢¬IÀ$ ´Îð¹`Ù£kÝì-¹¹-Òo‡\ÌI}"¤Ž7W]BÏ…¼>ãµhí!±ÄË_#¬x™L´O}V§~~×e›·®ÁKYÜF0ßn¹ô°õ¡)Ñí½ü©÷évÍÖŠÑ–1«QŒrJžK#kÓš$æÑ†ƒ#ÛŠ­š£T ˳X[²RïéÇÚ—8•‚ÝY×êáÄÅùoUðb„ÉM¿|¦.í!V—R&Œ9Úaœ•⸴ëôÕìï~-±E´ÌÛ k×G¬9¶R?e Ý8Íz/D@ Cµ¨Žën®ûv/‹«‹ŠáòAÝŒ¡È]Éïâ&Œõa†³¤ŽS1ÉŠXõøcY7…žç…Üì¯@ʱãÓ}1ùÄÏ¢½T¹¸ãÆ4á^œÏ6î03/Œ|}Câ?Ÿ/Sq„‘`õ1Ö¡×v˜ƒè­­°Z“,7vU ¦Êöõr}"N/óbÕßFä]²)Æ ‡0ô'öâõ·;×ÀùȬ@{éÜñ$Ÿð+öí\Ìvuä7„Ï À-ðıçßã‰ã$»õÙ¬‚MèD ËÝ(tµ;é Ô»d6Ž©æ¸¾ô–­¥åż•õ•˜ØŠ4Ð9àY æ0ŽŒ¨Èa¥3:§na&_\.šË<#Fäš±ÂÁ¹¼ÉÉá,òÒtŠï ”û#  ÐÏ5gîÙ¹•êí¢?y"YŠK§á_úõ‰¼\OƬè°ï”÷+e#¼|³] „ñœ‘M<Ô~Ÿ"ßÈÅE²Ê‚(š„@&¶ê ‡ŸtÁïTt·Ú˜»¯Ä÷ºqxkÛ÷ܼ†"æ‘Üöo[;™.2åÊ7“Þ0ÔªPspº}g./Hƒ_LO ¤¡$Ø;f©ynоÁôÀî¸~ßx¼ìIÂMÛ2(ì74ÖÊ¿ü~–Wñ¦ª`Ê~ † —Pàp>u’åñ1ºÒ’­Û8Ë«4$x­ïe¸dCÆ*œB•¸¡“ñ'ϰE”€Lb€!«ùvF»6­Â¦”þ7ÓÙe+ìöZjôΠu–`÷äo>ç–›ÄU!Uäûeð—ã…Õê1U9õôgã¤Öh§ýdz† 6#QüIsí;g?W§ F¹?™-\'!âø(TÜ`ê¹£ewwÞÐ6sšyñ_Ö—dÒü ;wýûÒi/nû­ “ê.š åü>½9߆à·˜ïæØÛk•<ç¿C9‡§È‹Ö$‘ñLeÞ!Ý/­«Ó-k¡˜yŽ’,…Æ~öÍfû'œg'ÀþN~’E!]ü©Aɯ #'`z!»ªR³ó*$út9Ô50'y~Î×8Øahç³6lÀ[?í»¦/D/Üm•—r<5½ï2%Î¥®'´a|öˆ+–Ú/Fð| 4^‰ôéÅéPŠŽB_Þµ¤³íZGšœuð;ÅgTNÍuî6‘Éu "(LãM<ÇÖyK çbáÅ"¢~·Â7`äoQ–*þ¥p¡ <¤ ¡L ®]÷ÏÅÞzª¦Rƒ)Õ`$|™KÚÞ…œìäO0\!¯õ~Y8V'Ž(µAÓF•ht{Á£é.K/Vƒ(XF²pTËojvõêXiøskyj¨;n8K–¶h×Êéz=Í‘0Ò~Àhæ5ÃÇ_óuºè¯TÉÕ…jÄXA£m ÷ÀŸ³-LµHƒ“p¡R›Æ¯ýÛ”ëùŽmô˜*¢‹C^ëîÞ uÚ-0±à˜ã ]º±yaH÷Çæ{ψh=ã.5«ƒ}­öÌûehm4êX§ðÒíƒé®Z³H6KòÅíº™Z÷„÷/«¦ UŠ"~êjú*ËΫÉlYŽÎ]D/T dYý\­ ø+ª½¹P-ËP y‘U ªi¿ ŒïðL³Ã‰ðð’yi±G¥Ò•ÏæñÉQÒÖ¨E8eU¦‡BŒ ÇDò=-½ÎZ÷‰B©oú €i݃½ä¯åV[¹ÜŽˆdˆ=L1xbÕ¦YÍø¯æ¾^Ù¼ëi§®Õ TZUÕm,$ùÉíÂnƒöµtª’£æi”©e…\¤m/úüÓz³ObyówÍ7uMx­cû<9.L¬”ܘÅçqa™³çÌûÙ¦¿¢]±Ócù2ž"â2«ZÚg©™Û6¤„mµ?¥BÉÿâÜþfHŒ:àyc¿¤‘Ž®ðGÆdzU¯ãá¶vÚc¦/åPº Ú¢¯82­lbg7Uï'˜9ªOÂ6r¶þlê šËpW^{€ðÑ×gàE=QÚ&ºCÞ€ÁPk–’dxÐ,ÀÅo_=ä—ÓDÒ»ë—iÐKJÙð—õ–-g¯­CUz¢©ÚV5æ<<Ýt§ äúO2 ‘ÍR›@Óýë…†5ÃÖØU æ˜$ÞlÑŽ‰†¹pžÌwcV™é&}÷£g§åã=?õ7yàt}ñÝ6NýC¸å¾WëÂá©¥oå*Qþ8UˆUåÓæÑDVÛ €[X;¥,~£E[ÊÛ}í Iõ7›iìfcRÜéE(Z Ítäå€ïQy k­ Ø+w‹:«§C&ó¦Bò=t‘âÏ2”MŸºw¢¤¨®¾;|íEvðÜç`BèÝqå’ž~…|ê÷™K©¼‡N‡Kó>§±‘îÈ«´ÕǺÀ2»ôö£Ïk0±Ýï«zÇ !Î)ñYÍc"ÊO%‹š#%q"Š•µ ïâÊø•º/5­–ô v;xÍ(p‹­–úðÃýh¶žjÎ ~ÜH(kêŒÁR+«ý:Hä‡,%Ö²Ÿ¡‘6\×L˜9ùõ¶ŸªÊV9ì~æŒt:׃¹ yGûï|ÏbÑ‹‚¼âQÀœpõ»”k(úé0Ttôv{¿®É\’Äf×ݦXö0W>ü{åetÞçï$Z/¿›²ÿ‡b2ÔBX‘e;z…•»k MþE´¨æwùOXsñ}\Å®ú¹^Úæ¡é £"¹MßHc}óQ´¥ˆÞS ÓyÁ…]×a¦Ç§ˆw´aRëA|&=“õ |‘ž4~#ƒêI<_H&2)R«-~-óE7Á1‰&|–Óof­O.ꊀbŒ:‘“ oÛó`ÁüUᦥ—D¿GÕÐÖÃ'F‚ÒRr8U%BÒüAƯlX«0ÝEŠýÖ Nãé™X…†É¬¬Î^¤Z?©&)¥E7"àAœÜ};Gì!CÈ3e»iž'í*ò·Wð Ýo|îB˜(%—wPí@H£ËRBx œƒNŽP\E1Dûß3ðhÆf·¨·AAgiÊp_ÃßÕgÚ ¬E#¿'×Lº[YH>>%â@À•Ü L’‘òÉ)Më ¬€8•‰YQ£0u1bãÆXÛ¥ÄDɯÏþX÷”ŸÏgÏ“>ËѽF} ˜LÅy““aA~Wß¼šo®´%CõÐåöÚ—äY§® Š|ÀªØ°4bî· ê2ª¤~È«ûL Åî§þ’¤O‰c~¼![/Hen˜ŽÉʰ…I*ú%Lj>m-S‹â­cßîCø ÍsNÓ½aÖiJðÙ&ç+$«ïJX ÔäçɰüšÿÑ™Y=†ý¨^€`õ{°p®6Ëš¬êàØ`L÷Âé^^ÜVž{þ¸Xó]Ä[ß1VkÛŒgR3öjv˜×„¦Ïô”¨\²¹«7}ØdµÀ$øGÈ+-_ÝÚŠâÕlïA‡ªÙ%Nø/苸?*i‹‚IW{<&i³péÍ(VžÔRšV8œK˜±J¶Cå%øªÝè„Á/Ô„JE6Ñ8Ìê5åÞ³=Ó08“¤oЪXŽs—+î¬åÁ£Ï<[' xÌŸ»r›œ ÆxŸy|â#ÀWÖX)•ædP²¶ÉìEQ]’˜¡9@³ñW=€üÓÏTå¬GÕ?ß/Ùþ K7(xðé³ t¨’ÑÍòÅ'C]† ›Ä°]3³àÕ> ‡†1ú³Í˜¨ ›šæÍPDcYr±°?OäòuÛÊgJ3(ÁA°Èy©Ærîqõ†‘ÝJ®z=.#oªbž˜UÞ+þË-åé~CÝçÂ~=†XW´š`IÚÔ4e"Ûð»KñhçVñõ¢©Hu¡mܶlŒˆœÞû(Þ×l Qi¶h² ªŽœr¦ß²b%“x7|D²ø³\Žs4!Öe/“ce. ÞLÙNÓSìÊ$® I…½ý–ŠÑlµ‡c»Ç Õ]Š8cÊ¿FäuŸ1Ö3¶—ÿ6ÐË–üì‰)9¸s•æí>/`gk;ÜHŸn‚°2éj$ßfÓF’UÄz€\Ò/Íß‹·(¥Õ:Lne9@¬ûÌID×㜮€¾¼š¿’Ž»v"ÇQH#W¾5ËO)]¼†÷¯¢Ñ»u`I©XÃi-}Ùß­€}~˜¤|õáLàp½:JK;waéÖjËNô8†ñ{8uú_ª8Ê‘~y5ùV£kBÕ¨’ðªU#ð-W"]€x„3FÐÉ’IÓ¶fD‰wŇçs„caâ­ÕŒ½­ÜCy•}K)¢˜ÊÏ—ÙÒE |¥ˆ«Éœ–AzPWó“éØ'^®—¯õ_.`ϰ&¹ímÔOuv[ùŽ¡…’úk*Ò ^^h¯HWÜ5̺ÔÑ`¹Õ½q­s^N`ãô>#âÇy%eÀ¢£ ÁØý¶!ÍeYçwÑÞš¾U¶Ùz?˲¥mYgÖxûä_䫤¡bd'úZe¥ÜöB"rí¶@N7±týîrád/“AYKàÝÓýi2w,ÎDŒð¨QÕ Us Þþ[Ã^^ÞLr,~’­" 4ÏúQû¤+iìj´“ç㟅¤¾õè$ª:0¢uí!xäsUÙ )‰X|’ô4ÜÑ`|#Žî˜N ‰ÖÆ^{:æïè¶äf.wʾ.A³–”JVÇßõ:|jH½S)[Ç|x® “!+Pž«k”!¤É¨…Ï/Ðìz»H9“¥‹<´ÞêE_­ .ßGm`Ìp´P_H}qµ[œÅÚV¯.Tm „Q¤Ðv…1ܦX×Lþ¨}ÒÎqÜÂ';PÛôAÒ´…Ÿj… eYæçrµ“šÝI‘W•šTƒ|‘)£i*N-8â u‚¤ìbŸaŒ¿É;¿«Q7ñ‰÷§ ÚÍþ9 Õ%KK=,î°âõ°ÖV/ÝóòÅ8dd’p´gÉ ³ù\‹d&bA´þ’,ˆ¾Øü˜˜ÃcxC[¯K‰¯C”w¨½ˆBÕ_¾Ç\"ú4lõàaBCAYMÄŽ”Ñ0\±>”Òº˜@DŒå3=?Ÿ©ì¶˜¿Ê:m«Óâk˜ò̹î¦Di¿aÞ™Š€ÈøîrÇuôô¾Íw>Örèaþ&Võ}­\Ÿ3|Y}6Yí*e‹Õ^ë¬ò®Øí:~‹Íúq¡¹Ò^æáÓýïlUJÓa‚¯¨»K94í}¹f8d¸Z>[Ÿ'·O: „Þì:=ùLƒgC ´!”Ê?nã`V µÞ5‰ (W„žÑ>å a§Å'??â.&šÁU ´™Ü\I»ß;è½ @&$ȱØiP½ÆY@ÂsIå~1iÈàŽehgG{ð¡¯Î§ðUdÁk0}3-BœŠ¡õ ¬-zàhÀõLÔ¦XŠDjóªrã¡È ì«bÁÁN"š“ J¯‹íl—üÛ:9*^+ÙqN£±’°ó`4%ÌNˆ-ÝÑâ*~ú¦:>¶ˆØ—Ì5ãF›”\±ê©»Þk¨(Çâò.aWDÊËÝ_¶‘ƒµ¿>!¨ŽÕ éw0!}ÙÁŽ« éjÛÊO¥J'Ö–_ad(ì£ög¯ ËÎÕÿ¤»·vv#Vóöy["StGßNÇR•„Ëê “vO’Æ!~Ù[g±Íyƒ˜ÑH¹tÏT60;xÍá|(¥šÌ¸Ñã ¬:'\Ä[¾!ÈïbŽqýje˜¥Ž/*·¼C“"%›¡ fD%î®"¤ƒ–Nöèb¯·p\™¾3£zE"a„wWÜ ß–jg|HÁ'Rˆã-ÒXCñ˜Q‹•ÐDÉÒ>IBBV”÷#^ç&Øö ®¬”“û›KwL«Ž¡Õ•«Ý¤Ýt“CIÑ.áL±’‚O Æ’ï®´¾I±„ÏgÞÔÇ©Tixº>\ï¥9.z|»ÂtD˜‹ E5kz¥×ßPÐòt´a}S½7¶Ï“iÍ^(0:AXð­ 7’ùÅ7?v¡¼•ZFÊ™¤&J´zÈ»t·FýÊήV®çùw©6Êììχzô`ÌHGœÄ€3¾‰>žB‹“|‚#ˆs„Vuf±9Ü×X®À˜­•@>@J°Ü„—’†ÄÂN]^Iÿubë”›´âû*$*®­½+_²s}TcnJÚTZ›¥-•±³¶¢£_¯„nýÛÓn*шÞèæþ/µ% ’ÕîÜ ‹Žã)zÄk:‰š_¥°2GEѱÁyÝ„îr8¯CÂ|Ì6;’˜/‹1óóÒöMê7üißJ›ÿôEÖžEi'“;~-›ðfù§uåx µ×%õE·~—Õ –Ÿ«þÜõ,{Ü”Xa.ÇO}Ü»}_8T k&rªJ*óàOœÝ:-¬߆A‹C“öô6˜½‰9ã^_uÂÍøØTIaÉl¢‚ތڱ™qºMÜ6_:*¤Ái h}J.4cÕŠ±oŒËdjdÆ.žzùÍØËJ#ÐÍÒ¥r@+éô—<ƒ#syeÇ¥_ø,fÏëø æ†mMëæ—ç½AÿrÂC¸IzuNCÃ)†€€v ´‘4¦_ƒx0÷ËÝ÷È׃ìý.?òÕ(O.HN"3>É>_æ&ÏÅy·Š‹—k±€í:ŠÅÅ&,ü+É=Ô—âÚ™[RUÇ݇rãñ?±HÞ`üpTvîÿZØö endstream endobj 58 0 obj << /Length1 2436 /Length2 20864 /Length3 0 /Length 22273 /Filter /FlateDecode >> stream xÚŒøP›mÞ ãZ ¸Cp+îîîîN€àî.…âNqw·âŠ(î-Nân'ÝgwÛ}¿oæœÉLrÿþ®×•„ŠLEQÔÂÑ (åèàÆÈÊÄÂWTce°°°3±°°!QQi€Üì€ÿ&#Qi]\AŽ| ˆ»MÝÀ4 S7°œ¢£@ÎÝÀÊ`åâcåæca°±°ðþGÐÑ… aê²(2䀮HTâŽNÞ. +k7°›ÿ<hÍ鬼¼Üïþ¥µº€ÌMЦnÖ@{°GsS;€º£9èæý?&h¬ÝÜœø˜™===™Lí]™]¬„èÞ³Å_ìø²=›oêjý§=¿%œÝÁ+ó ûo!/s;Sû¿ì€Keù‚u,ÿ‚¿!è/7`ür²ÿ†…Á¦XýÁúâwÚÚÛÉèð—˜ö—}p;lþ‚àšÛþÁu²û ‚‹øW&àCù/Ë¿Sqüã, ¾dÿbƒSqúÃë:/:‡ÿië¿©ÿÛW =èÇ€ó· Ð㯂q‚¸‚OþÿúÇâj÷wÛYYÁ¡ü |¤2»Y»ÿ¤>*˜Ý<ÿRÛpÿ ‚«ìñGæù×€µ½þ‚`óÞ¢«ú]þ±ý?ç¹» ¸Žnÿº)Àsúü¯{ôš#-/8šó‡Ù4†uÝÕ‹z2îNÎRíj¤cô]vév@…O¥«ËÙp¹Mùòvu[’öZä;é³ïQ{3|DG²jç£ß“q¢Úôn'ÒÒÎÐdñ‘hÓ 1"£†Èžß³³ŸV°-t;äg9ª|gwT•BÌ;Ïi¯¦ÁÊ•±÷ »ª{u\òÈO•3Œ±š1ÁesTf9óxäpnŒÄôg^hs×7³y“¯¤r‰ HþDZì%¾z›lq÷ó>kÕl®½ø”øzxÄÐ×cÓÔ¾bir¸‹¾å%«CK^í%¤y(ïÒWß2°eÕÔ¢Zû<Æ–{XwòSa4„C˜;É ­XÆ.äÊØuÆÑ˜n ì¶@âƒK±µ®­>K›t¢ï¼3M¯€9]룠¶>ßÇf絑Ñ!ÆÛ÷éw#­£Û}þÚeƒÂDVž¬ô’ÖúQ«+|YÄé°•nãP½Ú@Y¬·ï¾{¤ò<Áš† ø†œ‹=°n³°èñÀöCŽýâÌ÷;ó¯i9SêÙœC_´óÄ0 } åRz5ûœ~vVŒoÍ.i’ýž¶ K²ïCÁz<*#°e.´’ùl³‡I¢Â›T!i‹séü+e ‡¬tÿîbíH”Aˆ·Í=oÓ/­ÍÎÎ"U²ãhÆñƒ–àa±ÐðÞ‹ƒ­›ª*1#ìCG­û³ÏH±‘¸ÜazÔ‚òz]•3¤E£¡Jo·çûÍ(N4Ä<Èì[ãÑ´5¶Ç{KŸ<ßUÊJPÑ”,+Ê Dø-.I”ôã׉à|¸n bkìV¾£…x“ýv-Ž*{ošÂ¸{¤Uëü…@ŒP]ã®Âg¢qf|›ìÛô6¥¯]AÇ@€Ÿ3&Ú.‘šlq;(órC6&êÅ2ªm]ÓS´ù;QCÖÜpãxx FŒ¾›KhâXP\A“½®P€õºœk³å“#ò@ß&XëjиŽu-^%hЫ I¹!!©ùqjï‰+5oÆKº@Õákafþ¨Žj!%—lqI ·A?7ýj»$Ž,Æ2\'Áü=b[`žk¸1>í\ S…]²€,Œ‘¯}óeJU­Ž vy Q\nBέ‚ºß—.«‰"Xkªª÷h~Afž0ù2š¹bUjÑÊ“\Š3%Þó\=’Ò¯î5ØQU`b"yÚ$à“@ +K…BÆ«@ç¡ýisëýRöúˆõÙz~ØFÍaBKK}}–ð¾‰ð%wÚ²ñX>¬~jÏ%—Šë<ñ¥ Ûä Äw4~®Œö6ywÖCöÇLMqXP&N³Œ[Mãô¬ÿ7ÉÂK€9 •Á†óiAGLJ2‡™~–ñ­lGaýiSÝöОy4ËqÕ…¢Ü½bž.KõX¼‡/×^Ê‚Á]?Ê%Ïy$“œ_ð:%GcFWžiUïç]ß;\éQy_ÓHq=1éÜn$bïZÔ(yš÷ýlÊ,³Zéµ\²*””v‡ÕmŒÉ¸›ú½xGsÀñ‘軯øòc‹ÔlTV)fu>·rcøïÜ«Ùð|ÖÝ“2ò#æ¸è3;éKQ¿ TM¿ò~ž(ß(’OÁ1§Ãû6yûÃàÌ*Ï­†Z¸b;nràêÑNãȳ;y¯‹2ßBÉ@¼-«xâ»òͤH‰6›é3™‚ÅLí6¡ãQÀ«s̰˜ HÃÕà KRŸn£Ï=åÉýÔí× ªšÏÓ#Øž?Ö°:VöojŸ…e­¿¡¹ˆÙÿ’‚ÃåkaEEXJ³@¨öæº éý¾ –B)ÃSpàT¿ð¤®`<š­¯Ë•ö¹+çý‰'8L“´C;öÜ%ûd­ôм໣Œ@ýÈ ­àa>o j)í¥&W<þi? Òª=ÿ¾¯¸Æ~ÕŒŠeù#,ýƒIæü@AòF2³åuŽ]7•Š$[‰–(º˜óž@%Î/NÌJñ\Ñ¢Û;þZßo½È´K:°RpIäÑ!Âx|Õ–“Ðû4piwŒµÏ ÌD_åý‘1fÓµÑ7^ˆ»Q¼w^¹‚ªòŒºoMÝrID”Y ˜s ¤ÎIS>±{“Pdl›£)*.NIê=¯éCÛW¿vøLÆHÂ!bØUÞW‘‘ªÂ,i‰Åk‹ùïkÑ%À¬ª¯€ðßœ¯—ªaà´á”twJÅ~…®ìVÉa:èV€™òºƒ)Å3ÖŸÐÑŠDÇýzy¥š_‡¢¡òzíš}—äôò Ô%Ð:ù¡-ÙçµJÝ[È,xÁÇÎd!‚–gÜÍr ¾ˆ#ñ|ð•)™Ïc‡DòfºÐÏ÷jßZscu¯©Ìyã$:ɤÏQ¤¾Äfæ™Â`p«(þ¥$À³&’k\C$Ï # Ø©âgàv2(öôÂÇÐÝ’F2Z㔬_=HÆÆ Ê3^mO“n¹ýqvDK"úÆ®sú«É*;Ó¥;ïíeò‡K!þ³w>#ÍÓTHŽuù^ÐÃ,]CІñ1´Y¥¼&ͤo¬^ìc[¼´¼¯[¾£OçRøॗ9y\ô¡ä¿O̸Y,grãý@‰íЦ+^#Îú0àVi!Bq[Î1e«1-_™?Ë_Ôûs.Ù.Xêy)Fé2ª’óG‰ÅKõ—ÇȤš·Oô]QC¥©Ûƒ˜a˜•Ю½¥¼-Òñ(â’&áZ‡…Ö›Ý#TتÉR%Ž*Ý[Óè¡”• ÝäÓìì]·*¼‹¼Çz±®¼4&Û5$(Ç.ßµïù”RJއ]g¯¿¦´’ŠÀY†¾É6"Ö^ §fãðĨ°óAÞ£œcÓ>Ålô\ :Væ5{-Ú%W.ÍÃ×î§—Â0\c.5ñ®ŒQm@ŒI1PMýB(E_¹ ÒÀ9=0RÜX)š„¿W§ÒhRê环C`PËX8Öw)ÉzÖ½Ä<Ñuúw5,½)‰¾]ÒPy¸Bš‚Ô”,<¾|‰9›•íG“‰¡il‚ÔS™@“]Þt£*ÚyMËîÕ/§ã îyÊ·žpÈyÓ€¨K3·búå“Fÿ9ŽKàÜF³ƒ±C*´œB2mšRã~þ·©L¤š%b_y©–•3#Û,³õ/?pNLêŠ VL‹5 Û}Ÿ"1<¶N6V;ö 9’¯X?RbºØ‹¾þrMÅú2tyá ·MŸÑ)Xq;êz}昅e&â›è½€Œ@&.Ní‹kŽÀ¤¨÷éø”d×yÓÌ)ÜÊUå£%ã´ÁMÝÄ\µ‚€ÓgŒ64;Ì7~íYÍ<L#iŸ,'¾úx—{„’þ0nó©È{Å&ì1×…õó²ëz1¢_Ò´ˆŒ¾Ø)m%ÏÊs™  áŽ{W QWÄ[Ý0à<<*£æÞÎQlá°:‘4-ǃŒåþ 5J*%äÍÚÆÃ¯ É SýÈË;/Ž ™ð|l¿ŸðÚOÍü+Ò$ Xou1ÉÙÅ®8—/Ÿöî~ÈÏVéÃݱ¦Mdœ˜s*ÝñV: ±qËuŠ=_¶¸2CÿŠúdE{Uá…­úÕMH¿v‹gõ­’Ö%¶! ¥Ø}VuJ|OPU8K×ûþSÎ}KÉs,?¼:>Í£ök¾7¹‘ŸX^UM¡’´,Ò£üØ+ªÎPq_Sl5hÚÏrßòç $8:,í¸»Ù)ˆ³&ë 8Ç Ò‘œœgJ‚…w¢§|XÅÕz‰ŠTXûH€#PêlãÈìUriŠßGÃRíhÏg[(H:€#ª»ÌöNJœòÜÚÏÀ(?ºþ•¥uÓ(™´šM½öÙ:÷XMÿσûýWåpXš€i²oiº©Ç?!îÉæ eætwËgóðå©ÌK$ëÚŽI³Å–Çd ƒÔáXo—³èÖ¨cúªï¬+uwCÔ]^¼M9Š*ºÒóSü$¼0±D†0Ó¸º8Ñá"œAIkÓ´HwÙ‰BÈ9ý¡­#26…o·]±]°C¼VÓö˜¾}ŸZ¨«+˜´M¯ýn7ÀOVG!°ÿ=³þ˜@g9½#å“eÒúµU¸p[ÖBD]†:Mà¶+é ÐÂ38G*„È`[«¡’}ðÒ´F”ÌÛ»k›ï|–+ל¬Èë 6-ÄlRñeáÃfKö´¬£o¥_·É¥+ÒÐköi,]§áçHu´5ŸýŸnZZÛtZbƒæ©lÆ¿’ÉôœF+˼"ŒŒÇ‘ߊæ?ˆhœ¸LÜ’kó(@m 6\¥@¢\†¯D[¡·æÌSÇ{&î<ùVßw.’\ƒ—ƒ˜iZÝĈ•!hõþÐÎQ²Y‘ö"΢µŒbÅdqÊA××fÅ„ÝÏʃ¸Ö¤ƒFÖ.ÿ9p³l·ù©àzVåËè±oJ!œdÝÇSÁ 'Ó#äà%ë#e[h“œÖ_YÛNwojxܽW)].ÒÖ.SÝ‘,ýÍû8û¼Èž× çsïzQ5ßÕ@ö¸åÓíød.ò{”ÃVhkÊÁ‡ÄRZ=/ÎSp8BÕÒ|‘?Î|Bsódô ¨â5O<–z!HP¦|¶ð6¿‰ÅzS)ilµ}bÄ«Õ9Úî²Ç’KºÃaœe»%m/ʈ¬&[N†>Û¶©1e]?{Ð}µg Wßþ"Ñ9¶lçýDŠÇŒSøF÷zA¼©fÐþ°ÅèS´;§±tü]Ë2ké#šsÌòËfìØÖÒ&LJÖ‡HãG !'8†=ËJ+^kJâ‡òAHC‘SÅ{“%£™viù~Î*¸bϯG0;².Χ¢åì–¢!ºè ÷ê /k;ýÝÓ§Ç3º.o‰%®oøâc¿[À«•@¯’EÝÐß1V&¼Ã¢Ûê0­,é°ÞšäÕŸ+sEPßÇã•OÂCЯ0Õ¡Üñ´Iþ¤ª¿åMèü5´ÈF–6íµËæ-§BY¥l¯y‹¿Ÿ5I/e¢Ä¦æä—ÏNÝX†îçj=¾‰ÓýÉâÁlÖÏ-vËÀ^¯&îw"P~0ÕÇ¤Ø «‘Í™CFß¹1çzO?š˜ž`”£Dé×MD½5™é“Ì-úâIÐ{lJBø¦Àgt¡`†‹áo3Z¿?"òXµë?è…¶wªý >c×!Â5î#‹=ª/j°^…Ïetê(F´PÛ>Ï ©£Â4õpÑ#ô©¼‰À¤¹œê‹IZ᯵Ï#®•Âß:‡Œ ž>Ð'a/<«Žþ(W2³ gXéêYËu$GŒ].ÓDìò™¾´úi5nº—àÞ,•?!Щóà½ÖVY=õ@ß5luóÊ!¬Ãfø%£\&s¼0kñ"Ó¨oÊYèØ˜ýQóä2àÍmIRÒ·Ð1ÝPñe†ŒCB.TGy(7ù¡¯Ä{5f6·gÅÁ¯W÷®»¥A¶¼²:WI–EÙÓÐ[¯“Ï ˜Ë¥:’Çeio‡™~þ ëê'§—[QçFíý®KR‡@t{j³‚BL=³,z ÷†¡j/òs{'Æw Áë4¥‚]J.Fu ]ô+xµ é¯e¿ö…³ &inŽoûï¥p-´¤Þ1ìã¸bWK¿o1·U87÷ÃŒÜ@ZÇ ¢Ußßë‘Tû„IT¸‰¼a¿bdfå(.‰ÙÐ h[›ì(vE ½.޼ünžhIÈÍû¥wtè ùO—‰»D ãi ‹ [â=C},JJQñç‡"Ÿí’¿™¥˜HÔÈoE•º}0Ø'x¾5šK†&&€FæÏ@ìõû©0A‘qŸãzU|žúH“£Ê>lÆïEÑGÊ~Iª>ÆWic†~²BÕæ&—Ï,4Z üú+›ØEk=ø ¾—a};\—ÝÖi^_‹a]Î=4§cÏ[ýý‚))þ·ï”¨Ëi0AQ Jª}²2sãWS¤½Â|'"KR§¸ûJ¯Èðê:}k, ÕtÖ›Ú pÁÝ–íR­Öñ+LKMPtFê°ÓŒ$ážþÏï#õ”Ú$±§Ä!¯ên‘u‘cÕ`->_ænÎØ»f >-¹¼¼èþà4‰Ç• 0Îòê‹ …ÖÌõ‹¦ÑIjù&-‚¦¹ð–KÙ€4#lõFè‰oDËøjDh¾”—^–Åg¶Å€Üw á”]ó»ÏÐùˆûŽT¶%ôØrqøµÊFù(dB$ZÈ“fIöƒ']í™fÔ7lç9¾M Š’öºUìHkxç¯È%Ù]®Ž(Œí"IHòÐÒÜñÛä5~ªSP{•»½ W¿ðÿ¢…ùfų¾ ËÊyû®²/¢ì§ñ gêÈl‚ø‰ úM­KÉOrÙ«M8$5D8æÇßñ‚žîÑüྐྵõذNe‰î»¼Da×öN×¾àÉy|y'Oê‰xÑ>™ªµžôñ)K5¬Yu]@³O‡&äûа9à Õfâ‰R窘:'ïû»Ð¾&ÈØÊ\òþ…ì&«îµ#³ íA¸ÚFzJ@Œ'\õÝvTLòv¸–™KM¿®O‰@É{j\—t~Q YqBæÈ`$F$—Á7» ´±ç¨lyúñ`yFüÅ5íõ]•èñZÑJišÐœÕܦ©™1ÉS¶¥)î>Ü:’ÝíÆF© IÅÄEÃÌa±[y”†çlTªí{[{ÐÖF?H¶ÇŽë‹¾R|ÿêkQêä9F zý$uDàuI×ôuðùæÏ¥nx)¨b¼}‚ü¡EÐ¥ªá¤Z¢¤¦_u«ìXü¬­ÌUÉd1ô£†@„l1ý@Ôˆ w_¹vU¼•ŸÊmul!¼“vO¡M»E¿ ³€rôÆvBFŒÂl–‰4@úY’?®’d8F)}Fª »÷jÇÂñîÞzPbÉR¼U†]Õaki,hçݪµ>Ž85 â§Á<PžßËË%ð°ÙUSš@| Ú•¿Á‘˜5bøZ·VƒŸ¹tV†ÈhμuØÞ:% .啼™øGù¸Wco$Pû¹bÉžX!*1Žš üþiµ‰W[A¼¼|ʹ#H£ÛK†9A’~Òý‡t©ÃGZ¥-d1ƒ#Ö‚ͨĢe¶0ûNPO£01ÈZzŸV‚sè·:ùkü<½ù©ßûT×È'Ü£œvó.ÄcÁŽq{Ù šxÿÚ·X®ËYôÀÍ&5EJê¸ý>ºJú\]oß§èÝ,O œŒajœ¼¶·º¤ßäBX=zó‘ä²R;®S0\÷„;®ˆ7ªBJL;OaÎ%äÿ²‰ –].`×Mu AsiLa†>³O)TŠŒ¯íºUlý&é«JØpã ;[¨?#X¯¯ÕU²”q®mÄ´¤‚‘1¬‡–ÛÓl0×wwÍq$ïÔ¥C½OA$ˆ4á©Y¦r²àAÞePô6ß<÷=Q²„ï=Ô³"IÙsË ¹ˆ‚ns†2…q…µÙ®ôˆå.·çÞÒ%Ï£ó~ÕÞ齩˷úGL*ËïŒ5½þÉçüŸm>êI#ùå-ÅÝu߬êIàæ¹P}Š Ô~I,úáE{&{(ÿ­ìª”c{\»XœHù79þ>2œÛºbN•”3]4wTç— ŸYêv’ŸE{_Vª©"¡”ШrOØ+²ð's»‹âÔs’%ÝZÒ®D\åz)p;ì'°©¿8A{çÔïiž7¯Ñ3Ènï° =§G~•Ffe¡’1K Akä‰ZŸR4õ00Ó_÷È‹šKŒYÚª¯¿˜rË‚s`VÏRI^?Íݸ¬ aÑ€OÎbÍ礨C)©-ÅöóëúÆ3ÖuÞ#±Ç¸FÃs¬ÆPaa­G>o¸äôÆz#=¦Hýå«\?'%>¨-¦iz›Sê0\Ués_Çr…O\6ÌüŒ%BÍ¥úŸäšÅüÝÎ\Ój:¶ìу‡IÎçQ^D$·–¤ÕÝÚ4,D“˜ìùš®:’ÔΣs.<[*'¯×$)´·¶°ÏpPp/;(Ýþ"UÉPLÔ÷ž.–(Âo°ØB>Â;Ã|tg"$ÁU•+%k­î Cý†®šw$ 娮F[su_íA‘Yè¼ÙLhGd_Lfw«ÐöYÐØ–÷éŽ!£Ã.«9¡ ÔÑÓA‰¬Q²:@‰ÐåuÊ ¡óŠ Hý2¢SÈ>ç}y ½j6£Áƒ=¡a”I|úr·ôîl®Ç%o#ߑר¢Ö9L2‡lÉ'¨ÛðMñÌ@ôþ½†û`2Fôy'HgêÙtá­w³ÕÊŽH‹í6¾T£x &RZ&o܃'Yä½í:°_gðß•ÓérD‹ÝÛî†Ë9A ©Fd†.»¹×iUVÀ“ °±Ðü¨AÈ&l¢ˆy›¦Ö€ÿKà+Ø—i"LÃÈóyQªz¬¾M%'«Ó"ÒǼõLô[÷ Íh /‰Åäüì[<{¯Ýð{ªµOŦLc2rØ`ßç.‡¤\âõº!/rÄnèK+«»•¾É.$É¡Z7íTŸ.Ç“ Å¤N¤BÜüÔ¤ŸRrÙš×Ø(ŽÒ…,ßîâœF™$^ÁѬŒ ²©J@ÄÛ?ùmZ À¡pÀ«aU¹rEV\?U<Ç`á'd¶é¸½fKѬØEƒÛ™nNb[j*„®-‹¦9|Y,êÚQ"=gO§Á_ÊBzícW›'—÷ cR «#·š?»[²xÄÑÝù´:°Q±Ä™5Qç òôÛÿ€³¸úŒ¹Õ)NÁvWX?h¨fu÷×ví¹c¢Ý³NQÖo}†yÈ=êLçF›¾ìÞCsïZÉÚÓ0Ml¡{ãˆú\a»º‘'=Î)ô6¦k± fNb5A›é›ÄâÛ¤¢I…p{+îXÇ>lÍeZ&ç®0‹ýŸp¡I‡Ç¸~C»µWwªó´_7•;Z8¤òç€>ó°Ò íw&±‰›Àüº~Å‘ DŒ˜ÅÊ9ÿt¸ñÓ«$[Ú¦ÙRñÆo—ÍáËôÛ¸ùmVšÊÄ™i³êWªË%°°­s¤Rxw10àpa¾Éu *Å4#»C'ÌîSLºÚà£Pưù à´ÊÄ×·r;ƒy|>@~PßCý±Yî¤GX£ª|IËëcå϶~’ÕÖÈAýŽF89sΧà3‚›^(²è”ucÏyíÒß®‹)äiæ|ÿ³öí·õ®íÏjJ†!*jIV;Ñíz“[â_Ý$Ôk ÔuÙÔ…aõdFî‡ ¤æŠº×ã¼jÓ‰ëB†GmZ¼‹â* ÒÅüuÌ™+ÁK¼«!2ŽH/Ç£#é3¼V%mÔ»#[lަŒË÷l?Z0­}Ó3˪&!T,ƒràYÓvqá´ÅôcŸi»…Z<]Èüà*¬ɲìlnúä÷rÀöã[Œ+È\¢Ïu T–õ±r_­•`¦ !4Ÿ¬}{WŸO8æç›x1ƒ÷½œo]~5§$áuöýjfƒXA!±€Œx&rAÿÊc²°ä{À"¥ØùØ©=æ —82š'ã~J>;G¢êT‹€#Q_™+]?rÒ“1|ý ”`O~u‘¦pížÓOL[žóF:Ú?äK B²ÜŠÔiƒ?wïd…‘|ìòˆû:T×.(pÁÜTB(Óp žÓ`BYщ{$‚Ïu1i¨úb±ÞdÑæ[E²S)œw u?ÒW™oµŠ2üo¹Cc‰’Îs4™Vš#õ‘uø*ˆ”XG&É À¾£NÞÎÃHS5шÐõvN®)ÇÃbg0cêñ=ÍžE©ˆ_’›ÃCÅ5‘WЍ—N°H®ó3žnÄLÞª†´¬ d&í¯cNâ:x9T×j2/¨8l»¸(2vA\|`Ä Ø}§ £W!ÕÂ> Ûô‘ls(›´ÆFð•úCŒ‚0#ÉÇê9YŸ·[ 'Dò Ô.@쾑œêƯp!~.L.„aÁ«~¥å&3cŽ.÷vü~¾v4¹—·1·ZŠ‹vn0¶ú‘êßý?D¾@]d"´üÏåWe&^¸©ˆñréÛ¾~/LÆ™CçôÂÜ-*mé;²†€Y¬0j7ÇGMñZƒ±Ã˜Cðz[9¦Ì*Ú¨SßöÝyÏtxOÞ® °ÇôãSÌ·‹G8Œä¤Ø÷{¿>ÊJ-Âê9Pég®Ì™s˜é†I:‹`’ÈDAÈš¶/zgˆÍ(/ˆï£Ç8¶ÉT¦²´íÖ“*qÏ–¢Ú M,A˜ZU¾¤Bƒœ}Í[Ÿá†vŽ[Ä ²vË0vkN± F…Ú4Ö<¼ôÆ´ÕêB÷)?Sÿo¼S]ZœŒ‡ §²:Þ"™í(ÚÔüK?y¿„ϵ·Àë‹Bk9ÆñJCÅ‚X CËF»¸ÅÇ ½ëµäà"“ôò¸ÑS¸ŸŽoEÇSà'Ôp‡§y–7&ä_0¯c‘ôéýÂò?xÁW1ÝJÚ7%¾TÑåVyŽ¡â”…«Bè<À°" V$3gC9-¬Q‡PgùóòG+¯ÉF1nŽ›¿"›§¼ß«uó™*;%@Ñœ~¢nïÞ+ø¹²åùšKþ»ˆkp“©…h‰~ÛzÛWwœåKeùK™”ˇ¯e'‚]TmsµÕ˜™Æóc«ƒ{:矛¶Ã6þ3‡¬.«¯Ê!å‹5¹W#‡ÜòŸ“fÏG»œçL»27¹4ÌŸµëgÝežÞ(ëÞ ‚°DÙ•—¢n¶;O\®7Ÿê¾Ùö€Mý–Q@@bºl5GE–ƒ¶Ôš¥DJrŸ]½ãã9È yçÕkímr·øÎvþ•hu÷Â)pzý=*>=F¡óé!}/a<‹HSüÅÂ{žKÊÀ©‡*²-;¸–X™Å©¼w€Ì®¹ LÁKÌñWy¹ ‡-R1 ágŒY6Õ¹šÛ·Ãø6Œ}%ÌH$~£ÄÖ†“È’4ÕîªDׯÜçÅåÎTÖC¸RÔy wõ´ê)ã¤Ý…Œ»f sŸ‚[ªX¨ñK~Øo-µÐÅO—Ú%ïJüçç·óÊ0Z‰ûn~6{«”L ÈÝ–;AŸT~ÊT®Ü.Á¹Û+J,ðF“žA8ä¸x`¹qù[Bç1ŒÇütÀ/¬G`†íG>É·çQõWÚ|÷hNÑM©Hí?b©ÏÈ&œvøDšŸëô±>*aYðÞ{»VaÎÑû¬6«]ÔvdÓÚìê)’©Œ1£¼ÌÃŽÄ14*\'K¿ÆÔ2¦‡?MÖ` ó ŒoxûÄöÎÓ²zbs®OK+é“Ü Áv>ݱₘó\ÈøCvD¶o^ÌOHMü(ksZ,%4˜è‚Ïê´¦ð)EwÎ-rf¡Ô:æ)ZXßDņ³Ž®öÞ~¼ •\ÜäMJ™ˆ¤Ë‘pP¯Á#\Ž™éÙ™Ðå4±í¼a]ຫvo íN¢ )®jÍß@Ñ2Ñ~Ž/®fqIÝ4þ”Ïç;œJÈmÀ¸N;ÅÝ IVoX±}lžâ[ìæóU©ÍT×$sì/Öu@Bð<Ö™±2"ŒÃÏ~µÚÝÆ³M.*ÕÇ£Ì"ÖU!"s×韜ûò€%?Mª.BºßGÂÞ‚xw0¨½øP†m‰«ÿ£!‡+N“BóÓC»6ây¾æ ìcS¬C zlž©éô+‚3aë«vTSèä4°U …ƒ˜ê¢Ïª…à IŸÕ¨¸?×~¯IêO{¤½*Ýõvåýw…az[*üQ/H¯±A2ÙäFÈzGz$_OúëHÁ´Lêä®Yé—Sb6Ž ]i,?ógLr³l­Þ!"rRvÚI›Äe<éÁü³ÄWÆ¡<úw:åó‚bVØDzMÏ–VŒ $c–,WE–‹‰;Þ@‘¾o1! ²›e3uq‡lHZ?_©òèæÐ4:Ûyý D|Õ5±-“¸ó¿œÈÚ¸µ2æç”dÌŇªÆôn×àШ´{»‹(•wºD°¼;vŸ^à:Íy¯¸9M’4îr„Ìj'׫[Ê/pf§zà­¡¨DõòÅpóÅëÑn!اG`Ÿ—„Ëø¶Î ²©‘j´=פªž†@j…–—ˆõMzG-IS¶4ü²€÷ k”U}§Aœ #Ç…EœVB‰QL‹»]wÄKÞÆš¹êˆŸL˜B×E¸^I¿5š}ë< ð>äXN>Å‘h­¹Èq„%VöB…­¼:Ð|@,hòR¯Ìþ¥”}9”A¡ªxâ^îÕ‰Ã'iar·ŸÜib¶R0få¹€Jn6'âìñã­ 3­ÜHœyY‰ßqdÈË0?v&SC©˜^gòwzçî`04“@Ãéü%µ*YᆹIêw˜ŒÃ•!)Õ›˜7”yz™´d°_µ\x0Ã<ÔˆkOóP·÷uúk‹)_ÈQŸÄ««œ'‘1:tó»²QoÃ}_ÐéÛ¿zmZŠÑÏÂM÷iÊ j2Ú$ã³Ñ/Í©dé“θ›X*¯1±½{e*¦žr¸„xR¥g]aªb:‰÷œêé k@釜#¸škàÄkh( J„…^‚«†ñ¦­¢OÄ(OKùA‚¬£ëPÔP˜%µMV(É”|¬õ8?5F±ìÓ5"å‹}wŒ }ã•ÀVô5º0á(Âúá®ïŒ[™?¯¦ˆýÓÊL¹zY1‚ŽlK·ú¹QBuAòÁFd•k|ÆV¡_gK(:hAè»gaœÃñ—èÉ‘«£§-í½Ì6ᇉ|›5ÙÚО +{;*Xß¡°¥ä¾¨9.£œâǪÌñ0Þu'§1ì#´þ“ƒ_×¼…9|@ØjãïqF6Ù:2—XFÆ‚â¯{ŒS»WΗ6H‡‡õûRûz^Uê£Ìñ‚~žX™óÍêDž@Z¹÷Ê6ùA*2º-»Ðø\p'D÷9PÃd0<6 š5éÊ IcJÃUlz¤˜.¡®JÇ$½¡Vü.¢¼±Æ*œE 4Þ•5¸ÃRû_¿Ž;⢓‡0t‚ÅÖø¤ÑuÚÛm5žš53¹)nÊÖeð,þçšÁ©å ½Õ ÉSEׇ({­3ÔæŸc~õë÷62ú¿âËHãä¹y!ô¿•1ïÄÆQFnn¶!Æ=÷Y S,§`çÇ~¤ÿ¼C1dsªibcV¨—5aЕê:U(ÇôÕd"~ª¦ÃåþSƒÛX«åß›XqŠ;Iò°6e“*‡© MëèŽÎMMkÕ'[QP”ΫŠáO¸g‡òr‡N¹-3.ŸñÉ8tSsà 4j#¼0;‹¶Åóº6V$>)KI râ™–óg^«„´ÁÕ{>˜_ë23§Y¯fnå¤`\o o†ÔJÆ2‰Ýc{è—Ÿ3ý¼‰ ™¹­ß¿$4&-¯b‘ÊB*+wÃÈ ²¶Uñ&+sÝ=d6B$Î lüXa_X1Ç¢9òëCl§EÚJi~±ÜDšpë5„›V±÷]”4ßž8ç¥\"zb:I”â‡SãÔ}Њzsû霺öZ[nû[F’hÉ8XŠtÂÑÒ§Ñ Ÿ¸¤#žÄærHVº”Õ$Ðþ^²‚5ù&žn¿*øŸ1E«W‹øf)/¸B%ÞR&ßB >ñŸ—Dúlä°Ë¡þÀ ~5ïÆÓÜ/FÚwYóf_‹±ýϿމtâk— ù±âßp-cc4µ>a ")Ë¿íŽA/ôq&xC7±ç"¦Î àF´½™ND¤¸&uð¡©.ãá•ÌÆJ)yY°²VÛO¼õ5ƒž@¹W¯^€ñÂT®îz· w;¨= à`:5I $ WÑô&Ÿ€°õ柿õ> RÄ2PúÈ}’ã«ê×S„‘ŠåO’¬s±‚¥þàÛ<1àçÊù‘Ã*HÖÇHhEõaãæéÁGQàh^00¨5³Á6¶÷P¨ÿÖ°xöÝeT{Ý‘Þû’ÅÕÄVYZ˜8f$/o™~¹Þªé¡Å«f² x—JÕÜBß'QÕeçéâxþTÆÞÝÌo­9‡ï_¾«nxë#‡½E½Kñõyþg\×Õé³2Åù|rކµ -&3i­ÑžBlhÊ ¿/â+Ð Æcp»|¹«ÁRj…i âkt4ßÉ‘<»@UÑcÜÒnèÄW|+˜½ëÑi?VJ\FëO¶qþôÙ¯Ó˜’]ŒIð&®–wR;‹M°ÓÃÉVëbK|ÎÉ@¼|ÎÁ'Ň*fC¡ùˆÀ½Ý3XýE‘¥îþî(Á<²ÛÌ‹r«7ê¾u~‹@º+üÄe…ÔNûM”ÛÎÓin¹Ø;¥hñUB¨ê~Ê[}(<{)ä7º;6¦ty4ßÅÉ 8ËPÌ`kFs]ôeO3ñ-ó®c%ô ¶ÒEl~ĵdøtˆÔr8™6|«³hÿk¦Ý³•ÒÎrü;F8%³bþ--„Q›¾ÐÅ5”×+º€€OžOO}ž×¤QLg‚«¦ 0b?!a'´‘‘užNc¶¼ö«¥@[©9× @·fæñæ«!X)Áâú’M½µ¿¿jØ©ÛÇ[¸78'c=þݯý•iG†ÏŠ&á<]ÕŸ\jœÖÊÏ'Ûã]Тð•¹F‘ç±!1¹¸Bãã±ô£á݉Íz~Ýtÿ_hèüiŸV ÑËÏcÂceÌJ)çÑ¥u=Ug±Þà«÷~‰dpôD â÷ ^´Ê“¯ÎeÄñD…¹ …V7z'á+›±0Ó†÷§2õæ ÏHGµÿÅz>¾ ¸\C" ;5]ÇŽ÷Qé’0ÉÚô_/ Ò‹¢ãm—•") }ì¸)1¸ä‡*¤Ù ¾ÉP@‹cÿÔ~a(„xÛV¿!Ïö¢*å~ÂÊïžäZíÇ€ Æ¥£rËD×<‡ž4™cZZܳDNIcÚíÈ;~J3F°2d—ln…]±ˆÏF!‹“hƵ²’,\îlö“JåKMb)ל×AJƒš°ð®úäþÍ»çp¯©$”b‰ÉwðAêÓ‚»–z¤šCڒט¼6‹ 8Œjj³è•d´FÍ’ý­&©…Ï·^òCÊaÓ›{½«¡(¼frf£Ì°äç Ñ G>»²&æÜƒ²îñˆlÉ"¯{i¶ÕoØ nù”W÷¢[[ÉU’ÅÍU‚nÑ|d1ç/¿çÔ=Qy}Yå—ñÑñ¬¹Á‘ש‰=à«Gã4r?ÇNÎÎVxÓ!«6CÒJi áªÊÌfø?…RT È”s ,®Äë̾W&Uw€Éh)¼$`çj6béæ7 Y¢Wú•hKAWN»œQ4d‘ÊÓ[únö=™Ê…®9=þ¥ËÚqp0Õh•ÈiïaAÐãfÇ-™)Rˆ¿ª›Á(±µ—N¸Æ9±áé8Šãë÷Q›×!óæ]mŠê°SAàæüx<Ö‰Q ¢,èë…èÕèeøhbàñ¬êØ'=CZ92-ûÊž.Cª)Ï7 ýÎ¥}Û$}3{Ú -GÊ *͵¸õÈÖÇJú»l¦'?x4EÕ»S/Êá ÐOѲ>öuæÜª•¼8ò@ÔHÔoº+¥ŒCäÕe8Ù¡+^=·“É9ñg´²jtÏÆ².FzÍë­“LMÒpS®èKAö>K–¥5“o¦j »7M’yú¦W†=—Ú¾èYCÀ®;¨Žêqů¼ý)nßrƒõ]OíüU?ÅÛc74$™î‰Rà hP +ÖCàö¨÷5¾†Ÿ‚y]Q*–Ã~Kæ¼K+ïûqì²C–¨U³=Z1Ûeå]ŽhƒUGþE2†8,áeÙ÷zrâ'Ç*É{Ø[ìòÅ ÒМt¤÷>JI뼂²ð|*¦jÎ2 ´Ñ|{tQÈ_ä:ŒN²l^¥‰õBŶ©ä¥+?§?†Xùq4U㸰™~üµ¢Ó¦“1bà®òÅeŠ1<Ó ²ÌÕ_ÈMÉ•"Dʹß0LŠ\ÕS‰ü(xÂËÿºêl³8ºêvV]³"£É¥BX·šèiAúæEÝ1zSÈÉîŽ÷P ƒBw6ábëc³Gß¹z†Œ‚Ç¡%o 2ÐDY€ ¯k=\»4u îѕƋi&% äëm¿%©´+™hz–¥„²Y`nöR “S˜eヒGO”£Òš`‹QÀ°Ð14yµþÕg*‘Ù9&bÕcù=ésÜĬäÚáñÞ˜eÕ©Ë<…©*ß«‚v}GÏ·¢Øëèó úÖ6™W¸‚ôHR»ˆ‚(Ê{çÉÑL¤îkŒ%ЧKaßÃÝo…Y­Æ½Ï¼¶^KÒ¸ ©´f¸‰¯£Ág¥ÍrŸüsTà7¥mäG¬)ÄzsÛ×5Å%‡Ð—ÈQÄj¡( Zs “1& sWš¿~b¶i§º&4‘¢iXd¿bíßC¸00áà ¼aû,Do‹-puÌ[§Û~cš§²}'cjåïÜüeJj¥ü’‚2ahÀ/f‹—µ G¿¦(¬‚Õ=C6VÝ$KýQxA¼^?ûVFaê•…è´:K¶%`fñFc ÿk(¦ê=û¥ây&à¸Oý{º˜Ù˜Ó-†éLNå\äêeÌ‚7»Ã=dƒò3]ZæfìãnôØ?î}ޡĿóÇX;BCßáÆÿPƒªÓôÝoí)²ÖpÝý¶¤"­?‹Ë©™´z­`DÖXfÊç+ñ!púM{Ê,â[¡ü¥SÒìçÈ IÛÃÌ[Û·ûîa»ú²†‘CÉõ+'¨ÞÊ£¡eIˆ‹Wd.v%°d9ãÐßÄÒïãš]gý\oÃ(V=ºñûÈ{r¸çgš¹Xu ÈM‘-|u´©š\A{oö ~‡¤Óµ.«ã2çPjá/1fíH””J>n8¨ñÛ©ϬׅŸM‘ù“VH|ò™$¢·vM7+ÖðÇ¢èØmüõ¥¢ÔeþÃ/Ô }J4>>&О`²ûŽØGoI+-…5êtÑ%"qW*c Ð’ªÖ.b?õàQþÛ•þ%ð½šñYÌ8UI?v‘̓ Ë-ääæA´±uÔx(#kÁÍÚã‘ûãÏ$¬N¶ò/ii^dÐßVM¥+èR!¿˜[ OTXår|Èr¦U1‰•¯íØBMÏ"DK´WREæ=ß.UB A b4·éoNa—‰ -ˆk+u•¾åÄ¡[¸¤ª\ ý vh:Qðéaãû)À†»¨Ý A`w9ã½pëâøÌ¸|OuNsòœLÓo.u‰`¸ÄÎÕ$ôY¨W`ÈrF÷¦½"ÚüÊo4a7¤inðJrø}Úb¤îî Hsc=‡L¼ÍœB\V;—¡HùfŸ4|dÖê õ¨Ë+Šî~&ô0Ü]©ù¾Ïcèꊒެ¬‡Ä½m:Ê‹ /–î~ÓÝÌpÁ÷„$ö¾cÜjìJH~÷,þñ†–v^ ¼f}b€ãè—@Ð04˜Z|5üípà…ŠÕù‘vU8’m-ïËmHëú€a?ÕÐc¤¸IÚ‘4Á…™ÏË)ÿÈóþ×¢æ<÷‡PÅôƒîÆÃX€IŸf¸,g6é}Ô€+gÄ=“PùìN*Òf<û‡¿3÷ì$ÝȲ<•ÍÁNÑÙü(@ûˆñ´²ç]²M'ÖäÜŽ[‚iþ=(]YV™raè®ïäÀÝx½;ãºq]¨‚âõÈk²•ÉêþAû“Ä2©wÅD}e®UAòE’Ø"Æ3e/­åÉÉr)§þ­µ3î­õ'›}ÞL²J¥:êôÛÛömFì2_89*3k}ý§ºšsxÎo0m<œÍ¨E°¥ê÷½QÑàOÖ.Ì>%‹¨8ŠÓÜÁ™»«ætŒ˜Ô ìêNgnÝÐG>“æêjDDª” ”г÷ä!ˆ™ê‚a†é³õmÂC¶¥´®í*ölIŸ¶ô o¢.(Í6NJå¤×ÁmëЪ¨:â1ûœÿ¡r:|abyg8üÿK^°ÓçgͬÝç®iÛh|ZÖêa¶Òª dÂ[ö ØD´Åhd›Ç¶©}Ö701ú,Âߟg,Sê¥Uñ€ák¥ð„Ü­IÀ·ËX™2“Íb$ ô°:NÜ qÁŒ™|“ ž'ÿÜ.0o•““xŽˆ—z”FNYR"^Ò¾ÿ!1T‡»Ižl|5l Zãa‡|_*ç—Å fD”&ྠ&Àá±¾7Ÿ™–(™A˜ûWKÔä¥èüÍ%“µpöno†,›²²ÿDíÈj"_έ€Ñé²4v•t; „§B Ž’¨ë&}•½>©8Þ&¶„õ¥=á!@û/ðôÝK^H­¤Q]¯áy©+c)²‚_C=-¾¦Å}Z½½½…(}Ú^õá¦SM~ÿjû?Ó“¨- ~/u/©íuD0Ï)þ™ÃTc?œøRSíFÙ³àîJ}s3Éž»!"ž^åÝÏ]µ^A“z­ƒåÚWyêAS=ÓlÓ^zt'¶<Œ­‡ƒ’ΆœŽÎÈ…}¦¹ËXqU%é‹ãš3LÔö˜dYÅHŃR–wƒÏ5ê°)‡›ú$“ñn|‚¿“‹¤ÒaqRÏà?0fY”‚Âú'±Ù°Ä”Ù×)Q`QÔu's숉w\»ùÁîA—9e)€`¥;ú$æ¡,ÄÉ0TAIû¯ N¡Xž¢ÁÉÉfì^õ3=›vµ*”ÂÖt} >&EVýešþ¡pS>Xv ]Q”ÐMªx‹þi $ª±Àê?,s#“q×L —"ñÐõúöRÏB{¶­ž0JÁÙƒ)±©Ù 5™Œ6ú’¸9¯›}2Ö|„n°‡ù¦u!Ëc–A[žƒÜZðiÒÕÉcmÅ£+7½pö¸o”Ë„¦Yb"¶eþv—»åu‚½ÐÅ(m?Kº}—6òy5jøDWö[¨`s^(hÙDÀµV´/lv“ýÁ‰IƒèG gj ›“‘b-Ç~3šºäXÆ'ÍÖ ®!„+çžõ±AdÇ«Íâð%Õ}P>^SÏœòZÐâ¬äÜm¦ª‹'q§ $"0"3`Ìc=$œš¬úE0°%ë"윾 âX¼é r¹Î¡• #=9c~š¯RK¥æâ2ýIã³îK«/h;³ k/EçGœI’ 겋¥È& â5M(=hDÊ€@€¨À°–G.Ù²V¿5}Íôh{j •™zäñ_&ee° føð‰LešaW[ï\»áœX~²Ý™ÞTŽg–è»®ä§"HpÙ­BRn×âîý€9a‚Ó<´ÈÞÇè—¼Ù6Òþ‡d¨\“ ³ÒòŠŠg‘‹¡eÂM©/ÁI6¢Lı7ö„Ö`c10·ªwUCÁ´ÚŒ˜K,I"¾ììÞÈ;‡ÃywÎuçëZŒˆ­k r÷ù‰ §´TðtC0däM|E'ù©7˜X‰§ÚÚ3öczˆ‰ á“Ë„w,JŠÅŸ•”™7?d€!Eƒ#c-Âõ·0˜”îFj—Góæ‹ƒ6yÌ”\oB]Ý‘ÖOÜo æ—‡thŸýO›‘`ÎÊ¿I)úQuú—û@Ë]Zk!Åøº'Hß“häíMkl%»_u³šøÊLE ù†Z)Û ˆáA-‹‹õ)ÓXl)—¬X,÷𢒚}Á€ZH½Ø,QŽœª9.…§{Òj“Ëa€©d”nÁJ¶Š¥ø_«Çv?©Ëm[ùuŽÁ§éj¡Zä‹!<Õg¥J×ÉÈkX¥^áû½õØ\#رªÎÓ¯ hæ ÉJ ôùܪ;.ﺈ¨@‡«{á½]ç´!BÑÚúƒØ”¶W7{ƒÙOÓìñU]ÝßÂŽR (âJÈ-\Ð8A¯;üÐMZ{Î;n-¨¸}w{>нÊbún/€¶©7<ć¾žʜєIÙ¤¶xœykÚÑÞ+ñ8V(cNú‡)xeÜ[-õéì—Í+#à÷·€˜½DØ ì,‡µ´‡äSq%* ž<æg5‹ +ÃÉ,õÛ“ªæC^bmgœ²ÔÈ/ÌÉÙÏfUy 5D>¬ˆŠ/k°}ÌŒn$Û½¹¾¦½¥ -#zÝ÷Û3 Ü) ù¾‰‰-W ¾Yäk.žÓ™ù_ÂÒDI? (¢3Ò*¨f:²têqÍ/–‚GßjÏ}%óì̲åËéòH\ür³Á-QDk3úšÌ_ø²i_£œ8™Æ/Ò‘ûh‹œKuc£—% ¹ŠÁê¾ÊŽª-Ü{êe¬ÄO èîØu4‡›Të8ÈõqqmÉ·ú,Ó:°xlu0¸…,`UÑô‘Æ·oánfá/9 lœWµ®'³TWM Kä¼´\ ›õ´¼zmкˆ>ð?©jV‹¼|òÔ|£ÜÀ©&¹’¸M'áÿ{SžªÒf~[GòE©#Ó'¿(byÚjöR1º¸J–ÂÙ¼TœöŽÈJúýi.ËVVŽgDô qF}õµ¬VLÊPE@©Z¾uuCUž½~TÇŠ†{`åï¼ÐD¿¯89ã• aWúçàj1$ ' ´†ØÚ»IѶ¿ R2`cã(Mx7™ë@´Jÿ<îîí²…c) bªÍy@§ã%üU÷QÔ»dñÃêç6mÛ¥ªvdÇÔ#ç¯÷jjd›8SÉñÆmVEù1è‰l@ºïWé8³KÄY–æò…¸”VŠÇÂ( {zÜÀðéû3ªw§Ààh7˜¸Ó5L$|•QþH¢”q¢Ð¢ÒŽ £còÏvÍrC‡ôO7;&È¥‹ŸÖøT yvS ˆGg–ËÆß`  (öâŠ%¶R¼¾Zj惭šñl¾+è-m¶é€ÿPàé{.˜p;Dè‰â(X›w4Ø™øºaÉpYÉ[á!¨Îã;ö9<¿‰0™“Vgä-,“Gdœ~RË!KY à?ï ÆßpñTË¥ùà›.ñf÷Ÿ•-Sg ¾º ë$ŸP6¯dOi¤9ÙÂd@­†X%ÒŸ8 $Ùà7~ÒõçèÏÂñ´™v~,××érŽÏµâh™óg0ÅãGÈ] ÛÔ±nyG[»ì” »qGÇ­^!í=!áXíÚˆÛu™ ±ÜíŽÅ“IK‡^íG¯Žvþ|²šS…a¿%ËîÁÖ÷š–Ö{c’¶ËÃàÚ´Glüz™M<¸IB݃ˆ}Ø*º {è½T80ÆÏèÜRƒ,‡ÒNV˜$»n½¶`G=æòuIó÷êí“…q`V;ôLj¿J÷0tFÊËtðÉ`‡OÖÊëß?þ sdsV†NK=Ç&v%lR‚‹²ÒJ8JZþb„ê>Z¯Aó1ÌÁŸÿ/W¯ùò´Ó\<]u??g 8Ú+Žn‡×"é­ÛÊèέC_ºS;B™†¥¦¢±QÉ ­™'QW"­aö §m+ ¾ÂEWDœ«·SgÓ]TÍMy9›6+8šr ÿGW©7I½‰|׉²Hñ ’érám>aøÉ–¦:N‘Öӯ줪 —N݃)ú§/S41e¹³%$j–náY \Ô:wšÍÄ{Öå+=j,Rœò‹¾#†Yh¾ GþðÜeµùµºÇ~9çm¸Z÷Ÿ$Š£9Ó]07§oä…§æƒáO;úaCšÙÆ¿WW(ÿMÛÆ×h[P‰¿B™ïÞ«»—k¨…w°Û7 ™ÍØaÏ0ã¯àY[ÖA…kÌuÜ⯢­‚;ï}¨É|œ6Ló \m™xÞò!C;ÞƒYKŠÙäRªÇµlîD ÐV´]wnøÏèq8ïqÛl±Ö¡!ר%¤‰Þ\R_SUh7Ê òNÄ?ËÞÈ»ý;ž­ 5™ðé&—èK6Úu” ;pÏ&OÎÒ(C'Q%Fiæag@ÚÜ-3@Fž1D=” ãýÄàci ãDê¿üÊö4®z´SƒáÖGqÒ–ohn˜ÐêïA‚eR`kR‹‚R‹÷œ4´½=”"¬Ôõ±„½Û^‡ñš±˜JõÈáéÈ,À¬œWÔ¬!Âñ=ÈÆˆ}No;͉ŒùgWèÖŒ Å´ ÄÄûîÎ&@q=AÒ»ÃUΰ-pD£ÃÑâTµ@ý†P.¸¤l‹ô§i æë—à¾Ãf%ªXÅHš£¢'ѯÞÌâ¤uã{»Ä‘… ½]S}mbžÔœåú„Ï•¾üÇTvl¤®ýAÜÈD~¡ÿ3 €®Án¨_C†cñ;/ÙNisrvOT’Vé Õ¥3Áª©c>dèFGÝ×üoîèâô4Zû’¾ÿDzµ°WB5ÿÉtÒDVÏàǃ´[\ñ Èɹ(D¥#æ/U¹ùg^h4ÅÊFÌ÷7—…Õ‘Šôò<+¹Õšç¿áÈÄd#©’ ×±½ZF»7´Ž!ü ¬à7Pf•Ý õqqÎØŒÔ=—ÌLÎ’í÷|¯§vŽEŠ™¡t¢ÃVÛ£…ÂØóÁ—ÖæêG°yˆjœ„nÈk¨nÂþf¨mÌȾequÙ’Š¿óÍZõ÷ Ðs«Xt†æ€™ê‚í×­ m—'=IW£ˆŽë4B| ÔOQ/íûÉúøRhjWù2±ö*Ul­oCÐ ['.fü V"Ó0œ@5<<‚!¨åƒRà‰ÏÁß €»8 ‹@YªÜ„ÒZÅ7>¥v ¿õG±Û^t†X§oLt&u_aR¡\HÄýå߉žèÂÍVf±‚mÖ´¾sÞµ|G¢fvQ%‹eeÝl˜•Ô;»¦ÐGxwÓ<Ô#‹zÙ¶õ¸ÃNÇs8ç*líI’´±ïgðZ8ðzßÐàÆ±¼CšœRç®çÝÇds0¡Æ¸­YJO³ù ú>j¦r[íN¼»^¿½–°Kþþ;“m¤ª­ÌûÑ?7Yký‘Åv)íª®Ér”«€‘Iœ«ä¶sÇ A"£é Ö¶4¾SÄЈ²0)ËåŽv"‰H|m™>tç#º˜·'×W÷e»¥Ì\Oò¦u#22© Ðj)Àü˜ãü{´f?`Ù~éƒs©RbL–Î€Ž´º„Aî3WK¿tŽöÿŠ„Ž$8H9%šûÑ\Œ+ÇQŽÁ}YÒn|3÷"V~'’Vè;À«,Þ{'ºØðc—Š˜RYEpw ÀSéå³êèÍtJ‹‰(Ã|Ç NN¼ û/®ÓXa†ì$è{ endstream endobj 60 0 obj << /Length1 1995 /Length2 13571 /Length3 0 /Length 14789 /Filter /FlateDecode >> stream xÚ¶PÚ¶-ŠÜÝšàîînÁݵq‚»{pw'xp‡àîÁ î‚;<öÙçžìsÿ¯z¯¨‚ÓÖ˜s¹h R%U3{ ¤=È™…‘™ &¯Â `ffcdff…§ P³r¶þÛ O¡tt²²ñþ#@Ìhìün7v~“·d]l,lN^.^ff+33ÏÿÚ;òÄ]­ÌòŒY{Ð žBÌÞÁÃÑÊÂÒùý˜ÿù 6¥°ððpÑÿ+ bt´25ä-vï'šÛTíM­€ÎÿU‚šßÒÙÙ—‰ÉÍÍÑØÎ‰ÑÞÑB†àfål P:]f€¿(Ûÿ fiåô·]ÕÞÜÙÍØx7ØZ™ANï. 3 #àýp€ªÌ'€¢ôwð§¿èÿž €…‘å?åþýW!+п’MMííŒAV €¹•- (ù‰ÑÙÝ™` 2û+ÐØÖÉþ=ߨըÊÖØä=à_Ì’"Êã÷ÿÝž“©£•ƒ³£“•í_-2ýUæ}Ê 31{;; ÈÙ þ/~âVŽ@Ó÷±{0ý}³6 {7׿¹ÈÌü¯&Ì\˜ÔAVŸ]€2âÿy7Áÿ±YÌÌÌ\<ìàgÐÝÔ’é¯òjÀ9Yþ2¿wàãå`ï0oèce|ÿïådì 8;º}¼þéøoÏÂ0³2u˜-¬@𪿛æã÷Ëw´rè2¿kÀü×Ï>é¿ËËÌdëñ'ü_÷ˤ&¢®,¯N÷wÇÿñ‰ŠÚ»¼ØØ ¬,nn3Àç¿«([ý›óŸT¹=€ço²ïSú®ÿ¾êï à¿k)Ø¿‹ þ£q=ffÓ÷_,ÿÏJÿWÊÿŸÀÿªòÓøÿ&$ébkû/7õ¿üÿ·±•­Ç¿Þ5ëâü®yû÷-ýïPMàß;+4³r±ûß^gã÷=YØþgŒVN’Vî@3%+gSË¿Åò·]ý¯%³µ•ì¬þzU ,ÌÌÿË÷¾Y¦6ï/‡Ó»"ÿå¾/Î)2µ7ûkÃX98ÆŽŽÆðÌïBbåàx±¼¯¢Ðý_01‚ìßSïíùÌíáÿºQN“È_¦¿'€Iôâ0‰ýq±˜¤þ V“ôÄ`’ùƒÞóäÿ “â73€Iéz¯¢ò½WQýƒÞ«hþñ¼#ã?ˆÀdò½Ÿ`úÄÎþŽÞß…?ÑÍ•Éìð½à?à; óÀwÿ€ïå,ÿ—¥‡ƒåû“ø'âÝfõøÎÔæðªí?à;W»?åÙŸRï© ÷kÿ‡ÿªýŸÓß“íÿËýNÝáû½˜Ãûc²š;ÿ±²üÛú÷&þÇü~ÕïûgÿÁ°¼÷êøøÞ˜Ó?à{ÆŸüwA29»ÙÿÃýNÞåd}÷»ÿ¾³ðøsøûP<Žgÿ—žM]ß©:ÿëÅyûÿàýûݦðË ö¦|ÁÖuÁ÷5"n {“³{ši4 ^ËŽ.ȰÉ4ÕYŽ·"É#½¨«;Ô7Â+$/^'­ °am‰ÊíOÞφñ*3{íðKÓØƒSE'"õDp„ jÂûÞ/Ÿ½5l [Á»d)ò>»p#+`Ü»õK¹×”ÿ]ØSÞ¯æ”Cx.ÿÁ­¥P2G‘o’=ûÆ™è-úow”¹›ÛYôÜ©7Ùx:xŸÓh¶b/MÖ˜‡yϵJ5V§nlpƒ¿8ÞÁ,W/©®ÊãÎB+‚“†yh†ÛšpÁ–šÞâ_GÛ’ y¥-¸Ò¬ê½šø~u®_ÈÀÌÏXn¡*Š‚¨`›„¡“£÷é|6?†ƒÀ€§¡ó•·D®Úû—$|`ð›W_`%Ì¿|A‚~/çésŽ`§5³AªûIJòÀ+•VßÑ™ßQHD‡âž‡ÎˆÒs ÔN¶à[íÒªÃ1K€ªF ¸Oìï×îõrã+3fËþE/??~;˜yV·³ñÜ”ÇP¬ ŽA_ÈÀ”ás0ø‡ B©@Kÿ(ß©³«9~6Õ\¤³® Ù>°œOÄŸi¿ÜtQ„ìð»Q±ä9\(ˆ»6Eå f1Jõ7¨‚:ÜuÎ%`{Îf>1f«~I”¿&¡ ýÉ?ÒõËõEiÖÒŸŒŸE†LÑpÏi­¨½XòÎÂ¥¦oç ÛdÓú'3Lèý[bBþï'¤\;¥ wªÇXJVN/#óK”œ»©gTØh+¨‹ "ùŠ)½vØÊ+ä âZ!ŒÒ8î?2µ€¿L ¬¾ŽÊ!ö¤£¥r­ö )§Qbg²CôëôÏѤ¾ÑáGØ)÷Saðpn-ÉpÁsïQ‰gAÌ|$Ñb.¼mÉx ;ž2[Ó=JÁü$4”`QbìÑt.òµnÆ`·¹Ü&ö†;ƒƒñÈ*™.(ê»,´{5"oOáþêÔ¹¶ÊÁ”ÅÐoõô‘ó‘­ú#–'i¹>U˜£Ð“­9ÉÆ½wÜüí uĵTÏ`—0t;"È&Ó×óx¶9P¨x+Ø·¾#›U/’K“3™V¦ÛÓ^‰™A¤Ñ'ðíqËqƒåô»E%Wý®åƒrˆ¹£­#.*…a3³¬Ž#yÆYæMÄv¼|‹tÕ‚DWÚ_Œ°‚O¥GìL}ÓÿTˆÔök´ürå «†âáOF̵~®DÜÇ&·©D?¡VÈb±y)&˜\Û6™9e7²|•¦Hð ÝI´WHËÑR„Ý_‹6—½3æúˆÃÕ‹ÞL'¾m"”ÚÎùm%rþ°Ìà„Ö…c]ÞhœÎ#&ÑÙtU¿ÎضíPò¦ ÑjA¶ûݶX‘M„Ž.@Ùa P‰}-Óç”lHT8^—íã·Œïí¥XÍÙß3š*šËe-s‡€æ?œÆšŠÿRoŒñ\Bø–&ÌÅŽ ü†uÌN@ØÁ(þ 0Ab”óÃ’Ê¡òÇ{²ë„–OhЗq4©64q"ÞÔ@ƒ¦ØÆüÙÜ8ZÊVö ‡%Çm÷èç- $q^¬%J¦ð…ªáx̉˜­̸ÏJÃgá¶™ àÛ:œpmûëWؼfïH7ùHÇDSLž %,êÝ~îvØj¾är®!i¦ÄÂföT•iÜvŽïU³þ0 {êX‹ô’„Þ´«AQÁ•‘Æ${ȸA ªˆÓº¤”UÐq„{ä~Þówv‚jS65[Ž]Bj´…åA‘Í.“šFBø(ÊÜD×nRlDþ¯|fÅ0-«¶£öpø­¬/ˆ‰ê[W7.ÝœÛib _(ð´Õn¨T®£†ð­€åÚ/°É—ƒ|>½ÆÒýNäU|md7º:˜ÍÙÃÃao‰i¨…GEü2>L^¡m<_#'ïgÁŒ-£œ„wſΓúäêÍÆ ;'µ¼+ìýY+4Jpk$ÖA§Š¢†šXëÚ5ytŸË‡¥ܶB¤Znå›?8 g1Ú‡”ýÄTºŒÇÍZÛøkCN¬+ª¼“i¶U#Z5Ä&êoeú9}(œ*K º=ÔU§"²¤lN¼Ð@Z‘‚ãÕÂqDh×òa §…™(»Óéò FÆŸ‚ˆa½á³—Y‚Ê]a”­)CîßöNÆ[=AiÈ]á¡6b¥5æŸ5–“_‚8?GØÃÒ¦9Ð1ÞZXÛêä9[:ÄN ±ñ²;/_,{PH¦“ºñß~&9G>«i—d”ŸEöÂ-ÛâËa=š$|ríÇÐV˜¤“MÜ©Æà¤Ì#?Í|ŠãGð@?&s@¹›ªá][×0ªí&#€Ý„áRV …qïÍåzÖdKÙ‡ótÏ}4À©zðWA-ÂpóÖ!5@ùî&a¼h^ûã×H³•Ý[V4;;“É7…·S*üô[H¹Ÿ½l{h—GµŠ–S²"Céõ‚%è<®¹³q´°T2vÎ-³ßxÍÈÎS “>Í쫸çõ‘†<›‘vª €'eú…g4Ò^äÁtíòìçá¨~7ðYêÃO0!¯yÔ¸ -7pNa÷±äƯK+„;öKWÉÓ+6ÃrRmãmM¡ÙÔ“„f¿…*(㮹ƒÁ*ªå|ÐØch§n&•mlÖù×YJª_í¹¸*¼ÈØä_ö€…Èq6Å›(þ©ÃÐqYzï¨ê³E`I5±ãAI ê¡ÅDµEâw@¥ÙqWJ·{ޜȋéÉ@ÚhÛT~Á³¶Ùu0-oν}´™QxAˆü£’IpT(¨(>ì¨Ú*AŸ¬±8bÐÖ‚rdXšZ…r˜bÂa†1C«(LJT«®fùˆ·fëÚÅÓN`Hþ0‡7Nî –™1xB„J7NÝ"7ß±¢ô¤ñ%AÑ!tÈÙT̨m>o›93¦¿Îÿ7G¥¶õÕQB8‹l "~5FÒ¬À(HHÂ!&fËø3ZA¤^õÃ1͘!û2Ñw9íüjºF³AÓu/œƒèÌ:ƒÎôÃ/˜x4•V·@?€"I}Õl.í+·(Ç+íþ`pªû&¨S˜¤Gâ B\éM°-HŽ)sóóÔ ´ÜmŒ ].ˆuŸ_`C¶ùÍM¨Ç¡o«ám"±ïe˹1›èg7©!  c¶ß̱ΫøÙ 1a‡>ñ‡ð¥>ÝRS|g¤ØOÏ֚Գʩú·.¹‹(BzÊù€•#’=üÁª¸ÕÝé–O¢5Sèd²»uà¿»¯^û.Zˆpp˜±&˜`“Z!²èhñr\:ŸÄ?&_›ï˰” ÍY;÷Qpó2QGÉäHåç®À€6rFAá…¥!ñÑåו9Þî³Üs–¬ÆÂüªÛyºö<ûd¾“B}lrDJ)ti)8ˆD¤°Ã¨;é›ðYăG3|â=ˆºcª\ù6î€a¼”ò…wgšÅýa(érx×(F£;ù‚RB"7@¨ÉyáQìgêP:WƒÑÐ29Tu•¼ëì1Ñ1¶w}¥SÔ$â7N-Ò~ 2ÙT˜n¢Únpˆ:Øl9&Õ~Ξä®Ä0ä4Éžy‡­½"¼–p©‚8Ì]>N¦Ë¾óºJ—ÜL\íXgõñGŸuóf`¢•·"㛉*D›Ó€(¬gÜÖBÆpic«Åoˆ¢…gLÎRÔÍŸ&°¦Ži$­—?ã5ñYýD>n€´ Æ/XÁÛÙC@ñrD=Šb Ø "e[<8gšîãÔÐî ýo+9~Ä5=+ÎÜ8¨> ¹× 21â£ÁI i]Å$]¤q¸#ëÐþõ}À›$1»ºã—%Ô0X‡íT¬øá¾ö:²c3 ›o' B·GªfQjbyi,€}ƒùiø†T޼Ù_ÒÈ”š®i €´Ç’¸Ä‘ëÏ]OÀ„ض£Ñ¶S]r½\ÒÚw˜Jk éƒOB&„¨D°ÓâÏó|o×çký;)+ფ+¡Æ €$Bî÷#*¬•vÁí`̵N`d.F4QDø‚÷>:%lVçòº¹#ÊV“¿r‡/¨™ð‰#B}ïŒI7¢“W71Ð_QÅX‚ó,¸'Õ?…Pì öXm­4VMo­y„­²òÍ…«á3Òiû½µ-|hd[.éêo<Ç„àüh+¸#T#—~nÒ:rüzS¨ÒuõôgÅQ>_ŸŸ¥(-h÷f‚Á‘ÃøF‘ÇSH,Ú¾ÃBƒÛª?Û»¿DæÒ¡ÕT¦ðè'/UwáoÆ’åT‘J½mXsµ[ß­Ç¡¸yèkÏ­ÿ©_) $æü¼jȺ:4górƒ‰Ay2pĬ•âZL€.–I¹êdV6ïaj‹¿…‹|"ÁÛ8ÏÌ’ªn{³Æ$JNóE/İüÙxo¤¬ãÂîh|m’3Æ—å©D¾”üT#lÛ7ðyÔ5û$êÒBí–6<@ ˜—’¥E9Ÿû®Íìy(æ>hßUl¥å/'¬ œu‚™vg^Ý¿»öĺ^SÔkµà˜Ï5JAX€á93Á^ÌõåŒ{¬ù,Ø¿6l¸€ÃÆwž÷µ3[ÜÍácŸõEÔcܤ‘ªŠ^“6r-ã7…Mœ±Úx›=Œ ßui²#jnä)®z€+Ê“ 1‚ßa"³¼9¾×R“$\btª›4ÈÁG¼–ù öñäÚãfר¼(È¿4K„MHâ6†Q5W|z/ ÚtJÝ[ñ²”®Uˆ¼ÁqF`9,"î(î¥á€Ãh<¹~s‹iî·#‡>ÏèÒoAå,…E)+Ø Ê»¬x‹ÔP¯ÐYC”átæ9é Aê±ÍÎÙbOïLµñ† ùjsÚÂ`Î^,ØqÓƒo+ü©Y*® nßh÷äQ–bîRX5 ¾"Z… íÊŠš‡äáeŸ©|…€µòÚ•&k`·D Ï#°ðí€çþΰ³xÒ9)|Oísš‰û±­–£† Ÿš¬TUhtî0žoäåõJ•™îŽqŽE—‘?†Ãÿ‹ßy1G!c¹g|ÃVFš(oÍ)»OàÜ ˆR3¾=…sëWÄ";Y^F½XÓUöI7½¹?Ú‰ÖÔâFxÉ×ÎÆÜµTŒ{mt#8 ‰GÍAG8Ö-h|²$8ÏŠË¡¹‚µ)9PùSE=W]‘C7R3¶™ðŠ~Ï›¥ÐŠ®]ˆƒÙ©{a*3=”Û¾½gýefóê™mQiªSðúß18šåXQ¤ZœsÀ ­ö+&?§{‰ìŒ3êŸ><,gÌÝ»NËiáý0w.Ðû×85NC…RHázE8b—²`y4”ÝvòP;yã]°zP"]=Óã—\b•y…ŸJЋeE[™onÑn2U| íÞ8ù‚bªà\•¶i0¤ªŸwö”p)PêÌõ)büb[÷vpmÍ;Dl`·óÑGrC¶I|š„à §šëvÞ Ë®41)Ji É&ýk2¢žèòqÈGâX÷­ ÓËD³=]¡ÎõÅŽ†ôMéýf¢Ú‚ iƒF¿íÉ0·v„Þå/êM ¿ÂØ’ü¼Š¹²kÓ€/£/VÓí¸‚)\^0>§pŒYù¾Ü<)@—ÇxälØøƒžÊÿk¯ÍÊÕtIŽ#í,Ú»fˆÃ"y¡ºCTSéL hÁ 7*ÿóJ^z¼±‘â¸÷ç³4èüõ­I¶#|ÈÙ¾6ò­c¦+ïò*Í FQ›Õˆe|œûòÞ¹}¾—þwÚ—rÝZÛKjûŽå7˜¹ üt[ ”S·Äæ–ž…`:ÿ(?‡k³•z-[ÌXq-Ì‚)„VbVy`Ò^-pqñ~ÃJå§–;w«ÈaªH=YO+&Ï#Ubv % ³2½7ˆÀzêY‡÷\)±Ä±ºiùZ_ͻƅ9ÝsSS>Rõ@ &.ò D¾†ôÃ&ó!óµ?úp.šþ0ºþžD¤pA2{ɦÉDïfç°ÎUfín¢2_}ý*\¼vîøêÉæÞ)¥ÀR°)Ã($>×8ºL©†ÚךBé[hËðˆ |£”¦mò§€!ˆôì°ÝÆß_o¡ÖâË({"-]nÀ(;‰EÀCÉQ££ç(cxLõ”?JQ\GØ.x¹i Ý:…éîÁLþ æE•›J÷1qן s¥%¨\c[:2=áf¯1šEj%÷NÚ=r[¥ãH­nâGŸSCŸƒjLEX}]Þg^ø@Hw )Œ¾Ê¬—§X-dX›‡b›êã>ݦ™Ñ³ÑÇ%@Ü M|– «H¦à%aAÝÚ!žŠçŒ€9ZÄQ ͽ…µÈZ"•{˜‰¨’õN±ÁúYR”¾|×ù=»%YÒøZZþoµLªuýòã<>ZDº?UضEXé‚ÌN>n»‚IrÂIÜ$èž3¢äOÎê((ŠíôXâbEz“ ÿ–;äøñ…‘ƒZnëkV64ˆŽ½Ýc^¸k:;fZtþ—®6IKGÒ \%ƒÑ3ZTrDRüÅDëà×X‘ÔìBò°€ÐEZþ:¿~Rjw"Jr—o¾kQ ɺ!;î\7 ;˜s©-wàÖµÿªä óÉŒB3¬4ñäòs| ŽÕЉZ‹ziE=¯Õs_ެv¹a>l ^l¤Âjê¯uè ‘Ÿq¼½â¢ÓÁ´þWq¬T>¿y“Qß<¾ÂŸíuÒP룼aQ×¢f{±=Á“òG…E†§÷ÜÎS³ Œ g;“….Ê[0²X7÷гAAjö*åÃK#û9Ÿc_ÃO%Ö¬>W3¯ŒBš½f‹|ªl„ŠÃ3Äut¢§Æ úì"ø ˜!à„œ¸!ëO&Ÿ¾‡bC×wsÆ–÷¸–ØÉß »T¡T!yiDvlņ¬5ޝ\eïè+ŸZÚ]ZE/×õ<:ãAzsæG‹Ò¿?©¾HN}"@-[6¦g]d¹X‹UàôEÐÌöéõ-eÖNÄʼnGQöΧ$9é…êæ¡·¬}Ç2 „Øw $Ìdê&Ø…ì°Œ†Š1±4g\òßö àhòì!€·˜Ç*0· Ð¢§LGVƒ&ÈKñï*è•9ƒõ]² >òæñÔŸõW6ñ}–ÏN]Kn†ºÿ>VËàNf6-ÚBMÚB‘P"™•&•C^ÀÒ·˜xS§ïGQã´t°¯º\¬žLaMéŒûYo­;ÀÝ V­ÉW =’a‰NãM¡>|ÑkÊ‘PÈ3˜íîÔÒ%ï„(QW˜Ix‚ c¿Â…‹IÌßÜÏùÒ´òËåkº}Ap…P{ù¦Ã‹çåN“Y×#{›ë¦ï]z` i² ´}Îy=#]¹›ìÀèšnÙ8LΣü1U¤kÑÚ~ì²=ÒeYrœ>É#Ôª¾œ_Ã&57#.ø$’ŸŠµ”$nÒ/vOä;i^\ÿeujy&[Gç–{ôI<¦Ä·  ¡{«ésþ¤ £r8,‘êK@1µ)Ÿ¼MU‹÷viãùD1Ñ*·{—6‡H–Ø!ò ¬;¶ûGy°u­/÷5ô2—h¿]Ò¨2¤ÁmÓYÊØ2–ø£¢ñU|ø†–)*¡FFßÌyßÝA3H˜Ú?í^|kj"oMŒbÞ;Ÿr¼ÛŠÇ¾TƵ¦†ºÐ>¹æ/ôD¯ÛòÔWLßùuW´×µ…ús?pi¬#«±rykÛYÉ< <γíÀ(·w†¡±œŸ|K®êrlfž´íY©x|¨Î\˜‰tÍ!Y°ÓAüS2 cçƒzµÑyÄêOÓÇÁ<ÜB­)²ëqX‘Dî|š|Ï(*bÏ G’ÓeeÞ*£YǼöIW¦9Öí¹š9 §v”¬tMv·ÅÜ<+ï±°p‡HÁ×Âtñ(‘oœŠ’ŒŒ6Åòãø£…Ÿÿ½Ž’àÒ‰7ÝüØË\>,O•Ù/E¹¿‹}2ã»@VžûÛº¯à⪒ç6¼©î¥¡g„Ê$¯æÓ,â~ÃöÑn/Š¢SbÆ­Ðîøràí…wñ¯êŽ‰Ãˆ)‡T¿HghÒŠAªrâÏï9DK¯LSÞ1¿àLøßåPVèÃÁËEX*Žg±Ä[Ø• ß@qƒAmoS ±Åˆž«ÅhÊ 5’•Cò¶Çà-¤R/æ–‡ÖûÃ?J¦ù¶@<éxÜtÄ$tØúS6ÍÇ M5ï)¸ÚŽÕ5º¢=Ð{pV`Ì‚»­+~I¨^Ï:g@×^‰ô¼ãÞ.f8DÞ4Þä×Lt†l›$]‘ 5FïÔ}»šðF‚=,jäD)âi3ÓÀ؆îV©­K–雳¤œkìó+5FA›¾P1”XÛ³8-Þˆ¸f ­Êð-?®G?Ç oÎûâx»ÂCĖܪþ•1 ¹çëæïOréØ\ÿÊuàýµ–¤vçoúAxÑ€EîD½ˆÎÒZèÀm>ªá>׆Ž%»¨bkoJ!lì°Y É$"°¾àUßVFL|+\¾G¶­—ÜN¦/(šcŸàØTFm>æôX|HOo±îTÙÐÍ_ðž u½²Š‚¾¸,•~š¬ŽfWóç}8Ô5Ãøj 3;qÏ멟vû7ë3 _ÞÎräMÝ÷Y:cÚ"ÇPŽ Ó$w|ü¦ûׇ*’-A.Ík¼…xâ:»Žü3'²>äUž„HÔ•öŠëû¨.*E˜úÿÇÉS”køûŸªøæŠNëc±×q¶K®ÝYœaGì¾WÄÏ+´ëC¨<QþüññðJî»$‹ýƤ„ÈŠ¾Œ†sò«‘}×¥GGzžYœãwnÎБôˆjSL­©†ÜЭxQ^DyH,- ¶ÜÝ0œÖÖÃhd=|?Êò/Dªµ:h%I|%£øÏJpqÒ¢k3÷¡³‹”l;vÁrð¿Aò­TÕÔ…õŸÂ$­©—ªÊ÷<é€HÕóN­{)n!6zŽÏ¥ë» ñŠ8½‘»ºGáÁ‰u*“_j…±´ÍìNÆGœzä·P “¡Á4•¾ÚÚ¨0:q­¬;û-—tíž@¼„ò}*Ê—(‡^67mZ‰Еòªü°˜->ލNKi `þq R«8üüÚUR”EÁ!¾óM'N}t}³øso!…îJŸdh9l—nÇÿJ¾ÕµšO1ÛÀ™L)'¼ ÓV¼­kÍ>EGÌ`O †ÜŒ ËGânÑ­aĉ:ü’)*ë2!>Õ9–N¦©¼— 0§Óæý†¡r7™®|“1 Ã„J9›:Ï×û-‘)°)F´ô'ò 4,'íƒ<˜{©óGO‚(¿ž¢U%ˆ­ò¤ó'L´si‹—ýk„˜\Z1ž¥/)×¼UÁÊÌ87 x—²¹zؾýºûé¹ÎÝx«.F ¥^ìwQœF‹öÓ˜@œ^†ðÉHŠ~áÄ@Þ *[Jáho¯9ãÞ”öç"«®rX]º`”ÝÎ÷-[üˆ?¡ÞÈÄSgæáÙb[©¤¬¡@¨ÒÄ?ž(§ð¢”å›ÜØüW›Çò`üºVpêŽQ›V¹‹²±Of£ÁÚXÐ’{OBû· İbõ?^³gvuõnEyÖ1‹À Ú 4Ô‰iù< ç‹m+2&pÁî6=´ÎƒSQ>¥Ÿ éK'4FCB¼VdxÜÃÿ’jÊlW‘g.§õóa: ¯üÙ«Ì0€E[âóç…nÀÅ%Õù§ŸÄÓ8¦><ÈM¦J`âp4îW:4Ì{)¶u2y[¿K,>…¹ÕšÏu¦x×@8Fˆ¢ãÝ;¸ÄÛ8Ü¢÷/´ÂwZ•Ç"t¨µêIø/¢D¾’¡ On Ôìe‚{c[îµäÔK­ößFXV-¹òSeó]V 2Û›$i–P7":oWÜëP»ª¸Yõ¬ÊãE5ÕG!5椾pNwú鯓C`gS£Qà"®Ú…ž¼ìó¤Ÿ>5ú¤ÂëÅGS~ôbûÝ´h&ËýÂ%T»³(¤hµ“¢Œ•ÕÕªŒ›wÍLªJ}ZlÝ/+÷>,\èžÎÙa,øÆX¸p_ˆïëûìÙàñLñ£µ™§sý²RÁR‘ó+RßÓÍSl%Ãó÷cŠC_ùn2¡\J™‰»8Q1@—²s›×°¢`ær Éêîa¬ºî‘ª¤Î#<Ûª„c•fÀŠÕ„«pò¨ Z!4bÎ}Æý….å_îÅ8•M'Xq…ÑšÒ§»VË?eCÌu)‚ý2œÌ#[FÛ™tšäjƒ>(^ #UÅî»IDÉ`hÃ<”ÁØ6Xú$rðÀËE1o½zO¬ä¡æfÍÖ×uׂ¾z­2´DH|[q‡U…8 I9áÌ‘IÛßX¶˜ðR7¼ßw—䌾‹9ÅâÁV„V\ ²,4ù¡ÙÓÀ‹«†ƒÏ$¯±B‘S)!6gÈ_G·Æ?èlò ùC»æÂÒ7¤º¸p€FIiÑ}jd’J–£ H2Úªƒ¡aÏ’(›ùHÄh¤ß䥸ÎÊŒÇvÍ300»¬­”ÚÂÙħ„¿}KÀ®‚óóíaOð~ÌÊÑÖô¢"ž7„Nã[¥÷<ê€4x‹{ÂPÖ~ ­­ŽÊkÚM?<ê°{»±€ÐÈ–.tšŸ°á,u²'–qªwOÚ†k\úx%,î{Ohô\ØfÒmv"Çq1Ú–»¶L´±êL=&ð16g£6-òó`U> F: WÅ‹AbÞÌ«}Æ'žEËpϳ¢UâWª¸Ôœ m×ÊZŒ·ã’£”y&!@ô땤U§ðC¸8ÓþöSŒ^Q/žÀ -—_ÂXpW5uúðcÄgMu\ëˆyœ1©±°º»Þ’­úá ²Ñ°xÙ¾ö– béÅhÔ8†vòó ìd>!mQuB1 .Ü`°O…>)_êP@+Ÿ)蛕ZÓ\'FzÕn; GëŽ}°¨ |ÔI£#Rþ¾Ix9….ü<¾a3gñlVÈ Hð#¦=QÈñ!ä™ùªþf$i–ôºU¼&œª˜ëG­ÔÌ«qÉw ÿcDgS¨•Ïs[‡.~1κ¢ª|äß¿’ûò¶ûÇíìqÙl7bß0Šlö\tß4]ÍÇ7eއpaí²ƒùŸ˜a!·¡â»fí&ø½v â‹¥zTL´ÄµjáçNpx’ı±dÂZOêVO°§†ƒZùAÇß/àö< $9ð›kO©Æóš6÷Ø|\ì0µ.½ô›"~§&²¤ÖE2Bí…Ý: lŸ2¦žàPÔ˜¢~+l@\v²›óZØÂ+$!ÍÔeR˜rýêæG^ øÅãXÑÊCø]^‹ÍžDFÁ£Pm×Ó"äy´FÔcÜ3ßMáOÞ’Õ-5#•êr6ÖŠl\!«‚ÞúÈ’_)I®>×eÿ4‹ŽÓ˜XWš(c§iD„uýU0¥{ླQ\f¡…,ACªÕSm&+1.JOªèÝXeϵ‘!–™·ïMå R Æ¦¡çø)ü»Hº+,^*˜îˆ˜ú-¢ÏQ¡» ]öC!ÝWxC˜ù}îIZuÉ+ÍFÅ @¦m]ã+Iïë~oG]–üxæ¨ÿ‰gäFÒ£ù¦x\DÌÑ>ý\K›ºKSö§Æ¶ç©Âó‰IêØ™«ÛbwgªÞ:”€Ÿ6Ý—÷“*ÚôUaìP¥¬ Ë!¦}«H æÉ‚Û0‹“•ç}Ã9PõnfµöëÒ-ŠAçå«JÚ=<¦9â"m …ûHêXÔº.OÚJ &Þú†ñª)äˆ5wé†%%ü £+¦KO4Yô£=DŽŒÇE?G— ¤öј#J^Ïêc6&s³"ˆÑlyW£èxêù7„{‰&“Â|³¢]s(?ØÞGÔ…æ™ÆÃª½Œ‚Œ’È0‚Àƒ/¶§1BCOë *•Ú²A?^iÅB½Û.ðop"ªÙ—w´ ݘ²ÈÎå‹Zm½à½tg½œ¤Út³ãÃO-Ik2§×|ÄØC—Y6¦§¿Ï¸ƒo(Œƒ¤Æ””'°õ—Œ‡If<œT¸½6­v°x.uôŸ'£õ§w>‘®ÐýJ®†øÈ·íú ó»u’þè{Îf]npJ^€=•~Ð;eõæÙòJËŒ®@J­Œ:>Ê´º\I[ Ÿ°™7š“ âðÎ(•Df"vç‰+^Æ3Ç(ä·ì[KçÆÂÐÐL!™†Ê7ÝcZ *›ìCzŒý~ 0Þ“H” •&2¡ºtïë/”8ꃲÀ¶Èa˜åÝxJë—³ê`ÖCõÕ©¸òÏ—Ù“}ìç-zdD(¤WADÕÝ.¹¿9 F”ÆL0¿ß%v:&K2«WãˆYž5p“‘4ÔøÄ…Qåk˜Õ²m„äÃj:ר¹¨%nÒ;R®y¨Ñlf|‡ôÚ¬nÖCʬ?>ÜÉhP¨$íg¸¹iÀ›~v¶ìà7>ZLë¥KZƒ~µhÉ[‡¿¨d%4ß‘9Þ¿^æy$*ê’nùåâ|ň5y’Áxä¦ô¬ƒrJò@viô_Zˆí´–A¶JÎZbaýÎnáÚvÖ.Ï\ãs²Ç3Y¿{Íç”nCï™PYa|a}Œ5~Ò^h2V6¥K‹Œ4Ü–X™d½ªÌº: Ý2 õÅ¢æùWMÌŒÁ©)š–½{€üyÅTßã STì³}£*á y>œó¯L…ï¸F¶†‚G›Ø¼§9ÛSUÊâ³ÜúÑžMôxq!¬q-Š¡s¤R¹tŤd¸¦èÕç¢u( «·0¥O•kQÕUõ: ¡ê¶N^†ºý>YDÝ J»Qy3ŒŠ€ù2‡Ý»3"ŽdyÌxÑɰ­¤U{ÜÓÐôyùœ!Ôg-^×{x‰1JV8uÉyLL èñ¯ÞŠó[VòD~üÃȱºbê/ëÔ#Ú2÷¢úOË^ϜӾhðe—ávQßÐí™»?ÏJs%Àe%ÞP£§»^$ †U:OÀFa>&’RÛø‰1šØ[„Kü{fP~È…ø”{söXµÊÎØùàXíxFLҚէ禋¹øq'^ŠêНÉNøºÞNM³ÊÐRºK‹·Ë³;RÝW&‘ž[ãqñv\ŒõA®üÛ%i¾ËïôêÅ2²6ß½ âªTûœ¢_2UWÓv?|Ü#Æ'ÚPVx˜u!2~º­ S†’\coøA±hgCÍÊÉ_4XˆMÌÂ2µú/­ ¿\ž˜MvàXZf\gÁ¬œÈ­Khòáʨ°H"¨¿J†qœrp犗÷<•¯1vm=Æ28ù©QÚx8R/‹ÝÒËA«Ð7X8µËÈe-w fcñÝ0¡.ya2t“ä(×îdñ&)Súuø¦qtwZ~aÐÍ|>àI`_½’}°Nps<šqA±þëeÖªò˜þèê⾡ü5Íäz¸Ë$Ç”ÉëœZ²•ûo#Z ÿz ÓCŽÀDp¦E|ÄpÈžý7Òº¯c––d™ê¯¾C88öŽR¡«#IWV‘xn]È׿«Óc\Œþ–MrMã“ÆÛÅW_)“.OÏrY»à¡,˱±bTˆ°ƒÚm¢›+ÑIõ~»/•ЦÈ7ÑYR‚ý|þအ÷{ÛAÎáw=œ„Ï<ä69|½ž! F{¶I #—ÇÑœ±KßÓ?Gc©ŠBO£Âs4›B…Éâ¶£5â—†&Á 1Ò4XÛ¹ô¹ã`ÙÀèÌAg˜t¥›žÚeK̵{¥­¡PUÀ唲S¼ž‹—´Ÿ†AÜ}«¾¯ÿÄ!e0mcßÌBís°ëà1¿¡þML|Ày€Õ­”®d9Åí%¹ø§šçRLšzá6©Å_ØÜ•&R«ÏB’Lœ†¤Fbäw½)®×~£L¡r1ÖZ…lö8aFUs'ã™´ôô¹¾¤*ý`·…ôf›±ãpìÝÉ­‹%WWç^|ïldrœ¨zÚH(£¼b:ô ,Å]v гóOW\K’ M,„â šÃÉ)fã¾j.¢s®:¤Q&´Jê?øˆø_Ôe{1äüxXð<Ðî_#U¥ýHþÌN"«FÛã4ÓÒÛúåá+›²Fþ¹ºõÆnü*¿‹ßõ<Ì<´JDè6¶ê,L ¢LùÄe§³x]@CÛÿG7£T endstream endobj 62 0 obj << /Length1 1570 /Length2 8424 /Length3 0 /Length 9451 /Filter /FlateDecode >> stream xÚ¶Pœ[-Œ»C xãîÜÝÝ]h¤‡î$,8 žà„ Á%¸w—GrïÌÜ™ÿ¯z¯ºªû[[Ö·÷>kŸj:* m6IˆPö`ãbçH«jqñ89yØ99¹Ñèèt@NÀ¿Íhtz@7w,ôi7 ¥Ç³MÆÒã9N(y:¸x\¯„¸ø…89Üœœ‚ÿ „¸ d,½@6Uv€ tG£“†¸@Ý@vöϯù×#€Ñš À%(ÈÏú' é tY[‚ª–ö@çç7Z[:´!Ö  ô¿(Eì=<\„88¼½½Ù-ÝÙ!nvbL¬o‡=@ ètóÚ~7 P³tþÕ;@Çäþ—]bëámé<œ@Ö@°ûs†'Øèx~9@[Q îÿ¬òW+àïٸعþM÷wöo"øO²¥µ5ÄÙÅ í¶ ' @]N…ÝÃǃ` ¶ùhéäyηô²9YZ=ü©Ü '© °|nðïöÜ­Ý@.îìî §ß-rü¦yž²,ØFâì {¸£ý®Oä´~;”㯓uC¼Á¾[ØÆöw6ž.º`«'PQæïgÚlv@'''¿ è úXÛsü¦×ºÿ8¹~›Ÿ;ð÷u¸lŸ›úƒlÏ?h¾î–^@€‡›'Ðß÷ŸŽÿFh\\µÀ h£ý‡ýÙ ´ý ?¾È`Ìù¬=.çïÏ¿ŸLŸåe;Aÿþç|9 %e4 Xþêøß>))ˆÀ—‡ÀÆÍÇ äåðó üÿ›EÃôwœÿIUÛB‚û<¥ìõ÷ù3þ½L€ÿæRƒ<‹`üÆM8ù8­Ÿ¿¸þŸ•þ'åÿOà¿Yþoÿß‚ä<œþ¸ÿøÿ?nKgôï€gÍzz<ë_ò¼àÿ Õþµ³ª@§óÿz=,Ÿ÷@lçôï1‚Üå@>@ ‡µý_bùË®û{Éœ@` ÄôûV°qqrþïy³¬Ÿo÷gEþqŸç¿_) ¶†ØüÞ0n¾WK77K(糸ùø¾\Ï«hôù£a;âñœxnÏ` qCû}¢üÜ…ߦ?HÀaù$à°þ7ú]‡Í? €øøLdûÈà°ûäp€þNÿ\ÏÌàÀgfÈ? /€Ãíð™Êýð€Ããð¹Ï?ð¿Ædíéæö|Qüòó ÿ…ÿÜJ@ Ðmnb-æPÖv]%IêͶ9":A·©ŸÎÄæ;çÖîy‹…œÊôù}Ȳۥdê÷nœ…uYÆ ‰yÊß½æ:äÈ–·š­w~÷æIZc›­h³?ûF‹ö$k{ÉQÉØt$¶ü\ýô‚á›a¿(Ñå¹z `ià_{÷ÈûÔö–ýŠ˜ÞÔÜúüJý¾lœ-^7Î$¸x’.ß*{ê%5’9 3Þ±öäÅå^îè¥R šÿ~<Ï_£î77S¯+t¸Ý;‰i‰^’Ã_à ÑûJí¼S"šñ-ù‹ÓÄ>È]ð ”mº6«DÅh¶ôÕ×aD¹ Yfÿr%†ÞÑ>¸uä„rüZ®Mº%vK?ÜPÞ§K´ÅY“ ’ò—ÝÇèò‹ßõ”çÒXq&!È*sˆ³c©9æÍ ·44F!=£œ²DVOíÙ¢®Žc»ÙÞh·¯õR†¸†X»Ëâ©òóy8ˆ?52ø;°³º¶ÒÄ,bÍQNÛZ‰ñâ…±œCD®ã¦åtT×îFõ>”¨*À­ÒlÌŒAäÛS¬Oâ¨tª VÓ®üí]³ðˆö1 “.|+8$}£®Å®#´ÂnÞuÓ¯ëOí^ðï\2û¬Ù·-óFÖ“+z¿sÁHÉ‘˜>6äÑ#ÚñÊü6öéA‘‚ý>üòc1J·ÆTUîkëýòî·ýޏÐVÚ)J«o(Kì:¶H¤¯nqmnɵa$D˜Ê¦ï YÃÈ:µãÞ^Õ±äÚYŽÌkR¿9v–Êf $›X­Å&©è6—„AÆïõù, "7a—µÛ?¯;ÜΈÖp߆w¶«þÜö¾ïpÞGV;!#å¼Ó¿€Ö.Q¶Y <‹•µfùñ% ™Õ°û=7¤*†®Í¬k„VqÍ¿º?O÷A\Ûõ–Ý6E»Æ>=Ä}4¶) ø4Fôu©)e%üâ”ÓY’IúÇ·I¾1òýW17Ø¢^slbß1 ð¾ÝãH4!ÇZãd˜›*‘,¿|Ó^'ŠÊ‹–w2ë :ž¡PÀ,¾Èä&T늴¼µ±=ŠŒ)p™=%³„‚3Êæ 5KÕl!HYw&æ òN”Ç’ \ÁjßÉFó#ê4[xq!÷Ú\Ñ;o`Ö6Hânk%(°òžN$-Aÿ³\»i„bÕèy/éÍ+ܱØwoªÑ…—xêÌ‚ªãÔpÕù^RøqŸzWŘl‹ßMýmÆ¢¥ækI@íà\ž”i|á”8O2Iµ0mÿá¥/ÍŠ`[ñ|Ò,éF+Ì•«Ìçx$Q, e„ß›®›Å÷D‰9¼Ñ­õ=/ʇ˜'|[ô|ýúˆ y¶k-G(7°=dh?Ñ‚Œ—û£°Æ½L#OÖ‰„Ÿ‡B8"É8ÙSÇ.]‰œ~[U™C¬QYJã'Çì²j6£ç=i. R9u1tô3ÓêÆÃqÎwƒóóõ‚ZÓøkæ3]±ISò’í—‡XØI…¾·¦–xëêt/õ…®zPúMŽØ÷³ 3³Ž=Ì bØCùÊ­yè>àÚg.ñ;°ªûLD6^,¤ëxF5Íw(ÕÔLõg83PÔý“à@T‚”<ÚUúÐ7Þi½tY%UÝ öJ¹—-ôMíëGFÚ´¹Ö!²MÆUÌ­a{BþeU¾b¥+Q¬"?üÜ-ç(Œq¥EÎ7=¯P±câ—xA¸A¾Ú¥þú-h“¹ /Tÿ=§ò L„NF9zí|˜üÁ³©Ð£´ Çk$þ™¯“lù;áëÕm³¤˜ªáuǤmeØk1õ9}—Ö¾ÉO~< OÙA­Ç—ª¾Š—Â×"v¡±o]DÂÅÚý²¦l¾yñEõürÎOW&f)£|*–yø€¹wî¸k\Ãǹ#§Êb÷Åj¬À(‡Á€ñ þv­0)O ïkîà«ý½Øj@K¨Ù•½ h÷´®ë ð=¥Žp‰ñN!WÅÔº¤b‡ò|M®“ª;„¥«V½ øÁIÖêQ}ßc@áàÐ|œà )’IqciÖ<ºËÈnE’_y’ÝkxþÄ`;y…ʨQiÆà7swóuáìlù "+Ä(ÃÄØÑ­2JKsFV³µò“>Dêô­‡ð!áß{ ûÖEWpûîg$w1¡ÈMdÀŸzfrΠѢ…‡a”Á··×HqÅ™ŽCì—Gw›29K^.òJˆñXŸ¶>- Ô Öm¢8Rw_ÛÉMçµZà{¿ngkzJ5a€/dNZÑ{ £õeç.ïˆá—'B¼ænïnÄ“òKÕÚ÷6hVCí]‚Õ°Mžo£Ýu ïí W)œÃMm»E>aðàáWþ¸(»Hˆ},l¾Ìñl6Õ‚œMZ#o/eÒ»™"ù`A‹Ï!öºÁŠ.çÁa[‘‰5š´ Ùá½[ïî`ŸŸâ.b’Я,dPÓ-_§ˆ9›Ogá•e*cfÒH›3.´uªcžÖ^„9»ùæÝöÞ0™0ÙISÊüddÁpƒî¸²Tà,NT”îLÚ!ÒÊö‹-ïÇWœ‰\ä¦4Î-Ôúçª$*‚qYö¹‚$ ±0$û¼Ba8zGUÓš/MÍ8@Ôú”…ÇgÅhÝùYÊ)ëý¢úòv.‘ëvŸ†Mlƒ'éÌÜ&ÕÃH•uêN±ÒôÅ|x9ÐP²ª7Ÿ¹áXˆkQJ;å5„héK¸ƒU‹ÖXë¨E^Z9õˆ¿éËDJÜä°Ÿ¢Nm†9nhîÁñ }lxž{Ó£Í1èT³X{úà î„näH©Zzòëâ©¢¿JgB´¶qGå­½ræòÎÛ^F•òÔÝÊØ4ü\öâ j"~Î\§4½Ñ¨–ñ â«Ï@—àןzé>¿°›RçƒûÝÕ°« Ž’Q -õ”ãñï£"7ZœþÆ‹üYùdŸWMšë¼Å=©gÞ÷FÉ•WÇM’±AZLį E/Í·xlfõÛìLmoEMì4ÞU¤4ž@™öëqˆP˜1¢Ä;ì¾Ðýxû3–:ú0€ã4ò³òµ ~úˆÐ{±uû‘,Òfšh/ÍÆèI˜eåÓг4° Óœ&GÌÓ?ApAÐ~µTP3/çùÿ…‚PmFb8ÊÏ䂬ӽlQP:ìǃ«ëu?€ÝðJB[›9¡ª[ ë”ñ#$ð€|Õ]À©°éŸì*G~¼– @’¨¹ÀÛxž(%~éõS7RC/ä`(íæʾÙ‡3iK¢W8E‚Ô´ÂUŒR0G·J‘«Ÿä…çËé‚f3E+‘IýFˆ¹W=ó—p¾ämcHw—Z¶&©ãJùn„A.JSË¥!T<{¥dp®Cµìž€Ø‹m}Þì³"ðÃ8)2|pcš¸xF¼S¹¼1uö°´0mg`~ R™â>c½æø<´äq¾†ˆ%ò³NÛÑ›¦óLë!™HAÅV)”‰‚+cé¤+œØ«PóË©éŽ5îT±¡‡Ðãy¼WµÙY%p¸¾C!œá­Ù¦ôµÄïRôÂ*FÆ; mˆÕRƒP½0Ó½ÄG+u èòñXÈRº6»{¿ ÄÊ$²m‚u¹ärUÐObFûKåÔÄw¿}g°'7‡o0žò¼±öÖ~í0lÕ²c–'¤< }ѸŽmAе´§;VêÖI5S"|Ã¥à´p·$¼ùìÒÚØèR¿P‘1i‘•D1W¹'Š0¦¼¾Ú$…w–\ FTí™ØÊ<ü`ƒ:ÝóÊlߊp¸cPU¼¡y;¶ãDZS!« 4¯)Γ®|ÍeûÌÈ µ°ô’mÿΓfšßýÞCùk! |ÚM“Ü“"ÅÝ»‰¹Ÿª°Tºôï½ÞŽÃäÛ5Ø/J—ñRŠ#S¬>U6 ãHÖ‰õL¥¡i½(Í èlOˆÖnUí¬n¹Ø ›ÇÚADTR H#è!‘W×åu°©l«µø.C¶i¢± êj,çYJÓhHëWËçO»‚ã¶+/ë­Ôªá«vr}GÒBÙsw&>593¹ ç _¥X{Dj^pޤGMÿá¹û “ÊÅ\<§ËÚË ˆu8À!´”Gùá›ÈǨi|V"òAÓ]¿Œs1©ÊPœmÙPïåÑŸˆ´OKº‡ œÐÇ  įùNñHù¦ä„X>: hC­WN÷]°OFb KÂF5QÆaXï÷¾0q\ç– ?†zj€*øvSUiýºnxîÖÞAE—ÏûÛò§î½ŒÞʪÃ#µD·[…U|ù Û6î×™Á÷Ú'WÁǺ•š} )£w5³¸¦uD¹LS%ÃÙZ†6fç¹Úû’1$iñú浌`FYòØz9‘>Õ¤Õ ¿6t~Ï[+ý’ëõâ§Þà˜âƒŸ1û˜Mì)“WŸå¥óz Sø5›Y8Uõ"ïÑ¢5)ˆpqgêÒo Ü©9`1Ö:C5UΆmfdš9â&ˆ É>?átQBëy uoTa0Ò`²÷'~z’Ôõ¨{аg¬4çS1*Z+‹#Ñ’ØëÂn§Ñ]¾f%ø,¯ù”kÇ;{c8”¬Ó²NáØ××CRzÄ u¦F× x\Öœ†R ìaÇPIÕnÔêÙà^„{MGÜpŒ@A•´ï÷[’or%/•¼.X÷2µã´æS>Ío´_sômÐz÷¦ÉÆ—Å®b9ÉÃôí]d¤¬ÙÇSÇn'×|Ñû¤j<¥«+|Q¦ñŠ u½¹0†Øk{e ‘7™²Ã…ÚS™lï ĹÈÄ·R,°8Ê—¦î *X¿ÎË~‡Gº>¬}yVÆÇPB8rœ 讣ÌÎÔËäY§O8V Ÿ ØYœ;¸—ŽØoÒ©¨kz$¢k¢_¦ÿ××!e­x’Ù!iðšh¥¨ÒÒ¬ÑüêEÛS?>)”*ýjDÒò&ÓìÛy§1ºÅ2ˆBj>ér´ÎJ¡CQÒ—¼œ@ëÚ/‹Ñ7ŽtÇšÃd%l9ÁpAü2®F <èKÛÒÈüýé×ú. šó1ÅÅ‚}¢Õ}ÆGí=ö¬×Š0O¢ïZÍ,Ý^7Y`½£xx•Wþ*‰l,}Ô³ÜIhÞÒÇBV+z¥Uÿ€ìLÔ–»‚Ss–Ì3¬¬°ú:៛n•‹HPtØ;ð¬¨ÀªÀÛõ§5¤ Õ¼–³‹œ< °RB/Õ`$ÒÀÅý&¸ÊþQ§ù³g,a>WŸ¯Àh-¯MüsÖ Ô7êN妧šõ™ºyDăˆw{F#T‰“uKY­™87¢Xr{e^ôE/Ÿçg¼Xùœq* þ¾é#cQçªém[âò®]®ìgÓM £R¼¼ÿnÑC) *‹û¦ƒÂÛ§`uˆóG¯è3l‡èº°Ð௜Îó‡• pæùwD $¬ˆ•Ý/c˜Šws5õƒK¾H:×qÔZÞqOÊ$»X<Å™:½ò~2Yhxœð è/IFTr ~Åκõ Äöós._ª¹¹µ·9—X*»¼º4„L¢ëHãÒG奕 ÅE †%í5øf¬ÇJ¶!i;ë×n¥™òQÇ>¸ÉzÍmFÑÜ2’ÍÛŸä¿ïÑétÒÁQ2hhÌh'WŠH»å•Š:qí/eI‡:Ú÷Ï3‡ä|v¥Ò|qäá*¢ÑÚéy‚~¡´ÞýœÜ`ÍØa/=¼Y~-›Œ‰Ý›ÄdÉ™Ê hÂ~€j v ßž`Óºl ¢Û·ôy¾¥9]êNòg&‡à8©N.EšÎ)$Ò6;õ¹ÊT­§ ÷åH#³VÛ^™½.Œ—yy#CÙè¢\D”ö‰]§ÛѬ‚“+ARG>8úaÌwko»–~\£élÚš²:Œg!Øß yN©"AZNÃ>¶b¹ÛRðÞšªö Bþ1gÙ ]š@av:…ËÙ}²*õÁ¼¶¾ëËSçýªô÷4c³IFá¯5$ËP"r7ûLê‚ÝbM mWqw,ø]ê'ßš½C6ý<š&Yl¤+gãŠx3´©á¨qºû-Ób[*‰w¥Ñ/ÃÒ=—_(;x?ü||ë™#;éy¡8F1;=ÿ‚R9âí(â`$¹=¢•p0¬»ć™|®2äñsœCðG!ÌSæ7´D—e `”åÚË%ôçoe*B¡Uð– »ÈMÖm[Îøa§§ªB~WNf6}åEµå {“tg:ÜÂ¥QªXª™2êùJvŽrž2¹bþ*´ôãšÚ y·j˜~¬¦³N[ñ„þ-MhH‡3eß™xÂÞ†üˆÜº†:];¥ÏU.›< vÓÆŠãíy£2ÚtÑ¿(jV­Á3ìú9XŽtp>OMôúƑܯݤ—ÒVhŸc¹n6Ý”å¢blÚ÷íäªj*6ë ±~ÊD'5 Œú Ê%#i¶.äEw¨|̘¹œÿ†c}c”«x ±iÅ9§âZ¿ÔLܹ’¢ŠÅ¤zϽ«P1l0Û²#á3Uo]oX„°°ms¾u'9ÊÓz™»»;Ó6=¤_è?Ãf„tß້3…¼?DŒÂMo‹ÊHŠ9·9§CFÛr¯}TÄœ˜Ÿ®;ÛK»U#4lü(Ä3·íÖ¼° }~Œè #¦òA©eK¨Ý¯¢d›ôN ³0ÅÚwß?м§©À€&„kÈyvT-§á»tÛ®–4jËÕ­ú%æš3-’x×Ö§ÎíæÑïÝ¡¿x\Û¥Íö%¢ý™HøÃRpµEC¿‹ßéÚ×_=Ò±"ÇÏM» áúÉED°)‘ç=Ad†òjå§¾{Õ J3†½ž¥_´n‘ç¯TÔÝH¸ë ˆŽG¶éM9ùÌúù}Eý¨ÝãîÔû¸7ÄÜI[Ê3Ò†–*(QªÖ’ª±Zt“é0å¥_ ¤†w <¨¯rî|ÍÆM2¹ x/¤Täs ¯av»½àízÝOMÅ?ÄU jÌŠi&aÈ®J^‹¯{/òoÑ.fc\¹öÀÏÁT‚—»÷Þ ”qz¨åÍj¬&ºdi8pN´ä"ÌËÔœz"À6B±¡yân׊8¨ÅúÁ¯}»_û>Õ Ö5†A\2z~þ„!ͪŽ/è"E¿õ™•ÛÄù‰]Û/%&׬€6»CúIJÆLD6ý~|¬o©Gìhßšœ…’þ!cÅRA¦ ª‹9Ž/?Ì+éÛ7ÁhúkÊài™É˜ïîéú×Ülñ>ºÂÆu;3Ø;ª;±µ§Š5÷ù"•ÛàK·áÙ)Ç†Í û4¬"¨,Êc â¢üÍsYŸ„áX­]º¶n1}ƒQCl=v¬â¾üSÔ‡‹n÷9™„ß²H “Ó)ÏVf9û$"v™$õâ²¥ É‘3Ì0²ÓJöý’MU$¥,)Éò5Iç¾Ñ‰]pöò‡·"…ÖÆ÷ýGøÑÃk`‹‚—Ñ#X†¡ºÍVàòïÖh6©k4„öº+[ײk|»âÅbQñ§)wÖ9;²-'‹ÊiG8Œà°F°ã¬CŠO YÊ´ok`9±}Ç0îý„YXŒ·!ÛmÒ»)3•wÎË}YÖ¥m·i\¡ ·º´œKQ§ØüE¼â§Ê%Š,m*ÖH4G‰ÕÂ~_Æ¡½/*ÛWù{ˆ5<(„©íÙ{<®K¥EU"¶Š¿sc&züŒnCAØ®˜#e98# l_7%“{¦Ýh¿!ÊRõÞÌ?˜FõÄM:g£»@’uÙJížyA÷Ο•ä¹VôÇÁÌ­ÄŠåýhÏ™²{,Ê„+±ÃÓJÌmÍ„‹.S¿µ’§œxw'¬ë»ß>^‹Z\]1ÉY)éî5_¡#ºõ ×póJ½Œê·ž’S-0jÜÌ%|;yÕ›)fßÈbRŽn-uRÇïH*êPû†ª—.}@šÑ–•ClD !=c¶ÑÿJçE;}©ÒãuMãË vþ¡˜8G¾U}7¤ÒÆà²+e2Ÿa·dISÒHéWZQllÞÆÌЋLIã2êîSž)¯(E"ÊôêïZ²l&±ÃÉœãÐ ™зNÇóqpQ!nç=â$ÑɽPoœ“›kD€ ì®?¸*%Þ®€Úv…ÄÝ;›ÛbeaºQ÷;ðŸ°x¡•ÒjÏÜõ^Kq(˜Q,ÝQKœf^ø·•?ìPf™p³È*pðäËÁPëøH¯ðãá<´á•ÔäÝ õ˜ÍO÷É!!mo\³)1 b©Úçß |MÂÖ¿ùŠð¼i©“›a '³“$·Ç@ö~‚£ «­*«ª:Ï=l3¼ßZœ©µkÞPß™oËÊ¿7,Ÿ±¤ÙÞÙÁ,ª(bÊñ.ìó»…âOœhù2®ÇBb=ˆ}æ° #5_vð{[óSü¤¾Eb³‘7UÓz0\8.5·Ø…ÅáŽSÇe•1ð•Þþì¾Á´,?ï”b”n£aZПŸawêé[ÅÖZ1'>ÐÝvÛ8¾w6%Ô²p2. \sxGqç—Y9Skø£†ˆ f6¸Xˆ_Ô6˜.€ 6X?æVè£nm‘X6ŒA®É;!b}P¢}AÇÝ,Æ};zÂnó£PÕÒø‡½(^gÄo‹)µ>0²5Çñ nsüiŒ-KÉÎæÆáœ.„#¾½ôÅ‹U¯ÔlâI+^ÇO £ 'ß«=jÞ|ß:Bôa\ãáp™B,Sï…b½­Hö?–Ä/¯;¨‚hçsˆY÷?¯¾eD~á-˜?ùG¥jÄuŠÙœF v_}â2]ÙYʃ UH“¥ùÀr3zirHI8t†x“rª³ÎöH«ƒÐ3/¦Z7îNv~hõàÙ½ÅÒßàb‹™®Îô.¦ç!º<öaã: Å_Ä â‹ (dG$Ð%ƒãMyÿçõ½)ƒ ‹ñ~µqøqv˜1˜ +˜ Ð¸,zÃX„nŽZ°;¡_ñMÁ×ê8‡ÌKÔËw÷HÏõÉ©ºú^=ñ%}€<èñÛ’gfËÓfx¼œôËŠ†@D½E…ÈÂ>{\;Š0]‚/ÎÓê)€D’'IµøC‘¯ðm²Nº5Wa§þ‡ói¡á7f¤䊓^ ¥Š§빨? §5ŠˆµîVPï81Åv*¸â_ÃdÒl‡»R&û¯ïTñò,^íMEÔêRlj VU‰l¯S‹ÓDPe‹k¶ç}&‚Ó£”#ÁP® (‘¾^ŸB;@ëƒJ©‘{£âœúâÍ/¤â1‰T5sª‹*Cª…Éú& U¨ôýûyûÆVÚ ŒÇtØÖ)/1ÜSÊKNs£+]é1:yªÊß•YHç´=€>A”rEm+*t?žTþšÞô¹ Ì™m®‹K8§‘e 8lû5óí±º²lÞ*e¹„¦£îTU_e=ä&£ï{ÂÒ×™ùþ™02ÙPù7³Á Éûý­Òm‹33!&#îÈ6™ˆB¸Vÿjé,w,¥N¯[TÕi¢u(’.Ѧ4-ï ýïµoRÿôü¦ó¿èúg;ËsT4˜~ÞQJže´nI F˜ÆÿÈ]^K¯ù|–É`Ào@k,àûá—æfoF÷Õ©N.bP-õç­ L”ë©­¼óãp£bœ’£—ßžâM(^²@3="õÇ„y^%ãëÿ<³SÎ endstream endobj 64 0 obj << /Length1 1737 /Length2 10797 /Length3 0 /Length 11901 /Filter /FlateDecode >> stream xÚ÷PÚÒ  N. 0Xpwwîî2À 3¸»wwwww—à‚C€à!8#÷žs¿ÿ¯z¯¦jf¯öÞ½zW ‰¢ ½°)Ä(;Ò330ñDåTd™™LL¬ LL,Hª GkàäHê@{Ìó/ Q{ ‘ã«LÌÈñÕPH;Y˜YÌ<Ìœ GG[FF#ˆ½¹5ÀähP:í¦€?ZÈÙÿn‰ jrøK¡1st1²^Ö  ØáÕÅ l ´¼f¨HÉl࿌eÿ2 ü}9fæÿ†ûÛû@ ðŸÎF&&[#°l0Y ² Ž®Žt#°é†FÖW#g#µ‘ñ«ÁŸ¥$„•F¯þÝŸƒ‰=ÈÖÑÁdýGŒ„y½fq°©(ÄÆvt@ú£>1=ÐäõÞÝÿ®âöø2MÍþhÃÔÉ–Q ²sJ‰ýmó*BúGft°311qrs€v «‰ã TÝl*™ÿ¿öàåa ±˜½¶ô™_<ŒœG{' —Ç¿ÿ‹˜™¦ G€1ÐFú'ú«hö~¿=È ÃôJ?fÓŸÿžô^f [»ýcþ爕5%´4´hÿnù¿Jˆ+Àƒžƒ@ÏÂÎ`ffap¾¼þ7Ž¢èï:þå+6ƒ¸ÿ*÷õžþS²óß ú{A¨ÿKòÊ\ €ê¢ë2±3™¼~1ÿ¦ûŸ.ÿÿXþG”ÿW¢ÿߊ$œ¬­ÿÔSýeðÿ£7²Y»ýmñÊ\'Ç×-ƒ¼îøÿšjÿZ]9 )ÈÉæÿj¥^·AlþÊhzn6Ž¿Ä  +ÐTähbñiþ’«ý±nÖ 0PâúãÐ331ýÝ뎙X½>"¯ÌüS|]¡ÿM+6˜þ±k,ì#{{#7¤×Q¿"v€óëRš]ÿä2€‘ q|u¼¶è0ƒØ#ý1W.£Ø¢?Ë+™þA¬F•€Qý¿ˆ›Àhôzbüâ0šüýÑ£é¿ 3€ø/øšÒì_ð5§ù!;€ÑÂÍÖâõ-úÇâUú|Mmý/øšÛæøºŒÿò}]@FÈ¿àknÛ’½6hûÊKÈ¿Šef0Úÿ ¾ævø|õpü|½§Á×Êœÿ,¯¹Ýþ„ÿ3@'{û×ÇìÏE{îðŸ/'è 4AZY„˜ðZÖvÜÖp¡ß›baCº »FÐØÇw>Pòå»»bÍ_î™ß¿ÛÄ8f)yi¶ÒÄa²Òÿî™΃³ÒÐñÖÌx‡ÎsƒÿþØAÖ‰e_¦»[ÛljŽsÖ¢<âr-4GKgw±·ñC’ïSKÕïJ‹3xʦè©Èpx<..VèÀNáŒÎ*‘Zá¹ù.)r‡K¡©Ÿñ„?îRÇYØ·‡VËè @[©ŽŽPme®3ŽRû8d §s»}âx1Tþ×~TÆ*‚d¼™R»Ó—Ÿ/O35Þ¿ÌD‡ÜÉÒ½Gnà¦Ô/¶k*È`õ°îÊ ÆÊY…ŸýÆþÌÖj£l¸‹ p£¦Õ!ôÇòÇ—+G½fOÖQä³¶A\mµ€jfJ ÿ¦Ùµ Å©Ù)a¹ÊPrã"# jP1ór†_åh©îq+xSuãè¬%ç:'4Ð#|éµU@GðA5¢¦Ë7%£¢³B“›DH ]›Q"é·WÕ., »5Õëu£ÐÏ6œÑ-mh#†#ý⊊10äÝîoÐ2Ù¶ª4%´Â0Tûé-ÌnE©5ÝwqÅþôÂ9‚zy¦š$?=4ú£{ÃÃîûq\ä»›~­$|Só¨ÄÅÄíõægDvŸ˜´ièžíxo좫àÕt¥§ïï  ÑFÊðoHÏøiç ætJN’³<#û¹«/w ·œv Ê´‰°Â7÷`3¾P>#S^þŽ&ÿ®rÎlóËé8šO ìUe0‹ Â(Í|#Ì3ŠÍ¨ÑnI&Á­È‚ùrõ0¸ÊÝÒy\:ËL† GÑŒV#=ow4w“({$ifÔyJ{L˜ßþq®ó¤ ³÷±ÀI@Åà¯ü¥S¦µ§‘¦by\ÒÇ’ o~[bæ­óp›ª§´^CG½Æ;F|£ëg£öŽkýÛ"î­^Jƒè2.õ±ÉFbøˆÈçr’®T¸3xë¶ÕúÅNèCåÊ&_LõeºÏ/ñ|Ýé°àMÔ4%lc†‹‹'æ(X­(Cmâ+jœ¾°ÁÝìWõ®'zÊŸºxwîi+¶´‡݇|V­¼ŸåÈ$Ó¨Ò0üïž ˆöøank41ú¤w¸çŸÎa´~ž1TžåIÜ;esYɰ¸¨¢7ÌX¢Ÿ9u×øiP¥£‡¥h5>MÔ¼Î÷ã]‹]IlT+š`9UéæRA@*ÔF9c~·Za\Kiòû.aéÔEÔ5Fóü†nîbЃn©lfÖ (FÜcû1£EqÜ¿‚6_¹½üX&A†&NÓF«Dø`+"üWÎ#”>H˜fAø«7γIa†ã‰ÎέýãdyúuÒs™È÷-ü"|Éð÷dURw4øv|GÀy{¼UJ—ç@L:ú‡¨"ì–÷ #Ø#zT,¤V4Km<mí™Æ_ÖväØÐ)PX–ùÚl¿Æ¨E¢.¤…ïßýpX¢~^KOb†çõD’:·Vfx'cÓ¸®—0é1(VޝKù…ÕN)£`è<"Ý¡}o„:*… +´¸ Ö³Éõ˜xyò­Šf²õ;Õ >#"ŸF&[‘W¶Ž+Cfn©%s¦n„I±¯`e¡hËoñRY•@¨]Hõcd°+ìûíÏ~Û‹I¶»èSà›WÅe€«§0iïwŸ•f•øO¢mV•Çá-·Ô¿ê…"XòýÜ$Éøô£ß/‰Ÿ6Ãkï/õ“òqêïì5 À„‘DÁZ-eÈNnƒO—ù{FÜòºØö x¶7å 6Íåw-úðÌÔÓƒbæRï•õŽòõST1Mµ\i…ÖýKí9‰Zu_¥›ú¤í*Q—*Ù9ßÓV{¦/ÒLš›Îú”švá°LÓYÁï—ZÕÓÜñQµ yr'äÏ|OùB±d߃”þ¡Ù%‡?F$æ6™]ĪÚÃoí@Ô 1<ÙøÚÝ’L~;2ÍR½b1·ÐJÑú‘̹¶›sH ççÉMÖ‰B ­eGÝ2É’XéØSÒäí…wðÌÖÇ(­!Ÿ½€·Š¹0ÕDñ§ÚŽŽ¬Kõ’žõó‘¹4–†MOÅ~ðÕ r8Ag=#ŒÅö'-7Kî._×}îˆ)ªhývÿ–_¹©¾EJbŽ)— åš{.LÊÝ¥Àv ÷qJ߬x)€§zöguд»ïÆc&øMÅë¸,ò»ÝÊ ?©?Ö+®øŒ9_lzá—ãws (òû¢6ù]®~Ä•»†)u;–ˆ&£»î†°:Å•-q°_– Ò²…ßdœ\•w•§ò齋ûç‰?ÃÃVÂ/p´<¥Pd°PìÜðÂw<:´àO,—oš\%d‡ Ω­ÌÌÔ"e¢NRš?°<'Ç‘D^\‰sÓqÒÉ~&Èß³6Œ‘_ùx —§½'籤§[x­“³hïÿûæ@@&ŒÎSDš` ›’!`d×ô¥¯çM±ÅtzÀ”ɸ–ñã[<ê¯ßª]6Ƽ§§‚1êškMA­˜dÜͼì¡àGî}„â­¡B],žöÑ>>7«ÙÉÔ ‚êÛÛÀNÔ®·?º’Ö ÆUZä»ßYÌqñeºã¹ZAw Ö|8A€7S“ðOµÐ—ÏeÇ/¤ $ÎRÝLUüþ©®Ç¿*]qòeYËÔn áÈš€hHœsñA8¨L1!p€•Ž˜6ÆPÈ_]&ܪ^›*Up\lkڥϞÅlmð‡iùº~ãb™cáûî3” ¯ùã£Ü<Àäö|JÆ&-½c™…YæUíÇ8_Íy|FÄR²ÜÍ ç§}Ïââè”x†ºœu¨±‡ÖZ–ÕN@ÄÑ:FQn?o9ðÇ‹Ó5Ž‘† ‚8˜JoÜ´6‰d“­֋QŠ% ÷úpDÄöl¼òsн–´ûq i/‘àï+‘‡ðŠSЀ雏#v¥ÄjOÏ•y¼X§¦–¸ŽÓË´…R#ÍU?8:ßÓÞË?—÷ÀV&:±ï*u²P´'²”lNp±rQPeÿM3‡…VåïsÑÌ– ¥D'ö6P‚ØfŸ=¤7’.åc¥Äå¾Àµ)"cÈuP–Õ"ô9ƒÎx4[œ~úáp™F“F›Ìc"©àñ¶Mr¶zÈ@#2Â8GFEr~n…Þ;|YÞ½ÚD±ÙŒÞzBóEd®dR$Onú] .Õþp×”Q¸{>^Ä7÷‹+2Ðã~]~50ìïû²[ņe 4ĹCU‰‰¤±IýpñY¨Ÿ©^Ü'ÌŒRóœ®úç3§°']¯‡Ê:‰Ï½í)(ïíö©ƒ;¶ÅeBBg®“0îGø¦&Q$‰ÛØûE­ìñgÚLUµ% 2Ô¼29Æ×|…†‹¬3¿&“ÅÖa§þ~qÁ¹MëáVÆ `dƒ^Ýíz6݇¿ÖŸ^Æ,…F[Ö¤Cþ¢ð›ãqïå8å€r÷“ì÷€ó±:Ú^n'×Âü`K†¯«*ˆÕjÖÖÕ ¯óð…sï{x>üÔr°%nÌSúX†cC!Ëál‘oÍ‚m]þÔdáâ››ËÏâì7»ß1ÉH´®(†XuVü±Èbàɧ>»3òRñìhº¦¿þ!O¦Ä-ÛñndD³Ì|¹ËéÔ¨Š.\‰¶4®³#¶™ycÆâ²„ÔêØø˜‡ötî?ž€{49ÎËï£7sö¡ÝFHJëq2ƒY9dz7‹ä¸ºU8Ð40Söô‰>ã^ïr6_–¦!fççO)2Üpü³×ëü—‘Õ+Àv™÷¦*ÂÚð 4¢«à¢¸:î…²K÷~Äûã%æé±”b6Oì‹ØF)ø­¯É™Ö ”·>ã„´Ú3¼yoˬéC!û^I6/É× ¬þÚ|ødûrl« ×À—êµÎM7wc¹ó%7씯ãý×`öw™RúP_žÞª'—à¸bé&V ¯RÏšÆYИӭÙôŒÙbgboëѺ‚ôÚéî¨H÷á ›ô²‰l:ìBªXrQ¼¥HªVèú†¨1¦j'ñÒú&:âd¢< õ˜¥`Uàz,h iç¼8<7BP…­BžoÌ·ugïŽãª®D ªP²œ¾ôhõºc"J¹lƒè]ÚÍΑõ.PÚKP] =·Ñjà÷,—AGA+×fCtx‹’¿¶‚^Vl¥èWÜ5Ü´ÞmÔÓ߻꼔™c¡¥<¾ðÄPnœÜìžoš 4¥¿ÃÍ5fŸ®t¡Øº qžäDtwP£uo’ùDÄ lDÍüm`åXß¹”"ÇÜÂÉæúò5î<˺)ί·¨è{Ë.¼]rV˜/!<{tUÛrå…cc«fÓxG¤ÍÔ—O}òø[œ…1ƒõj„/)çÁ¨5âM©¶=n(/ù(ÌÓ†lµË¬„$¶ƒÛ„>YÛ™µq“™°g?°¤×h 7fpW“0föÔj—ÕÕõÜ÷RYcÏñ®lR+ ±Ü¯xºñT¯´‚ÃÍsôwÁʬœBmÃÑ“w{n$èèqyKfÖW„Eo²Þ¾·w# "@«z艞â<ÝÜœá ×|Odx¥ò‘gØÔA×Ñ•ï 窮äÙûb½á˜œ€ì 2zÞÔvlôe‹=²•™)²ÊÃË\XqvãcßL½1WØG<§¼©¤½_ƒ6ôh%5 ÖûI‰ÉïU³Å{êóÊ­$z%ùç¼$ÓÂÓLE6Sù”*·'áx>UÄ]×^î×üú)ÍýÜ,Ó<÷Y¢½{ªXjáü˜·¬ŽYͨBø&3`«ÖîÐ2ìI8Eƒ1  ÿ»µŸºMVT‡ to˜>žŽSá|D”NM°r\ω¨šŸ!9ÿÙ{ñé|ê‚/#„c“ÿS¿aCút¨&BÎ|¿³~ÍÆä[wrž/OŸ¨X»ûé1‘4§WqÚì0–à6:hqO)ÆõÃI”æ ù1ÎMíñ£X˜3^C±?×K 0«¥dŸ"æö/Ǒ͹‰®&9pmêý0#=鮄]/{Å3žeƒÂ>K6WLèêäùï¾µYÑlŘûÙbÁtRde¾L‹*îÚK=-¢= ÓõbìÁ9ž4;Q¼·ß˜²å\-ÅKNã9¶)¢¨JP&Q_ðy¨O-¼êa_ùl0x$úydV¹ˆ—-„jÄ f#$Fcİ4CIk‡‹­®Ô——ª.£Í¬ÜFÝ( tÏñ)÷7IP#à0Bxc6–?ýŒcÍc{Û6ëçx3˜ù°·ZŸ$—=÷xØöãBÅU=¸dq]uQfª§îWHk‘ïYÃU½ËÃû–pvËÛ#~ÿ¶CúmöoÓiüoàI›¢½‹roì x \%·iËðvZ÷¹¿H¤À”_m¨Csôñ˜šOÚ;g8q,„"èqo`EâÿÔcчºŽõ€ë+ZzÛà wȈ®GEz<¿öB/¥_ —Û…ëʯˆêb}³±á‰øR,— üшà/ô‚"Æj;0‚“„‡LbJÌàót?RU ûöùŠ ÝLGãœç«“œ%h7Q: nÙx‹«­^e¡°Ä„ú¥dõá.à”™•ïIi­›z·ñs‡‚¿5×§/=8ò5¢Rºä8-MìŸfD¾ÌkuÜéÐG?žHÄh 'oM+Fe#æ¶«ólúO 'µÚ"(äüNǬ›n#‰{®|O+q›ÇŠ}“®ï ˜æŒGI#Ê[,«éš^ÌThÃP¨™Äù;²(Sîuï즀ŧ³óãÞJ‰x€.‚…sPM´r—,peÇ5±Û£'<6dîr›mBeû\‘rT'GfUðx*D^^l±UŒVV+Ưèaé—ÀöðøofÔPFßô 0‚û®ñNøD`uêÆæë•¯Íäy$,E¯£ ë¦ã»G˜Äž¯ ˆNg·½(Æ•%¸t¶í~×Ð' ÒL¥B+úX÷ÃIø.X,põúC¯<Õ_ý«ƒ ‹ÖdSý¹h5‰?,læïϽäaŠæ ¾XèÌz¨@软yx{’Íî5z.ùà»,ø!çèsØ2…3«áßK·?³ܘ±Ò­ïùú£Ž£˜°}¾Êr£¿9A6ƒXò6Ÿ<óL´Dé+Û^AÕŒõ¶Ìþ„ª‚ç 5¡;=xuðrx¹–˜·ºÛ­¢Ðcç»í2l×>"ÍŸP´éç0Tlhîy1ï´nÖ½=íBàà _w?éyo\<ft$[N–íOGtl5{é°"ðÎà#mÜ!2QÍ}OB49kÍá¨Ý8üøƒ×ûàDR>lÊz”>\†PÕgoiœ”{K³™ßw™çúûZoƒ³h%ªþþsÞüªšL}œ´Â`|ƒî›Ó’ýEz¹lq#4òzØ¢IL‹y©wþŒŸ ôÀ™·Ñ]#!Éð±·ŠôQ³*&ñI¶Üh€àE[AÍñ¶ û**yÙåcG%Gž]no¡¿ã“NÞlìæm2Ò‡d^8•ÜG¹(á«-2XuðHŸM¥Ö¹Ã×Q_èóšý”_¸Í*C.±Óûf0êY6õÝ2VŸÇ î°¾_ø„Ó¸{7o†ßë– žêÑf\¦¥7ïÊ·sˆÒ+ïиù[^ÞÀîL¸¼ð1¡û ŸqóLmk›`é°_ÚDüìØÎJ¸Õ]a™_™ŠÕº@Žìª$Ò¢-ä$ »xî1w_®‚½Ù¸"ói,…_Ö,­¿û ÔCM 2÷ƒÞÇç¯<Žï`Òž´>ÿ­ñÛ8Ð)+zum@@ûUŒhMãM†nÐØuÔäPdyàq«T9tÈÔJ~±2àZw#P$ÜwÙÕ‹1j “Ó/VŒF7Cðrµ™ÖÅÊ$Ö[µìܶÅe _¿-U©.ù,RM.Ì#£¤Àµõ~ª WëMxSÓJ\ƒqJ€S‘'+xe'–P–GhE}ŸÈï8¾»^pä¯ÍbHЬ6M™>üv*wö8Ñ}÷‚fõ»ÁjŪ’1sïL¥?)^n_éý‘CŽR½ *yÒ dOLKµ¼I5cyà‡²pÐ%$퇿áÚ¦Cè²c]w¢3u@µå W=Ÿ(§.~Ÿ·¥+SóØº”IýU Ù^ÉÔAö\‹”`e'î÷ÁE[kœ C^Äã gÖ~7"³ÍØé6Ìø{éoätiH[ñLÒSöÐgªÏÃ|D*r‹B¡®#³Çlõ‰Ø•pYçoÍ­ñoßµ™kë~yKËí8ÇìªjÌŽ–jàE–À¥Ä´Á/‡ÞÕbf˜_÷ñ-¦¬èлÚ²¿[Å0’Ž>g÷I赯ÿ$›æéj™—¿wC <ÚÝR±ûà/íÕEGÈÕ÷é‡Z,ª“½d(èîDZa ø®|ÅP‡ëÚ$ÜÞêÇxßüv®ÅN=WR—`ôæü#wÄ™v‡ƒlÉÒß²a€q_ÁËö4 j•uh-Ê_ŒÖãÛË·“b´kæÒ5œÈ8Í„&Gªþ•ø¤O)ωHýõt{–FÊÃ˜Ü ½|êv"U¶EnÇå­>‘Á®HíAä;Å6|ûKÄ :oF. Ö:'J2ÍÉkjsà°e¨>‰|ßúzž¨r? 8m$#”Æ«€]Æ-$I´—•ÐjR˜ ˹G{-¶]÷>>-”Ç:XÚ÷ue«‡™ gogFÇ(®=FêìÝ\ß¾Àî:_bì³§ÐdÔâ"õd–§ˆû¨‡ͱ-,ŽSŽ •á”âèÉàLfïð³ßð(âæSaE‡T¥EQRÖÇs‰-—çÛ͘Ÿéöo8r@??TÃ8#{¦°/Ãè£$Ô}žL´çVÓâ&H"ß"3 ¹ÁãW¬z°h„-X˜KrEòï}¾Ž87®ï=ôþbÊY «CBØËtûA´MV&ËH,¯ƒBÌ(W˜}¼þ ª¨k_[óëj—»Û§*jÁŒ%…~¥'ÛG/¤Q¸‡ç[R+1øø]Ýdõ®F·ùlGÜóÛˆ˜}Ýóò#VFmáïHÍýš &ÜDªŠÑ?ÎW$Èk/Î7È^v¹F¦=o•{˜šÅ¾f6Æõö:.­I]O¡Â^šG³Ìš;¼“]A\¬Ìsoçélma†òõ€Fb“lyì°á³èU™1õEÜÅùsŽ>vãpíj˧=R,–—²ïáXZ|xn”¥øê³N…ŽneXšÄ’؉ú‘ Ø´ûˆCl6ßRƒ¡ïj`cpôhÊ»KY™lï"În(§ÿgÆm+Æå»Ûkí>›ê‚¬Ü·Pê~õÞ—*^º‹v} w°x¬(RDȸúÙc+EZîÙ€9Iþ¡«úØA2‰ìà~gmª%†þnòCÁ4¿ý€^4ó’tïáúϳçÆ>1h~=mü4·¦{ØÖö$¨7G&FÑùe^å¯3 E'MD^YîÜòµì‰F¼´‘ öAX¾vý¬êË| øm{1Ì–36ªÕ_X‡*pÃý.Ëæ¤ý “à2çvIº•Ãeë:õÍ@#ÀëA¾’JÆÊ±ï¢P\íåÙ‘ñÉÓS z‘´“0%ÛÓyï5ú±|½+à%qÑ89áÌXf ôƒä®Óýïlã¢îÉbHò½à Õƒ¾ åã¶Åý !l=‚RìöOúÚ^ý†w>xítûÅÖµv~%´÷©˜äNdlêp„d78NГ«•žIœ–²Ïù0Å4¹ï.ºÚaµ õ&;N­Ë‹0âO„éï3×,/Ùˆ·|§ZR$3šÙ%kÕôŸ/:’w“X$DÛ;~,ˆÑ1yxZØ0q ½q€ñÎÄS|P@Ž2‡§Äë}­š*bfê.ìþ O-{ó©Ð‚³ÐŠÉã üŸbT‡uçX*ŠtkNüÖ¢sœ—ú ?”úOâœsÍo½å¿½\‰–õÔ|†ÎÞ~™½¶oOÞµÀuUƳ=D‘;ÉZz‚7ùšß-² ·”5 5¯Ëël €j—rYAדÒP0B‰þ´Qt»YîOAš1wßڪƆÉ\sS5²âçÁÆ0³ $I,-ÚH´4Jë `Ÿ¼ïØx4™äØy°âZ³·”#h ¬5ŠGÁTÑ. ¥+ð…ÎéXdˆ2@DÇéù«žªÏ†pÚJd×)šüˆ2gšÆ³K®ÀΞ¯ß1ÉR'_GjX˜×û9ðÞe•[ 1˜©G¢ˆBÈ<íêKž6òü•ì£ .⯼Oòg«ùãMûOX ’Ýt1†ÔÏ n.^Bñ_aÍ%ZÊ—Èë(´¿rq¬ Á´ÍÀ<È¢êðá˜ñÛØ¼¶‰é‚ëi·ë Ý^þíòûu?ñÖŸ"ªÜiJšn÷´<ؘ¥·SÎó·|kY0×ÈÞi¹ë$“t…ÆKe5&+‚E»3Tgª‰–[X~ˆ"¸¸öK4íì%”Ó CçS2dŒ!¹f?—_¥Ÿd ¸èiÛpÑPmOØ28Šœø°Ò&mÒ=;›ÿŽUsx€w……5‹¯fÆHXùù"±›T ˲EDeí}îýڰ̆Ïú„2°nHa›3£H^.Õª*| š’P¤mË)è}ʽs0v„s#6ÌÇý)KÖÝ™*ŒdFõwNÉy]’AÅÁ¡—]Y P QMæ—iÅ2²ß’à,oÉÌÅ;adô«ƒû¨Tͦx¼A£[ˆ¥;‘²L´g4²¥,*K”óµµ$•³’èS¤–Úå…¤VR_Ö,trÔ”L5] „&Y¹ fÓh¬˜¶Åíp|J¹ÅkÒ¬M#°¥ûgœ>­¡ÏƒIß‚qJLWE碙|Oƒï<­1éY¾°çÏh½CÓ2yëøfÄwœáÅRÝ9?(¸÷DZMŒê¾‚?*âªßH9ý•3w sÊ÷|M½jšÛDº}wõ2ðë²É ¼ùìÛ"ÌzûK·}x}%bai}ÑhÖèrV¶% 'Û©H©á½ÛàžÕzØ=CcýȺ·(¤¥ÿ©êËðb‚Å2A>±w£9ß•)©$ª¼ß'qƒ_tpÇT{Šd³W®öL¦â±BgS”ÌÇ"6±šNš&ÆÔvÐWž±äQ,³M-tÍz_ÂïõÉãóü¦ÞÈ|%¡À¡ñLÒW—Ú&Û·ú²+,rñy—§ú:*”§¦ë:W¬g{û©þ4mBz퀸;ê¦sY'côÖ²'•ê$#yâøªpSŽ”#tqÂèž–\ŒF' ÀpÜôP°É•gÈíìy‹ä‹U`­›Äß =C9ì7L“¡èV´5ÒòQ8 Ô×9­ªqµñI$×Ô*hSe7ßÁµã“¤oñóć%ßI’²GP0~㑉a_ Õ¦k‚¢§I>øÁ>J‚‘sǬö:?ÙÑ©\ DÅ1~‘-]ȯϤ£2È~rm²øl›Ò;íAõ]V…SŒ?¸Ý3nTì¹­‡ù=i> #øÇ†›­Xõ§¥9è½g•ÜcS»A¨AScW--sa:»]7ÔÇu"ß2HZ_é€C¹:ʽ¥GÎKÑ“ŸO(ÁŸm=7ý€†QDw¦3ŽrOþcG1—UIôèÇ ÷󺨨n¤À0©‡Š=ޤlÛugÙÀË ø(Peô¼ vNðÛ›ûÝ 0¬„!Äшho· zÿM`vPÇvõm• ¯÷w'#QÜ–2totõòšÆ6)©pŠJÂé½Yù:^Çß îãXbph3u¿C`r|‰{6¿‡@)Ž×(/ã¸ci¾p!­À@/Qî+ x ––ß¿sß…Hª G·+Û& «$Ò ¡&“þ^å h=ëE²ŒN Ò,Ÿ­~ìÆ \ˆE«^¦.}訕WÔ©‹«W}wœ%Z>‹d'yÊÊy@=t|“_h.À5jP­Ì‹+RÚîj‘‡yÊA¦ÿ8tй}oßtP¢oÿ-áãï4´Ñ|¯Ïwh}_€ÇAE·¸ª`ít ãEQ}BtØTÉÍ;È&m£øìu×(ìKæ8Oñ‰œyÇ’Þ¹¨òÞF©@£h¡RÎ{ã%¬ tÃ7‹aðþõ±&R†w¼½‰¿Ñ Õ 1·Í¿#¹oÜTõ+ù>^&6QÊ|/†Øl3}loÀÒAÛ%uU°”èL¥®Sà {0å#Dåd?ÿ´¶R]ªUáÛ·êÝÜçÅ\ùÅ@ó)›ÓÀýž^)};t¯cüàätª(¡¼Îá¡Ô?C¼ºþÏ–\ý endstream endobj 66 0 obj << /Length1 2144 /Length2 12844 /Length3 0 /Length 14132 /Filter /FlateDecode >> stream xÚweT›k·-î^\ƒ—àîîî ,¸k)ZÜ¡¸–â.Å[Š»-îP ”âpÓ½¿³»¿sï;2F’¹|­g®çMè©5´Ù$­!– 9ˆ³S ­ª­¢£äprr³srr¡ÑÓë€=A4hôz 7w0ÄYè_6Òn ¨LÆÂjª q(y:€Ü Ÿ_ˆ“ÀÅÉ)ø?†7!€Œ…Ø ÊP‚8ƒÜÑè¥!.¾n`[;h¦ÿù `´bùYÿrH:ÜÀVÎU ;4£•…#@byøþWF;!ooov 'wvˆ›­+ÀìaйƒÜ¼@Ö€ßmÔ,œ@ÿ4ÇŽFб»ÿ­Ò†Øxx[¸P#Ø äìuòt¶¹ ùÚŠ*uó߯*°þ3øO¸ÿxÿvþËÙÂÊ âäbáì v¶Ø€Au9vV€…³õoC GwÔßÂËìha 5ø«x €œ¤&ÀÚã:t·r»x¸³»ƒwÉñ; tвÎÖÒ''³‡;ÚïúdÀn +èä}9þ9bgˆ·³ÿlv¶¶ùÝŠµ§ ‡®3ØÕ¤(ó+¨íÌäàåää€,Èò±²ãøDÇ×ô—ø[ í#Ðßâ°¶ Û€ hþî^ €‡›'(ÐÿߊÿFh@ Àlå°Ù‚ÑþD‡ŠA6c( ÜÀ>cN( Î߯¾™Byf qvôýcþ×AsÈ*Éê²üÓô?j))ˆÀŸ‹ÀÆÅÍ àòøùÿHÃüŸBþåªèlþ]/tPÿS³×ˆÀøŸ=aüw,5”À ã¾›pòrZA߀ÿ߬ÿËåÿEößQþ?øþ¿k’óttüË‚ñLþ/ '°£ïl öô€®ƒ*ºÎÿÛTô÷«‚¬ÁžNÿ[«èa] Ig[(µÙÙyøþƒÝåÀ> k °‡•Ý_4ù[¬û{íÁÎ ˆ;ø÷U`rrþ/t׬ ×‰;”©@ÐUúגּÎVëß;ÇÅ˰ps³ðEã„’Š‹—à„.§5Èç/>8Ø!P´Ã@€ Ä í÷Ñòñ8$‹þFüé?HÀ!ó BÉöâçpÈýA|•?Eõ‚FQûƒ QÔÿAÐ(€CëâphÿA<?Zµî?HšÏâ‚æ³üƒ€Päfaå‚^ú6äÜÿÈÿ¦ê? hVÿ hZè½æô'üïCâ°þ„&ýÒoäê åóhW6ÿ@^h^°è_z¨ÄöO¨µíïg ”„L UØý© Ú¼¯‹ô¢þc•ÿ¡3øS´bÇß$ú£‡6éôB¯ Žƒ^U?é ¶ÐGÚ¿ÔÐ]þ¨¡¾.ÐgŠó—øéº.Ðí‡ü 7´7WOˆÈÚò_CB¥ÿÚÞ¿ÊBÃxü BÏßó_Ú¼×È-Å÷OP¥Èíïöþk“¬<Ý õzüuéA×ìð_2Èd…¶´±~cßð¦ó¦N’Ì›mw‚ûŸî¯gO*´8GØmv¸~5b‡ 2žÂ+´\Q;±cxUx؈D2Ù8~oA©.vd¼áxp Ѥ'ì"jB?P§Î9?‘&ÅgäJû~Q8£®üeH/&=ާ!¦¸1Ú?pUG"£¬,®FU?X`$¼{ý]]Á1HSin¸[†ÓþudºROËô-ƒ’zËÉõ‹ê.<óù…ÏR˜¼Ásòh ?,É +ø…ºÇ{¸î캔%ºa{YKc ö/YK‡¯—Tà“ƒhéDÝø5 ?Æ<ûSËdäNǹ:®ãÔM]¼Å&ç˜ß-`dÚ£Úx‡—Ú)|÷Ëã¯8á›U`<ËÇ›éÏŒ™Óî"䫽!Œ(T¤‹”^Ô™´=T—lo=!ùõ옴+Ç—ôʬ‰Õða>uÐc m(DGI¯r÷ÄŠŽè¯.Ó7k®la\L~À2Ř=z ¼¹€$+§c,Ð l·ðIÝfX*ž²ú<ãejô£¯j¯Æ7¢qRuÈ’Taàw2ÝLé5ÉmúZ²šäé#h2ޤéÔ"ì>É\åÝù¬¨º‡ ‡|Õ¼õIx|z†10«ØnÔóüArŒÑ·A¬Ý*£ E|ïvÁz…á¥Õ¢1¾ ©õÚcô!ŽAX†€ùÄ$§YÅ2ç<¬%o×õcvZ†¤ÓØR¹Zø³ÂIÂ*L¬fvbDä½™˜¾G:ÔuU>&_š¨ºÂؘK©›ÈW¢švqqú°°†5„ÐH§>G{Fcùð_ü”R’¦ðFµÒŸù í›ä<µ@b)®N»®³„w¶½§Xذ<+å(øÖdmk4_Ý©q°µ{u QïgŽÖm¢éD4 ǪW÷ŠÃtvnˆ„žX›ÝzÎÑ}P….³»;›‹ó錟š²r”)êª,‚ owÃÏö…'7ïY'-©‹p²n2™Û§«_ —Á»5Håþð¼Ä¨5­'0òiBFF :wš¦Òª\XþH9¥êïrø]À…‰ÜGQ;°Ú$I)™TÙ±Loúµ _;´_÷©Á»6M¿ŸS“¾&ðÚîü$i,å^þùõ×O­ÜíWž¨IEkÏ,h™:Ç!i|Ÿ^ƒ^iûÓ|~ì’aóI ï­ëaã¢õ‰»¼(H"@Òf'Vê ½Ê_…Å¥ïšÀ7J•ŸXÅÌ!OrÖŸ N_P"ù=Ô‚€+eˆÒRÝo¢fb5»SÙÁ åèYÚ¡g{-ö§åÏ—j]íä9â-OúxQÊ¢bÝi3ÝD9 aMlºœÆìÞb¹ÏZ¯}u~À‹zqÙÐØ¤Ã÷F•ù4œŠLIÃÂå,D4c¹…ÎÞ€Éæ0%äø—æ›‹é_°e¤Ö×=ìçû›Z¤yÈ~ }\YGÀ S©qß; 1MjšØ©Ô¸Uñ—ÖŠ@ìøƒX[o#x³% IÍvStÉ,1=n›®Ã¬êÁÚ9þ¸òu E:q]i_ÑH쾄 £T,A+y þOšWEå$HÈ¿€´G&Øâ8D± JÆn—oÝ'Ë$‚Oë7_š63ÕW?f•ûdåб/ó0¡ŒN^³ûI2“ÎãÃÂ'~¹[ô²28H¸ÕÀÓÓŸd§ôÒ1îÝ;û·¡¨Qt¾ÎX§aK%¾üø~kÍNIŽç¤¡IÁ5ã8Có 7WMªUèÝ×{Êiʧûæ]?¤ÇkAU©"×Wxb"ÃHã£~Òé›,¦¿xxs}ŒNkû+¿ý ¾G¨çtǸìÿàèÒM•Lì«öz%=¨/Û7­ËKëiÔÊ^ŽŠ™c´áv5 EîÁØø5ƒIèë«–DF ½eÙ㸙Lzý‘4—„&ÞU¶ïYÙ¯ÕL†EjNˆ¹ÐòùËfd-\M_é*…]ï´u¢Â:ƇV08ÖI:V±6¯ãC")óýØ+óŠ%^M7t›}|ïÑLÊS/ßÅåNÕù°¦ò$]%˜rtÌÃ~Z^ýýÀªD²Np:_+É+&iM¶«ÏeH/4¬“0¿¸µËn–O„b$Èb‹ÓuÜÞ,œ¤Yìön €ÐE’+†žÔÄ9K3BŽa_Ÿ'‘÷£†ñO˜|Ê;Ò•èñ=‚Cö•Ú !˜ê%Ãò‹(¡pJ´žó•Ïdü¯Ï^ÓqR,>…Ö<œªIG3Î:•ÌÞŽÁÀôg×sú]„ž•L;¥ÇL÷KëÐtJØÚ> ˜•ŠÄy¿»Uè¦:”UÍám»Ì£o:›ÿ"å¸ÇžÑ7ÏÒÊÓ0þòÅç.¯®Þy´fòZËÑ}‹ êQm“‡­˜*ÃjD¦²©G%c™Ù²í:£¯¸˜³¦NÌöfqФ;]è ê,K&-žˆ¿Ð14zùSg:3W~(J½Z›•Q8§0(ó±R22å˜ô-~Ú6ëá÷ƒë«~¸ßãܼhPU0\ ] ½yì–ü-¾1ïYwòŸop7² 0QõŸeý8;2‘A°´>ûøÅ›o¦^[¾GZ“ͺvxÜÄI’ ¤ˆO_¤¥öl8!xÑÏõ/Á´7|scPÿþD¨IÕeAñ¤6™sxà¿1ÛÓñ@–‰<ÃDŽW¬i)˪ˆÀÁà’Ãʳ-®SuÑ8:ÌsüL6»₩KqUã¦Üв+L òyL?Å=K1ÆúJ꫌˜ê„©•šónŠÞ—øõkïxþf·xíYdÙ$*~3M#BOI+fY}>]Š»è©)C!e!IèÚD—aÞYüž;õGâ#á¡/·Ww× üOègYV¬~ æÂ $?ÉÉyðÆE'=»ÃÅ Cr M¤Àã€þŒr#ï?ß·x9½ŠR>¾yýóÈ<—HqëPza[8<ìFèÄT²øÔ÷ªç"EÅîÆÀ 1ÏÏ ´s\\^­ðnlbü4VÏöù\ µ£ÓAÔèK'º'ú÷Pi¤~žîü‰ÑÕ½3’Î1mžZ3n qy۹˷©Ï7’vD† ÅÖ”2P®¸JÂþÝf }=¬˜a[î!1¡Ýä9 2†(ì}ɳ)¢xL0_ãòbŠâä×Ïûk¡ÚÃt³i4ô{Téç©Ud™ãF°a‹~ÅŽXÖÝí6£Eç\Cz]ßÕ}ã´…‚Ýv­*6ÊÅž/ñ#Ù^Ì!EÅÃrßã ÙogX<"Ú'î—¤Ì) ô½M¶]óäüÂÿÕé²s.nÁBƼÞºúÒ¯ø+ÌÎv¿ ³Á¹ ëe¤Ë÷­«äW “œEßq§ÚgæÓÂã>=)DP‡¯…³²L£ ?>L‰Ûscc¡ë¼ú>l¢ ¢/Ÿ:¾{wòÉÆÒˆ¿‰ë¹‰d7i6£z¯<¬ó2Õ_p^Êq7Nâ—’ŽžYÖ“+)ë›ãÜ\U5»a†;¢yB[3_Q#™ÙãÕ ß£%g9 T…˜#ìÒ®™] ~j®e&&Ÿÿ–É&³8½>«.?&š¥K‹°8uŠg-ØÖ˜ òÉ\ÂÃh33¦µÇ0ö¤aï‰?ý¶fè]+w×iø4•TLB#žÆNŸ‡pÈ›©OA.î¼ySÔŸ<¬èßÞŽ¼hÖ*¯)Z:9D³ ¨3Ioô](÷ªYÛdBÖ‰\'/Óú³ÞDj#ôP .-¢ "UåŠyûK´.‘œe#¾?0q·öü½ âkYÌgD-—™i‰¤$:¥R#˜71]W'×´Ã@ ùlÕËO&?ɲˆ€ëw£ 4E1àg;ÖÛ?æÔć€ٺ̒Ÿ”èÔÔÊCº~¢<’îk­º|]nB²øÐ8ïtÅÏIªFB‡„¶UÚi!-«ì”-ŸÝ-÷Ù0îqOôfÁKàX¸ÞÇ'Ë?î0yÃnŠ©¾¤C Ž­ÎRXù‰QSÙ>_kÎäÛ tUö÷[[F£/Ù£Dy¥R„sR,|{šï^6þòÙ™ˆûa‹µ¦½›·åT­½P af‘ºíÃqÊú©sÝôÌÙÝ…X‚ÃÉ®c2¹— ™„ó¶ô”o«îÉ(«(µ ]j†³o6Aoœ°› Ü…ráhyT„—Ïõd›Wž“ÂÕÏd÷<4Ê!ë‚•pLUÙUý‡Ë~QÒ€ºmW©\gñÅŠZeXŒÖ$˜‘Ö˜qˆ‚x>U& ›l ‹½µâ/RÞc½_'J@GM‚G›R×D¬¬ÂÞB ‹7öÚšŽm—44(GT™S¾\Z€M§›•3ýa\ÿt[oþʬÏ~»“òÅP¹ú‰WÌ“Éìè*Œåá:ÔȇâU“ãõý,ro48×/™;roƒÌŠ=¹ý>‹<]ç,â¿&2€óMÔ.A¯.ÃÑ.H2íá—ë|™¤N‚×óF¯Œ¯E¢æ6ød’lÀÚݶ·\)è/ Wx7°š°ÆO|Íûj8ÚëKè–Š¹«îXô×ee˜æÞ¸h.¹÷MØõ¶ýžÄÑÈÝæüÍA>L/Ohyù°"‰¬ )ÅÛiÌ*)h¼ 6$Yá»Ìkë.ëísR:LÃS ËÙfÜȇkªJ¡CºmvÅ_¢–¥ÝÜÊ «˜ÊW5L©î.Æ0íè/PÛÉ K¬Ö*äøµ¸äª©‡KÕõ²,kÑ»f0çË£|pqx ¿èù¸èÓ ˆŒ>Žsó¦s¯Ä§LgFþìö­3v‘ò8PoÃw?XçäUuØîÞ_p‰’† ïêºZx'vœ×ªr£x6¶â<­pÖTz\2b0™÷§GÙÜbŒ±ãc Ðlìúø\äü8’ˆË…ìËÓz~Í@xÌU]òÖµD4l‚ãMß1|Þ´ý˜üÌFŒ ©_T·7U†œdu5RѨçO¸ëRêÝeagï ÎÇ5ï498F6kÍ_\¬¹ sÛEÖœBÄ2wIv1Žt’†‰åꓵdØâúÙ¬úM>>¤l=ݱÚÐÆ»³oÅ.bƒ5‚MíÙOLŠ1µûÄc÷Ï#³å˜Ew;gO¼ò•U“kQŽA+* ä8ou9Ûþu®0¦ò¯_›]°Ä³o ¡~DC¾°A)ÁçåEE”ô.íE}À›_üÔÞAHކ;cánÔLæ'±µ¬÷‡K¨Ô^8g•Dsÿ2u§‡Šq_!}]µ|1¯v»"³'4d¡;x-·J®¶\óàräÄW“ún\÷îÌô{ôP[mgì»DžcßÔe!§)aºø OØÎ0ŒÈTÃÄq7Pø½–¸SžC׳žïĶÿÍ”¾¢f„øÉÊüÃK?ÏÀê&OaŠf¾Êi–¬·!Y“‡Öö[[ë¡[>ti5FÖÛ_/‘¸?" ¡­.gX_9÷¥¼¾h³_Ða]‰ÎJ¥ú°ee¨˜¥ß•;üëõÒªž™Àü¨¼ þ%f¹ˆ#fGJ£QìT£®Í· š*3ý켬wup¾DZYrÈkñëóÉ?‘1íØ{äa?²ðl1€tw*—P“T‹^­ vs½fCÁŠãÏÚÈGxó ÙûÖñìÂefk·D¯ÕcRèç»ñ÷#,ýófï²Æ;V¬¹Ö0ã:»e³§f<[m©1XcËÛLñ;WWïJÝÎ[Îëf¾½©ýØ©0»RLÛÕ¦3Xk‹.ãIxy™—öþ ®ÚÞ”}ÁI<É'ñémUc a½xÞªœ‚]cãÞ1c£_åÖ* úŸ$”½žëå~ñik˜,I !K Ék°(’§žgª>Ææ"=î`;²46êOl:?âW‘R²0Ömè¤~ÄèÁªƒÛÑÝ*¦sE^¼õ-[¸m¯´[ŸŽ"“¾¼×|LCàô½(¦/Íé+†ät|}Ø*~.15r)Ñ&t;î”hÛÖS._¬z?«`BÀ‘ XQ¸ÒóšðSäš jôWϼêñfšœ=iÁ73OC<×nÌâöôZ°À¥SjÞ+k/ÖÈÏP/2”&ˆm@i¨À“,d>®áˆÕ•þ1uúY|äa1x_F›a¦AÆ&KÒ*y ƒ%ubx³>”Ó‰'}Pî m½C(­æè‡ô†&Í~)óF˜Ì¶9j:Æku®úɹ{›_¿Bp|Ü9.*Ð÷æòo6mÜgék^r±p³a½0Ð";¿«òñ}è €Þ¡ßÅË—wß ˆ)Úåè':ܯf&'êÊîfz-è×u‡¾S˜/nRúñ VN½³qñó‡m‡žk)õhoÀFŠ1sŠkO}ÊÍF08Ñ9hŒH™‚T®^óÎ}—§¢Ã—J%š Ùìã(rÓ+Äb Áó)ÊNïioБQ¶Uóî캛\or¢ÊBÙí™D¨ãlŽ)¨«ÈäËñ'ígžCf? ‹ÏºÑ??61ó)õ@ÖžLŒJ`À"çþ´­A˜ƒÝÄüð«rMU"¬Î6¥Ž81ø£èɵ»‡’¾wÎ0f&Ƴ<î.ÖdŠ6ˆ? äSZ9”1ÃÝâK#3³æ¸Š<ÆŒ“‰Ò-_\ƒ©¼¨|Ñ#t¿ê$ÈRV÷wJŒ[ˆwÐìcŒŸÐqnbãrö‰'ö‘­a¦gEWˆŠû^x›BОýÉŽ1R”ÃP!…‘˜éÃU‡fA= 1).ÞÕ´Í®"BmhöLFƒºÞ•€íH«Ÿµü°ö¥C.Ê„4ê)u.)Yþ·÷&O騇bêüX_pNÀÜ´o¶…CED¾bùuecO}æ°)XJ7Ôèñaž{ìÊ™â+7r—rú¬2E«×òÒ²¦2ÝÄyæ,G0œ]ûV’K_Ö†â£=hÙ¯ÉT[§]l`µy†Yw{³¥ùL¸Ôx·\ÇXçâ!åèÜW=]‘¿±ÏÑðüG1‰ð2o0çReÒ´~Í-#ñqz¥Î·løÐç Ú×c‚ö„±&†”H+ãzù·Æ­œ\«mle†¼ ,±CTˆÈSe£<ûs$¹žCÈìß”Pf*‰éFfËË ŽÐ;Öì">7ŒN¾RÁ]ÎNÉñ’h2+[”uf›þ$DÄåoØ7r£÷ÙäûÅŽˆ²Ø&cUБ/ ᯉó•·õj¯2sª‹`…/›èTÔUFÎuŸfènUAOGo["ŠïK2­ÁŒ¬¶õpÕ¶V‡_šóbÖ×äƒc¦?/*PÍLR—ÎGDbUlù°¥°›X®(i™p%Óhæ ½NZ%ÂØ÷D„0ÞÙ”J±ÃR Ã=îd]éûûá¡Êöö½wfè½Åß.úHé½ý&?ùTi¡“'n‹§ý®Í¼¾dß¹@ÇŒ §VÆxÕÍ5U(~ÑTìiª+aÁ¯r³m…WìXSA¡s+îu(T‚n›üQïîó»ø¡Š–2RÄ3XãÀ÷ÂqøÜ ¡w¢éâÅówªÄ†/HÌÜÏ,ÔWïÙØÇ(_n“Ú%¿bB7êæš(‘ÄË¢Bqì5Ï´ÅJï9ãkÿ¡‚Œ´ò¹pÞBâåÊs:)Öyê B纴£çÓÖcàƒëF²ªbH-ËU @Ä}ݽøkF!ƒIè;…Ãø“b5}e0úãƒ$/ÞxÏFœ,æÝbv|WÉs Wb¬ÊbÒÒ®#Än¯PA;}=€ÒÀíÖâ_÷-ø¸¨&ÜÿÅÖŸ¤µ”ÓFWèd;Íê=¢Á§†h o3j>€×cÜiBÕ]"ïC­v÷+Ze›ò µŸfo¿Æík°•|óE£¡X.D„QÛ)þÂmh%]ÛöÝ‹#¿âO&ñ#Zj·™)” ’¿ a=ݺt$IÙeêÊ‹]‚¨e­ñ†äøò4Þ‡â»z1­fý:X°ŸÜ–¥Är3ŰÕwiXÆ’‰=C…tlñ-U†~È_n6×2ð„ m·a/ <€Æœ:U}dð'êPÍ­-hýQ=\ο3BnO2%ää>Uýb&.Iv½tûà?yþ&N(ņб«B—›¡£©P°õÃ*ʹW»à Rü \5Áó”î»ê*¯\ÍöÖ„nkñ SF½ÉÚ„³Z+Ðæ<Ù`¨°u@aËAȋ˛öq%±˜áSrUdÚQ6DZï ÄÖÅÔ¦ dO½,t**®† Üú9äroÖQ[VÃû¥èh‚0OÌ%8Ð;=¼ûuýÁ-ãX3ù½$º1zHÏB€k¢¹ó>ðÈødÁÖ"ωlœ”%–j7ŽWO¶Ñƒ†P –óGHŸfø×׫`[f·}w%oY-’€8Ô©¸›OÜLJ%"ä6pœsÉßvJž5îB&'±G•zè1óRü«]Kx*ÝÍ‘í)¹,»“W¶ï{>Ã]­=~AóÿÙÝú¨Ïýú±ëÍP- ‚v !Zê×̾ó‘v³“ŤÄw¨JÁ޻ıás&ºÌ¦x{­W4âþ‚‰3?mŒLàÉhÑF+çcÃpô!ÆŠ<[Ë*ŸÙ9)}îµÌTbª4®«±äø§¥#PVÈÏVOÙË‚÷hvGÆìË¿[æ†Wú ™’ô¼¹Vº^®n!}*¦³RwíIU;‹ ntÊ`†=õòœîÔ0§€©VÄwêÀAÔÀþbqǹµË_ch D zti¬}ûrˆ ÏîAðb¦2`‡g/§":²ç8rû_¹I‘&Pv]’.O0ÏÒ.øÎ÷Câ,`g ññå.1Jj÷ÙÂŽy qÿÕÁ½‹n݃¨¼{UTô-·vóÁ!fèYÞî»QE%•ˆ$%˜8=*Aw©º­½žñïßgqÂ!|š^œÛߨÔE™¾;v‘„ë‰Qî]/2?G}ÑQÊuÀ¸l¦èÇö•ç×ÚÜ(žÏŠT£Ù] aEÌÔcžt0vd2Š_/ØÉsÕ,öáµZ<­wN>Vœ°+ÉÛ ­¯ÔÙ?7uîyn¶?ÓÆè!óÝ h½ßïÂùß§¥Ù­F\†b±˜wŹt¾Í+Ä©Ëa_JÍž¥t|ná¡|Ti7pcvâVíšÃCñ÷XŽ;t-)ž.Ê•?D£w4½™LáÒ¼H¦ä̘j ttŽ\yÅ–,•¸±$ø9¯SV4Þ;èi³ýyš¬®/â)«TGL~>552étfø~ã6ƒ«·Ý;Ñ3´ËeöbQÙìÿ°M”ÍÂ|:(ÕË€³—œµ3ð•åS«·úéCËÁZÊ“RÃíÖ‹SøçX£Wâ01¥„ßï×È~M?sžkn´2qZq-~¡š/#;ˆØ!žh~`ÒÙNKÔCP0_80h#X;>§ˆ}1\a›‚—!ãïÚc:¥íÐÔË»c„!roÕ6î l ï‹Îï»ÀÄ=xª4£¤‚oš~²Æ+‚lnÝÛ2Ä< †¶ø˜Y9:î×a1Î5Šøµô®ý´Tj¨ë‡ E‰1Èêâv9÷ÜéÐÖ*“†àÌbµ-ùeèkñK/|¢¹™¶úr?Þ«å‘9<:•즮É8†mWU*ÂÌ~£È)Ñç<ª5õ¨ º÷Ù?ªjót#º¼”@cfÞ*U3#e­ÑÐÿž*‘ÖÁ,wSv¬·bY*Hú"ÿ ||Cä§÷ªÎU™¨³m^ÙpIŒÒ8Re¥«iEí¨Ì~h”chÓ•L®àôÐG`ÓOdrï £Ëi în©6éùäo86II®<Û÷Øëœ3—ýèÁŠ*à ÃX76‰èœ"º‘âßéÕ†¾Àqqï™ +WÞß…þÿ½¨ß+qª!cÑL¹HXDòÃîž4i½]lÖcv1S­à¹Èä–$#I7ŠãrH!Û̦N`çe§Ìv!G+R–êÙ+Uùîö‹ é<û`ná¾*®À|ðªÁù»ƒjDP8>÷’aãÁA߃cïQËY(Cõ@æCöñ--Èšg®£¸CÜÔT߈oh‚KA+?†u€Ÿ¾óãÛûâ3mêpâöôsõ³¦ fRÊ8 òØ­ÔûuKƒ´ÊФ…£Yú—1LaB`˜EÛNDgÌìNSRé‹ï¿n&3ª7ò;c{ú>Ê;ûbu§$ÚGÆ?ËL…9—NEªêÅçSåWÜlµ’¤(]éM85Ó=ÑD%pÎ4ôíQÚ¿—©î“n™£ˆ¥—òÍ#¸tt¹´«5»ŽQÞáAì—Ø›§‰ääÉíøRyFÀëô ?Y•YØ®“ ´1Ù-ŠvÍÆ“°õ$—ÀÐEÄŸ1Þ–{¾•t¤1ÉŠ+."·µæYf ?ÒÇî§(Öø‚¼£Ü?ªÙ„í䇎K^çŸÔ÷^õâÀ£•×p´cÁ‘wɤP=_ãkÆì¼(CÇ÷Òf¯n*4Ë~Au{¶À8óöífµaý¬‡ísh;ÕÝaM`g=HƒPB·Á¡zÈ`]¸¼9\17/º–y)áà-CÖ¨‚Ù:!mùdíÜǃ \ãW‡ZÔçžêO|›LVS>^[åÒ|ï®óƒñJLhMÌÉõø­³ëéSؼ2L•ÊÀA™ýÚ<Þ¸ÓõM¼7Ùu þdÒ·2I¤jZòƒ¥ü ãÈækï/}ÖÆ¬4Tàg\öÆi×€lSމ³€%ú“‡.ôŒÙ Øâ,6OsÿN«bö¤›vúžC€%%6ïÆ½s€Q¬ä>ZÅ€ÄúÔ7--Œ³÷f°:=½¼ë8‡Çgíz }NƒxÆÙÆgºŸèxæfL ¾ ãÿ yS8Û·á÷‘‡PÏ—\”&lÑ8§Mq˜yT¨Yܽ¬Ïº)ÀÖ×JÙdècWŠy¸jÄãGàv³÷Ø+®Ê}Sí bHÓAgfˆµez}Òþ@º>jì ö‚Zï¦aqO9·‹í([ŒD•Rýù}vÒ}¯^1ÚAÙ ýRƒšÝbÒ-ljß·•Ô 9?X {8®c‹³«+ói;˜>=s÷º¬î…Q}°.†DÖcè²<¶émªûúÁâò m¾‘@–Ù¬ mý¾_Ÿ/ëõ›¾Ÿ["S,¤ ê…>{zKÅçþµ ¿<¹ü¤Y›â|¼6¬óËyäÍcn«BŠ^•úrôDKýëŒO• ±& tU–Å„Ÿ¾)ˆ§­ÄWÔí`'Ó š+N@Î/#}X¹¶EyGÇw|òÑÓÌàu:à›°‹Æ+z•Ü(ž‡Ýo±´,sÏo?°H‹ìšÆ‹7·³þ(óÏÃD麌KíÍ=ûòÝ VÙµ¹ßÁ”!%š.#”´^vÞîuÔDÊ$¥ê”«Xßž7âÓEÀy@ÜG¯n/f!£:=Œê;Ë­~HŒ"YžÀmnìƒÚÐ?$² œ/¶úx›ûY¸|ú{ôþ'þ·w‡ˆ‚WðùÏOR–H¹[™lãý[ Ïâ,çÓ†\þÙ½~N¯WCU£Í| ó!*òWN™iqd3k“–¯$?ƒèв©c£BË(a¥˜Ú¢¸)r•àx0ªtMùöqsìögÆùÆú8ø1n†®Ãݪ@¶’d†g|x©UrEqUì+ÚOà¹ùêÕ†í›Å7vZ¸î6QGÈo(Jaú%1¡-| à{Oé;RQ£÷`È·`&1dÒTcëÿüÙ*wœEí–(0Kà/›pŸ¾u_™ðfù‹á%`˜'èPÐ!ÌóSI“4Fr„£‘—ë ]7Ã1‰ÅÒV—`˜7¼ /RiÏΚl,;o¯éñ$ kMK„·°'Vî½â˜Æ:δlF‘ÕzéË!Ù»ƒ—]J%M®ý$ÛÎÂMü–cMÞÌ܃¶Ë4o[¸Mꑼ$è»eŸhßí^;Q]òÀ}K.G–òÇ“/+4’C3|Ý"ÿ>À…é4áÚJáÛŒÞù]o+ZôBcR×_c™HÔ;OÁN‚ÜD6ú]´$k#KŽ5pUè]ÈIí]êžôÞ_}—©0÷ö-TÒ¾yÛjR>—âXDS‰µNA”î±®Óz ß— ư›‡ûPf|€<.’$è;Š7¼9ËlÄ>*¥M¸s´~0y·b¾ ‘L~µmpÁgÍÔ‹êf¿'úê8}ÌzÉ#l…¿¼ à£tðٙ¼3 àHäµÙ©w¤ÙŽÖôò/Qp“àÚ´fåݬoŸÎÒ7uJàŸÛnÂ;E˜xi¼ç‘Ó{·ù eÝþª¡KkV Ïñv;ʹ†ce¶+/%0êZª,i#€çŒÎ®G™O&§“ÀÓ—©[ |Ø™ðÊë”y§wJ˜ªy–;r_0;hY Ýñrû‹Ñhþ­mÎñ®¬*NV£µNvÊéÜc:CgH¹Ó•^)­}1y|„ úmh¡;; ÙÉÑ£:™H½±…xR%ÎZ ÞÓÙþ[üüŸ3 1Šá«AgßjâýbŸ5s ßòbw»ñþÁˆŒ endstream endobj 68 0 obj << /Length1 1616 /Length2 10030 /Length3 0 /Length 11086 /Filter /FlateDecode >> stream xÚ¶TœÙ-Œ;wi®»»»kC7и;$œàîÜ’ Áà'8Ip{dfîÌÜûÿk½·z­îoWíªSuήó5µ†6›$ØÙ"çìäÁÆÉH«êèp@ 7;È…FG§õp€üÇŽF§qs‡:; ý‹!íy<Ûd@ÏDUg'€’§€“ÀÉ'ÄÉ/¸€@ÁÿÝ„2 /( ÊPrv‚¸£ÑI;»øºAml=ž×ùÏ#€ÑŠ À)(ÈÏúG8@Òâµ9TA¶Çç­@mg+(ÄÃ÷¿R0ŠØzx¸qpx{{³ƒÝÙÝlĘXÞP[€Äâæ~· P9Bþj c uÿÓ¡ílíá rƒž P+ˆ“ûsˆ§âx^ ­¨Pw8ýIVù“À øksœìœ§û+úw"¨ÓÁ ++gG“/ÔÉ` u€ÔåTØ=|ïÒ öúKŒà¿s©9?ë`üGæ&@^ Õóçÿ³ØÿùÿÓøï,ÿW™ÿoEržøÿ$üü G¨ƒï_ŒgÝzz<Ï€ªóó$8ý/UòçàªBÀPOÇÿõ*z€žgAÒÉÆáËA} ` ¨‡•íÚøÓ¬û{ΠN gwèï›ÀÆ þïy¸¬ìŸo÷gIþá‚<Ïί(ëdå þ=d\¼|›È ø¬$.^^€?çó4‚!>ˆÀÁîäìñxî.`íì†öûHùx’¿M"~‡ôßHàÐøq8´þAÜí¿‘àsè$à°ü 8¬þF¿{äÿ r8 ÿ‚ÏdÇàó¤p8ÿ >—àò/Èàpû|nÅý_Àáñ/ø\¢ç¿às^ÀÿÚN+O7·ç;åÅ?ïõðâ±B[œs¶³«ûxýA’Ì›m{\tšn[?‰ÍÑí“ç-r SmVèšÛ¥dÊPöò–,ã…ÄÕƒÿa[#rd{’fÇ]À½ù;­©í´…IÂþ‰¢CɆ> Tr6‰€×½{ø6ØN%º2VQÄ%:¸- ÊÒv4p„|Á@ÍÚÖ¶pa®w9Äš'f±öö–“/Po-é5ÛŒD½Šrc«zÅ#«ùúïl,"\^(×Oû ¢¢m,·H­È+L#HÞ®xøùÝZç—)ÐÅœ»ùÃx\ÞÒˆUÓŠÓNœÙÁu¿wüдjË4³cë,áŽaMþ¨ØÑ2&»Û6^U‚‹µ?P¦+€O¢r7w?[Øô€£ö³mÀÄ$ûu©–$ÂêÐåxÇ,‡K³ezýå`¢˜•Ê–‚ÓG¹^Ò•“JŠìÆ´N²¥ÙŒ!î9å Eˆb~ B~ÉeÕ;u ¬hvá·+ ± Ã,ЃºáÀúÄëü»>Cƒ ÙyÏJª…¼.ߤ?–läGkïËóø]4?Åñ­`2ffyN=0½ ?}I}Ÿ³ÍõÑØcòëÐpêRalc–¸í¯ŒèVˬGæE 8Ú`4p±E4-@##<óP6MhÞY9ÓÛòÝ©“þòÃXk"¤›J<™¦"kìpZOižhUfïØÏ¯XÈš‹~ØñwÓB%žÒ7ñq˜H`§¢Ft8`&]£Hu—'åöë®Ú8“=d'B¿¶¬H`Nq’|)¬—3ÓOeËNÉ>ï)q% C6Ë¢¾‹…cæ´ñ¾ÚåêÊçá(ÉËmåÜ4ùÀï÷ äw0ÅI¾Ïô¦…}=ãôgHŒ¬O¨!Cû`ÅÕo”œÕ+Q'»Ï\›Ÿf‘æ¬ð÷¼91âî¯G÷¸ƒ-sM#³¢»}[ë…ÞŽ~Ç´žË/_>ˆIÚ>8+¹+³¶9¢ëtÉ¿žQ©S22ø—ÇÏÖ`yW>jïhC2džÎF¦È6¸&Æ»v)¯äªƒ¥fËœ§ãÚßã™!¼îNûaT’‹„~¸øÄ’P«î?%—þæ h#S0äîzB˜w%«HÃQrf%Oo~Øoe¸01ÓŠ*’Ø‚ÛL -Wà žôǭw¦WZZ™ÒÅÒfµ¼‚/Io°ˆHÒî.0H$Ö •2ê3ì°ÂGÂ6<¯MÙûs^Ê¨íø¼Åöûè÷½¬¿rJÁ›¼ –#–+®¬© —ìõ¤{}¨jjNWF½òçˆÈ®žc”-usªÉpŸuÛÜíÏ€9àáØ‘ƒ˜{?¥ân´~1Ü Áw¹iBãüQèü"Xæˆ1õÄ_’tÒ]u‹ÂB€ƒ‘Ó4H‚€{ž… ƒ[föu©”ñn‰³Ó+WÖà~AÒ.Záçk{^у~úLÃü#(å¢÷ψ‘¢2âì/éÚ ˆß–BÝjܾ…ÜëM%duê®ÊG¿þü/…²ìL…’Y,iã΋É(5Õ˜Á¯ýð-v¨_Uƒ˜kÆÀ»ºè ÈLìŠä¶®ïæ£(™ðy®?‚Ûë‚kÅvÃÔtQû¸û©ã×ù_ô ª{4sÄòºá¶¹ô/<Þ- ¯¥è§“»Ê¨ô}úøÖùØ¢?iìL0†)¯ôäÇÑ’µ§ÀðÙˆ6ìÍ"LÆ‚êx‘@\ÑúÛø~#ìѨ·°E("ükþµ*.«8_®ô˜èðp7ö쇗¹_‡eè8ÝK;òˆµ=¹ ~ðïp#êÖ5z”­È›‘¯Ùð|Ç#¸$¶náLâúyŒ#´h#hËg35^;< ËZj¾J¼ËàŒÝsPÊÇZÞRöžÝ[_ø.<Ùµ •¨XJÅÞ^¹ÉYÐÅÆæžÖa? ’þ#YðWb,ÕAsá íåŠN¯—ËÍbGüºGVû¯Æ¾4#WÂ0´ÀáG:µª½˜ÌçŒ}¿[¢¿Ú_ùˆ\¶ø¤·ŒQž®æ‰Ä"ø´ô‚‡'ÆWþãTÏãÚfYfOW4Mµßð&–³zyr@••Ñ#¹ŸTÑÎk‰£`Iª•Ï7ë BˆOê6™¥x®Í8A ìÓ²aWøÉ¬9¢7û:%6Rx‹q‚ºrJµâ4¸Æ.¦=èKˆÃ.ã¶± Ýyã]‰egºpÁ³I~ù»«ŸU"Ä‚oá’ˆî…(•c„k.nóõŸèF ’Ã.h"…ˆ,‡]¶åÌÀï¥Æ)ç”´aùŒÙt˜y¨³â¦| a,T›áÑ«Qj‡Uñ*5$_ö[¿ Ö¡±&¥A‡\¥=”Ò'j}î•8D7-ãR >ˆè…áCÀ8·e %àa[¿¾LM¢&jF5z`·¨ŠôÅÀaý>q®ÑòÑ´ޅ苬qÚmC7ùùvh?óÔA`?46vËU¶Y½Æóu<5÷œÛ„VÎEÏ)šdGÊ·´‡GÆ€ìIµo9W™ÐÁw„1¯_þVDîÖ0%Âý4ÔtÖF¨Ûù˜r¿—ҮϚL3(ž™r¹e)•;1>x?/°üM~"guåÄD¸]ͩɊ+j­Ö‡Ê` ?“Z\h`§-u›Ðã%å—uBqº¶©«›°híNˆ±ßû\ÈLx `xÍ–)íVæÞŽí6%¬÷H=ŒÏëÂϳ(AÂò5jL*ԧѪÅÂí.½[×.9&–âÑn§™v6–9 I`vF!´îÍ“Ê,röØI¸4º"×Ί_b„¯ªÄÌöý|X—üÍǬLÝ27ùœkZ"šÚ.òW«<’t–¥&æ)ê“®›!¦7´ø¾­-RQÎ*n˜Áþ¢Ö]ÍU´ ¹ë¾@æ‹×t7Ì1X¸··`÷ÂLiX5?«æŽ÷¥w¡¼òAc¯zZs €®Šp%9.lô–©¥»Cç ÐöH_²±"CŒ‡˜í®„×sMT‰Rçü¶:I!ŠV Ìp£…ЩorÂÿ¾Ž-äÅÇë[”§¯÷Gɽ¥J> )óƒÓZäß6؉Lü‡-˜ÖdùŸF©nqù•Ö¦æÎ®Ãæ§»®µ³Ë·rIÉüüIÚ-›Š§­ì5,Q‚úC`Ѐùu=_®KÆüȶ§«©Ã0u^·ÆhUÐN³S†–qäíIÁm†Ð˦# t7d Ü«¢zë^eñ¤-,õ²f­CK" -ÖÚ˜! ¦†,q®¿Tß1·dµ“uù-·‹¬ú†èîHOJD¤Fåʘ¶¬9zfGŒóݪï“ȲëX6á̶½Ç1çŽÖ+hüA’‡ÔHÆ£µOOAXT?Ùtpº……Yî¡”öV¿›¾“ -@߈ñ&ì`RAuHb'rœ¬þ¥pg8ÌU?&ôv¹š(GȈÙ‹’™FUd‰2Ýw~ÓNPMÑ1½Ÿ7Îvvå›IRôÍwJË‚nÖZu'‚¡ûÁŠêj"RÙ™aŠÇpNB@äVDÖð+’Ÿ £~€Ã.yÙtÕ%ÿ Mþ¸e°ÎW‰TM²do—q»L‚¤Y[Ϋa†‹²IÄF³×"Œ'?_ÝvóÊw¡‡A°Í@ãfèdTæ¨à4‡JJ(ßs/7=õo„(° ÖH©»¹¡¶¯×l82|:ò?¨M÷Oër‰ŸÛHÿÀ× ©Äé‚h¨ˆ¬lzÓÅúyúJ2ýDKc” YÃ*r¥¯»T”€¹~hå1®•¨ÑÿÞa4þRÇŸÙ¨ƒ>FI7(µ´§–‡Q/ŽLNs!y:~T9mL6h¬Ã\Z½‘I©!m‹BpÝ>òÌ'ÍJÜ€?W¤u'ïzÕòâ7À´gñBze¬µXZ¬D5 pOî•ß Þ P<¾Ÿ™ttŒ ~ÿMX¶ºVå ×4K! Z;©çC]d àåÅ`'rgééõEcéöŽþ»Xõd7[€Á«²ó0m>,›èSHÑîNjnØ`)Cìt¤%¤º@B|PúxThTnåœ5AæŒMnIÇÌã©A‰9‹G2g>“ä–ÿZÙõÉ–Õëùyï*Ëü4˜5GÊÓ´çµ¶ÁÌT»jÇD ñ–xvC‘åøpÞŸ<ÍûJá‰ÉkÄÜÅæŽ_Dá9öÖ^ižýjð´þµY€vÆ ˆª†Pб)>„çjWq–Á„6üöðýt=KNs¾$îÀŤ[ÁfG9½Âd²´àœ£bÀ~·zÜŸò7tB‘çJnÍ"«Œ•6òf=½ Aµ>ûHfÁÊ›«3$DÝ놷†á2#I4×­É´øâÖ-ü”iKï×qÚ F¾;ïEírßg1l²è„Cø/‚r}®å>'Î! 'Rð=ÑFOÒ“¡[(¶0h‚j8$‰á‚¡žP«÷±ïJÁÞhVFÍÝøˆ´ÒƒÆS”ÐèyÈékô¯4W0wõëŸû…I¹$ßùÉmò\ê#/h“+ §+m”æì3=áÓe•Zw⨋9ë&ÈÕòŽ\k­æ-§ÉÑßž­Qäû—lÇ`ŽÞS¾ùåÈKâ¯)ÉäMªtàl°þ–Êí«…Ë §“éÊûÎ2€2èð6QdŸë®‹ŸÄR20<$7ƒ´XÅ`Fo÷Ž›ÎBˆjbæì\cgã»2ÊvX‚3›XEXÚ yÖ”§QU ß@ùƒ’,¢aq&«ÄeÄÕŸ©ïŸø7”Ñ©¤†?½u€b_Æ€s¾ œ lšõŒòBSn Ó‰TP|ªV‡½—m@5ôêèêrú MÓfÊÍyKáV3ª!ÃËïl7³ïõ)$jñŒ'Þ¢l`Rny•Þ.^©ÇP ò9% ÄèaÞ#DȇíŠQÝ––þUð¦&ùø•Æí=™{渶tÍœhÏWßaŒ;ã™»3j¶NÚ¼*É«ª88Šlb„³ÉŠ?`8#Á/³p•Ðg(ÆÃ‡¿¬LôÙ¼]k¶bîÂe)«&X–ÓwG¨NhzfXjÀ·„V{½ßµ7·Qµç5rîu^ÁÿN^–Ûðú«u^|ÑÚ±|ØÌýBÓáXk.v^Uú–ý)œ¬^‹8|zî‚(JéôÅîøæ>¾ìÖc•‘øWb+@Xºd Æ䯷5kp¦ßÊ ÷Q:&FðòæÑC‹²] %s©ÚuJÛÒ׳}< já!öó‰²NæÞNÕ«ª“ÝLËÎäUe’4¿¥³r¦âŽÓ”n˜s+™÷¬ Å9wb÷l\À;­Ï®â8§• ·å/v`µoçœçôLz‘¿™BçtlcâŒÍ׌šÌì"Èk “y‹jËÔ¼_¨FÑ!³È yÅ¢EÓ„Á·NWœ”¿ÚH•ø Ob¼µhpĸªï‡ÄŸœ”ÓXº¯C]ÅX{Iœ†)¾”erïôÄB&G‹ëlúœïJ>Di´nZžr›—«nksà7•„¬ì*¸KB­¹ÐÔ›„¸¥¥„Õ±‡½hD^6HŠ ŠvÎD‚§‰ýñooUG:¢›/éUvœ<³K^ó|ËÕ3]ga/0·<(£õ¸¬Là<`oš˜î¶šº¯õÿ^)"’éutµÜºF8ê7÷+Rä} æ,up²pFpÔ i•¸)„&Ë2Ô£9c8Öw²b._ï`ªŸJÜ3÷uF¬pe ŸƒsñŠJ£H*±EŸ™ÂHz¨?3Z=ç@;$õذ< Š'jŸëjGÚÖ«ãO¹b äŠ2ÓO¸ÈÖÉLÔª(]ïÏ9Ø!¾füæhg8/m ¶1¾t¸ú…š8WQÝ&ÇÉÏžåÆÉ(“IOyÖ—@22]âQþ Ø—î7i|lK„€¦ió9챋a¹.šø»ðÚ—û ý’¥8>¯4.Ÿ²W>H (ÝW™|o{r¾ªRµˆñ«Y°¹þ8'îVè"Ól ð¨¿vµíQŒèâ®Ì‘Q­‚Áüž@Ø<þÀF¡au¨Ì®Cyøß,ûÅ®0Û„3;h—ç †Mý¶]ñî‘Zü‚ÔŸÎÄ-Ì™ÀeØ”;œwÌ›fj'2Ç8!«òýHd'd1õà ÷ê;]ÑÛ‡·² ݬšfoÙ‚«ê3MqjÙC1«ùWuë½q°½©À'Yòµ×Œ-Ή忨EM¯7×èÅú¤N’Œ‚ùÛăû%>A!âæ”_?Ò¶x*œítCœ®XãÚà{È·0W”z,«J¿íÍJRµùxÝ!úÝEø'P9®­Šûv+ÃÒþ0¡ô-BúD1KÌ妿ø¹î£ô¸%$âÁ;‘Ü÷ÂóA‡]ùš}‘5?æCâÂJ²³±Q6§H½¼û¦çˆ›%_wÚV!M´°|TC¨ŒV‰•pøÄ<6€a-±Ì2¹²ÖÖçað‰Ž!8nÔÄ¥þ•øÀ-n½:8Y!ª “£äþ,Ý€jéY\ý’'âc\¨û~iÖrSñž/ñê§O‘NÁæúRÑz‹Ùk“{kO$óbîAwþÔ7M)ú‰ŽSsR+A[*OâÇÆÞTYÒÂÍ”ƒè‘¶>þ®yêÌ"èêMÿÛ¹\s˜ÓárÕ§¼íƒØ8¥¥!>×™ >ÄÕ±lùïÖ™tÃR×Ro¾©TJw›ŒƒKôû¤—¼”H@˜Úà7³¯+fщèÔ™bµÒtlºSõ^‘æ…-›\ tÕKDŽAvRɬÎ1NY3T0[±˜ùˆB5\*ªÞ»â’Ú`9.—t#W;ž@.íã=‰[ßì¶]#Â5¬’Fõ¾÷ÅÄTl ™>íQ™ëkh×ò˜IÒ÷ŽV ­Š[çôúc¿, ŒgªpGÉbòÛrä5¢`½ã6@ë~…ˆuxPúé@Ô15áƒëÍzzEQd+;íj…àyÖô!þ@ZÅ~œÁ +ÌSt ò*YϤG›Q”Ñzc–êÑÊ©©À>ñÊ™[¥—½vå"­­™^qTj½œ4x^}Ìˇʑiiâש ‚|¦3”"š1 ˜#7ãra¸ë:“o#Á¤‹J‡›j¦Õåuû¢J,ÜâÖYÀtk‡µ:G%w•h!ÀÉ­£ÊóÿX¹\û UÉf‹—µ6æ ­@xñL ‹üÇvN9à¤{b"M¼D-eFÄ'àB­oº¨¦{0Ìâc†ÎÅäªÃ‰*äÝq€“¾W$?D‰Q%‚%¯ë[4¤¼ß’qØÿD93§ðá) šÂùq`ªËÏêR+]Œ±Êv1jU/M 6caÒĻ̴W!¥GÔî…Ó/„ ¢ ë$« tµ ×Ûa æÏ¾Âõ§Ð’óËND#- ƒdô ò²)Ç÷lNàÞœr3^A˜ áÁ¹BEK`ÅÙK:âUýª³JÅ9Ív9R ½ƒfÀ¢¦A„îFCÂ׈lCï7K³zZ1¢~åÎS4Ž$ĹòÞ[îüÚÊ{Ü‹iØàÓwö+Rê¡Më/–·o»’§¸ña¢ô u £VQgŸ¢cºaS|UÛíëzmoËÖâ „õçrT.g¡<¹´ú¯IËy–O~ŽÅëÙb^¾ú†X]ÇD–À9ŸÈ®7ö ßé/ò†XYÙ®„d΢N‚}+éŒaü’‹ØÌüé›/&ܰJ–_äpín%"ðÓš!_›[c/\Ñ}æÃ2j&ÿ°±òOÜEͬ*) 9ç|—7š™ñlߨ¯@ï‚[°£Ðpõ»>`]nq]¥}ÊÇu—)ÍÜœöîGÑÉã¢Å2…Ó¦¾ÞÁÈgËá\^WWù¼Qe܇Øf–+FóN•yD¾%z–^ï®oŠvµ¥Ïqâ=bßÓ[Ó©+ØŸ·¤Ð•ê˜ Thnžlíáâg®ª”ãh‹‚è†]ßöCû\Íý¶“ðãHª,`pÏvøhšÿ»+‰Úèÿ"^õ*b”c\…Ê8žoÔ¢÷®Ì¾ËŽ ¦ÓéëZ±G]0QB"¢¡ˆ%ŸœZŽFb¨ô‡Œ24ë2b¦ÆºãÂIù¢Úâ3áœÞ—Уcn²—íê-±A÷?ï‘¿))¬¼¿’—E|ÕT`tM .®ì+䇯,=4KÅ.¸Ñ <1Qõð‚è—ô=}÷ÜHM2\$h‘YÍe÷3~QËâKö…D·ÆÎ¤`¶«†{¿&²Ýu/f]}$×ÃO‹ (­Œ+~ZY<þ1"âôãörÕÎ_L!61 »ïj,֮ȶåÃ>¯DÖQ4ļ;zh”z!qF³XR±<Ï#ôã­ßî8·ã8¼: —º´¼·†ݨlU‰Çii²=-&)åFèô¨N4©Q{¤a>°7ÅLdÞ½É4¬B¦‡ µ ê zj)d¾Û`žiLJžÂK~…šàCIëB ‹a´ÕT¶òõǾF.i U+$1 çžÒÄ|zÐìôYéâº×2ñ¬»–î½”ßZJ[±œ¤#s \]•¶](k»À/¶ôj }ë¶Š°ÒÛ€ïƒFúK9ܯé_cƒqXNûÄ/ ’3?·D>[;VºŸ\™pçì­ÝO¨KPÛòæÝš¤¡5™#ýààŒ½½øI6¾èoìÌÌqn¬`ÍŽëò&x8o2øøžƒï}4°‚>ѲßO}o&våù ÑxÒ-6]ïó€Íù‡û@Y<¶:ŸXm<Ôë)^8J0r·P”¦U¿ÌàËt}FÐv( 9·¤:{ ¥’ÉêËA!˪"CÐ)ñ„ E2SM¦Jô9è O²¿¨8Þy~³}Ëjô¥^ª«VÕ»[éQ½ú½€ÏÄEóUno§´]¨G¨•¾¬r£0 õ~,WÂuªs)ÒqScf±b}ú¼;µµìbdâˆ^ñí¯0‘¨%£Ðãà”9¡,¶>ß&ÂôøXUìïéC¦ðÜe»[­ˆÎA²ð‹•g½Z;Ó0HuªLû/>>yxfØ#¥D6‡Ù¿qø¢ô“ñë/ìô—ê“~WX^ݽ]BÞʈ‹u/ÜY$¼U`BÄÙ˜ *L$™<áÄÄ/šVeÎÕj}¾–^Âdõ©ŽÍÏ ïQì§b Öð€F>(-’se¾U>h0¿j„ÎDb`à K¥öט*ÍÒ‹QÌS…ã•ÏM×­—T#C>£|€64q 5‘ÙØ¬K‘HF?ž¦ß¤![Vä £T˜'ÉÃt(nEz `õ9ĬKda´i®°2ÞzH†D€mtÆ^>Á~T6i­4%Sâ4Ú¡`%†ô†Ã7¯DOUŠLÂä>»fƒýÞáÚYF» Çÿ¾+ íYÛÃþ˜x3F" Ì2ÇWI˼ëŠD¯!À»Ø• ¿®Z8"'˜Kp̵EAùTªv åÇlN4¹^Q i\¬N«ï ‰ù:ßzf½¬„Çû¾•—ÖNø¹Sf7/\DÓ¸¢³Q©ÐƒZËѱÏh™(ÎÉcÁ¾ÐöÅ]œ¥®uªQß8œFóE2´Úø8B¨Ÿ¡íÝ&]ù`MÐRJ5OÔ‚ÅL¿x "4ª%Ühn4¼7Ž”– k\bå'§ºûd…EWíæHE'ߎïI3¾”$ËማFÑâZ¼¤ä‰f™£}Ì;ö†Õ Œ(½Š´÷]têèšÃ,¹6ÅÂŒl¥›Z¾‘týÇ$þ¡Rw[¢¥jG.6•žú×½WeèL WÈ,Ç/-"ŠýeþáU}¢ endstream endobj 70 0 obj << /Length1 1680 /Length2 3950 /Length3 0 /Length 4992 /Filter /FlateDecode >> stream xÚw 4”mØH˜’²¼ÙªGK–™±eß÷%{$ɘyÆLÆÌ˜†,!¥R$EÉRʲ“¥‰$K!ƒ¬5–7J}–7ïûÿŸó}gÎ1ó»®ßµÝ÷ïºÏ—°sT0À’½@S2‰®€RDjF6NN(%‰TVD"•`p¸NÛaðà •F “470Œ¨ šÙŒÑtˆhC&–þD¥  Ô4Q5‘H@ ‰ÔøM$S5ct Ø(–dHƒÁÈ” *ÁO‡êüþ È`d”†ÆAùá€/H%`Ð$ÀMǃ¾PE š8’1ô¯2Úx:¢‰@0 E´/M‘LõÖ••:pi 5Äë#‡Ð¾à¯ÑapÀ O ýt8’qtš HÀ€$âO‚Tª8ZX¶ô“lý“ ü:¥ˆú'ݯèõDÒ`4Cö¥ IA’7€#AÀÖÔZ‘H—Ð$ì:M¤‘¡xtš@D{A„­£S{ Møk>†J ÐiŠ4q}FÄzè˜MHX#²¯/H¢Ó`ëý¨ :÷ įËõ!‘¤àßG aqëc`ý)gÁÏ´0þÅL°?6o¨"‘HueeôÀ@ ±^À)ˆþp¢ÖÍÐ ¡Á2ÀAc€¡}Á‚iè SýÁÐàŽ# ` :àzH°?Ù!3ˆû‰¡û§£HH~(¹þùç×1HaX2‰ô‡þãŠFGœŒí ä~üÓÐ+ %e@UIPÓPBÿÅMøÕòO¤ G4~6 Òï†~)@æ×zÈÿÎuˆ édþÈÜ©ŠÄ@Pÿg±ÿùÿi|=Ëÿ*óÿvdêO$þðËü$ü?~´/ô‹éÖŸí€ ÚÒ©.àÏŵ±ßÿz-èhh HÞÄ’@3%‚X;ƒÿ¡Ÿfçõ=#H ™FXYù´\èõ A’üá¡ÝùwE†Œ]_2%U5M¥¢ƒ`HHIJªª@0 ÚF,øCÄB‘D¦C!4](€#SaëWªq@ ×M?‘:€ðúƒ4æ¤ ù0d"ÔèoËz¿ìˆà¨ p 2€ðÞUþ¨¢ ¡ ܘ²6@¨>q„šóýQP+b¡C7@¨ÊŸbj‚´GÞÐ; j‡ºBµi C¢Ñ4ü”ƒþþë:0þT*ô&ýØè®~ã ‚Ø›^2F+êDqTÕçB1†³CI…·™ua‘ÛeÌD$`Â>ÒH{eA93¦Ï ®gï㓃I‹éu­1“<¤Øeû0yº¦Ã‹µŠÛúE‹[N-î¨97Ï!a½³t5m´ð†ë±èNá¿è«ø“~n'²n¹?4ò/`·íÊxû/Øì>¼“ŒD<]YÔÂñp/Ü祟,=>w­÷[—ná~I+”ýuáLîz”ßÖðטsŽo5&ù4Ý–¥o<¾Û½½ óNg¢Æ†5þhmñîaêÒ ÓŸ»WÔú–ÆÑÿP±<©XjY¸u\ÄbI€ÖÚHVy(Z.‚Ĉpr„mÊèš“EÜ0ÿ*õÚ¤ÄTä[ôS¬—À-ùäŒZ~Í×–V-3w‚0†ÏLe¢œ‚ÙØ¾Rm¡‚•s¶×+Í”¦Ñô÷½-Lƒ¦¤Só×D¯´Õ¤Uâ¤ÞiZ«f5ÏV&›–=™j¸ ¿æ@\‹¬«Ð9ð›ÊD !†ø"µ%¤kÓšïÏ÷Ïj™´=ȤD¼v¸•hQ7Ymp7ÿt±tûRP«K¢å¾bÚÛƒèR‘[új CûiÆÌ$> Oâ|è†ÑUÞjñª^} ›±Ž¹Ûã3ïWù÷ÞÚ\6uaSqvý}ŽZˆ¾i‡elqí–ô0ú#)–í-¾­^róÍÛíqo2¥c➊ŸL˜¹õ¤òh¿šÈ‡Í.ìgè'ï;Ñ’<5Æ<Ô[V–¥ŠC‰îNr.Î ':«i„|asèHñT¨;¯®°bÝJ >«0)CÖ0R—zpÝ9 VèyÂTžV5›òѲüý$“*êWÀeÔߛ֔Ós75üÙÙoñ¡Ê<ó3{Ä–b"©Ò‰fƒ’&ÈJßÓCV’Òˆê)Vˆéç“ä´=A(yWRcpgÛ¢™ÊbµˆÖ=—]²Y¢Ì[n0É-i·Q:1©`Ú‘›³2ËsD݆‰¡ÙÕ;t&ÏùXËQ‚ôuN"n‹´Ð_ùl÷êÁàLÁÄé.Ê-ñ—³ª_L¾ \ž ÄëX‰?k=Ö]š…{óíç3ëM“u>H†Â¸ý¯dÕ\êíOynl®©Rƒ£Í¶Újë‚Çu{H‡œ éÖû»ä;œÒÚßû|æÐjø`'{pÌÀÜRðµòu_ø‹§E¢GûëCF,˜\¶1…Ek‰^ù·{Gf<²75#Z6¹é\:FÊ­OXû"ŒýÒæølî•‚Ä=J••Œ-ªðñèçm!çÌÖZˆ÷ñÔÎ »ô¶ìzÿÎSª‹í»$ ŸhpYO6ñÈ,â|€òæÇ®X~kàe°ø¤"“]kOLk~ü1žÜ›,J }À YõÑ}ãs<Ø Þ²(jXÏðëš“;+4_,ص¥&ä°ýÓhÆÔu]]¥‰Œô¹x‰ ’W›Lün4Uáu}÷ŸQW8O'å®)ˆ½½;t‡ÇêJø~Ô„Ž¾àüël÷s3öTó<ã»2hî,rµÃéáÍtÙœÝ ’6å\Õi‹QQYyZ¶[f<'=2ÍÒ¢Ø7 繜ÞqæÎ.K#¶3õàœôÃ(!j è‘eIÞœõ¬Òòöüû©»¾lš}V;no/H"ݨʗ¶[©éØ’®ç¡S¼¤:Ï(±(Ʀ —¦—;Çó½©ÌtãÍ—]婪¾;ŽQº©ú1~_d\m÷C WZ¨d×¥vï®Gîf=o‚…¾Ó^²8•*ÑS¿tÑX4HwF,¡ïšâb†å©¿§­¾žÑ77W· ¾ÈýWÇþƒÑ[ö[½Ù~ñQ^PØ\ØõÈ—¥)°216-‡’É/ÇûNwpU—ßa >Ò3﮳>¼—FbV$ÊTfÛWéM ^h†YG×åÙ˜kŠ?ûªi{LSÁ³-úH>´V'íp°é^‹)<f\0ìs‰nÌSn°UšçŒñ¾Ý66ž;÷ôÆ­X©F?uÌÐiÜ#ä¼Ô»YœÍæIÞú~ë¾´^¬#µ¹ÀÂò嬓øŠ5ò±uH[¨Q6>“Ïo,ÐÅLa& o°yæÞyCÞ¡èTöÊ0>Ž­sÌKý¾…{$š»Ï–’£E£HÁQŽåÖß»ÈS‘[Y‘:‚þ®Ü6©=ÜóekæßOÄrÕ8[BL/É\ÙA“Çrˆ]l+ülºÏöˆsbÙò¦Ø)>7Ê˶Óûì£ÅŸ® _ÆÍßy¯–‹|<ßå²ä±”Sêûzí«ì¼jg|ðà²/éÅQʋ⽮ÛÙu݆ä¬ÅJ¯ºË<[½M}Z"Oʱó¼?”WŒfîýu†?QbßJ>á1rÏTø..SĬRmÒðÀþhíhOYåÞçÎ'"ŠQ.왋ÂÔát-3ƒ‰%Ž¥ó™9]£á±¥Obn×íJÆMé»õ^ém_U\*³vmšUðˆ2 ã?‹T~oÔçJˆÉÖÎoúfxÁè E3l!.Õ¶"y­l9Ñ*þ•j澿ï¼â©X.ËG¤2ÜbÍÕ%-9ç٨‹¬iç€PSc¹QšñÄ[Cåû;Txò2L žá€¶è!Õ{µHYÂeJ ï²¢j4ºMpŠ$§tT .¶PÕ©y¨ßñ©æšö yëëäÀ'©dÿßêááR÷x*ÉÛô¹¶{ÿ¹qv}ÀÐÒ<†?Z‰þ(9æPW9ß\ëÞ—àxGæã+÷/33ž+)Ü'äÄÏrà|¿èêœ[°ăQ’ªÏ»f“Ôí0¡Œïsú-Aǃü ·#åV•™nF™CµSE³6Ñç~+§ŸdÐïˆqp”6 mYÏê¼Âr…/®ß›h';¬;íS¤û,ä«…4‚fèÝ ÃÞ¸²7Uc…d‰~å~¹*L~WPƒçpRNß”¹žÎR§:|®ó~Ât^ïLJ{Žõ¶z4”9^ ~ºÙD<âYJ’)òjϰ,Ρps= x+lÙs° ´óû‰ã wß–õs?d›öf y©YÒBGküß¼³¢z?÷°òÕ½ª€î7ÉçVoñÊ Kž fQZmwœ~Ó-±+ýìçö»‘±åvE"Ïxšn¤séù.Ö¾Ñr˜!Væ³Ì'…_É_bE¦¶pÎ÷|$Ì(•æ[9Â>ôå`ÌåɨΔi8²oÀGTCÆ“ïð¹ªó£^óå¢s%—x‹¶‡2"¯­lâöYæñ*ªIIÉro[¾}‚*I”ó7 a¹Ÿ«;ftÊNtW4XÁ²Ô‡ËŸ¤;ƒÂ‹÷(*U¢§6÷\i×Ýf•|„~­€xZk“BeîôÚ».Ô4JÎ(+ÆŸûœ)WE˜Ý’f‹_‰øZ› Š™k‹·WzÊä¸ (èâ¸',|[‹–gµ_ ÿ’ À½¤ýw¯¾0Ã×#ãòüz—íG=Us´Ç(Óf—ÆžæqÓï“7oÛ»f‚ñÞ LïS{k~ ³ìñs[„ÈÜŠ¾|AŽaÑ W%Îá%ˆ§±ýC“/õÚvÇÐ॔Ï{©"‘vÑ)Õã°Õ59gã=÷¨¥ Œ'‹E—é†Å¦ï”{bÃò¢2ܧô?¢îwU]éÔV›ág5ý Üè|1q©ï#{B¼µz²QõBvdò)ÍöØ{KµîTØ #•ç߯Jï¬q_k wb¬Ö<Ö3kÐ98}r³®ÚxîùºÔ}‰¥ÑTmé-Øå¢y¹üøÛ’;J·uÊïˆgeÝ ü¼¤ùæcgâp·íåä´|ƒ†Üú¶ùÈ!‡¥ÅŒy!³â×+W3¥k)G);†Ü%ljÉÁjú‚êÏG”='ã—¿4]GZ(n^Áž—=9¢{(KªkeæõWÎ.aäÇù箑yÎßù-*-¦Õ‹r„ƒÌL³®šg:²ª„u”Øýè‘Äñç¹Õ°esTÎv§÷+›“‡MJVÞìM8t$ªÿLµî¨ègb£ ½W­`¼…ÙØWº91AI»Ò=nB„Ñðy:‚“®Xu/D£¥™Çº±‰ß§Êþî ÷œoó—"÷²Dêcè#˜ëC[ÃMDÊsžÔ ð»í“òîÏ “ÅÇBJê'Dez™¯š¹˜Ñ¨xðük ¤¤ —r•ýûðHSÔ51T®-ƒ{—¹oµ«ÂiÒîô¼Lxp[òmê0j.êðƒ¬WçøÌ“\\@c%±5ì.˜!í—Ò_ò9ï|ßè{.yÈ4‰]],Iî~ä×Ñ<á·WÄÓž½XU˜hnÌdΰžW´ª†Žo÷Ͷ·9+c.xÏŽ½ˆmõ—M‰œ¼½U=6ËÏ+n”Sч d\^°ïöVĸæGfwVepYì)ãÞ¥T~Åum0;ùï¯ÚÕ/ll.vݰ¾™½’ÙË­òภœÜ„çY_±Ç²òaNœšT™ÎU¡†­m³»œþѼģ6¨Ô÷E–¸‘$ ¯sïœÓh<'ݬ²¶™Ïûpà C½Šî-®æ™]Ÿ>ƒ©ÿ¡ˆÚë endstream endobj 78 0 obj << /Author()/Title()/Subject()/Creator(LaTeX with hyperref package)/Producer(pdfTeX-1.40.18)/Keywords() /CreationDate (D:20191029194128-04'00') /ModDate (D:20191029194128-04'00') /Trapped /False /PTEX.Fullbanner (This is pdfTeX, Version 3.14159265-2.6-1.40.18 (TeX Live 2017/Debian) kpathsea version 6.2.3) >> endobj 2 0 obj << /Type /ObjStm /N 63 /First 477 /Length 2884 /Filter /FlateDecode >> stream xÚíZ[sÛ6~ׯÀc¼;!ˆ ¢“éŒ/q’&i_šÛø–h™$ª$•&ûë÷;%‘´ÕØMwfw§3çà|瀤`1K˜Š™aIÂK&b&$.‚‰Ä¢Êdê˜PLthfb\,³ÊŒDÊl3ÅR©™e©I1„9 Qƒ‰D‹ÐŠ)FÖ2¾J & ®ý ãt¤ˆu‚q”$%™0¸ÇPa­cZ3‘jË4Æ;‰{Œw:e¬„ÕLY&%¦£R\¡K*0Ñ"ÇÌR@‡D`*˜œ…Á$Jap¢™¶V‰ŒS–PS* µÊŽŒ``F1“Ä3VQÁŒsÁ<Àš˜&dŒÁ$˜u“ŒY*MX)7R†¥ Ì”f©NA«A? I­Ž KS±’¥ÌR\,Ôe.”VáŠQ褶# Ð Mk˜Ó€@;öG?eüIyV2~ÄÔù¸)ÊE$öØ?޼>ÜgUy¹ªÖ”¬\5³"¯ê½Ãdöª%dÙbÂæE]‹)ûœÍV9+æËU“ñ .áO1;aü¸h.ˆÏ£G#~öu™3þ:›æ#~X.š|ÑÔ0<ŽøI^—«jœ×äT¾åe>)²ƒò û“Ûº$’P9Õ"JÝÅ|*0 ì<õþbQ‚ÝGra’ óû‹ö—‹QožxÄÊj’Wž½¸àOù3~ˆ8óÍgÜ@ +£<m¢XkGi Ï12Ò&Ýéê²Oþ¢X|âûy |ßcÈOùùÉ3ú?¸nšåœ_e±¸*«9`×Ñ| NËÙÕi4ÉùæyÛT[1ÌËÙ=óx3s­\dcøœ1‘sS&2ÀOÙ$Ò±Ý>š~\4ytýþ©XÔqp›¿d.B=x¬J’(¦àV*Jüˆ‰y`÷\fÅ*VÝyt<ëÝû°µ‚9}‰Š¢{±šÍ.vÒê@+L;y'Ú$Qœ¤w¢•NEÈF}Úc¸¹Wí˜r¢kc·u ÿu¬ …IÙÖ)¥­›e ªƒ+]•ãÓ¨"”ŽŽ?Ë¿4CÿĘ’ÃSñwÅýƒ{Â[÷B ˜(­»(E®Fj4r»r"²BÝêWÅtUå‘8‡ƺt§s¨[Œ¨AìÝhrƒ»©LÓÈbÕÙá['~•Ú½ï`¤Ì÷øƒŽoøCú'ü¡qXå~8Êšœ=8úAÆÂa;à°dc°¦ådgßYÑÌÐsžTÙò9‘aÁÁ²²ç›¬ÆyE½*2°u+«¤¦½¾†fÇõ¨QŒøÏÙœúP;ÈêÜÎ?dË«#,b—YS߃‡ìòxšÏ>çÈàÙˆ?^ŒË ­‰Ú¬AùÈŸñÛ j¼èYÓv›Ñüm±Ø_ÔŶû¨¸ºÊ7™ç#¶B|^,V5s†ÿ¶*›|–_5XÝ4Ÿ”@Ë2ŸVÙçœgãU“óqQWó«Yþ…7Å ëÊ<WX.«4’ǰ$Ÿ`iÏë¢æBw’_ñ ²ù^0›e›ÆëÕbšU«ù,[5¼œ–‹ügį^fã¼çýîîÙ¹GkÒHaG)iÁÓoF yþ»W—¿R¡žgsl»Ü‚â#v„‘‚aã‡Ýí]ÿaD·´¾Ô˜±lËÐb±ÿ´1ôÁ†Î·b™s̤q„­c"‰"EjòtʈÛ Aa´‘´©6mÛæÇ¶Ü•¡XJÍš?ÑáÀÚˆÇßb½UtXë@à™i­#Ý««˜ê[Š0¢m•[©‡4Ðl©…Oâ´ý&~Hƒ®×u_^Œ>&’¶åÿÿCÿÿ_„I³´ã‰¾ ¾™v‚¾ãÕƒ?EMt'·!M÷.¦Ðq'V‡Ô¦֥ Î߆@÷ûþÎIÿÅ9iëäa8ù%ì¾¥RKòKã%H’/é>ÕýÖ9š¿§M1šž`C óY¢ƒ™àÏØ*$’T -&ƒàd縱‘h RL6ÄÚš´õÇBáíiSa!@{:,„‘m5”&µ›Y“œ¶vC#:;Ï”öWAè+­€s §Öµþ¾ÕQZœ²-Å–6ôë„æ¡1VØRQÙBÐäÿ.ÿGJXLyO ¥NÉE-’ E1ÎSÒÀ!©z¾§$.ë é‰M[†‹ƒ–'­#ËjŸÖŒAݨ6ÊS)l`éa€ò#¥ÏbþRk~adË[RrLô†7Qá/âØln—öî×¶wÄF«„¦°©I$‰n¿áÛäi¿F iÇn¨BßÚ4ÝPßö…醺/a4 ÁŒT­ %%D:RR÷ýÔF+EJê†:xS*ˆ)9(CéMáÜiI°&aqÐÞŒIJ8ܨw[g“Ûtë‰0íÑ´cC»_çB‹Å )Öc»<= ‘ôÒ5Ô}ïÅàÈx”×ãªXâXN*áÀøüààÍ«ÿ<|yðNàìx<˦8xŠÂ~ˆ#ÑC™ä]ò¶„ˆì×t, <Ì–OóbzÛÔàÀ”‡¾‡‚:Ÿ5Ù¬ï/¦8=Ç8´6ùüx—騄E <®³ŠÎAø>?䝸k~Â3~ÉÇ|Âs~ŧ¼à3>ç ^rëø’W¼æ o~/ùŠÞ ó=. %1ÝCÿ·Ôúüùãç¡þ‰ˆo×^Ǥ=íâoi¯î¤½é*¯ÄPù#~ÌŸøçOÏù þ’ÿÜÂqÊÏø/ü-﹬²ñ§¼¡Óõº^G 6.g8ñŽËù<óèå‹IV_óü·U6ãù—ñ,›Ñ+úÜÿpÔžòk~ýuy/€ô¯üÓí%=PñÂ|-ˆZæ•?œãZ”þ›·IÏ‹0ƒGx”Å^Ïh°Öu•çk›ñßùþ•ÿ+¯ÊžùÒû˜ïlÿüÍËso¾΋üå×¥é_c½ä¬wûÛ½ÜXímëÈ[ƒ\õÑ^c½(õ·' »®„&¾„ï÷Þ|xç!´; T!œFFÕîÎ&; LT¸ áS€ÕyòÁ ðª§£¼Ž'ïŽß¿}O_ì s0 a.èíØŸw”‡n­¦ÝçGÞ3~¤¸G¬coÙ·8Í×ú> <þéɇ÷焳³]0HÊv0¸ôÑtP¢‡ÝnPRîBá›Ùî1òÝ0Ï;Çm²›OkWë,6¥¯yÕ¯§çŽwJg>ÑSËÉå¬ úÍ»×"s~ôêÃé ¿w`OÔwÁ~×5&à¼õ¾¹w¶Â=Mï•ߟ½ÙšîHÈ‚4ÅFHÒ®6ù~MwçãuÞ¥Åè–¼{k”µKU»#Þ|¿}ð>ÜUõ£÷øþ-ÆqQÕ Í—iØëEÖÞ‘ŽøÛbÒ\×ôIÀ½…ö4Ùn(›€ÞÊ–j+[Ü_ö`AîË6â†ÞqO¶ØÊÖ÷—=XɲÕP¶•=ÌíVvrÙÃf <¹¡x²CqÞ_øÍÌ>Ãß”Þ%ÞÝ_ü0½ „ßp8“ìrö?iÃŒÓnÅ/ÌV¸º)œÞGÖôBrQŒçŤ޼Uo_ †·Cƒ™þk”õ’Yðf/˜µ¯0S¹ç_0¶Xí³ök“öE¦m_Œ¿®òϬ~Øú½7·û™Þ*õüв54ypTŽž6YÕì1Þ•ßø¸BšN{û¡ƒ´¡mûv[µ-K F÷¢{/÷èS+QÌ‹f zMu1œžo‡< ^óó/1éµù¦~ ÄÚ¡Ô5ÇÑFj°® \¬þ£ ß2^2{Kî8̰p•ÓQp§ÍW kÙšÉkm7¦Ÿæ/ËIÎÏë|ëN¯–ù"|´C_ĵ¢þ ­Þ¶- endstream endobj 79 0 obj << /Type /XRef /Index [0 80] /Size 80 /W [1 3 1] /Root 77 0 R /Info 78 0 R /ID [ ] /Length 214 /Filter /FlateDecode >> stream xÚÑ9N‚a…ás~GÅpçyé,Il°±r ®ÃÄİ+]€»p&&6.}oóäý’“Û|’4L¤Döû‹¨)¸C¸€…1‡ ˜„¬¥\\IAáNaöáÎ`гŠõ8P‚i+]ŽgŽ¡G° +0kîc·sP†yX€E«ú“ TaÖ uhÀºÕü‰ñ†ÕÿŒÚ´^o£¶¬ï§¨¦ùjÙí¯¨mûñ9jÇÜEíÂ9\A:p ]û­÷ÿ7úh endstream endobj startxref 114757 %%EOF pcaMethods/inst/doc/pcaMethods.R0000644000175400017540000001313213556147214017633 0ustar00biocbuildbiocbuild### R code from vignette source 'pcaMethods.Rnw' ### Encoding: UTF-8 ################################################### ### code chunk number 1: pcaMethods.Rnw:102-114 ################################################### library(pcaMethods) x <- c(-4,7); y <- c(-3,4) distX <- rnorm(100, sd=0.3)*3 distY <- rnorm(100, sd=0.3) + distX * 0.3 mat <- cbind(distX, distY) res <- pca(mat, nPcs=2, method="svd", center=F) loading <- loadings(res)[1,] grad <- loading[2] / loading[1] if (grad < 0) grad <- grad * -1 lx <- c(-4,7) ly <- c(grad * -4, grad * 7) ################################################### ### code chunk number 2: pcaMethods.Rnw:118-125 ################################################### par(mar=c(2, 3, 2, 2)) plot(x,y, type="n", xlab="", ylab="") abline(v=0, col="dark gray", lwd = 2); abline(h=0, col = "dark gray", lwd = 2) points(distX, distY, type = 'p', col = "blue") lines(lx,ly, lwd = 2) points(-1, -1 * grad + 0.5, pch = 19, col = "red", lwd=4) points(6, 6 * grad + 0.5, pch = 19, col = "red", lwd=4) ################################################### ### code chunk number 3: pcaMethods.Rnw:253-255 ################################################### library(lattice) library(pcaMethods) ################################################### ### code chunk number 4: pcaMethods.Rnw:258-261 ################################################### library(pcaMethods) data(metaboliteData) data(metaboliteDataComplete) ################################################### ### code chunk number 5: pcaMethods.Rnw:264-266 ################################################### md <- prep(metaboliteData, scale="none", center=TRUE) mdC <- prep(metaboliteDataComplete, scale="none", center=TRUE) ################################################### ### code chunk number 6: pcaMethods.Rnw:271-277 ################################################### resPCA <- pca(mdC, method="svd", center=FALSE, nPcs=5) resPPCA <- pca(md, method="ppca", center=FALSE, nPcs=5) resBPCA <- pca(md, method="bpca", center=FALSE, nPcs=5) resSVDI <- pca(md, method="svdImpute", center=FALSE, nPcs=5) resNipals <- pca(md, method="nipals", center=FALSE, nPcs=5) resNLPCA <- pca(md, method="nlpca", center=FALSE, nPcs=5, maxSteps=300) ################################################### ### code chunk number 7: pcaMethods.Rnw:293-296 ################################################### sDevs <- cbind(sDev(resPCA), sDev(resPPCA), sDev(resBPCA), sDev(resSVDI), sDev(resNipals), sDev(resNLPCA)) matplot(sDevs, type = 'l', xlab="Eigenvalues", ylab="Standard deviation of PC", lwd=3) legend(x="topright", legend=c("PCA", "PPCA", "BPCA", "SVDimpute","Nipals PCA","NLPCA"), lty=1:6, col=1:6, lwd=3) ################################################### ### code chunk number 8: pcaMethods.Rnw:308-311 ################################################### par(mfrow=c(1,2)) plot(loadings(resBPCA)[,1], loadings(resPCA)[,1], xlab="BPCA", ylab="classic PCA", main = "Loading 1") plot(loadings(resBPCA)[,2], loadings(resPCA)[,2], xlab="BPCA", ylab="classic PCA", main = "Loading 2") ################################################### ### code chunk number 9: pcaMethods.Rnw:335-337 ################################################### q2SVDI <- Q2(resSVDI, mdC, fold=10) q2PPCA <- Q2(resPPCA, mdC, fold=10) ################################################### ### code chunk number 10: pcaMethods.Rnw:339-349 ################################################### # PPCA does not converge / misestimate a value in very rare cases. # This is a workaround to avoid that such a case will break the # diagram displayed in the vignette. # From the 2.0 release of bioconductor on, the convergence threshold # for PPCA was lowert to 1e-5, this should make the method much more # stable. So this workaround might be obsolete now... # [nope it is not, ppca is unstable] while( sum((abs(q2PPCA)) > 1) >= 1 ) { q2PPCA <- Q2(resPPCA, mdC, fold=10) } ################################################### ### code chunk number 11: pcaMethods.Rnw:353-356 ################################################### q2 <- data.frame(Q2=c(drop(q2PPCA), drop(q2SVDI)), method=c("PPCA", "SVD-Impute")[gl(2, 5)], PC=rep(1:5, 2)) print(xyplot(Q2~PC|method, q2, ylab=expression(Q^2), type="h", lwd=4)) ################################################### ### code chunk number 12: pcaMethods.Rnw:389-390 ################################################### errEsti <- kEstimate(md, method = "ppca", evalPcs=1:5, nruncv=1, em="nrmsep") ################################################### ### code chunk number 13: pcaMethods.Rnw:396-397 ################################################### barplot(drop(errEsti$eError), xlab="Loadings", ylab="NRMSEP (Single iteration)") ################################################### ### code chunk number 14: pcaMethods.Rnw:420-421 ################################################### barplot(drop(errEsti$variableWiseError[, which(errEsti$evalPcs == errEsti$bestNPcs)]), xlab="Incomplete variable Index", ylab="NRMSEP") ################################################### ### code chunk number 15: pcaMethods.Rnw:445-446 ################################################### slplot(resPCA) ################################################### ### code chunk number 16: pcaMethods.Rnw:456-457 ################################################### plotPcs(resPPCA, pc=1:3, type="score") ################################################### ### code chunk number 17: pcaMethods.Rnw:469-475 ################################################### pc <- pca(iris) irdf <- merge(iris, scores(pc), by=0) library(ggplot2) ggplot(irdf, aes(PC1, PC2, colour=Species)) + geom_point() + stat_ellipse() pcaMethods/inst/doc/pcaMethods.Rnw0000644000175400017540000005554213556116437020216 0ustar00biocbuildbiocbuild\documentclass[a4paper]{article} %\VignetteIndexEntry{Introduction} \usepackage{hyperref} \title{The pcaMethods Package} \author{Wolfram Stacklies and Henning Redestig\\ CAS-MPG Partner Institute for Computational Biology (PICB)\\ Shanghai, P.R. China \\ and\\ Max Planck Institute for Molecular Plant Physiology\\ Potsdam, Germany\\ \url{http://bioinformatics.mpimp-golm.mpg.de/} } \date{\today} \begin{document} \setkeys{Gin}{width=1.0\textwidth} @ \maketitle \section*{Overview} The \texttt{pcaMethods} package \cite{stacklies07} provides a set of different PCA implementations, together with tools for cross validation and visualisation of the results. The methods basically allow to perform PCA on incomplete data and thus may also be used for missing value estimation. When doing PCA one assumes that the data is restricted to a subspace of lower dimensionality, e.g. correlation patterns between jointly regulated genes. PCA aims to extract these structures thereby filtering noise out. If only the most significant loadings (eigenvectors, also referred to as principal components) are used for projection this can be written as: \begin{equation} X = 1\times{}\bar{x}^T + TP^T + V \end{equation} Where the term $1\times{}\bar{x}^T$ represents the original variable averages, $X$ denotes the observations, $T={t_1, t_2,\ldots,t_k}$ the latent variables or scores, $P={p_1, p_2,\ldots,p_k}$ the transformation matrix (consisting of the most significant eigenvectors of the covariance matrix) and $V$ are the residuals. Missing values may be estimated by projecting the scores back into the original space using $\hat{X} = 1\times{}\bar{x}^T + TP^T$. Optimally, this produces an estimate of the missing data based on the underlying correlation structure, thereby ignoring noise. This will only produce reasonable results if the residuals $V$ are sufficiently small, implying that most of the important information is captured by the first $k$ components. In order to calculate the transformation matrix $P$ one needs to determine the covariance matrix between variables or alternatively calculate $P$ directly via SVD. In both cases, this can only be done on complete matrices. However, an approximation may be obtained by use of different regression methods. The PCA methods provided in this package implement algorithms to accurately estimate the PCA solution on incomplete data. Although the focus of this package is clearly to provide a collection of PCA methods we also provide a cluster based method for missing value imputation. This allows to better rate and compare the results. \section{Algorithms} All methods return a common class called \texttt{pcaRes} as a container for the results. This guarantees maximum flexibility for the user. A wrapper function called \texttt{pca()} is provided that receives the desired type of pca as a string. \subsection*{svdPca} This is a wrapper function for $R's$ standard \texttt{prcomp} function. It delivers the results as a \texttt{pcaRes} object for compatibility with the rest of the package. \subsection*{svdImpute} This implements the SVDimpute algorithm as proposed by Troyanskaya et~al \cite{troyanskaya01}. The idea behind the algorithm is to estimate the missing values as a linear combination of the $k$ most significant eigengenes\footnote{The term ``eigengenes'' denotes the loadings when PCA was applied considering variables (here the genes) as observations.}. The algorithm works iteratively until the change in the estimated solution falls below a certain threshold. Each step the eigengenes of the current estimate are calculated and used to determine a new estimate. An optimal linear combination is found by regressing an incomplete variable against the $k$ most significant eigengenes. If the value at position $j$ is missing, the $j^{th}$ value of the eigengenes is not used when determining the regression coefficients.\\ SVDimpute seems to be tolerant to relatively high amount of missing data (> 10\%). \subsection*{Probabilistic PCA (ppca)} Probabilistic PCA combines an EM approach for PCA with a probabilistic model. The EM approach is based on the assumption that the latent variables as well as the noise are normal distributed. In standard PCA data which is far from the training set but close to the principal subspace may have the same reconstruction error, see Figure \ref{fig:pcaSubspace} for explanation. <>= library(pcaMethods) x <- c(-4,7); y <- c(-3,4) distX <- rnorm(100, sd=0.3)*3 distY <- rnorm(100, sd=0.3) + distX * 0.3 mat <- cbind(distX, distY) res <- pca(mat, nPcs=2, method="svd", center=F) loading <- loadings(res)[1,] grad <- loading[2] / loading[1] if (grad < 0) grad <- grad * -1 lx <- c(-4,7) ly <- c(grad * -4, grad * 7) @ \begin{figure} \centering <>= par(mar=c(2, 3, 2, 2)) plot(x,y, type="n", xlab="", ylab="") abline(v=0, col="dark gray", lwd = 2); abline(h=0, col = "dark gray", lwd = 2) points(distX, distY, type = 'p', col = "blue") lines(lx,ly, lwd = 2) points(-1, -1 * grad + 0.5, pch = 19, col = "red", lwd=4) points(6, 6 * grad + 0.5, pch = 19, col = "red", lwd=4) @ \caption{Normal distributed data with the first loading plotted in black. The two red points have the same reconstruction error because PCA does not define a density model. Thus the only measure of how well new data fits the model is the distance from the principal subspace. Data points far from the bulk of data but still close to the principal subspace will have a low reconstruction error. \label{fig:pcaSubspace}} \end{figure} PPCA defines a likelihood function such that the likelihood for data far from the training set is much lower, even if they are close to the principal subspace. This allows to improve the estimation accuracy.\\ PPCA is tolerant to amounts of missing values between 10\% to 15\%. If more data is missing the algorithm is likely not to converge to a reasonable solution. The method was implemented after the draft ``\texttt{EM Algorithms for PCA and Sensible PCA}'' written by Sam Roweis and after the Matlab \texttt{ppca} script implemented by \emph{Jakob Verbeek}\footnote{\url{http://lear.inrialpes.fr/~verbeek/}}. Please check also the PPCA help file. \subsection*{Bayesian PCA (bpca)} Similar to probabilistic PCA, Bayesian PCA uses an EM approach together with a Bayesian model to calculate the likelihood for a reconstructed value.\\ The algorithm seems to be tolerant to relatively high amounts of missing data (> 10\%). Scores and loadings obtained with Bayesian PCA slightly differ from those obtained with conventional PCA. This is because BPCA was developed especially for missing value estimation and is based on a variational Bayesian framework (VBF), with automatic relevance determination (ARD). In BPCA, ARD leads to a different scaling of the scores, loadings and eigenvalues when compared to standard PCA or PPCA. The algorithm does not force orthogonality between loadings. However, the authors of the BPCA paper found that including an orthogonality criterion made the predictions worse. They also state that the difference between ``real'' and predicted Eigenvalues becomes larger when the number of observation is smaller, because it reflects the lack of information to accurately determine true loadings from the limited and noisy data. As a result, weights of factors to predict missing values are not the same as with conventional PCA, but the missing value estimation is improved. BPCA was proposed by Oba et~al \cite{oba03}. The method available in this package is a port of the \texttt{bpca} Matlab script also provided by the authors\footnote{ \url{http://hawaii.aist-nara.ac.jp/\%7Eshige-o/tools/}}. \subsection*{Inverse non-linear PCA (NLPCA)} NLPCA \cite{scholz05} is especially suitable for data from experiments where the studied response is non-linear. Examples of such experiments are ubiquitous in biology -- enzyme kinetics are inherently non-linear as are gene expression responses influenced by the cell cycle or diurnal oscillations. NLPCA is based on training an auto-associative neural network composed of a component layer which serves as the ``bottle-neck'', a hidden non-linear layer and an output layer corresponding to the reconstructed data. The loadings can be seen as hidden in the network. Missing values in the training data are simply ignored when calculating the error during back-propagation. Thus NLPCA can be used to impute missing values in the same way as for conventional PCA. The only difference is that the loadings $P$ are now represented by a neural network.\\ A shortcoming of the current implementation is that there is no reasonable stop criterion. The quality of the estimated solution depends on the number of iterations. This should in most cases be somewhat between 500 and 1500. We recommend to use \texttt{kEstimate} or \texttt{kEstimateFast} to determine this parameter. \subsection*{Nipals PCA} Nipals (Nonlinear Estimation by Iterative Partial Least Squares) \cite{wold66} is an algorithm at the root of PLS regression which can execute PCA with missing values by simply leaving those out from the appropriate inner products. It is tolerant to small amounts (generally not more than 5\%) of missing data. \subsection{Local least squares (LLS) imputation} The package provides an algorithm called \texttt{llsImpute} for missing value estimation based on a linear combination of the $k$ nearest neighbours of an incomplete variable (in Microarray experiments normally a gene). The distance between variables is defined as the absolute value of the Pearson, Spearman or Kendall correlation coefficient. The optimal linear combination is found by solving a local least squares problem as described in \cite{kim05}. In tests performed in the cited paper the llsImpute algorithm is able to outperform knnImpute\cite{troyanskaya01} and competes well with BPCA. In the current implementation two slightly different ways for missing value estimation are provided. The first one is to restrict the neighbour searching to the subset of complete variables. This is preferable when the number of incomplete variables is relatively small. The second way is to consider all variables as candidates. Here, missing values are initially replaced by the columns wise mean. The method then iterates, using the current estimate as input for the LLS regression until the change between new and old estimate falls below a certain threshold (0.001). \section{Getting started} \paragraph{Installing the package.} To install the package first download the appropriate file for your platform from the Bioconductor website (\url{http://www.bioconductor.org/}). For Windows, start \texttt{R} and select the \texttt{Packages} menu, then \texttt{Install package from local zip file}. Find and highlight the location of the zip file and click on \texttt{open}. For Linux/Unix, use the usual command \texttt{R CMD INSTALL} or set the option \texttt{CRAN} to your nearest mirror site and use the command \texttt{install.packages} from within an \texttt{R} session. \paragraph{Loading the package:} To load the \texttt{pcaMethods} package in your \texttt{R} session, type \texttt{library(pcaMethods)}. \paragraph{Help files:} Detailed information on \texttt{pcaMethods} package functions can be obtained from the help files. For example, to get a description of \texttt{bpca} type \texttt{help("bpca")}. \paragraph{Sample data:} Two sample data sets are coming with the package. \texttt{metaboliteDataComplete} contains a complete subset from a larger metabolite data set. \texttt{metaboliteData} is the same data set but with 10 \% values removed from an equal distribution. \section{Some examples} <>= library(lattice) library(pcaMethods) @ To load the package and the two sample data sets type: <>= library(pcaMethods) data(metaboliteData) data(metaboliteDataComplete) @ Now centre the data <<>>= md <- prep(metaboliteData, scale="none", center=TRUE) mdC <- prep(metaboliteDataComplete, scale="none", center=TRUE) @ Run SVD pca, PPCA, BPCA, SVDimpute and nipalsPCA on the data, using the \texttt{pca()} wrapper function. The result is always a \texttt{pcaRes} object. <>= resPCA <- pca(mdC, method="svd", center=FALSE, nPcs=5) resPPCA <- pca(md, method="ppca", center=FALSE, nPcs=5) resBPCA <- pca(md, method="bpca", center=FALSE, nPcs=5) resSVDI <- pca(md, method="svdImpute", center=FALSE, nPcs=5) resNipals <- pca(md, method="nipals", center=FALSE, nPcs=5) resNLPCA <- pca(md, method="nlpca", center=FALSE, nPcs=5, maxSteps=300) @ Figure \ref{fig:eigenvalues} shows a plot of the eigenvalue structure (\texttt{sDev(pcaRes)}). If most of the variance is captured with few loadings PCA is likely to produce good missing value estimation results. For the sample data all methods show similar eigenvalues. One can also see that most of the variance is already captured by the first loading, thus estimation is likely to work fine on this data. For BPCA, the eigenvalues are scaled differently for reasons discussed above, see Figure \ref{fig:loadingBPCA}. The order of the loadings remains the same. \begin{figure} \centering <>= sDevs <- cbind(sDev(resPCA), sDev(resPPCA), sDev(resBPCA), sDev(resSVDI), sDev(resNipals), sDev(resNLPCA)) matplot(sDevs, type = 'l', xlab="Eigenvalues", ylab="Standard deviation of PC", lwd=3) legend(x="topright", legend=c("PCA", "PPCA", "BPCA", "SVDimpute","Nipals PCA","NLPCA"), lty=1:6, col=1:6, lwd=3) @ \caption{Eigenvalue structure as obtained with different methods\label{fig:eigenvalues}} \end{figure} To get an impression of the correctness of the estimation it is a good idea to plot the scores / loadings obtained with classical PCA and one of the probabilistic methods against each other. This of course requires a complete data set from which data is randomly removed. Figure \ref{fig:loadingBPCA} shows this for BPCA on the sample data. \begin{figure} \centering <>= par(mfrow=c(1,2)) plot(loadings(resBPCA)[,1], loadings(resPCA)[,1], xlab="BPCA", ylab="classic PCA", main = "Loading 1") plot(loadings(resBPCA)[,2], loadings(resPCA)[,2], xlab="BPCA", ylab="classic PCA", main = "Loading 2") @ \caption{Loading 1 and 2 calculated with BPCA plotted against those calculated with standard PCA. \label{fig:loadingBPCA}} \end{figure} \section{Cross validation} \texttt{Q2} is the goodness measure used for internal cross validation. This allows to estimate the level of structure in a data set and to optimise the choice of number of loadings. Cross validation is performed by removing random elements of the data matrix, then estimating these using the PCA algorithm of choice and then calculating $Q^2$ accordingly. At the moment, cross-validation can only be performed with algorithms that allow missing values (i.e. not SVD). Missing value independent cross-validation is scheduled for implementation in later versions. $Q^2$ is defined as following for the mean centered data (and possibly scaled) matrix $X$. $$\mathrm{SSX}=\sum (x_{ij})^2$$ $$\mathrm{PRESS}=\sum (x_{ij} - \hat{x}_{ij})^2$$ $$Q^2=1 - \mathrm{PRESS}/\mathrm{SSX}$$ The maximum value for $Q^2$ is thus 1 which means that all variance in $X$ is represented in the predictions; $X=\hat{X}$. <>= q2SVDI <- Q2(resSVDI, mdC, fold=10) q2PPCA <- Q2(resPPCA, mdC, fold=10) @ <>= # PPCA does not converge / misestimate a value in very rare cases. # This is a workaround to avoid that such a case will break the # diagram displayed in the vignette. # From the 2.0 release of bioconductor on, the convergence threshold # for PPCA was lowert to 1e-5, this should make the method much more # stable. So this workaround might be obsolete now... # [nope it is not, ppca is unstable] while( sum((abs(q2PPCA)) > 1) >= 1 ) { q2PPCA <- Q2(resPPCA, mdC, fold=10) } @ \begin{figure}[!ht] \centering <>= q2 <- data.frame(Q2=c(drop(q2PPCA), drop(q2SVDI)), method=c("PPCA", "SVD-Impute")[gl(2, 5)], PC=rep(1:5, 2)) print(xyplot(Q2~PC|method, q2, ylab=expression(Q^2), type="h", lwd=4)) @ \caption{Boxplot of the \texttt{Q2} results for BPCA, Nipals PCA, SVDimpute and PPCA. PPCA and SVDimpute both deliver better results than BPCA and Nipals in this example.\label{fig:Q2}} \end{figure} The second method called \texttt{kEstimate} uses cross validation to estimate the optimal number of loadings for missing value estimation. The \texttt{NRMSEP} (normalised root mean square error of prediction) \cite{feten05} or Q2 can be used to define the average error of prediction. The NRMSEP normalises the square difference between real and estimated values for a certain variable by the variance within this variable. The idea behind this normalisation is that the error of prediction will automatically be higher if the variance is higher. The \texttt{NRMSEP} for mean imputation is $\sqrt{\frac{nObs}{nObs - 1}}$ when cross validation is used, where $nObs$ is the number of observations. The exact definition is: \begin{equation} NRMSEP_k = \sqrt{\frac{1}{g} \sum_{j \in G} \frac{\sum_{i \in O_j} (x_{ij} - \hat{x}_{ijk})^2}{o_j s_{x_j}^2}} \end{equation} where $s^2_{x_j} = \sum_{i=1}^n (x_{ij} - \overline{x}_j)^2 / (n - 1)$, this is the variance within a certain variable. Further, $G$ denotes the set of incomplete variables, $g$ is the number of incomplete varialbes. $O_j$ is the set of missing observations in variable $j$ and $o_j$ is the number of missing observations in variable $j$. $\hat{x}_{ijk}$ stands for the estimate of value $i$ of variable $j$ using $k$ loadings. See Figure \ref{fig:kEstimate} for an example. The NRMSEP should be the error measure of choice. But if the number of observations is small, the variance within a certain variable may become and unstable criterion. If so or if variance scaling was applied we recommend to use Q2 instead. <>= errEsti <- kEstimate(md, method = "ppca", evalPcs=1:5, nruncv=1, em="nrmsep") @ \begin{figure}[!ht] \centering \begin{minipage}[c]{0.6\textwidth} \centering <>= barplot(drop(errEsti$eError), xlab="Loadings", ylab="NRMSEP (Single iteration)") @ \end{minipage} \begin{minipage}[c]{0.3\textwidth} \caption{Boxplot showing the \texttt{NRMSEP} versus the number of loadings. In this example only 1 iteration of the whole cross validation were performed. It is normally advisable to do more than just one iteration. \label{fig:kEstimate}} \end{minipage} \end{figure} \texttt{kEstimate} also provides information about the estimation error for individual variables. The $Q^2$ distance or the NRMSEP are calculated separately for each variable. See the manpage for \texttt{kEstimate} and \texttt{kEstimateFast} for details. Plotting the variable - wise results gives information about for which variables missing value estimation makes sense, and for which no imputation or mean imputation is preferable, see Figure \ref{fig:variableWiseError}. If you are not interested in variable - wise information we recommend to use the faster \texttt{kEstimateFast} instead. \begin{figure}[!ht] \centering \begin{minipage}[c]{0.6\textwidth} \centering <>= barplot(drop(errEsti$variableWiseError[, which(errEsti$evalPcs == errEsti$bestNPcs)]), xlab="Incomplete variable Index", ylab="NRMSEP") @ \end{minipage} \begin{minipage}[c]{0.3\textwidth} \caption{Boxplot showing the \texttt{NRMSEP} for all incomplete variables in the data set. For the first 7 variables missing value imputation does not seem to make too much sense. \label{fig:variableWiseError}} \end{minipage} \end{figure} \newpage \section{Visualisation of the results} \subsection{Quick scores and loadings plot} Some methods for display of scores and loadings are also provided. The function \texttt{slplot()} aims to be a simple way to quickly visualise scores and loadings in an intuitive way, see Figure \ref{fig:slplot}. Barplots are provided when plotting only one PC and colours can be specified differently for the scores and loadings plots. For a more specific scatter plot it is however recommended to access scores and loadings slots and define own plot functions. \begin{figure}[!h] \centering <>= slplot(resPCA) @ \caption{\texttt{slplot} for scores and loadings obtained with classical SVD based PCA. \label{fig:slplot}} \end{figure} \noindent Another method called \texttt{plotPcs()} allows to visualise many PCs plotted against each other, see Figure \ref{fig:plotPcs}. \begin{figure}[!ht] \centering <>= plotPcs(resPPCA, pc=1:3, type="score") @ \caption{A plot of score 1:3 for PPCA created with \texttt{plotPcs()} \label{fig:plotPcs}} \end{figure} \subsection{Using ggplot2} For using ggplot, the scores and loadings should best be added to a data frame that add other relevant descriptive factors. For example, after doing PCA on the Iris dataset, we may add the scores back to the original data frame and use ggplot to visualise, see Figure \ref{fig:ggplot}. \begin{figure}[!ht] \centering <>= pc <- pca(iris) irdf <- merge(iris, scores(pc), by=0) library(ggplot2) ggplot(irdf, aes(PC1, PC2, colour=Species)) + geom_point() + stat_ellipse() @ \caption{Score plot using ggplot2} \label{fig:ggplot} \end{figure} \cleardoublepage \begin{thebibliography}{2006} \bibitem{stacklies07} Stacklies W., Redestig H., Scholz M., and Walther D., and Selbig J. {\sl pcaMethods -- a Bioconductor package providing PCA methods for incomplete data} Bioinformatics. 2007, 23, 1164-1167. {\sl Non-linear PCA: a missing data approach.} Bioinformatics. 2005, 21, 3887-3895. \bibitem{scholz05} Scholz, M. , Kaplan, F., Guy, C.L., Kopka, J. and Selbig, J. {\sl Non-linear pca: a missing data approach.} Bioinformatics. 2005, 21, 3887-3895. \bibitem{troyanskaya01} Troyanskaya O. and Cantor M. and Sherlock G. and Brown P. and Hastie T. and Tibshirani R. and Botstein D. and Altman RB. {\sl Missing value estimation methods for DNA microarrays.} Bioinformatics. 2001 Jun;17(6):520-525. \bibitem{feten05} Feten G. and Almoy T. and Aastveit A.H. {\sl Prediction of Missing Values in Microarray and Use of Mixed Models to Evaluate the Predictors.}, Stat. Appl. Genet. Mol. Biol. 2005;4(1):Article 10 \bibitem{oba03} Oba S. and Sato MA. and Takemasa I. and Monden M. and Matsubara K. and Ishii S. {\sl A Bayesian missing value estimation method for gene expression profile data.} Bioinformatics. 2003 Nov 1;19(16):2088-96. \bibitem{wold66} Wold H. {Estimation of principal components and related models by iterative least squares.} In Multivariate Analysis (Ed. P.R. Krishnaiah), Academic Press, NY, 391-420. \bibitem{kim05} Kim H. and Golub G.H. and Park H. {\sl Missing value estimation for DNA microarray gene expression data: local least squares imputation} Bioinformatics. 2005 21(2) :187-198 \end{thebibliography} \end{document} pcaMethods/inst/doc/pcaMethods.pdf0000644000175400017540000112715113556147213020212 0ustar00biocbuildbiocbuild%PDF-1.5 %ÐÔÅØ 39 0 obj << /Length 3015 /Filter /FlateDecode >> stream xÚ½Z[ÛÆ~÷¯Ð£„Zcr†W}°Ý&q#‹d‘HR€+qµJ$Q ¥µýï{®3C-µqÑ´Z’s93s®ß9³oo_¼úÚ¹YZ›æÙìö~f“Êd©EflUÌn׳Ÿç·‹tþ¿v±tI:?.–6Ÿ¯hùО°££–5¼ <äfa‹9XáËï‹ÜÑצ]üzûW_§õ,MMç×\¦…I³ÙÒV¦ª-/ùÓ¢B²é|¿ûþ4{ «üï'üfê%PçQ[Ü!®/Kà·æ9ß¶iØÊsÃß·k:lü$½Üâl™×¦¶9ìÍ™:«xSïÞàâK<÷Í7¼´š7½ì )·=“~ˆ©âó,Ï–§Þãñz~‡LÜ/·aâ´]0Ò°cêoµOÞmàÏgîø%Éøºyÿî-<à+¥ó¤Im@¬|œ’óƒˆU™µÑ¤ý’ÉÝ  |¿À'µ½{ˆÙù,7eYŒ¹d@ Mma²"ù€ƒ>ÉRx”æÀb­@¬Ä™÷B%pò’ÈáùåH»:Ë[£LÖˆÐ?©‘ ðýyº°vó™v^&…±î¯“ kVOÏ´oX ÒùžO„ÔYñ³d¤ø¥3e}Ádî É_ßWø»ÃuÊt9°P'¶¬Èò@@¤B[é;.Ë Zδ÷#6 꺥%&l²¬ œuiqeÆû%ƒËuw`î…Wt MõËðŽê—Ör^;K3ã²B L<-«ÙÒU¦PSGº gࣚñGž^ÍÀ [àäv“š*6ݲCâUêhœML•Ãhu$mb©K?‰Úw¤›È¸'k幩“:&Òˆ«ÉÁ·lPóJ;ÿú+¾³ÝöŽÑ3¬Õ•¸ÕM®ºýQ}“:/èö¼jø»9¬ÃŠä–ÎQ÷ îÇïp MYlÆ9G‘dh‘$˜—w”бW¾q@„ŽGÒðÅ’·z–BÇ(Fî}ë†ãK,oå'– Ò|ÐÅ‹•E`¼w‡–G5à +ïÙ|rfFsŠ>d_HÔ³ˆlE¾ÐÞû ÷²Uò]Ä ¡Òñ$™K«¢V lïBT‘Ã×Ä9PšÀò‡YA2ÝŠy ƒÉ0››Eƒ*£`‘'óñT9«T‘>á¯Ýy«`@»dW„§VÆ^õŽœõwTÍÛöÀ¿!‰­÷ªë@°§¸9ûõˆE0iCûoÙÐ&4M}“u 4uN~góSÇÏö“ !8´:5kÔMæåtV9ucJðAc*6Þ‰- 1ðÙN!Êú°ái:øÀê‚ ÝùÄŽ#K“ùû{i<¦Ðbâvð}OŽî4b€xSíFQ%îc%ñžý·«YCšµ×{VB ­U¼É^üq!âãÛ³'‡Ñ↎§’^ß«ôékÍ=ªÒ8CÖ9ªú$nìu©wE “šŸ=ì] ;ÒÀ¦×X{©WŒ® ‚Øð›?˜WfU Ü$ ¤÷‘'ûØG@úÄÆE‹6ÃkÁ/iÀÓ´6ŽðneœÐöϧ>EDá4Ðÿm¨š§B0‹–¦„dà/Ib'ˆ%&¯‹0Æ;[&¦p~ÅO2"™¦.]Eà Q6ËLšfš~=Y ’4íј¿L,EoºÒ-øtPÈ›‰ѵAª÷'­èÏöã»3SØ<°ˆr•Ôg) ÑÄÔu-r+.âH/n? ÈÀ%u°ü=÷Mɯ„Ü2)þP~)èÊŸ#?gŠ4¿ÊMÐÎ’Ëhí¨&=\X#Ÿó!œ¿S{Øo!“†!¾©¯¹S¼Á3§ÈoZ/˜U¸LÏ8e6’I/ƵƺW%äýÆÒÖ ;¶{ÐŽâ-í‡w\Ä@éÔÓÍA^^f‘VOJ±.ƒIOœ/1¥óèþÄ#ÊH‚ì’ŸT^”= “a†ù ,¬(§èU&¡RBk¿œÞk|ü(õñ7ž&ÔÏB*îŠÑj¿O™³Ë]¼Z « øï ÄéH–«z¡„‚z éþ „v{…“ æ`aÐ…DÄ”Ëÿ@†Ç§<Ï äT_"âž`úAHˆ³ÿ… S2,Mš§$Cð–èÕÐ[&TZ¹¤­ª¾È¢î9_áJ[•Ò¢>å7Š{^uÇú@@<.€¹Žö  ÉugºøÁ$ÄH”qiû m¼ª±‹¡‰•®!’•$áÞÒÙW!‘¹<ª:ŽÓ¯)]sàsž kÆe>&4}X,¶*øôV±] ØåLÎHŽ‘`ªÇú )NÕ²ZÏHišå¬.BžˆBáG;tÚîc¨Ïã()¡w©dl£JšÎXc|^ïÃJÅÔJ-ÔK¨ajn„q÷áa*½ŠbÜA`«•DíØˆD±á<(ÄåÚm!¢ÌgÖä©‹þ5È­©r,q…aÿ@,L^eÏŒš¼Íÿ_”€/Ü_LÁ5ð®øR€˜ÿÁ’…±à²ã%1˲Ùü»c¤Aû¨ØÂ90Yt” D½¯@‘¡¬D»°Ó|N$D|Ê^”ÆÖÅX©b÷XTÓ…ìˆ+3øÍõ"I·,¸orEä÷à-ú°V8ʧôªId¼±ÌÛÅ¥;¢ 6Ú‡LX/ñ*ÂUÔ¨–‹sÙÏz« q‘·FI1%ÂS•ÿyÅA|—’ æ(µCùÐF[î¢C5—;î|úŽAkò§°Ò–êh4FámOÕD_'&4Ø–€rûœoƤ,¿ô͸4| "x€í¨œºû, µBÏ^Ž5hØëŽvTõv¬Ç ‰;k½¼¨~eÓñƒªÀ6 éŒM•eåt±Ô„ûŠ Ù©H®ý¯LÉôq¤Ck¤Zs¹´Äçžwø”ݱvWV©v+¹Ò¢JòaÄáC^9Á…¾Çp"V²Ö*u(Ââ+—lÿŽzˆÞŸÇ>8ÊGÞÁ„–´Î>‡WS9ÞµóY™:ÕUZ)<Ó¦ä k_ ¦m©\5]ÊbÉUßUÖìKíä1ÚQ‹„P¡Due9.VRº”*-R½€NbÅâšk‚ùH¸ñ57¤ˆ‹•¾”¥%5>-± VW¡+û|jçEœ{nµ,¸:¥»†FóOÈÜEþ‰7 ?þMn6sº³¥'ó¨“+lY5CÈh–®ESk£29®&b-¬IñZ?’§N\@󩮤»ÌS;ÿ¶ EÞÓËÎ`#À¸Á%$°ÂÜO¥|PLWFÒÅz(]§˜¸=Ë×Üã¤+¥&ÈÓé"Çq¿zÑ…I„‡°!€´m¸²££w? ¨×ï~´<Õõµ{ ¬®"Ž<*.gT\‰«¥×XØ4«‘;j4¼MÆAîGG¬Åb6]\È|Ø~Ç{%dK†Òû\âÑÅ  ÐDÚáoá}èvÑ¿l;)£ês+—Uþž‘ï¨oÍhHTàú½Ï›çö™ÿ­ëz!ÐÀû½&tàVІ:rÚ'/`¡}+ýȱA^pгµ¯‡»«·³ÜÙ‘nçÿy`̰1aäÕŒåÚ}cžˆŸÅIôódjo:Œ¬T`ÃIÿ½:®ùëéÂÔ|ßÉèýVŠúƒ¦qH9Ê Õë$A‹Ï§ñ ”™Â’·|ïÙA A»QUÿnlH·ú(Œýï¹`Ùô¢ŠA—ì¶Ý,¤ÌjkSUeTzñ÷Ûÿ´&¿€ endstream endobj 64 0 obj << /Length 3042 /Filter /FlateDecode >> stream xÚ•ZKoÛH¾ûW Í&›¯=,Ì&@ ›sIr $ê1&E/)Û›Ëþö­gó¡¶19Xd¿»««¾úªèww7·l´06ˆm-îö Ù £E¦‰ŠÅÝnñuiVk Ë·õÊ,-üt'ø¹øíWßï~¿ý/Š H£' ëÈÀ49§aõjÉGTø9Â_ uÙr×sS‡M8-ü=báÌõ%?¶-ŽmZ©ÝÒœfYöýPQÒJ8ÉŽwe‹Ñ¶LEd`{´­èFƒàﳬí9Llƒ(uPÙÏ6u^E9m¹D‘œq’ŽÛöxBy×W££öxÎZܸ2Ê-Ln ð H^ô‡žh{kk²åá×ÃIJX>£ *žÚÓeSþW4ØüØpõ·Ð$ØK7'‘!]&žãwÜã¶ñe´m,>â–+\7X­S8Ç[®¦­<<ÀáRlçIðxg”ñE–À«Ã–-_Î廦PŸ^_Ó·0 ùa®/**Ðe¹'‘VÇ4AGÚ²x:‘rÐÒÐñË è^p± »=¡Hª^ú‰pÆj öêXݨA„h–" ®m÷<Å©;Ÿ…JÙ¡{Ñ2.”Ü!ñ€‰’:‰®ù 0W…"$ú´û„¢*½öjçÆ§NÉðjq'øxîʇáW¾t¡V5Ö5flqqë,î³t‰iPdqŽ]² …ÃÄAjä(¡gž(ÈŠd±õòjäºPaÛà«.;ŸšÙ(ˆÓt hZK†ƒÇ¿ƒ$°‰¯DØ8ò|ùñÂÒ!mQ3{B%¬d1‘ÿ`e 1@o‘M!ÀƒcŠ–¥çˆI˜Ôþ<Þ cÚÍjØå_’VR\ÃÕI\ Î\’6rL:®B z>ɶ#³«Äêh'2»ËØÞhþ-Jî~•¤ËòP‘”}¶‘§àw¬Ï8>ê­>^ªŸ2‘¸ÈKa0¡91üÒe„•q‘-¿üù/‡Á¸¨¿âÎ%{ÒÎ/Œmx ]( £j®R'L,½DjžâÂP&E“r¬ØG` _Éû¡<Ï;žTáßIÄç«{Gäxl9šŽh»ÎÚÉ„›ÅÔðÊM=pßË+ž¢ˆˆÿèC¦Æ™ôUd ãü™P'_C&T%F’IÀæ>î'£w8f"`-€e™MýÞij¦jl8ÎÝ‘þòñÂ!æS+±q·tx# UäHö ³^Âz` 65³åM8BcàSY<Æb0Í«Mš(ÈÍ'ÎlmÓ„m;†ePÉÓ)*UïžÛNÃ+ÞI{áw”¸r ,?ó½|aù3eIy7áS}¾08TÉÜÉ:öÚ”¢žFàÄ7ãË/ÐÀÜêŠX%9 «p¢¦ç ¥I®PŽX£­Jº‘gãþ÷ò2gïê·`‘ŸÃ‘ eÃØ ‹UDà9×7®ÝÑÜ%/HáíµŠÅ(?Gþy­ã\a4I6©v4’À’IÔ ï')hP†õŠ”¦QÚDÏð 8³D;Ýð…s¼bˆO‰l|ÇuÁ’à½X~žBœf ŠÂ%FpK­“*ðýÕõšñµ¼£0 2›³øC&ÿú#XäÉðd«×Àü–œ*‹æêònuÆe}zÄ™_’-žÆã”bºìbñßq‹÷xÓŒoÒ埫LœÁf•òÑ<Ä8…™:Ç(Žþ^“]]1ô¹‚ΚíSŒÒe+ù¸0 Hßzi,Á¯ðG,5|ಠ’¬ÿ…šû8øzGw KÿXÅ£/‹¥xHäµùÙÈë‹ÆÌÍK°©â\~™¼çëõc¼áñï† •Ú!å¹ìÀ´ñ3=ÒTå …° íï ÿpTð²Ù8nó8Áä—¶‡mÍ€©•0Ïîþ¥€¨õKß"þWÄ ‡OgƒÓ Ok3½UÄÔâ…Zݼ¿»ùÏ ¦Â…qÿ´aÀþ²0_l››¯ßÃÅA.A\ä‹gêÚ,Lœ6ÇÏäõâËÍ¿oÞá¿€DfùƒL¾Ù&Ažg0Y„ð>ˆÂE‹ÉúL Œ‰«x궈™U¢d’!x­Òˆæ —ÙÔé®MD1€Gð"Ø6G‡ë$yé(¼kº“¬jhI ÆDI¶î¯ s¾°ç廬7 o€à4H33Á¼´ð‹=â^GòWÿ€ßÛÛš?lD‘úIƒ8»ÇšÛÿ=1ÓÿŒÜßz²‘‰¥ÿþÁ5Èó|´T¼ÿ1ž[ endstream endobj 85 0 obj << /Length 2430 /Filter /FlateDecode >> stream xÚ…]oÛFòÝ¿B(P€"šËåg°“—¢Í¥Wã^’wWÝCswvvvfv>Wf–À?33iWI:˧e1[uW¿]Åy]”†0‚!-Ù'€ë™½ë¯~º´Pš‹€èíýÕõ]5«ãºH‹Ùýƒ;8+ëØæõì~=ûÝÍM´…ÿ§þ¶ó…ÍMd~˜/²ÊD{vÍŽÁkÁðgÉòЮ=Vƒ€çiô´„ ¯ÓPú’˜LhQí`Ü7tÔùb"‡¹¬_œ†Ä÷<\"B³š§Uô-ê:ºç³¾ÞÿzZ×yÁríiž–Q?_¤u íy§ñ>EÔ€„§ ß4¸áw„¶ a¹ŒNǦ“is…ïGay@M­Ž¢·~ψxì€Ø k,épÚÛœFOûÓ[ßà¸&#ßeÔ k{VqµÕîÏ¥ÏYz¼ؼÆÃ‰¿-éãÁEÔ1i"ƒ:Eef&ºß ô'Vn„Çi¿Gägžto´3,Dyàï†ø‡ÁÏé"è¬PœVÏ ¸ÉŽ/iýÁÂëé‘ÙNi"”!É£5#õYX8‹è‚:Æy±å ž°_Éà@ÛFTÖr$.¶¨I›EïB‰Ä„½!üQ?^aa‰jFí}ãsIÍ(?gBtÚ…^gSy+N¶cØŠnlyvd½Zk½ºüšôH“¤Gi䘲ƒâÁNp6Ä\àTH¿áÏ.4˜³Kõû0]¸"¶²8PøŠUvkâ*5,û½j/‡ãv½JA¢u 'ßk;¹ŽÙ€énáÛ‘¥Žÿ$NfÿçÆ¿$y‚ðë;c‚Ä`ãªÈ—pþÁg‰ÃÄI¡ë&ùÈ 6ú•¢…ÏQîoÏ¡Ô]bÎ¥_в›-)·á,]€…Ü’y<“ª(R5{^Ðxˆ›Æ¦1ŠQG1ägÆ$§†¨aÙ0hǃØx*ja²¸NóéØÀ>P'¯q +g\Ü#`ØxK›*yIUGŸ`öö&æÉýf+j:¸š{CXº…ñ§·7ÂÐÈ ~ÞÀúƒ£°ž² R­ÇYáÙÍ.Pc *à …ƒNƒ$[ß#CŸç9ï9‰oÕUè®9óÔ-eÍ'vF£n3ÌS8§À¨È”˜$Ò(à]X ;å„Ôdäg-—ƒÄÐÒ²¯˜è?·wdìù¶Dz3UìàVPMM·Ù’7‚²³$‰Ž=þ¼ð@Ö`†«ãJ­BK3„b@"R’e$mÒ €Þ2S†±€ìFxÐp㖜ع ß5ÙTë™~Ú0g7ºê;JAƒ”ˆ™­5c ã3;ÊA0È…²,¡ ¿è”Î1 ùÇĤ opu¼´î ÙcQ†g<:VžL?¸øò¸˜’EŒšáê Ê{­µ€#UÐÊ¥~ÉPꬎþ)‰™si€{É3VH¦99Æ¡xÍøÕkÆñ-ëˆ'‡†Ë0e˜' 0ºÅ~ýwUWƒÊ)óH ŸJŽ“—€W1âã—ØñJ£ªi›õƒX¨bw7íD\EóhÖ-OÁqzÄQÉ Â²‚¯P|˜±˜«ÊD{ “S•6²8b `Ø-Ž´Z†…ª°Ií‹+;wTIZ_”†IÁ73-»ÎÂTyôÅØŒ6ß1Ðåf p/‘ˆuTé™±÷šR9ˆ+žØ§ةϵ“åõŽ+X¤‚É#5ÎyirÖAˆÏŸÈ`°tæ„kÜ:ÎÁ£Æ'©gi%5.›Àpì\S“·©l²lšŠ1Á0ᆔŸ·ªœ1`Z8ZNÇæD&k~P:σYR¸•”5 «e&½IˆIMÙeÆh=áäç\‘82ص3ºgào•®¶Ù¸àctÁ­&©xÒâ} б3O¢ªùr.ëóÈeêXŽ’V­øV{QÈÉnºèHA¤°adkMòK“&©•ƒ°‚¸—É…¸ÔvÆÚTã84U˜"1lö§\^jÜ'î«]%‹­ 9†_WÖµ‹+jß0¥ÔRGßiÛ5Qý[{棾z­þ2ÛQ¾¸z –É׬kׇ?Ò½zSÒ£éÔSeuÐTU\ØT»ªåœ¹XiëvÞ¥YlêZ‘VÛŒÜj¯ôEîàíŧbÒ_.ëb­×¨K®˜_0ÍBžœ¶}D¹¾+gE\—¶¢ö°*â " #J²—"dqV”Є{¤X´’ÎX¿)¢-Ò dÍAéiW¥´ëèî,Æ{à°´‚Òwßï\’¶´wuÍ!äl©Ÿ?þäŠ:­£/˜DO®b[ɱrF^VÑçô+Jgeõ7}Ó3ãÒ#ˆFÓf©ö[%Ðøì‰«¬Ãî”ùãìºé.(Ï?OW¹3À kùá ¬v²8sZüw¯ÚN½èòDº'£Ž mÿ\øþ ‘Hg§Eƒ{,X1äs¼”‘bÒ…Œg†–wœ–"Ço¡’ûÓèyÝ3ær«O4ýãdÒú SUÍ}úŸ˜m;I¢ü¦yCã¥L¥yÅÌà€06A¡&Lkî”=©r± éáóÑù0M?8æ>Hè‹zär]2£ p_¹´W^=éx¬m´[纒"© ‚tv `ºÒ~vǨ«g­Ú[E›ÌýV@¯¿.Ã!¸òïŒ&¬ÝF ÷%:ž:0Т,…|5>•XÎaÂáàÞd|F±¯Þß? èïÆ&±)2üäó×d¶†EPPlëjöD¨ ”qVatÞÍ~½ú…KIÍ,‡ Wò›˜Íãª*¨saaØ+êY ˆ‰¡8£¢Y\ÁÚÐo‡àïõõƒÖñ¾q¼¥t½˜/ ËzSÕÿ àúûkóqÃÕ9¢¦QDûÕ:^O¼áå&¶I8‰Ó:ä•ô?oî^8 endstream endobj 59 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/pcaMethods-002.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 89 0 R /BBox [0 0 576 360] /Resources << /ProcSet [ /PDF /Text ] /Font << /F1 90 0 R/F2 91 0 R>> /ExtGState << >>/ColorSpace << /sRGB 92 0 R >>>> /Length 1347 /Filter /FlateDecode >> stream xœYMo7½ï¯à±9„á ¿¯ ÚZ µ€ŠžÒö`Øbú÷û†»Zí*k7zɦF#’Ù÷ÞpÅ}tâîÝ×é^ïžûù½ûp7BpÛ÷»¿âӚݿÓºàþšÄ}Äë~û‚ûe*â[uÚ| îÑ¥V¼È2|pwWæe$¾ÌVIÅW]Íçáj×(>—Õ~®ö(Å'YíçájOA¼^üÏË}»ÜËêWûûÓœÏÏÁyþüez÷“"’§œ¨_ óÛ<ÎÙ‡èÔ·æNé;ÝO?žÆ/¿î*±›ÛÅW¿ßW±#Ý̾ß5Jôa3í ³Æ^}×‹ë ›MM}­×rvMÑ+Ìã«8 1Š:†ãwæXÆo]YÛÎÚ®¬¢;³\ÿ¶”½½\Ù5ììËpcß/N¯W§ûåéõúâ~}›½o£º åÛM„5ÛÄè¥Ì9ŽäUçzq>L…WE.Þr»w.¾”—øuç¾:ß>³ÆÕùöMk]o·AØâœö5°€Ô”S½ ÎÃhT/Y³ž¿|7ð/·ý<‡4R+¨OÁ=ýí~w_¶¤PJtÛw#qx™Žqû¼¨okkÙĦ¸¶8+/PѪ,°âNO®úÔ` ó?š²%Ÿ¤îC}°ˆÂ|zZÏäwl¦cò„µ÷ÛÝ¥ŸaÖä{!fsÄ2J@‰Ù5/U‹Øí‹’+Ìyðåí³ü©ÝH(Â=K©ÜìZ,äRÁŸÌ¹çj~›ñàíî—FäQ*Äìà@,¾8™Å£”Î=wŸwî‡E&ç{4¹#Z}dÎîYGè Âˆ½7Ÿ¬d‚WªÞ±ê<꽡³Ù­â.|ysΧ>P#î’áV*1;¦m¤M "­Ðíø»ïLÉTÀŒ•L7СN±w(«BÌ.ýMÆÞ©Åi›•L³Ô'ŸTf¤ á.µùÜË(³wœXCä)•Áºn!—ŽcЦŒU ”OdB–©³JLL½C\Ôj C,^ÇÞc¥ÒFóˆ™hô)ØÐZ%ªŒ{WChÓ[‰p7œïqt¨•Áy„κqç.5›¬¨ËÂ$-Ú!;¸T ò¨‚Ekm](a†Ðå6ÜìÅúd«œÌœ»qÜ‚óLÅØ+H ê=1@d_¸Š‚‹g=?.‘(m3ø}ð4qpy´(ÀyŽaûX|ã*N¥ÏšV|aÄ úps\d&ë°êfLÙ(y -¹™¢¦'V2 )mRš+®q z±œ/h™N WgEÍ0¬‰“FHZ .€6Ùà¢Xê—Ã"m:ðÈÖ0KâHhÚŠ/ª‹ Á2sk"ƒ«Ò¬…’Z(y`«¶vteªÒ:)ô¢é&ìÀ­“jA[ÒZ3bpÁTp¾ÌËÝ]ä!NPx™98¨Qƒ‹†äc ÒöžÇì MHkWT%^¬d”¥€™ÞFHí]»uÞ&°(Qšú¹â*%ò|k„ã§ÚÀùÞò ’wRGÅ52tØ´Í.ÔíŽë>¸w¦dÐûFò€Ú;ºÈjò pÚFÇEé:¥ºFZ3Bt›ÛÚ¨+/µ[Ή“Y˜±‹1Ž3~§Ôfó]3…´(ØR†4ŠT'•LÔØ…[§ VM*¸VMšª¤nJUg†-†„¶I.ì‚ýˆß¥ïŸšÌOÃ{±þqÊègs¢ ZÍžÕ¬žåðÑó·ŸÚÑž >"Ñcy)m·ÿW¹î©[j[bæ§é?O_¾r endstream endobj 94 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§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 103 0 obj << /Length 3316 /Filter /FlateDecode >> stream xÚ•]“Û¶ñÝ¿â¦3¡f,š Á¯¾%©Ýº½x<ñuòø'ñ$N$Q%ŸÝ_ßý@‰—¶òÀ»Øï]øû‡WoÞ5wmÚVyu÷ðtgr›6Y~WgUjòöîa}÷Kâ‹eÞ˜Ä]Î “ K7Ÿq€É:YÁ°w8ßÃïË"o’¡óäßÝþ9ÁÏí¢É3B=/`÷€k¿ñé«aG:u@ Ýš†'ÞëøböW%ჯufà"ÃïoDʉgŸ·HÝ §¶| !8y’ÇÅç‡ÜewKcÒ¶,åÞ#ÀM‚·Û­Œ)’_Ma™FœÆ¢ì–HGÇHj¸’Iþ´0y™¼ز&Ú‹3ŸÖkÜsàã‘'Ãa)'õÄ"wâµ™ëÀ¬C˜µÓÑÃEÈ9Âïrþ£íÄÁ^¾ËùL!`3Ç’þbYØÀ’ÂÖɉ®MûdzÈû²jPˆ¸O&5qébi›6y@^t¼ˆdn-Ì9lFžÆ3ðz8~$y üØ û½hâíZ.ÑÉ6:áÉÍT ` ,ܹâ;ÿ(D#…¬‰æ 7ej}a.”Qd¸JÈø‚úIf@ W'­™/üáNa×(¸÷(ÒÝ·€dC*C|_3è3_Îã'Öíð_$pç 4B«üÀCÉN`3RGv^N^6¬ÏŽŒ Ù¶<ââ€4ºG4PΦA9àeä}î?Ââßñ׊”SéâÇÅY«ÎåeX¡šÞñü~^DŠw k„Ðït[5ÒÔÉèö4ªØO‰ Ñé,<»h€ªŸ7âHhHL ƒ@¿‡0ø.åÑC„…@vßxÔÿ×Ìê™:\] ½`%:ËáCvS‚+¾ygLäÚ¯.ü‘A&ÎߤY[ {?ÑÄ¢(ØE!žù›–P¤ƒcä„Åâ‹d œøÆð.3 G^ó̇_ÙâR(šºcë6·d¢ <"áÛ&£¨aëà©pÏJUYØK¢"¨žbyrÔ² F§²d0V#<Ë Â#ñrc€ö0ð_Z¥è9Ü£†Ž9·CŽljY@ Y©õ;¡• -Ìš Q`B¿ã Iå{æ*‹CåvŽ ;uÙdZÎKÀÆaÇqDÍ™w¬Yðd­0FA ^)aLlDïQ¾bå'OÐŒöjÛ'Ï|B?ò}ËÌò}…»EYˆ&\vì pmýÿî)83¹Ä± ûBHÁC†=.=om âD×ÉÁe–1ˆ;JSBßÏ‹Æ^IU.Gòçü…°ÄT4°‚‰°ócÛ´mdœMžÚFmMä-Ý„¹Eò£ ¢žñƸË,-2£°«¿BaòÔ´öÇa’wòÍ078k“š¦Ð£[®9þŠ˜½éŠ·Qk"Ÿì9’+rû.ÖT.ß1‡sD¸Ì ä­eƒ:mjÆû::17<³. C7$£Äš´hŒìEÒŽnÔü׬ÌàëûîI†vóvÂ.wm?Hôý­žkþŒ )ðœ¢ãÈcÂÊ}Gª|æc>ý.!ѱ3 ¯üŸÕgÞÓ‡8¡…nµ–/i*™â–¤¾Ì[6‹¼Í&¹MËþhÐŒ˜(a‡äñý§…äm˜ÈQ™‚ðdt½ÏÎ[ã3?#üW2œ8îËñ”IÀøYýÆ–Üß&­¹ÊÙf®b$™“Ô«lºÓŒúKtvQH<຅'$!§O”^íùÃ;HÌÞGqéÃQyîÎáEr`ÏOs~1}hVg1B •AÕ¢ãUêBùtÖ8 p3¡'2Ó‘ì}'ÚíÉÏúÐ82ĆâA'‘Mã›Døá̃ýpºµrRGX,ÿ,š c V3›Ú‘ÑZbÆ dë7ȃÅ[n“{¼Xy'HŽ€X|Íá#)¥8ä2¬^ôž´Zh…iV“K”,©ÿͳ„úÇ‚ƒ‘lÕs£“'Mº¾hÕ¨Á V¹.²jĤ;ÊÑ=/­¼ ¨ÏxûƦ­-Õ;ï¼õW?M¶çBŠ5>^F¨Ùüð§EÎNòqš¹J@ØAZ¦EÙÞ&èÁi`p¦pŽeB›ûYÇv>¢P! Ø] DRdw};-_ÝæÏ¦†q¢ã ;šÔæ¹®+rô4g>™¦¨|Û†lˆ¼ÄiœÒà„&¢Tû"J/ T'Ч:Ÿò‰™Lú¥˜PŠÕªÀR¦Z(~¤¶)šÓ¼XkÜ¡¾—[äƒHR“„䳤„ûØøqÚñIâÈlÈMUmémáÖ½ê‚Óº§=æÐ5ê:µ°kÿ={yJ÷$›(¥=u–U·*œ“uïžaìGu›õÊå4ŽðfrXº9L¤xN”0¼Æù*ù4á©;íÕâQ'>êŸZ®Å ÏøêÕà«™Ý4_î›\qŒÀëVÈÔ>ª{H µE)t M5<±—lg§9Jù²9!ŒV?…eÇp‘t×4Êâ˜ëŒ/Q{·¸9áî|Ì[Q&T‰|zn˜¯’!œÍCÄ|¤ {†u²ºÖF˜¯³‚Ò­DÛH° “ªú3¦úy›¼?ðºæÃã9F*ÕŽQ׈†³æSü“V~®|í¦~ûšÔi‰,ñØÐα÷åïoú%U9 qŠWY•AEοµPÃå,ú;¹ŸœuÃõ%BŽt"ãŠÏ Ceâf.»¶Ù¼ŒÍ ’\$Î÷ðû{N¹m`0°Ùé ,©2X7Œ)¡|AŠGý¬^kð¦žtš*öÁ“—¼ùæ$¯Ð»\Ã=ƒîjJnÉãÊ’{êBÀà· ´±‹ F.­ûÝì)ô²Ç릛ïLú…l#Ý!jsKûÜiãoЦ%LÎß¡žÜ.j^¾1¹»êÊK—}¥¢˜ÕBqë;öÔ2Ý Rd)þ“Ÿ£ûÚÚ¦…­´¶.ò´Í¥¸Î¥®þ)ÏÙ%©¡ÏNüÜ™s): –IQœ›´É¥¥Ç„7Rœ²õë¼Ê8Z- |<ÆtbƒKéÜ‹F›6µÒxX4%19Gùƒ†Dêqäy”E„ ¡ç\-Ìq^ãË‹¦Ö^ÚÄ×ZŠOtÇ…WMI{tBîD«ÝŽÎ0¢ã¹á'] ‡yÎY®÷ÓOœoòŠo?á¾@I¾ïCNÍ~Øw“WäÇ1¼BÐ ©Ím_!Ïë4ƒ¡ð|+,Ö§é¿Àïüž£_Jy·ä$ nâ&×âÜôQyæÉ+‘bœ°Ma6„ëF3ò,m_­‡²´±Uòø…Ö:ùY~ñøý™ïŒ»Óy®¿’UÅO·4 °}CAßô'½RPJ±¢¤¤št%&Xª:­¬göGéǯè¿Uðxãƒß ¶J­ius¨ê/ì¿ñŽçð°|‹¼,SHâuÿ{‘–g‹è#W®Ëøaªf¦–™%ê«‚W¾ç©]¤ÒÆÃ鋘Žac?°]ù› þ5#bÑÐkFïUŽ^UÙSÃ@SºÍvçGZ3˜êúù8*U'oˆ…)¹UpõxKÔy¬ö¹¡vi!t ’ÿ.¿ñäð‚įrƒ˜d7ß°œšˆØZeÒºi¦Ùµ·ôû÷½ÏU¿¾ù—&9_¥÷¡V3’zʶš£0%äNš³«a¿—$m®Y¤Pþ#ÉÿðøŸ6þÊŸ¨”ø„—ø}§ÜÙת*~ ã0Ö…¶ð9"5ôJ¸†º%|\u纟„Œ3,/Òúú½ Sñõ'NsâÞ"S‡³”ŸR&5ÜØG…?|sÕÝ”^³I½NŠvÍxwLo½Vô/Ú<{ôãÿçš*-‚KÔ°Åîxc¯ÿ/@Ô0‰f:nNÖ‚Þš?pÆMZþ}uœdˆ¾ž1Uš³ÇñÆ2¨Å¥Wo^ý?HkO endstream endobj 118 0 obj << /Length 2285 /Filter /FlateDecode >> stream xÚÍY[oãÆ~ß_!,€,.‡Ãá¥h d³1ºEš.¼N^Ú>Pm«E…#­ãßs›áPâÚÆ6Nò i®ç6ß9sæèíõ«7—E:«â*OóÙõÍL¥Y\&é¬HòX¥Õìz=ûWôÝ\E|ê5|mvðu;_d¹‰wÐn°G{\°šký4ϳ¨¾Å©?Íÿsý÷7—eÈAåq•ªY´¯ç¥â‹,1Ñùê Ãä‰DV4Œ‰ ´„r^!wøüƒv(¿ÅF©í¹ &ÓL…D@ü´ñMâ7,ÆÆwÜ|˜§eÔa Ÿ©Ð±Ñ¥#wuίŠSãe¶(¤µhÏ zëg ­Tœ•él¿•1¼ö€|`Ñ~¾ù¦ ’¦±.=q´#ʽ$QÙ0ø‹4þ˜„H=Çf´\M¨RŪ,»˜Œp´P©ŠM¡g‹ÔÄeQòÊ¿!§-kbLÔU½mƒ¬&°’«¸Ð™ãów£5*‡dÄJ™ñ!Ý è¨f[ÄÝŽçá÷Üf mœëÿEEœhó(ŠJ¡lœÝJ¨ò©[oU‹œËá€Áp:íO°Ð-…@M~HH5ÑGÕ»ûYà<Ô½ ìŽË½ÝQ†x¾(”‰.çh¨ž4¿ ‹vïì|ÃYNd;^q 'AšÖDkÕ‹å÷£#€ùîf ¶$ãÏw9Â噥H*ñÇõ §Xhfy¦Ç¼“³Ý 7qˆ×â/! ^Öò›G@–ŒðÿqlU +%«>Ôð/LœA,v‘òqæ×&‹lÝ¢°BG˜ I}CÞKÞby¾î‡µ«® "9ŽÜo<Ü©OMZ­£="ºDDB´r ÏRƒ³æ>µ Õb[t£…º¤ˆU^ŽÆÅ­ƒì}'{o„F+ç´ x4çæ+‹¸ü Zøø±CçKS‡_0‰'ypž£ÉŠG”Ý2#Úr3¸šÇ?n$ß6=¯jiG>GÝX7v™áÄÈ&"ïÈ$Öé7erU´²§,þ¸UÏA—Ç&÷ Û€²*Œ)ص„eéŒÀ‡“b-•Ñò8´G0ƒu*á߯xúÍ-nh,jÛ¡ÿ ϯ¡PL›(èÁ1ñª<Ît.ÆËYx «Z§Qó3žb½åÞZìa47K‰ÎC`vÖNg*‹uöË$Q wêzÌ ¸~ìZ±Ja$p"FÙ7'n wc™”ãHk=$@(§wAhŸº Ö»‰•}¤š'‘ àO §d,÷Ï⪿˜‹b¬ ¯J¦…û\œàWN<ˆSàÂi}„ú—ùÀ·ü%¹=Ç>w眘ü{ò‰{>5´hãã[ßœãb]þþŒ£lÙÏT’ÑŸ)nÓð^N´ßž_j‰ ¦jP8cüE.àÚÉøëaÛJ†v‡^¶^Ãç >?ÀçÛ§ñÔ ô¾yq•Ÿ:üßÞ$Ó0¸:òƒÛÇßqcOwÚª¦²ÛP¯¯/xò-t>ø¦VÛUv3@Ò?wCÕ†JÍØðy6.6A¥§–’ àOZˆ³ÓB윉¤õn5pÛØ€Á‹ó ²ŒÑTv–\„x4—_hå Ù½•´Y¥#t¢uð  Ùd–×5ÖU|Zw\@Ù¾”q\rýIˆq W\zZEs(Ùn”Éqj#RÈCOJçC ·ð; öÅS–0éSPàEC kõµ{W®yÖ™ D×Mïo‘”çô†­7$¢qGJ(ÚYZ³rnŽî®fÍSõR¼Ù×ò.¤TÝ zœ] 8¨ã) \;õµfÓ#ï¦çW•êYƒ™‚ ›{/ÊKÊiÛúÒðiýÄÖ­¯|cQ°Ê²i\¯ÌùÜ&‡+ÍðþíGýÑ [š!´W¤Rïa&kà ãhÝÇBLoœT>L“€"ó8ÖúUôÇŽT¨Gzç;³¿™µRƒ. ±Ö I-}sî¬t°¸[5z‡XÞH¸lEw¶#ڈʻÊÕ TÊ—žlî¼+ 2F¦¾ÆÓ-ëÅÔðí±Ù†…câÉ›[÷ŸÁÝp‹‰šõ­×Ê]þP´èD `/1€P½H°¸)àÁÝÑV„&“þ,ásã‘%œÝ®³¿5(ì„7”ÒîÏ¢ Í]‘}¡«$º'Ù¨}Ç#C-¹¤†ñ^¬ÿ“¸ÈÃädAXó5›C„>÷|Ø ™rp; ²[Àc4Þr1Š÷òÝjW:Uñ³rqö¼Šunf‹¢ŒK%{ N½úöúÕÿH|ªK endstream endobj 128 0 obj << /Length 378 /Filter /FlateDecode >> stream xÚuRMOÃ0 ½÷Wä˜â¤ÍRn 6 ĤހCh˘´µÐvð÷q>ºnBS”Èö{¶“縀€Ì¸’dZp@ª}òðE‘ë…gœ˜‚˜W÷{Iîºä×±©&;)º,“«µ!/´Ô¤ü K®É”á¦ÈHY“ºNnqo=žMÊ”RT^§,Ë]M`Ó¦ÒП4Ï©Ýaà0s4Gܽ‹Uc{Jj‡ÀêÞ#ͺzmƒGßmD>CFíüWÊqü|ï1°÷.Ø@»”É­‡ô­|HVåQI 0‚+|zX(Lx6/¬. |Rô¿ÀSãÜðL—Öš*T8A;ëÜ:‚ƒöQ#´e@*§Ÿ“¿rïìèëYgJ¢¿|Bûö&8_.¯#£ñ)8Ìݺ>Ã|ŸÝ ó=«¹s2£ €ã s¨=VÄñsþ¶­ÝÙGZ¸÷%A›–k. …õ4Ô?ë5¹ endstream endobj 113 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/pcaMethods-007.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 130 0 R /BBox [0 0 432 288] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 131 0 R>> /ExtGState << >>/ColorSpace << /sRGB 132 0 R >>>> /Length 837 /Filter /FlateDecode >> stream xœ­VMs1 ½ï¯Ðu-ùcícCÓÎdÚLZ˜öÉ ¤CÈIû÷+¯½»Þ !@{ö¡•ôü,KF8„[x(¾Á/¤†R ­Ai%# Áã~ºx¿ùþùFãB )%äßãÑyA‚ ü).¯@¬@8ãÏmáøZ”(JDJ„U…,É caY’BK@i„ XIÞG©…/v>¼ïù_d8Œß"$Eù&¡¸Ü¤ø.aÜ3Gd­ hMôks‚­=-§¶'ØÚÓòšìföœNLÒâ¯7[¿¹æÍúÄÛ“–X$CüŠØº@H—a{'+à&·Å養¼Ûًɶ¾´¿/Q)¤Í|Õþ¾JÚ.g}€¯3]ΦöMU﫲Lg€Ð e*¡{f®F­%/J5æ3»Ò¡¨{„™Ý’°¾µG˜Ù½ ´{„­=Ñ­íû\˜Lw™Hš+ÛB)¥ ÿ¹MÖÞÞÂÑÛìd›*<"9–ÝÈîH ‹u|Dv í“{¿¨âé-´ŒÇÞRC"'|hui»°~96º>îèKá=/6tu&{98]üš¯‡À…<øÍ?Óåó|3¼‚ÉÙž ð#3vZXcŽŸ¦ëÙôq³ù+G^LŸwk¸»‹Qú°aƒ;‡ ØÐä¾5wÐôæNÙ;ÚtçNÙ;ZUÛQqL *+»ÈÈ&fGe„§,»âîë³ìŠ•tYvÆ¥o³§†StîÈeéó¨Tp]¬êª`toúšŽ 6W_ãJ˜—2r©`žÞ ꤯ìYz%T¾ N`Y¥Ç2ä9=…MJ¯ƒ.MƳ„ΦmˆùBeû²¿v¿ ÏÝ—êsŒÑk>.2”J(êqŸ’Üæ©Å„á H jpuÉ8¬ò뀜󀇀‡ÕnЇatfñÎãZudÞ"•Gޏ¼w}Õ Ú·#îk¹%`õÐ Z=vCsÊ·)áÿrSÜ$U£„â »}8ànÝÑóÕ9¹åï0´Ààr”¿åkOôÿø¸XÝ??ÍbXíä|q?]nà(*Ü_“’ç_²ßŠ¿¸˜3 endstream endobj 134 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 115 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/pcaMethods-008.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 135 0 R /BBox [0 0 504 288] /Resources << /ProcSet [ /PDF /Text ] /Font << /F1 136 0 R/F2 137 0 R/F3 138 0 R>> /ExtGState << >>/ColorSpace << /sRGB 139 0 R >>>> /Length 1588 /Filter /FlateDecode >> stream xœÅYËŽE Ý߯¨%,R)?êµ$AЉbM%Ê%³à÷9®ê¾Ýw4y´CÄbFãqû–ÛåÇñ¹ž ¯ÂÛÓÏámÈ=& U¢j Â±r œcæðîðkxszxûËwÂãg§SJaÿûÙãñßšÃ?§ß~)¼8QxŠŸW'²Â§GW§‡OÇ]ýi¿Þ…µáÉ4ÿ ÞbÉ¡sThoÂW¯¿W¯ ½zwúöêãÖuXkärÜš“ÆÖC«±y¬ ¡ª¡–X’çìggóßåyâÐZLâ;[Rhð 9bŽL5Q‰©:ÌKZ`Þccßé­Ó«#pÔ`.âîcuòàp 2žwÇâ¥yô¬­Ùã9ଟUÔFLóíj(kðÈž…Si:Ô¨E¡˜ÆhK«£E)wKuKVòf³5HC"žl•3Jd³Ä‘p•­N=ÝY…,Õê ¼R _É•qbd^P<¾£+£CÉ*îøéðZæ\s• î«À\FÕ¸N·œ³Ó=å ‰ÖhTpsl욊uHB—îŽî.È9cÑ1™²g(k6'ž‹C“L4x¹ìÉ:ÃƦ6,Êd{I³õÃ弑±XkšËùf`€jua)ƒá¡É±7òÿ ®FhLE#-\@Ó¡æÁ+&OÒbÕP£%S쎊¼´1l˜TžïY+Ê EÑ.\ïž[Dâšp‚B/F’I_€ðj{ÎE‘g$;ÆçŽ~7NE¤wÖOq§Çîc骟âNoTõÆ -â¦W àºq:‹¸Óò¨›~Š;½}s·qR‹¸Óï_øüþÿ!g$Il½¤m°FÖs%]Ú§#ö˜WXÔ7ûcæ*˜ñ|a¾y¿,ç=Û߬»:s5L2³çòÄ¢­âNOÜŒ‰=?°Èû'ÊøFc{bÊû'úÀÐÛSÞ=±øu~bçç'Œ™i([ÖÙ…{à„Çbo˜‰WéµÚÛ—Z²ÞÄq{#Ü—eçÿËrsÒ*{kϤÕ9ž—ò%‹µpW_Ç’ÜßÃcÝ{/ï),]ŸÃc1V·Y§9L¤ endstream endobj 141 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§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 146 0 obj << /Length 2100 /Filter /FlateDecode >> stream xÚ­iÛ6öûü }”ˆÜm $i t™Œ±(Ðv­Ì¸k[SËž4ÿ¾ï¢D5Þt[²(òññ]|×ëåÕWß9“h§¬+M²ü˜hãT›¤ÊK¥O–ëä§Ô-2 é›ÃB§]ý"³•K¥IøÞ³Yã舣þö‹_–ÿô>ñÊ—¦Dìy’ økÆ{ `†Áê È)*À´áãœ.÷=<-Þu‹ÌT)ÿ¯éHXë Ø¥»¾šþÿ‡–§NHz»æÈË1mh³©é„¹Ü7[[áþÀ3Â>.Š2mˆaxÖÍQFž®YYût‰Tn„‚ŧ[Ÿ"^:ä=É„_±òEÁ\·}%<;:˜°Ö²dŒXÛG$¼ÝâŒI»¼Ò#2rZ ¢ÓaÄÁüÒ°Á—#ñÑ)òÝ·´A³õxzÇ ÝC q'tö3$®ºû—W2h$‰ŸpsF‚ÈE%‹àôZ¢6@^ž69“³ ›µœ¸¿ëYêU>˜gÏ€ôDƒ]èC8I§;ü'ÛYóâ-Rù™÷w¤ÇÇ@‹—È‘›©$加3ö+&úwâílU§ŸF…Ñ÷¬{^£ïÙ«*]/ïèæ '1<}dêD^˜a΋\¸S-ßÔ¢ÎÓ}wäõ˜û÷·ð€š×qú‡àPúl΀‚5[79åàj^ÅËÓŠDÈÜø~ˆÚ*òë—‚{~÷+DpÏbeF·â)”QpLWïAÿ'Ð'Èå=˜;hp†göò‡pU)ÚôjæR8«œ+ÿ⥨•ƒ$ ¾â.­w,ËŸsíöÁÀ`²‘EBd3‘_±¾e C˜ð!ZïùcÕNƒq8c¤™X×àOäS‚>ÈéU 7¨ðª2ndz´R=©”Ìé.vµçR. Ux¤üã¹½Êí ÒSRjUp¾±Êk^¸¹ù ®Ó—rF„ÁB¦#¯\Y<{ ¦n>É" Ê9Á´>äT£<2‹B•…£¼Xå¿ÎyË\WID;³1>Í)­Ý³6–Õ0Ä›RÕÁÌßx{só¬H\¡Œ«þN‘ÿŸH€­XÕÆNE’ç3då`1ƒEüg†ž,W¥µŽÊ/ÓQ1££)MÞƒ4Ñh •çú8Df/DZc!€E@/YÓzF¨â2P’_)Žy±åÌhYÇP/ϱ8åëá ¼ä,Ln”+@Ö@‡øŠeHÎ\ÎÉØïäåÑÍ$ rùÓ@È“âöΩƒºJiûeà®ë²».§U“ )¦Ú8a=ˆ›æ?ÝK˜“¨Fk»–¼éd+9]Z•R&âó 8šýjäv3–Sf½S•»ä55/]OK?R‡µ`Õ3‰‰òTài.Y$Ä#)CÍ2T;Ö™šIÁë“Æ ÁãŸsäÛBÙâõXÛNM 9²•MŒ*t}áêÐ`…¯ì‹b ˜J£Ñe©ªÍÖ“« Èo é26ýªmÍ9<˜v}ÏK_S FÃk’ØzQög)µÅ;’4”ã™»$?¸ôLïòbgœ'T¾—±¾ú3TN7þU*¯Þ.¯Ð9ä‰z#àÅÀ·&«ÝÕoW૚¢!-…m2ñÕ÷;—|Û]]Ã/,ee6â|M™‰U…c5ä•ai}'ÖzÊz¬Ý?Y™Ûô5¥^X!RŠÉv\ IÞuÞžH ŒœÑ³í°Õ<ºƒ}¨×Ž¡Ob£þ „:yóêŸÿn#”5Û^*Ìam(8> NaP2=®‡Œx½ØˆÞóAX¦ùr4,^}5$Š8~r2‚·œFJÁ„3TP„áåÛ’áEà­{ŒVØ3ÄÊŽ=Ï~uÏĈ|x±Ù¯çîKøèÑퟚТ¬¥Ô=„j&¨kï!Þ¯7Ò)â–ó~,ßê'ør¿ÈNÂP¥×†¿WáÔ±}R×cs?‚ÙÀpZíÎÝ̸d°NíÁP®7w2O^aàÈhæB£\d]Ò,&^Kª“aÏ»¤5VíÞ‡^ÕÁ÷¼<Ù°™¥Œ4d†dÌÙ6а¶m'Ò:"gŸ‚¤¾•½Í–?ÆŸžmdËÒ¤u#^;°Ô2kØ›­iÇfHäôÓìóvò:,ù¶ÏQ{±=ÛÒì‡v´^`fð¡!uMt •T™ñª®%)«BNðêÊå§ endstream endobj 142 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/pcaMethods-011.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 151 0 R /BBox [0 0 864 504] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 152 0 R>> /ExtGState << >>/ColorSpace << /sRGB 153 0 R >>>> /Length 867 /Filter /FlateDecode >> stream xœÅVKo1¾ï¯ðlŸG¢¨%§R­€‚øûØÙñx’¦i¶]Äa7š|߬óÍèeÔ¥ú1œÑs¼žÝ¼}¬Îo Zk%ß7ç×óW¶ø¬¾*c€ñ5Ú3x$cu¥žœ.žªÕåðr5õûâ£Gb-Ágek× œµï×oo8§µoꮂUÖð~íjåÖÄ£‚†ˆÊkHZaŠ`PaŒúy¡>ªk&‡Y,wf¹xWþ^ý>}VZ}Œ:)Ïå`Ö»y;$¾ÜX0ŒºRÍŒà’ú¦–ƒ 0wṴ̂Þ ¶™%F0Øf†³´`°ÍŒY;^i9qeïA²ê#jЀî0™Ï²ïøhv¼?ÆÉì8š12N¦Àc‚ ðÑì¸s|bœÌŠOɽ•7ë·ÍËõø&ͲžÛ%ì}ÞÆ–±{˜ìžŠ¼BsOSÝk^ÙHîIOuGƒ`Rs·“Ýc »»íÄž!OÇjÀiêkQYÚà–¤ŒSN6¼¥(㔑 o Ê8åcÃ[z2þhé%Ñ*´u[õþÌáwoº(vß)q«uXúâᾈœô+Î@éë·Ärõ¿¥&¯Ë’¬&¬ËÜ£¦=«i\K¨ ]Ù¤S'~ÐÚ‘°!/¡§šVªš‚,Ö"ÆŠê„ëý—¸ÜÑm‹ë˜Ž\ãÀéÍû.ú6´…”S}7}·}Þ†Ýç‹W==h²korå¡ÓÅóYËÒÞoÞp%•Nø¯Ë„…b”h6Þ‡:2ƒmf„¢|'l3#Úh;ƒmf¤2éYÁ`›.š.$H%¾±t-?µlzõ…󄺰L%Î g,IïtžP9“Nà6œ·Kç¬2âŽG*âŽG¢ixÓã$™†71N‚ixÓã$nø¤žŽ£ƒœy–Me r˜%œ§Y"ˆq¶1Ú<Û} %O´Ä#-1x¦%†j£MµÑÇZbð\K 1Ø^AÕ)Æ¥g™ÐCÞ[ŸöýœzÌ;‚Œ%V>ö wEˈEÔ;¡…¹ü$×Ã>s[Ü;«öjŒûWxg :‚u²3.?¼8zsõý÷¯‹Yûãþ,oÕ¬³á/¢ r} endstream endobj 155 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§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 2 0 obj << /Type /ObjStm /N 100 /First 805 /Length 2425 /Filter /FlateDecode >> stream xÚíZÛnÛV}×Wœ—⇞û¥0 8vsÒi'Ng:uý@S”ÌšU’Š“>ôÛgmÒJ$Ù’¥Ê( t€¤ÉÍsöuíµÉH&˜eZ0ÇŒf‘ùÀ¤`RF&5“äóLâ_Ä/0…J2eô@)¦~–©H§L+Ü‹LÛ@Kê(™ÖÌÐ ÃŒQ¸ÁŒ7L;f#Ä<ó^ Œ`Q*f$‹ZX( £­…,)sPG™À "u4ÌB‹°ôÓÌAë ÜÀI( S”’Â3RÖJæŽÑ2”ŽÌEƒb–h(ˆ”ö‚yçƒä´š¶ù´mP”$8HÞåM5¯3(„Lì®|›‹ôEõ‘]ÂΈ­=êÜHâåëÔXUÕKŸL§–» Ì£‘Øt¸¬lÞI ’U=Ìën]y™¼NÞ$§ørIŠd-»^qK•i·Èb@TÍNqãäÎçW-ÖLÞÓ›ääø¸Û!9éÜ”œ'?¼{C¿g×m;û*I®Šª˜Žªz‚¨e ŸÌÏÇU9Áé˜óäz>¤êÝ6¬Ûg³æâ³æÊî†Æ9Nµ®ŒçÞR?€ƒÜ [pV´9Gzg7e‘7Âw!¾Óf)˜?þ÷'”¼†#Õ?(íé¼,/7Êš^V:.©»ÈZá¹EëÚ&«Œá›ÚY€k+²/‘Y/»–x—†@ N`3fqNŽº;¼ûÅeˆ‡Å£ˆºXœ##äâI¹XèÞ¯%’ïë*;Ï $ûÙK–¼Ï?¶ë‰¸VάWSUí²½A¶·Åõf8¹¥4öÎ7ônA,‚£Ùpj Ö*îuØœnm]}J§ÍMú)r9áöU‡¬ù¢ŽÖŽû®í Þ1!£9š¥5¨[oTçõ¨ªZì–ó'ÔÃ8„'¦9Zè–æ  Õ©¼PQ1ž×Oª„Ò†;A ébˆKJ.ˆ™!HÁëG¼¡6*²~êÈ…!šã8ŠIÃcÇ%<Îí Ÿèá5/¦u‘–³¼á£:ùýC^_åùÍ2p&œ^…:· ªÖe·AÕC²ÈJ4gY"wr'Q#©õ¸deŒÜÚGd‘r‚ؽ T‘»É"kAž7ë¾  úé×r?¡ƒ_ˆƒôúÅu0a¿€jLÞ²Á®ƒlÐ¬ï¸ Ý¡W:¬¡ëiwLí,msöìì+%dÄÐ1ŒÄÈDßGW²4ncä·U9tîé Tw!1â{P`*Œ±&Q±mÛ$zSL¶TçþzR*zà Àè A‹ô^P÷A;hBy,´˜ÄÜÀ `H[ÇÊ&§ýÉðíí-¿*ª¬š‚ Ýóªo|°–uÐÜŒ˜œ,ªØq ¬“òÉ•fÐ+×Iú=áŽMê2i'aF¿›°rÔeΘ5 +þ8Ð-£µ2áÂ-îá–tὯî Jª»£9 ªVÉ„´H6¸["”G?u ²R!µÞ6&«'¤4 ÞP™BåºJP˜ÐÜ·)¡ŸòÅ…æ‚\à\»7”¶“£”qg-(Y¥ß¬ /Æ`x"˜Ý„]\Y¹£0½—q;ªaAñüú(ü¥À–*fã ºRaDÐzO~w¯ÂÔ©°†Cë6‡Ý½?e8|t°£¯K÷';©ÿV£Ý_?ŽÊOß —²{‡%õƒK>*úðºË9fäß)Ç$•߃9­ƒÜáKÁEM|DUwÿAàøø w£/ÔwÜ› Éö Ô5îê³êÊž}-›»ÏA»”°jc ÷÷þ"Pü¥¹,—‰µÿ˶ó¬¨WÝËrÍidÄ­èÿoXÉÙ2;r¤êêÐJUeöørº&¼á{Äÿ>kj endstream endobj 164 0 obj << /Length 2973 /Filter /FlateDecode >> stream xÚ½ÛnÛ8ö=_!ôÉƬxYl÷¡Øé¢ Ì¥Ó<,03 ¨±’ºíŒ7íßï¹eÉÎt°ØGyHžûMyuuñüµ¯‚N¹êꦒÊ_«ª©*TW«ê×ÅÃ¥\|€ßº¿\jU/>_Z·h÷k˜jßß;øuâritX\!`‡pÃU‡Püþþr©šN|À¥-.ó‰t<üz†Ä¥ÝþlÚ»´Ò>ÄÑn;ßû€_"º‹ÏÙó´¾»á…û8ß­ðàkøC7Àñ¿_ý«ª«¥”"XË¡…G‚ÜTZŠÚ*ÞUï.Þ^¼Bµ—uåDh´'½R4ÚVÎk!kÃWnBÚïû¹«pÚ¼LÕ¹»•ÎÉsw§³+ãfî† z´A—åø[]× Ô0N(`e‘S±RFUZ€ÆD'µ‰2¹&{é{~cm$C\ =Ñ$z²¦ï`ªxæ~ÐGY Y:)¼O8"ÁÊg‚K\ÓdL`ÂÈÒáþí%¸Dbƒƒh|û¤“Zh¥H)ey­;¸5"Ê´>Dòà&´*£¿¦5y°/-; ÚONî·Zšmä̈C°ƒp~1æF:DOxë“a.eÀ6—Œ“Æï.—¶ ]ZŸlt¤B4" qðˆëÓ” (b¨ ˜—SË“µ°€ Ôçä­°‰s‡a.€h¦s>…މé¨Î&8k…7ç,G‚ý;ïÆ¦3R ~Ž2 4ÉÊ~t%ÂÕ`Ô¬üÌÌ?«<¾RA€ßã+o§|ÔT\¨Yñï¹-kt`K%…ŽLü8cïZdåW3‡X¡]6£NQ àûº!%ã%s*êÎE&S›"œ–á)t2ÀO *Î!Ï„èå!¸Í¥d ¸/ ;x›œœ­gÌG ߨt×—Yd\0#»X'›‹”. cð©þÈ6Bvªÿ™Á¨rZ3BhÖ3ö ê>z„–•SWL0(—dtjjò9Ÿ±p”ªÖgL@ëFÔ^M`7ç•„‡²¤õã ÏÀÃíP?%Ñ€™O(<º6rÄûRñ (¸Jñd BòAÖ*èBùäB%öWàÓñ M+J=Šz*ÚÁ¥URS4já¤=K#˜OâÂ_¥ý¯ÁØoÃé8 DåƒðÔ¨§†¢„„ãxŠ&×·žR;r#/å¼±õ>ü_|J¼–s^à(˜I€®Bzu>˜àÍg‚I> „ü·0ë3 霻˜Q´±Q‡ê=@½œ»0x5¤¡'¥crŽ7ëz¥³Ó\öHÂàU%’ ¦•†À˜šÆ/Ša*§p<)§âücÊÕRm¶åù«®}i׬ÙKé Œò~\–7”%±ƒÔñõ¥7”–îs¾º§yÊ-ê.›£X~¼Ìü\qÊy&žÜaVÂPH¸ïb– cÊvá™jïëÝæ>¡ú³[Xž­îû9leYŽÏV|;çiê ÷#!k‡Òaƒ‡AÎÚ…æµ+²vÓ,Ö‘ø{OHFIueÕÂtàr¬zª”gˆ©ê|Œ—‘6þc<3'­“lS0¹^ñÇõJÏœ'bQÇ"ï–!wï©´* ”\`ô‘¬½átirœ#<80ö¬tçjE©³”[nÛÌä>Z5” ㉼Àœ R5ÔÂS>ï¸ÝÓñøDÑGkÈT|nR©Oº‹ Mâ+*Ü _qkò 0œ5ŒF(Lµ:ÃP õÙŸ—ÎϦ“ÀÈfðžOód: ňó%Kûîà€@#¹©¿C\f#†áX¹±lÉ–Ǹ Frt5ºìºùê·:ë)ÊL2kÞøÐBsGCEìksV©ëv¨qA’¾“§ƒš0{±OsGù¡=ÄîÚUV4j 8n±uSðšw‡ÔdžY~äÀQs æ¨á€än²£1aψŒ©KêU£?þòÃÐÜ£™Ww‡;î¾âLr»ôRø(ì{tÉ=i¬£-um_ `di°znüÃnÍA–z’M£¯¸[J ±å‰»Fw©Ó¦ kpÁLWö¼_4’{Öøì7¹9ûöIåèr©FZ•sØ8äÉpË£aX?rÁïï†{6䓾¯üɼ¦Vr„|ì1ÁÈm2»|®®·ìÖQ=ö¹»Âon ÷î°µX³–ÁÔ:.!½ÍMçšv]çVüö–gQØmÏ/í}ŠÃkî©Ó™—¬Y8&uáx½éRó¿æd ¢tè“Ö́Ǡÿ °¡™–yе«! ƒV‰€‘ÇzðïàU¡ÎÑ&FË£&6 ?M d”]Ê2d* 7D )îܹٓŠK/ùñì’¿>ÜÇÖ>žùl8 #uæé;2\CT^ÂOÂïüì°ei:¤Ü.‘À‹“7qúYÝǹ>¢}×Rõ€ÕOn{(p”âP6Pl@íóÇ…h¤«ACZJûâÄó7[ýcµÐÛ¼´Lg.‹C'>m8Y]i!ÃÄ‚ðuú¢rˆž"ï¾À´R.^íPõ¾$¥„ ™–‘Ê´ô˜?XÙXRq¨Ê '1ýž¢¡t©Ý·~Ni€ØÒ>£‘!‡Ú³So !ÓvŒTʘ}°§2fZCGìƒKaè<œ CEÑþ {¯rÉò{_cÀ9qì™=èžðÒb»>á»Ä‡Ï q¢sôñÀrâ;cîCõf¹¯²»‹oç |ŸeÙ1Ñ׉Y4?YD¿4`˜»Oâ|D€:PïoNñ`=ùf™>ÖàWŸóGÌ!q¤Mìûf^±S4J‚] ´!ÊóBþˆßÇТºSxÎÉ@Ìè5b-$Z$›N}¤§ýáD×m-û ¦ô;vùéÓ+âç$:*`kV˜íÍÀΑð`wË1s“ Ü’ Jié&Ë[=‹< †‰›4È_¤×ÑuVñ‹¦³¢®ÍXE\X–5ó8ÞGër½”ßdÇ¿… :çŽo§Ý(JGÝ7êÖŒóO¥Ò²Z±õÆ4!Grlì‹ |˜†41ù­Ÿqâú¾ãÇýëCT\NóW±½€cfËzÕŠr.Ë™j-)™W-'Œh€gþ¯À¹!}6õ lo(W"niJ·)¶´Nh]dòE£õê—¥ðø¢\ÿù{$…²6ÃL› R(ë'áþU—ÓÏ»¸?–* õ1S‰;ê=ä–wì(õ×VFß:IbqiÉÊû8·O:O¥ÅCtŒ°pË¥<¸•ç ÜJ6ðf0p+šñÚ”_£ûø±I¹SL=r‡Üe)ª>È‘sµzˆÞVû¹ 7á‰Ë›öS" _ûÔŽë;¬1œIB«¹´†ñcj8& pr]¾æ5çÒ™ä/ÒÉÇ@dèÚí)o¿™ÿ #]î¤H»¸'©ÝäÈëéX$9ît¼mRµâ¤CÕøþ‚£—¶ŒmJº•õN5Yý÷$©äYÚ9Û[å¥åÉÕ!­Ô꼪A1‘ €%,Ž ˜Ø1ì¡87û ß´ý^gþïF6B òÿ;^—}Õ¹z¹)*zá}üºàSJÿ_™ÿÎ× endstream endobj 159 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/pcaMethods-013.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 169 0 R /BBox [0 0 288 360] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 170 0 R>> /ExtGState << >>/ColorSpace << /sRGB 171 0 R >>>> /Length 393 /Filter /FlateDecode >> stream xœ“MOÂ@†ïýs„ƒãÌ~Ìî^!jB”¨íM8@QŒbâßwË.m%jRÝôí³ï›ÌG&À°…·â.>çûû«,÷¡3ºç~¹Ëx\FLDÐ=Ëñ´6Yø,æ@°*&ñÙ\_€›B ³(4£S ”Eåà} £xË qß0l36 ­ù޽ –Œ=¡ð VBŠ=Š?Á„†U?t¨÷2^„êXaéHšµA²1ª¬^`pýºXmvOû!TÛ⢪ƒ›ˆŽï¬_ƒ³ 2uÆÃ`zS^ÜÂlPƨç5l>ÖïC`‚Áâcóº› ‡s¨&u|;±ŽÄ¤cS_ )ã<> 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 183 0 obj << /Length 1467 /Filter /FlateDecode >> stream xÚ…WÛnÛF}÷Wè‘Lš»äòÒ·8há4n_š>ÐuA$ÒÑJJü÷33KQ¶œÂµ—™Ù¹œ9»2“”þÌÄäi’ÙrâÒ:© ;™m¯¾]%®´UÁ£!oÕÓ…›Ûbòkõ‰þÂVlÆ#£·W7wÕ¤N꓇Å$Ëi¯Ì&E^$Ye'óÉ?ÑÝÔDkú,;úßNck«¨øegµ‰nû©-£´ñ´¡ý^¶=W¼õFñô߇ßÈØ˜¤v…˜…Én9«£=ÄÙ8 ÞÜåõا*±YNÚ¬õIýIŸßéó™>ès/ZgIUgAiÏvrT77ìP\ºÄ %ì–¹5ƒÂö ’-ýÛKÜyí¢ãÔ¹¨ÙiJšÇ â±]ˆöY´j–#$.+£94±Ð`.Éj1O¦±I]I¯rq˜¤ÏróÒì—Ôä´·óìeœiTâËÀÑB/a‹Ènz¯AP.X}hÈí¢,Yv‹B4%úï°]DóžJ_F’Œ ;…‡s™Dܲ=žÃP/ÃmóuJjÏ6Èn C[ îÌ0XQt™< w|D¢x²ÔRI–ê P—¤%õDe“Ò1å(÷ÆæÑßœ›C£)[{®ŒuÄš•Žœ1*æÑUÅ`×Bô¥½?ËàsÉw›%U8ŽÊmôÌO‡5Ç™è«ó3±SdáHøÐÍĕ…¹ÔÏ‹ÒÓ¦çʼêd­¢†ÖP?3ºÑÈå;|%yžË‘ØZ‚¶°rŒÒ”4Èõ³ˆô ùFè³~§ÃJƒzÏe,þÂò«Ë øã×Rbç)t„tÛ¹´HnÓèauR]¸éȽ"ºï.ðÈ èx&DÕk¾¤.•/ó:‘¹KJ³ÒJÃÎOÓ€Ô<5Ñ#üb õFÖ<ˆÃtï;â é|aäÀ„xfh‡¯P|åÖÑ‘«mnOž©A?¦“2ðñÀ j€åaÅÜYÓRIÌ©˜6b·B®Ú­ÇC vZå¡ÙYâN.Á6¯”TÑ‚0v dí†jì•SÓb€ ÆçÈX3Är>£B]fSl'¸Êq‹¾Ã)Ï"ÙwoÑë=m¼—(žs›ÒÙǃ0¯,q¦hp*>M¨gx6[3¤˜«[5Â¥ µ k;qœR¹× <‹÷ §k@,Ÿº Ó±w-s ¬ò깉.Ä:\:™aÔf†Ûkî1‚Qãˆg&⌻f/÷$*īË“µÞPzù^…B2nZA’êò¹r 3Cu¡ÎØÔ6SÍlÆið#Gú7! ™¹°vÄ€ (dBe›€?Ü'ÜB° ZÅ®•5ñ¿Õzyó3½™ÿw]ÀýJsh+{ži™ ê3†sòšëLJ†Sˆê%ÇÝóm#$òs¾«“²Þ^Ù´ÃMÃc¸0Ï™ oF÷³r_©þYËïµÊØh–M¨ß Eg’¢vçLÞ6L‘+©A?4 t=z´"ðê9‹Å*yëÆ.éA<º²­^Ùñë ïWܼË%?€9¿öïßS§YÇdêGì”Q‹-—§ŠÅ’”kÙ±”_ÀV¬Öù~x#@ƒ¡tPó*«œå[ŠíQrbp>"Y,„æÃž85~Ïb¾P’n¶í›OaQ¡û, 'ðx\½8«jáƹ*ÉÀÌf~ô’î<%Ò_Lª;ó~\*Ò½ÿN=ï~+Sh} ¾29¹h.i~ Ê¦R>•íèÅ•![æ æHà‘£‡6Þ•÷¾Ñðsã>8µä—µ®ø6®ªÑªr–Ô³pç@øàߺBËz¨çŽ!9U¹£ =«Ï†óžÕ—j¡QSÔIV€*“pãâØ{W®þ_å¥P endstream endobj 161 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/pcaMethods-014.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 185 0 R /BBox [0 0 360 360] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 186 0 R>> /ExtGState << >>/ColorSpace << /sRGB 187 0 R >>>> /Length 803 /Filter /FlateDecode >> stream xœ–ÏO[1 Çïï¯È±=àÅÎï+ˆM í q`ÐICÀ6˜¶ýùs^’÷â'6­Ú¾ôS;ö׎ST§ Õ½ú>\ðëÝËå‡Cuû2hÖ©þýåö©â£ c­µêß7GçÙÈ©_ÃÕµÒên@uʯûóÔÙàè ‚Ÿ”mZDêy§‡` i“*Œ„üs,0pÒmÐ@ F°ØCBʾFÈNÈ H¨î™"è(Ürü-ZÔ‰ÕÚ7²‚Ð5$ä\ê)guƒ¬éar3$òÒoœhÈÏÒ´)ˆ! ¿ùç4QAjˆ°Ê„Á ½lÙp ­¨MW7t$DŒ.ëV¡Ï^zÈéµ=½†$$t!W²BA*è!6ÈÕ÷¢4žSoñ×V(èLÞ§ÀÈ}#Ýê\©  0PN¼Â”·èÝÚ©1Ù\Ázð¦Á^È07åI[°B>~ Ø`#äsÜW±B´€R> F7!ŤڻDÜŒR¾¹‰"x)Ÿ›ò$cÀ ù¼‡”Œ`„|ÜM>²ÜOÂmDÀÐ`ÌãCvß” çí" 4 ðaQlÓòô¬–Ïðyhn}c—ò5K.<ʪ˜a…|Ði‘g;ùÄiE·[±¹‡½¬k"$~¦LmÏäsp=tyþŒÐð\"½hMŸôµ«̧¹FkxnG!ŸIä-„[KùLŽðpûÊý0^ï‰ïšíg>PAy+k$Þ?xÚ“Ú>ª«ÕÉÓí×Ço»;õs­ø\¬nž×ê€5[}¹ùÄßhµzØ©“§»ÝZqS¬~¯¯Õöt8Þæ ¦íº=º­ù1yžë¬­Éû­Î/Ï6Ç×j{Ÿ=Ì÷á/<Ç^ë¢GUVÄ—ŽsêAm–çNž)–p®ËŽóéà)0ñ²ìx´yN¼,gN¤ó½Ðx]vœÏÏ䉗eÇK:Ÿ³ËúÏ&þÙ )_;Á«|ÝXþ²•æ?Š[¬Q›<±Š9íonÆÁ^ÌíþæÁçû¯˜û½Í©üS(æqsnžÙß Åq¶óIº‹á]? y endstream endobj 189 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§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 195 0 obj << /Length 271 /Filter /FlateDecode >> stream xÚMPÁNÄ ½ós¤h(oêÚDO/ÆÛvW“]«[¿ïmvC†yïÍ@¨h! 2ÒU ´ªeã ?²o&ÖÞ'ÂEš \dK¡¼?6°™Ø­kKqîyÓ±²uॷÊB·;ÛVZ:« à•·òŠýï‰ö±Æ(Þ\oÝCÙ!7V:S“]ÒÍÄ>P|-çDñ“U—žš¦Ñzí"ñ”=æ~ÊŽH½¨¢yø¤|Èhj†8Y¬îÊ´M6ÈCBÆU ù_|GDÞsƒ~™+Ìs¤Æk8dèùeäÛ_1qlNÉš~N J_×yàGÂo¯e" ¥h U¡4ÖgVdwû;Kk¬ endstream endobj 176 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/pcaMethods-015.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 197 0 R /BBox [0 0 792 504] /Resources << /ProcSet [ /PDF /Text ] /Font << /F2 198 0 R/F3 199 0 R>> /ExtGState << >>/ColorSpace << /sRGB 200 0 R >>>> /Length 5165 /Filter /FlateDecode >> stream xœµ\YsGr~ǯèGaMoÝGð‰Ò®´’ɽdØŒŽ!8KŒ<ÀpÐ6Öáÿîü2«»k ˹)Ë¿œêèá,7‰hàȧÈj¦l‚}5c%šxΣ_´ʌ1°(×Ôýsµþ}¸ü,“⛂„; °-4m¢MÅ1%_ží¯6×»âÇê†XFÏBtz\ÒèÌ? û¿ 7ç›á¿N‡ãõÕ ‘‘þ³]_žm†ÍÉàÍpü?vëíåæÝÉ/ëŸ>s¾Ê+&AÏ"´/¾¬‚ô_¢· q,izf!h~ÀoæqÉÀ(¨þ°!¯°‡ÞÕ¦x§Ç¯owûëÍp±¹9ß7ÒÜáèßõhxs^=ùæDA¨RÆdO@12É«¸£áø–xa;Ô^‹Ú"Ž#G®6 î› çåõêåörýa{¹=ÖgÛw4ÓNà=̬wMmíçW›ý„re¿ «ÝúìfÏëôß ý4»±~Ì…f#‘{p0ÙBûÏÖ ÒéLË’ÊJõñìì‚|6”F›˜—V„þåùö?·_¡5¤¯° ác\ï¶g·»Ç¯‘äÍ‘~áÙŸ·o¿¶P³öDZb™ðêÅíÕþýîãÍúâñK·IL¬Ï3Ö7dÏVùùì'AÊ„-6Úž¿8ß_8ßÏ6p^©^õ2bWàNM«V˜ØËíÍy¯ÐO†§WïIêè§ÿýöÅÿu¿ßóJøÄ+z™Šràx…Ç?ŸÝ¬ßmÎÖ—ûeßV-¯ÄšÌñOnû¾ e§-4‰Ú¯~¼Ü_ooö»Õ&öúÛʰšY,±ÏCØ“{¤`)²Ý´E°?ßì¶o·GMŠvGjBZhýdh¢õîæ«Oc-œ‹§X"ˆ¡{¾ÞýF¸??Ô'áv•ÐYüë~¶ú·“!c»mgëUñ‚Y:O&ÔOHOH¼|œ„I† ØÑJàDXÅmtë|2þª¥ˆ5QÔ½"§m‹¹™íÈ©î>QoÊ!v#¥ÀZ;M÷âjHûÏ_}HȾ@#“§Õ¿8ß\6Ú­wëËDzÕ53ã'—ølõœD{{È…ð© Å4f쟭žm>ž=rëQ|4£4l?^ïw_€Ð@~}2³Sy¶úãþ‚TúKyN”Ìžë¡¶¥þ õÀ™þª8Pt!òFN‚fy´½°¶ÀÊyG)”i8¿»½¾Ù^-Æ>9ª%"˜€JqÐ-Ó‘X¥.4z`JuTã’Eòï «x?åWÝm̉ÅÏ’²ü°–páMÒýì ù8’çö’e¨ª^E4àmjFîôøûëßÊÚÝ9BY—c³ßOh?žÝüc¹[ÀV‡BU¹wê¯+¿Så6¡ÊÓýÓùíÍùÕ=y¹Æ gö¢äòj¼ùÄ( ïÕèWq´™ ¹aÿýæ\èrûîj¿¾>Û_½–ÿn{±™æ&“úæøŸßœKè‡ SLϵio<2{žý»í\¹¸'0×E[¦rÉTvv|¶ýtG ŽÍÓ‚öéñø·ÎV—Í#Ë"«$†âíæfݹ,µkàè›÷M÷¾Ý\þM‚Øßæ´ ÇPEÒ)¤kŠõtb×[²@_»tB.…Ê}Q-¬ž¶ý[ÁüñæöS–êÉðCÕR僄.$ÍåºÕ¿œ ˆµ67RÔ¹;ËÑÛTW%|tÍ1Ÿþéé¸_49ë }Ö̼ƒ ‘ƒïQÞ˜0 ë%Ó¨(¿Mªé jj$’d T„HN¶aQ;ŸQ¨((ÞJ0Ÿ*R8´¬¨–Î(¢Ššâ (³H-’†(û¶œ½«¬¢E˜,N³‹Œ¢(P·)lœÒÆ ‹Œ£OZ*';"šª}ÌžŠ'ùýö½¤^ßîÞ/èê]”0¥‹¡ÜæZž wO ³t›Zÿ„i÷€ÜnEx¨EC3G°R25s!%;"ÏóÍnó·m[¹µZªä$Å•€£AÁjˆ(èíN΢m5’×v||š‚y2ÜÓ·ª, £¨FÄž2ÃÄÒý?Ž’3XÉ\f畬£•Þi+S,“2¥ …iÓ*Nc¢eÚ?®~¼|·ßmVî¿z·Ý_näî›Á®Þßlf1Y-çX;œÃÞL§´»ëíîv÷æd•W®fÍÛ­¾˜¹8]}2´ÅûÕúlss÷n¦Ï]·»s«hº“²\+úÄe¹éš .º‘žp 8öóm†ìªn°ADŠd0—CÈp®€‹/¨µii¦‰øS4c“œ!ãf"ÃÒ§Eh¢<—ƽÀÍñ€ý¥E<ñ½D³d²CÄ´;¾¨‡õQò!ëE}‹ü囆×G‰’¤P°çb\4(@¶œ’z†áhs_ àÄôˆ±á£¤œ‹ ÅÀ†R#Gl¸Pèc‘‹@Þp7-:ý„~FQ’•÷]•VèĵÀ7°|ò(¤6p&> #aJeV_<'þ™à,”)#öÙ§7X"¥TeHîÁOú9ñzW=|¶ÐJ‚i#(ìd'ëwD˜fþ¢7™Ÿ{¡?ÎË O()ñúˆÑ0m9´õ‚¿4?-Kð{#=ª‰0ÁŽcîYåõSHzœy8£ý䌰„ã*ß?³<όϦoβ^h+?/|µ`ƒì°}m‰Ü283l8ŒõØÌgp G† »aàÃü–C‚Ü $`#púã¼…Çg¡7:AÛ~<Ï¢¬‡ØKñ—ã°T‚™Þ©ÍÅ´L¿Äãq"•fú Bçý‘/> Ì/–4è0}}ÛEã8ÉMQ¨H‘å¡`?µŽ^ÄÃ`¹µqÓ@é4œÅ“p9,–d•e»ˆm(NÀÌ·"ˆDHOSj`dE%úy(vfÜ;¼›˜~P V{úÙgÖ¢Šw“Ô¶ŸƒíŽHÑÈNæ‘iF> "ŽKmx7r¥Ëˆ,*2Ý L¬Ÿ‘Ř@>ì„90üÔ²u!‰µ»pY>Ê­à#Ï×Ë`{XWÉ&òÅ¡ÜL‹ãòDˆ¢™8òòÜmÀÛ—œvOôVjL°Æ,։ˤLE´ÚÛ*J3¦·rÇL²xì`}³0|¥v”Iç$eA¿°0äé n³l° ¦éS\Pa¾¼L䜘ú¶ é˜i«P£mÁæ¦×²A›ÚD²}Ī6âHAt&µ|=v",Yy*d7E¨Ñ˜b²(lc™ Þ`c¨9d7 =ko²Ù©2)ÞÍ‚äj]1cƒîg!t5¡3vQ‡`!Í kϤñvÕ‰áwÕŠ<‹jÀ°MÅq8l,³Z9Ü̳ҹÒ’¨$ÜûkQXW¢¸ Qg€¬ MÛ¹Ý4.ÖN†YÚ¬…+Íû7kØÚÅÚ ±QÆ‹5¢ìUd Y+ÀlÝ›5ƒÓªf±v€C^¬!`SkIY9_¼ž¬)àk ØùÅfõmÖ·Ðe>±æ€‹]¬=`ÙŸx†ëâ-x|\¼ ãK‹·Á|Ö/Þ0›žæ­‹·oXtD¼ö+Š-Þ0³ºyKÐÍ`ó¦€SX¼-èoëâ‹–‰·¿ø^róæà¯¬W¼=à"f‡£ȃà—hò"ÞV¢ H——è‚£ Ÿ˜£Ȧ«ÆÑ D×JôÂa $[Ö+ÑŸµ¤ECÐ 1ƒ-¹Úä±ES–ºh ZVæh ¯Ù.ZC›½D«ÍA…åýí#ÞwŠ“ýMÑ¢iÑÙM¢O`‰6Mš¢M‰Fi‚¯E«¦¶[ä-š•òÕí’š%¦eVßn5DÑ)š¶YLÐmÛ"¶~ŠÆþNÑ:Ô´æ]óþS´ïB£ß” ¤¶Þƒlaº×?|Éú‹¿, 's&“‰KfƒLæîˆž?ÙŸÏ#fxA¹,ìÓçkü¢ÏkSþÎ÷5>Ãö¹¹£(¡Z¹Þ#…Ì׿\q‚@Ñ ‚œ¦“zý~û”¹å°Üåù (¾z"› 4¸´«m@à Z÷5 ŧ“òûÄW](Ln œžBd€eÙ½ßú”Ïñ ?"êÖ™óÚ*˜=wŒ9ÜJ›_×01RI¡–M¶]¨a &Fq$é9OGÐ@ `"V@ZÄ-[P0$¤ wÛ P01¢ÑŒŒn»‡†‰h Àì^\V 7‰X”#åDðuÐHò ÏÍìa~]%”âûc‘›÷&)ˆœH[8q;#ÐH2q/7ÀãŒ@#ä™HƒÐÏÜŒ@½¡VMªlÝô…œ×Ö)˜ˆH¢HÚ!Þ×p!9ίñµ±Z 6$Ïòmß­AÁ‡„‹ÈMT³29#’çâ·÷ƃ†–û÷-š¾;:h²&†2Ýðf L…ËfÈôÛ·Û^;Jâ}ø ¡ŸßWICâZ´‹v6¬N§”)pþ?yétZ™Pq1¼†vG‡1h¤¡ð=Q|¯µä0•^’Uµü}Åv—ž1(¤![9D}t2NEÁËl³tãÔ…‡÷5¼Ì†¯U¢ÊëÜ‚AÃK¤í…#I³ìAÃËL1ZFð1‘n ^bíæwë¼e ^ViÐ2~ºÐ¼L…ch.u»PhvöŽ‹G9Ï”¬I# ¸°Pù¬Ç›ù}•48é¦Ä¹Œ[0(¤!“§ˆr²“—=¨¤ŽO¦À4Ò€69Æš´TõŠ$çSó]Æ Ñl'ÇÀ0t/fiø²ÚÚRFK‰ûRÆýµµ®`ùÖ"e4&Î9!<]ðèÿžŽÃ) endstream endobj 202 0 obj << /Alternate /DeviceRGB /N 3 /Length 2596 /Filter /FlateDecode >> stream xœ–wTSهϽ7½P’Š”ÐkhRH ½H‘.*1 JÀ"6DTpDQ‘¦2(à€£C‘±"Š…Q±ëDÔqp–Id­ß¼yïÍ›ß÷~kŸ½ÏÝgï}ÖºüƒÂLX € ¡Xáçň‹g` ðlàp³³BøF™|ØŒl™ø½º ùû*Ó?ŒÁÿŸ”¹Y"1P˜ŒçòøÙ\É8=Wœ%·Oɘ¶4MÎ0JÎ"Y‚2V“sò,[|ö™e9ó2„<ËsÎâeðäÜ'ã9¾Œ‘`çø¹2¾&cƒtI†@Æoä±|N6(’Ü.æsSdl-c’(2‚-ãyàHÉ_ðÒ/XÌÏËÅÎÌZ.$§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 205 0 obj << /Length 237 /Filter /FlateDecode >> stream xÚMPÉNÃ0½û+æèìx¼Åî­,‘àT$߀J E¢ Ð"~Ÿ±"ËšõÍ<½1‚¢‡€ÚÊ 4h¤ÕÓ‘}1颰L4ipÝ[ýÝ1ÀÍÂè] qá éUbý Ê赇4:'‡0€6^:e íá‘òw²·Ÿoò¯0Æð°é„µ†okùIÈÙr®õ2ç¨ùiÊÍf7”›šÏ”©õv×äV¶©,¼œ³ßב߬ 7ÝsºïGÙÚ ´&Ò¡Eð¿˜u%Óg'²'åT X‰ÚóEtÒ9ú+ë¥ò®²ad·‰ýCWZ\ endstream endobj 178 0 obj << /Type /XObject /Subtype /Form /FormType 1 /PTEX.FileName (/tmp/RtmpBWQQBY/Rbuild32a06b1c9fb0/pcaMethods/vignettes/pcaMethods-016.pdf) /PTEX.PageNumber 1 /PTEX.InfoDict 207 0 R /BBox [0 0 576 576] /Resources << /ProcSet [ /PDF /Text ] /Font << /F1 208 0 R/F2 209 0 R>> /ExtGState << >>/ColorSpace << /sRGB 210 0 R >>>> /Length 7870 /Filter /FlateDecode >> stream xœµ;¯$Éq…ýû+Ú” 6óU4äp! "!H%—IC_qNDTÖÞ™>»kìÌÆTUWVf¼2+¾Êúøõ£>þðøóÛ¿=þüèç³´G?ê³÷GÝ÷ç¾ç_ùŸÇ¿?þ„³Þ~ù×ßþó¯ßýî­Îgü»ÿAq+ÏnŽñÜÇãÓ÷¿ûÅøûǧ?¼ýã'þò—.=¶ç6ß]Ú¾õÒZìyÚýÚòÍ—¶ó¹Õû¥ß~×­?ÏwûÍÏZù,ïvÏKCõÎóY¡PÌj'º¢}vþó‚]>ŸÃþñÛÿj°SŽãoŸ¥•‰û|ø,+ª´bªvü¬Q%rE•¯¨Òí¦Û¹ÂÈ{ñ:;m¥5Áó²õ¾mðÉË–®ãñ€.ÞŽs&Ë–C¼7Ïj¦twñæKJ{Öõû!ÞŽ›‹<–/ñvÜ|_íAþ á­èUšóù¢u4ó°cúµû«×Z~b)¹ 6çÞ4Ô›éK“\P·dæþÎ}›ýÉVêIšÁR”ÈfÒp2»IË`fb­Ùy͹Á´ ã7æó˜0-“ñ˜Hzåº=÷~†lq¼÷†~wÙž®[¦³m”·ç뇲iyï|ä¦tsÿd&wÄn™Q!Wk±éxu¹!Óéæ»ŠŸo0)Oÿ½·ÜG‰ö[Ä7_Ù‡µ—í9ÎgÃñý„¼?§]?zÜ߬ô°ç7ýõûãÙÊÓ¯oøŸ>¶èSÖãì“wƒaíß)Ÿlï Ê@Þ‘!~¾e¬¸ÿŒþØíyí|KèZØ>ý¸ëƒÙRóë}<Êÿ¡Ìã…N™ãYر”ÙÅÝ~í1ÿ3âç1ÜÓG­¶Í†§„ȇµìíìÙ÷fj½òQÙ•fg£È–œqíá?ez=½_Ú}°/tóÉ£º"ŸÚ”Jm"otïÓaˆ¦¢ÅÅŽGÜ7oÕp‹2…ÜÅ1}äš™B„:VŠÏ‹æ4j߉ûnnì&6ÜÈ΂ímá;ìï¶S,¸¯i"5¹¸%˜m; ƒƒs;ñ“ÛlfÄ6Ü­ŽZcJyî4JöF°Ãµ™x¸ÉmøW݃Ù/²¯ °…lŽŒöÞ¼7L}­Ï ¨›»Ax—Æ_.T~»=ÇN´Ò7 œ Ûéʶébéôt;E}k Å‚‘ ü$õÞÄZéFé†Ì*àEL,ÍEt»µµð§†ûÐWLqß販*¡±Òá÷vO Æu²5Ž£?e­·ÁF·Õ œÔV#[÷áŽG0¦%ŶŠN¹ßÌy¶Õ9Þåêº:}£cëéËÑí5|H Šåº 91dus݈µ™D ·MÔæ¾”ÁDv]¨Š9 ªY(R­‘Xºš™ 9ÛRBDçc©¨ÅÞ~S`›#íc©·¹h(¿‰m.Ó(»{©0œ²ûp‡Y!˜÷et–€»Ñ¹IšØË2X›ªÇ2gÌÜeìæÕØá L¤i„£°ù%܈‰[[NÆDöU¸ sˆg]ÊDªJ¸/©„îÜ2–ír}%Ü1BäÜmBdOºS…ÈŸr— ±–å‘!³³Âc·i“ÓcytÊcy|Ê·ˆ@y®ˆÁßÛVDáý¶qØœ[D‚<úŠX=Còˆùì·ˆW2âGD,áD2bšì+"ªÉï"n©‘ÁDD6Ù#nDlL/ü¸Gtt #ZD|ÓϨ"#0™Ÿƒé“ŒÂäÈ0<ã@·úqÏHÊ÷Œ¥d†ÉÓ3$ÏxÊý‘ M!3¦’ãU™™AyÆUÙm+#«löÊØLöñŒÎ†Õ#|d|fòÞ‘šƒ ̌цê–%¦ÒC\„{ÏaÏJ2 “óž³*—º£Néõ¶û”Ù’€CQ·ÎPËMå_¿Ü|îyz¾$´½šËÛ9ê§Òøâi²c¥ñx‡°ùL@7„ŸâSЦè\åD©ë²|v©ëaŸŒ¥Syv ¸«MQîn—smçD=20b‰°(Zw08™î6-8!«B||Å9W…„Qçì‚^Vyrë8dtÐxårK0‘7Îw…»3q§Ê*^¶! ‚®Â¨Y®ÉкIÖfÁmk ÌU n;×V01h’ºžw䑯÷Ûék;Lîõ›ü K3ÝËL)£È‰KÇüP k›gCœð!éŒìÊÄçØ¹2ª¨Œygä6òSÑW¬€l>±V.÷W,P¸©Xj÷$tH‰ØÜ°ÜÏ9¾ÐtKþé%†f¨x·Káʃo˜rp¹Cxðýô1«’ƒ«¾¨}“–¸ÃwŸÒl³\®]»9Ó ¦Mø—1Ëï˜Ïkff*J@=cÐ6mžÌ0ŒnÛ¥Ù^(û)%ÞÓþñrD1´éIÔ)-­œ=;Û( ,E²³ÃÓæ&¥ûð%Ò"9·}úTéÐR oP ¤´·úb®rõðõ¤Ùš4Qƒw¾¾Sf:›çÌ»âÕQ“â¯!•…_³¾zÛ³„íÕ¤·x¤¬ ÌŽ7 q%}êžùÒÕ6ÏÙý}¬²(pϽDïÒ]Wizyìî›Í¥7á]¥p²3|sùPð.xÁëwŠ¢o•®m4m=aÇk¾QDó×÷›¶Üm±ì,^쥥}§§Ë’wj»¯AYTüb·O/ܵ‰‹fí¤)Ï]òM–21Ce”W¯)oDŽ=N%–Ó+8´ÄkFýGÕ<ÓÎ7Þ<‡Òç˜Kc1@™R²(—ªÙØ¡/üÔÉÅ|86%mCꄪC”d*Ý«ŽÍÊ”^¯}ä†8Ÿ^’£$XöüC¹Úz}øB£ôºÙ2®êUÂÊÊO5ÇZQú<¥Aà -') vÂå,gEð‡Õ! (ñ?¿dùŸ?'å^8–‹%~õ™˜''Á¿•dl~|~Ø%´…÷æQ—ÖÑÀ]®ßúew”÷:îâí¸ƒ¼×qoÇï½Ú¾ŽcÀ4Ô‹*j÷¯2½·+_Ez×¥¯½ëÊWÞuå«<ïºò3œw°ô; ?ÔŒ;—‡¦¥´Ž"'Ÿ×Q—ÖQ”›•çáoÇ·†^Ç]¼÷Æ\ÇWÛîýER7—€^&ÝŽ‘KÛ¿¨/^Šj¤Hô_…äPX|¸Sù®*ó½V}“;ü ¼Ü5Ðé×ÖHì_ÒÓ-ºÓz,ºaj»Õâ²Ni_µº¨qôZc¯åÅË”m[µ¾(¡<ÏU Œ—¢q>k…YbÞW-1–RαjaªõF×Õ–µê^«LGXW-síY;íù\íý]-4ó±j¥Y0·­ZêÚóþ^k]ûéXGÔbã…tÝW­6ʾ¼ÖÙk¹ë`3®Zo”´]‡òÀ ïXQ€z¿m¬Zò:²?¼ÖoþóZô:o‰Zõê_W-;4·•Uë^AçU ™µÒQ+¿ö¶jéaü^Ëïµö¸ÌY¯Å‡ìýíµúÕÂ\«–2i«¨õG3Æ®«‹¾ã"#Õ¸Ñu]_œ5Àc9-ç,do¯³ ½–݋ґ™x-½³Vtòvc% }±϶X Êçb1(ÏÅjPÞËÁß[tÄ}¿@Þ½\˜D‚/‘@d¾#&x4& @‹®ƒÈ¢y‡WÐo¬Ùw´b[à zOéX D*‰C33‚>ŽÔ@$>áÀ ‡À q(ÀaèYGy †uÐÚ6]#Øät—èT›]뀖ìøDB'ÁN¨×Ž.A;K|–­Í[åØŒ’JÙÃfŠÛ¸#S0ivŽU5¡7Ç­ ÇcÁXð¸C0´º,«zÝmb^Ðqçç€À7ÑVƒÆ—€í¬íðtÇ¢ë ÿ®q`ákˆµÒ:lÇ<ž> kƒm´ äèC‹ë°¯º¡šˆ.Ç9xô„œY À9nâvñs<9`ãü©†tèºQ­>ÜÑ 3©ºèºZ΋Ä#”Ã=c<`Ùã—ýñˈfxçñ¹u ÿöÕ±ŒæÕí@³Æ5&3LÎÌÌ’è“æŒ¯÷øPO>DêÁLª’Jb»×Kƒ0õ[ê5‹¥®{çôAvÅÄJm½´†].•¶IŽÑõÝâa_ÆpŽ0ZÊÀ›Ñè6†Wçe€gu´Ì­ëHã2Ý#¬ÏíúˆÌþOäÁœÈQ.waq)÷%'É »£98–酎ͽ»¨csSqÿu ·nwn&ñ‰Üóü»Å#lÙ}æ&åõHÜÝí‘ì¾{cÇÍY=b‹ûò£G(vWd$öHptáG Ä‘£¿ 3GL…ŽAƃ”‰Î“y C¿í+Ä[Fµûù‰ºû3 'rÓäó–Ò êàö=ÿ©rgÆ[Ï— ޾1šåÝ· L?<ñdÌL´úÉÌÕL_hO‘ÊÍ-¿›ÀLÏlÝ?á‰àŒ`yâ \7ÓÈR£Í4¥2ùчꠄ?r¤©æ®<íˆ4¶äÍ#Í-Ù ‘›/¬7$Ž%7$*Hž§Ù5‘ÈLÃ{"x‘¦èÆwiü7#qÕ‘Ae¹wóR”50Gâ0gª;‰Äqš§Ô"‰CÇJ5àŒVtª§#qT¥åDâ¨2JI=J(ˆ0dDâà蔦‡DqGâ8ç”Öš‰Ä!•š*©lUKç‰CÏ5‰ãhWËØÕÒ ;ǬKa›‰‹¥ár÷þÆ]u"qS«° "þ\qAÄÑO(*çD\õÏ6w'‡`£¼7"î¹)w"—žÝ‰8Ì)$°Ë‰86©ëœˆ‹åAéHÄYrÐe §TC;7%Q Äa®ª”g©žò5x8LÁ•ºÞàáàà•øà<R©¦31®nx%²m§@6ÁÃaž£);y8ô›T_rúl¿Ký<%Ë…¬€H ‰ópÌï•ˇãú‹ÂøxJP‡z5–Íê!±§ÎÃq""t»ópȨ¦„j“‡ÃŠ_—ÆœÕÐÐ8¥h1x8˜ªÒxçáŽ)ÕC9Ç—ÿJ6â8œVÂÆI˦9Çá¸@¨ãp ë mB7ïJ­(q8ä¿ZêÎEÎZt޳d‡ƒ²KU*Äáb}A©‹Þ]c¤Ìq¸êß+Ræ÷[g¨T¹:Gÿ&}À„ï™ðæ@*…'Wýc¦ÂÍ‹{öM²4Çá¦V4–.%j•4Ì´h-ïÞm»`¦NÃ!¶(•ÉAÃq5XÉÙý–…r… s.Á—ž²¾0SjèÆîog3# ÇÄU¹š4#±Âʆƒ©*1Ái8ÌìEUm•ÎIòçƻö¹§á¦Æl8 §¬­‚†ã+å»ÕgISâøæ5ÍQ0‡á0­Vœ²ÃpÕ¿µ÷zB¯|unnÒ˜³p\†ÑŠªÍ3Ù4GÂt›Ï4ÆÆQ8Ž—ŒÂÁ@¥9‚£pߨKYþ”³|'áX %­Ú‘„›UúR‰ƒpHU¥Ï!µØõ¡ˆUÉœTÁÆ”Zráà $Ø• ÆLbªvÖ%}qŽ.YaM®Ê5å}P€p3dã²<¬ôG€p(…ÐôîK@ÊÒ‘spUüüspL]ºh÷²¥]ãÙcÝè”>"àŒMb²Ü·ayYéóé%ȇ¶€Á·î^¡4<܃²°í_)‹ÃÅ‹_EšÝKMQº*cpcצ›¿E©ªðíWÕhORp¬@S&,ÜÅÄD¹Ú)8(‹ÂŠ:‡RIe;£²zWù=›¿Úo(3ª(š®Ú?‡à°H©\í\ƒô J‡àøÂNœý>U¾ÚàŠÓ§rs‡àp¹RFã^EˆÓàê«&ÊtÐ!8:U‚ÛOi. ÊÛ¤%>gàæ)u[ pç¼Üò{}iƒ5?Ø|e¯Ëüµ¯mq¿ö³îqö¾)ß¿Å&M×¶e!_»šeG|&ß¶9ûÆĘü¨ýèšÛJ›åû›ïí…ýèÚà6|íh¹ºzmHWfþÌíi¾Â?æØÿŒä»†\c•Pcˆò˜½ÿ™¼ÈØ[.A¢ØÊmAŸq†±÷ÜuÜÅÛqß{î:îâ:{Ï]÷wñvÜ·»Ž»x;~oðÕþŸ „Œèn<ã‡ûÑ}x©oDw»ôÃíè>¼Ô÷¡SøË؆N0c:ÀŒMè>0¿Ít~ ƒXTd˜À‚å>Ö¡,{l]°\ì±uÁr”o°7Ùê –ã_mÁrÜfk.XŽŸ@o –ãb}ÁrÜc¬.X2ŠY–ã^[ç‚å°¦º ËAÆ»Ž„å ·sÁr° Ìv–ëNq]°d8¬~ÝŸ;R]°d¤Ï Ëa¼·±`9l³…”%a9ÈãËa£-¬Í',™É]ÀrØy C–°(zÂvËA¾Ãrø.:^%&,‡Ý·Ð ËAF¥[Ârø8so –Ã\ÛX°¾t~l –Ã\óËAFUvÂrø¢á·€å° Ò¦„åð ÂdËa.,',ÇoÈÖËá[•³.X2§ôË̯7XŸ)ªç‚åð%á¶-XŽ»D´Ëá»@°–ã^ 7XŸkØÊ‚å¸Ïþ`9~ ¢.XŽ_]¾ÁrüäB_°\| î¢åøí×márñÛ‹—ë ÿ0_þˆ9~>m[È>·}HfŒècAsø &ªIÍásX1þ¼X¥*‹›Ã;„œƒLø/È|'º ׯû¶Ø9ìè{Ix®ûüú¢ç —²ð9ìÔöÅÏaG¡ÐgÞâú‘]w¼àBè°cž/ºî8ÂÑA&Ÿ—í/¬–½0:ì öêúZ¸ÅÞéŠSIWºÃ¸©?…Uö‹¥+^E~Át&ÓŸ$MW|éát¿&OÇ»æê oÇ"ê¸w×¹:îXUoLÝ$ß½ ºÉ²êEÕaÿ®yÃê¦û“‹«s\duÓáÙ‹¬ÃŽ\ç ­ó»[—rÂuy~Òuù{‰×Åý.¾.ÚsvÑÞ‹°‹ç¹»xÞdì²?²ËþJÊ.û31»ìïÅÙq<.зë"í|4/ÔÎûbí\.ØÎUå¢í\“.ÜÎ-y»ÐÃîBM“¸ -Nä.”<™»°„îÂD’º Jì. ,¹»°¿ó2_šg’wa½‰Þ…q'{¶Ÿð]¸†¤ïÂs$~Ž%ù»ð; à…[J/¼V"xáÔ’Á Ÿ—^¸Ä¤ðÂc&†×/ŽÍã‡ûÛñÂ'‰Þ:Q¼pæ‰â…¯O/BA¢x)Å‹@’(^Ä™Dñ" ×þ–ŒR‰âEK/b\¢xÅ‹™(^ÐDñ"¾&Šá7Q¼ˆÎ‰âEðN/b{¢xúÅ‹Ì Q¼HÅ‹¼"Q¼H;Å‹¬$Q¼HZÅ‹œ&Q¼HyÅ‹Œ(Q¼H˜Å‹|*Q¼H·Å‹l,Q¼HÖÅ‹\.Q¼HõÅ‹L0Q¼HÅ‹<2Q¼H3Å‹,4Q¼LRÅË$6Q¼LrÅË$8Q¼L’Ëuó@‰¯Nð$;Q¼LÂÅË$=Q¼LâÅ»’ü@ñ®IÀÈn¼O¾Ńm±ÖX<~«Ê0l9‰ãq’tÊ4½Æq<0ëÎ4¯(#yñ…d•È£?=e$¾Œ“? Ƀ¥kH·O¯2“‡Ëk—™NÌã݇LæáòYe4;’OÍÃì1>ÿ£9‹o)U6’øê¡ÖuÕ{Íã¾á‡Ìæq9`—ÙÖ*"|\8è2ÃÇu¾*C|x5g׉Wõ˘ l¾0®a|ܦêT1>¸ ú)ããÛ˜]æøb'•ビ=Oäà 'C»FòqÛÇ]Fù¸â²É,÷(Õa>¬ÊÇ‘¥L¾³V…ùðÞÜ<•óÅÞ°*ÍÇi—q¾ÞâëÏ×ãsÙ"ÐÇ]u§Lôu_V‘>®Ó™é£ÊWêãΕ›Jõáj.ÏjX¼4¿~¡q}YH(‚}蹬³•.?ñDDû¸È7e¶m/C†û®·)ÝÇÍ»‡Œ÷q™Nçû˜SðCãã;…Òr \•Œøáæ[“¿î5é*äǵ¢)S~Ü߷˔֊Î"S~´·&c~˜€áÙEÌ›hV™ó#­QeУ“~¨G³³Hú¡”ŠõÊéÇ:Â]%ýPIˆÕô Äïk¤_ ~_#ýâg¾FúÅÏ|…ôË_ûÚF|Ñ´ŸD{×ïß6À&ý"ËB¼À³ìËÏä¢ñæÚnÏ¿[¶@´Ïއ¸@¯á•+y<ÄÛñq"±¸Ž»x;~¾ÉB\Çß5èjßOšaÀ,~*ÜÖ(ÜÒâv釛œ}xéàR¿€|³¿ßÿî›ï¹Áï•`±› ao¼T¡µ?ÝßœÀ‚Ÿµ{^ÏíóâøÚ?/1Q±8ᆊå¹…^ž±öÐ{ߪuÆ »èmkÉ¥môâZi½¸VÚH/¯UvÒËk`+½¯{´lëcÌwtßßhÕTþæ»Gx»ù,N¸_óöÿžj endstream endobj 212 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 215 0 obj << /Length 393 /Filter /FlateDecode >> stream xÚ…SKOÃ0 ¾÷WäØj$ËcIL‚ˆÞMÝ«ª´­@·Ãþ=IãôÁ˜¦éS<Çþ>Ûq'i4œj 1Š+”®ã#2¦I­‰æc”.Ñ{|—`ÉEü•°xáÍk⮗řŕԅŜ•w³ä3}Faƈ‘²Ç—ëþ­5W‘ƒ}FçÊgTPQ ׫¶ — ®«j’æÖ> /ExtGState << >>/ColorSpace << /sRGB 220 0 R >>>> /Length 3192 /Filter /FlateDecode >> stream xœ½[ËŽ·ÝÏWôÒ^¨Í÷ck!`  …á…¡È ÉN4óû9§ŠÝÍ+·4wJW^LÏÔ°Ùd½,úåÛÅ/?/ÿ¹ûÛ~¾zøûŸ¿^^>ÜùÕ9·Ìχ—¿Œáç/N†_<ÿ+ÿ[—ÿÝ}ÿÃâ–Þùå[üü|ç^Xþr÷L~Ë#†Ä_Áþz÷jùzÛÁ’üâýÚýâ{Zc\|mkñ|íË/ÇÝÚCXæ'·øÑù?=†[s<0pïa ßNu-uy{b^sà›»c‹¥p +x {×W?MðôBì@fzAáé…ê.60àã…àüÅÌr^«Ôy; ô²{†¤}|€ó ¹¬±/(8½0¥úý…Î_¸‘Á­°†~ìWÁŸ¹‹}XÁc¸ÇµN³œÈYËÓôO/€ óòž^håbæ Í­í ‡B3² ½8¿Ð<7³¿ àLO µvÐSÁéH~ÿTäªhÁïÿ«Ò ^\èç×÷w_}EYîZÂrÿnÉkòuúGÁv0ê@qŒ¾]¾xóårÿ3FïßÝýéþŠÙà„}-åd¶­Ãž@l·†àÆÎßÿ/wþÈJ>–5c©”V¶Q·ÆÆŒÖì÷ÿ½f©ÖJ)Š+¨údšø’ÖÒ cÍDRZ4ï„1OŸNKˆÕ#tIJzk”_òêÓÓ§Í+ >±xúâ)¬{ïÅD¹à•Ó;訳 ^×DÊÕ óõjø˜ÚdÌZZXû•n(Ê$I‚‡J´%Yr4EÞUêƒazZSAm´ð#R–è=ã?UC7 É É¾4òˆž¢e î…Ìõ>3<1i±§cƒ­l5,k K†Ìöè>±w8övF¹[*B\#­ ÜȹÊÝΨÃUa–/Jn°Ênõeé˜m1-¹¯!ɯn²ªuõyÑP×äœ5 &5,aÃ’8‹b×P xoò¥¢Ý›1º)ŸœJò£¨ Ž~f€`Æ ¾ç³G7­‹$C( .:À¹æF=È&»ä$·hg“C‘ 3éÁ$½ÍÔ¸6XÅ,›0®¬©/Íf’‰:pö¯œÉU1qº…rNìÝ#kkg¨ßPj3øÍ@”åöŽ’lŠm,ÒsyO61QI¯k¶¸“OÍ0¢–%$Õ5 ÞÏ1e즠²jdäápšù”×jÒ„È€ØÃ£žŠò5a“°.‡î· ñDwê7o‰QñÉÛ–[âÔi†’3éLX?`ƒ§fÒ2‡ÅöChª[˜&3„ÉÂdƒÀk([÷¦< ²hºô¶ƒ•»t0Q6g› 8ýU2¶îJJ>Oßx×òMZ»amÍÍ%CÈXŠØFà],RšÔ0@—ñ¢v9Ô‹uÞ,x[";¬„ÓÇ´VËì¨Ñ¯-C^â _-Â8 VÅÄíÈRQÆß–mcMUÎäÁ.J¾™Ò"¥§–M…2„"¥ÅT"Ìêmrà‰ÆÓgwÑΘO[ø%Âf’q¬“àm±i¹Š~59zúlHØ"eÆ¢!9HÈe±+ ™òÒ³ÍñÂûQµ›Í¤e„áýŠ…_•~é¢e6Ý_Yæ[Ü_a$&5K„]<ãzº?w^»Ý© –‚ý0 %•8«Ç±Ô˜w/ܵDò¨J•€Þ¢Ì<×dqÔY1ÿüÕL^‡«SŸ±&ç¬Þ’ÉÕ³¨-–¬qŸ5âG$ÉòÃys;I5>¦‡¦´Êš–·,ImòÁ„n+’"òa™3—ó´òQ~FÖXJDK™3h96Ú4=Knգ픊ÄÅS²R@9¶wÀ1z“"5 üô–^WEL?Ðßð¨ÜÉ®ÆišÈiƒ•Á’#Q©”˜JöT…$ …‰¡¡ˆs0­žªD§>Ú sÙóœ‚~ß"òÈ2ëy«©L‹0/h`,µ­ÔÅ›ÇÀ’Ï]ÛbïÇüÏ Ý‚hie¢…&A£öbît]މʾ¶S¼ÈÏÆ|K!áwq02P.ƒ«½î"(Ç]>àÒ\Ü…‡'úe—¬,EÑMìà…\Øe’GPu—Wì¡5… ËøVì ír®Ý»g¶š§ Ør÷Dꭀ̡ŠÒˆÇŒBÞÑ&MQIš°G…£œ)y˜³,°ú¯;eu[:J¸HF ñ%QÐà ‡èlhd¯ø±@À#P[4’ì}`MŒXÑ óË&ÙoÒÂ*°Ù¡ãdBÚÂÀ'J¥G:,£Àröî¾¾´Á ¾mi? ÒÚdÿAš|o*}ld`£7l粿p’ýCLåûU5›ñ|áûYåL"Ï,¾. ÆMš_ã GÕ£í}vú§é{<6ðÓz$s™ö“‡lûe(•&|`™}™ð…*JcûFèÆûƒ^ìH=Y´ˆ½]W²ñÃÅÁoåÈ¡ìPvv ôwnye† \¥˜Ð!+`¥Š†ŠRsC2TÒ îJhDj·?ä´JÃÀ.ÆM ~»”c!EBu ÕmtÒU¤³Ãm(iLoìžÈImx áw+¶”t<÷¨Êv¹œ"× ÄÄr5Kçµ×ÉN´À5åJ€µ“Þ¢ö)€¼l) ÒæŠ÷®‡`Am3Å}”Qºþ{‘*0# ÂM‹€0­QÆ›˜%ö¼©T“ jë«H Ä­Ѻ®´oê8ŠÚ/I3¼1xÕr'¥|¶Éþ˜Øw ‚Š2:‰TÁ2ç<Ä‚V°~A4óÉ–)*^¾§‚’Š„t`¯âqn®\À^mèö~ÙÇõ{7Õ¢±ì±sÓ~RVþlû…7O3>Ñi(°áK—å&zø<´|Ð lTþ(=ÙrC+äÞÅàFqÆ`VíêG/!ßj€•Õìoè‡$Ð\†CPè¤Ã!Gy|jˆYnzIhH!]h;d£±Ìc0wÈwÒuE`.tƒ*ô©wÇ*w³¢ã­¬ã)×Ǿ Iöý&˜<ôM®¾ì5ÒgpÊŒ%÷(ôѹÀ{Ba®{ÚÜî¶.ÎgîúuƒæÐ,#èÂîúuÇܼwÎ>}ݶ·»Lø^^å{NÓdòñ<Áç›nQ\Ìv/k¾¥EIÑÑíZÖ|IëÝneÍw´ŽÑýRÖÅ­i|»“uqEkß®d]ÜÐzŠyÆE¯q“o¾6ÝúÚ†/.M—¾öñùØtç+ì§+`3ŠŸ¨pUbî°Ÿi< W« "^ܧ^¯l0‘n^ôê5CŠÑ13½/ð·Gn|m{^‡gßóŠ ³¾4ëV,þî¹Í'¬1}øÙ´ :ƒýŽ×wÏÃc˜]y[7À¥Â‰@¼àháÎÈséö’›°7'WÈAO[ö=/þýêåëWgèÌ·xsŸãùÉ·À,KD$‰¬<ç›Æ[8)ØM⺻”óë÷æéËУ³’щ‘º®ŽrI†¨×#yþv LË¿Y^ü¡ÌpzŒx3Îjš!Îæ›0Ãûó»`'Ì8MÚ¯bÕ=LÌPøf3mw/Îúˆ':œå\·àRýÓrâ +ζp+ºF'œa0Œ§Ž é&Kã<,Ò ëëÿþúðãÇŒü–$Ær>òý¿} Ú._¼z÷ðúå¯o~}÷åËý·Ÿwù¾G§¿½~÷¯×¿¼~ùQœý¹û?«;Îõ endstream endobj 222 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 225 0 obj << /Length 1664 /Filter /FlateDecode >> stream xÚ­WKsÛ6¾çWèH͘ ‚ ˜ždÇvWŽ'V›é$9Ðmi,‰ª(§Múç»/ðaÉÎdڃͰX`w¿ý°8ž¼xyfô@™(1V&·¥Mäb=Èb)&³Á§àCy;TA¹Åëi Ÿzøeòîå™äQnµ¥•&Ê„ZÇ ?ªú2 Mž×;øQL‡Ú÷Kh­†¹Ì#øq4 S¥a³Ϩ`'jw¨å‚7­kÁVÌA¨–ßÙθ£Q¬gb|èLP,ÑØœÝ á× 7ê ¨ƒL+22^âoºÈ‚w»mM×ï42NB¥¢VŒ XxÔz›ÿ$¨q{µ¯ÅDÏk,ô4ó ¶ Ùç8A°ðUð}©ŽÃT7ø÷èx#ð?CDù`”ÄÄ(úŒ¢L]IÐÈä E†ßW¬›rAÆyïÍAÄÐNYåÃ…—ÍO5óqßùì¡ê‹=Æû<³¿£K ÒvAœ‹\–ÀÄÝ8¶u8Ã4OU€Âo5WÊxmò±ã”èÛ~¡”ʯY{'3 tÚAn±k 7¯¸’Q›·h[œú0£jgÉW‰æãùZ–Ñ>¯ÙcÑÆ7,B´çŒ éËdp\u4ßË>7#D "Tuà9ÚvšÔ)wEDá*~±© öý\1F‘K þ|×…Ä0 £=„*fýâžJ¥X‹Å·à´õƒÜÎŶØghñèBH3M[Òùm=—Ð,xàú ÊÌT”¦M½ŒXù¸½®è-° æ…‰Uèפ=â”»"}–;ÉØwÒð­Ü€(ß•Ü]PBå¨=÷K¬ß”M Sh¦ú+#•S>n1þç^1aéRaE› Ū.ÇêØ¹0·?ê#­@ö#¢ª±âRä@øšà´öÔ¶"Ïš€‘|}”È,ý=i–<‹ÕŠÛ´Šú¹¦©Y£Á(È[ÏDÅ®lGWíX.Û…7žèá Dâá“ýõ6Ml™÷ùÅHÉÃë~¥ô.øó›Ny¯F¬÷vÍã9ak KÄÀs$.­)ßä½±¨Ù‚äëtÖ¿ÂÖïji@çÂÛ£r[dEÉñQï!'M ‘/¦)c¦¹«êZzÐË?|3š+xƒÅ?¸Š3AÉã LtÂÑÒ{¡p^ù"½asyb°fÛz¡|E ëöž—¼p•‹¡ÛOz-Wûj“G°6ûöËG°ÅÅ[ØÅZûö‰]·i¤sÀ3=½ªïøÊá-a”ÍO[æ\¦x£¤™é¶ØT$°Ø¿½ ‡8@8¤;WpØ<ãWqu×-ÈÃ7¬KsMع[A§ç4³d¶)÷ÍéŸ3‚bM€¤Yä•;®ÃÄFÆ8ìŽÓ(Ž ÛW N¾8¼øZþUÄ endstream endobj 250 0 obj << /Length1 1686 /Length2 11632 /Length3 0 /Length 12714 /Filter /FlateDecode >> stream xÚ´Pœ[- ‚»C€ÆÝÝ‚wk q§qÜ!¸»»,¸Kpw  8<®ÌÌùÿª÷ª«¾>kû>{íCIª¤Ê(jjo ’²·3²2±ðÄÄ´XY,,ìL,,l””j–`пä” 'gK{;¾Xˆ;€à7™üf¨`ou±°²X¹øX¹ùXXl,,¼ÿ2´wâH]-M LY{;3¥¸½ƒ‡“¥¹ø-Ï¿ŽZ+//7ßîQ[“¥ Р[€lß2šmªö&– °Ç… °ƒø˜™ÝÜܘ€¶ÎLöNæB´ 7K°@ä rr™þh ´ýÝ%@ÍÂÒù/…ª½Ø è¼ l,M@vÎo..v¦ 'À[v€ªŒ<à“Èî/cù¿ _€•‰õßáþöþ#¥ÝŸÎ@{[ ‡¥9ÀÌÒø$%Ïv3€v¦mœíßü®@K ñ›ÁŸ¥R¢Êà[‡÷çlâdévfr¶´ù£Gæ?¼]³¤©¸½­-ÈìŒðG}–N “·{÷`þ{¸Övönv^ÿBf–v¦f´aêâÀ¬ngéè’‘øÛæM„ð™9 àdaaáæâ€ w æ?¨y8€þT²þ!~ëÁÇËÁÞ`öÖÈÇÒ ôö‡àå tÀN. ¯*þ!°²L-MÀc¹¥Â¢¿‰Afá·ù;YºtYÞèÇ `ùã÷ï“þÃLííl<þcþ爙?Š©ˆkˆÓÿÝò¿•bböî/FN.#' €••‹Àývðùï8J@Ë¿ëø‡¯Œ™=€÷¯rßîé_%»þÍš¿„ðß±íߘ Ðü‡èz,œ,&oÖÿgºÿéòÿÇò?¢ü_‰þ¿I¹ØØü©§ùËàÿ£ÚZÚxümñÆ\ðÛ(Ø¿í‚Ýÿšj‚þZ]1{ÓÿÕÉ€o» jgnóïk´t–²t™*Y‚M,þ¢Ë_rõ?ÍÆÒ¤dïlùÇÓ`deaùÝÛv™X¿=ÎoœüSz[žÿN)igboúÇ–±½Mèäô@xòâx±¾­£)ÈýO˜™ììÁo.€·æ|föNL”›í8ˆþBìf™ÿ .³ü¿Ï›Nõ߈—À üâ0›üqò¼!{›·–þ%ù£3fÓ@V3è?ð­3Ëhß2™ÿr˜-þ9Ìÿ4~+Äúð-·Í?à[a¶ÿo{ÁüªÞ6‘Ùþð­‡CŽ·îÞjÿºYßr;ÿ¾™€ÿ„ÿ5'§·—éÏ­yØ¿ðŸÏ ä2AXš·7Ც 긫%tcÜ›œ¥ÜÓL¡eôZrúæò€›D[ñyÃé·hÒp/ÚêŽ$ÍÈ2ɳ×IklH[‚rû£÷“aœÊô^;ÂâÎÀdÁ‰h}?üF5‘}ïgGoëw­]²”9Ž.<(Jy˜wn}Òîõýe+c_æ÷”÷«¹äŸÊf£Ô#õŠPægÎᑽ3ÁÑa\¸£þ¸ù=‹‘=ùJ"GàsÅ^襳É}?ç¹V¡ÆæÜO¯ƒGôîclšÊKìð«,î‚WIQä†àBwXÌÐrP_üœœ>™¯žSP)£`º¿|&Ew¶›Iº“†É¤"€Šš¤Î+˜²DÚ,¡Š|¸Ä0óƒ¥ßý39‘j¼¤)ý·âîÓ3õ.:ïø<>Ö¨ƒ [âß‚˜Û™¾\O÷ï™ËÓÒò0IfëäËŠŸpfÆUs}$U†£ ¼tû_s™Šn­ÞY…xmõÖùúÝyp•&•Î>vÅZ ë6Ö`Ý`N¡vò†îLÙ^Øš¸/ƒdØpÅP~–gò ɦl°ÌaÙ:P‡ÒœR~0Õ:%ý0Ý/Œ2ûœÙù³LxEù ã”ã'H%âc¢þl8SªƒŸáÔøጀ]ž³,>û#¾fu—‰]ÛTæ!`ãìê¾ßý:ìÆO7½jÉ؅(%…,¦¹ûÀ¬JLúÒUÀF×xJøþǸa$ŸVl[À/õrØ– ™¼ka]·r(‚í1x*£büéÒìÎõ$gÓ ©¢djœ:çýC䙩5SÞbÀÍ‚´ÏÁ˺ÑcÀ8Нæu¬íOf¨°z‡I°,GðÜ?‰*“è4iOò¸Vè䣤¶nëlíäív£N[Þ¯Ój?åaô§¼ÊµŠ^8ú™¤“Ü0›*vW¼•îØÄŽÍWÜ®Œ­»¥j²+ŠÍ¬ø£M…ÉcmÈè¹NˉzÍ¡öc¼îoÐéî×Ï~&ãO#b2¨*µç=VPQ—ØUâëÕé+Gò¢l¯µörÙàÛ±ó~±Õ︸d8a^.)ÔÓ¦à"Ð¥P— 'h‰TÈýtJƒ¼‹†ž&é:ÍÚߨObã…¹n*Éšc8>:Æ:Í¡lôÛàHìûCÞö\­ÞÇßcmLÃߥ£ñ /Øp˜&”ݨ¾4X¤Ë²® Î(¨>)ƒ¼ŽónûS›^;9\_0ª&œ'ЂRá﨔¿;|> £ OU ªãn@oEÿ ÖÙx„Ös"òóæLR(ná$«ÿL×YÚEtº|,P9‘$YÏÐ.efEŸw:Òù­kò÷Iöp&™;Kï­ù»ò ]‘eŸ­Ã%²ù@„aS™8I:x€U×¾þ~¿+Sؽ»þ<ÞZ²GÛ&¯ºÈÇ4„‰$‹xP?x#›ëW$HR*·ƒÓ(leW-Cæ¥Bˆ´*Ñä}GÕ‘úýÆZèÐ2³£wé  Ó½&òĸեW˜°èËtmü€!õ\£Õ9Ó³;éiOÔÙ¯Yšx)o@ÀĦtà|²‰ØSUüc³é»Zù¨.è̯càþn=¨Uý)­œ!óïF¾YÇëkÓÎC;-ÝÀðPŸ· mlrÌÎi7‚싆íð£îk¢EwBhÛnš‰®Ï–ž¯†*Ì{§ã›.~Œ¡Î,!޾Ş ã3o²ûÚW¼-‰WAz þ¤ë>6î^e×+îHÞH®fÙÆ„vmy§¹À|i¹G(,g´<œ•Û‡¶ôLøêsdÎ?ÑW‡Z‘ Rð-RhþcµEÏÉãýy銭ïkêcÈí^Î#v©.Z<dÅÏ~äïH`+nD8ö~…5mÏ¢ }†-a‹f‘ý,%1K¡«Ý©“ÂóÓ×P ©9=Q ëZû”¦*˜M´ h’÷P©ÐççtﻈÂe”]‹y9 <Û_Ã8nmPv<´D™ÇÆúƒKY°¯çãê4æ{7‹÷ë4FŸØ’#ZŸ ðV<ÙaXvGoñÔ4j¼:ªáï ?žÇF„èv‚0mÄ··‘“£‰ž8"/ƒ»Ý¡8­BB]‚ ìbvå„)Æ4j韥xÙÑä7²pš¸µPBn½Ç` ðgÕ*>{T¢ Iô÷‰üÅ6hÜÔÞyo]•´­Z!Ι(Ì ²L§æ$w4‹* ÒkQÕ¡*îÄ;5‹! Z ֚ýÉpÀmP4M’¦©³ºÌà¹þÃl*š‚÷,LZœ©ìžv™ J¹´‡?j3‡) Ú†bÌ·/2›{½p€¾õPBêäù…è¹ àäÓ @c×Àï7Ýž­û è¶âÃ/A½‘Òû—œïbMYÆ<þS­ÓZÈó© ¥5ß"+1‹·1,¢7S"hÎK‘r|w×X¯Ò{bÒ`øýLª f¹ ÷Ì2”º&J¼/ª¬;TxÚ¸¯‰…,C߇6Õ‚]Aõ™Š3yDÚk,03ªuŤB:×ä)تÈè8ÄlùwÐQÔØÙ»o±ËÐnz¯4(:uót{"ŸƒSÌÔùñ9^"¢ê²ð•ß[µ^Å?ÕÇ…¿7o!1OÞ°__E&¥$R'Àêî°+,m_¥ù‘ `»9l]È®ïª Ì+fl5l¦~û„§;?a¨'­  fæoBj$ÛWùPi»Ö[•N[õ»+]2ZÏiH“©gv:Þ>±p Âcg>l‰Vl,üB¯SdÙ<à(rRVØ é踔tÍÃa0?=²Ú5-‘ÝôøÄË´SÐÒ=åU9\œŽQ®CÚ`hÄè´l³CwÞîù9/k÷ÃRÛbx. Z^Ò|ÏÎX>ôA¾­Ï1† qÐéÀŠÓ ÷]™¢ŠF½,êA%7{óW¬]åéÖd ¦ö@’ËxÞï+ñ t(.Ç@KÄŽ€#™&XL?W´äFÿ)?†¸î†Ð oØÍÃÃû#V®×ñu_›®1à‹Â%L D§?ú™óÊBëâ&ªátGßãÄœÔDD_r¸k¼rȰjö :TüCU&6e–‰ÿ;ò‚A߯XÄäðÑOû0Èï@P3aq!ÔÀ%UÖ´_“ À~¯¢ ²³%ÞÑq°Gƒ[,Cbc /¼øEa¼ï¶Ê@¼Qá³Y ýÒP…œu²2'nXƈL`™Õè“ä—Kc5Ú3ÔjÔ²¨/õBnGk®õ[Þ¸LÇ t"³ÎÂ+–w¡“ ÆýÛzNðQ¼(¢åß7uî]éËFÞWË‹ï¶'MЗ­ð@Å`}ýpïÁä.>³4Mõ¹K½$ÉgIOf Ô‰$säaWÖœ)³{‹oEB tK“þ‹½‡éë)g”– ’óÊ÷/X^£*žMlÔ˜ˆí¾e) *ùÏcÞ†s/LÃ9ñ®ÀãMÕ½d~nuÊ1kþw;òHbbFg`ëxÉÞ=ÔF½Ð*ÈãdR©Ç?µÐjB¼W=&Ï]|÷”Ñ-Ê­»\94`g r ±0ä4Û&2y.KýzÙ1£!Ñç´Ÿ ã*.Ñø¤ù Î%0]ÅiZ–c f‚!Ý‹…?”øŽeè¼ã¨œ5ž+¢ ¹ó¯£ #û†f˜0ç|<œ,§'Ž£³=jbäžBÂÉRMc;þBwÙ/™8hïHÌß:×>ɼÝÆu¬[‚$6†fßÒCج†ôk®4W*,±NH°ír‘ç¢)œ"6+Ût/‚Šœ…dÚþ.ì¢Û0«<>ö(©>Hú¯Td=ÖÓMòõꮢz锫­\s! ñ ¦2<òÏz$JwªÂb„ùëJ´3_1‹mÁM©†¼ 26½V}†]¿ÙZ'M ŠÔ]¸6÷ÓRÔò¢Qb­ƒª0e‹¥÷hÄ=ØBîÇXàž[Að>c†×yxÊ®‘àóZÓt™Ç:ÚBp·)³ñ™„¯.mõ=¬9Gq.ûáóš4³ƒb`/%jÑe8a2™2Ïáðéë¾×ÇM!  ¾p²Â?Ÿ…ɸ4yÇŠTg'ZD ¢`ùóïOy›´¯Ïdu€•‹¹\ÌÉÀî5u[û™Ô3¸%¶¨dZÿÌÏ„M>|ôÚÛó›• ©{øûÞ‹þ@´Ò¬ÝCiyTµeî_’+çÛ/ÝŽaš¾â y÷¶ÚøW‚ &Sʤ}¡×"þ Ì®˜æfŠÝKƒ ¨!í &"©ÇÕÄÈ´lJÈDéï¸mXV Tg8êTªe,¨Ç©„‘å)ª•ãæz;-½á©'kÓÐô0ôýcÃ×¹)ç¦âcó¹ü¿Oˆ¥û:kòý úuÈEâñrKëÏgÐ@Ž’­B¨P]*òeðŸ_üåE(¼—$4+S#f¯™•ëßL%±Ó±ÈÝìõ=¯4Óú̾«’=R0'›¡à|à¹4ì4/wfoäfËÌFˆ›hŸ˜¤ÍÙO åXGz@Ye%‡ìµo+ÿÚÚŒ‹-C¢e’©t»èÖÌ È?‚þú¼kPØÏñŒ=ÌE4›HkEô©hˆ è"Ø{cäí®Ò•-õ©ëGødy¤¬†ß„.o—m!?Tè©S,Šiètü”¿ÁÝ`æ,pU10e,õ¨"Ί[³è˜<½~&þý,lâÂHýð`›˜†4ߦå.®î®L±¦¨€HËáNº` ×oájÌ#sôH dµéÒŒÜEgyΛ%ˆ‰£8y’+—pÆhH®ysC×(ù}ûBÉê–<_Ú9Rv­Ë ·oˆòÊ·{^·=)à P"áo“=F¸bÑlÑ”jvÔaLﵺ5PA©Ãu­36¦kÊ£9sây_É2wű*IÀK×j(ѤïYÓU®h¥Ò]Kk€JÆ>˜;1ºû÷*úÜb‰¬É6ß0½*qW`û{@ò¡°xšìtÈYŸ*¹ze[?Ãúêg¯jƒü¹Çéºá‹ÕÝhŠÊB“Àùà×Xëœssynº9lϪÑκ\>vl¯=§êGdr¹lT\]*¤… ¼ýk“~¤|˜_ü馭$hÐßë6¶D—¨è?ðˆxx(«ž¸Œ Ÿ'Njy¾dÙ-±»/\‡YÊ‘FU:»DLÅ›FÔ/ :á‰×j _Ùt&iøišz¬­pÊ¢ÆÖOZ¯ž°œLß!ª¶µ[¨®ÉµîFѶ`Ãøû/<”„%5Ú´o9_Ö¼xAžðÉx5QT^æ~ ÕAæ¥ÄUi¯›¿â-½ ­:›‘òå/`»mÚ¡,ó£¶&ŽDDäàÓ„Í£{¡ã¿'Cs@Ôñ ÑöŨŽÃ÷›@tc¬•¤¨¨¶NÊŒóXsä_a­ªÂ}õYâgøºr<Ò(pœ PU7€šˆˆ±w¸Fkx5#ã€ÏÛ¿4vÄ:†Ôÿð»SsžNãS ±B%xyžAõÂÜÎP¿1‡%8¼ê˜QzDkîγþ^Ú;.×Ì0ÞI_R:ne踒m[l ?§ß¨`—ÿš çJS#QÇÜp½,o\Žîw±`º}¾¤RÉ¿"., ”ý®¡†ºõâap¸RùÚlžoë»™fT§Ãy¸5ê3åµ+*k’¬¬rW€ähgJ³Žº’á‘ú«ù·½Ä¸‘Z™¬x ôêÕÇOàêõà,‘ÒSÝsþB"ÚÂq؈û@†¸HûKd‚‚Ä€nLˤpÉ3hJGíø:†>¸TE7Ïúμ²8ê ;p7ÜEVæˆbPk<QÓ¹±4O·óä‡ÿrêt{ºté]ÅÇ׌¦£+ö©GÝexõ£ñPºZâ#JõO . °|G#Zµ¬1d#B‡Ì™¢ žø|†c#I.a IÈ_<ÞÁ& Ž›:ËÖ,‘ªÁ~„µ«”/xßòe&¸}Ï…¨Ù¶g ¤8é™ %ÂÊ{<Å>óÑ¡ÉýK¨„ÉÞ )d¶ gTvmëöèñn±¢n³K?ÁGžÎÑ~Kváç õzé|ŸcwMC\L, TþàÑ}œÕÙ°´ŸFmdGg·S‡ivAZøvm›0àDÙàÒÔÄs„ŒÊ¾éðõ‰]‘!a “ên U—]ézËù}xžs²¬§E ‰ÿà,þ± Uœ;Q_B¢Ö…e 6i®@ùIWÐÓÿe¤P)—NN\Ý1Þ:"1p‡|KøÀºµ™Vþòç£ôâšAMd2mî‹bÍ—öÕˆÙãùš“°óÔz‘‰ú9µ5EµAákÜÈ‚•³`»l¹»‡o+þ‹Ccž7r›0çy þ+….­ó³¢©@&ytõÔæ‘z¹Âž)dâTcÉ.å“{ïÑÀL½f\§ó¹ßձƑ³éörÝwü&³?Y.Ä5;1ë™ØO†©ò‘CKÜ ŠÈýÝcËÆh.ÚÍ:BªD1—'úhûi)¥¼,Ç…N)6§“¶å‘ľw0i!ix9hê–åÊ€LIO–Šºj‰Ûñ½]iBeÕõú¯ÃÁ˜tª-t-ÇYŸ˜çgð°’XÃaDjWì ÑwáoÈÂ~Æ4¨03Å"&g¹ÙÒÃK>Gž}{Ɔ¯Ãck a³‘¦ü¸;hÈKëéÍx—Ÿ.Ñ% dΧYÈÜÔ{±Qº‡ Ë̾úÙÐ9£ È6|ÆÿÙ‘—ÑiF0Ggâˆòɇp1Ø÷{†>}3PNÉ‚‡ÁÕT‚¬ /5û…;=¬¿Û6qê#ŽÚwd¦/_µêØ _®r^óêN†º §ºOŽz® [™•áZî½Ðþ‹B ÏÐÖ£o‰ˆ*ÃN–è×8©©Í ˜¾i±ÐÇJ*¤­ÑQÅØ!ÀlÌÿ`|Â8³qâ$Gwóí æs ¶Õì‘aK ´,hbO6båêÈÖZ‘/Õ4”Ý,L’†’HPÍÑ™0ãAvߤ"ZjHírÖñ§‰1§~„~k"Þ5dŒ„6R|_‹?‚¿;”/_ŽdIhÕ-tX„O ìÄqœâœ?L|à²iø•õg±È/ÅÜý ­½=ª¹–¾’ø¼þdܫعʔÖi#wsÉ”.DM„Êœ"˜-ž¢{ºf÷óÏOíÓÊ<´ÐbC×|d¾OY?ƒ±¿õ—/âü…£r¿ñ}üB*4ø9ÂI4­j¾zÖ²-G‘xeF)è“*ÄÂÝèv¹2ƒHJ«w¬B¿šF¤Å¥Ôñ÷<3¶S¡¢€½žà&£¤z¡¤’[‹à5ÄÏ—mïí¹zÔ s¬é†jƒd%Ð9‡H2µìPÛ(\ØÑRÓq¤µãÚÎÚIW‰c‚…«ÚŽ˜ø¢‘YŽ+ÏëÕx#«µ¬ ?nG$ÙE†÷»¨¬kcм?çYÏÊëø¼‚†"¸“B?°J/ú|)AJ@Ïé» äX‡j¿.üfûCIFì@£èK2»"ºÑöi³Î°£Vñ0l|6Ä!D<Çv¾—\ºB €:L×GÜb¹¤öQÅÖõèˆ&‡ ³@±vðý¨´ÕøZ7|'¹MÑhÈv¸uASà²Ï]熎9}B-¾\ÝXU91—¹¦D«I/9‹J#íǧ¹ |ᣢ/6Èø!ŸÈ·„¿G;5~súb¾lfW$ʈcèx6‚ã¾™™¹Îilнîè@ÏÙwõÔ§:'ÛfŽ©W®Wâëi–€GŽ¥ë‰všc­ÜëèÃyz±;Ê0Hù]\åÀxUŠúžw­Ïƒ’1Q…¼,1þnÿð¥]èkÂò‘“éýø‡›Úï„?¥›ùÝ.ÛoµbhMÐÑ2îX¾àjü>·*ö—“a\3Q»q—\‘-ë_²¼N'[ðåC>Œsg.Sq ·9ÉŠáÎèhóU>M ÚdJ¥Í¿ÿà@vš‹zÛ¼/Ä ²ÅÇí‚>Úi…J+ö¬þÅ+ûZZ{hI´Åƒ]ÍÜèâÀB¡ðΊt‹­4ƒ—†§îA8I¯›×ýΊ5lf0Ñc ê\®Hý ¼ªÀú&–Á¤OöEn€‘œî«Sxýƒõ{[( Ìü°B±ˆ¨ÒL  °¸êîïÔ\¸sr!!=ƒ\`vO.N´'v!e®`8Œ³ÄmADëŠgtDÔ*Ûñ‹ÙM *ÔDXêW-“÷ßÐ(Îôc©ç?ì%D`&}鬛ñ‰^xdeîÃÎiJïƒ Î½Þ;¦Òeï1ø?ÑñrõVeåD\B77Ú`™OfØdÑH”l˨<]|*Ûõ^eú =ä’Œîq—¹Yáq{TÓôËÒôÁ¼vÝÓIµr6FÅó]Îz®Q4Oû 9?l)âP'xÐ'öÝïΙ»‰hÞÕ<‚4â{túôó’4ÞåäâöˆfÞÓÎdUKÄÂèKãaç{ :~IëŠ&ò·Í¬ú}oè‘αNH—Ã?$h|gãÊ~NTÌ=)Òéðsµ¾Q9Ô7â’FP½t‘¤ËxŒ—o)k<#´ ±aÍ=€ ·(ù(+…a*’ÉT F»1î}?ãÔè„ÙÏ.”¼¬ØÓ;‰„6ÎÑ ´ˆ•üP¾F/ÿèrûY"‡ùKiµO±™Rb«¸£+-Ê iöû×®E¼¬q©‡.° V±tàh–ÿS <µsdr§§ñÅlU$¬E±9ìê u§$c¿> ”Üן k¤A®‡²ü-½¼rÜzs)-ý~̾'Ö!^qNßä~çV³´‘”"!ûÁׯYá \ëµj„—D„Ô3­V?ýd.͵µ•†uÚìiu×Qr9-®èÄê·‘Il.i>Ѳ­È#9UíWB…„5§;§Ÿ ½—¡“%3÷©¬só>¤HR¹Ú fב²ªªhõ¦BFöL”Ì‹¯[éÃxɤ[ÞÒòK¢;òY%{'»jýÜù^}fe!:ß Øî\ytêαß3·—eɧë⬮’üúì•üm>Å–Q¬°üÖ±j¢bEŠlÅÛ¤Ù °ZNÑŠæ7ï=v…ÓÓ—_2[šxdÛ/š„ÐŒj‹"C‰§æÖëïégw#2¦K ×EöøÕïöhJá)å& ®i›–BWÈ›ÁY])“Ñä_ܾã±õ'b ?áiœû®Ô€Š?ûÍ7ÜúF_4áß®ž™ZÙƒPN-J-Y8KÝ ¿9Ûm“I¹ôq>kr“ªIg™-~”zÇB,ÌÕ˜nÛÜœv¯Ð±ëÎFÒ|PU&YGf1F<¦u¬QÆž]û[Q±„´Ë¤ªRTðåå=œÕxýR;¦“ªí5ÉhRÑ‹OŸ‹éŒ%ÁV^_è„MR·[u}ñ®¢+D¯†1w®†Gô–Eƒæ}û~“Á&vÒÁïEIì¿*¤?¹¢Ý ˜Î@!h'çCPÙu`ùä thx~UØIã‚.‘Âõ3gDàÆVþjlçýŽ×þÜÝ=@cM›\Î9ôÈ6?ó÷*×­‘ì¯~Cú/ Ÿ¿²›l褂Þc“TNð†k{s\ÿ¦;Ý3}ÅY—v6 ¾¡Åÿ¶«vPO¾ò‘`‰é ±4Í7&ön–Þg¾§+hÁíy¼RJJ Nq)8ø:Kž±ôª…¢¢¥o8C·’¯Y$Ön?(³wG¾¥û…-nh?_2Èë»Ø+ö‰Ùh]Db¼Ký…˜1õ +¡S'YãuÒÛ$L:¹t>_Í>3K:m)¾Ë_Ͼ%†þ+¤ÈeyiùeåAD|'=ÄÃ1¡)/Æ#¶™HýúÐIm¨6—|LEÍàc¹X#û]±£}Ÿ_ææ ãƒŠqkd$mƒdžšÇ‡\'×>^ïÔÒ”Hï¿’}´½§wiưi‹•^E±"Õ1*[c “în! xÓ~°±Òší¸‚+ГLTç&fî ×ß[áHû°rI“ƒÁ %‹ø"^òä³hgöÓ3™üÆ‚xÇkùIia[-®œtL¸¿Tcæ—ĸ+•Bë]iEKæcÖ#ÄVÀýpœméIàCñ=¹z“FªJ-º­"X¬÷­Ä©ÀTuò0Žë'e'LB‹a>œ;kƒC‡½+„oÕæâ©÷_ww:¼ðÄ€²*ܤÂÉãëVÝË©¾L$œs’¨^{ºDÖgFØû¥(ÒŠt«Z-ø;' · j¦AŸ4Æ ?z4)²¡ÇTÕ*æÔÇÑ&Ãȳ®A:èÆ£ÃÔÆmÜûüÔÌåU8Y=LB „¥Òð˜çÍ»œÿþÉw@&Œ³•§5¯Ï]Üå=Ñí*åùÞûÓ×YŠ ¹#*W‰>A³³ð'h˜Ž™´- l…6PyÕ¡¾¹˜%7IQÖv.4j5ãÖ³•þMiF)6ø”vª{wzçÛšÛOh-¹`«ÝHI1Ö—%=Éßô&> p’Yýpq.³èÒúÀwÐZŒÛæìhƒcø=;öåm&.ÿŠëªÂ¥¨¬í";¿³ÛÿÂŒ}ØÃéc>ðiŒŽ¤?Æöׂæ™OÊGœ_+gµ yàÂdòÅgô‘„ÌØãªÜ¥1в#ë!¼AEB®Ó†tÕ-B{…‘” Gí{›»üD:ûžwñúXðt²¼Gßl7G³kÔ–ÚQ±¿Y¾ðR¸$Éû†-8?ФŽõŸ½H:dà\âž±}íKм¤#Z´µ$ÿ¼rUÒ‘ú‰œ–Gɳ*<„ZÞH ò7ôþË:fÏÆó¤ÿI]G0™ÄAÀ­¡¯ÐõÁÅ˲å.• c Q#ÁŽKSnεîPÞF´4‘`\ꊫ™/°¢ä!ËæÉôµ‚|c•‹–¸ã»qÏ$”å—¤âðGjæÚûIoõ [ˆ[ œ™¨z·!òI…U¾îê'™=(Çœ)Cü¢)³Ù¼L3Ê÷®,tWi>¢(4çcVËk#²—yéM ±ØØ9²¿Éœ'›v|kÜpéEãáãÒr–™1žÝ—Έò/[kã¹/¸x¼3?FÑ|¥!±G¿FTÒö è‘9µ_6~Öö7WQª}U“ÇÄækŽt4‚XÎ?¦„C@¯é«“‰æ¬‡îã Cáž6+!fãö”À—Àa¿%Qzá™l†o6À<†Q”˜«¶ß¾|1x¼êÁ¦QÔ:sv©rägi5m˹úÆ¾Ø ¿æ®_ð~-Néªmˆ…Q»4‚r<߯ÁºZÞÌÔ”L;´ç_¾¾§ï„UÚÚÕ—Ê—É}Ö&¨6Ê25Ñi4BÓŽ-O³"ùŠBWÚ¸Q~À«þœñ1žv1Õ&%gèîs[_cö©ÌÛüs‡"·7M¬Á+èÊÏÕ+d5{t67=™ZŒ±8̘‡•ý24Ú'5ÌðáûbÉ^X_xóøi%^šIhÜß+9×GBr©%0·¡6D^ÎҀ߂zXKÑ3Óðy.• #MC´wÏðù…óÑܨIÃÈÝÂM±¾k_‰'˜÷*¯T»H`à&j“@IšnÓ²’ÌQþ‡,ô­§²ÇŒ‘„—«WãÓ»üû’l8ƒh%Mâô¸ý|®¾»Ðd7Š*y®¼š¢o ~èN' &”j¹oªžä›TÓÔ&¶ÍžBµ Ù#æ?$ªK¸9´Ïs¿#œ\Çxç$ü‰¡ãE¡µé‹ öBWšöƒQÉ!.Ò¯¢<”€Eüü¨û3:ž°zÃPý±@Zô½îA_ÄyÖ2¬ÇÌî…ƒ û’oÁ¿á2Á?v©•ùÓNû }…k$ïpü°¾¶í-ø‰Q“@¶­#²u…Ô, ¦÷§‡:o† ĺa 4˜’w§ž7]Øbé o:ù–îÁܲ V«k%ÄÇÍ¢¢˜TV{û Ì´wÅ«ŒÌcÅ D¡ çÁŽ".|^©nè,ܢ௠þ¢që¿Pî·a§>7;Of÷EiÖÍRì~†–Ýõwáé pÆÃO,,‰x]^E¥£ðÑþdõµ endstream endobj 252 0 obj << /Length1 2075 /Length2 14295 /Length3 0 /Length 15549 /Filter /FlateDecode >> stream xÚöP\kÓ€‹â‚KÜÝ=¸»`pww÷àîîwî®Á <¸öþ$ûûï­:§¦Š™§ííîÕý.ȉUè…Œm Aâ¶6NôÌ L<9aMf+ <9¹ª¹“è?rxruƒ£¹­ Ï?,D@@§w™(ÐéÝPÎÖ íl`f0sð0sò01X˜˜¸ÿchëÀº˜äÒ¶6 Gxr[;wsS3§÷sþó@eD `æææ¤ûÛ d r07Úä€Nf ë÷€V[#s“ûÿ„ â3sr²ãadtuueZ;2Ø:˜ PÓ\ÍÌÊ Gƒ ÈðWÉy 5èߥ1À“TÍÌÿ¥P±5qr:€ï+s#㻋³1Èð~:@EJ `²ù—±ì¿ èÿn€™ù¿áþíýW s›¿FF¶Öv@wsS€‰¹  .ËàäæDÚÿe´r´}÷ºÍ­€†ï§ˆ )€ïþ»>G#s;'GGs«¿jdü+Ì{›ÅlŒEl­­A6NŽðå'jî2zï»;ã¿®¥­«çÈÄÜÆØä¯2ŒíÕlÌíAR¢ÿ¶yÁÿ‘™‚œìLLLœÜ=äfdÆø×ªîv ¿•̉ßkðö´³µ˜¼—ò67½Á{:]@'g·ç?ÿKðÌÌcs#'€!ÈÔÜþOôw1Èä_üþüÌÝÚLïãÇ `úëóß_ºïflkcåþÇüïGÌ(ª$+£¡Nûï’ÿ«¶uxÒ³³èYØ™Ì̬ÜNv&€÷ÿÆQšÿ;¦?¾R6&¶î¥ûÞ§ÿ¤ìòï ú÷‚Pþ7–¼íûä‚T]‡‰ÉèýóÿçqÿÛåÿß”ÿåÿuÐÿoFâÎVVë©þeðÿ£Z›[¹ÿÛâ}rÞ·@Îö}lþ¯©è_«+lkeüuRNÀ÷]²1µúoÍÅÍÝ@ÆŠæNFfÿ—ÉÕþZ4+s¢­£ù_W €ž™‰éÿèÞ·ËÈòýúp|ŸÉ¿U ÷åùß#ÅlŒlÿÚ2vÐÁèÏô>J,ììOæ÷u4¹ý=ÅF[§wÀ{qÞ[ø¿ž(;€Qè/Ñ¿ˆÀ(ü‡8Œ"ÿ%Nf£Äb0Jý¡w?Ù?Ä`”ÿCÜF…ÿ€Qñ½ÇTúC,Få?ô~‚ÊzÏSí½Ÿ§þ_â~Ïø‡ÞO7üCï§ý—þê0£ñ?ðý|Ð?ð=“ÿâûÆ0š˜»üѳÿ¥¶uvø‡Ã»‰é? Àhö_d{OÙÌÝÎìý ýcñ.3ÿ¾gnù|OÝêøž»õd~Ïý¡Þ¯FÛ?‡½Û¾¿hþ¡~OÖîúÝ×îý®·±™8ý‘2ÿ[ú¯%ý¯ø½»vï«iûN½¿êíÿï•þ£Ìïe9þßü ÷¾ôŒNf 4ò='WÛ8¼÷ÁùøÞ—à{m®åÝÛíøÞýoüŸÅ0rvx/Ìéï«ë}kþÿ‹@ 7üÊ¢­oÅ· ŽûZ!|Wú½Iþ9ò=TjzχNçGdØ$êšÌ€M‡[¡¤‘ï¨k»bT7ŸW‰^<[`CÛ”ÚŸ¼žõ¿*ÏìµÃ/Oc N Õ~  Wý¼ïõbï¥îo Ù Þ-MžkïÌ…¬˜ÿñÞµ_­~ üÇxÈâžÒ~ ‡ Âsù,}´Z”Žɽ|ŽÖãM¢3 œºÇ×áKeBÏR ƈQ>8–0RY§”ú‚ÿ ôâ^®Æ[k†².Rc&w?´–V¡µ—¶mPh]œÿH€”Žuˆ&…å¾#Î:Ý|ÿa2‹Ì¡.©ìäz@@«âl|Cp*f‚kGQ Üú0†Ô$ªx=“^ÆÑ†ÿ5ñ»Ø…ßL¬G_“jÐ=.Dþ>–НäÛ˜}‘/C“­ºý#q좦HL­Ø'éÛ˜Š®fTåy]Í< œ¶äO ÕÏ+V4-óãø<_éìF8o“ZdÛž†–µ²‡D|ꇬnn*#ÐÛÓ˜ 'xý¦ÄtCðkÞÑyÈJ“´°Ü¬êž€®$¥h«Åc‰êƦ$N0b^"pw{ÚíýŠK¤Yé†$_¡ch˜û.wŒÓ «™Z„ЍÔlµ&‹´ všúaN>ödíXe-2>³\‘ ×ÔN¢Ë'6E¡‰®f†Ý¾’¬æbÞAÚ(EÖ'wIʘ÷d„sÏÚb¹lžƒûí<”5Žé?´û—õ¥ ÷¬ÍÊA&¿);Zü´í3šKà²E˜œ ÔòïO:)Ê÷~Ð Í|p¥'¡e«'Û(˜m-Lå'­}–¡Ñ€è ¯q‚Ò¡—îLB0$œ,-¸Ò2޽(3 MŽûÛ/à&Z–¿a·ëYœÕ*!ë·¢Pd&p•3g}=wœ[š…×¶*Í{!M«c ?ID ûÄ3úÝ÷cÌ¢µ?7 Œ<"À@t¶W»óphÆáê)„bÄÇ¿Qíâ1Ž¢&R´ 7Káå(´”rb"3H{]ÅLì_]’ bóP8°ëCí«F¢K˜Ý¨c~AŒù =až9j?–j- ©˜t憹Ө<€#{߆›K¨s¸›Á±·É¬¿…¸ð¡zh)áëÉ.0ÉÿR¶O¿%þAÃ9â,«Ð1Éoè;ásö¬¿.ýáE$@Jý²AÑñ¤áÔÛYÖ YnÙ¯¼‹„3[¦kòìî ý÷Ïu.w@¤çïßq…^¹.ÔŸ¾3ˆ†°æ­ì'É%áb~Hëбú~KOž]wìËÅnßšS~ O(ú®ŠÔä7nT[Árò5ÜÿÙ]TG®M…[rùB)ºc!ËÁäZ¢DïY‡=4‰twÅS;øËT5æ–ÌO•"=‚›Ö´*9<Í':.ŠN Ò 1ÙkÊz‚ž£ºÿ²ûT·ow?N ÜU%[‹’Œ'õ ÿX°†h‡±@»-èתV†ÝJžy=å³b¼ì›a¬Wå-©4 G·L4²,‘¢æØ¦‡!rqŽVˆÑì÷›ÿ)59¶ì²Œ‰ Êzµÿ΀J~¾A-®†¾ôÖéà›õÛQ¹¢Õ=JÕÒ K rë›äéâ÷L}"õC÷“s|3-†UƒñÍO¶Æ£2#YÇЬ{UimèJ—I¼çàý(ApùûH9Lš6¤Zyhóó3Y“+ñyÆA„°ñD[ÔÉq}àqîÒŽð/'ÝNì èÊDeÆ4¶ÂãÃc,¹ÇÒd¬°Œ%qWßás;eŒîoß­ʼnOvÃà1KKMÅ5‚ÊqÀ ”IÙJzHüø)#L´òȵ;V ï"UpÜÊôÕ¼X›Òa=ˆØÙÜEÊ\úŒ(¿ ¿5ñû.W9Õ¸QGéÔ1%ü¶†ÇÞ@OØOþª]Í×,Õê¸_«å7gcž~£øv‘ ®šEýþAF°D·}%³Šwe{$“²J"™5Uƒ<’ h©`™F-U›à¤”®£j:o'ʘù]eĬ(µ·}Ç`•«4n1"Nn‘SªØÏ|ÖîaÈ2R]§á·!Gk-ÓÔ:8dBG²ÆÈêSÐ_Ãût›Ì™ÏÈÙ[Å?fK Le“L -ÜDðÙfüCˆéè»Y†²8wmŽê"¢gU>EêÙfpœ/ÏØµY7ÞÝÛM¬Ü›EÏliâ2þ yœ8¶zÐÛj˜ÄΰKsŒ  ÏÖ14ÈNyp¼iåÞw‡;£,*wlï˶ÍpJSÕõÐDÌ1ÆÜV×öb ×/²ÇnRÜ4?ú‚%!á¥É†IP^rlTÎB5zY@¢¢+¾X#wÉÍ™|{Ð*_þKé[‰…&t»¦#䋜Þò/±š†“¬Ùg’CâÚ bp5¡ð¯S£ˆt‚¬è=»nœÉ—¤øÝo@â«þ°]ÒáO/#I#ÎEØ‘†ù-Œ ›Ýõk¤IÅ;2Ãíݲ[þ.ê Á8Æb¼;î¾ ÛÂj¨¿ŽW8«rÙËvq„›v•¥iõEÜ@u⻄nh̦û%Ç•ö©–IüÉ!'Qô-æŽèšŽuIíóêôŸâ@È6—ϳæGM®³î¤/ÆéoßÇÎCPÉG7}2t Aá÷sºT1ç«ð<{¿,m…‚3×Ö<ƒs÷v%r³á>ã̳ª|pîs®tI´ÍªG˜´_K£uä­uKó´žÅÐ žÝèòÒ_cÁHx«}z!ÞmÛ".5Y:Sž»f ¾Ô¤¸ûƒ‰ÿJ Ú°ëk`ù„ÑšÐï-E³Ñ€-,2%à.˜® ?ÝÁ–‚þ̳“Tg(v«®ÄÜ4«~ñ`HH*c+ÎdlÄk;Á}ÏÙ0a“Я´Œ§ò<è›àYƒ‡Kk|ÝÐ¥I7/â%ù/bÖå<›Ü9Ò¢OæûF<ê7:fdª|ÃÎ:$¤:•º›ƒ×]ÎŒ;¼¶’7ügÅ}6E-Ê)Úˆ¤È—ó^þQ:Ìš#j9lf­“ãàl,°yÉ{gÞ*„¯ßuŽ o+®?“fdy#Ÿ1yOdcÿ貄¹@Í’žô#-þ'é)ÇÃŒçë9ÜsÅÁNu“ޏËÚ‰!B·ŸÕ<¹´¶iÃ<êi°‡OºÂ<æI|¼ž¥•È·sw”¶5r‚W"êir›&ñLF0 9Ô4÷*ߌùŸ+-ûÎÓ²U"s©ü`²â‘1Õ ·Àòݧòƒdêk"ópp'6­Œdu°y]ë½põ9°¯« O þ~Õ7Ô¦`毭ð¿Éé!¤ÕsµMjU59çð‚B±"ýáÔ ÂèîÌ8ϹäòÙ¦~|ð8äæÙ¯ß¶xµvp¹ÉP¼Ž!EˆÏ…3ôsûݾ«×^<&Û×ÒŸƒà^RžX{=ÅŽ2£ü™ÑÓäXIäÒ»Âx´nxéÖgyסþˆ¸’­ý%g&î¦=ÊD±À¶ùÔ¯[nŽV’.ùŒCŸñ‡†§ÙòÄÈùcøLú¯q‹ƒzïÙ×SéP„Lj…¦ŸÛ‹ØJ7ê}<2d=í…tÀ”gÍȨ#y<äU¤e×Ý ,ô²û‡â Áã >QϪNûLËhA a>*2AÇÀH+bùP øpïPôm¸ôî轑ÀBwQw†j²àó›AÊÏåQ—¿•‘G3Cï,ºÀ0¯™ºø±qÔsúKÞ~Å:&~½j”¿~éµÈßQ`#,íÄXé2\Û+lwÊÿíþ{¿~ïæ’âÜóÊUI›V+B:>ß¾ž¶uû3Âù®µ ݦ…œ˜‰¢5B©²aTUÃä®ïé÷òÁØ¥oY2¨<øºý…ÂÄåØò‚~Ò˜ç”KB ·–=ž|ŠÃßS÷È ówÉ”üC¶m:ˆ¡hw§çqc`pE­É£8ë«fýJgnB$¢8£ÊëQ”8KyÝ<Ѿ-2•5’ ð €Çc«À^TU|ç¨Vs½Q [zò%œ¾ø}Ì/ÛÇ7H”‹2£MìÆ„†¶û‚v£åéÝw½AÄ„ƒxÊ56DÙ¤õ¡¡wBHnjŒäÓîg¸{(£1_q¬–ëòü6?rT>+$ÿ–N 0 :3ÿpa%Ðô|Pµm´[ÒMËh%œ;ãƒé|G¿Ÿ‹Ð%©ùQèÝ!vúãQp8v” ïwu:c?p2ÞâJ'Uår 9®^Ñß–iÙÇ/žá¢ ¤l÷¢žÀ2 Þf2›"TãÑʉ 1ÄdJ»„=DàܧÛ5}4lpæEcdäçÑg ˜òIw1Ÿö¯‘JohN?Nòå¶´?KF©&ËpʪWHC¾qá$ô[¯!Þ/;_ª»­zÄ=—Ëé0o3÷‹¸á?RÒ ZhÙÛ®5À ÇÁ‰¬=£CæÀ]õú­†{øM›e¨Ìó¾e|oßóZ¿ž â}¸­²:´{¨nR×Ä*wƒã2)dˆ˜œŽ¨+‹âPM.næ¯8“`b¬flÒå\÷4ù2¼FºïóZ.6\qXHÛGy¼Æ¦;86ò–Ø$j‹æ-.}¤‚ÀÄbЬ¯ç/Ó]ªÂk é¾k§»•øo£~ þÜœºå@ ‹ÝÜCùHÄ:ÆcP$Ä)ë£I«¿ŸÚ¥L*ñSH½ßI7Bt髸Ã"“ß¿ó&ôkßص©µøÇ_fN‹¿ÉGÔæ4‘_ÍàšÂýÆSÓ¤0›ãˆ[Vè:¸zܚטd©Ó[ŹGö¨9}˜ËçÎNV¼¼¸›t꫼‘¬<1˜Üâ sHÀ¿  '}z6Ö<‹kĸ‰iÕó«Ë< Þi ·ëæXY)•}äx1·- òç«ÃcO{Ö‚|]ß‘'ÇÚP‚ÑsÄYŒìóÌ «ŠâÉ·ä „á_¦—Ö5Ky%Áçø,¬fԘȔ™[ºbÏÖ§–ðÊôE¾Í»í’uÇŽÉ«á«ÏZ%{•å Pº¯½ºØù|ÄM˜fÔã€>¯ mIÌr«µÖþ£Ú›¡ƒÈŒ˜’óÀ//%íÉt‡Æ3¡:Wݶѩ§5óË=‡­xbß¹À$´Qg?"›4”¸öãÔÔÎED‡=2F× äh—*}’¸‘fLÒ6&/ƒx¬¨ÌøØšm€Vñ4o«¡ÛUÇ1V ñsÐw Fë¤ï£Ýƒø)b[ÌNçX"Ø.ºð¬‘r ´zwËSwfÙ]Ä *¾R[²ª-ÅØ,îôfpaËçkê¡Ât «]x;·ÙïëvÚ×­i¥§×•ÆXnªÝ·(mBTøÆLöÖcf»ÍSàq•Å“±­êäЯ4ŠeÝ­n8iíá×ÚuíC¿ÏCݤâ\æ»ô¸3šø¢Àƒ\g`A´yô%VÍÂyÆ™.Ï~ —ÁƱØmÒ&Áƺ±XÖ¤~[9Z¾²1Á|¬ @¹®2XZ²Í.@Ÿù*«· OC APœ^Ö ™@HïÂ*]I9j[]ß„¹K„y=(¢[<ƒŒ}j]Âfý)­súþ­÷Â|G{µ@²õÙg‘_¢äçÃÍ…$vd‘iòQ¿ËÛðLÖåFC# ì™V—wÎ*tj…F0‡^ÈŠqM•Øú˜DH6åãŽA…Ò‡‰zŸ:%9勈P«¡«ëñý ¹X‹dë³"Íð(Qé°W ÊeÞ&¢Ý­ój¬÷CŽí‰S¼$ÐÌ]ËÛXòÛc ü„4Yú´(˜s10ãj¾.íÈ•GÜ?½ž /Ø7AŸéqB0³ô¿ ŒÊ®¤Ât‹—y«rÀâ;fÝÇÖØc ùðÚæH]3µ$ÌaúaNò–N£Çe©c¶œ|•(HNÉžyùRÝ^,-­€[˱ּßì…˜ ˜ÁÊØè øš‰ÏÂXK ÈŒL®A²8!Ü0<žñì=‹ó½’ï²'ü˜?C©µ6¦v>¤_m u¦ia4q™Ðn¬4„q%êOç[ï2{ÍŸbe*r=$Û«Åñéíè²îuÚ5að[±^·ÎMYl¹l¢|Çv"-Ú^Å=l†«™:纬ø~UáÜ:–8LJKH¼¾$„KXX^_Éq¹²§˜i†Úyè5xÇÖ&ÖB-†LŸÞrƒ °ÕŽÝ[›ånH;Pçßnʦ˜Èµíêš(Í6Qõ¥¥½@Mråáa?›À¤›0û°ö/•Ó[ì£Ë‡ mKLÆhüÄILÇPaZ?gp%ŸXÐÙöÿ²IK:ô×B5úÿZkÚu·D¹½ƒª'¢OX7Á6NDìtwÑ£wÍ…qF†(wõotá¯IÐ69vCLâ$E\:A_úÝH*é®Í3¼ûWøý„_}@“Õi1â!þìÆ¢fìBò;µÏAMdªmIzdï÷e@gQ¥Ùkƒí™ sö€ ÒÜ?ŸOÝý°Y’¬í͈hnÓ4Éë³'…¸­mÅò…pzäBô½ä/…H¥!!"¢;ÒDìpœS$N~ìaHS¨HoÀeÿü²Ø´>ö}/vWGœ¥Q¯$*ŠgØM˜Â´ Ó=«ÿ¶þ›ñA@ùóRáµKB!µ+îäq4³© pꪀK‘œž8‹àAïÙÁ:8«ì^B+àŒá«‹vžgÂù½~hk3ÀÖûm>5`ÿ²"uͼϾ³¨¸ïW<·8AƒÊOE„¬øë<Ï:i÷âTœu³ûéú ¼àQ¸Lu¬ÞQK¯¼–6LZGwÜâz1a¥É_L®aÐ ’ÞÂâ:¡{bD•¸ûO¢Ž:ÅGŠa|.ÎN¾…ݵ¾I%èþ?.p˦söåå6•öqäœB í/;œd«G„˜õnØ©R á*¤xÌZ7£OªzÏý=Þ,å>a'¸nËZ5|&ͳÚ'— °„¹õ6y^UMk‹MªÂG@‡7ãà´%±ÎõüÀkÓvöz,’Ž_0¡ ~Ê |mœ7¢œÆÙ;]p×3ˆÂ&CFfˆI“w0¿zÎÿBØý %´OiªoÈÒ'#içMÊòš³²lE„äu÷´æ*›ÌJʆS ®OïA± éh@ÇʃwÆy±þʨïhõæ2² Ÿ  ‡lâSR‰,JE{äVŽ)7*Œ/ŠQ•²ò–.Ã`Åql+µìgw5^>àÏ{Z)Ò:¢ˆ-¯³ÿ˜>€ÙcL·´°ÌqdÊëÇ|»Z, ¯@Wµ )Tbð–ÈÁG¡õ•¸…¿“¸Éêá\ƒd} O㾜*¨¶–󗇌¯¤´Ú[úé÷¤Eí®£ô¯Œ2üœ}¦kèc¦ø&{µLòÓdgÌxоÛÏ9D¿Ï’ºÖ>§ý.hÂç]¢(u°hÆœŽN1›üæf\±Ê‚ƒNJ^zž#D@ o⇠x•{ý¼žåZŠý]‹À-‹~†Àv5+½À‚ÍO 𳃪NJÀ éõ5äÈÄï°o°¯°öÚÎj¨Rîe›¡‚áð~l\¤7Tä„Òhq˵WP×–ê´)†Ü{%ËK.HU~l— N0ænh /±,²N7Àõ¦²âHe‹ø$©„ÐZŸc0Ywv4EBŽÁløÈmGOhl%›úi̼tçŠjõÁ®GÉ`ìØ¥ çý… $póh¸£´I)$ì¼ÀëM¤ø<…4âi?v›U ‡G« ¹’®}¢öj‹QU-ÞeÇPȪ2`ºÑвä§Å² Ë÷;wí3Þ-Ž!‚Ïñœa‘7‚̓Ñ/&Ì͇snèÎoÖ(ê3B Æõl“úü¹ñv¨”6k…¡àyÈܱEŠÏ {ÄSÁ ôdR 9At_¯çZ¡ÓïGÒ*—+7`‰Š&}3í&¿{xäùá­­tÀ| lô6©‹LAOצû¡Ša׺(¿‡i1%Œb]øWJÅ€¼æ \w„ÛD$^àµN) *ihQcÂüL°Þf „ºú¼ð ÆÅÅcÞ_-ñ»®D1Ò0ZÙ'Å'ypSQxh'ÑŽÛÈÚÏaâ$_XªyÇ?=½=œ¨«wVÄÁé(–\u{‘ b{ƒ9Ã(~°XERˆœ@Ë“zñ†¿ÍÉ‹æHÊ”¨%¤‚žë6Z‘xô½=ž%]({€ÚëOX›7î–å>`¤nÅGòvÕ”9¾r2hRˆBŒ9þ’D>N´«ï)^±ÅÉùSeé¼7È>ݹs¢„š‚’BªÞY¦ÇR,¨1ÊW§£iïâ³н©ºüÉïövhüú+Òö~ ›ï—8fâkè1­ïUP»K šÁƒ'|žB|ro O®ïbÎVAµBrd¨àò5Q×­¸[Áçv†¶*¥–P;©rw•F¤·/L4QL¥€»ð½œ-Šgî“·ªŽ‰|5Þ7jóÕÈÔôîTêêêl³mr.n­©.Ù¢ ®øFx´«Ÿ4Á.“’ͶA=¢;rjÐU8wTI}:â…!a ÒûÕ^Ñz`±¹kM‹ýb] ÿ$1+6ƘrM’E,íü³UÂ.ü+›eÁœd‚Qg²ÐÇÜ÷’úoô‰ÈñVR‹¢Š,çñf2ðqÑ:P{‹ù}ªÎa½¨¹býÜž:ƒªð7Æw¥ÉG™ë¿¢‹ö*ת;GJá½£`ÒèŒr),gµ¡ÏÔ’Ž‘³wx^ƒ’^éÑêCćƈ²Î¯¿A‡‡;ͪ~ÛnÛN¼«û0H­ÉÃd=NfÊ!ÞPB4~ö¨ŒŽ½ÿnÔ¹ôõ¨qg‰Iš=]rr(ÔHÅR›¨£‰†‚ô  Þy2Ö9Jÿ×"…x1ü=coˆ¹ÝKßÏ|[HŒàÌW¬ïîvÍ>jÍ­8VÓ©Ë\&¬Ìðdjì…u½!"±¿Œ__¿p€mÝñ¼*d#Íó3QçÀóÌᙓK«ZZ—Šƒ4³ â ®šž3C…”¥ùÑÍGM‰.£‰o&.ŒÉëˆ$M£º“óÌ…3iŒãÖac…nÍ£xÍf^AÈ¥’ûF')Æ ™„nîQ4nømY±Ã#ÔB2$åîÈ fWŒ=šìè9\Jΰ°.M?o_~Ÿ¡]œF žH~™° ÂP1&ʼ‘(:œ‹/2n]÷Æì”k“„ýÿ@ kç1¿{sÞ:¾Å¯ú˜}Òýù.6‡c¡¿!Š?^ØxÊ –@2…ë¾Ó6ùz^,dj¸ˆ .6aÇÇ×Ùן´›OªS@÷¹BR‹ßß‘Lå V)÷#ÞíÙN”xº¯½ÖÖ›óKBÔëñöeaÕüÜéÏ"…ÇîVU ÓWƒ Μ«´m:GB4v7'F/sÒ@pËÝ«À…Î’pU§ÇeæóФãA4hóÅÀÔ^ÕQ‚-ÒndÈ¡q§¨MúDïµ’l8"Bå›k{¥í©LG²v/¼K3¤³5?/œžÆôƒ²xˆâ‘›qÚºFþ"Shñ]…x“€ìWdÿ!Œ#’0\.ÏÏÀ©‰Þ‚ +´D…ê£Ùoe“$Ý„ßÉÍÛ-~*㲋¨¢4¯(½~zEÞÉaѤcÚP%ç‚›öÌ?!Oþr"“ÌóqøAWàÇaƒPÛ8ÃôíF³Ÿ6ÍÏ!“UD+ú½ë§üPˆ¨øÁógˆ¼Ä*þ°ÏœRöþ$%:—”?­• .úh÷Ϫº•0 É›a%?¸ Ī1/—v1è¡Là@«~ÿbãªÍ£Ô’LiW¸.ߨ!‘;ØIÄZU“YÞçTÆ3¥ ûñ8´Ü‚ÐG€ MOèþêè—ˆƒgåÛ‘ >0Å»†ó÷)û0;`1XI¡î8ÌâÆÁµÃŸçÝÄ¥‡W~³5ÿöC¿\àO•zв ø¥7ÍF3æ›VqÖÔ°”cÄô}¡‚3£ŽVèX¡°a$©vü^ˆ(®‘A=íqƒ´ý]ßx«{ûÏ}7P¥CfÇ)1;´£xgÓµûØøXo `ðùµ-ÔXÆ“Ÿú4¬àG FZЂ­òÊÑ“©‡ëð¨åL„Q6OXå Qýèû¹\åǵîv½ŒS…×¢þ-mì)±Y‡ï—~QŠŠ™ñÀ#Õèù2R"m¢ãÉ+ö#,8Ϭ™ñöôéoÁbý âçaKxÃ2ÔòŒOSÇ@TDwùÇ)B¸ó-*çOg?hˆvý ¼$ÙYA±›ÁB7¼Ô>—kƒƒê<F¬£¿}ñ¦nçQ°×ŠdfìT,c>¾JüÿØ›‚òu¼wwËËÑѽ,0XÏN!׌<£ï˪QEJùÚ Ç5~D ›YF#±¢™kGÂh*¥JP^¼~óé¨d¢릇~ŠÄ”D¹6ßùÏ’Kò[FèÜxõ‡o¾4àô€sE1?rNxާã!1ûÃÒô‰Ã¦$Š0ŠG'_Ù]ü©ŠJ‰Û«‰“¾•Ê-–jû`Y"´µÚX ·Ô~H`W ~» ³i øeué•©‡‹äñ‰­&w­nrÈÓÎè­kþ¸Œr…ågóyð@­åÞ  qÁ±œ]£ÐaIˆ:‘ôhÜ$úº–m/Áàœ{­æèþ4Æï+—|Ž.à’‹ä3ß[7æ1¯û"[WXR:®ðÎÍb׺HÌkÃWFìjÓ|œ(c–P³8 _ÛkÉá Qç´$|+Pvt\¦:§¤Øëb×áxçaçBD*IH§£ù¼©Þ¦×ì‚ÎÞ¸4µi¢´ ­@^”\bQ,ñC›å„Í]\½¾òݘôÔз[þ´f©:¹×L¯ÜmÎtèç¬âÆ’A6Dη*à1íÉpÈò_¤2×ûd‰¡ô£ÅàgÔÔÊv%/Ðè~ª‡ë¸±îQ¶7–B!~£13–o£è—±pL ÚØ1ÈûpÆ;Édö‡üÔÑÌìQ›¾ qMâ5 €Æ,Êä}°:óP´´šá%–N™>G,v),ªÝM°>¢aÚ\w4Â甥ÊLòŽæ¯LjZç¿xyƒwé#Vö‘¸/-kËX8o@JU*»F¢9Éó¢Ÿ+óf ‹žp¾É_Iç5%á= nBèDÝÊgz,ðö45ˆÀ*[áh_%øV¨µšýz„Àä@8¹¹¼&hÁv%‹¦ys¼p)@pa­<8J‚TØýÜOâ=ÀdK9^ŸÞbøéΣ; ùßê>!žƒ#ŸNÐÓ‰mÁÓ,näæ¶‘kÆÌüÆc(ƒv ÔKuo²èušX·èÓŠ"œÅK?âø2Pg̶LnMY« fVº8“ÈÚ€¯ªdÏW™a…¼òTÅyVXÏÑ–ì9TS¸|¾¼çØ”´*ûá)¦Ôi·qo€ÄLjÿªÕÞ¿z-ìÒØ}UANOf­ËèBC32zØý.3ïÅ}U´_¼Ù­=¡SÐTD2¢ÿ„°w£/Ç+.þøæÇYÁwá£:›eׯ¥âÌlZ*ù페ñšÑî >.*@ÍÙPƒbbLáÝ̦¼Áb儃d|k! 9ºú‡'~²äô¼S§#Ïg»Æ)13<ó-_ki¢ö»w'YQm)ŸrÁ¥GWcjV†O&Rnä¾h™ø±õ^ˆ7ó“ÿyÉjÀWÛ¹VSrøNø©=ôˆð°ß^g\iÁMH:’\*R²£»bèuR'*Ëa  z'ßäË}½ÄñÇá²b±,g·™ðÅOöÃÈGKÕùt›Ø$,„¸ £ƒ5é>áÈ`,+]¹Þ‚I¦ò¨½ . :i%Á'Ê?`EgÙg¯èu=mú.:K…Ù76¼Bå!XÆo2v ¤¸`xa6;Û²ÓóW‡n¶Ä s²¢IeFÁ™ cŽßÌjWnwLœµ@†À ~ úÝáµçç¯Ï69R¹iß©HfÎšÝÆoèðàa)Qùl°™ªG{Ní†Àå¸]âYÊø½ 1î±€šX7¤>“¹ÑÕhìþ§B¾õ$Š<'ÎJž÷¾<߯*CC–4x}ºW 5¡¥—šÊã¯Î×½ó³JvjsK-¹î‚µÜ9¢¹°çÁûš»iIú í[­÷*QÁ#(\:—ª W„¶¬Â+_ë~+GJ+f2úË<¸À1 t‘ÈK—¢Gf”XV––Ç<-Q–ÕÏÓžtÒ¬.ÝšÃr¥×Ž=cJ/£ŠŸ¼a(v Edlv9–G+cxË:a>×6­4ÌP]ßmã€D+ü°¡Ÿ;Ç;E(¿µ>ÆÝ3'»Èà—xÄ¢ÁŠ7¬v»›Âû¦oyi+ÄÍé²^ ¤Žñ¶Ç«ÏXf[ÉÈ1YCí> ‘> è0†½Ú9Y Ã6oZ¥žrBºr  w÷¾™;Vãž)!õ˜ÚF9<.Ò»d§J²~>àI˜wwA~§ûP[ÉZË&è~ýØeAjsŽEÒ!žÐ¼ß¥~@_¡~œúíÌS÷`æeø4͸4¢Âž3ZIõ¤Ôºý„hAðœÑ9¹wv{†Vt´'ο“É[Bƒ“¿âªíÌ<ß‹õaD\K8Q’âàKDt–´#8rHÖæá©“§/0Wþ¬m*󼎮·}\5ÎòTÒ-ð43ÍQ±)îòuß²Óíé·—¾®ëcŒý:Áî.¬pÖ¯ ¡0¸9håÑÔIE§’Ï—ü;…0ê#¿E~¡­ÕùO„&["RÍÌuFÉXþ¶@‡ ·¶éÆê=7nWì²VE3ï6¦ 2c­¾¶1€—²Ê-¢d ¡(…âM½+ÒQqgP’1ØQ;½K{C'cÆ€ÙÈ&?ïùþxª›%4þÃf©Î’Yc©($h’”L35Údo©)k u-ÒD¾zÞà$AlÌPÎeqHeÚÀiÈWö€’É^b›j'Ì‚u··p ùþ·Õ(qf´"ž ’´óv“H0'g÷·EQtQlÂñÃŒèp¶©Y?¤dÔJÕ©æ&³ëì¦Tu÷NpËHÇLevªóU°ßøk¡.mFïñCܘçXÀ¦ØAþD;c_U¤F½ÜeÁ„xʘÝJAª”•äC[ZË0ñ€e×ʦ0°?¾ì½ÌÅx¾åY~ü-€<™qú9âl°öˆ§ôàËGJ×H±&?óÆ$ЕÌ/$+#vxe½W~®‚þÐ-#ÙsÖò³×@ÇÏ¿ VÃ:óoáPP'å4÷އ‹ ¬já³Ñ"#»$å>YÉ~ás)þY³‹ßjxnð]g«“ŠûG¡e€æá4Oy$Ë~Õ€P§Ïsëñ ܶs¾ÙÑ•IǘñÓ”xÐBš•TŒäûÅÖÝQûö…ÊR«Md<·Ê€û…P'ÙC€õ±län¨ä̇ò’³Ÿ“|yë7&%[ 6æTã9åR¥aÕ»éÉDû­š#7ªÓÝ–äÍ#¦_Wá~öôxªÀ¨ýŽZ¡~Û´#úzu'U*K>Ïp•…0ÇÙG(î©#(hö³Ba[†ý¡©Í‰^ÜÈ…îQu Øgq›!"Y$ g9³ÖîZô〆¦Ò‹oà¤ËA>úasÛk§À §1 #Åy^‹ý”t3ÉÄÖmðAnS¤ýE,$3 ÉR) ºÒ„x¥Ìr!•s{ü¸_¶é`åų¿Y\èAÿw·qü'xÃmþX(|öŠbê5&Ý*Þý |:Æ GÞùÝ:ãOÅ48ÃWøÖ°v½<`&S¼9iÁæ`r}Òàƒû—D<¶ÌÀ¶¬,’xÝŽÂ(ê6¡špßV)í±ÎÅ+•°œt†ka:’ÙeìDo»E† ÔÛn¶c^e©°›ù¹ÚgïpÉéŒÕjôãt|´edáŽæ§(çÎ^ýg q‘ŽlýÅæc>ŸL˜«o²À§×e#P{w½zŽÓO©OE=íî «XHŒ©æíz‚# ²ÛÆÕD0s#{BpŠƒE±|ªÚ0Ççãž·Û“¤Ì´%œ7—‚ÂÎRìÛ“wÀ󈈥ý!,ÙKPB’Iï!bÐË íÃCy+œô.ãÕWV¬l’¢7KPøFfEÀD_ˆÄÓï—?Mp)anÚLò|\|ïzI㫲?ÑÉ)ˆy…Ë„f—V?+4<V‹îã9ÉÝ O¦cñéò•VkÖòuÎi‡ƒ}¸í “HËSÄj›+{íÃU¦"Œà¦§‚É“ÎRÛ–aM±eŽÆ’'þ!a]S§)mzàWˆª£ZWD]¼Ô|<¼æuç(Ïüyd~ >H›ÛY~hLŽÁ’Í™µVx™þbµ¹¨€f“ªflY*¿Õ¢ƒža-ÚÆª5jŸ Lªæ¶‘+FŒŸîæ{¢ÏîÜ3ÔâÔõ˜ýdd`aÔ) Ç­{1€öeJÔÁÇßJe·6;Œ°ÿóÀË þq?u']l'€FH*ɧçNZ4aKõ5ä[feÉHJ TÞOÖAz¢$(–Þ•‹µU¦n?éþ•È’f¡K[©/æ·éV(Mß©²ýèÇ`9" «\0ãy!MzÝøpˆÁñ·MeýߊÝ)†„¦i)H‘Är‰œD¹ìähÈÆžõ÷¥iy¯ùºÓìêmt„n«»ÆQý‚Xꮙͳ4"â£&¼ØÃŒ‘aåØ0›NiÊ“bSå †X„ÌÁ @JHˆÏ\ih’Öÿ-CU5 endstream endobj 254 0 obj << /Length1 1510 /Length2 6478 /Length3 0 /Length 7479 /Filter /FlateDecode >> stream xÚt4lß¾Y{ïZAÕ±÷ž­½g B$DìÚ{kÍVmJm5ZJmŠRÔlQjÔ5J)_:žç}ž÷ÿ?çûNÎIr_×oß×ïæd30P# 8J((,PÕU· „…E……E99M (ä/œÓ ‚ô‚"à2ÿ°PEB@(4¦B¡ upÀo( JÈ%e„…"ÂÂÒ"25 ÐÜAÀ!^„œª$ÔÙ…Îó×_·#(--ÉÿË ìABAp€.åqGgtÁÆG(åÿ¯Ür.(”‡Œ¯¯¯ ÈÝKtVàáøBQ.#ˆé~¶ йCþ´&HÈ 0qzý&ŒN(_@0¨#î…vñ†ƒ!H:;ÀX[ ïÿ6ÖùmÀø3Pøw¸?Þ?A῜AŽŽwÜ w8Aa€¾†Ž ÊÅÁÁ? A0/Úä‚Â@hƒ_¥ƒʆºÃ?ýy9"¡(/A/(ìgB?àǬ«"ÜÝ!p”áÏúÔ Hˆ#zîþB.× Žð…þur‚ÂÁN?Û{{™Â¡žÞmµ?6hˆð?˜3–‘”@<?G¡Ÿ Lü= ¿HàOÝCP Âà„nu‚ ½@> é ü'ñï!CQˆ3NøŸèhâôûŒ¾$Ô`-Œ– üóó÷?[´ÂÀ8Ìÿ?æ¿®XHÓRCÓʈïOË“**?@ €ˆ@@DZBHJŠ‚þÈýSÈ?œµáN€ôïzуú«fŸ?"àþ³!<€ÇÒC ¥ pÿGé6ÂâÂŽè/àÿYï¿\þ2ÿåUúW¤á ƒýâ¹ü?<È óÿc–®7 ½ºô2ÀÿÛÔò{wu!`¨·û³Ú(z”áΰ¿ õÒ€úAÀP”£ËoÅüÆMî ‡ ¼ ?_€PXø¿8ô‚9º¡_/´,Qôþü;¥:Üþ¹h"â ò'D_3ú$¢7 ñû%d€ B»ÐíœHŸw D?cBHøçðU Î?É?¸Ä߸ê°Ôß0Êã°äß0Z(tο()´ƒ—·»û¯ õò€üÿC ÿƒD{ýNô¯½‘Hô¦ÿ!ºû¿Î¿žÄâH87p”t}ÙvV§Ìè+°6*ÿžsÍ<›G pÙîý/ƒ§öIø'ä©rÆ`7ùüŠ:÷‰ÒÖÛ-x1­i†¯.î_Ú=0_{E8;FÛÿ®d[¹¡™€IÀDiýþÏûfanØ-˜w8 <½¥H ЍÎ|{5ýú*>¾ž^3\¯•¸KtY1!dšhV6ÉYè;EÏ~%ÀŒÏKyàG6yrúž2ÿÝ5ë|„A;I¢¥VK"ÉçS U&"^ ·¬è™±O(ߎßTÙȼC7X^ºX<É–ìäq­©h}j†ró~Â%N«Ol~±¬QTÉØQA§¼5CàhʆõÍßf<©RÕßÅcÂŽ¾mciÈáSEâ>h»¡..Ùªü©˜/\Ôº¯4ê±d¢h_.­Ìmo½B[•U‚; ¢’Ñ+Ÿ‹nnÊϱX’w0´I5ô…pݬxT±žº…¥T>â•,µðöñ}ÜK’X6}æ|{:¡ÊôršO­Ú\¹´oéÓdË.óêÀ”A®Ÿl1§-2—B©d ï:9Ÿ¢P/£f)tw-Øéû±Èm©Ùó@Ÿt Ý㢳³Ï/Øs‚«¾à>Üò}ìSõy`2æŠØ´}Õ»­Aì~ÇÂí&N2õÔfŸ8F-/)v‘¾ßzØá[¹Ç½»/K3á¾,gþCßbžœ œÖ#Ôò1 òÆGZß™e¥4lǬg\Ú4ã»éß;(‹ü_„Õ÷ñ,Ú{kS(gŒã9#v8ñNà™k)œ—ô<:µ5"´±/Êìvèm³àfü“ß¿Q1E´»ZOÒh=Ÿl0‹Ò}èê U:¢2Œ:±RY{Ì}¶].uz"ÐÖdm@ YÂÓz&cG¸Ÿ´`4þzkˆÕUYÔFãÆôñ°vÀ7­¨guþ·²Šn!q‰ç…êý\G¤DsÈñN>î-+D(Ü\o'“þ¦7þÄGàÃWŸâ%žK“b5^æWÚ:AùÍžLuŸË?:øÍ(|;eö& jO­¹½)3º-³bÙy4Žyñ¢1f«D=-¡ú±l×:Eù^æ»§$±ÖYBdkêF#²o6“dÓIÙû%¸ûž5ÜeTTeQuváy=E¬ jÙ'†>c]è” W¢[?ˆÎå¦b•û€›”ÃxŸ/š'hÊ 8†Zäö³z0K¦§»ãå'ŽmR—çÍ=k§í Xãæ¼9âHVøê—·÷KYcü}’Ä„k.º·m}ôc¬(ßÅ éqhïÕpaÚHÏû^ôDÔ`{¤]á¶ÚoµC«wp¸WE†õp0œJ‹¾Ö&*· =°V|r>t”ò^Rï#OŸê„§ä^¥º6÷½ù‡ºÖdŸí/·MœÊ­_ow0,lÌ!ÉU§Ej:,?«ö• à–öU±Tc;­ý ;÷†W›ì9›¨8ÏÝsjÔ§.1¯"Š"-zwÁùKÔû`c6 ¢Àxz®×•:“qFYõ”õ›g9pÖXpxºˆ/¥MøM\ˆÜò·2Ûæ·>y5ÓîxVßöJ ‰sjY/äÛÍÍ\,Én]V³‰„|œoÛàÚÖCJ0Ö§8$¸•ØEÐx6¾‚¿YwE›O×ùìæÈò6W§R,°ŸfY´)ô@÷îBj¨Ýé…Ý\¯&hÁ¤CÛs(~Dت\%pµ¤<½Ö~CäÎifͪÂ@Òè=~ÜèÝ*ÉÄeœì™”TüñYº.rÙ7u;Õ|Öb©*G¼LøžbÁ²’ jqì(sE,ão¯nψf”/Zë¬ížWžÜã L¼4à1{Û\¨ÂŸ¶²e6¹HšR¡†ÈÊ+½—\„¯• ¼Ï2Å2ϯNVÐô=œoègí‘© Éh«$7QôE”Í]‰Gé—žÝ=†?¨` 2t5Oí=w'ý÷¡ùGvúůq$7©?ëÞeë¯ãöÚ¢w’RÆ¥å»hµ­àÜXÖ;¢u£”ÖȪ— Ͷ&+E¤ôáÓ#™ÐwU··R¸Õ"„…_”8¹©2ãQ[‡å×dÎû1$.pÉûEw™‡h}}F &ô"-‹§¤ùQåsž”TÜ _Òj¶UÍ™¼­³Dÿþ;çà †À)ÿ…l, †KvšY õüUÏ~.™5 |Ú+ ’·l+ ¶&˜ù5û‰#ê¶]¦jß­¼•\m[µÙìþÊ ®¢”È£û§jš1Æ7z oM„z`2VLEvàIÈ]9« Æ;`WÛìpžÊR/Ñ"&q„ø¤Žj–æ-sÛñnZ<á«y7¶UW}:õ:îu—Œ]¥…n×üia87eY0êF°_bïÆ±c9_…½–ãƒUjö'|cØlëÎ+ÉØÌ0rzYN·ÑŽðÄÇ>;ìá/9?ê>ëÆQ~(1­õ†³5Ã6‰^í⹽캂öv[Ãzqÿšs3Û“BUcèÆlO6ÂilÀ™.Fï-ƒ-ðÏd6=X(š9°åãÒvÉâîp¥-Å9U*-ç<ò™ý\eõ£Ö•HOå¹mD)õ ÿK艹hÏx`lâ›[ìT²Œ›3–—¶•ÉölPàaàåóz«È:8Ùý#ŒªH3c~ŸZ‰ù¾Ï‚ ÷[÷W8©çÆ'>ïò˜„¨v|Éœtc/~ ÊJ½‚x®Äkd:ß<-»*ÇÈ!Ö-†oŸè§ÖЋ‰QŸF}d Ô”à‰ÉŠ®ylëûe1ï$GU¹"/ƒËÕ¶Pþââ®X,î1óžºr|Iý¨Ì°®þ$¨‚ÏÌbÐzå©ÿ³;hŽ·%žþሄt–U|íPÉ«¶Øˆûµö,Cï©iP>‡E¾³£Hžþ1fÖ8–i¡Ul}Ÿzqì>Š.%ÿg áÒÛVޱkÌ™ñ7çG˜~ÁÑûlÃŃrV'ˆdA”ÀŠSëÍÃæ¥à¼¡î¢:Ή*£Øû6Ä*л¼zœÁṲ̂•v·æòœrL«H¬ð[Œ '£Ô´ÞïÏ?Æß(Ìëq°&¿l9FŽÁÌ·CiÛZÍ…pȆÖ]ä5(j%éѽ\Ø]ÇvaÌï×Κç€A÷7üÒèæðo]#©>É5¢r⡘ ž‚ÝwÔ7Dòå~4d>„ ||:—¢½5Nòli†&1qÍöóúº.~l£o2uZ…³§a–$™ZÆ!{4P¦7nÆw2ÈdhR[®ÿˆlýªÙ€°‹hn…a ´^ím&°`/§ˆ4ù0DÓô]Üñ²ž Ò¢›>³[;T룃cBô”2£Ã†b54•Á_csþhËÌsa`åø9X<ˆmã‘ mkKƒ^¾úÊä)"3G©Û£îGÇ«ËÃ1IŒ«tê1“‰ÕÒ©S´õ†Â+É,”AüBEìéõæóc0‡6â’|J'#©Ü!)³DE]YS»®5¬¾"W[Ä\”3sQ/òÆÞ» Þ’ûgm®Ó‰Ãdó¡|ÜþNKT{.VjÄà°–ÌVáöÝõWÄÞŹ`þ^Êxœ÷”b±4õïœd†®®w¶Ý;H?™xÈÏj4唺Ó|¥Þm“Ss]Ŭc÷yfš5,+³¶omxhô~Mz–É ü:úÜ·3cÈö Ô—jšá³$РýG=8[ü¡i¸S±J2¼¸Nå~ÖåVæþä`¸o›²·÷…@“>5[ÉÖëíÁðòN`Úy5þÊ9Šæk„®UtôC7Uá7>~:+[©=E-Õ¾Ž0;õ›;³—/GC5±8Ïg&ß×wu7SnU?–·µ”¤ä}’<Œ0¼°™οDöTœ:c.u¥Y:ßÃoò¾2xh|Í”Œõd)Þá“hdå¦J«„œøT hä¢ïŠ{E‹unó-{DÅßÑ—4;€6—´È@ijzÿ^kF1µo“²‘ Ã4?˜Úcó­z8ÙÆêSÁ*·í´š%G1g¹éÍWt«U¶ZÏ­¶m:°ThX*£Ä]Âs‘ÍÑd êÅ‘4à{T}¾ÝñXsЦî€"Y¼bzvž-B[{»zÚgVz<)L›ÍÑcßì'õµsÛøR¤Éi\GMuÀ|±MiNÞémÃ¥½ÁõN,=^š‡º£c<[F–"˽ 5þ"à8yœ—£ ¯LÉ2q !»Û³—_[0¤¢•xåx›V×ÙNRRwüóm \’éô¦Ç*ÇTo=C{öב–TÎ{Ïët“> %RnuÛY ÎŽ®¾z½ôæRþäá ¡â¾io‰§à 0ÿj¡Ù&ÞýöÚŠü!IÇ!m%)„äÉUÌ5~xúÉá÷'&-gEƒ»ûv1QÝ®÷Öǹ¬á%p}«býÕ«IEÓäôвÔ×¢²HÐr9NYPªIŽÀÒ˜Z>â9“þª!WÂP½Îè©R6ñ³YA6S¿±¯Ád#WySùÑR5Ïõú»O?=¼ç-½y3â™3ª&'ó ÐT›@±¬ãÐ.(Úvxé¸z…ËùÁ϶r&N§¦º«›ø€®} ´ÛœmäOòÑêƒÌTRìAòëÓ{açð™rR¶a!ä¥iâS¬©§{b<}.ú/êâé‚{GÈ[Ö†\MoÅN…,.â^xêè…Ù÷p”+·«šN ÈGËÍaÉØ*zdñÈ|©Üñ…:Ë/ÁHé—„>Å-G—¹¼L¶ï‘¬Ž„Ô–8_£ôÉÛ?~w,¡<™ò${-Ž›Ÿ•iÓü=“»ÕŠ6åÇÓ{Ÿç¶±¹uÊš‰€â¢s®>1Ç.ñLS›Ò›/Òkâ׳%§©¬NT-Oo`®%5„¶¼ÕïZªág{ñð‡)u”ê-Þ.ÖšaÝœ˜bIúÑǯ9øJüÓ2%ØÇŒŒ(d8-t8;,óhŠRÕâÀ•ÕüÖ‹7.«jjb4ŠL¶ JÂb‡-ÃÇê;ÚeJÒ16{‘Õ×m.œ>Ò'\(Z¼SêÒÆRþŒÛß5®=ú&K?NòK™'¬mg5Ÿnk´¼ÓÈËNë0,Q7oSS0.î•Ç!hrC Hn$¹_U/sqk!ˆ/É»Ymž’—sŠºXcêÐñÄÆV½ÜÇ ï~¢ŽâyI²~MÉät@#).Þ/c„[*.¸|NŽèù.å‹G ˜{óÀ!”ýIç]ü{Pñ¨é.Õûßl™ï|YÅ¢+ô@¹öÅ÷yˆ~CØ—ý “(^MÁX«h'9k`nŸŠY”kÓ¹ÒÝÏZbò*ð†©i6‰‡j‘[LÜî÷zù€Þ÷ ïáYVáÃîáUYÆ­U/îê$•¹×„>Š3þNÊ«‚zïqŒSÒEA™ËÃãÌØ$×·!+k^,^µê%ýk¢ ·¼® ¶r|ÒyKÝlñÅO1Ô.$Ö“«üÖE—?r¾«U?*o‚š]w\»ö¿/¯m)K7(Ž`Ñó³~|J´?ö ~Ba0óÈÌaŽ7æSš£,µØ!QH{™«³{ˆÒÜs7¸ÌõåÌmOáp~u£2*ñjLCf®sû@쇌7•ŠÙ7%“þZ†HiÀc8[‘½Ç+g_P¤æ¼äUú‚<‰ÃfÀ–×äXI{I¦¥ReRs¶p| ô¥ÊžÄ­¡¬É29¶øäÝæ•¥ùtÙ¥zã`ÇÜQ}åÕ‹OÕšú‡#ÙAf8e­òn> c •C´p1yZŽûå.?‚Ï3ľݰ“¯X{KÀ4@ÇAtxùþüp)†¨\{'¥ã°qÂS‰:¶÷}àùé9Ô&¿Õ.T¡ x¿³»`ñìμ{ð¡(v²tÓÉùK¾aÌT;šþ$c¬7zû¾B^`Ds|…±Mh·<ûÝ$dÍÛäñ»¤nJ'oÄ/?®&=rfh,~P-‰=šÅÕ[Íç²4ŒÊ¶õ•q× o_$W‘uªè=Îmø|ãiÚ1Ž`úW¤}ƒr‚ ˜ùÚ½â~õå³#Ûyî› ´~—8®ºvRÜ·ðâ²A6-N"̆Ä×fµîu-Nq’QrÈýçá‡I"ë“1;'лdï7Dúª³8¿ÿ()3f8yÍ1,¶Š3Я¯É„"Š¢©_’yn~JÛò½ ©H¬2ƒ Z›úHû†wðRÂêrŸ‡«cyåÖù±û˜vBê¯>dÔ󘼴bk=:¬‹O€¾—œ™×ž¶y “;ìB߯ð®9{ªO§ú[Ó{ë¹bu•x”¼¡MüK-¯G@<’Ôæ.F7xovð6ñÅìn8×EÃéò¹&¥t£õ»GT†£³wíNj[‰E†ˆÈUx=É[a‚>_' üaàE·Ð ¿´ËUJö~å£[î\– YúV­ÄÛFä°É9>&`°÷Úš.ÙöÈIËÚæéŠ ¾‚Ç4$)~Œ¡,´BõBß²•ŽH÷£÷ゔɉº‚ñ{«úÏ`ÙœLÛÕ’ù¹9Tþ nŠÎò:ûÞ‰ºŒëWÄ ¡Žû½¾`dЕ…çfݰA(Œ€ãsâÓÔÕ™½VËJ˜—ˆ(š=ó;´÷ícó„š…ý3jðĽ"û™ùš´¬~‘æ÷M׈Ôfª•®ä7ˆkúÒÌoöÎ;…Uêt~ ]I>É49M;9Çï·üYGó‹LE~úf¥ÁÆ7;Hög±½õµz$åo.Å]/–—ªn9—ÖŠõ°²Ä¹ðubuvø³–ã÷œTfô®¨ÅÚ¥«2Z† §æx鞊qí5ãH3è%Û$ãý:v#ÉßlÅb Gš½øC®S&@ÍòýÙˆïûÄhÑDFAþñÿíD…Å endstream endobj 256 0 obj << /Length1 1786 /Length2 11545 /Length3 0 /Length 12689 /Filter /FlateDecode >> stream xÚ·Pœk.Lp·Á4H€ÁÝÝ î>8 2¸w'8w÷ànàÁ=Á!¸¸äœ³›Ýýÿª{kªf¾§ûi{»û­o¨É•UEÌÀ& I°=„‘… È SPa€@6& ™šZÍ b ú—™Zääl¶çý†˜Èò"7†¼ÀöY[ €…“—…‹°<ÿ"‚xâÆ®Vf&€,ØäŒL-vðp²²°„¼Äù×#€Ö”ÀÂÃÃõî/s€ˆÈÉÊÔØ ` ±Ù½D45¶¨‚M­@ÿrAËo 8ð23»¹¹1Û93,éÞܬ –3ÈÉdø]2@ÑØôOiLÈÔ5K+翪`sˆ›±ð"°µ2Ù;¿˜¸Ø›œ/Ѫ2ò%ýßdù¿ ïÿ€…‰åßîþ±þíÈÊþ/ccSS°ƒ±½‡•½ÀÜÊP’”g‚¸CÞŒíÍ~mÁ/öÆ®ÆV¶Æ&/„¿R7Hм¿TøO}ΦNVg&g+Ûß52ÿvórÌöfb`;;=Äùw~âVN Ó—s÷`þ§¹6ö`7{¯!s+{3óße˜¹80«Û[9º€dÄÿἈÿÈ,@›‹›r€ÜM-™Pópý¥dù-~©ÁÇËì0)äcezùAör6v N. ¯ÿTü7Bfa˜Y™B& +{ä?Þ_Ä ó¿ñKÿ¬ÜºÀ—ñcþý¤ÿ2af`{[?ô¿ZÌ,«¬"¢¦ÀðOÉÿVŠŠ‚Ý^Œl¬FV €ÈÎ àzyðùo?ÊÆVÿäñ¶2öæ`Àog¿ó}9¨åìúÏÐþ³!t€ÿv¦~]€öϤë9€¦/_,ÿÏóþ—ÉÿߘÿöòôÿÍHÒÅÖö/=íß„ÿÞØÎÊÖãÆËèº@^Ö@ü² öÿKÕý½» 3+»ÿÕÊ@Œ_ÖAÄÞâe¤YØ™€ìË­œ%­ÜAfÊVSË¿Çæo¹úï…³µ²)ƒ­~_1/V@àÿè^¶ÌÔæåq~™Í¿UÆÎ/+ù«‘¿1èe©þ; {S°Ùïícåà;9{ ¿4ÿq¼X^ÖÔ äþ×t˜™ìÁÀKÍ>s°òïFsò˜%~‹þB\,f©?ˆ À¬ðq˜ÿ ;¥#n €Yùzñòþb0«üAlfÕ?ˆÀ¬öq˜5þ —xZÿFAl\ä_çŸÅñžU‡?«™h˜_[jÊ‹åöÙTC† †“ÖÐIâ6læï¸Z¦xªoje¼yÍ~ñ,pJòÅ)í³«Çrµ¥}´Üb­·Þå7íp·×„ ‡ùlïèû#\9|˜Ê yŠÞÔäPp·7`¡ƒ´÷Ó}â`8KY}ä'†5xŸ£–Ç™‹9»}ÇÂç6¢ aƒO¶'‚n kZü[ÒÚÑ5ƒ°î®Ø¯}üw$©2(2LpÈi=mÄß«˜@øê‡Ð÷­ Zú³lPv ®Ù+gàÙïÒ.×ëWž­Eè3Œ,¦Cß—ÕÖ!Ñýà ¨y¦N$æ`²÷E˜ï¶O Š´O#©K‰œ÷+ø.î®Hh?Ú *â8µ½Íð²*kÚ\ Qsí+ôvIN‰Ëpõ&>Ú†9¡Š¦šã udºä®qîR@-㺼¶†N™É»V¶î8ÐeÜOPb֦οŽ\k [>®³éöÓP¼úÇšëŒm€B¨í½š4ñ°ýô“GF?H}pðù—s°ìJf8˜lÂ[Ÿ P­Ÿ¶óȳ5`U±Š&™0IjNE¸1„‰™šohJæÕÁ¤¤gUQ§¿ú7f9hÛsNwÇ\$ä:7ŸÚÒÒÌõôÔ÷Ë&É6‰Õ¿´k=Àî_›¬«N>ºçשv¤Âèy4ÒÙ¿jËÖØû ÛD@zÏ@6Ldcœ«‹ç¹2nžZëýqÁ\†~Uל@Zm†Žºýgif¢cYU“-\—£yiö¡l²4±Óûë¸F’÷×ù~jø&ò>Ì‚6ØGø7ïÝò”ž¸ð8~…Qñ&/¼ËS7Ä“WŠo}£(Võ]UázÎêtmI0kpò&lu÷'ý‚5¥f_–ø÷^$ÕùŒ÷y6y³dÄúèxDËáþ’|"Ÿ¡²ÙûYª¦¶”>íaÚoøxv*Ž{¯oòeuñ|Íç1axÞ¥”sÖo½1d¹ÏŸñu |>,%V}'Ä+ˆˆr1/2;°îaKûa[þUë$äs.Ì©zöÉÑR7A¹ð4=ÞCœOܲD+t¡¦’ãRÊ|OFô¾Ãw|tcixR D=Ï‚Ç^LæéÒ k 8¯,–ð –ñ@µÚd ùÇiM9oOÿé‹ígÞKM„kQ³ý` *ÏE¶¥nµyYßöïºuOslF®¹OöNê¨ še×L?^k3BöÒR÷k*ëX=³æ~ÆB…"ó«rBXmûW>›¤€Êiƒ’9Ä›"1¤ËÒI»˜5aJ «BÐ ÕM9޾ˆm G~p`¹Ë‹éIhò{P,"ùª‚<Î4tÿÎ*N›òaÜ>·ni>sÙj%ºlrª4Ê­o±=:pëžÉi ±†ôBB·”¦â±ùè }/°$À³#e7Ö© ã`cšªP´ Eš9"îo‹(Þ`~mÆ_Š({z‚Å…¨ïl€œp™ìÞ–H[‹Çž³^L»&Z÷ȈeÓþ¹5{0Ŧpóòá³i•]äªðšaÈÛï Ý:m3ÝØˆï¨T@Hv³âº0•ÑÝѾcgUgá(ðZ+‰"Üv"ž‹W?2aÍ:ù(vÒ|ñá§J:ÖÛ¯?ⶈI¡û'*~Í©‰é_÷æÈ!œpbv‹¬ßÌítZó2£@¿Ò©õ~R/¢¦¿ç$ÒûvÑ$ú}3q)žô`W9ŠÊƺҿ{pß­Ì`U²fUU [6›ÝükС^80væ¼§Ý–e쓇¬× ‘;@8½#McFßÜ|Eò²õo£¦ÁTðª&ê<˜|ôîeòúhZj{ ŒÉ£-Æ[òCLe8ùBÜB¼¨Óòä@ ù‰i\ŒÑ@%Q– )mE·þf­€ ®²®²k•aK¤v·!KãõçÙÿa8ÁøÙÇÆW~JøÇqE/Ú˜”@¿A¹å·:›êŒV+$n“™ø[agƒ7·îRÏLñ!ˆ‡øJ7!é:‘'ÖÕÛ’8j+í_ÑÞQŸª,×Ü;›'5ä|Ó³ó7µa×2|÷C ÿ¾ëß#qjrùN¬Êñ¢&äQëõ¡L~Æàl*¬S‡ÞrSç­ØÈ5a³ÍÉìƒÐÑälUu¦ùœØ.|mé|«&ܵCÈì… |^àİH.&í¢¾ÂÁ&ÎõJœA@åÔœRr­íêäWâËO‰¨Oª§c¦ÇkÀ1Շؑ3hËb'g”àŽ0#8Ù/‰O6‘í­ãR–³4{B¸Á€ü×UDFÉYÛÂ<‘ä©`=aùDcµú*”# Ϯ귄ô=7ê|!ŒS„»Á‰àÏj&[«];ÕÍ]¿:X“:ËäÜÙú 7F !Å$•ûq_½§4¨Ã€ùãß%â…«Û’˜NY 6œê yshI¤+@{KÊã«õ³|Õ7߆>ùØifî3þlTØ™ðt6r6ñ‰ËŽZ|ìsnÆ1Ltng†þ˜½-Ö„¡” Ñ[þÑö$rÍROÊÈäÕµ1)wAz3íÚù°:ÛëjÇô†Ù«Óí 18Šòg׮Ȗ+ë',¨h¡Ôœû•1’™æïL%μúCD|Ò¬UÁuB¶] u¦8Ø;OPÊ…7—ä ÄüåFKÌ7fBæ¡xþäÄ•G›æ]¯uøý©$pÆDîë Ø/ƒ0ÛG…ÜRÔXáí#‡ƒÍ ?yg’ïí{©LUý‰8È˲ۀÉ[KÂÁä)M¾;âH3”Ü|¨~ŸÎž/œaIr j,ÄÎòâV…îö<ø¢ õÛ|_þsK'V‡3C~¡ô]ìqŠ•rIYÄ´¦Þñb܉ ðP]!ü:ħʠo挿º¶t¸"D"Uðí¡ûOk0•˜ùÞH=%R7'  Ö„Yvš £Ý­j"4ÊQãÇ[ªË“*1ç~X(¾róÐ_Ÿ×Ë\qfu´;yÊfúV(×tÍçuu¥ÖÊþ3¤;§7¶X1ïts¾Óæp£ßlýGÆRB`(°8¬P?Aß  –ÐíƒÖv"óáá8ƒu• Š­06íé`@©æ#ò\‚©Ë£G«ºJ€ÍLÜËIãh8Qù—€® ïv-åÙ\ŽÚ}ϵ©;/׎˜>E}ÎøÏtSL(ñ´ÚqÂC>ˆÍìcFÖ; uqqœdÉÝiêíêqá+¥ÚªçóZjÚ•‡ ñei V¸TF‡¹Sßšp“íK6r‰‹÷ûIiÅùùê­kúG›ï×evïXqHòšÖD,B¤¿(ªE˜²Îþj[pÂf YÝæš‹1Í݌ܫÁXuŠ›2J9ÇÉ´u¬¤eÕ_íJSeÉí%ëIބЧXRn~¢D§ž£ãàb)ûª>“ È­> IÜ}—#ÊÈ ©g¸ÏwÒ£bE¢Ò]ˆbvŸw`ë„:ÜÒVÿTá×Hm”ë(½A©LTçÇÑ;8£ß¡I^ÛÚ›Û¿g¸çS‰²ÏNgõ˜ ©Æ Zlò!ý¼'·ÿMNC-²œ›p3*;ÙA‚·Óm4+{ OÂ%1§¯¢O@æj?¼XvÈèl$I‹NÅýt-a?¦ß‰cŠiɲ„±M\šà g8 mÚèÛÓŠ'[ˆºÒrPÈë‹78lé!·Úð$Éǯýœó5ø—oŸ÷yyó:Ò,ÖëK5f§®Ç"²êS‘~“ƒö¡ü†‡Û j[ÏK`º±¯óDqe1Ev9½¦ÕóïB”?Uì§o’)6m(Ws.„Usï½eQèž/¥õ`‰àq|Þ%M»ª ‰óÓù¥œêd<œ$>K¨ØZÛ—ú:„Õ“Õœ®Fµ'£p½é •ÎÇά@â]ÓÜE™#`ÄI8 A{=Ám3Jj•·Ïor—œ™Óú y;?Œ”O-VjÙÏÏ““KÕÉ&¤Ž*Üæjç‹ð6bcQj¤±‚XoÊÊÑ:÷Cê¥Iõi *D>”{ªøq4Ì{ræ S‰”nÝñë.¦d2)E):_À½®°m.r í[‘á÷çDÌG†ÕÓJ<»!">ãøŸôP6= 0Þ ¸Fõ”«Ã}Ží^ÿÄ©„j-º´Í-,ôcWîs4T<±nÌ¥‡xÒb‚f‘!e‚I¤ñsŠ·–ŒÑFeŸ•±±·ñ s#“l–šÛ°ˆ0ÍwáŽc¼­™ÌððÁÐÝ)`æëXÈ!Š;JòÅÃ8=WÓ©ô¤áí{þK|Wg%ÿÖ@µWõË6sÍ´œïS¤cjÍl4¸v=T¦ù‡H7ÝQ.3]Ã`IMâøÔh·žxš¦Í^VO2ÒYAÿ´Ó˜_§­OOÙÖ´ˆ,}õä4!À\‹ÑEÃ"©æ#¡ÿ³íÓz±*51B‹Gt¾¶XA’ÉÝvKQ4’ˆ®\>òM€¸2>)_‚s‡ÄVÓ)”žýéVaÔP_-,Í÷mªk5ɃÐ{dëà¯Ù›I~=(©ÞŒ5¦ò˜°A— j°X9'ž¾ÃXXЙÏDò÷„!\³…•R :Ýc9¢Qµë]-ˆô4ãj\åܾ~¡q”©Óƒ]¶É½Üd¸Mdu-–²ÍE=¹Äy´ÖÑÏ&»ôxÞ(¢w¦Ó‰@ì móiMÃg~67¨Š!–餛Odôd±’çKú–ëï°ü<ÁÃ…+åõf0˜~‰£4}8¾ÁÜúÉ·2 Vç¶—TÄP¡>p9‚ÔÒ ðŠùM Q uÍkÓ55M%£âé3v&%w*÷÷X?cnµZr+Æ,¸å¤p O€õäóàKÏ"ÍÀ³ñ1+Þ!xÍv†sØrõí]¿æÕ`Ý~”:;†ªUtZ~àtZ:8—O¤áÕÔŽèoO‹l[”4Ê› F…‡BË!d  ?°Íé¨*òrÊÇ{k»_´Ž¨.´F·3«bB=õ°ŽjöOÝŒàDÛîò¥KERJ@./ÒrãøÂL1¶ém“VB " bÆ_“-ï”BiÙH$ºêÉ&*\)-¦Ôö©ûÏw_1ªäÒ ÒOÜ#M— ³r„Y¶ï¶yË0q¸ô½Wο™ÿ@gô³ôãû÷CMû´¿nØ9ð“ë Šׂ>à(¬WƒÃH9¤è„¾¨H*” môù?Í[lÇ8t›/­CÛþ-Íòú'ɾUÞÖ^L¥‚}*ÊáDrqÞ¨H,¨`j:hcüí ÷ÄÇá! Cù÷BĽfØô÷¼¡óÀSV}.wÚ£6v”*eîvzb&XW’lç1þ«˜z}ù÷!ÀSúÛ&¹ãä†faAã¸l&vïIŒ Š6åÓ ¿í(æéà”ž/ .•k8E¼0mÉ"dñÉ#àî wÏøõ²·}N" KʧÒÒ|Ã;oK×HqŽ: \eåÛóÝÎÖp¦WRy'ˆª²“šcÉãe„Šh"°:[SNÁw[Ûä#àã¼£Šœ@­GÉé>ÆÛsº\X*môÍZo«X€Jé ,dz–’ÅÁ1{cnOú ÖIÐgÁ ‚þìÒL‡ Sšä¨W2²öÌyrž¸SÏ&Èã lóo?ÿòÒU¡±£7Ë—¥ò-?îLõ}ù±e/Ói¸‰9šžhÞnIÑpBë2ÿRgW¸Œ[â·†ÄìDXßó>½O žâ3LÊH}ç¶US6r›ý³KÄÝ|šÁî·`Ú%èâaõSã¸îÉX¢Š½6)ò¾q³¿Ëë 1 D%Ò¿ÑÏ^•h:ÕþTî1éÍA?¤¶ªùÞUcîÂÇEõØÔP%y7Ë`–›œÔ5\}]à‚Oüâ~Bi€€$2µ ¥_·ØesêÕ»ˆ¡m»hH¤*Tzˆí<ûc¾ èD®O!P¿!±jrû›˜,ƒùM_päˆIQæRÏ£EÖX¨ ´ ExH¼³õÃ'‘²€c·ÉJä£ã€‡Weâ ݹ´jw²}@è^¼>=6éËèíÃ?ˆð‹T„{µT‡¨Ÿó pŠ­–‡=`Iò“$õ<Çò5:;¦~¹v ½ú*žÙ\*‰®"bÕ²ì(eÆ/Ó¥BõÙ†ßOøÂðx¾ß£ƒÞc˜çêå”Gwò:6ñÜâ#3å_áˆ='À3Sn°ÎNû=RùUkr0á…Þ©~Jöf÷â}·‹XKÄs ÞuQ8’ÛK eÀj56°õ‘–Hˆ½ø sn-Iéƒè–d((=M¿ÊFÞk¯!õy ¶€uÈP¡ùþ€ì]§›Ò~Të¡ ÿc½Ô~1j‚²‚]$O@–lÝ!ÂÕ©”I =s,_bBÏ·@zÕ½ÒhôâÂUr ågE!~ð]­þ‘‰_ü3&QÒUÄÔ&ý¶¥ôáùœü)w*›"<."‚¤F1½´íæ§*vm›\ŸÍŠu¶ò/}Z<¿]*PD™ÞIëበµÞ¿×Z-ë ó'|ï«’óUMfxí 4«Ütš›G.ÞŽ•0BwÅ (Ž#î} èï›÷ªÈÛè‹ÊÓ²I­s_Ïâ;t™½ÝVλ¾ÖpØý¢ë)›!‘NDˆÏòÎ?pÆ¿‘Rçž„iÉ…M”¹q_©ã±}k¬“ÁX¸?åYgÈBDW!¾ðPã 1‡³Lyó å!•Wbi±-O|oŸÖù¿†ý” !†…šÂ˪˜`ÕïA&,,“”ŠG².Þê"¡7Èï륜ت˹DžÅ©„<ÜlŒ¿Â龺 ªP[ÄÞã¢ëIÒÊTazhuÆ0ïÓ1û‰Pc×ýëM0‹ÐCŸ'Í …§ÊAZ–ã]ð·PÛž®Â$ùEX2ðXl½tÃðåŒù¼aVµŽÓ’KšñnÔ°ÿOP×½»«dµ-6OP ÎBU®'AæU©§–c¨]‰Qâ@M=2ª|”åÄ‘°¹ûe0çWç¾F,ÜÎ<*Î’ø” )€¨Q?X²_|Dz{~°ç{Ò@{F®4Ø«˜W˜¦Vó£æÉ9ŒãSᑺݷµŸ—t×Ý.2.UþÔÇ2ºÙ{üÐÖ‚¨ŸM°™ÝÄBl ïÎÚº8)(’žº ‡Ûí½<&ß°ç–ÜáB/'¯Êß…®ŠòÁ6¡:¨¯Ü´ª;•½=Éþƒ­ø£Ú1ìsªàêRÃÓ³¬cË ^@q¬2LÐó%5aégqB­!ñ”4´rB7Ãp zóÓ¸I­ÑÍB¬õ.?æ›ãWí´€ù‡ÝQøsê]ªtÉtŠW1‰âQ;4d_ÚÞVÝÁÈˈØÀµë(Ø9|#˜~ú*`¥S¡Åʶœf¸ÕRÅ !q' àò¹4óÓN;²=î¬éª8žJe씈ͲñIc4Çâkœn]PÌY )N¾Í kàÛã%ïÿYrõÑt!þ"§"z¦ðèÈds£!ô„¿öB(X¦'‹µÀ_z±©ÖÔ̻ڎӿ“‡ývsyÁч’Igß%ù™×C&›­ÆÅB¹2Æ1‚™?{DÑ¡ÙCK£ MüЩ2/CÝ+€ .ê+õòÈWœ’ȯµ"ï~Мq†Ê”Ç^¼ß—,ü|ׄÏÛñª™ÔlwA`Ô© †ñð™ÄäÞ5«z8ßÏwFÔI¯R™°ùx>¬–w¥wé¤;&iŠß¸+ÔIí­µÎŠ’{ ²xϤN,Ç5uÅë ªÒ•­9-È…ðæ‚ÿè$8ícƒ_%bÉ»kJ9eŒq ¯ÝÔÃG.*òoϬ*øööÊçYt+ô¿´äPkÝu½8Y?o…j.±h¦°òšè”l>/½x÷3ª_“"²­1zpªâ&çˆT‰¹ZŒ2J;¸>)‹N; ¥ÝÜÝä-¢…R–×E†ù–y£~o8Z—G¾¢ ëLºé÷þ%‚®ÚE¹5­Á¬¨毪vØ0Ñ_Ÿ²í2í°üVºj¼!_4ƒEµÉ¯›-öšhÌÑ|ްÂd—¨·}ò[§rpŠ«GKJ‚²XйbÞžóÍ_ñùhsΨX8×3z$Ï1Ù¶ jbŕβ´Â¤b*©ªá7äÓåŒêÑ›ÆSÿnø<ªÅf¦ã«eDRêÑB·@lRš^ü@ÒXRbºÜá*QÓ_€TyzfŽ-vÇðHÛ`ØÀ¯A÷˜Þ6à*'ÔXP÷ÃeŒÛ¥>Þ›wf¿®)~=a†Ä7_ü ž/±‘ŒÀ&™-Å4_5“e{B>^ÊJ©†‹ÅŒ§ØéhÍÎaæ~]ùäÐ#H9ÑIOiFž·æçkÍÝT‘ [±#lõj1­1¼âÌ®P»V{G¤Dr7 .{ï ³"95¶tq¡s-}nöD¢.$$œ·×,Raž•D´ü@CHxu7x&ã‘6/€Ã+ IœoBZKëêRRJžbdloWÏ·ô(DE8;~þŽlþžëþdÔM=²1Î×Ó°­êìéôÙILöb›}«Ü>­ÈÇoçÓäglóÜãv»*K¿5ä§7ljÑrÑêpKî=’9Øjyºí z:©è¾Ù¥²?vü"ÍçòëUôcñûiûúúËò¤n@ªöžÙ‚ A(Áö‚ž¢¯¥ÜÊPÍ\|_ØBTnÄŒÒ~Ê%Ié”aóÂÛêZÄ'K…Bì ³^(‹{h$ïæy@E‡Ì…(öð¶UqÃãdËtdáSÓotŽ¥ç-¢VîWéÎ="CÍ-÷©–öªÜÛÕà3Âî_°ùÒË3q+—±•XÚ[´Ñ0ÄuH. Ð"™fXĬn6Sˆ¨QòÊ;ŽL.cõ¹¯Oµ»íS>¿œ\³ßR_<+á_S瀭„ôҒǓ޲˜Œ%Ö÷̽Ýr-»œÇÓ=ÙŒ‘qg@5gÿµ8ñ«’ãËs²`–OÅ騣Ê($/=Tø‰éŒÏ˜šXšx-Ƥñ¨»S×á ¾Ö¿+KIŸ£™gÄnì&î&â¡ýœ~ɧ‘yN‰@>‰`ÝD÷ÊÖ)¡'0É/.Që*®AŠØËiOú9§ø\çµÉñe¼õÉÅž‰“ò;‡½:NÀ,·zÛÚˆË>Fi—s½ þ ù²§°á"C‘ ]Ɇ=­ Çœ ”|)ÌjßYbè1q8–'ký-¦éFXRG¤¼WtÊUÔGÐ2KðEE를l3êþ¾ …¹–'—Œ2$°RAïêÉ5jåø#ì…3äÍ‘ß^”wÓQ2–ï!š8Ãé"cÐõ "T·,ÚõŒxÐ-´¿ äÕ’;†õŒTÁº—J}Úü@È+@Z£{ª.Õ³$sÏ‚jà «ì!$Åa”Ç)9Ð,­'ù™AwˆšIùñúd ãH`Š* ) 勦‚%6ú1Ãk#ú~v|ÍáO»[ä¶Ï£›cè´lß•V‡Uô¨. =#£ë°ÊIÝgÛ?óEÞ'º|Waå˜#è«yK|(”Ï˺¬bÒOò%Ö-ÂÞ$'¬W2´B³Yùœüy{×Ùi}ò1ýM ñ‘¡AÞ˜Zt}a¦½8$ÖSåø¦Ðæ:…µZ嘿ÇÀC§ªdeÖ¤¯–¤B„xš¸¦9­S¬‚ ç#‹RS´Çy….x*#”5³OÙ€‡>5ÓŽÓ¹±h³„lýƒáëÍžû0s»„©*U¼›©Êú­cM"çý‡ÈsKz[tøº_ÐJògBÏ*wpõêŽp¨ÒA¥-îJ­Kõ—Ò{k¸ÄXç BëÆˆ:ø³·8——Ä;{Ê*æäŸëV,Ö–I6Ê"eÇBب¯Z±¨ØZnAoôr7§‘Ë™V˜ ìTÊî7¯¯¨ÅŸ¸]ž°Ïè¦~zÙ+¸)‹ëd}çÿ Dl]_“f_ÈÒZùK VmV†ˆîl#z)ýõg‡Û}çwLµß3¢‹·ñ*ÑcWHß@ÒtÑg®»"¿Gi®GœÃJ½…vj·ô.¨¹xIÎÉW |€“àjxåR·C>ì|à¿yÑ+(Úx§¯s+Ú–EHD@XV©Óñ> Exm¹Æp0 þE>YOÄÇv”&Ýäna±*àiµk+ÖbEËê=u¿ËsÔ Œ)m>žõÚŠ³ªªáçµÇ|•œa/ÒlÏ)`z”Û'¾rÅ : ¢¯ñù6^†ëã329à*I€ê]ó<5}Î Ì´Àlw è.Wa0§ó*h}GÄKR"r®ÿÖÛ}qkût½®R¯¤Eö1i«V­g>[¨¼ÐPc¢ ´"| ‚Â];.ºï!í[¸(3fÂX-òÎh€JšzÚÁQEXùÑÜæRpë,œá~ërÊB"½(RlBÔ›bs˜EÞ7K¾(N!×±¿ÎÍQ“ÑR™ã±T!£9ððFk~pþ<þ¡ ëF Üö–B boDàp$eôDLL‚‰mNW9ÎËgÃêK)“Ò’@ùëð±QsUóIt Ô·ÌÐÃ'˜^‰Údïdô ¾ª‘zÑd†CØ@ûµ³d8âKRÉ.€S ‹GÕNЬÀþUÚ¢Dz¯XÛ†1;Ç@¬® •š„dkQãœXrLA¸§ ü³øZÑœ&`TµÜ°œ0›Û¿9 ¥. þTíàìµ}ŸPe©êÈCºÔiHD¿¿ŸgýHŽÄàöœµ5ìcZbB*‡u/‰îÊÜguúœ»¹ ’Ö”’#Wd )òºü=`©ñú[Å"^<Ü-nº£÷ÔÖC~öé)UHl®7áÓu/JêÀa$ê#ÜJµ] kO¤”Ô’œc°œæÕ†)w¢pSÞb=pšÓ!Ú<Ÿü•ïmù ê@]øcÖ RfÛWCº‹™¨ccÒµ„\·E!‚ct#Ü…XAÙsv4¾‡7[-áߎ‘]K|B~Ö=#ÞuߟŽodèÒ®7zD¹bÖ«†ò‰îøŸË͚ŔH¦ÍGHtv6Ì_c %SU y.·ëßy4Çù~£³·ôÂåõ^pAêEABXÛ]ŸÎìE¡;]È4¦?fy‡-æ‹ÇlCECÏô‹1hµƒ‡ Ôwš/$mVÞÜn …Äý>±Šf'µ~À<¤ôƒÅsŸ§þBËsñwƒW+¸)Ô®þËšçš%ùLJ¾I×e¼<³7{bóÆÞ±-}¡Ù<üê>òÜÕ–ø4˜Z³˜!·¼”½ò?œ kG†2 õŽ*ª:OA"@Œ†I,·® „@R‘§‘qƒHÐó!ÆHÎÃÝ„©å~¨ö›záÉôsü¬ž½}²Þ•ûlQ£ë xµ²nC‹‘ªÓlž¸ ¦Û†Ã›Œò8Vdç¶vÿšT—x/НJeôê«O.X¿5}îb”ruØg@Nî~‡=ìÕ—Äo[5îì¶ÈŽk¤Á=:ê¶«æü•¾žq­W&’Å"zènu¡áðõüË~Òý˜ÞâùESjOršä*Ç“Š®êÝ¢XžrõHD2ûŸbK˜ß®^¸Åûi&Ä6˼b)\ ]Mý‚€Å/juöµ%Ièÿ’úü…kWîNf?$[î¤lŒØÜT[}¹ycå¼ÿ³¹‚¸Ù ê¡kR w,¹ úJ o[Ï-ìaOÂèá,PÉ}[F²â‡f"/;Iµ“ï»æ¿ÇǬ.ÍÍž×ÐÍ3:W·Ç›5„B((ÞÈ£¯)íœ?²—ØÓçÛu¸ìÒöf>L.N½6Ou¨ÎGŸLpƒçÍ¥üWkÿó¯×ÔY«!:ÚÇ—AÖÝo’ô<”*Ç-½|䟯·Çöü]5¸†ë³rµ]BãuØT£:“Ë*KkaHC­É΢ÑÝÓW~N[²†ì”Á¯Ëæ_2üÚݲxâ˧h.%³~³‹Aoݺœ«Î©‰•´`è³…(_»ï(ù?®:c!lÁûub><8X:ê2ù–¥#+•<ÊaKè²áuȃcÂ…£BE¾¬„1ÏKÂO¯»ö¦{ÌDÎH Ÿ£Ž Y|I!Þî"üI'r☙Aö:vÖjøט£fP_:¦V)ž­Ôµ¬ž¤Ó(p,›ÛDý$j,û¬©hõ»ü':ìÝö$¯›Ž« /Žä)ÇžGB³Pól¸RëΨ޾ƒIž{£àŽòà‘ÏQ?<ßVP?X€øØÏcvsðk½Nlxn¾¼~˜l",të]v!ôGJË’–h©rXêø]2#ÐwÑÕçñôÅ}Ë‹ˆÁ:ÔÑ–å5V©¿~šxª™ŸÐ˜g¦ÕM»¸ýï;û2e;?Õ›É{-iÅJ;YD8¢€þoØÔ<µÝ¹Õ!ªâê²Á]f×)]‘Ù$Ámr¥ðš¼:MW_[!Ul˜bÇ;»‘)Â3;¬gÏN•‡ºáóå׋L‹&§:ú¨m–¶"Æyúyu©¡¸@•Uˆ,w;׌TÜÔ÷S …é@G:^€Ó½‚Ò0#y[ ªn¨?yr¼©ÒŠO¥g6[|ÝӸঋ-òþ•€W0”ï`§Éå÷™ö?´®7cÎý¹›ƒ”w¢U nã_ðȹÌfEÓñGæ±›pFòÕKõàMàBÃÒôøÁãË&b8ÚÉkæôaØ‚£´ò–ª´3†–¡’슩‹ÀÂ8YQj ƒŽIœçñO…•²D`ñJ¢û„|Ý«Ü]–ëÞ¾˜>Ú=( [yÒêÅÌøm*“TÈçзN{×9vË{Yæ$Ln‰šœòí|¼mFä¤{w³²"g·J×°9sÏLÖݹ:m¾èì^(~» Í]ϋޫZNã¶åZFÉ:Js‚Ì–©ÄÍÖ&ÿÿ ŽÅA endstream endobj 258 0 obj << /Length1 1385 /Length2 6006 /Length3 0 /Length 6956 /Filter /FlateDecode >> stream xÚtT”ïö.)*%9€t ÝHw§¤ ÌCÌCƒ Ý©€ " ‚´¤¨4’Ò!Ý] !üÇ8çüçÞµî]³Ö7ßÞûÙûÝûÝÏó±0êðÈá6e8 ÁÃÏ ”(hi© €@A^ PŸ…ÅŠp†üuã³CÜ= p˜Äÿ(¸C@”O„@á´à0€º§3€_À/"Á/*€@ñáîE Ðâ¨Ãa|¸«¯;ÔÞ:æ_¯v[¿¸¸(÷ït€œ Äj ‚´@ˆ êD[3Àn … |ÿQ‚]Êp•àãóööæ¹xðÂÝíe8¸ÞP„@âq÷‚€¿hƒ\ &ãÅg:@=þø àvo;€r8Cm!0T†' q ¨it\!°?`Í?nÀß»ðóòÿ»Üßì_… °ßÉ [[¸‹+æ …Ùì Î€Ž²&/ÂÁ ÁÀ¿€ g8*ä‚:ƒlP€ßƒÊrzjÀ¿ãyغC]¼Pç_#òý*ƒºe%Xîâ!<ðõ§u‡Ø¢®Ý—ïÏf`po˜ÿ_à ÛýìéÊgƒºyBÔÿBP.üÿøì!€0PLTP €¸ >¶|¿ÊúºB~ù¹Qú»Â]v¨! P;êßßä Ü=!þÿ;ðO ŸŸ†Ú"6{( ÿ?ÕQnˆÝµ|w¨Àˆâ?øë÷ï7K½Àp˜³ïà¿÷˧n¦£«¦ÌõgâÇäåá>AQ€0EW!q€(ê%ðŸUtAп]ÿ“ª³ƒøºE]Ó¿:öúKö¿âàü³˜6ÅZ€ý?$· mQþÿoªÿNù¿1üW•ÿÉÿ»!eOgçßaößñÿ# r:ûþ Hë‰@ @ Ž’ì¿¡&?¢Õ‚€¡ž.ÿUC€PBƒÙ£ÈÌÃ/Ä úã‡z(C} `](ÂÖáeþø~IÍ ƒèÂ= ¿¾-¨, ð¿b(}Ù:¡¾(^þ «äh¼yÖ‡¤°3x† ¬p]J£O–S ²_Íj¼UžéäW¶rü¤-ï†|½0õÝ­–¡ÎW茇AyËžÁ&iòö }<Õÿœ–i’¤-×$Mž^¢Ø½ M—.–ô3Þ#°}— I[MÇtª7+Y‰²¨zøퟫJ^iRЊï­è#šlz ×…P±å¬¸$†®ÆYD§kg2iH}¼ÀOñ®$]êéë»ÛI×ûLÝ+dýèaâ®’d/­¢N¥ñ\ñuN±×næ=•û²»à®²EŽ4dö‰ÃÚÉ®%#§[çëªNܽBM•”§™¹ })²ƒÉUß“-&Ë#a d>üCŸÁ°GF¯„$ïnha<Õó%hû6·.txÞ6î™{V6ur ñÖx~|áNtXXBCˆŽÊ »KÿÝDœ—µ%‚•ù§OïÆ¤»E`ÇŒEBö¶7ßáºÏ–¯:™H¯š8vI§ä XÏîíÄÛ ?/ìF~)ëèx/÷†ãË’èÕUÀ¨«…‰¤B€ÎÒ; ßwÀœwN Þü¤þsöP“£‚GsU«±QQÐ mÁcÒ—Ó¾˜Q¢ ݾlS¹·âj˜Ai)?“¨ž`c2Ÿ 5=½ÂË?‰çe%(먂ìQet‰u=ÀÉÿÑäURûi]Æà 胉ѴXÍGMÜìøaôÓZzñ@ì‰9þ¶ªs­çÝ .Ü¢”ÏÉN9›”X欤Cn¬À]ÙϦ?”†ªT0NzTbu&“ß÷×È †¯:Iç…™Q*¿¤ÚÉ~–=_›ù´Oš5dˆ™ŸíBÆ«íV9Þ &ް¡«cp-€f ;S·È5ø}«)Úñ,î’¨O=h:üü¤$qðÌpŸæƒ|VBÜ ‰1ÅW¥}gêá©ÊÞÚÄqM!||KgÃ3É3Ü?»—n«H_ªò«B1¢þ»oºâÌ9?PTi2:ãÐWÍþÈZ¤‘#1ú)[®9>9óÕJ²·õJÐKn‹„X)Â:DÅ‹íNVåà½!¥/i}‹ö‡¹€#†xÛ:o\ †S`ôýn ±œÛÛ1?¬8š-ׂÏL J›T+[×ìxF6 6J/)‘«ëÃÆ±T:Ãh(¢(S‡)³ü¸áÕ  ¶Ýö3)¼!åøÉ‚ç{Çê|2aJ =JlüÈï.‘rÒ`WKû}E]dG˜R…¦X}«æ1»’àu|§h 9ÀÀÖÒTÑ]åšJSNvÓ6šgˆV[’ïñWž¸Ng‡¶¦K¡03×±éG–Œ·sé•sÄÞ°Âú¯‹ö ÏÞãbË [ 9_ƨ·2º¬ÝÞÍ×­ ε~¾¹jF¨ †~E´2Û>8¸I~Ë€ÅÍsܰÐ)Ùü“æaa’ÂlWÇeûAþ´§‰¥¯z0º}‹±ù2#czo5K6…¡¥]?Ÿ£8„tõ¦p”€†[üaJù;HѺJŠ<Œì3 {ïQùŒ2’Fà ç…àˆAÄtèmLxÅëÀŠË} nt»êË2¡Ÿ!YLd%b„•Ó˜Väv{hm8Ókùu&/^?ÓIÜÓ{ÝYT·ÓD;m0ïfQ0“Œ>Ú):nŒô¾ÎýFÊ;ï‡'޽oˆ¼AÑ‹éL4·t,‰=ÊýŠ—:å?&õÄy­s’Y–û±Æ³¾\Všøjå¥Ú‚)ªccfê´@ÕW#Ð/­g:À2}làÊ­MRصôÊÛ÷( ô\œô w¶CDàD.óÙ[æ^“}QäÙÍF­×ñõÛRó¼Çä{÷öЖҩ[ž›ã ëäÔa},ÕT4ä?;²ó£9EYâ‡]Å¢¼=hÌå _jSJγö:oÅ8èÖ«Jf 0¢ …ñÔʦÛéwºã·V#­šò(_u†öt/õˆÐà²Ë ©‘!ò¤±jIŒ#áÑœ Œ²¡_¢ÍU"RpMÑNWJLø,+põ¼³ ?9r\Φ•¸é\©ñhg(ž+ý1F©Ñܳ`“sm,?j®³ Ö—÷Ó£Ûk·7G·Y!K?¿ NL©m©-ï›tgl·aæýdXÞ{J¨šÅ>¦0|:8‚ÇÃÅÙ°®fû¡Xlk&±ûà‘?x IP¢s«#øÃ¿ƒR·NÕZÚï[æ6_)õnËÑ ]1Ôh¨Žeç®ùÃ5ƒš¬ì0\ùîäºÓd¤Ý"›QN 7˜wWËP!Í¢èÁ+jT Ôs×óULÚPmk‰q­Ë^_L™ghèâ½ð"üÈá*ß.;o#wמÀU0 ¾qMÝ]×›¨·ø=Yš°íJŵ·g^#B\ûs}§ChŸÑsOÅ»õׂô§.“£*—k¼iÐöñ¤P<‡uµw††Ó‹§Òu£žèÆÉ3d -ÕR<¥Öéy邃5aòÈWáÅÅ ‚qÙc-g/Ö)d`ÔM’ˎѪ3šjhW.ùê*7/ZxÄîYŽå `Ò“ÈÝ"žÌÉóïÚ7cT4?"h¿H¯5žQ?é«ÉÔÿL+ ¿©w·£ë€°÷;úp’îxíò•‰þˆç Ái]V{>+eNUN¿=–] ¥PÅJ´ôÓˆfCÎr­Z³BÏé‚`òAâzB›~sÝw›“ßiøaˆ2JYs'˜~ÁyÁŒ©x-¥›Ç5®aTƒ¥’¢›£{ÚÎ`0s±}dûîãXÙË$³ ÎÖ¢kp²–£ûÑWfY©éö•eö†ì5çÐûnPPrÜAƒæ4½9¯¡*xDXõØÍ(Õ5 4Pø¦…iŒËœI[’fKý”ìýœË€(-8ôÇw¥¡o.ÓÆ, Î2,.@* Ú"£øw/uUš…ŒúMMN–ޜڈ(}{ße¬{7ƶaA"SB½á¯¾Vƒ¾Ùy… ¾18•+J ÕäÖ`Ç«5(ÜlöÿÞ0G¾ãJ0ÙÞú¨—] ªEùÅujÐéáëéÀ!—ÍÔöêÒ %Lý6oVò'LDF”¦Ëf5mi ú½¸„£ÁÛÙRÄÁ퇱Ÿf¬ÑóXoö=îÖÕÉkùEzøæš÷ú޼ÁBkªÍΠu2âüÉrg¤F„ê2scšLjêALXH¿œóMSŠP` ®™¬›dkòºùD‘0}ë âU»–úXNÜÞÔdïýtc|MÌÂXAÄP÷ã1të UÛüÔ˜hš÷¯>b´(ÑÐÔÔoÄÔÓí†_XI%eÅÚ­šŽC‚)hÙrHÏñ¤æHñ &äÜðÍBstT‰–“œF[aË·£\Ó¤!1ôA!Tmf¼éýÝx)ù\Á¼o¦vMlî[IÞ1žðLŸ¦Êë€ ÄŒš0’Ðëlˆf “H3}ª”+÷ŒÌ²Ý%N⛨¢3¿öËøÅý·? âRNab9çE Š6ª?j4ƒÊÇ‹Þ~»Î¶ FwÓÇëT9Ï Þ«ã®<ç*•ºãχ1Ò T0§ìãG̬»UÊî—^À»œ(‚%³4ZŽ¥‚Ë0÷4‰F-Û|DÓ¬¬Fn™ô„ L Ù'¿X&¡÷lHsɯaN4Y~´’h´ï˜R¬ÛÃB˜•¼’žfCÿô@T€ljöYM¤Y†‹`°ç­Ë2’ãHuÝ <ÓY¾M¯VHÊü\B \fl2I*aˆ}8><²[öP†O*ÍÞ׿΋²ÞÚð£“á¦Ó¢­yëq‘À£~:6Å#Ð%Å2(#R.ÿAµ±6V3#^šNF>k\Ùšîˆ]±í)»š¡±:mDD±ë´|qiêñ·3Úë‹›¸‰‹Žš—qÆÊ‰<ÈëîT3i&úŠ7mIà}·žù7U)˜¹TšÇòº0h£\Ni¡f½~ª’µ7ÂVû††ï³XÞLç6ö·ë€gÞ[Qt¢ß‡|¡Ç ö”Ìßÿ5qJjð•Éhù£ÈËXV¦ÏžY“NkŠòºs^íoÚM7ÉzÖ7¾ÀÜf©µ@ëßc÷SçÀe¾ò¥¬Ï Üo®¸ã«–ÊÆ¾•uÐWJ·Z™uäÚôxp‡oÔ+÷–›<Ù"c#èÍ‹–Œ´@²ª¹ÊR^m^âO=ÓØáå‹ìÓ…íõ|›oÏø®’N¡Hu<Ì—.Gmî‚v"k¸‰2\MuAéÑØœnÃjIqst³4®¢É¦V§Æ§8ÉI&o§2Ùë¥~n¸­'ÊpþO–~êŠÈ{À–•:ã<Æ•sÎXSäMÉC š™ùŒdŽÚðœôÀ—=–õ™­›Ikû‘y;"gÂæšÝä¨?Rïe¸èÓ&ñºÝ1pe—"1R «=Ù<Ý@ƒG+Qj"áM'”ãßk´uþ܆æì Í€ÐSžÓ3Ài;Þª„æ'ú³HÅ”bùyÁCГﲢ“Ö ÷¿àú¸$1aÓ„è³>fË穲„hÕoŒ‚ê¶*ñß1 ÷è©hä^[wDx1–åÍÕÕî °.U3¥Ãv^&úéËÙ¬«,N|ÉTx¿mHŽì0ÿˆä\ï;Ù}¼ÄhE=ía‚kgj&f‹ä¼cw¾8ðïxò‰T«R„°Ýí~ÄœõbLm:ò –sÁT_Θd"ER)×ýÛ+v«c¥´Ý[Ìï°š¥âŽg¯Ï«Âõ9xÔTù°ÝÝn0x £À3Ó{LEq̸¯+n¤÷ÁaÓrôÈúŸRæ‹f½Â<ËôÀ¯ýnKIÛÃMeÜ•_I80A’·ÎÎA¨õÕ u;i¨j塵›/\£æ¶ èé å¦#iNÞ'pJñMæ†A±jÛHâ{–ÓÚ‘ÇmWN|vA+˜¬Üv6~‡7c)5b õ]æØºý!¼ÙånóʶŒÁ‘:KùÁ Š–±'!ô¯Õî_(ÒÍ´nÄ{¿s×ûìԒɈ‰õ:Š­ïèÄéPð̸(‰Ùò ›y@/ó2dsÞ”ãC25ÍAT@e­~(òžã'•¼øæ6è ‚}4Aq²)>õÎxÃIuYœySÛ%æê£DÙ¬‡ÉÔŸ†a?JÖÈY¥|r¢ïȸ}N&jÇTQN n¸¦ßšú‘Z}¡¾.Ë·À¼ðiºÒ7é 09PÕÕ¶ØJ½‘2@tË}ëDÓ±4|Õµ?vÂärdxÏÕ~!‚*‡À4·‰döÃâH^o:êàç\7Ž4FÒn½UÚýoq,^…™ûàí'O¿ÊWTð=}VN(›`*ÝAÔ×!´¿}À6#ДˆE47òb’Üg –NÜÂXª©ßªC\9¾&]Ÿ-•¢d… S@ûÖáòõq.èË;˜FZ¶C^Óh?g†§s‚óävÏ÷/å|µó,6Ó[uÙ•µ¸ÇÍ B5Ôcš6»÷Ê÷¢žF&ø¥ž9”âY=ɽëx›Ø±Âä Ñâìžö¥ Ábž•i[x‹A|pÂN¨˜«wÀ÷¨ÍÛÁ·3î¶ù)*‡ý©ÒótêPk+9èxžÌf’ž>óbäClíÙìS_ªk‚n4= ó¹¬â×^RL_¼¯@ÊÕD¬7¹ÍÅÍ'÷ïÇS°'<äA¹}{Ì(Ó±V¼“-¼Jn òо0y|Å ÎëùŠ#{`Õ.ÚÁe©sn•UMMD·ùº n×N½kµë#n×OÄG«ùªŠSÞj :ՊŹ܉ð#¶’n’¢áSÖd^ÓâØÇ…IšVQ'Þx`@yrIĤöyzC­cæÝí}AÞ¼µ/ƒ"ÓŠ2½Glõ´‡)ÝE <ãåÍk¥ù§'¯(øEˆm_$ú™¤.ÐbUU¸ R5ö G“›úX²‰«Ëgkø±èC\}ÂðTsÝÞúI±ãTÃVœß¦C‚ibr§[+¤P~Û\½sœSw̬€ÛÇÓ cL½[öMÝ}‚nI.qùtjüY{Îülé•uðÏ€ä/^]ôUý<Š¥z-m>IÀ…å÷\†o<\æOÞÚuÚ°N,õ¸Ú ¥»êi{‘”§>¿bäæK×a˜ˆÃʧ?•ˆ©­p÷ÑÙ°/«âãà)8+¶ >è$DFU(§X¶¼ç"–ÐK´ ²gÎÓ²lªÙ=Äêµ®f¦o1“nH8—ÞwËØÎùë|‹“W« wc åZû"aA¼ÜÎ7¸v©èî<ÏEá†x!/‚›ÇbR²iúg_²ós¶ Ęè=?Ð÷§–U‘Èáë}“ ·+&™¥Œ®œ‰¨”ì—o 8ÿ@¦±ÐøœPT{‡P^qçƒnŠ`Ò: ¿ØòŽ%r˜±úsšÈãÎb'"-ô _î¾U´¬‡ZÇ x«ämºé¡²ÌéfiÃ'0š9çGâ¨û/ðqsƒÃ#KпÎ{ROo±%Rã˜nYÆà*&µT!à°·"b±¤÷ÁÒüÂx8Q*­¸›ðÂO2ÌCxÒd Rí ô¬_´Ìd÷Bq÷ªøÎç±öH±óÞÌ!Æ×#ñe¾«m±ŸåŒEÛð"&fk÷¿\¸â–!>Gb>?¤®||»ç*5f=È(tìñªnÅUèç:Gfœ¶aTRð î†^œ8©ç™âs“ïš ÷Š÷­©Z6ÙÖǽ£¥ÃŽMp4¶ÉÄ«>ëYŒ™»¿ûf:]é0µë6Fk!â²Å/ËæP®üIÀå®}tšú}á§éÒ“Ð}Þã$'ꇅ/»^5îlç…²ÎÛБ¿¾s[Gqäx7¼>5ÊŸÒ­{=5‚Ü÷Ûï¡€±> ¹‡\‡ß!âÁ§7»_2©mÄGNß»¹"(1÷)/Á N£ø¨ _ˆ€³)O;\Y\eRZ‚ÎÈ¢­TmAœî.ïÕÎü´¹Ãï"î}Ñþ=5ÛÕ%«çé¤fIÇæÞLbDë•f ÉJÀÖ"yðd­b0ÚåÖé¹~¾3ºMIØS Óºs®ù897’ø*éÆ)tašVñì,J‚ªÁן¡]|ûÇöbÏDÉmsÕHƒ=-Çbê_ªâãeÊíõuÈN«n íW¾]24éêÊÙY¡MÌhô]Hß/„aíGt­<¸ÈÐÈ?q JVs ß[¹g7|úi¬÷g¾Ëál,ZŒS0éȽæ;þn@' ÿ4êÚ mÜ~Ü"1AÊü„ÝÞ6‘+õƒs'MêöOñda»_f1ùмÙf›^JJˆ}É8×IBY›w–2=-Å8Ý“U³ŸÍ A®%5&ƒhD3Þ,M0Òˆ;–®áýL¢‰z endstream endobj 260 0 obj << /Length1 1546 /Length2 8420 /Length3 0 /Length 9447 /Filter /FlateDecode >> stream xÚ´T”[6L "!))Hw§tw‡ä 1ÔÐ ÝÒÝÝ) ÒÒJHHwׇzÞsÎûþÿZß·f­göu×¾ãº7%™²£ˆ©­1HÒedebáˆ)(ÈpXXØ™XXØP()ÕÁPkÐ_bJMƒ#ØÂ÷/1ú$BŸìl!Y'k+;€•‹•›…ÀÆÂÂûC[>€8Ðl P`ÈÚB@Ž(”b¶vn`s èÓ5ÿ9hLh¬¼¼Ü ¿Ý"6 ° PB-@6O7š­j¶&`Ôí¿BÐX@¡v|ÌÌ...L@G&[sAZ€ jP9‚œA¦€_6 ?•1¡PÔ-ÀŽäj¶fP ð$°›€ ŽONSàér€šŒ<@Éùc,ÿÇ€ðWo¬L¬‡ûËûW 0ä·3ÐÄÄÖÆqCÌf`k@IRž ê e!¦¿ ÖŽ¶Oþ@g Øhüdð;s @RD|*ð¯òMÀvPG&G°õ¯™…yê²ÄTÌÖÆ:¢üÊOì2yj»óŸÉZAl] 30ÄÔìW¦Nv̰½HFü/“'Ê?2sÀÉÂÃÍÎÃÙ@®&̿«»Ù~+Y‰Ÿ*ðò°³µ˜=ò›žþP<Î ÔÁ äåñoÅ#VV€)Ø 0™ƒ!(ÿDƒÌþà§á;€]º,OÜc°üúý}Ò¢—©-ÄÚíóßóeVSÑV–•¢ÿSñß:QQ[W€#+€‘“ÀÊÊÍ à~:xýwe ø¯,Xþq•˜ÙXYþdûÔ¦ÿdìühþZZÀS´}b-@óÉõX8YLž>¬ÿÏTÿíòÿÇð_Qþo$ÿß„$¬­«i~ëÿ?j  ØÚí/ƒ'Ò:AŸ@Áöi ÿkªú³´ S°“Íÿje À§E˜?‘™‘•ƒ‰…ãì( v™*ƒ¡&(óG®ñkÕ¬Á²­#ø×ÛòäÅÂò?º§ý2±zz?ŸxùGt|Z6èï1þ §uúï<$ &¶¦¿öŽ“ tpº¡<þ qƒÑ“­• úš¯Ð¡F™$üåççB¦æ½j”žÔ5Ýs±.T8E’ܲ­ŠíŸ«ŠÒäñˆy5÷ÖT¡MÆêœgøÔ™k6Qþ?>è…&*¦Ë ´_£Ä¸T¾\éXÂì!é ÷•uöÛ8zÅ”à ç@ªÔ\ÊA¿Ë,tÞMÁò0–•¦Õ¤-ŠÞHíÕÚEð®²ÙŒØtÊåïò±Ü-¹KùPGÂ…a—f¤2Ò·qP[}ê»${Fú8„ïû¬5ª‚ýH8ãõ^Þ‰t+ q¥uO):@cø­Ú „•ï,åù,©»[$YÙìP%ÊÆHI?lþ&×Ò!Õ¾6™iœnÉûðnóØoU“×ùa‰æ»ïÔM_ ÕyÑᇊÞYM P ú|6 a ÉQP˜kÎãyfÓÄÌË«Œ˜®Ú@úÂúØ,·šñÌTßýyÙýr¹„F/Ñ}á—Wï;‘¦ïÒÝX^â”_DË }oÔLcõÑQø' rJ´Žô§ŸÅõ!íÛÅ‚Ÿ§çŠò2yk¿_Ž'ªPÇQvå|x1^\J]޵oŸ¡Žû|Zþ\ô#e`RøXË÷°Ÿ¡/O^ GÙ€7ý#ìueÐèu /8س·øg§67©²jõTö®ÀrWhÌÛ„tIãcôÕ2ŽLäTÅ~™[€¦.ÞЪ*Ç+2KáôÕ8žhËVÑpU*Û²ÆoÓ²½dÿúeñFGwlnÚw“œç¨çŠfmŠ9'­4y~¹Ã¼È›ðw´ÆZîäŽwïNy+¸£©.U>áRÁ=lŽ;_Ý1ƲÁyܳ/øïk Âd¢î6бå EóßïxÔ£8^Å ó,a?GàãXÓ¼­êvy³hŸ®œéhÒе½\˜I)QTjN¬äyàÎæðDef}J´ÛäÆsB1.BtN¼Üíý.OÛÐÎùBŸFöÃsúÔþø¾ò¶ï]!˜;Ì4ÏA~eæ…sàǽC4[‹4”d>-瑺Á%òDz ZˤìI•Ö툻âi—ú”l2PSh8-Òœ¤2—tR„Ô;íG^g8Sû‡-n1ÝbŽ´ -}òãXÖ°4Å»0™{fJƃ âe˜[AôÚ¦Tªá‹O­Íiº£±·ðò…Wðu:£ƒŸÉ:0ä™ÚÑ%VA†9b&§2ä œvÖAoÐ[×;ƒ)´îºñ‚wïÏf{+r4é° Ó ßI[èG_߇(t›ŽSU¹˜Ùú¯×¿«5tMþ¤”S?u%e«»J žåq7ÊÀèj ÂV[yÕäAc9Î$§ªóá¼ã—Â)è ¾©ëG\A"DuìëL\H<è“ënÇ×­Ìä€ì—Z¤DmlŠ]/–mðkþ:l;#v’jŸ+Í Qá.s¦iÐ +‹89ùÙš½Þ¿ÂïÂDõÇÃ5˜2‚ùЍYçjȸ䇰œh§×»ÿy8qdÆ5ðÕªcJ¢:Bkua®°™uÌuï r I>†×îϵ Ū2],[ÍJ:o>„Ç?âPt‰Äo]å†r0ºÆ)Éy2vÉJ'RŽ&s1Y^JƒJͨæ]ª C§=ƒ8Å´ˆhs>$úåûÞœ„ŒÎr%;{áp°‰†¢ÃõMŽœÚ—Q¿ÿ!P^î:Êa4$bÛ=–m]Ù¯:§øl6ýÁÏúné{z¼NJƂؠĖ ÍôIôðâl-:SH14íkùê&!¾¸e ë |jÛ®ÃþbÑ[Zuy2d‚ ³X¡¹9¾òM¤¬Zrq´¢EÎê]%ZŸé| \âÅ€¶C(óì(RÒ#+ü6Ôq^IαJ"‘Ï+~߃;¼¹ ›W¦ªíÐ7„³WKƒ¡/è§# #CKßwœ¢ÿ0à@kax•4F@‘!«o;ÏV·—i{CW̓Î}ïpE:ÜD؀Dž¨™'Üè¥pgÃ"°K²£mP‚˽ eIØ••”2DÙmqÓûv´Œíé9Ñ]¿Á¹iàÍ; Šà,+1Áj %3UÀín““›…Ï&«S¹ÖËyÉxÁ»‹—ÊOX Û¯‚‚áÒä"OiY)°j?i%dHšå'>õS~;T 52Í!Âò3”]æ]²6ü×+¡=©%ÓWÎð 4Ÿ‰§‚}jnê "°Ãx¯A~ {´!+QåX]ëÓ';ô¥6Ž›|%›p,)¯àÛú‰™šbÂ<E š!x°œÚÃ=ŽÒÀe¾hW–m„q¹¼ÉÀÞ=*9ö°aÃÐ}UñBäzë+?’®†·û=6¤ŠójåH|ZˆÝÓ|a|¶71¬ÏxÂf’²nÁÔ¢çìm\Жìð)£•†;+:!5ã·p1x9kù?1‰ò[¬hyr°H§ú‡&Q·ê\ Å £ø&7uÞ‡Rš`¨E/0.^-l¤a³ÖúsžV\ñލVM ¥í}u¶wzã)Ò0ÛäÞܯ¸)^…ÉKd€St{1*Óä'••²µÄGî®[ñ7J¨œMÇOï?†ùñ哜µ-ºsÁ‡qvÄm~(F-™h8…³BËœUÅ8âÛo‰ý ”µ|Û8–>'†ió…È4uDZyú“‰æå½Êc?~0l Ø‚¡Çá%4­uqY@‡v{4¡¶òÂQ0k,¶ó¢è½ qøú.Ë»½9Šãäð»jsöË׊‰¯©}q:øÔÓÖzb‰ü3¿h@ï½g$É>¶Ã, =ÎÑe#íó¢gT~Az{¾½'¿È+¹ùº~ÍLó[äe^;Gßà @Û»#ºŠ¶vâ˜Éœaxx¾š.ÔfÕ‹&F=úYû¸NÏG ƒÉAAúÏù²¯%McñÇ)ÛkÃJIF¶øÞü4—P_‡©vzT‰±Óôų£âÖåµ_ÄT±•9¸¿«×L@ˆR`ÓöD]:i`HËïV}î=«Íæ™×$£J®‘ã0T5ï>œ"U<‡½ñì)fŒ³›s•Ú¿?„/ňHGò¢FÚ 7ïpŸÃ³®)Š ¾:ÚaÈÚ «¥h,Hiù$&Yžœ c“hðbOøvÁœãLsô¡ ‘™>ñܬ,fÖ䔨K©û*'üæí¯*x‘“nbQ­ZŽóx?к«Ö¨Bk|eÔ‰0%G}d÷ÛAç?#¨k@o4³Æ­(_Õ<ïYÂÛOp¶¦¸ æ:N}^ËsŽ&½Ü.â7Çš«_ž Ádé /¸¬o„0»˜é¡ì^ÁÚlÅœøÞÔtÁ`n¾§®M÷‚7:e@èÜ>™¦Ù\g-Y$}ßµí>­j)âråTä¡€*UÙg"~A/ôË{óö{¼?ÎG5¥´ %Üœáåáõ-Dën˜$¬û’½¥#H Ýþ´a’r¾>^²LWÂ$CîË ¬íq?ÄÙßÀY0ø²àv¦ÖËgx²K ®+‰‰?Ù)Sq¨ë—%ÌYéÛ‹¾'¥Ê÷Ç ©¸ÏˆÓkPV×Òó¢næDïÛy è_ËtO Ò4}ÃýŒHìh•“÷Óö©Ž—,àjNÇ5àKâšD}\ýËrÝØîµ˜Œ‹FYAªòPQ¿ãçøj&)³ag¯‡¼¼9Å}7lÄ5Z-¦·å¼Yh­úŽæi•/)™­1=¡»ˆI&m·¯¼cWÕQ ->/[TÃGX$w´.}$ù<(V_m 8[‚¨¬¨YÚ“ê÷P£«œuBl­x©†æØ/jq*8Ùæ´ûyîF½·õ¹¶Xé$^Á%Àçk[ ~\ì7}fÄè¨õ£î£Ñ6‚Þd8ußÃöð~ÙšdÝ×/ÉfÕ<¢3fó=K…4´0µ•Éä*+”Æ_„ñ{íéc ”¸D&ÖHKz¼V”ù¤h-ý¤éd½€äFª Ó†™õPôÁÞf)˜3â´ä]…eRqºU†«uJÆÒ‡ LÕ™ï3Þ†ã9Êî½*#VËöϦh5¡<€vƒ-¼K´=k¹¨Ì¸r?(7…þúnN –«b¼ˆ¦¸²%4X¥^¤* S}žgx’2†ŒV"Š/1Ó¸ahR¨ù2•ÍY«ÏZò{ª)u㺛çæLîö<;ÑIµŽ¤m–Q—¾#JZh«^q:ËmJÜ€5¿“Ý4n]7ÕUóÛÇcļÆ@Þ„Fß)Qrk¹Çž&î[¾¯î:Ô0×±o ØØßŠŠTë$N¨[>ÒÅ0ûºVjÙÌ×–‹”?d#ïä­|šNÂ÷í¨œÿ‡÷³d€Q´Š'šdU',1¾¨¥u*¦§¸ïì@‰S]»ÜÑ!ÁùÛ;Xz&öW÷-Ís˜L»—¯ŒÒ6ãÄÜcÞ ]{F6}U¶ÈÞïñf[£ùýmQ2xã+ÅhÐaÿØ'ª·„6ÑÏÇÕk.·Ž>|y;ÚUTš‚%–Žg³f‰v»¹"¹f^sÜè¡3tm]šqfvY¶ë ò8?Û5oñjIMâ`¯'•ì¬éf§™F`09%’!+ÊÌ} <^ ª3í’ öJPU¡S¥{3;mRv×¹0†Ù%<‡Êݓ곪üe„JÞ)0›/Ûä¿ÕÇd „rm´Xy6-õlœWKhßnf啯Ø3vÔ®ÆÈf 4Ì“Æ= |š\Ô€ÑO‰+þÝóæµ½µK~fÐ#?Þ‰#åèD&"ž5¹Š-­/CÝÅg²²²üƒ\˜´Å瘨÷Ò)‘ÃM§*|¹ƒ$8¨?YfÄ¿³®­Éñõ"Ñ¿½£9› ٠о|‰VZWÍþcŽQÌIn]PÕœÍÿ¦vWèJÚG¾¶êÕcqTÀò$¸„ÏOK ]ÕóxD¥Ò} ìÓF«I/ÀIŒÛ$dõcX–í:)QºQ­+;hÈo-c)Àœà;!FÇ?áCoÏåm?³„Õdÿ‘ýIÇæ˜Â‰¢)LÞh©A ¶P¤Dð ćB2*bõ¨8ÄñjÿS±â(Þ~³•‰’ýÕò0?¢B*Z™…ᘲ¬ù(64o‹t98ì¼íß½vÚ`Q6dþ8_¸+´@îÛ´ø©-_øC rìëÜffXB(¹1b䣗©&ø‚h•’Ž’³'—icË;¼hX3ÆÖì"Є‰?¯sâ 1ì1ðš9¼ÒŽr`ËÛº(ëqñ99¾ÂˆÅ5WÃú]Q(Wï÷|AK³3UÌÂsBõÎù­ýÞÂô:ìidGtÒ^¸BË(šÎÍâÒØWŸØ(·Yzý˜>ÙÖz(ú!Ð^€ŽœÌ±?ýċХ^{ucL@+L&<@Ç ¸É›gHDÁ`ä¼@ªôpwø¿eˆ@C5`ëã2ÊŠ(°j.²tÐ_>¸ÈÓb¨É[†7¢Û¡8;¾T>²+7f”Ž5üàëz2\×þ€ÃãrÄ@‘ªw"U12Çʉ¼±ô–‚}«Â¯›–m^ªq»MÂ*îŸ{9*`¸èNbÒฟò@Kžþ¥ šŸwã;Ö“ý¡D9€ÉÒ§nó’ÕãàN&€yiGA‹a"V^TBБOÀ4°«Uä•=Ûu¬vcàƒ¡ £kUÖa{^²–ž\1à3vá¶f>§.XzvõæJ·<K;ÖCœuŠ{$$Wá|¨öxÈ™mž´üž–ü°L>ÈAé!¶ëwjÞ[¢ŒßžáÇRÏV 7€c’VÔLš‹ >éé ‰_bQL–§g¶nÂú¡2¼Wnß®3‹Rs˜êröeÔ@k×¥çî¨DE\?µû½0Ï{š8ò‡w“÷鸜)rä°“œâê§(àÄ$'B² kIË#Jom|3W±­ñÚ|/NÓ(§¬}â[ì‡ç¾ìA²_Æ¢í´7vOL;z Æ÷ìß `47ZèöF&%qÇ{"ßÀà5FRêäüœ÷YÍßuC=Uºë÷ÑĆ;ñ3ž,eÝŸ‡×çõʽ F<ö~Äv7]­.‚]ž›ß:6tú:â.$Ÿ€¤"„úÎ?䳚ÂöZðàêš§—¸—Ü*Ù\L~: ejî–¼h<6+êõ¸5EÚN/^¬¨‚e›¶ïgïöÉò…ª‹]ˆA³·‘ú®c‘ný'ìÑ…Ì bäÅÊ寉 Ë5&ñŸÃXN…h/Ø”TÆ€™g–&jI¦Eu×¶ºúûäG»^1õ¾ÝùvéöñpêÝXÐv;¼’v÷Ÿ=‹ë3˜-…‡ðº§¾ Ø¢úÆ‚í¼1á¬:NÓ¶ 2âÙáV=qSµ i®ã0>âÆ¢)QQÿ¬íN,ÆKU¹³ê‡øUÓd/ Ÿ—(d=2žÒˆËS'e)05.æDlDm[?ö¬2rî?sÂUà ðö)Ÿó%|ÌÅŠì’¾ÇÿÐimÙj»zsÜÌDGT¾¶áî6}`˜pÍ<âƒpUSL6~âç2V;Wø.~‘ðÝ6Hö ®Ûæ’/­†“´‚7rβ¥¿’tÚ[7Àö,á‚„‚æª= †•údCŸvö]œ.[[0Z&©H!xŠÂ¾ŸŽŽÐ÷pxe5\åGYk)“µÐÁG®wÅç›ÑS‚hâBßÄ%ìæžKØg˜0 Gä ᱄ùäêÁ$úÑa£ºm7ÜãºÀºæ¼Žp#Yó¥ÔÅ<òž¸í@£&®V¸À<¼ˆOcuŽK& ËD:´ŠkfS Îújsµ*x;N¶yQo¦¬Ï6ÎEoÖ? ÁÙ8„PY_b3H>(±S¨ZU¾3[?YâD4êÖ%ÙsÌ…sý: ¾:«ßëÇáwj©q¦zÇwº96×ÿ˜ãî;KâFÅÅJÒûÈ÷3V‡·!‘9ÌöÓm2¢i„:¡†FùRœŸ=ïcdÑé,L!ñþ6ò–º£ Mä縄dN!øŠÈ+âžä†"Ü¡înCúµÎw-~èzÀÛ‰ ÊIâ»5”ܬÖNÝwn°ôÃ*Š”| ÛñæåÙ;£HçÊȰ'ÉyS*.°<êÕÏ1¿}^»‹ÙØô¹aóL½ZÚ]â9©ªŒ}3l\ ©s%¼z2‡Jjcr™Ýü9Q¥¬4€š/¾&'žO¡¬ïGot^uﮃª™qý}æÏz¥-Cž™g5p[éJªmGóΟ`ã^6¦EÓ´ú32fÙ–arÓ±dFÃ.>Hžúr¥œÞ3£É“HY4 ?×„Ç ×Ií²&G¸{|»ZzÜ`ëHI¸«Z@µÓå°û1#är¨üXX|‡†Ç%éÚe+Õ¤åTÙÑël"½}­Q¤\9Úw„ÒésY$ó¾—6mÓá>¹§ X*%€¤¯Ãn ,’gêoàDûÁ‘Úáy³8 òG”k¢T.î8zšd{Î08ˆƒùµA ’ÀpS ÓÍ\aÖn’L–+˜ÚV¯å)t·—š„w,‹Òú>"¸S“ËT¼·8Í>e¬k*ÔòBü¾6üµÁ J9ôYádpú ª,¥¾z–Åoa÷5g;ïT¢6ܼ>Òq»DeNCÅÅ-jÚ¿¢-PÈ%¹l¯c‘X܆:¸¹üp¢äçÑÿVò Q_`ývz>Qì3];Œ3ÑÜ¿(³^@"çZç/ï>' ¥‹«Õƈ ËErÅFÊFø¿%þ܈l×·68Û_K!Kšä®5ô¢>¸L * /Õ®Xùäg~‡Kr¡=Ö–Ë^)I‡¥ˆ‚‰ßÆËÊT÷eød\² Çdlóü ꡻ɔ?ü4,Åk¸Y‰!”bAT¨ÑŽQù•íÛ)Ð-RƦÀwæ—…´ÖLƒzy€WL^}ÈÜÏÖW±ûK$øô°ˆOiiŠ}rh[ŒD××É Ž~¢ÅRR%Å †­×)ÙÒøll$’€DƒóC¶À$Éþ ¼Ÿ ô×"² ÇHë‘ïFXŠPùhsVÂ5 ^Š‘®2¥ Uk¬˜…hÛDEöÝ ÷<ú³¤RÔvx¥É%,³ßÍføm©¾á›èû6¿ØÚ¦m1cññ­g]‚0JØB'>ò·B.ù2_ pˆÚx•XŽÞiÿ¾n)³öÐæ ºäREÉTŸÈ}à2»²§ûz¿ý6£aÕ³~µ/kƒËÆþøõöp¡Qhs¹€€#åCmLíè§þt±‰pמ¦îêüžl¯+@}YñrÊ™ò>Rb„‡È*6¬7*O=²1Pȹȡ4¡mGÀoè5”¬)¦É²à‘tîBPâåX—·75o“ż×]†µ<'韓¡‘’½tåÉŽº”3ÛN•%W3ƒ8_!¤r'|Vo&ZSãt^É•¢¡û/WU|èinVÆ`<Ë´Ws‹FàbÔ§~.³ôs©¡X´Þë© õ=@ îžÿ9±@˜/£Uò§xÎ|ã<Óí)³^uÎS4 \¾™²ð´[ï0'S§k µrÓXˆŸY}…ßœ5ÇÞ1Û"§óÕ Þ.Ѩ֗djÈÝñµ&½“ ¦ér+˜|€gPgçåmD e•x rññ¹©-©ØeÆlFeÆfo £ðQ=ÜÍÍ)Rë§øø—a^V ‘ V±m{Mª“{(õTKOT ‡Âx1©ã Ì6 KGäNÐF hµYžM$BOж;¢e¢©ñ‡ ?™3*ÙÐŽx­ ÕødAÖ¡Mu§ôÛ¾SÈ_ûr-Öóýoóá£#,g‘¡tÕ s>êÞº…ÚŸÖé †f\ãIž/Þ:jC‰¨Zº02 “Ùƒ¥ 𬾋æ£[oŸ'¸L¶fÃJ»¿Ë¬ë¾rUÆÁ|g`JÛ[¦²0G z9wœ2 Ç«žHTOV–ãY)“Þ¦¯wõcGÖJëQH²~a¤tjð>ßÔ"ã[ƒôJ1¦”þqÖ¼Z·g \Jv¸‘LÅ]âÇóC/¦hÀyuãËFíŸ|7NIÔÀûa!°~´4z&ù-7Fƒ ªùû‚Œ™aÅ›ëT‘"ƒ”ÛØ7EyY•¹÷YÞ>§’ð›Ï&0ÙÂÙë K­9ì’ÐB¹–‹‰bâÈZ²Ûu—³5r9:媷'“7Vn×Hq‚Õ!2³ÏH(jnw‚Œü¿FµY†Àª¨Ã¦Œ6ä6²¢îO¹þlR>‹ñ& ^O„Zº[¼H¨g:Q¯°½'ÏÜ5fxgýuåj<äþnÛ¼ijt¨2ì›#ÅóbuåÏ@Õ­¦÷ "¹ŠA^ìäS…uõ_œÅ¾Õušj<§`]X·'ápoÿâ7ñšoN#Î(Ì&I=N ³ßéäÌ/³e£Áqõ’á1抃`ô®]+)|k݆ÜoG'gº!Y4ïp F—G“&|Z÷±ÒÔEË1„ü1AÖ"ã¹(Tí3“N=zX7ëy§ë»x¾Lwm»u8Ä…uU®àç 1ÑJ7æj:†¯sÆIíhè'åQ°H‘åYjÚÝÀÁ5bØpGaÁ­*Z¡þ6.f›9oÜ;44§ó’Üt‚”> stream xÚŒ÷P]ÛÒ ãÁƒ»lÜÝÝÝ‚»³±àl\‚CÐ Á]‚k ¸{‚»»{pòö9çÞ›œïÿ«Þ+ª`îžÝ£m®Ù;u&1 G3 ´£ˆ‰™• ¡¤ÆÆ `eå`feeG¤¢Ò°Ùÿ#F¤Òº¸Ú8:ðÿa á4e’¦ °’£@ÞÍÀÆ`ãægãáge°³²òý×ÐÑ… iêncPbÈ;:]©$¼\l¬¬Aà0ÿ}КÓØøøxÿ>³ºØ˜›:”LAÖ@{pDsS;€º£¹ äõ/´‚Ö ? ‹‡‡³©½+³£‹•0#ÀÃd Pº]Ü€¿(›ÚÿÉŒ‘  amãú\ÝÑäaê€v6æ@Wð 7    .§Pq:üc¬ø#à?µ°1³ýÏÝNÿåÈÆáïÃ¦ææŽöN¦^6VK; @EZ‘ä b˜:Xüehjçê>oênjcgj6ø›¹)@ZL` Nð?鹚»Ø8\™]mìþJ‘å/7à*K9XH8ÚÛ@®ˆñ“´qšƒËîÅòOgß;8z8øüXÚ8XXþ•„…›‹¦ƒ³PNò?&`âo™àbeeåáã@Osk–¿Ükx9ÿV²ý%gàçãäè°'ô³±‚ÿ ú¸šº 7 ŸÏŸŠ#D66€…9`´²q@üí,ZþƒÁÍw±ñ賂g Àú×Ïÿž Áãeáè`çõÛüïþ²èꪩ©h3ü“ñÿtâ⎞&NV;+€í¯!ã?øýÛÍ;S›ÿÐøã¬œƒ¥#€ï¶à2ý—±û€ö?ËAø·/eGðÔ´¿‡Ü€•‹Õü‹íÿó¨ÿ}äÿß„ÿååÿmÈÿ/!i7;»¿Õ´ëÿÔ¦ö6v^ÿ1­¼JŽà5pø¿¦ÚÀ–V haãfÿµr Sð"ˆ9XÙý¯Œ6®Ò6ž@‹w6 së¦å¹æ_[fgã|çèjó×µ`·æÿèÀ«eþ|u¸‚Gòo¼9ÿ)å`îhñ׊±sqL]\L½ÁM#.€x-€ž1€…ÙÁ>§ç°ttAü«£Ü\±¿Dÿ n‹øoÄ`‘øx,’¿€EꈇÀ"ý±Xd~#v‹ìoÄ`‘û8,ò¿˜‹Âoæ¢ø¹(ýF`.Ê¿˜‹Êÿ/˜Ë»ßÌEõ7sQûÀ\Ô#0ßÌEó7sÑúÀ\´#0ßÌE÷ˆliú-Í~#033Só÷@ð{Çô[Îñ?ù?»ó?صùo¦knãbîfoinñÅ\àæŽvàqú¯„ó/CðEù›Ç_sÆbñ3þöÎø¯Àlì`JàÉ3uµþÝÿ¿Î8»÷õ·pq-C°åó/hó‡O0þ ¹8þ‚î¿Y°ý%øíœë/sG7—?¢ ¬þ€`ÿ¿Ùq‚»gíåd tøÃ,û#>+8MÛ? ¸Sïÿ€à2þ™¸ôöÿƒì`2ö¦æ.™ |ã²ü†\`oàýýC.‡ão~`ŽÿRƒ]:ýVƒ9_Ëÿ N¶ÿHÿ=àXN@ðë÷Sî¿e6Ž¿»Ì ®—“›ë1ÁçßófàìæZ˜Ùý+,çoÅ¿#³ýÕÔ?Z®ðoÿ\à]ö6ÿžF®¿l€î4† ìÄÕæ÷s‚óqµûsÖØØÀéü ~‰°€¬]€Œ,¸^ Ç?€}¸ýÁÝtÿ‚™yü1‰àÓž@°{¯? ¸.Þ¿É=y]þ õ¯‹ØÜÍÜÐ߯Jð®üÿýázÍçÍBlëBÚîkÄ=˜vÇ„¦©vµSé˜|]ÚÝQß|¦«Î Zw¹û<܃¶²-E{#ºDúìsÜ\ÿ&¼%Qµõ—ï“q¼Úän+âÂÎÀxá±Ø×~b"& Ñ=ßgg_­À÷ÐÍòT¹În¼¨ïò1ï=úd<¿ö—-„ÍíªîUs+ =•M1ÅhFÏPå™eÍâ‘Øˆáé1.<ßÎÜÜNc䌿’ÊÇ3 úÄpùèm°Ç>Ìz¯Vh°»váSâëáCß`ŒLRûˆ$ËãÎû”­ ,x6 ‘æ 0¦¬0¡1°gTÛ¨E94öÖº,v°íä~„Ð`î$Ö–6b»«`W·Ga‚j9Þ‰:,ÅWÛnµº-mSˆ–ø¦¾¾fPt­šº}~Õ;¯`º Ki¹nü¾Ýí§],Ò/BdåÁF/e­¹²ÌŸAœ[…êÒÊa¡1.¹æ}‚5 ô ú)þHÀ¶UÈÊªÇ Û 9BtÆ•ëû3Ö¯²áB¹cc}ÞÎÃ0ø5˜[ùÕ¬3åâ¢ßšCÊ$3Œ6/CªûcÞZª@°a&¸Œåb£‡Y²Ô‹T1a‹káç e§œLïî|Õp¤A—í_ó™V¥fë'®U²“(¦Ñƒ†À!ñàÐ>¾[‹ƒ­Ûòrq#ìCG­‡‹JñáØM ž¼j9]®*i2bQP_î¶g{Ì(N5ÄÝÉìãÞjk=nv}yò`,““¤¢)ZTêW8 ÷]4\,êů ÅùxÓÀ^×®rO œ‰¶K•¹· MaÜ>ܨõó…@œP]ã>ŸÂ{¢w~jt›ìÇä6¥]¥A1g_@€+: Ú.žšl~; ýj]ˆ6:òÅ2²iMÓC¬~‰¨6cf¸~24#Nñ¦ƒKxìDPXJ“¹¦XÄ€õ:—˜m»å%úH×$TåjP·†u#Z.dÐ¥ I¹.)¡y„0±÷Äý9gÊS&OÕa°œ0=Ǧšj.)›l~Q‘·V?;ýz»èY´e¨N¼ùz<Ķà, ÖP]\òO5ÌwRyd!Lüµè4è/b¨òo«™aûê$äÇ äAeÔ½>LÜ PõUvNc{l®m²:–õ®ªŒŸE䬼u;È·?“†Ãåo`C…_H¶€/òâ¾ êZZK£”„à):p©_zP—2O×TgËxß—ðá Ñ$ìÐŽ<·É=Y+?ÖÏùì¨`PSÙ ÝÍgí±am¾tQ‘+Ù÷‰6jφUâVÚ¯˜Q±.¦ÂÒ?š¤¿éËK\Od±¼éͲk§z'ÅžÇO¢%F§.®È²'X†sÆ…Y&‘-V°~—e'Påó£ ‰–¢eAÖA.> ""„×GmÁ1½»O—vÇXûÊLì僂¯£’1æÜtŠ6úú Q|;Š×ŽÒÁ+w@yn­Qû)(›^T…í8% Bú'iÒ7/Š´mó·JJó2ýzÏ«úÐö¯-ÞãÑÒà0ìÊÊÉ{ä…«€0 ZâqÚâD~ûZtŸ`VÔ‡–mðóÎ׊ÏÕ0pšpŠÚ[¥c¡ËÚße1´+ÂLxÞÃ|Á3ÖÓÑŠ@ǼºVÍ­FÑx÷zãœû€üΘàô²DìòÁ:ñ±)Ñ çç§h;ø; êÝÜ'cQK\nÙõw¦î“Þøš£ØÆ59£Xzål$ôe;ƒ1ÕqJ ƒfNa>«e#•ª8ZZ ym£Ó¨(U*Ì*Ù6Œm¼¼ås$Á'½a¡V9BB¢äÒ Ò÷yf¿Rã;´cíhí¬ªº³–̦*cVÉšÊf“gl>?˜rª²áÖŒøÊù¶œµŠšŠGŠ ®FïOó]-b3P½ã»áßTìòÏ)f{ãÆ¢_”¹ Eå—»+KvD)ÉððQÊ)Ú¼P'Š-Xn:Øq{A†Ë#ÄI«4\ * ì¿-T‰Ãûî–øamÊ.Ÿz¿á^ww¾žfq B{ib7˜º+Ÿ5Pð‚Îbƒ`³8å†`–MÐ#È;Øç-&[4›Ã‰èÅ|©ŸëÙ¼µ bs«,ËBvgÖç,P_`7óHb0€J Ï”xÖD’#u«Hýäid$þ;å ªÝK/a¸ÊòÅÄ«,Zµ!b8É@õs¡4}ÙŠÎϘ#eÀ­…q’Ipa˜:•ÆW垸jxµ´¹}—¢Œ§!Ý›~ÌS]g _[írQÔÝ‚nø»ÇkÄ1iHM©ü“«—è‹i…à®Ù·²Ñ4u Âzª¢coå7@T;¯É™]ú%t<}ÔO¹ÖcYȵø‡º43˦=ß4zÒè¸|˜YOcq0vø -¯˜‡ƒO›,ŠX·Ÿûc"±²A™ØGAºa}ùÂè}†ÙZÑ?Ω‰bu!ü²i¡fds¿ÏS†ûÖ¹âúJË^>gâ5[*åL{±×3×ÏX=W—zëÐôi­B¥wß]o.3°ÌD}â½æàÉ$$H£|pÍᙵ?è};9'ÙcÅu@Ú0s µr}—jAÉ4ip[=6“A­(èÔ ƒÑôÖÙ·Ù?£ž÷à‘y¸/ùƒåØ ·W‰{0é¦q¸wÕ÷ðœ°kv÷™6x8¨£«¶›ùð^HÓ2^üÙB{Ääåp<+E²7ˆð÷<»Òº¢^ê†þ?Cs Ò*ìÅç+H“³ÜÉX#¥“‚W×ÏJ¥ÖMõ#‚,ïßyrȆæbû½Ñ~ªX–!™ÃBÓÅ$ç¿æZ¼zڻߔ5 Ÿ.ׇ»gKK;5çR¾ç+sfç‘o¾jpe>‹üfE{]ꉭʈ Ö¯ÚâM”qío ä'U£ä3䯢Žx´÷ª¢³iÀt>´í¹Õ.‰RìS7"V7µë<öüjœ;¯›Å_Ç¥T\ ³¾w¶Œuù3¶µòmᇒ_,¶…¡ý‹yd~k&IgÇ’ÔÚDCe;sª:Í”p²ëÅûoG@r…•m†f™u‰~Š.8`—»$£_Ÿ8¬öÀ£t¨Yié{Èg;ÚŸÓ ”ð$-ÀaÕ]{'e.Híg`¤/]ïòši¤lrå†^ót5§[Œ¦_gß~ïu –&`’ìG²îç“#ˆ²H9øÝÝ’é|ªó"©ê¦’_Bâ‹#r†„êplwŠ«étkÔ}Õ0Eë2]¦Ý u—/S.€Ò;]™ÒÙ >˜"C˜ÒI\]œ¨PQ®€„ÕõIZÄûÌxa¤¬ÞàÆaIYÛð|´mWlì Ï•dÄ=æ᥉¹êê¼ñ÷)UKv}ÔÉÕ‚ûKé5—ÀOt–“;ÒÞ&ƒ"ùÛræ¢ê²ÔÉ‚¯°m g0@w^ÌÀ,é "ƒm­Ú2Žþ+ÓJ12/¯¶mþŸÓ XÙædž°ÉAfãJ/s7Y,ý8Ö“3Ž|Ü&—)MF¯Ü§±Tr|3Cª£­ùì÷tÛÐØ¤Û×c„03Oe;:H&Ûq¥"û ?<Kjüð-Jà ¼nìòCü–|“{j] á "å⛲·Ë4Ö"\i‚!êxÏÄ­§?jºŠ&Vâe!¤›VÔ0a¥‰}0 ;´sÔ†¬W¢½Œƒsh,¦X6™ŸpÐõ±]6áðµƒr'®2i¡‘³Ë}þ°QK¶[ÿ”w3ý®çû‰OR>œTuê¹Ð˜“éRà‚õ±Ê{h“¬Æ³Œm§ûGäJ^7¯J—ËäÕ«Ïnˆ–~æÎ½ \Ý^þdÏ«†³Y„÷]'¨šŒ• \º=~ÙËÜ•оeÚ*c Jà!±´VÇ‹óŽp…ÛÁF“ ÄæÏ>ïàìÐ Y=ª8ÍS÷….” ïm#¼â1^TÊ[Mß8„ðªtŽ7DÚì±äîq˜¦9Bîˆm¶çeEW>ZŽ?¿oRc>θyv§ ´gÆ Ußî‘lY´óz"ÅãIÄÉGÖ½™“øZÙoØ`ô­ZÆËX&ãÍ}Ã"Û—_oA,Ñ3H/1#[ˆœkÙ*"ŒI û9Á1ìY–YñYS?–ôCŠž+=˜,M5Ë(Ü p•Ãz ·Âìȹ8Ÿ‹•pXŠ­ë¢Û¸Uô¬îô^¶OžŸLa躠KÞÜòÇÅ,Y¼Q+‚^$ŠÐ‘™Ê>1bÑmµ˜–·XoóéÏ»½Â«ïãñ)$à‡ ÷i(”šêPîxX '~S Öß‚òŒ"t .ð‚‘£M~m³EãR,.“ë2oðóµ&颌÷–ÜÂÔïùäѪÃÐþ\¡Ç?v¾ß7^ØŸÉÖyÙ`÷aØE8iâv/ Eà SqBŠ=·qPŸ>`´ÄMxÆ’í5ùËÄô"ãËJ¤~õX$šÉTLjTvAA7ì‰) !rž÷žÍ¥¢i.†Ÿí÷šýaÑ_‘»~ý<ßîkƒ}“¶ëî›Êê@êƒZm'¢Wê}õù;F”pÓ>ï)©“â$õPÁ/èsa€I} UIrþÙjç°k™H‹ûÖ#ƒ§ô ØsϪß7K”Í,C–Û:ÖB²Éb‹5Ú¼'¯¬Ž¬FMwb?¹ÕKçN¦ öbêzÐU]XaFþÀvžáß°°¥(j®^Áް~ã<ˆT”ÙæêˆÂÔ,š€¨-÷M^é«:µWö¸ лôvõ ò+èSœ­WÇŒ,â°,›µo+îã8›»PGbïŸÓÿjѸø$Ÿi°ò‡¤’èÇüd /àéá­/ÜP‡-Ûä@†Ø¾ËK$vU×dÕ þˆ¼{£©/$Âe}èøg­µOŸ2SŸ2TCêõWÖ5»uø‰`‚––EÌn©6jN•[WÄÕ¹øÂ¿BÆ”e“÷Îõe~µ:Ð8l[Ý12+Õ[¦­£§D{ÀUÜÐaGF'n‡a)“¹Töêz …Q㺤ˆYÈIš°Dú!2Ù ¹½ ~Ø]¾yŽÌTà$  T`Ÿ_Õ.Rß}5Z%V&C܇ӷòˆûubjDêœ}aBöO7n5É©îŽö0Sô0†”Rü¼aúøÇ­J ßìT‚ªÍ{[þ{ÐÖF›$Û‡#'5 )–},¾8yL†CØ<Ú‘:Âó¹$Žjú8xÿðãV7¼zg¼}Šø±AЦªá¤Z¤¬¦_y§âXø¬­Â]Æl1°YI J¶?Ÿr fDP…»¯€Tµ"Ñ(@Z™ meÝSghÒîFÑoÂÌ #¿“§0›fE$õ—y@’ˆ-#JƒQN™’nÃîºÞ±p¼°î—\ð‚”h”åPuØÚE ØaAµÖÇ‘ FAøÖŸa“ãûòr<¬­¤àdŠ¢¨ŠØ$©ŽØ*›1>Ë”oM ”8 Nަ$’µ=Ï¿8k˜™¼};½Ô —L¥±-jÛTÈdb§ýŽW»“NlÐ=º1Jï+ýH{{Ãö”š3̪} ] Ìäa–^ëË5å¢vµ:mPc ht|Û2ª|ãW‡¨B¾cgV.õŠ÷ @a¤%åîö!¦ttgU\ÜéÂ5Fø¸Ùˆù#Êgñ¬Î‘äl·Ô¹yt¬IÓMö>BuM¨FH¿AäÉb4”ôxCú¸#ovÎRó»ÛAENÊŸìËÒµpåÇ‚ &[%Ñ© ÁŸkïO%ô7ü žäšÔ“¨Ò× rÊÔñ#_–a,`v‰MÞ¡O‡«rB°Å~ßIøR)»WdG!€ÒS€?M®ý×ÛFV€Ìù1bŒ!ŽJ­\Ì:–ÇJ§F=Sç!©a"ÉéÓ#Ž*Ëëª6­&Ûúúç–¤…CGÒ«‡-UOŸnF¤E(ßNO½Ò‹{«½üë{D„÷™DeäA¬[³Ç‡|¾øn‰Ó:X—*¿ÆFa˱¢=:[]Âfå²ÝùIï,Na O·¡²¶XfÞ…Ý`?:wúæ‰{+Iõ‰¿õó-ÕÃÍœ>•¤ôrŠQ„aÅýzš¸×dÕ×+x¦…ÏfÙ)½ô~C áD®»º<Îà نI8…²1=°ª,ÝêÍÿJî ý—Qøh }Ô‘`ã@¾¹®Æº'Z‡„2't®€$ÛêçLD6ù ÙåZ =Wpùöåâ¶BåþóUÜÊDVÀ0õò7‡–5&ÎrëI†ÚnMàÇ9õœY|çÝ´ðÇîèþ+GÒõdÒ´$c€Ùƒe™ýñÆö­„ÉZ[ÎÈL˜cÓá`wób |–VV]\˜t»¥àÙæ±% LlŵùížÎ±¡yêÏ=¾ß PQ†üŸSßs_:Sr<œ²`'xK3oÈ25è•ÌÒH‰Ê8´–bÀ'`¾æ@ŸùüÈ¿èK=˜o®A¤ª¸/Ç’Ä#ýP¯ê¨T¼wä¾HÚ6­ YŸs’õÉmÛâÒÍÓÿÊd_6„™±P °?‘mΧÀ|EZŒ[ûeï­æ5DÑ—>§À²båM\b˼Ù\:“ Dufž±'kf4ÚsÚÑ–Žs5BJÑßu¬ÇÃ}µ@<ûº7Å&/åç¸?š{UN6Y¢ç2ä×ÙÞ§±høò°œåvˆj­Ï'ÝÆò%V"%w ¡š_r<áIvG­)Û®>–a ¨Æk|.*­LS5®Š¬Ó=vÎ& …f›E½Üï œRPËú*›ÓéäñS GÁ$Ë•PtkWQïH‡k„fŒ_ ¶wçƒr‰ž9ÞY$—ª¢Ú3]è§N²ô%Ô·z¸ÃQŸÇ½¿T]ÐJ)ÂÂdf<_<î+_ß9¦š×¶®„'#›5œýlà<¨R»GÖïtóÈx ~ãÈÞYñ¨ «ÿãgw’³8™zŽ æÈ v©ôLˆv%²ôEqö/2ÂÌÇRà/?³¤}=ËH­náIRBQ„š|ÃÀvz¤²º_ÝF³ôntƒ1xó—:‰H ¸ÅîÏP–SÝL5”üç5}£<6þ{£‹ =‘pXLŸS^:ªõ–ÓºKëF9Ìoñtµ²Ë[XwëyGÍý2'Îf\Z«¬î,Ô6{UúèOótÞsQ¬. Û]Ç˼=CÞøGŽŽÉ«F‰Êí#,Äø§ó0Œ‹ý·MfÅòÆÙ¨ûç0nvA0¾Õ«ÈÍo47E:¶`Ř½vÒƒì¢Ã¦Èö{1 /EÊs#iv “dY!ñ¿èvEðC?Ñønðù~Ùâ…¤š0Hç>=eÛÕ«GOQ4W#XÛ«fìix'ÎºÞæõ0\zuä]ŠK@«5ˆ³æÞeûøÜ$‹7•6 _§²…3âé¯ðÝ~xå%ɇ~ÙQ÷É:³ÅÕÆ¦WËÏ\rÍ×A6»LФÝZ²†¬Z+®'„ÀPƒe‹Ñ¶„vå¢kÿLíD/±¹®” >nU0-$¦•…ì7Ë—ÝÑÀTpÕÅ¡ô—9JQ¸›_`§²î¡Ë¶÷ضä8ºr=?©ŸgNQVߟè1Ååžü¨’±%™v5yàRˆåKx8hdaâDwÄ€ˆgо2¸•¾ -ŒûakHÈz~“gÖª*“¤îã.â%ýÑ¡ í|ÈÅnQÀÐÛBm3/ôš Àf‰Ï W¬‚Øz;Eÿë²§o’†Õò躷ÐC¹ÜßÚ­ø=°T—ÒåO þˆ¹ñ’Gs—)â/à£ì+=wì¹çŠJ&[É-$êÁû­ø‹áŠx#dÌãcÅó¨¾oÏŽŠ–CDÀļò®UlrҶѯ›#éJ‘w’ zòëw’Úûr8©‚wÜ_u"½. LŠ9$.2ßà7×ë17âÃÑéÝåñíÌ¢ýèÙDi¬l¢É6BÖÖY8O"NçƒNr$?hTÊöRo†NX#E³ïÂ~'TëÀ¾´ùKÝ~F0óR Ã.ÿEj8š/;4¸WzSk¿wyò{óHÛζ4ݾ{éê­Œä=£våk†QÒžv[ç€aFéYyÄ´í«ŸÑ;ZT-ŸTú¡—:-ÄÙ elž Âc˜œü锬…"ú™Ô¾µ}Ù’~ ³Á ÷Ÿ?Ÿ¿'ª,X ×`ߎvüdvÓ®èKVX&¤„÷ÖÔâó—5Ä ß+#ÌGí|Å"Ã6@ø'8×mÆêLöÚóâ‰Ó㉠|ãûÜK™¶×¾ÎÂuÌ¢- å#ºï’´©¶…DË ÝKOÏxñ¦ú+†øçl­zÌ„™¸¢n'CÌIÍDõy«"LÏ1Dz4 Ûˆø*’} N³Â·‹‹ÖSæ#¤2ø8Ü=sõ•Nwñjx™®·.'PM} SQž©LŽD;o”Zl"ç=jïéêB&üw±œ‰þæÃ«<B‘Ï÷Íù(y’ØŠë(:qøßùÞþÐmgÆÆòoZvê4x"1Í~¯GRŒÆ<¬¦k‡?f#ÌÚ¿ GXØ]8jiî—fgÏŸø1VÆM"hÖc\ô§eÄ#mJçÎæ2ýÚ¤?:T¯BoϬЯ·ä äADjË ‚¢õªïâ<î›ÒÚñ¼òtʲFöÿ,îB#â”üé8ÜÉ¡~;>]Ü¢§§býµXFþÜ×c£¾Ël÷a·Òå³"&idtž‘H0F&BÍÒÐN4ïø¤{=ã<ßòñ+Ñ\ÏJˆQõ»ú&“iÍ^½Râ·.w©¾¾Ö–9B ©ê‡ÌàŽ¥¯aõ-~; Šm¢ªîÝ-„°ok‚ éëãó`ìmh™µ;Ξúý¼\“ÓPÜý­Ï$£±RýÌcN|ìé§Œ@Vyþ³áSÂ(¯8ôöÐŒT¸+UÙ¶ïÍÇשâ°ò¢×Ðo÷Å OMßO"ÉLßD ~R÷"Œ}>ißrE·ê&•Ñ•«äóšº.螭fÉ.iZ >ÐB 6¿Ÿ¨ü^®©ë8ÇÚ™S˜ü¡%$fÿn…{¡}Ws˧÷,êË'öRu”sþ7¿$. +>ó§•d)ýÜ-iý–{R«#ZÀ­wä¼ëÐFÞEg ·†=N¨ä}Š5Üc#­”ô ácsŒËÒ‰-–I“¼ÚšâÅ{3»áÞ\a ‹»èâ9f ã.èùÎb˜3–v^©k8,CÜ/lõb0¾4mJn<²SÕÂÄì&ÕE´Œ C{j—pgëðysÛXwýÀ2N›ôŽX7×&·ûd¡“W˜WRè&çôÆ£eÓËÃ<Àí* ¼4Vdƒz¹7áò„´ïcð)mD®ŠÅ¸š¾Ó#CP•Â&ÀÄßþ Ë™ôHV»ಚÛg×­N:‡äá- T‚=´ñº•ç¿_ådy?ùú° ˽?ˆÂßY·^p:“Üý´·30ñê¹ÜæÃP2÷öùÊ<´›‹¬YKÅì Îí¯6Ï“¡£¹·ûÚùrÉ| ¯øŸÎ743‡\Ï:’•zVÄÈûjõܳ¶0°Ä9!•tÆŸÜo 8-ˆÌ’TÄìçÒ0°5nÇ4N0g³çYã;;’K>ï´{±ß5Ê•Pç¯5ž«a`ŠÚ¯;o‰äNy@R©Õ# ^}²¤™ß#v)ÚFPQ&Ø%¼{O‡Õçä,f ËR¡Â:‡]þA?{­ícÔKôca{±xÒ‘ÔaÊ™0vâüÖœz.L¬ÇdÕ3J‡–U¯´¯ÔN¦„ÙKoQŸ™CŒFI¦#!7É—¶|Ú#‘¶ré±[…1d>'Wná^zb8iž†,Òð'„=Å &¢ïù¨)oÛ‡òY!MvZÅë{=OÛ›»IëHÑÕqX÷®U´†äÛVÿb.í^UÍütéDšO´3MµÏYYÁÉg¾5õxYIåpEÈ„°Òû9$¬Ùëô6ÓÚzŸùêVÖCD ³{G.ô‘¥à`¥¼ÿ@þÓ‚Ú¦éƒ|y½•RUÿb!Ö >Ue‚[^‚ݽŸ!•½ÌsOJå!Ÿ•NOk»Ç·µxpSÄ¥û°Ý–›—šF_I‘p–{©Îç’‡jÊ_^­ «V`â•k<o 7µ±Û…`ŒÄ<óÖÔÅ­ŒLºå'9ÉÇ[d’k½>ÀÂrÀePk¼Æ‡2 ŽëÁEý <ëc×¹3–lƒÈµŒb¦o5¿?¼þyr}‡ª}–Êø8ÖŒBöxÎ#DE^o&LïÀTkŸð-åDò"ëKýIN²#?z9@c ‚Êjµ›²ZÈe6±Í¼[?]5àÅv¯ô˜¾ªÛVÒÀzXú¾Ì•¥1‘£¶EñŒfB’IÌœÏÙØw$GhZT‘n˜s½]ÄÊ#Ѭ$HaiÕ[GÛÑJO‰VmMÕH=/–·ì¯"å΀\°õî² ×ùó¦V¾=ÿÍÆÀÀ:ÕR£TRMz'ŠœšR8L á ’4æäæw޲jŠØêáøÂþ¬[!5œ¸æqüÓÏå]øÃ‘5ú¤¸³ó÷øÑŽYφª3}‡Íƒ1‹Mñ¹êæ»2"~*Í¸è ‰ e°P®›_Ç' nÖ™¼6( ˆNh!­ï?bcˆMDw±P.û”£¿^âD·0ͬw±Z3$+“S¨ s&ÊRâNK KI+R[7n¾²yKæ7 V©´0Ë—VÑá|î}›nÔÂSøZ°"lêsÊ¿8¢wðA+˜Ï¾RÔMÖ#Òâa \p&¯2Ý,+d‘8!ë(ØfRhëÒàooo×ç>ÝÀ,äƒí^GUcSÁ!2!©ÿXœ¤w§÷K›‘ü½ -¯ÂÞ“d€]¯P!Lš‡oä÷j]!Î\¶iâ/¬Ûa ÒWëí·bÛŸC*D€¹(©º¨K-Û§ Hœ_9¸JSÄìx&¾þD ÿüuêŠh³2œÓŸC€i0KýÁ¡Ó¡yö–´Ù‰ÎÎjòsOÿ§Té@M­è–=s½‰–±à—ßB¥{³*"j— K&£®E(®ˆòT&šqÒ\ÌÚEíÙéŽÈ´j0ÔîªG1Ó 5–ÿ¤fV+T@¼&ö%êP 9´ÒüÀ‘W;zeί Fç>ªôF)ôg7»{Ãͨð~„9ʧõQlÕ{‰] ñ‡¤¦Ž—Ç©¢ÇÔ½ãåÒ俬$šëå[ºÒtWÎ!ÓG ùº\aͳ“¿6i #MôÙ‡ÅZ‹ÃœJÔöÜë Âl£/–2=2mÃÙFWÙoÙꑽ֔PDfH=òƒÓ¦hoöû¯{H¥¬æ^GŒèfw!ç±¶çùCëoÑ^Œ$yï…RWzKôì® L±(‹Ã’$5…#-|ÇXkfÉnT3û¥óâöŽf;Ë@·ºÙBdÒRÝ«¹UÎìÒt/è=»×‘*0!íÛ”ÐÌ¶áæ§švŠŸ2eÎêV$áXHº¯,NÆ«ŸByT™RxxëëNµ¡ åÖ× ùHà¸ï¾D,v»¿$Ü¾ÏØ'‚8lÝWlqÇ®âMñ&¯÷ \h‰J¬|{G©éDð«<â½ÊÀ®¢Áؤ²É?d,O?Ò nòxh[¬8<ÂÑp™Š×-±ê+„îtA£t€ Za<ѧ«¼>Ãjˆ ÛÂ>ŽNRö ÀÛÊ5±ÛñÒ®_pž.¶Ú[äg¸œT­»úHÍ-/C‡Õa?0L}žJÓ‹¿ž… n]—Ö¤–†mÓûô÷ÝRΨºK7•ÓÆ%W8/MØ/”Ô&éÕAØ€|3MÈ|î+È£¦óDD# Ó"‚‰ØªbX7¢åoɆGô -B¶o­K·çV?¢."iD…Â׬¦;ä¬fIš!‰FÏE( b˜°KU|´H‘‡á¶Ò˜7r ±9&U§qÄj’ŽwY+:8 /¾ï¦¦z#9{fožjyRâ í$Q’ቨk^¸öf.w"w‘ÆvÞ§ø ¬ƒ^©aŽ?XBqõÖÃxçÓ@Ç“k!f÷IÕ>±`æ\4•ÿ8žù"Ô$¿Û‡_P<8Ô¸Õà‡ÅÕë®ìê'*"AàòìûáYšG¼’gnoe9T³9ÞÑ/‘{„vŒtàÙÉK<°œ~I«Ç;nó ã—|:æ} Pø¨™ÇPZÕÇÔ¯y‚}&™µ–Æ54× /UÀµîh¹¹#…˜¯èB>ÑõuZUöƒÇ]±#<¿ÝŒÁPµ_]+sÄ!G™ÜŸ!Ç 68¨S¤îò…4ÜjÊ1«§PÔð‹_¯ôß)o5!ƒB²vßÄ«c`ÅNÙ{yFt‹lg}@ÒˆdiÞ[„þáúÆ/ê½ ï-Ž–Ü83{¸fm†~Å] Ã»½ý¦ŒGCÛZã%ÖÎÖkæ óçòuzvE¤4Ö`èÛ§# H’€¹ÓÃr«ûª’jXzànáÛ­Jæª!y€á“ô_XeãÛœOЂۆêyiØ_¾¬—{ùÌË·ø}°Ï9»Å»™c1' N®<ØfÛk¡˜ëÿZ³i(öf÷”K«¬ß+Ãp¡¿³auÎ×ÞÒYà2.,¥ ?ŽøŒŠj¢%ð’ÅÔˆ²Þg.…¦:¬Äxbº-)ñ1öªåSZx°=‹*Ä­lϬ¸T Óçð¯A-t)çÕû ”õîwwL˜Aî°qXÍ øYjHn|Ç4òÔe1ì0ß••¿Àpãb5AJ”Û¤[í†>fm“¢Kû°î¯ AÜŽä ÜPÈ9¢$±‘ E‘5Ú5%RòS¨3¾¤1Î8«œÅòï&â’æh؞Ϻ³Üé¡ÇÂØ?!Ô)ô‹×Ñò;• TOÜ6N3Q°3‘>ÕzØ|ÕxRm†5d°”CŽŽëg7!‘_õ»‰­SšQPÞŸ©à•4;êÒ-<ž!L ÏWqz^GÕdð¼JÅþ¦Ÿ^äŽ ÅÅŸó­™f<±8´S|ëï0S¯ÄèdŽ’“.h² ¶¨çZŽ–1iª¦Y*â4‹¬¼ŠUyÕ £gNÐmü„í›"ÆíxÛÝÌÉ‚^Y”öõ벨j`LI)˜”dœC-±’A‚É4‚ƒz­<•SI’ŽåÐ!(_+õ À¯ì¾·¶‹êzU6­|¾2¬~†Æ}$ùå‹Îhžolý 4ÿ|0þ ysOû‰5Óò.ñ]-I6EÆ¢*…]K¸‡Ãý¢oÒ°/ÊíöçêGH·Cç éÐëW”‡~P0-À#äªTa‰ãê×#ÇR›Ùö”vCGbâ2»êGÁµ™¬ûSQ{Jx7×'éÉÝÓþ­K'Á¸ ¤&ZëëUgþLØpýá‹G0= ñΑ‘“8»ÃãrÆVšbÂó‡ƺïNŽbGO®…û«(ÕCøÉ‹¥?Ju¿Ë¹ò'´‰’jË–àÿ«á6’†_„}0;»=‰ŽÃ,fÿæ[©©E2ýE,4ª0SiJ•Û:‰<' ¢ÓÉ0•\95Ii­Õâàú)+ÐŒ`À¼E€¤‰½8޳réO¦[±žÆÜ•ôÓ/é7ó3ͳ¶B܆ĺüðÒ‡BQª!eè „y/ã²Ìw)å#:$_lYØ'Èð”w"8"µvº¼uXèT)Ÿö³Ý7ösLíæM²á1 êCÙ~(Ÿa°ä-(Ý!4”®"~¯ñ’>UÄ+,êÈW‘áàu#GvÀJŽqi`\Šz:>ÎÕ›|ô·µ5ÈÉ)U:jVeSFªGŸUQw|ú˜ª­WÐæPJ‰ü®XAveCclÓ•ñ¾ª×pâil>zCS2Í•lfÇûÓÉ/Ð  Z%è±²ýÇ©ëUÕõ¯½bUÆÎôÛ9*:Ôæ˜’u‡3…Xëq¾Ò"J>á>+ÊÇÌ2GÁò×ýžÈH— ;_RÀÎìIÔEÏxî wPOŠ5qý»wÐo­fxb¿öK2~Z|‚Ä^П‡S#aÃýB¨Iôщ¼Òé2ûÉŠ4ÀFò˜<Þ3þ cKÓH,Im&£ƒRΊA ÌB¶ª]Ûf¯’t]§ÉjO‡ã |vÁF÷"9óÉ<3·X:­ÀÚ³¼ /à]ŒÛ7¾ûT“Ø-Çâ+B¶{#oKÐIæÉ=|wÀ0É–×’4nîfz­µ-§·â…ê0ëŠ/Ï“'W¤RI/u†°Q{þ“)à!Z'!§¸°‰)Ø Í™j‹á™ü.«ðªê\¯¥ø¦"G” öFKJQñI×3©Ý…A¦² p>ãFöf‰³­° ½µ7÷6õS¬¯Ïäã:2‚l59Æ9àicÿJÓ) O¥éLœ$馴fOµ4M¯5SY ¿.„ÉÁ£Fv2 É¬¤uëòá¤c画IåÖ0ru¥ûï}jÿØø¥™Io”/pH¸EJG #ŒÄß]óíîo€ÒÛ˜Gié2Ù¹C²}²¾ §×•îRPñ÷լŵÊD)a¡wkƒf“ÓÅQ§Q•Û–»4Û;%ûA]TíÏùYqý&¹ÓÐNFAëái6 c|‰’?zÒc ºÓý[WR"«½ÙݸÄ=ATAÍTQñ½c϶‘5JòFö‹‡¶}"¥¹WýÆasØõé£Ö–è íeg€*UˆQÅ÷®»Ý—]©WÔUʬR¡aÃ..Ú§­}"·æ nÿHÒê–¨p.xß<ž0ç%ÃwŠÙÇmî¸ MÌi˜»‰Ø?£O= ‡`d'òñ™Ï}½ÕùNR$ûî YG0Ãyói+Çg¡øâ^ü Þ» ¢ ÙEkCa›¡DÅÌ”Ýèñý^I®Ù;Î^%c$jý²lÓ+âÃh¾¹¤k†Cô‰GÏ6jA–îÆ8÷ÉÓ$¬¬Ò–B'ˆåy×Ú‚—ß”…ÃN»GÇ‹ù¹ ›T åQžÈǸ|«:…èA:k“ÔO¤ ,Î_sª[ç?AðºjÀ0Û˜ Õ¿Yˆ×笙=@žýA )§à`¾q“éÙߟß4fŒ¤=ðB!*Šäµ›QÑôÈðã…øhQù™á|Rw1 ½µŽ{ê¡€PuÆ'F±GúÕpö³i€v ¿iÜD ¼ÚWÚ¤ªku?Í—ök…a6M‘ů©Ó"x¯7,ôKf'³ˆ;9xÑ€]Öö#v~#~w8 äÐÞÂú£0æÏ¥%Vhô§ ór2w’Å7ÃÉÃ+ß2œº‚ –MK£0ׇ¸—ÞHD0œ•¾or´â'§[vÙ¨yËÔ [ãß°ÒÇÝŠ—)®Ám?9^IhhÓØ·Bÿ‚OúÃØm'ítÕ¿G(Ö4}ãqi΄¶tvº„²‚o”ö"Ÿð©ÃkžmQÕÓE£6Í|V‰8UÆÃž!“xûc[á² ïõC‘™zaöµêQdƒ¶GOžžÞa»§.¦¼ÁZ$t êÔ]`µq.§{:ºÃÖÇšyŠ=,ÖIõtq$(ˆ°ÈëƒQ—O}!Æ#ØÚÐÝövŸ hSž›¸¢ÂÓ‰oÄw¾H°¾çØôUÞìÙ"NÒ:/5üúQK±2QmÔ»/:ÁÎ:¹!iêtä*ü‰¿á닜]§F_Þu¼Éšð[3?ñ¢&«zVœlb—•ª«žÀi¾_îøÏÀÈ ·¦Åkù :†Ë)ìAØçE'AšjØýp÷n¾xŒGü3[¨Lϸ4¬‘<ýìðDØ©2m 5m@&då‘„n×ÎJÞB U©7mÐyÉcŽî•6¿Ï;ñhUª íësɵÁé|Ûá"¬Ö}yÛâXB£¬â'û¦…L9‹ºÉä¹ø) ”³Þ6…õ{àdÁ7Ç#Ôô5,j²ë~·÷ÍË\i>!™Å—]¼øúì„ûë+»¼ÕlÐWPÓñs÷ˆ‘?)–‡Þä(/c–DðšŽì/w§Èk‡—Ü ™­øb4¤ÚžÜÒ¦Düë½úL~±£h‰«k‹òZËi?Ý9, ³Ï¡Û\m¾Jêõ]NS@DB^gMÑ¥ßèï©2.e(Y}_›ûû…pf˜¤Ð%^Í}™µTÿúÊ:lã¯êÑaã-uâ¢ÏåF»¸Ý†C¨fº~‡–y¹‘î)"!$Ã1:§:?s/¶Ch ¦œ Û7ÑN\•—m)‡ƒY–“ñu ÚcàƒœôíwØ(— ÐÅúÃQ«´P%Á˜§öõ†dÛÝ~Q#É‹+*YCðið,`ü¥1öÆkÉ[ƒ\Ky7å´½†(Ÿ¡©ŠÐAðN4YÞýKŠvü@ø |J|DíŽÛ^i®m>ÇGN!˜diX-ø¶\ù`j1tæ^q…|)à—H¸¶ž&ÝäOì)ÐpÓ÷¿;oÀpù‘¿‘°†ÉZ÷Ø©fâ~Óm•£ª¸¹a ä[’ýñr¡“ü&ß®­_@úMLÒ–à‚Ï®ºÖn9Ô%=Hõ}|ºû Š=ݳ„¨x"Îkp=Œƒ[¤Y¬q?­Z“è1Õ5md7¨ö~Q@PO)tõ —^£‘úÓ>„ “ ï\öiñO÷ÈÆ5w§+õòeŽãI©‚ca˜à9Û¾} œ؈88þ‘NBdÛ)É÷Gò¡™%V<ʘ8­÷r÷ó˜-Œ—)³.\çˆF¸ñ¼8ßrÖ\ˆbÈýü`šG#¼&PŽ3WaGnàÑÚñ±8eì>Ü÷:¾ŠäN4{ȰKrÁy“œPãÂà‘­ZBþs=ÙV핺ÂÚ”¥`_ŠðçÞìÿ éß Çd’×÷PõLºa¸wÝúM=šiÑEvqê¶_¿¼tƒœÅÔM`ד¬NVÑ6+Ñõ¦Œc¸”+°È ŸÆø£>jÎø¯ÂÆãC^?%¼ÿD$ã?þ&ªNƒšÄ'ä[–—«‡:·Òôx0³BŒktÞnø+kóª_ Á±_! "ko"0Í<‰tìUÓ´0 Wǽ×n£¥8Z)Çá䀙¢‡àc;wäIH< µp[_†Êµmˆ°fŸýCøieË"Ò¨®¢áñ±¸‹WR9H°ÏI'j…ÍbèÊ“[¨”\Ï6-:-ŽXÄá»Î]3³)"”6¯XW¯_Š*8nd¡ÍoÐÊÒâ\v«cq#'äë`Ží¯ÜeЖK¾æçÂ~®fQ  ÎBtõ/‰my É\¼_^[ ÎÕŠh‡ÎîWmAªwU’SðËÂtòxUÉàpt²l >øÈ›÷Â]¥0͈¨‘åS‘cÞlegžñú" ú ßaN(¦ûiÑн—®1º\ó¨¡å,Ž0“\T¿M\Ï×GpFõ@™Ìeià:"ƒýtË­ z-€M÷OÀÕC´©^²!Kà®–SÒ°A'õq<â«û?Kœ›Ž]YZ ¬á¡v‰–Xæ­üŠPrG ƒ„¥ˆÇ]Ú÷»lÁµ$d²¹E |oæÎx¸ËÅõó4ÇAsbëJ0Õçƒ`; _ †ûÊ8îrIîϼœýkõ3w?ȨRÍJŸ‹ðV$2°65oåáN7åøÄ.‘DއQ7ÍùJT+q¾³†ÐrZ«¯ï²“uë©Äñüe¡1hnzBÞ„G&¢"øç¡´?†ÀïYª$¤HôŽ¡ 2òçÀ+¥£G þªS$Z9Q£gAW…³—€Œð]`RÂnoŸM]al)à¿•y·«Ä™yÁ@¾!·Zu–*Ð1eê‚WŸúRg{‹pêËK Ÿº ¼ÖŽ¡pïÓkn§aŠGtY+ªI’S¯þ3J×<¯T:q»ßrþr@ÔÓ]{ì"„22 —S Õl4dž:H³Ì½nË3ýˆ7=qχÇ/œYó6š³ϥƆ5ƒUbíu°V©<Þ‚“»#l Ž^‘Šf¥õwú°jNŒŠgÍ=YkÉTš‘¡gú޽zˆ«r„x脃i‘‡HªIv lbc'RI)TëÂý™W;jŒ»¤SæïEkÄÜÈ7x)aÝŠÔ‚ÝñÇ‘ç¦q¾¶g€v*£æÂ“„â9Ak€L&/bƒFÀÇH .ŠA ¨‚@O\ K÷(ÎB%úÒ4 9˜7˜:6ÒâÝÓσ(À?¼¹ádþ Ç#j€ôßm—Âè"òÈ›¯'Ÿ#.›¬ót«} Ù—{äµbðárÛÇD2•Gu¸aKuXí¦³ÿ9“ ÜýûÝpÈ ÓT°â/‡L•©€´_L‡‹Ínx™^Ôc3$qƒó6¦H¨½$¹(wJ—è°oÍËQ Οe†Óᜠ³ßëZ¡Î(Ü&½£ö10TõBäÅ´Ó‡ºÿÆÉ:øþ¥A^·ËX+ždö™õn~^O³ej¸}Ç\[ø¯y„×,³Ý Xè _£å®€” \‰(2LöÊÀ‰3´|+‰ŽU¨LIΛ3 õ³7©ä‚«Tí¯’½"º’v©žÞ:y…½Eøi6I:5ŒX²¤qAƒÜ¶^x‹d*Æã b¤qKT©rßÝS*Ù‰·5ì¤aü;0³U<ª«£~¤Òé6ÔÛ³ã´+ÒåU+½ÎôS²góâc ¨~fsÄ@îcC&ˆÃ÷óÁƒdûi ÅÖÞ ƒtgø£*/Šn×Ã’¼$É!˜.ãÇ—b_Ò 5Ô’A¿°{Âzlýëï®É ÙŽû+ÿ’êÂ] ïºt•”¬¿³¡;½IímQüŒáÝz)©ÙSh=úaÛ~~ ®Í2´VI ‘¶n?ðO²&[\%ûM§mEh¤Ä,·m RUÚ‘¢þ;ü±ê‡1áÛfÈh ^ÍyånROt ø$HŽF¸‹I)¹V)ת`VqriSPè‚g»´‹· ~~ƒv¶J7Òø ³L†ò¹TÀÿЀÖAO#å€Yé“7„ õý—nG àd?À/‹ƒ°×W=>D0±D+žyÕÄ)´þxaö4Ì…D7 è§?8Õþâ·Ÿ?7¶×/:IëO¼°Ú±øãZ"<_ø9 ëÍXäL SÜLnÏb¶[R‹+꿼6_-€]ÙOÙÝ!~hX¯¥ïüì—x:i,äô Œtðãc<ôî.óÎãÑÕæ±Vy(gÉ…Z¨÷oSò0‹ˆÔYý‡Ú&ú䳜êA³‹VHd˸ gíu?Ž ®/Ô¼CÞ’áqõVmEL¨[*Ý› ÿ(@¼Ò¯Tý­6ÒK“j$= ŒÅ¾‚xOŒ˨Ó/€ÜÚÍ.ÏN‹v0Å!`y¥úùkݯëBõ;KDI¤?}jr¢Z~Èûìls±o ùiß«ÎqNaü]¿p$¾|¤L'x?Noq<Û-?Ç´Õ³5q”aÜÐ`z…ðܺ%†ÄŸÊ@ÚcÎêMö[ Jmp9|‰«‘•5™ÕY™q#ñ¹ø2†¹®Ì*>Oush|5ø èa¶¤Höw (' 速qÓõØÝ¿=“#ÈÚˆ¢ÆZˆÌÞš™%`›t¹ïlˆ9rë3¾³Í´ôIc)‡{*è$ЍŠ\¹oN¬6¸@!®Úâç,0Æ[[gÁÏø ñEù¦| ‚T;RH3Ç.‹¥ýbüDÍ03“X…]Þ~¯FÏ÷nÌõz¢ï’…ŠR˧Nù£¾” ñ¢¿7&ô¤Åc¬"{Ûkàáʼn…V5ñJ(ú¤`RÁ#ž hôΉ‰ý°ql›¸GÝRF‰Uà*o„îB¼Ùó F!ØdX±Ì¢´êj‹hµO?f¢Ï˜0Ñ}S¿’¥ŽY)÷ ½\È7œú f`Á8Á=»€)Χۮ93‹@)³ ÔF´‹¥ü¦üôQ%Ž"Ý¿vä-à'¸ØÝñ‚ã8^ÿÓ(Øöà®b¥ Bj ílkÓÔ%鬸êYÞ;l‹hþßÍœ,ê:ÁDõçûD%ûØig¤$HsXŒÕ(Ã‘ä µ©‘ ´Øþ9ŒD ‘æKŒ¿‹¸É‰î~!5£½”Ra\8¢á¡!oFÜ¥G†˜Åè‹_Ú0ù@0kVi.€ÝvpŸ욬Eçý B-1ð7&pXÐ㮫ZÙd<þæS¨(°¿“@?6Òk&oÊjY[Á UÜM³ `ç½Ðö{Ç?G\3«ÏiYÐиÑÈTÖ4ñEžÏ|PRŒü;YâÌ|ûí³TÞ8ƒ{Â@¨Ð„¼†;ï(RYº:Àû˜Òo÷®ê*‰ ŽûÝ M¥r]1¬Yížãµ*…®±¦ùšMõÆ.ŸÈ¹èKRb¼Ï —nCuž\yšû#d)¥x}Ö,*]0Þ>oòÖ| I¯í õÃ*FJÃ…¸îUSÕt›Œæ8wèSž;·X`²bòì;³TSá­>š³@4‹&uOgâªE¦ë樔w)ŽC™œ¦âuf®¦MÒ~u×éÒ T‰ù¦núáW¡å’$ô ¬ýd…ÒòÀYtu3ÿ˜íi4ùÇ«®±|ÞZT^ªû¥ßôífhy|àÉnéÛƒ…1–ø~RKõâPÅ%;ŽºÚª±[¾.©å»®ÿº=Ù¬å:šzù˜Ixj€ ÂpsYAÇo›Ñ¯*ùB^–¡ò¸€>^!uSð›ìïäâìUzOW¸ãç¶vALUî+þ!üŒ;Œ÷Ñ'EÀÅJ2îùëàú}¼t!ø¾VgéX0F;N}û;ã:O6Œö%WÞ¹âüa’LIP;šëÃZô®õÖÒwhcriîjÎ Tl³›¬Ž~ tAKÉÔ¶@^tÈò¬ç/B ¤·éPÌ+sO6ë3„yÕ3 ”‰©ùl£N5U$LäÁTPyúfN‹rfv^Èyòü-Â1ÞïiÕMåœÆ×^iV˜ê“ç¥÷‘?ž_Á– z̾D°(P[8X7¤ô ½çæówDA}=KùRù©Ú,:¿+>’UzRä0̇ìr™ÃDï_=!õÊܱޜr1sR“ËILp¶RYúõ QÃR½é+È}ž@M?–¶zI``ãA=‡Ž9Æ­Cì>aq€cô¥#‚˜¸¯Èem1¬ê ±¸èW\rë@Ä„2‰ŠØÁñ>ØÑTÐf­äœa‰.íȯU1¡wm”ÏÙœæžI½Ô\ÿøùøê:LösòNOÆëˆãMìy¾´_<À—šGŠm1”¨¼0’Á¼¯ÚÈ춪‡èØ®B­ßB¼"é¤àyzIõ— ’ó”}x£NÛŸAh´ö.Ô™>)Ýæ²9ÿcÃÑ1g:gü® «…pÚñeäœAu>¶¤MÔ,vt¢gRdNmdJ¶…„íÑözÚGUr'¢„†˜QX]-ED ~|j¢î_âj›Î÷=¯aÓ…iÛ]vJ«ì½H•eù:²˜ýÓlÒ:*ð Œý°¤QÚæl<ÞüÉJп&pלï•~Fh²ÈX| ´Ü1$m_¶Ÿý¶,76¬ÚæùÛJêw‘–¯b8QÛkvÜJÿiί¾ÑÕ|ððgëíõ-Þz“dUSÜÁ-T·ncü)çÒR¨?æÊs[ŒFØkßVH‰«0îû }ö>;¡‚뤂Y œì²Š,2•ð\+·2¶„¸¬Ö-¦ÅEh2…t2=èô*´7(Z†¤`}ë Eà»$Û‡û¯ª\6Ö(DÚ#¸¸š½|'©ó¸æRêèÃë±Æò® hi›ªV¥u’ó)ìˆË6 ­h\rÙ­S¶ìœÅ13ŠS\Ü­Ü*}H¸Ï²-XCóðŽ1Ü‹eå¿áË+BBOíE=ÙÏñí¥;ï‘QuƒÜpVÏi`Û0b´ Í~»ˆð-Ô(ÐéƒÐ?P.}¿ ‡eBU¹²#¾*“ô§JŸ:™IônÚǼòŸñA#ƒÈÖ¨"…½ý %õäÕ]• RÔ(9¸6¬[dç8^é=Œ0Y¾|Òºœñ§k¥LêbáÂjónÿ£"Ãòc]Él`³–gæaȽ5"‚kg æ4^(Ýr”†pR|Ö wäöïð‰}O×Ózœ™@÷xVJaLÍÜFhÁRÊñch°šØaMãîÞ9XÆíì 4FêBœi%™ .ú“W‹‚%Ý=Ýd¹DÌ&Ôq[eìLwº¯—±'‚z_kpì;²×ŠÈí@“‰jåvýtªéœS dkâ9‘ˆß¼cì‰b®EIlMŒåð7õÎ-yÉÃÃØñÌ‘~Ëu7ZLlÑ(ôXæ­†ù¯o|š¡LÎy'â¯!É[uç·IO>H \T.¹fT¿{ø¥»Y•n'ÄUô;“þX¨µÆå­U,!OÀ’5vÒgpü|Q.ÕUh\ÊùQåD ÙÑþèKÔž¸~EeE›Zk£#·r®u‡ø^ÄœZŠ¿û˜ n 8IÉöˆdݾF^M½Ó³.ü¤?"‘XŽËQx©üóæ°‚Á'ÉÑNÉ…Cölf·ÌvThF+š­ÙøÛô<Ë>;;ûa×£^fVr°èg!¿»1ÁÕ•&DŒU9o8 :Ûê‰aøœ“d•²+j`áÊCy‚ÜsÏ4?àq—«ÚeJÇ&»FdŸÈ¹¾dŒ¥¹ðf†ÇX”þÍaKA\ÊdLÿü‚”Ôÿ$úõ…ÒdHÒ¦¥vLEI»Al :+LÆÖ:ZÒ½åx©¤R…’ó*ÑÃCü솠>´ž/Ž.^¢s¢XìÚ=--D{¿’ràUñ×¢Ù9‚Ç?‹:î2?þGúö ÍŠ«˜ú@°èržªñÚ ~Ñ&nÜ1$Ò:X8¢pÒþší^Û+xç·¶eúµ§¬Ýe!Š=ä0ÞÄØ5Ó Å>ÜLS’Ú F,°}PÂFù×™{SséJmŠ÷Ý©†ìP|¾áœ½¬y v‹1(àb<{I-?l/TUXŠ…“Yž˜|€¹…P«K´©X%þÜ$‰oÌ¥CUæÙL’ô¨m¬oh Xû{?KB´zœ3‡Å»é…ÆÏøgŸã¬2ï<“rkA¦ ÒþA"wX5šE¶»ž(™ Ø Hd„Üì‘X:,YõÿÅcbÎDP:[ƒùmÙ Ô¬L‚X£¯Çtï]öÄî…‡®;äYŠÖ“: ÙH´¸X5+¡h È!›&1šXl/›4ÙAåc|Özùo‹$tSv!ìcìLÞ׌þJÛ&›T~œ¾=¤î*ÙÐèã·Ù^<îä*.DdQ×–Å;Ïç1?ËHÊÉg¿)º½ àÅ{ò¹Öy¿ÿ÷Nû°ÁÄ »Çß´üÏ¿òA€ðŸLå/±°,« ÒYi²*ÂV2ž ›ª{a¡45›) UÍ^þQ™ˤqçŸÏ8¹Èe4(Æ9©ú+ű½Š> ­ø´š›/ÿfè!÷Î]eƒ¡ û’ªcÔáP¡Ú2ú /ó8˜h¼Të–+üIõ7~€nÚZp>WÂU»°ébãj NÜ"ÒpY- }‘\ÙèrÖl>œ¦Vf™®ü~yÛ$×´Ôa}·Ò ÐÃx…Šp¼ÑÔ¦£Í õÀ¢›Â³µ,QMÆÆ§šÝL ˜2¡9Pé¤ãæaÇèS°M :"Zˆf‚Mk—™ËU­?M¢lб"°¶Ž à61±¡˜Ä·œ?™ÛV„ùbHýhûuenªmQ÷üEä\ç«’+¨ÝÉyBºPá5--€G;¡OèU@œñy×UÞ']'H%ô4Ø¥èz/|‰ü—Ý8hÑŸµÎÝ(ö’Q['²FÙyÛp E’tç>?Dgù7›²U$9ßõÒH· —?^•Vaè9O»oœ3¡tÔ[ä F"ÝA~¾âû¾û÷B(ǿ˹²§Å¶ÞY|rYÒæ½ *ZŸ¾EtÉ·Š/aeZÎ5SVóÍvà> stream xÚ¶PÚ¶-ŠÜÝšàîînÁݵq‚»{pw'xp‡àîÁ î‚;<öÙçžìsÿ¯z¯¨‚ÓÖ˜s¹h R%U3{ ¤=È™…‘™ &¯Â `ffcdff…§ P³r¶þÛ O¡tt²²ñþ#@Ìhìün7v~“·d]l,lN^.^ff+33ÏÿÚ;òÄ]­ÌòŒY{Ð žBÌÞÁÃÑÊÂÒùý˜ÿù 6¥°ððpÑÿ+ bt´25ä-vï'šÛTíM­€ÎÿU‚šßÒÙÙ—‰ÉÍÍÑØÎ‰ÑÞÑB†àfål P:]f€¿(Ûÿ fiåô·]ÕÞÜÙÍØx7ØZ™ANï. 3 #àýp€ªÌ'€¢ôwð§¿èÿž €…‘å?åþýW!+п’MMííŒAV €¹•- (ù‰ÑÙÝ™` 2û+ÐØÖÉþ=ߨըÊÖØä=à_Ì’"Êã÷ÿÝž“©£•ƒ³£“•í_-2ýUæ}Ê 31{;; ÈÙ þ/~âVŽ@Ó÷±{0ý}³6 {7׿¹ÈÌü¯&Ì\˜ÔAVŸ]€2âÿy7Áÿ±YÌÌÌ\<ìàgÐÝÔ’é¯òjÀ9Yþ2¿wàãå`ï0oèce|ÿïådì 8;º}¼þéøoÏÂ0³2u˜-¬@𪿛æã÷Ëw´rè2¿kÀü×Ï>é¿ËËÌdëñ'ü_÷ˤ&¢®,¯N÷wÇÿñ‰ŠÚ»¼ØØ ¬,nn3Àç¿«([ý›óŸT¹=€ço²ïSú®ÿ¾êï à¿k)Ø¿‹ þ£q=ffÓ÷_,ÿÏJÿWÊÿŸÀÿªòÓøÿ&$ébkû/7õ¿üÿ·±•­Ç¿Þ5ëâü®yû÷-ýïPMàß;+4³r±ûß^gã÷=YØþgŒVN’Vî@3%+gSË¿Åò·]ý¯%³µ•ì¬þzU ,ÌÌÿË÷¾Y¦6ï/‡Ó»"ÿå¾/Î)2µ7ûkÃX98ÆŽŽÆðÌïBbåàx±¼¯¢Ðý_01‚ìßSïíùÌíáÿºQN“È_¦¿'€Iôâ0‰ýq±˜¤þ V“ôÄ`’ùƒÞóäÿ “â73€Iéz¯¢ò½WQýƒÞ«hþñ¼#ã?ˆÀdò½Ÿ`úÄÎþŽÞß…?ÑÍ•Éìð½à?à; óÀwÿ€ïå,ÿ—¥‡ƒåû“ø'âÝfõøÎÔæðªí?à;W»?åÙŸRï© ÷kÿ‡ÿªýŸÓß“íÿËýNÝáû½˜Ãûc²š;ÿ±²üÛú÷&þÇü~ÕïûgÿÁ°¼÷êøøÞ˜Ó?à{ÆŸüwA29»ÙÿÃýNÞåd}÷»ÿ¾³ðøsøûP<Žgÿ—žM]ß©:ÿëÅyûÿàýûݦðË ö¦|ÁÖuÁ÷5"n {“³{ši4 ^ËŽ.ȰÉ4ÕYŽ·"É#½¨«;Ô7Â+$/^'­ °am‰ÊíOÞφñ*3{íðKÓØƒSE'"õDp„ jÂûÞ/Ÿ½5l [Á»d)ò>»p#+`Ü»õK¹×”ÿ]ØSÞ¯æ”Cx.ÿÁ­¥P2G‘o’=ûÆ™è-úow”¹›ÛYôÜ©7Ùx:xŸÓh¶b/MÖ˜‡yϵJ5V§nlpƒ¿8ÞÁ,W/©®ÊãÎB+‚“†yh†ÛšpÁ–šÞâ_GÛ’ y¥-¸Ò¬ê½šø~u®_ÈÀÌÏXn¡*Š‚¨`›„¡“£÷é|6?†ƒÀ€§¡ó•·D®Úû—$|`ð›W_`%Ì¿|A‚~/çésŽ`§5³AªûIJòÀ+•VßÑ™ßQHD‡âž‡ÎˆÒs ÔN¶à[íÒªÃ1K€ªF ¸Oìï×îõrã+3fËþE/??~;˜yV·³ñÜ”ÇP¬ ŽA_ÈÀ”ás0ø‡ B©@Kÿ(ß©³«9~6Õ\¤³® Ù>°œOÄŸi¿ÜtQ„ìð»Q±ä9\(ˆ»6Eå f1Jõ7¨‚:ÜuÎ%`{Îf>1f«~I”¿&¡ ýÉ?ÒõËõEiÖÒŸŒŸE†LÑpÏi­¨½XòÎÂ¥¦oç ÛdÓú'3Lèý[bBþï'¤\;¥ wªÇXJVN/#óK”œ»©gTØh+¨‹ "ùŠ)½vØÊ+ä âZ!ŒÒ8î?2µ€¿L ¬¾ŽÊ!ö¤£¥r­ö )§Qbg²CôëôÏѤ¾ÑáGØ)÷Saðpn-ÉpÁsïQ‰gAÌ|$Ñb.¼mÉx ;ž2[Ó=JÁü$4”`QbìÑt.òµnÆ`·¹Ü&ö†;ƒƒñÈ*™.(ê»,´{5"oOáþêÔ¹¶ÊÁ”ÅÐoõô‘ó‘­ú#–'i¹>U˜£Ð“­9ÉÆ½wÜüí uĵTÏ`—0t;"È&Ó×óx¶9P¨x+Ø·¾#›U/’K“3™V¦ÛÓ^‰™A¤Ñ'ðíqËqƒåô»E%Wý®åƒrˆ¹£­#.*…a3³¬Ž#yÆYæMÄv¼|‹tÕ‚DWÚ_Œ°‚O¥GìL}ÓÿTˆÔök´ürå «†âáOF̵~®DÜÇ&·©D?¡VÈb±y)&˜\Û6™9e7²|•¦Hð ÝI´WHËÑR„Ý_‹6—½3æúˆÃÕ‹ÞL'¾m"”ÚÎùm%rþ°Ìà„Ö…c]ÞhœÎ#&ÑÙtU¿ÎضíPò¦ ÑjA¶ûݶX‘M„Ž.@Ùa P‰}-Óç”lHT8^—íã·Œïí¥XÍÙß3š*šËe-s‡€æ?œÆšŠÿRoŒñ\Bø–&ÌÅŽ ü†uÌN@ØÁ(þ 0Ab”óÃ’Ê¡òÇ{²ë„–OhЗq4©64q"ÞÔ@ƒ¦ØÆüÙÜ8ZÊVö ‡%Çm÷èç- $q^¬%J¦ð…ªáx̉˜­̸ÏJÃgá¶™ àÛ:œpmûëWؼfïH7ùHÇDSLž %,êÝ~îvØj¾är®!i¦ÄÂföT•iÜvŽïU³þ0 {êX‹ô’„Þ´«AQÁ•‘Æ${ȸA ªˆÓº¤”UÐq„{ä~Þówv‚jS65[Ž]Bj´…åA‘Í.“šFBø(ÊÜD×nRlDþ¯|fÅ0-«¶£öpø­¬/ˆ‰ê[W7.ÝœÛib _(ð´Õn¨T®£†ð­€åÚ/°É—ƒ|>½ÆÒýNäU|md7º:˜ÍÙÃÃao‰i¨…GEü2>L^¡m<_#'ïgÁŒ-£œ„wſΓúäêÍÆ ;'µ¼+ìýY+4Jpk$ÖA§Š¢†šXëÚ5ytŸË‡¥ܶB¤Znå›?8 g1Ú‡”ýÄTºŒÇÍZÛøkCN¬+ª¼“i¶U#Z5Ä&êoeú9}(œ*K º=ÔU§"²¤lN¼Ð@Z‘‚ãÕÂqDh×òa §…™(»Óéò FÆŸ‚ˆa½á³—Y‚Ê]a”­)CîßöNÆ[=AiÈ]á¡6b¥5æŸ5–“_‚8?GØÃÒ¦9Ð1ÞZXÛêä9[:ÄN ±ñ²;/_,{PH¦“ºñß~&9G>«i—d”ŸEöÂ-ÛâËa=š$|ríÇÐV˜¤“MÜ©Æà¤Ì#?Í|ŠãGð@?&s@¹›ªá][×0ªí&#€Ý„áRV …qïÍåzÖdKÙ‡ótÏ}4À©zðWA-ÂpóÖ!5@ùî&a¼h^ûã×H³•Ý[V4;;“É7…·S*üô[H¹Ÿ½l{h—GµŠ–S²"Céõ‚%è<®¹³q´°T2vÎ-³ßxÍÈÎS “>Í쫸çõ‘†<›‘vª €'eú…g4Ò^äÁtíòìçá¨~7ðYêÃO0!¯yÔ¸ -7pNa÷±äƯK+„;öKWÉÓ+6ÃrRmãmM¡ÙÔ“„f¿…*(㮹ƒÁ*ªå|ÐØch§n&•mlÖù×YJª_í¹¸*¼ÈØä_ö€…Èq6Å›(þ©ÃÐqYzï¨ê³E`I5±ãAI ê¡ÅDµEâw@¥ÙqWJ·{ޜȋéÉ@ÚhÛT~Á³¶Ùu0-oν}´™QxAˆü£’IpT(¨(>ì¨Ú*AŸ¬±8bÐÖ‚rdXšZ…r˜bÂa†1C«(LJT«®fùˆ·fëÚÅÓN`Hþ0‡7Nî –™1xB„J7NÝ"7ß±¢ô¤ñ%AÑ!tÈÙT̨m>o›93¦¿Îÿ7G¥¶õÕQB8‹l "~5FÒ¬À(HHÂ!&fËø3ZA¤^õÃ1͘!û2Ñw9íüjºF³AÓu/œƒèÌ:ƒÎôÃ/˜x4•V·@?€"I}Õl.í+·(Ç+íþ`pªû&¨S˜¤Gâ B\éM°-HŽ)sóóÔ ´ÜmŒ ].ˆuŸ_`C¶ùÍM¨Ç¡o«ám"±ïe˹1›èg7©!  c¶ß̱ΫøÙ 1a‡>ñ‡ð¥>ÝRS|g¤ØOÏ֚Գʩú·.¹‹(BzÊù€•#’=üÁª¸ÕÝé–O¢5Sèd²»uà¿»¯^û.Zˆpp˜±&˜`“Z!²èhñr\:ŸÄ?&_›ï˰” ÍY;÷Qpó2QGÉäHåç®À€6rFAá…¥!ñÑåו9Þî³Üs–¬ÆÂüªÛyºö<ûd¾“B}lrDJ)ti)8ˆD¤°Ã¨;é›ðYăG3|â=ˆºcª\ù6î€a¼”ò…wgšÅýa(érx×(F£;ù‚RB"7@¨ÉyáQìgêP:WƒÑÐ29Tu•¼ëì1Ñ1¶w}¥SÔ$â7N-Ò~ 2ÙT˜n¢Únpˆ:Øl9&Õ~Ξä®Ä0ä4Éžy‡­½"¼–p©‚8Ì]>N¦Ë¾óºJ—ÜL\íXgõñGŸuóf`¢•·"㛉*D›Ó€(¬gÜÖBÆpic«Åoˆ¢…gLÎRÔÍŸ&°¦Ži$­—?ã5ñYýD>n€´ Æ/XÁÛÙC@ñrD=Šb Ø "e[<8gšîãÔÐî ýo+9~Ä5=+ÎÜ8¨> ¹× 21â£ÁI i]Å$]¤q¸#ëÐþõ}À›$1»ºã—%Ô0X‡íT¬øá¾ö:²c3 ›o' B·GªfQjbyi,€}ƒùiø†T޼Ù_ÒÈ”š®i €´Ç’¸Ä‘ëÏ]OÀ„ض£Ñ¶S]r½\ÒÚw˜Jk éƒOB&„¨D°ÓâÏó|o×çký;)+ფ+¡Æ €$Bî÷#*¬•vÁí`̵N`d.F4QDø‚÷>:%lVçòº¹#ÊV“¿r‡/¨™ð‰#B}ïŒI7¢“W71Ð_QÅX‚ó,¸'Õ?…Pì öXm­4VMo­y„­²òÍ…«á3Òiû½µ-|hd[.éêo<Ç„àüh+¸#T#—~nÒ:rüzS¨ÒuõôgÅQ>_ŸŸ¥(-h÷f‚Á‘ÃøF‘ÇSH,Ú¾ÃBƒÛª?Û»¿DæÒ¡ÕT¦ðè'/UwáoÆ’åT‘J½mXsµ[ß­Ç¡¸yèkÏ­ÿ©_) $æü¼jȺ:4górƒ‰Ay2pĬ•âZL€.–I¹êdV6ïaj‹¿…‹|"ÁÛ8ÏÌ’ªn{³Æ$JNóE/İüÙxo¤¬ãÂîh|m’3Æ—å©D¾”üT#lÛ7ðyÔ5û$êÒBí–6<@ ˜—’¥E9Ÿû®Íìy(æ>hßUl¥å/'¬ œu‚™vg^Ý¿»öĺ^SÔkµà˜Ï5JAX€á93Á^ÌõåŒ{¬ù,Ø¿6l¸€ÃÆwž÷µ3[ÜÍácŸõEÔcܤ‘ªŠ^“6r-ã7…Mœ±Úx›=Œ ßui²#jnä)®z€+Ê“ 1‚ßa"³¼9¾×R“$\btª›4ÈÁG¼–ù öñäÚãfר¼(È¿4K„MHâ6†Q5W|z/ ÚtJÝ[ñ²”®Uˆ¼ÁqF`9,"î(î¥á€Ãh<¹~s‹iî·#‡>ÏèÒoAå,…E)+Ø Ê»¬x‹ÔP¯ÐYC”átæ9é Aê±ÍÎÙbOïLµñ† ùjsÚÂ`Î^,ØqÓƒo+ü©Y*® nßh÷äQ–bîRX5 ¾"Z… íÊŠš‡äáeŸ©|…€µòÚ•&k`·D Ï#°ðí€çþΰ³xÒ9)|Oísš‰û±­–£† Ÿš¬TUhtî0žoäåõJ•™îŽqŽE—‘?†Ãÿ‹ßy1G!c¹g|ÃVFš(oÍ)»OàÜ ˆR3¾=…sëWÄ";Y^F½XÓUöI7½¹?Ú‰ÖÔâFxÉ×ÎÆÜµTŒ{mt#8 ‰GÍAG8Ö-h|²$8ÏŠË¡¹‚µ)9PùSE=W]‘C7R3¶™ðŠ~Ï›¥ÐŠ®]ˆƒÙ©{a*3=”Û¾½gýefóê™mQiªSðúß18šåXQ¤ZœsÀ ­ö+&?§{‰ìŒ3êŸ><,gÌÝ»NËiáý0w.Ðû×85NC…RHázE8b—²`y4”ÝvòP;yã]°zP"]=Óã—\b•y…ŸJЋeE[™onÑn2U| íÞ8ù‚bªà\•¶i0¤ªŸwö”p)PêÌõ)büb[÷vpmÍ;Dl`·óÑGrC¶I|š„à §šëvÞ Ë®41)Ji É&ýk2¢žèòqÈGâX÷­ ÓËD³=]¡ÎõÅŽ†ôMéýf¢Ú‚ iƒF¿íÉ0·v„Þå/êM ¿ÂØ’ü¼Š¹²kÓ€/£/VÓí¸‚)\^0>§pŒYù¾Ü<)@—ÇxälØøƒžÊÿk¯ÍÊÕtIŽ#í,Ú»fˆÃ"y¡ºCTSéL hÁ 7*ÿóJ^z¼±‘â¸÷ç³4èüõ­I¶#|ÈÙ¾6ò­c¦+ïò*Í FQ›Õˆe|œûòÞ¹}¾—þwÚ—rÝZÛKjûŽå7˜¹ üt[ ”S·Äæ–ž…`:ÿ(?‡k³•z-[ÌXq-Ì‚)„VbVy`Ò^-pqñ~ÃJå§–;w«ÈaªH=YO+&Ï#Ubv % ³2½7ˆÀzêY‡÷\)±Ä±ºiùZ_ͻƅ9ÝsSS>Rõ@ &.ò D¾†ôÃ&ó!óµ?úp.šþ0ºþžD¤pA2{ɦÉDïfç°ÎUfín¢2_}ý*\¼vîøêÉæÞ)¥ÀR°)Ã($>×8ºL©†ÚךBé[hËðˆ |£”¦mò§€!ˆôì°ÝÆß_o¡ÖâË({"-]nÀ(;‰EÀCÉQ££ç(cxLõ”?JQ\GØ.x¹i Ý:…éîÁLþ æE•›J÷1qן s¥%¨\c[:2=áf¯1šEj%÷NÚ=r[¥ãH­nâGŸSCŸƒjLEX}]Þg^ø@Hw )Œ¾Ê¬—§X-dX›‡b›êã>ݦ™Ñ³ÑÇ%@Ü M|– «H¦à%aAÝÚ!žŠçŒ€9ZÄQ ͽ…µÈZ"•{˜‰¨’õN±ÁúYR”¾|×ù=»%YÒøZZþoµLªuýòã<>ZDº?UضEXé‚ÌN>n»‚IrÂIÜ$èž3¢äOÎê((ŠíôXâbEz“ ÿ–;äøñ…‘ƒZnëkV64ˆŽ½Ýc^¸k:;fZtþ—®6IKGÒ \%ƒÑ3ZTrDRüÅDëà×X‘ÔìBò°€ÐEZþ:¿~Rjw"Jr—o¾kQ ɺ!;î\7 ;˜s©-wàÖµÿªä óÉŒB3¬4ñäòs| ŽÕЉZ‹ziE=¯Õs_ެv¹a>l ^l¤Âjê¯uè ‘Ÿq¼½â¢ÓÁ´þWq¬T>¿y“Qß<¾ÂŸíuÒP룼aQ×¢f{±=Á“òG…E†§÷ÜÎS³ Œ g;“….Ê[0²X7÷гAAjö*åÃK#û9Ÿc_ÃO%Ö¬>W3¯ŒBš½f‹|ªl„ŠÃ3Äut¢§Æ úì"ø ˜!à„œ¸!ëO&Ÿ¾‡bC×wsÆ–÷¸–ØÉß »T¡T!yiDvlņ¬5ޝ\eïè+ŸZÚ]ZE/×õ<:ãAzsæG‹Ò¿?©¾HN}"@-[6¦g]d¹X‹UàôEÐÌöéõ-eÖNÄʼnGQöΧ$9é…êæ¡·¬}Ç2 „Øw $Ìdê&Ø…ì°Œ†Š1±4g\òßö àhòì!€·˜Ç*0· Ð¢§LGVƒ&ÈKñï*è•9ƒõ]² >òæñÔŸõW6ñ}–ÏN]Kn†ºÿ>VËàNf6-ÚBMÚB‘P"™•&•C^ÀÒ·˜xS§ïGQã´t°¯º\¬žLaMéŒûYo­;ÀÝ V­ÉW =’a‰NãM¡>|ÑkÊ‘PÈ3˜íîÔÒ%ï„(QW˜Ix‚ c¿Â…‹IÌßÜÏùÒ´òËåkº}Ap…P{ù¦Ã‹çåN“Y×#{›ë¦ï]z` i² ´}Îy=#]¹›ìÀèšnÙ8LΣü1U¤kÑÚ~ì²=ÒeYrœ>É#Ôª¾œ_Ã&57#.ø$’ŸŠµ”$nÒ/vOä;i^\ÿeujy&[Gç–{ôI<¦Ä·  ¡{«ésþ¤ £r8,‘êK@1µ)Ÿ¼MU‹÷viãùD1Ñ*·{—6‡H–Ø!ò ¬;¶ûGy°u­/÷5ô2—h¿]Ò¨2¤ÁmÓYÊØ2–ø£¢ñU|ø†–)*¡FFßÌyßÝA3H˜Ú?í^|kj"oMŒbÞ;Ÿr¼ÛŠÇ¾TƵ¦†ºÐ>¹æ/ôD¯ÛòÔWLßùuW´×µ…ús?pi¬#«±rykÛYÉ< <γíÀ(·w†¡±œŸ|K®êrlfž´íY©x|¨Î\˜‰tÍ!Y°ÓAüS2 cçƒzµÑyÄêOÓÇÁ<ÜB­)²ëqX‘Dî|š|Ï(*bÏ G’ÓeeÞ*£YǼöIW¦9Öí¹š9 §v”¬tMv·ÅÜ<+ï±°p‡HÁ×Âtñ(‘oœŠ’ŒŒ6Åòãø£…Ÿÿ½Ž’àÒ‰7ÝüØË\>,O•Ù/E¹¿‹}2ã»@VžûÛº¯à⪒ç6¼©î¥¡g„Ê$¯æÓ,â~ÃöÑn/Š¢SbÆ­Ðîøràí…wñ¯êŽ‰Ãˆ)‡T¿HghÒŠAªrâÏï9DK¯LSÞ1¿àLøßåPVèÃÁËEX*Žg±Ä[Ø• ß@qƒAmoS ±Åˆž«ÅhÊ 5’•Cò¶Çà-¤R/æ–‡ÖûÃ?J¦ù¶@<éxÜtÄ$tØúS6ÍÇ M5ï)¸ÚŽÕ5º¢=Ð{pV`Ì‚»­+~I¨^Ï:g@×^‰ô¼ãÞ.f8DÞ4Þä×Lt†l›$]‘ 5FïÔ}»šðF‚=,jäD)âi3ÓÀ؆îV©­K–雳¤œkìó+5FA›¾P1”XÛ³8-Þˆ¸f ­Êð-?®G?Ç oÎûâx»ÂCĖܪþ•1 ¹çëæïOréØ\ÿÊuàýµ–¤vçoúAxÑ€EîD½ˆÎÒZèÀm>ªá>׆Ž%»¨bkoJ!lì°Y É$"°¾àUßVFL|+\¾G¶­—ÜN¦/(šcŸàØTFm>æôX|HOo±îTÙÐÍ_ðž u½²Š‚¾¸,•~š¬ŽfWóç}8Ô5Ãøj 3;qÏ멟vû7ë3 _ÞÎräMÝ÷Y:cÚ"ÇPŽ Ó$w|ü¦ûׇ*’-A.Ík¼…xâ:»Žü3'²>äUž„HÔ•öŠëû¨.*E˜úÿÇÉS”køûŸªøæŠNëc±×q¶K®ÝYœaGì¾WÄÏ+´ëC¨<QþüññðJî»$‹ýƤ„ÈŠ¾Œ†sò«‘}×¥GGzžYœãwnÎБôˆjSL­©†ÜЭxQ^DyH,- ¶ÜÝ0œÖÖÃhd=|?Êò/Dªµ:h%I|%£øÏJpqÒ¢k3÷¡³‹”l;vÁrð¿Aò­TÕÔ…õŸÂ$­©—ªÊ÷<é€HÕóN­{)n!6zŽÏ¥ë» ñŠ8½‘»ºGáÁ‰u*“_j…±´ÍìNÆGœzä·P “¡Á4•¾ÚÚ¨0:q­¬;û-—tíž@¼„ò}*Ê—(‡^67mZ‰Еòªü°˜->ލNKi `þq R«8üüÚUR”EÁ!¾óM'N}t}³øso!…îJŸdh9l—nÇÿJ¾ÕµšO1ÛÀ™L)'¼ ÓV¼­kÍ>EGÌ`O †ÜŒ ËGânÑ­aĉ:ü’)*ë2!>Õ9–N¦©¼— 0§Óæý†¡r7™®|“1 Ã„J9›:Ï×û-‘)°)F´ô'ò 4,'íƒ<˜{©óGO‚(¿ž¢U%ˆ­ò¤ó'L´si‹—ýk„˜\Z1ž¥/)×¼UÁÊÌ87 x—²¹zؾýºûé¹ÎÝx«.F ¥^ìwQœF‹öÓ˜@œ^†ðÉHŠ~áÄ@Þ *[Jáho¯9ãÞ”öç"«®rX]º`”ÝÎ÷-[üˆ?¡ÞÈÄSgæáÙb[©¤¬¡@¨ÒÄ?ž(§ð¢”å›ÜØüW›Çò`üºVpêŽQ›V¹‹²±Of£ÁÚXÐ’{OBû· İbõ?^³gvuõnEyÖ1‹À Ú 4Ô‰iù< ç‹m+2&pÁî6=´ÎƒSQ>¥Ÿ éK'4FCB¼VdxÜÃÿ’jÊlW‘g.§õóa: ¯üÙ«Ì0€E[âóç…nÀÅ%Õù§ŸÄÓ8¦><ÈM¦J`âp4îW:4Ì{)¶u2y[¿K,>…¹ÕšÏu¦x×@8Fˆ¢ãÝ;¸ÄÛ8Ü¢÷/´ÂwZ•Ç"t¨µêIø/¢D¾’¡ On Ôìe‚{c[îµäÔK­ößFXV-¹òSeó]V 2Û›$i–P7":oWÜëP»ª¸Yõ¬ÊãE5ÕG!5椾pNwú鯓C`gS£Qà"®Ú…ž¼ìó¤Ÿ>5ú¤ÂëÅGS~ôbûÝ´h&ËýÂ%T»³(¤hµ“¢Œ•ÕÕªŒ›wÍLªJ}ZlÝ/+÷>,\èžÎÙa,øÆX¸p_ˆïëûìÙàñLñ£µ™§sý²RÁR‘ó+RßÓÍSl%Ãó÷cŠC_ùn2¡\J™‰»8Q1@—²s›×°¢`ær Éêîa¬ºî‘ª¤Î#<Ûª„c•fÀŠÕ„«pò¨ Z!4bÎ}Æý….å_îÅ8•M'Xq…ÑšÒ§»VË?eCÌu)‚ý2œÌ#[FÛ™tšäjƒ>(^ #UÅî»IDÉ`hÃ<”ÁØ6Xú$rðÀËE1o½zO¬ä¡æfÍÖ×uׂ¾z­2´DH|[q‡U…8 I9áÌ‘IÛßX¶˜ðR7¼ßw—䌾‹9ÅâÁV„V\ ²,4ù¡ÙÓÀ‹«†ƒÏ$¯±B‘S)!6gÈ_G·Æ?èlò ùC»æÂÒ7¤º¸p€FIiÑ}jd’J–£ H2Úªƒ¡aÏ’(›ùHÄh¤ß䥸ÎÊŒÇvÍ300»¬­”ÚÂÙħ„¿}KÀ®‚óóíaOð~ÌÊÑÖô¢"ž7„Nã[¥÷<ê€4x‹{ÂPÖ~ ­­ŽÊkÚM?<ê°{»±€ÐÈ–.tšŸ°á,u²'–qªwOÚ†k\úx%,î{Ohô\ØfÒmv"Çq1Ú–»¶L´±êL=&ð16g£6-òó`U> F: WÅ‹AbÞÌ«}Æ'žEËpϳ¢UâWª¸Ôœ m×ÊZŒ·ã’£”y&!@ô땤U§ðC¸8ÓþöSŒ^Q/žÀ -—_ÂXpW5uúðcÄgMu\ëˆyœ1©±°º»Þ’­úá ²Ñ°xÙ¾ö– béÅhÔ8†vòó ìd>!mQuB1 .Ü`°O…>)_êP@+Ÿ)蛕ZÓ\'FzÕn; GëŽ}°¨ |ÔI£#Rþ¾Ix9….ü<¾a3gñlVÈ Hð#¦=QÈñ!ä™ùªþf$i–ôºU¼&œª˜ëG­ÔÌ«qÉw ÿcDgS¨•Ïs[‡.~1κ¢ª|äß¿’ûò¶ûÇíìqÙl7bß0Šlö\tß4]ÍÇ7eއpaí²ƒùŸ˜a!·¡â»fí&ø½v â‹¥zTL´ÄµjáçNpx’ı±dÂZOêVO°§†ƒZùAÇß/àö< $9ð›kO©Æóš6÷Ø|\ì0µ.½ô›"~§&²¤ÖE2Bí…Ý: lŸ2¦žàPÔ˜¢~+l@\v²›óZØÂ+$!ÍÔeR˜rýêæG^ øÅãXÑÊCø]^‹ÍžDFÁ£Pm×Ó"äy´FÔcÜ3ßMáOÞ’Õ-5#•êr6ÖŠl\!«‚ÞúÈ’_)I®>×eÿ4‹ŽÓ˜XWš(c§iD„uýU0¥{ླQ\f¡…,ACªÕSm&+1.JOªèÝXeϵ‘!–™·ïMå R Æ¦¡çø)ü»Hº+,^*˜îˆ˜ú-¢ÏQ¡» ]öC!ÝWxC˜ù}îIZuÉ+ÍFÅ @¦m]ã+Iïë~oG]–üxæ¨ÿ‰gäFÒ£ù¦x\DÌÑ>ý\K›ºKSö§Æ¶ç©Âó‰IêØ™«ÛbwgªÞ:”€Ÿ6Ý—÷“*ÚôUaìP¥¬ Ë!¦}«H æÉ‚Û0‹“•ç}Ã9PõnfµöëÒ-ŠAçå«JÚ=<¦9â"m …ûHêXÔº.OÚJ &Þú†ñª)äˆ5wé†%%ü £+¦KO4Yô£=DŽŒÇE?G— ¤öј#J^Ïêc6&s³"ˆÑlyW£èxêù7„{‰&“Â|³¢]s(?ØÞGÔ…æ™ÆÃª½Œ‚Œ’È0‚Àƒ/¶§1BCOë *•Ú²A?^iÅB½Û.ðop"ªÙ—w´ ݘ²ÈÎå‹Zm½à½tg½œ¤Út³ãÃO-Ik2§×|ÄØC—Y6¦§¿Ï¸ƒo(Œƒ¤Æ””'°õ—Œ‡If<œT¸½6­v°x.uôŸ'£õ§w>‘®ÐýJ®†øÈ·íú ó»u’þè{Îf]npJ^€=•~Ð;eõæÙòJËŒ®@J­Œ:>Ê´º\I[ Ÿ°™7š“ âðÎ(•Df"vç‰+^Æ3Ç(ä·ì[KçÆÂÐÐL!™†Ê7ÝcZ *›ìCzŒý~ 0Þ“H” •&2¡ºtïë/”8ꃲÀ¶Èa˜åÝxJë—³ê`ÖCõÕ©¸òÏ—Ù“}ìç-zdD(¤WADÕÝ.¹¿9 F”ÆL0¿ß%v:&K2«WãˆYž5p“‘4ÔøÄ…Qåk˜Õ²m„äÃj:ר¹¨%nÒ;R®y¨Ñlf|‡ôÚ¬nÖCʬ?>ÜÉhP¨$íg¸¹iÀ›~v¶ìà7>ZLë¥KZƒ~µhÉ[‡¿¨d%4ß‘9Þ¿^æy$*ê’nùåâ|ň5y’Áxä¦ô¬ƒrJò@viô_Zˆí´–A¶JÎZbaýÎnáÚvÖ.Ï\ãs²Ç3Y¿{Íç”nCï™PYa|a}Œ5~Ò^h2V6¥K‹Œ4Ü–X™d½ªÌº: Ý2 õÅ¢æùWMÌŒÁ©)š–½{€üyÅTßã STì³}£*á y>œó¯L…ï¸F¶†‚G›Ø¼§9ÛSUÊâ³ÜúÑžMôxq!¬q-Š¡s¤R¹tŤd¸¦èÕç¢u( «·0¥O•kQÕUõ: ¡ê¶N^†ºý>YDÝ J»Qy3ŒŠ€ù2‡Ý»3"ŽdyÌxÑɰ­¤U{ÜÓÐôyùœ!Ôg-^×{x‰1JV8uÉyLL èñ¯ÞŠó[VòD~üÃȱºbê/ëÔ#Ú2÷¢úOË^ϜӾhðe—ávQßÐí™»?ÏJs%Àe%ÞP£§»^$ †U:OÀFa>&’RÛø‰1šØ[„Kü{fP~È…ø”{söXµÊÎØùàXíxFLҚէ禋¹øq'^ŠêНÉNøºÞNM³ÊÐRºK‹·Ë³;RÝW&‘ž[ãqñv\ŒõA®üÛ%i¾ËïôêÅ2²6ß½ âªTûœ¢_2UWÓv?|Ü#Æ'ÚPVx˜u!2~º­ S†’\coøA±hgCÍÊÉ_4XˆMÌÂ2µú/­ ¿\ž˜MvàXZf\gÁ¬œÈ­Khòáʨ°H"¨¿J†qœrp犗÷<•¯1vm=Æ28ù©QÚx8R/‹ÝÒËA«Ð7X8µËÈe-w fcñÝ0¡.ya2t“ä(×îdñ&)Súuø¦qtwZ~aÐÍ|>àI`_½’}°Nps<šqA±þëeÖªò˜þèê⾡ü5Íäz¸Ë$Ç”ÉëœZ²•ûo#Z ÿz ÓCŽÀDp¦E|ÄpÈžý7Òº¯c––d™ê¯¾C88öŽR¡«#IWV‘xn]È׿«Óc\Œþ–MrMã“ÆÛÅW_)“.OÏrY»à¡,˱±bTˆ°ƒÚm¢›+ÑIõ~»/•ЦÈ7ÑYR‚ý|þအ÷{ÛAÎáw=œ„Ï<ä69|½ž! F{¶I #—ÇÑœ±KßÓ?Gc©ŠBO£Âs4›B…Éâ¶£5â—†&Á 1Ò4XÛ¹ô¹ã`ÙÀèÌAg˜t¥›žÚeK̵{¥­¡PUÀ唲S¼ž‹—´Ÿ†AÜ}«¾¯ÿÄ!e0mcßÌBís°ëà1¿¡þML|Ày€Õ­”®d9Åí%¹ø§šçRLšzá6©Å_ØÜ•&R«ÏB’Lœ†¤Fbäw½)®×~£L¡r1ÖZ…lö8aFUs'ã™´ôô¹¾¤*ý`·…ôf›±ãpìÝÉ­‹%WWç^|ïldrœ¨zÚH(£¼b:ô ,Å]v гóOW\K’ M,„â šÃÉ)fã¾j.¢s®:¤Q&´Jê?øˆø_Ôe{1äüxXð<Ðî_#U¥ýHþÌN"«FÛã4ÓÒÛúåá+›²Fþ¹ºõÆnü*¿‹ßõ<Ì<´JDè6¶ê,L ¢LùÄe§³x]@CÛÿG7£T endstream endobj 266 0 obj << /Length1 1553 /Length2 8310 /Length3 0 /Length 9335 /Filter /FlateDecode >> stream xÚ¶PœÛ-Œ{ðà2¸Ã`ÁÝÝ=8 3¸wO!¸Kpw—à܃'„ Á%É9÷ÞsïÿW½WS5ó­îÕýu÷^½kh4u8¤l`V yÔƒ›( QÓæ¼œ@ ƒ.ØúÛŒÁ ruàÂÿ ȸ‚,ÝŸl²–îO<5 ìpó¸_s  Pè_D˜«0@ÖÒlPã(à 7 ˜³+ØÎÞýé5ÿz0[³¸…„Øÿ„¤œ@®`kK(@ÍÒÝäôôFkK@f ¹ûüW fQ{wwga..///NK'7N˜«8 ;À ìnй\=A6€ß Ô-@uƉÁе»ýe×Ùº{Yº‚OØu{Šð€Ú€\O/è(©4œAпȪØÏÀÍÉýïtGÿN†þ ¶´¶†99[B}ÀP;€-hÈ«rº{»³,¡6¿‰–7ØS¼¥§%biõDøS¹%@^J `ùÔàßí¹Y»‚ÝÝ8ÝÀß-rýNó4e9¨ ÌÉ uwÃø]Ÿ,Ødý4v®¿NÖ ó‚úþ lÁPÛßMØx8séAÁ. %Ù¿)O&ŒÿØì@î~ ( Ĺ@ÞÖö\¿Óëú8ƒþ8¹›Ÿ:ð÷u†9lŸšùƒmAO?¾n–ž €»«Èß÷ŸŽÿFÜܰµ;À d†bü'û“dû~:|W°7Àø¤=nð÷çßO¦Oò²A!>ÿ¡ÿ9_.eu ee¶¿:þ·OZæ ðåàåpððB|ü>!€ÿgÑ´ÿ]ð?¡JP[@è¯bŸ¦ô¯‚=ÿ>æ¿wƒðß¹ÔaO¢˜ÿ£ñ—@~ õÓ÷ÿ³Òÿ„üÿ üw–ÿ›Æÿ· yä›ùÿÿã¶tC|þ&ÁW¿‚wý@ÙêxÄž֗ê*˜÷e3ñzq/ƒ‹çò¬²æIhQÜ9(ÑXñxãÌ]^ÍâçL=R+'°aøÆóùoò¼¾µ^¡ËãÖCJOjLB‰x‰?>Íè+½Ÿ¬L¼è[ò!·…sŒ'ÿA9Ëô0cI™†Ùl/裯äJ/ªìáÕf £7(¢sìËu? ÚWò2 h±_ ¼{ÅÚœ´(`I#xœÖ¸ÆWÝ~7óGÊÓÅéD`« gãÀ®Ï­1¯Ÿ»¦b0 ëg—½õg÷ðÖY*ìíúa—/7íúQ°QÚÏû`CsVâ½ÂJ~..òªfš@ÎzFï—q‹Xs´³ŽvRüxl§Ñ›¸y]µíŸSúE%jŠ[t»‹Ó>Æ IÖ§q4º5†[)×þvŽ.BYŠøÄ‡Ïˆ.}+¸¤|£nÄo"´Ãn “ûw æ.ù‚÷¯X½·íg:6ø")•¼’Ÿ&ŠfË .ü0âÕ'Þ÷Lš®zP¢â¼¿‡}ZÒ«3U“ÿØÄ~¿q0ÔfÐzÊ®[?Om5„fñ™sT×…üųͅ’¶œ¤(KÙÂ}Á/ö0Џw×A l9v–“+ÚÇ´ïD8)e±RÌnÕãUô™KÁ¡ xWË)_rÊÙžÍ4ÍÖtû*‚èdW[Ý‘9x¼â-§ó&-é¢Ç?ŸÞî­¤m«:t ;cÛò  •ݨ/“V ÇÐaÖ;I¯´í_;œ«WÅE˜ÐqóÅî+U§æ!#Ìm*¶% ø,FìU©)u%âú<ä<áå æCѻ߅áëј[1Ïeñ‘gB„øC÷ø’-¨±Ö¸i¦¦Êd$c¯;ÄÐù0rO—¼À?©Ÿ_¦ó©÷FZ^ÊÙØžDÆÆä»‘dÍË~F#Àâðõ1{¯ÕF˜´ãDCäõV[.pK¸s? ÃL¸Çlíù¥ü+s%¯ÜÑ%Û ÉŸ_6ƒ+ïDSÞTËwšF(ÕL] ß¾À›Žtó¢™JR$Á×`R¡E¨Íó”&ˆ»¬Øg±-N†šúÛL¿‰–^©'ÿê„æð&-(:ˆ$Ź{PH©‡éø?ß ,Í'1+„oÇ÷L±d˜ª0W©1_æ•B³D.É4Ý1‹ï"wÌ­Wï'vQ˜33´îñêÕ êR')öv¶pN`gÈøá[ J<ž"š÷²Í¼§’~îŠáÈd3ŸÜM {õ$³³ly_ªÉ£aOÉQ›<:†à”ÕrO:}—Qä–$—×ÇÄ<×e2­m>ž&­¬4 i/`m›/öÆ&Ì+Hu^cã$´¿/ñ*60Òíû<ºtìNsæ77i?̆ÊÊ>ý°$„chïÃ_hÍËP„gŸ~DþYÀ]Ã{6²ùr-U×#ªe¥K壖Ö{¦sC%½"¸7ÎÄ%Ï‚Þ÷–> δÁÚ¯œ×¦ÈÕ‚=“~BËÖç "#m:\ˆ9æâ*Vˆ·q<|†2*ÇÞ±S…Ç•(ÕPW÷É;Š`]kSò/¬(Vì¿ô{{CºE½> ý8´Çš?j T“ÐM+Ǭ_ Sø4v¾¦;z’²ëxƒ"°øqŽ#ïïÄ_¶Ò,µˆzÓ2¶r‚œõÏ” €¾Ÿ·‡æ>‡'í£7H×^ÇL„Kh“:ÓÙ·¯£àato«˜oD^v«]\-ûéÉÆ|N+Ÿe8bXþÑ;£éíÔ•]ãJuð|+VpŠËpT—ô¹@§v˜4—!–×!Oðu€Á¨~l- -ÔìÚ^|pÖÐÛ Ê¤Ö)1Ù/àߑRêR)"Ðâ>­ù‰ô9Ī]ÿql¬½uÒ8ø+ `l|1N葎‰åeqsi6ö ¦óäAE‚2?y ŠÝ+D·Áv Š•QS2ÌÁ¯—g®4„srä)Šn’¢MâD ¶Ë*^6¶ZªT˜ó&Ö`l?F 0ÜÛļ_”:xæƒÚBZÕ7“wZ‹+8£¾»»A‰+Nwç¼:ù¹'ûXB²Î'i(4Ík}Öþ¸.X7Ö°‡æHÛwc'¿ÛnMHàõª“£åñýK&ÄÖ„M8}8íîmÞŸ¹—ÆLß<âµvñ"UHÔê3m0¬ÆÓ:{…já[<ÞE»éÝÛ+lR9…›Ú¬÷‰VañâT~º,»|+ö« õ*Û£ÕpRLv>gQŒúõsV•þíßÏé¾.ΕÙqžBIƒÊ&u½òp‘јó•Äp6 ˜K+‹fʲI¤uAgîüvÈ«#Ó(‘(¢e0çÃúbdÀñ€rg¨"Xœª*ÿ|Ù qP±?ZHkËœy»éDì<*?¯ya¡>øv¹F ©"í;Hªû+@jÐ3Žk`J-¥FbjÆF¢5 .øáp^ŒKÒ[Y¥ž·Þ=,l¼5¤ìäæ½éônÚÃ1|”IÏiQ;Ž4TÝ¡í/M]ÏCä’7¡ªyñ›M‡¸&uRßZÈ?w‡;XµiO·OY䦔ÓNú›’¼¥ÆK ‹ñ.ìÑaZæñÉ9!ü±ÈÛ]áÙsïp@:Ó*ÖY8:j8e˜~pƤl›U㻬2 ütsºlqÿT$¡5!û­R!­Tüì§;üR­™Ñ”‹3ªn߆E~é„/žžçý(NÓ IÖ75}Ò XHÚ§À#Ê-7AH`/Ò?ÑEžòÇvE²î÷Qhb¶””ÄsU/RS?äh8åö¦ç^/îœ-™~Á<JË&w1bpHþ2ÃUŽA‚s¾/ÐÃfž^2öµ0ë€Æà7eÐJÉ»æ¾^õ,-rÇÍòƒÃ´ HÛ•‘DÖfÉØr—ZÙ=!©'ÇΊYµ¨h†ƒ1¸9EB¢A+R®`B›5!#BߘW‚RæŽ|Èܨ5S„‘8ÃßñY·ü¼Çvê¶å"Ýz\6RH©Im6ÿÚ„K&áÚ_C‚ž&FÐX݃äâËþk=R€éSK$ûÉ3ªM†¾½'ß–g&vïùб• ÎGiÈ` G»%gpŽ7ö:Ôüj~¡k›ç½øøCèüõY»—%x¾ã!ÀðËö,SÆzÒä$ý°€ŠÉ™#›5RuŤæ tÏg©žâÌS•z„ yølI½{}&Ýc‚±²o9ö zÜ M9ª˜§1SÃ¥òkêC#LÖ”æˆM&óžÏn­½tE&^9LF5íXˆ¨CŸ7ïCcÛP¢Àm?R+õ$›ƒ©‘†ðÁI¸m¿ÆƒeÇ'{Ð-¸Ý»«|,`CN¹d‘?bF¦ú™œ-¹ƒë]öž!udÀËq‚òkÎóÒ ü¤âÈ$k¤ªÊ&” D4©ñþù íçù¡éÃBo"dtÚÕzjÛ.¿ …­`ï##++¦O‹’)hèñ9ØTvÔÛŽÈRì½ÔÜ÷6—ón §h6¥ «ç ¤\#ðØ•— Tj×ñ×B\’ÉÚ¨û/ãNÂ$æçç¶|a3[TۿЄ[ל"Ñ—RÃ?yT=£q6—Èîµö4 bGŸpˆÎ¥.ýµ@ÀNL9fzªç—v!.]Š{¤#êµ"¶ŠLïðøYï˜ èó+hWùc$ÅÁ"Ï”’ÛÐ[· c¼ýrß‹ÿh,¾"eVcž€÷Êô…‹ãž½°ü`ô!ÔC\Àð^õ-½_ï-ïÏíd±‹áÎ7yó÷žÆïä4QÚ¢Gì¶à•H¾áØÆ};7œæÙ®ârjrlج;‘3ÛxÕ²JhYG”˶Y2o÷bé<‹"^¨g–L£ÈH4¶nª`3ËQÆ6ÚÈ‹Šª%lí_úu` èzŒÚZ”ÄØXï?Ç­Æ>ÓláL»®^WÉí'‚$ hµ²C€jú‘÷ÑšcTÄx¸Ì‹©w„nÇ4ë\ðXÛ=¡ZªçãÈ6‹²­€l‰Š—šÈk‰Þ«„zh¡È¼°†×jpX)pYß™§~úR´Hèß•àÏÙé.æcTµ7¹>Gb8$p®7ž„Ý-`:Ìxã½±í]®ïä…3épT²ÃÈ>kߨKê÷in05¾#ªâ±oâ6•báL8†J©õ¡×. Ýk9â…c ©¦ŒÜ‘zÓ,9v¥ìyɸ—í©Ÿ¡7Ÿ÷n}­óŠkp—Þk E.¾,†Ðh ¢7øý&?-%`Û>ž6ökb]·~•J¨É¼žžÈe™æ nôÖ‚Rϯ›£Í|I¨Ô]δ*̨P{ß`îe:•R¾ÅIž í°xPþÎMŽxV2>ù΄ÎÕy?S Ñät@_uVº~:ï4uÖ±fâÍÎâÂÁ­üxÒ~—œAUCËý-©¶æUê*´±‰"«pÓƒÂE“ï¥v’]CÀ6Ý·X4„ï?ŽB ¤K?“µ½N7ºè1Á´ØSI¯$\ÁNvbB©t©J7Þл Ëa ΠüdÏf±±œeº$%‰«삎9Åƶ5³Ž<~Û9 ÃpúLJu¹fŸ€luŸÖí¨ó3c÷•Ü#—Xr»™¥ë«¦B ìdª‡¹å/(æ 2'ý=Dæ}Iƒlõb×Ú¨NÄ9›¸uç‰Ü1‹ Ê «ªÐYÿœ$h»|Ä%·‡/âGµxUD;è˜Á‚¦ ѺºçFV!ÄÝ^Z˜Dþ¤#$Z<üOÅ?ê,oéœ-ÌûºúŠÑöê¥öΛ| “݆3ùÑ…ùV–>¾&Q‰ 7¤û늓j¤‰z¥ìÖ,ÀÝ(¶œÙçƒÑ¤yéÏ7«ÓÎDƒÁ#{Þ² .Z^¶%ÎÉòe«-·etÌÊñ’LˆþŸúÄŽ¥©h,îOYŽ ºlƒ5¸`N<£Ïq¢ê ÂŸ@ß²‹fVŽ+›ÌyÃf£ƒe2ª'±}öb%¨uè°§›ÑƒtŒ„ã Àß ¦ôã¨ò˜ëÎîò!ò‘Ôl}úŽÂ÷é-I½¹"FYC\ÒW"uQÍ7N[%žbä–Îi»WCÅr˜õàZuÙÛ¶\izãطÈtjV:aRÇ×±ò$‚œ7ëqŒÜÖtÄ41kÒb(µ¿·—M‰#…G‘¡w ².IÛßQîÕ Êæ´•­p'Ó«Ìt3f±>:Øž}}6¢¿,îdâ>V¥˜Âå7g@’¤Êê8®f£'²³fF¿R†xçý$Vì"cG®ì#‰a)lb>È1Ñ2V*é–rjષüÉ3'›èlñ’`=éñÌ]e±Ðt?åÔûñYúœ(¢Öˆ[è#NÎüg ŽŸŸSùçºÛ»){› ÉÏeWB×WÆ€9L]<ƨܔRḈb¡°„ïM¾i;±R(:Nõ_R‚Lùic\å<—÷¢èî˜)VìOó2ûu{ º¸Ê†MmBi”ÊÑ)w|ÒQ§.Ã¥l©>޶¤Ã+¬!GÙÕ.4ZÏOÜ]C4Û{¿OT£=z@ÔQ&p!9sJ2¨IÍ”¤›^JøeþJ“Pª )"goд© Ô'ÞÖq‘pÃF< }ô­û~ÌaËF×"‡ƒríd²Ho†1?5Ãp!ªmZlK£(yä¦<Õ=!ÓÕMÝ…ÄW´úëGޭܜǥÒ4ÕÒÂÊsj•ˆ_'G“‰å˜íDca}…`þg‰ªãî«3\DŸŒðáÓ¿ðø”èñ¡­3\¸…ÿü­|OƒTCdüÅ„Sh‚!çúUÞT´ß_S¡Èx /»ƒ¹™lQ«N½ ÇùæQ¦'aíÊ8]ü½™ úÅf¿N¶J¾(*µbå:´ôöú"eŸZ¿¸A¬–“'nGñ¬Á]hH¾t"ÌAÛédŽÂžq䋚üK¸ñ÷ZsÌ:8õ'ˆ â˜?z†’ÝNcZ séÖÜÑN=ˆ7§³_ãbÚz6|pw~Êu²=(’˧šó_YbsÖ›™)öyüªMïj`¦9jÉÇäÛD¹Æ/] ¢LY’•GZÁ˜|dqÞÈUÈž«Âéw:.|·"YÓ±H~¥ólR Ïö=Hœ¿¾û5—‘%c¡ª8œ%ëçKø»O‘_n|On†ÎsTÊæNƒ]u°ãøú_ë Mµ\¯‹™ÕjòN¸TË“­¼ÁW»¹u¤ôë|9@m+|ȵѰ”jÊvY1½àûn\j]í=û1©AÒl-œÆÚWµéjHµ`¦ãFq*Ú²Çó1È)¾‚rHûŒˆŸóˆ0.àãlCÖ8/£Ð`ñ‘ x[äkØN÷²z õeΠeò6D,+KͱVª‰À·èTEk9eÅoBaAf±"S2oE¦¤þA80bˆníz½´t¶"ÕK `ü9)nWåSµUÉøýÌ%v)kXÓœ_C–}%ˆªØD•úcUbö3‹%¶­][Ì )½H1šš&hr2¨Å@ÒÛ{CÑ$ö±<¼¸º–-¦s‰ƒ¯}À‚êüÌ×F¬¼oÕ¡%Fˆì—[—šU\/TÉ¥i:÷å—[ÕIá8FTpŠ¿%ºlÝîr4‰MfùÄ È¾Jur__ÍòCŽyª °_¯ÛÉÌeã‚€yþjæÃV©?¿e_ßZcÎ{!Ѷך蒯ÛÐ,õJö|ö=12§Ælï䆆ËþM<î£CN¡âe…I¾/µž±Ê@VgÖ@§Á)„öXÞ&øÔð¸Å™&/⟪ m_©åÞûÔfÅ÷}ßjBÓÇ{Û¼ù•Å'µÏ_ùá¡“Ê'o\Ø»=Â]$Ëø¨­—éJpVÆfLáÑ3|§ßäúx\-&áùh¹»ÚD"gņﱄq:nV½“mçÓÓœ”\Ï›ÿ)›¤ÊavDÞ'Eæõ…Ï/x[è°³´æ3Û ƒ8B¦ Uu¿…„pIÿqsˆS—‰´¿žªS©XEbA ÆF!úºW+9<Ýânm~š¬Ömª±ªW'†BöŽ-º0÷¡„•ü­˜ÄÝÙ’æù|AƬï£j ­‘ ä+æÛ4mò&ø»±æCbam’"å¨>Á—e0Æ–ìrwYtz×⤎&?âKª†‚Ѭ,7PF)½lxYeµ˾32}_”6{[ŽÈ´VøüI'1\­íÞ·H%5A ÛTñ$½sÔÝ ¼•²,FSÓ©omL{ŒC‰Ë4®G‡ƒ–Áço^7}óîù´NÚ:z|_Œ¡ø3Ù=¾—¬‹­!fî­–Ñý7æœ (k½„^m(cm ßÝ[#­¶m)1&³ö©ë]eů*eóG‘ÁºT9Éã=¥³Žé#Þmèª12œ§…Šx@@žcq8Û2‰0dÕès6rX¦Ü†t’‰B fËyLcE¨ûÜn[*Ú)/cªÎ†JÒ¾n³?ÒÛŠGPг<îüBJ†æ¾æà¢’Ë%ß*ë.–#š)ê+ï¤Û·²£ˆ¶·.[ÆFdì‘™ÍäN³¿i%»H=ž|¦1³º“ät×ÝÁòʰsÝÞбÇS cüJ®UP&V­.ç™?-E°çÌh]E¼ps/dk TÜ5)ù²SºLn ¿ó‹¨ØÚÕRÆîTá¦ác¼3adeëØžžJ:­Ûl2^@ˆ„µÚüÈMv2H†¶*9ò%k°]7¼± Á½VˆÛ¹Á  Žýhµ£•Ö¦ lj ©²á4q•ßÓI#×å,®iÞ#Wì ñÛ`Ìi]i]°ûlé³gK¡©ÄSNÁg¤Ë&ñ)êO®4òMÆ7ãôŒ|ÿ’¿>ÐÚeä|„°»tHÐÓ(½ú¾MÛñò+Ãi^(=M¤ Åó »zsv÷ŸÔAn¼7¹Y~hš™;[‘ÃHºöQ#ÁÇûêm´OÛöë¬Q|=‡aŸÑž°gõgÁ'ÏÀؽWŽðT|µÇ¥á]T<ó±7¢¼‹?5}Â*ÂÉÏšà,è@Ǿ7ìÃ)ª÷œËçê}ü f?—Ä4¡“ÌÀÂL–14RpZàRŠ%D'ÚŶ»½Ü,½ Oáï6á£FömÅÔO:®ßy†?ó~ “Ñ’¸W™p­†?5¬¡¹«¡%Þb"%*/c¥«žò¬*ž­¿»2•Êö…Æè[¨]ü¨OæÀ‹9El}¸ßUÝïÒô–йo×§‚½ƒØ™ z¶,~xbÎæZêŸÏ^p“…|ô#3«Öó.B¥mL'ô% C?g./…@0Œ&¤0òb‹WG®õœp4¸J7$j_R·ñ¡ßbU³bX©M<´4x‡«ƒÇb+WSC£Nâé³Îõ‰œZYÉ­Ù+‡é*òÒ«ö¢8‹S©˜"„dÖeSÔÔý‚ödvTGDô^WHiDð“6õZ©´åô[˜³ <³Ûûò¿xX1³5ªðýæ‹¥æ!6™NœåªÌRL®^‹™¡%(hC4ï^±Lhz=ž«Z Ñã?r•ÎÆâa/€Ð{)ÛÃ;*gì/ð*»q>æ.H¼NqßOÇêöeq¾—K`M,üÞ+¶"3_kQÌÿ+¹|)@A(êÛDEöi=±¢½V‚“[ø«F¸¸WÙ»WZõIëé8ŽçIŠô Žœo›Ác˜umgœ9çúˆ§/=ÐL ¢îÍÛŠñ¤Œ`È»-dAå×#®iŠmåæóµµ8{SþšÆdo©…ö_!Rd"&Ó *ñ")±T@ÈNGWöî ÞB)|ú/~p°ë8¥&c³(/”†)(¯N—QìE3Ìœqয¦{êc¼ *ŒëíÜ®Ã,`¿áh^¤¥JƆìž-ÌDi|`m¨¶YÁ¶Tás᛫Wj_D¹ÚòÀh”¹dXs7—@­¸-"ß¡Øí³/³×)æ”òœnýÅj¥¨¹#Kuúý60oÙzçÝJXÉÞ¡Z¯WDåþškŽ—µ"oˆ®œ€·.µeÅ‘’ä7£ìÌ,)^fÏh‡—yåÊà^œÌ¨‡Ø °‰ïžÑGtôñ×d:9n3ÖŠoe•J¸¸>ËmV®Ý©Z!™»¹MA.¢9e:Ö ô.Ã5B׆@•VãÙ^8ÜQ M|ëáS†èÛD±6 ŠÿøT*å endstream endobj 268 0 obj << /Length1 1391 /Length2 6286 /Length3 0 /Length 7237 /Filter /FlateDecode >> stream xÚwTlÛ?JH¨0ºatwwƒ”0Æ€ l8lt7H—´„tˆ”¤AJKI4 H|SŸ÷}ÞçýÿÏù¾³s¶ûú]q_ñ»îsÆÎlh¯äw€¨ÃaH~! 4@EÏXŠÂ$ìì¦P¤äJÂnAxBá0éÿЫ $S!1fzp@ÛË $—’Â@ Ô¿ ái€*ÈêÐhÃaOv¸uvAbnù×ÀæIIIðýv(¹CP0Ð!] î˜Á 7€  … ÑÿÁ%ë‚DzH úøø€Ü=àgyn>€é0†xBÞGÀ¯rú wÈïÂHئ.PÏ?° Ü éB@À †À<1^0G€¹`¢¥ 0ð€Àþëþ1àüÕ€€Ð¿Ãýåý+öÛÃÝ=@04æ p‚ºAêºH’‚9þ2¹yÂ1þ oÔ ä€1ø8 ®daêû«:O0êôð„ºýªPðWL“Õ`Ž*pww éIò+?U(Æt-ø{¬®0¸ÌïÏÙ stúU‚£—‡  úÈ ¢¥ú—"ùs† b@ PBR y€ À.‚¿‚›¢= ¿•B¿`Lþ~p€¦HÔ ‚ù!ñóyCH„$Àï?ÿ”H„„ŽP0àq†ÂHþŽŽ!NdÌäPÀˆ!žøëóï“-†[Žp˜úoóßô44125áý]ð¿UÊÊpÀ_à„„¤D˜CÀ?£‚ eüÛW æHýIÓ¥%ìý×ô¹þZ nÀ?céÃ1Œ…¸þ&¸ P Æ| ýŸiþÛåÿÇî_Qþ‚ÿw>ê^nn¿µ\¿ÔÿäuCÿ¥ÇðÕ ‰á¾³°ÿ6µ€üYW=ˆ#ÔËý¿µZHf”`Înÿn"ÔSŠ‚8B‘`—?Tùƒ›ýZ07( b÷„þzPüB@àé0[vÅ<ž>þVA0KóÏ+Õ``¸ã¯í€š3bŒ$ð¬¡#õ›ÁA‰q`Ê 8Á$¿æ)*Ä‚Hä? ñÿ–ÿq5Ø À,Þoj`òú—ü{Ë!L2?Ë„?l ï8­Wºãÿ6*7žf‘ÅÍï7èô:'#Lç®Ë ]A|WJ|E±øYëDñÓ¥ßN[aT{ªÑËŸþvÉÆk/IæÆiûÆJv”ž¿½G|—ßTqÝÿò‘¿yˆ+nv·6{á#/I2ÃbªSŸ7¨ço+ÞEά­×‰ë^TNò'˜ÅÛ„”¿g/rÈ›¦c!@òß#â¹u€"ò}êVÁØ5“v2/IÀn‚H©ŸÕªpâÙ´ïRµ©°g=½Ý=Ü“[ï&8ü”73´oÏú=+]A¼Ëœ¦íùèž&ä¶É廦o|ŠØúö€ƒ}Lš‹žN—Z+.~¥]3Žœ9…@4ï`Q{¥õM¢ÍÁ“Ï zmà&шW=UƼî@}wöÁÎÊúDÓ(ÿ¶š<™H9RFƒ0^vMÆíPú5ñi©I:EñG…×R…c+óW3ëoœÆ3MÜ忊ÂÿYApê@;ÓIbézZ"gZÕÒHiÖ{K䇣Èú×H6ù­n/Ķ¥‘?Jª¤&®v§´Rµ•jo3på×°[²ùd2LUÖÈ=ÑÍ?¾LˉLÒg,è€M{PÍh}}±±&z}ô^¾ÂÍù…Q$+ìú!tI`ábŠ¡¿ì>Þ°T`ìçÑ ©×zžƒŒz¥$—¡º%Žÿ,µª { Zî*ÒÆlb•5ž•__i}ù}½æY’&_äHÂþ$#aÝÎJöÊ9ïЪûrgdäVSe„`¥  (hº ôQz´` ÁCNk5¡%â;ƒ>Î|:5ÁiÇdp÷y=É&™zD Vå>5î½fƤ(sJWjç9~ÁðêãmcÅI¦y·C,•ÁŽ1ßÐz"F;¼ÊZ(çŒÐ˜Ña¶ÄƒT‘vµ¶Ó©ÊýÝ ¸{Rl¸=Vߨ¦ѶnHÆDE!¥\]NÎ|<µN0û»I…P§Í’ü3­€£ïc ®hÖP=ýwÉÍâyfÕdfº¯ÇUúè‹ë@„~ ÅÊ—º9]ÚÐN$ ž”$CÆêÅj{ÀPJ 6±agྎ©Û®ÇŬ⼼-ýµÓ§y‹Éð £Çæ;i«Õ°ŠÀá9[ü-«$ûxÊâ½ÔHžuœ@ ÒM–|òB¢ÇâÁës?|î[‹pÊc/Íü,ó`Cv(­nÚrË2#[ûÉx}zúÌã©´'½ïûlI¬?7–‰DÖäÌÑå¦òÚ:'×oSY½§Œ¸ºy ÷ô©LÝ42 áÎm‘Œ«Å~Rzê÷–Gõ%¼Ö í°ƒ Íb…dÇÐï3]/Å™ s†4òb%ªoLêtÐID w9åÕ¨äÏA4îB^ ¡a·§Ec82lw6>dÕ9Ž\«/z¸oS¶šÝª#wi-œj,è¥gó)µ©Óv$”äí1)T“+ï¹s k6OfŸZP˜¬›)в ‹­:ƽðuò1ArÀv˜l î]vë$UöG× ø·ðû\É]ýÍ`¼Xš$ÙïÛ ÂdÔµpKŠZŒU²7tÎùqœNvœìe.ú޾l)šg\äîì^¸|Ž÷ä]åI¥7ãÏ­cZ!·ysÉu·WNÕÚ‘[ñb™|’ÁÍúl>%Fo>SIîLLŸíQ¦k ¹KÕú›[þwë‚:¹­Àû‹4³qrIécâ=pÞù”pB¼Þ;ákÖÝô`F ÷²x2ŸÍ˜<ó˜kôÖÇ"ÈDK¦~gÏÉŠ§¥¶ûêþ^U²¿5±U?øÜ%ÎÇÂË®÷ó´ø+kÚæ¹f’Dó~Û»–U­Â ì$5ío·¹cuŒÓÚ‚ü!r{ #ø•¸E1Øvá×Í@“Íð¡ ±öØ¡¹n÷AÔº|Æé¬&ƒ¼sJšúÆûÌÖŸCçç ñíϽñœ¾ù|4>O"e‰£%ä¦$ð¢ùê}cEͬ*»Ë&î……ÌN|>ä°æ†øK¶™Òòõ¸¤CW…·#è—èëhшv*YüïKËÃ)äT4Ø·3Zç{ð[•„Œ$C•øýWb„ì}kbš’2Ê.¬l(Šúû7 ¿ô{vg³¼pؾmkŒÎÎÎæ@ ¨)\tsL¦¹·«}ÅõÖå>eR]±p¶lf\õaúOì'Ì›ÎF1uP/;»ú2åÛ…ý÷Ƴ÷¯‰…å8’rùf–\*^ sí{<¿£â¤–"v®n;‡Î€¹øß}çHt›`\«;ãj£Ç°¾TYê…Ê5*8^Šäly›ø1;ÝÇ|R„Ø”cD:»S_ué²ñ¼ïRkòÅSËO&FÇßåoJÃaáUík=B¸èÊÚ*ñНk¿Ò~!µ±¶Ð´G$üã¡,³ ›pÕÓÌW_¶»ÅúÒzœ¯Í±Sú‰o¯ñS/çJjåëŒmŽsw ˜.òÊ#§’L3¶Ð,ý¤’Ê»«:ɵ5 Æ™’Ú/Êò„ÌÒGäù|5Ž’”Xbn­‡'{¦´Ñ9 Ón*ë—2öhÞw!dU~Î>JØ,yY’ýTÖ–¶{|á\ºpS#--Y¿^þWér>9eËíê+÷¢ÊÕ¬] B IVC^[<øõümy#ÞÝ]y¾UŸy”–àòÜ3¦…òã»yT†1«¸©ÔÓï"s à·fÈÕpº—6ŽÍe_Äsv(ø¹C«ÑÏç@šebùyý&a߆ª`µÝÿ”·èܾJr飢ÌkG«¢ C†Fœ;ËwPGœt«ÎJØÉ«l¸Þ=ÆÙq¨Råh«H†hj;Ÿòñ­ò®`¿Øê6^ñÜ•‰Ý+2]œb½ŒK÷$=ÖKc’ˆ-Ë'AM¹Ulã…»5Áû‘ÇxwúüËÄÛ²ý™Ö²“k<¦ÖÊl±ãäúè(@ÙŠoőǙüJî ‘Hiø{<­rç»fùbDaàš¾Ë ¥Î:ì/…·ø.Ê¿qoÐJCé‰úÀ.VÒnÝ[u“¡7dÕs$Þã òJÕuÛíôš <Ï} ·q½_BÁî(|]‹Ç{ëi߃¸O©r3T¥×Ãà iÃ-rõ¡ÞL}”ªíH,ãíW¼Ä–VÀ9Ôhö½m¼(k«À™Ð´ò/]Ù §G-âRI!Ê?ÔŸñÑi$J€ön÷šXÎ,¦üÜAÞ½ù 7ú»^7X^˜Ö‚õ™H5pc¡ÇV"¾¾©EëîýdüÛ¹ÁvË3>ª0i®Bv•öåý =h‹âþˆÍIžuÁ1ÉâIB 2}r•Fn3ÁLïf­lG3 œa‡›@¦óÊRÐä£Qµ0£~wÿÕŠE/çÎêÞœ.~o ¥tó꾑ÕO^vjâÇ;1<~·Çb§[cMbkKtf;p Ú_9à_Þ1”¨DÏtÅxx3º–›~x7kjÌËþVk©q‘9¦áÓi“:9ÀYcy+ϰ'NXÁ„ʿٽþóÕefñ®>²OZžNÍ`ÈR[{_ÿº$Å*1µrëZY“Ÿ+Ý1Iï1ý.v²ûsaxôêLã×›”¾‚Û%ç+"%«i’Kxz¸îic>yËÆYWR£ôÅô´¹åÎp%Š¥ùí±C–h ¨í}²éöõ”®÷²¤j¥Le;ß =ÑÙÆ}1P Ek­å—ex.Æ&FfìæàÚ«¤ý>\¬ôΣl<,,9ðÛ¹Dá¼´cèÂs(€XîQV CLP»ªØ¡’ŒO«îÀw³7?ú-¼ìÙjÙÌ "èöñYR×ú>ÁɈϕ™á5ŽFF˜Ü?ü,7Bj3éUÖ®œNøöë‘—Šw®‚æ,ñCäu –jA  ùõ|*¬ýoS©¿,c½¸““ÜEÓ†…&e_šo³Æ#m…ÅFÔ¾U³¶·¨<ã*Mwާ  óɈ#4øû—ÆRzÝÁk¾5kGÁkÜ"äI9{zß K5kò Qw^]›s¦” pMy֟ˬû¯Ø˜LçÞß™”²gÔÀ¿qÁfh1_±Ü;\˜sRc_bÀ@Ÿèá¨Zºä~ì;€_1óÆöcMdã‹™R|ßž(4^<”¹+‰aƒôñ±^;a¡ûĬèðE¥¥ykÑÙ§öÖœ6½‹âÛ²SžþâÙ§RŸ)£ñgtb¬÷Ÿ8Õf!dp¶Ñœâòô¥I^_I˜ œìáÄsºü ®ÂÙ7Ã\|´9e!Â>¤[²âEê0p¡óšR›ïóqã;ç4o´²_nÏ5òšºöÀoKÕf-H“¯ ZØœE{' Pà ƒ•ÄhÒÖû^9Ãr²Ux(Î}݇zR‰^`Ö,VçÌ7%-œ—½§žÐAɾá5­f¿WûkJyäóƒˆk©n02‡£‰ìAÎÑHh^(uœ~Ù‰ùØfÃB~ɘҀòMW^œWŠƒœcY‘ž¯Â"Û#ídÚ ³«[¨cpÙU\¶æI혗æÞ ¥e´ž fPÀs¸B¾Ð?(#ðïyo8ïÞðªG0Ú«n•UB‰Å¬ê5ws¸©Í¬jGó&^¸_0ñf$xæ}Ø¢*âj–~lÝ39ËËÀ·SÕ8󚑎Žo—È€ô«(ûJ>•&çÅÛJÁ[MWÍqÉx7â/Š{§ýùнåÚö¶"s™T¢+ssÑöâ–d†<­ëÇqSË=’„ìÔ~pQpšÐz»É¡r3³£!‘øN­»sxÍ´ð[îrAs…#Þµ¼^´poÛý‡A{d½ÖúÓŠé…ë&´j•öëˆCЫ«ðhs Šr%Kç‹L~\âxzØ;8]J\‹[‹…˜9ùÍ´à³³’]'ð.Ãb´ñvoe˜™ÉS{2rY¦3jf?úØ<²ñ^HohØ:—Jès/¥úWðèÇÝ6MÁ^Ö¤§égä«åNÐ Ÿªˆ²5U×Óó:"ŒöÚ'i‹©ez“ “>$Ÿ¨üþªäÍÛ½Ÿ ôSkç®A@;¶ØÇ=Ž›¡Ì²=»ÔúC7•‰ù'äÁ ƒø4³úä¬eýI‰ÏEÉÒä½ùxõ–…¶7z<×%îÖ †ÁW#ééûñ¡Õ‡QI­Î¥]VBÑ¥ƒ#ס\)nñ±Ÿå\Õ‚êªz„È¿äÎ-sJËpÏ,ÛŠ2ðíˆßöB7tç•@aó˜+ŽuɰcRX ½hó™ž|"Ì ”+S ·îI^b/>Ô:;75˜3ºIbvMÔŸ†^ä*U£áéù)ß]͹µ&}“ÑåfP¸ypæ’=ßæ‰7BÄô^¢é”ßÊu§A×Âå~ž]sB¢ìÎhø”ê./qù5ö² Í2CL£#ÕP(•ìÇ5¨Àö&ƒtýÔ([‘°*WHËe™\UÍ^Õ£¼:Ï•-oàVy7LŒþøH*€ÕB”ƒ€š¡âëäÏç›”t¸¢ dž+…ÁÛƒ¬G¬àÓÂ[‡cO˜kø µõ”xqî9ǾÉ/vu¸éÔ Çì‘Nó&k>ŽgÞþ™Ú]y!7* Q©Tp•WÁû|@#d‹->`&”ÅŸÅùÁ­ZÁ†(ÙìF5ÃJ¦ÙŽu_—õ§ä-9ªÚÑK4I߇ÂÀô ªýV‡ŸHÂÛ®­ £Û ?VRߺäü,´–e4íƒ4úÂ?$˜r²„à+mwÎ}²àÑXõG’šµa´-xšž‡ÝUªð2Oïî*(šâ„€ñÔkíWD­T ý‡Â/[[Uß®6%'¶u9.|_(Þ»êÃ{"_âarDaΚÚë%-ßéd?FÆb§œ‡¥1ûèwe’Ž¡UÄžŽ!>`=–ø9ªøT7Â\,nšqlçÎ\mM‰r|„xûÐ÷¯pËׇ‚Þ<¨K³ò´ ø•Sn?ø$«ÿVº§)«ôx$¡4»RŽHž.P-ôUªÙ\&³xŒx÷…Ý~FsÞêý‡‡Ü _<èµ@ß+gåÉs/=ÊÍDÓ†Õo¯ÓßâY‡:iù½vÔUÒOé "pÀ½ ÙTW¸©Ôjëà•7ã=ØRUòh)5ÉZDvLTOÙ´„ÆÔ²ŠÐï´®F¬\S‰Ú,”àlÖE*ÃûZß}ÎQ¯&}ü +ïrá@>e˜ó-¡c- Õ¡Ÿ™Ä ¥(tµSZ~Ò<ý<¬¾Ø“U¦ãƒÏÀ÷‘dlláÒŠØžÛÏB±RËms8àlàgXÅ4öŽQè´‡ÚÁäK’÷§=÷²×K¦wL˜_Ýød’b~ÕtzÕuœS¿£a+ŒS Š(£Œ$ôÇ=Rò}w3먥kù*,ëÇÓG=븺õù°Õ`TM ¾›mpÎq8—gÎ*T}0rîÙMˆ#ñ#*ï‘qeXwîNgt_0"Õ:¾Ÿuoói÷j¢Cn‚À ’4)á î9>3úXƒìË®²«=ÝC¦4 ×匈ÚÌ7åy]‚¬~ØJc×’ý&é-±º)Ëv_jÒ¥ `ñ~3],Æe,ñGL_X¸fFÞìè»åp‚i{.$àz^žF9e·9‚_©è•? OY7r> stream xÚvTlÛ?%1ºF)½énI ©1 Øctˆ ”¤”’ÒŠ(ÒÒ RÒJª€¤|SŸ÷yßçýÿÏù¾³s¶ûú]q_ñ»î3s15´3\ŠAÄÁ @ #3Y ,)K,X/ø pŽñE Q ÿ¡×ÀÀ¡X¦ ÅâÌŒÐ( ¾Ÿ" „È(@dÀ` ,ÿ/C4F¨ õG¸ÄúhÜ  ö Üܱ¸[þu „€yyYÑßî@5$ƒ€AQ@#(ÖŽÄ݃zÍÑ0ô‚ŠîX¬· EúŠ£1nÊB¢ÀÖh÷…cüá.À_å¡HøïÂÄ@ w„ïØíŠ €bà@à…€ÁQ¾8?” ÄÝ 4×3Þð†£þþ1þÕ Dòw¸¿¼B ~;Ca04ÒŠ B Ü€®/8ð†¶¡86+ „¢\~B½|Ñ8¨?áuÆüN ÔV3BqõýU/ ƒðÆúŠû"¼~Uú×d-”‹‰„£°¾€_ùi"0p®ëA ßcõD¡P!ή”‹ë¯\ü¼A–(„\Oó/ ø7æÇ¥Á`°¬<÷Âaî _Á-‚¼á¿•_0.ÿ°o´7ÐW< á ÇýB|¡þp ã ùOÅ?%tAÀ°@g¸øwt wý#ã&AoqăÁ¿>ŸìqÜrA£¼‚þmþ{¸ -s[+5}‘ßÿ­RWGCÄ$dbÒ` "!”ÅÂþÅŠø+ ð¿}õP®h üŸdq]úWÂþM_ð¯Åþ3–1ÇX8Pðß·Kƒa¸/Èÿ™æ¿]þìþå!øç£íçåõ[+øKýÿh¡H„WÐ_z_ý°8î¡q€úoS+øŸu5‚» üÿ­ÕÃBq; †róú»‰_mD ÜÅ…¹ÿ¡ÊÜòׂy!Pp´/â׃ƒ€Áÿ¥ÃmÌ÷høâøø[Ç-Í?¯ÔBÁÐ.¿¶KBZÅ` A܈q’40‚[CxàoAâ(4çÄ•tEc¿æ)‚à>~¸®ãà߈”<„{¨þ–qaÝ1ðÿ@À8$ý[þG20? ·Š¿É‚Ëô_ògÃá0ÀÌvý®ÇË»¯kÔØÄÖ†•ÆÖ¬ ‰…Ì`ZýN©H2„žçF.bŽÔ2ºh>®j ªÎr_„l5½"‰iN3m9 =wL5L2½y²¥VÛËIÆ!f¡ºzázóŽ'a~»¾@Ÿ•Iýq@N`moåÜ»è©5Óõç2äç•ïÅ-ìî”M:?ždá%ÆŠq’ ÓíROÓå\r맊¶%KBl—$’N&ƒçŸZHøv°ò³Ú²pÒ½»¢¾™©Ïü!¤¼d¡xÂί4Nòcž½NøÔ ™¬b"Uî…J?!ϺÛòívòÆA”¨Nlõlæ“õ ѪÛÕÞÒëçàÃüþÛ4»!Sœ•tõ/yÛÏŸÈj¿zĸ"¡ÏÇæ¤tx-{sÖŸ[y–[ç€5³š'DÉmJ©Meô*¼×„äb.DÜÌãåATtæ "¤é•)z‹ÏÉ\ÒDõªêî±³‰îØG:‚7˜™ÙÞ ³“}«‰õt¯'ð¸°ÙúÑg›€lßóªhÚº£JÄ£w?ba\¼i(•£ÒáI«Ø4¡¼´ÌÉûÜôlÜ„¥d‰žÝ™ëñN;‰¶ÏÒ»AÉæt”ä™ÂC*uÄ·a¡$ŸÒ׬zÝÐm©»©“lo«_:w"äÜá¾Çm5¦ÔˆÓ»ÍRBÚ…©'ôþo²#L(n3M§tx#KLôøjÅŒèÒF6ozSuön«.¼ž˜Ë¥·]Ô—qØ{¶2«ä®6[³8üŽX¦/í<åß¼À†·Ù]pPîàwör‚Ó6¯ÂLž«‡ô²¹â)-×»@â¿Ìëd_ é†|èPuÖ ²Æ´ehVß°M)×Iù¨´•&(àPÞ ’¸ÙÑÈ,ÔØÆöøËûÞ¾O‚‹û;ȽfV×fV’+tfI`§…¸ Cäÿè“)}’ჷ‘”ÁÓFœÈÜC%/Jr÷ϧÉD>7Nù«Õ¾ µêrWúslïX7èª |·z•ÿ„¤ÿtR9pV£«²}KzŒ(*ý`' pÜásÅ ÃtOÉ´º áÁ˜”¨Xq$¼ëëëñ–bRYÜêv–J’f­ܧ3³‹öqbi­XaA9i ÖÎ.ÿðãÃD¦”"aIôëåꇑ6"àcñÓ‹ºˆ^周ØkJwúôôÏ8º®öïb4x}½Q‘5‘²ÉS%­óYê-©Ã [ã·¼ç‡áÂwšU~â0=ÙÜåË|UM$_ÉvýðŠHíŠP¶I¹llØG‡:¤»önîtŸÄi|aúÓšÀp¤L±âM/Bîáà›ÉA?ªŸðj•ÈQëЮ÷~êièiSõh¥ÚxßmþÐ)ÅΚí2ûL¤Ä>Úk}ÍDÃ#}V‡D .·águ>«¬ÿ0<šñpŠ:ÿÇîºÛ×sên¥JÚ ½Ë;صÁw¹öi’Ÿ’?² EŸª&}¿E¥3oXÜ/-ÿ2»> á]= 9+8ç}Ûœ¥wé%£u´Ïƒg„Yd8ÆÎ¹s°9ï§–Ê\í°cÛìƽÎ8VþTEÍ-u•T§WSb*“SNõ¨ßPÐUÊ’ú±û¡ºfB=¶1—W y¯Ó“µÄóê©8ÄÕÀÕ]4ªÆE2çâ#iå ŽWqíñ¼à=‹M ^¿; eˆ4–bmö\\dÀ  Ïiò’ÐT¶çÏ­/ÛiáųÖb—›¯Iz_ì aw‚æÉØÑ;ÔÊ'ü²ŸPÇ¬ç æG•€[šªÄ3w±cz´Y´Æ[wñÉãìï$ºöžnÜüR=™Vµà!{êèø~È+·3¾Ã¼^Æ.vç”p¤°Áá|#kîÖW!Ú]Ýg¯I&Ô¬ fµ¶ZÒ{.Ë~Xºå+zÖù”á²vתW­O½öx:ÞLëN5¥—p‚ÿJ›ê¬R\<ä“%PéíT±SÚ¹~^óÓ‡qeµRä¯íg»§'úÚ# Ò¥)v6„y1»Ç™ í ×íäö¤ýÂw×n)p‹èhD(¼Ôá5¦ŒŽ³ÝÞŠ±}ˆdÍRêš„nä#çD”j*oöÐë[˜Œ.I~ =#(ðF>ó¶”´YºÈnÔhcí¿uÓ:o/C¼Ûú\ûm[f²>¨É&¼ ø‹ël]ëè¤cÔq0÷á=£%Óð ¥$“Á=Í/¼O†´Þ}6o¼õW(Xca ¦vñ<Ò>¯–/*Lž‡«ò âh6í¤ä­-Ç€ñf9‹n|ö{V>u¹{¶å©«-Ùƒ¨ññ ÂÇ›{A¼yq¤[/=øDúVúúíM…¤ ¥ßÊØ¨Ì°m¥¾q\Ê6 àMzóìûY(ðy蹞½­ËÓ\p]ºmÝå¦>¿juèÞ}®Öhærá/ålm9AûL4´ùÔ#<ߊ&©ÀÕ›[’Ç„>þDYâÚÕŽœÖj™…ÚaG÷¥­Ôƒh¬”'Æ%‚odø¬••Ýö©‰*N`çÍÞû‰ÑøÈÍ_¦ì»N²°sXL|\c£Z~ÒÜ_.}Z±IӢך4M Lÿ.èÊFäKô¨Æ«ùí(n5 ¥õ•Ýâ×¼f±´5€N7ÇÑðú³>âZ¡@úÝH–…Ìu„Šu¼ó;©¨ùƒ0ìx^ÇOB‚‰ÙËÃHøÖÕ4CÛ¡°‰h«ï9´â[ReIQâ»-CÂ%£œµLƒš3¡{¶*µ;weï&YAöñŸQSj*š¬B=ìOâ–ȃõÛè3 承ñ¥ÌÌ&‹xo±Š‡Œîø«u]}CËf¢þÏeïSk9ãEö®7ÛJ3(±*wçœd‘<ÚïÊe8.æclq7§3S«ÝV–èY Ž"½1©âEÍë³ H=¦ùåFtÌ¡Ÿç!¡”c÷\__^‰“q{ì)ظŠ$’¦ÅQcí錺âŠR•Û–Á *èÚʳÃ3‘&è¤?øì«¼xLŸYñ`y„ă¤ô«*˰ö‡Ó©ëÉSDe$˳©ð¡J¿–xýš#,ùý~+YÃK|Áº÷ힺ›r˜ÞÛÌÛ ¦>bGÒ{UŸÇVyè)ZïFCܺݜ€ ï¿^©÷Ë$«?[¹é‚¶hr¬Òâ.‹ŽÑÚjlÝqþBµD3œœÂ"TÆç÷^¼9¿D²)¬XDƒ$î c'¾ä ‹»¯xÛöU„êéÁq†½l}Ö*jóÞEÆz;•LÜÑ,œ‹ÏéÚÈ©QoéÓxVXï’^¢Y— +-þ+ µÊJÆgev [†{•é¾ BÞïG~,î]êÞ`Ö4ûà~×ðŒ .!fŽv¥¢B -úåHu®!%»¯ÛKlE оÎjÉoã!A‡EVAf¬,´`H°6ÆÚÿ3ÎYi}<–ÄÔXK̃þ¬cLl|ê8é»\Ùó§%-á¥y¦2Hühvãxê(SK±·×l Eù´¾üôœºh”S‰ VUüL{‰T©Øë¾›½ÒÕx1Þn³›ÅÜ!;é` ŽéûjäÔ6c-:3p„&.Í&£˜3¶.ÿùñŒ˜JÃ|¿ÈÛyjp¼½ÿÒ®$$ö¤—EB²XlqUßlOÃï*ÙH3‘¯Þ6œ:xÆ¥v÷óʸú1KNða‹L ûuÄDèôÍ—3 «C]iŒ)1´_ƒ Q’@|ùIÉ]§²˜ëÕZ+ùÁžq6ºK4é):ü[ú3=‡iY7ê” DîPrˆhÏ}'0ø¸StùÁj–šjöó‚•íŠ ?‡)MÑÙ¤YCiʆÅH^Ýà¼òf3ê('s'>áºö(—ÎTÙ×kmœ /{+Š]·™¬cFP¡z¯í‡Ê‹}@,Ò˜óf_D©1¿³wfÄ{ïEKçC0ß@j…®öÖþIíµ¾ø6ÌÉÏûGEË ‹{˜ÞFvY”^“kðÎxâôÍ(áÝ‘ïQ"V» ’'¸6¥ál "‘/ý‡ºÊïÙ[¡_ðKÜH‹^šæ¢õf¨‡NòØítpº·á7 e™cçÑ”dd~¶sFmDþ/‹~‚,¢Ñ£LbV’!w/3—+ê•z;§ŽÄG|{»Ì–Å29cƒ £5 󮊃BRþe.åÜ) íÚÄ 3YÛû·vöÛœfK’“þÇ ×£B2·ÙYÂym!&~ÌÀ¶†Xt=Ý–~S .kR¢Cðø¯ú:;&lÖ‚î»ÏÙÍ3íñ3#¯Úºß!¹NR[Ï÷JMGF]¥ž»·Dw wàa¸ZJx~ çX±=’E5£ÖŠß©[Ú¹kq›ú¦G»Öé[dsFµ7pÖ›žŠ+Q¾+ ­€Í•‘L?Èœ|Z4ë° 8ç§=›’^'‡].BÍn{ nøY–ð ñÝb ÞAui_rIÑÌá:°éõ vôën,kô…´ÚGr0‹HÈ2T‹‘¢? ÒO¸t´§¢¡¶ï\ÇHh} '³R1.oj¸QxÅ j¢í¤™dK–Ǹbû#õ¥ÐÓd9ïÈÒ[1TŸ‡„«x[n]­ò+R&|ŒÉöÛ"‹WzX^Ç‹ÑPGùºM²ÝÝÏr?A½¬¨Ô žUt~9\%ÿ<³Â.B÷÷ç2xäñÄH[*-es7Z©Iæ)úEùSæo'O&©ª9]'†ÓRl¶:ÁÖºR5"Fsº„Ø€àhBçýIRÁ°¯ˆï=³¯¡áêŰTITˆHyÈCç‘ïàîòìjht²=ä“Géª ¡{™å˜Y‰.þ`OErºža¤éÔyb¡»Ðæmãááò9¦ÕJerp¶Ôq•RØéF©‚ôkò8£í¼ìÕ°ÔÐã5[ã˾I[<|‹”¦ï[©5j2óôÚMáëhöäÀÐ<¢ø~=ª/4P>·9vgRþÉ2É·¨IJóff+-rŒ˜Eœ¬rî*ï]Ž”K“ ¦±îgq#øBóP|CÇ¥h“ê~7›šþ£€™âÜhx¹×ø½£ÄàNºUGe‹‹ÑîvîOxñkæsHx{owë"ùˆR¿R>-öÝuÀ9ô^ŒÊ"ÅZF=À†Ó”Xµ#ŠÀ>ç𫉄{ èƒÕV¤õWÿ¹ö4ÀÀ9á·‘Ž´ë(ƒÜEùFÃ;ô Õ UNÍb‰Š‘„ß²röXª=zGßÍé~ûn…µKø±ûfô›N‡~”«ëöZÛ–Ú£o®®ÊPiM þ ™¦/·”¶‹Œza©=P¡0ýCŽU~£%öªÒŸÈ~’Ú´r~:p|}ÊK ýx^K)Ⱦ k÷ù‘uÞªZ{Îm©§wöÏw2CÓü£W£f% íBÉäy÷ò]ZeˆÄeµ›=5ÖAKxi}i—œux®N÷‹¸S¶(ŸÈê̦îe<ÿH öÅŽÅ<㾑ÍPU»VÂ7$Ñp Çmånz~ëC9˜ú·Æk®!%ù¨|S®—¶Æo0dfµÉ¨üÒ0³Öȉz0åúçGÌŽ£¹ ²7º6ĨÇIµŠ==ÖG¶>"Qž¼Ø?H—y$Ÿ®:ßO>Ôª¾'ÿ=–¯S;i3W±j¬éý,Ë‚ÃBÆVàëÙ¨.Œæ¼ëQù-X2?wÖŽ(œNS 7s”È›qÞÊyS$Ù= Z‚}–š™ªÜIÅ3X8îP{›|YÒyO@M•¿ÒÈ¥~›_NpË7í¾É[+X©Í徂DƘñÕo>´DC¥¦ðÊoÚø×v½Ñ­'nb6F²ê`ÈET’«*d>)¯ø†ì_m¨®P¥´¦¥øtö’ïS ®®ˆ¦@‚¿m¸ÕD•›¤’½’„ÆäÚ÷ÆïäÊâ“nñRð¶ÔaOá¹±Y[C¦'~Jr³É&o¸å>î«Íz½eFädj ÷‚Ûîxw6#Âì~Xð¡X})ÇPx2(jú @>vÐîµÉâɶft“ó>ˆáäÝ]»{üx45mÚÎ{ß½ŒLñƒŹ›W ÷õæžãÕÈ5ºúST‰Í /Ìæ•ܸ3®nqó³nÅœ ³H‡·~˜9`qà%Ò»MŸU2º|ÀæÞ´È—b>¿‹¨NØÜÔÆ°YÝÄñV¤®‘:]©›ë:Ði¿#œ¶íZ¾àóç£VÙ*ÚOM.JóŸ~ŒãÜí„òËÃX:Š}ÅŽmŸuv©„•Õ½›öo©Ôn¬~e¾´ã©áã §î•—Wž²Ã8\¿Ù‰ûšQÏóTEn=‰´)È­X¯.D[ZÙ£ÞxľOÈ:©˜äợϑ”ÉëcÕ_¾$×gt·°’X¨âÞáþóLešqühuKB‹÷¤üÉÏ ®èDtgàEb%Ûø.¯ï®37?–£ís¹MfQâ­ØnÄ·¸=~í©"¤ƒgB‹¿·Ÿ/Nø›W‹põM µkç׎K»]&µ%ƒæOÂ=‹4¯çytI™«î%Eœ~ê{åM¾yœ¿;õ…@”‰ïø –Röx.—=ˆ¯˜¥zÄÏüD4¢Q äô¸eiÕfÛð¯èÔ ²ÛúÃ=³QŸ»å޶Ë̲Âod øÞ®&ìÛÅÀWB¯í°6NÏgŠÄl<\Êý|=l­áAÇkÁœä"£MA~eCúÜ}†ÀŽŒçsëªï-´ï]v›½xüµ{Òú?¤$õë~ËŽšèh‹B6¿U»¶»õ £†Ia`Ÿô•c{‡’w–9QÃ’_JƒÓrN…¹ÜêJ¤gè¹’Lºu´„Ò»©Ltx™/ÁBë…Á1jÌñJV-Å•ÛdúÌ“£7¶ð¹ìMã 7:õdk0Õ ¹»¾ç”ïYù7-QØÊÌ'Á{y’(d»²ó™øœï–ø”ŒFX»ÜÒ:‘þ$<3Â{µs“YŽsvŸ¨k+ÐÊ™äe~^G÷NXðm¢ÿzûw8svQe(Ãl­@Xúe7Àk,Ïck3,áŸóL|ÑæçÝÄó×’òøf¦Q‚î]YH£Ü†ÙžÈM›qÇÜØf\K¶“7Ðl:»jÞJ’Ëôçjq ùŒpP”a—ð‘§9Áäe®mCä¢é·‡¾ 'hÚ¾Q›Ë ÜüÙœ· endstream endobj 272 0 obj << /Length1 1779 /Length2 10638 /Length3 0 /Length 11755 /Filter /FlateDecode >> stream xÚ´PœÙ-Œ»[ÐиK£Á‚»kpo ¡iÜ!@p·w×Á!¸»w‚‡ Á™™;3÷þÕ{ÕU_Ÿµ}Ÿ½ö¡£RÓd·t0É8@]ـ윂IeWNNnvNN.4::-°+ô§Näìv€ þK/é 2s}–I™¹>›);@ näò rr¸89þcèà,2s[”Ù P ¤ƒ£—3ØÚÆõ9ËŽF &P@€Ÿõw€¸=Èla(›¹Ú€ìŸ3Z˜Aš`«×…`¶quuäàððð`7³wawp¶~ÍÄ ð»Ú4@. gw%àw»3{б£Ñ´lÀ.Š5¬\=ÌœA€gl‚º<;¸A-A΀çÜMy%€ª#ú§±ÒŸ¬€¿®dþî/ïßÀÐ?œÍ,,ìÍ ^`¨5À Te”Ø]=]YfPË߆f‡g3w30ÄÌüÙàÂÍ2âê³çþþêÎÅÂìèêÂî†üîãw˜çK–†ZJ:ØÛƒ ®.h¿ë“;ƒ,žoÝ‹ã±ÚA< >ž­ÀPK«ß-Xº9rhCÁNn y©¿,žEhÿȬA®^NNN~ äyZØpü®ååúCù‡ø¹~?GG€Õs ?°èùÍÇÅÌpuvùùü[ñß X‚-\æ k0íŸèÏbÕŸøyòÎ`O€ç3ñ€Îß¿¿OFÏܲt€B¼þ1ÿc¸šÚ:Ò,4ü·JBÂÁàÃÆÍ`ãâå9ù9üÏ¿ÿŽ¢fþ« Î|å¡V?‹}¾¥ÿìþ×ôÿZ &ÀÇRqxf,ÀøÁ 9y9-ž?ÀÿgšÿáòÿÇîßQþ/ÿßzdÜ ?´Œ¿Õÿ­™=âõ—þ™¯n®ÏÜWvxÞèÿš¾ý¹®Ê K°›ýÿjå]Ížw@j ùûÁ.2`O¥ØÕÂæOªü)×þ½`0¤æàþý Ø€œœÿ£{Þ* »çGÃ器@ÏKóß)¥¡–¿·‹‹—`æìlæ…ö<âgÄ ð>¯¡%Èó8Ø¡®Ï.€çöüVÎh¿çÉÇ àÿ-úñ8$ÿF¯8jÿ ‡ÖßHàÙÒìô ÀaþpXü~wÅaù/p€þ¹Öÿ‚ÏYlþŸËÿ >çü >'²ÿŸAÿŸ9ü r8ÿ†<϶ŽÏ²rýG üKú'gÿó=‹Ÿ™êðOÏÑœÜ\A–æÿ ÁÍóâ¿£Ï*çÁçÞ\þŸ“üÛø¹U÷ÁçV=þ€ÿ5~ 7ççj]ÿXÏgnüÿñÒ‚@ž ´…9 ¡`ÛšàÖ_Õâdl;c"Ót;o>1±ù,8·¹Ýb!'3U¥¿[s¾OìÂYÞ’f¼[¤|ð9lªCkNRo¹ó½7IИÜiAû6AÔ7ž(^ÛKJΦ%¶ëûàä«hßûU.ÛÉí–Z.þ/YÏÚÞÒ¥‘йõÝ*>EôûÒ)¶íhÃÀ¢ºóŒYbj$W6 f¼SOì™Ë«i¼¬ñ'J…4¿£îýu®Ø›Yï•r-.—Z}b øK¼‘Iz‰½ /æ}Š ÖœG>εnØ¿Bö½wT4~9ïŸÓÓ 2’sÀÈGE¯5Ë%ÚQ%"ñdœ.+¬yÊŸñ7™»°ª&‚ŸT!ü5X&=çÖ“µŸÚ\ø„v“»Ô>d|¬k¬ðsÅó¬cíwWˆäîуÉ<*3B:C»Y>~[ZO™Bø=CnÓ"$‡*8Âè™ Æ¡6x÷þI)…ì¬4Fó—<ƒ‹ù*YKÞü[ÿdH¾j_ᓉT,T ++ÄG‹./?KTëômuS3iÙ$ŽKâb{Â} fÔØòJfípab‘[‹Ú²‡éb~!úýµ\zuØJL‹Ð+óóEY$üýîäŽ$ó~ËuLcðmEãióE<˜Ô tõÞ%»¡gu#H† "è÷ò©Š{[-“BÙý|1¬Êiû|ï Ö©Â6Ž·¦×ÔiGð ˆ8ñª\ܯãÁH€h6#yÙwÛmVüÛ3…#g\ó£¤ ‹y~Z¬Çï}˜%Ýbç¨ôž„Dkµx¯lЂ |¬ÍêxÉ×”úg=Þ"ˆ¨6]([XÿRÛü|àž ðYf8Íû–ç]é/Ü`ïC¸êGŽió“ *’°öow:¨eŠ˜›‰ÆM=¡¯fÂZ„™”+:B¢C·„Qàì×bŠ…F¾H OÄ®ðSµékŽ|i¯d5ÊS=ÓÌDñÉO9"M­ppÏšk¢´ŽÙ”¹ÛOÅú±šëˆùQ¦Áï,Ú§¾HØ—œ¬ ªWúzt–O7O¿<€öèjÈ›øƒvým1òž~& yÉbòøq!v>›vˆT礽(÷XÝQI0ÊÑ Ix{^Óf1¾Î8‹Æ™¿ìƒc¦—Ô r¹qwOÇ< %"„Ük Gë^WûÊiaj£Ì¾²Š5£&lä& Lt† á¨YT¨UL$ïXBe­\¶]÷2N Œ}à€ ÅÀT«\ÐlíÒÔ]’B4°;¼ŠŽ³î78qMïì>Çÿ¬mL"šp:48iÅÙ˜9ü“ÿú*‚&ÀóXf¯»Í¾9¶íÒÔ¶ÀbЮŽÖM°I£”`!$,ý]»§e6g 'Kø¨u³~–ÝXþó§¥Pòx‡U2—×ù˜7ÃÝ-í Í;hU¯lÛ½Iä§—ö%?Ò9´LߗŦªB_‘iXÓ|’jý0L–9ÕÀ¨À4|î»0Ù‰"›žT´7{œ~H°n'°X3DpòÊ%Å}(æ`ae3çöhxÿ‹Î/cþ` €…{|è=»‘×DÞXð8:]Bªíõ åxÛá:×ñÓä>™‹§ø»ß”KqbØw"*¤Õ4Íx>švµÖp]üÜÑ™6Þƒ÷}±!ö°#$<;\ tËw´48>†Øu"RA4޶ÝÏ¢x¢M£Ï Ò)œ EåDÓ‡[9~Ó+›tãþ´©Ã»Ò·^mÀë¬æç„¢uêno+9·Sf®Üƒ'h¹qÜÙv…Ú££zHÛýž´XŠý²‚ÆwÒËþšg”uøg:ºUˆÌ•ç›[¬˜¶²¶/1Ž.ºT1|NlqÁÊØ 'ø9½q‹kÔÀþ>ÖófÏÁ/òšyÒzÒáé©@H˜‚ÎåÝg*6´‚ÍÁWºi£,ŽôØ‚œˆ~í!wÞ^!bT·2â¿@i»™NZP—v³¢óìÚc´§F…•@g4wvÒs"qƒ5+“£°î_è!ÎwäÏNÊ­ü¬² .Nh Ûz:;«—ü3ð‡º£/áD—ɵ¿fÇÀðõ+³$›:‡Ewjm®Xªò£"; ^uѸóœR¦ÝÈdkz…ÚÊ"Ü€™aÖõþ%[M#Bú!à‘ŒÓÄóÀóÒ n*sl¸v¦i¥òÿ£“>9j¼Xÿº%Â/(„¢ÑçZ±Gê"ö©7uŽÇuIK/û…@ê“® HÒiÂ/öWŸ\»§¤£@Ü¢™y¦D¼Ó‡BHeÚ§‰’çé8ç÷²|»C©ªsA„J‡—½=ÃÏVq㎂¨Àéݶf <3ƒÆ­H&V€`û « ¬‰_ ªñ IÀÎCªL‚ÈÏÍX !fý¬–˜àÕÎìHY‡hÐ5št켫8í`üüÓÌÍÁ1öóhº{I6±»ºS±‚Pä–)³«dÆ#H9èLYí‰ÊjÁ™G8»šë›.@˜Ê¯!ç£+»ñy·ú›³YDHG&® e„Ný)È ™9ïíÜ$¼jhXJ«K€rGJf‡æCJ±€÷;)©?ƒ.wúÐߘSV1©ÇàÎÛZA‘ã†mqv\+R„Æ ã—£¯^ °ú„×Ôˆ<À+`Kq¼©”jö‘êå_n¡è¼"cPXͼžØ¹£^Œm? NBY.esYç{ÇY“ wðý!.Ü­+ïd䟴äÖãäÈCLa®RÝÓÛ“KvDè¦ÙÍv½`Cã"*¼”%n Tmšnv^îÔ¼Ÿ—¤¼Ñ抖ÉxH¯×Æ`‚Õ}1é犛:¦š5ˆ-¸íè§[²³1xìÇ0Ï ¾u$ ØøŽQ¤ãón¯  (ÔÄ4½¸œ›Fã¬ÀšÝʲ>Û&é4A#÷€n|õÙoUo@„•4ÉZ'ÉË}ˆ6B–Ƽ©Æ:OpðïÕÉxjØ håd C6ÜÒÊ×£DvþZ›Ý¸þ¡[¥ItM9kãsŸ¡zȤž-U‹Á^ÒÐÝqƒâW)o}UŠÉ‰<&î»›¸Êð‹®8žÊIW±û½·ÔÜ]S½µ¢–!ñ§€sh^ÓFyG\¥ìÈ|þqÖrû¸-颃F>×ÉÓ÷Åür¡¤âg‰Û@“>[•Ù1¢:ÓÃ*y¬ŽjWZI«±,t+ëL!ùǶµÎdÑ\ê”%‰é¹i)¯‰ ¥#å×hÒ+¬½<…s¾ûh¤~ XáXˆ;"·ß« êG= ÍO¸îOàZ#ïŒwNŒß\#h‹ëðWzª¿ÒœH+ŒXqðÂjpaئ­öcÕÍ@Åàd°gRåé -}›“Yž4sãÐÁf•bÆ MÀ¢îeêî7áy’Èà Ìò€æT@¹Ó$"NQâ*·è|\_êۯȈsøøùd.ñÕ=qšþ~:éü,‰wX·çîã(¯¶ÃVì6–K~R |%Î:öóˆðëŸùáõ§×ígY|(<¤š³}[#êGO€c®›ÿ©Ñi,Åýß3Áþa,îæNŸo¶Pú‚?·Vr‡EP1aæ "éóÆÍÊ[KëR ŠìXó.Ý|"Ìè|`ñû/GKã«Û½KA@¯ÍÚ«;Heá#m’ÄãÝÒòÏ™#N@™Nu«l ÂzoKT²€aò±®Ö´ï”p‹;Œ½áäõÆÑ©äÒµtŸ1¬¼#F¶Œ_¢§Õ1$×Ö À6`jê¸FãBÍ‚[‰úƒrOôð°æBõs.ôɃ¦'­C,»Qµ¦9¶|s.¦uÆÛ´LðˆVèIiMÒʉBD€)wÞøc²ó²žý{pÔ5õ…U2×—ˆÅµÂ¬Ì 1U/¾]a/ÐÌëýt˜\3£·•\!¨ê-p;&¾Tl3-­vëŠE\œÀs\¥»Ýz˜I»Æd¶h T~ïÎãËÐ72›`£ïCl¿……¦‡10ƒ4lvذcØ_Ú ÅØ )8ô}+·¨ŠÃ¿êõ¼e¤<wGz 7Þ¨Œ4ºm&üz©ur‰/R ÛÌg ÐIùô˜O[ƒ}j@žh¿=vDLæ):© Ö±Úvíƒm¼µ¾Î0²B[¤÷2‚#µ+ñV¤¨v?|LƒÞ ßÏ7ï`’ñ™ÑG½ Ú·r}t§–!R{õ× ‘Þû­J`û‹ˆ„„2Õòõk޻ǰh>е2ñÅTy…(†86qH ­Ð{š õÆ—¼”––œ¤nÜ,¥”s’¥—úÓX“Q3¾À¯=FB»ØõÙ˜Zá/«›'ýÚ£V[Î}˜/â³ê©Uy÷1… ;øÖúfz䇢SìÙú<ÅÎᦗ†â\¦ ¾rÁ´F F«ÇI.˜øÄ’ÞõoIËeŒþU¶Åuó[ÏÈŠ®<>±…´¼/õEdÐ>Å/b¡i™É-º)P2èÌ;uZ•$¬éõÅÐ\{Ðþ¬ÖÞ§ß¡ÆçÛ ™=%÷=pµÖ–ÏyáYé¯uÕC2M36°½ >D^%jpS(Ü’<+lä³?5=ãuÅÝâÆ‘QÕ­¾n}¼hEUz"$Q…”‡‡ïô˜çË«TáO¬0}³ÙÚqŠó“B~àðûpÌ£O''!CÓ—Õ„çº;!¹QÔš{WÆ\1Šaî37âBò«S "õÁݤp Ùb¿ùž%Ô …#³)¥ãôx¢ AÃl®ÊÛéìHŸr+_ÌãÍ” Ú &#Ät<„½¯µm{"¼¯· ‡ >‡ÀJä>‘G¬qñ Rÿ¡É_UL¯Pz’f–Ø·/!_ö@Ï]±`¬¶aͦ±jñ¤{dy¹¦ÄË“ìu·ÿU%†&ö㹦KclËò…m)][ßÀÆÜÝü^·³ÓN–BuFà×øšwÕØÛXC×$;È×rG+M—‘p5}“/Ì©ø ®æôéØ%Ü‚x|ÞŽV1H¿Ù¾*\[¹‘Àà³îŠ„jbÓûÈ(¥ñ:ýt÷2ZD…Æåú«Ö(ÔT™¥yèL¦GÕ½"ðÙ,À6Ø Ó!¿ÕÀK¡Ñ1CÍOEi±ò!ºH½ùêÔ3ÈÛ«çÇt:Rç%åQ±Íi„¡S»M^ÇTáBœDE¶¦ì%” ÉŒ75Ë›÷–¨TEï9…ââ>pÀöD¿¦qrØ´Œ5ÊP}Ý$âÍÆTöTVoIñ«Ðú Ä€¢4[ƒúsK&NZn*Íð-„ì k*8q˜ iãç±x·»->íPaï½ÿ‰l¥D;´!®KQy$"ã^•â.:tq¿‚˜0˜pØ­#jã Áë,º’âžÝÓ¨ÿ(ÆPwÂÒaïõ‰w[쥠cæaïé>ŠS‹¶©×(×:JÎã[z±Ñ:W¡ èìÇ_"ðcÕè§q‹—‡_õXJû–Òã"eº{]C¼{8¼5ìüoŒ¼Y3!ÝÜ]ôuèÆ±e1O^w=ôå†h9auî,yîºøÍ‡n’\·ðò•ÙXGHo³.H­ƒR]D~Q3:]eSk¨wìi:=#F Y¿ï­à dßµyÒSŽ&ãúÛe}j敾/fÀ¸÷Ž•±ÿx›bTqºl¯Õí3þÎ&èߣ™ð‰±ïeÞƒ{º«jéKÁ»‰\b’4’ÓNµ§·3¹e{Ÿ³ B›Q×+µ+ÃÀÓÚ /y±ÔÈX*«Ó¥Ð‹"i¾¼y\á 1–•-ôìT“À]•óP1¥ý-yQ½Ýt7;>Ê>çuëâf³÷a¨D®N"78ß•EªEoݾó¼µá7³•}*ñ€ûš?œ7›Rý`þê›/ÿ~·ÿÞ ~2^¬øYÿ.—×bÇÎßìbéi¡;3cadЊ…¯¦wQã­–ŠR*V!Ù”UÛþ|œ}<4-ô|I—µ<k•#Ó—£ëx[º²•ªÚOšU:A%™ ©ô2Ñ4§Úãb:ÑmÎ7 ¦U×¢Ô!ÝÒÑœ‹Ì¯0>æ1—»ÐZŠŒé¾õ÷# K”¾$’ ªæÆã çøŽ´¥ ›‚á\Q¸~.Dy„Îû¡ƒ ô¢‚îƒÀðæ Ã/Û°$¨»FŽÀ¢é‰òûQ}Yg˜{»$‘ÆÁ‰uÖ…ùø¾A̪µØ_-?°ÝGó#("§Æbë+“Þ_b£;¹:LH˜ß$üi‡+x ÚBæìÍÌ+w\¸v69%•Å ?Ðòt{ÑãäË èÝo–äBšÞA¼ð”a"G߈~hêÏVI ÛÚ?»ì"_JNQՙ˰3h D7„IµÿÑ5÷C³R®â-XŸéÅu"w\°h1O–›–G¾ÌÒ$¹3­ÍR3?ÆéËv)$f¥Uª¶ŠO±ê²ƒø5›%6Te #h¯ú¿4ïÓj©ù4×™~ÅhInr!¸›;S³'BU!^è÷­1F@|¨1ù.a$jåTQbÛňNNFÎ\é© çøa@ ~Š&ñã:]$-Q#ﲇeeXu.w¹gÉÈâêX6U÷-¼3þ¡"áY ¿ç¥RŸ(Kf {åfeÊG§Ò›Zlf\svo—:@3‡ú•¹_+%nZa“€'Òr ÚX’³¥gþƒzÎ(¥ê‹òCý/øn¢jTÅׂtM Å îQ5ur°âŒ&þÕܬfXÇëc‚irBu½´\,kÚô˨VäNTá˜ÇyC^)TÄ!6ø–oÎJb•m#‰ ;|ã~AŽ‹· ú¶K 7 ¤p–ðUè¬Î‹wó_X5ãt濱ثÇoNPñ‹÷ÄÛ"ñ«¨ü ±–~¸ å_&(xzÒ÷©q¼±ª[bD5ýZvÖÖQBYÎQá.‹¸í@jcÑ·§˜•ÕÚа¤+ÏOiÍ+Su, ¼óÆÌ+Ñ4œµMUß@Î÷sz õaÿ\ý½V;füöÇ*a™#V~ÌѹA“i›/–|ÑÒ×ÜB`q‡éf)ðè£Ð±¯„GéGlö¯‘åNz«†nЍ-ývC¹ÌÃKåÊ¿BHRê™ö×_5JAùQ–_qÚ|‚-¶ëòûÞe?Ñé4_ÜÙªÖ¡²yßÞàÍ‚ yp›¾¼b¹!{Ø8߈Œ–r8Ð$f)kµìÀs@È`@˜™6dt—cÔ§ÓRzx?Q ‡ò˽ð×èYG€Ÿó¿›ë´å€»jü¶ø‘M¸ýÛ)ƒ}Îáð£þ”‹Ÿ©^0ËWÇÌyŸ* ò#áÚØhX²&¼îTJ€'iîÐU—D4‘´¥Ü@÷§H;ÔcÕB|8a’cd¢ðÚ_®;¨—ÊSfÉúúýUabñZ7étÍß§…ZÛñÖ¢·¸Ò["éYÙßÈÖŽØdæ#R -~p,Ø&cm#[N®9gï{a!äÜè ³qò á5Ù]ˆUÌçÚ»wþÏï™1ÍQ’dsº!uÈCó«üj¾ ^ƒ*Éåw¾I _x´á³0UÅ¿w5HϘq#/+|Wª&½Æ™b3ö1ZF½‚œ ɸý4nìÝ”ƒs¾³¸6´º~‡% ç rqìU%ÍZ ¸t¯7É. †JÝ/Lä—1ì7j½à¿=Ј¯5$5‡Ò¥ •/è<¢ÏmJí8ÕüGä‹òåi=EÚF\X˸^/‘9 [Tçzã¦Éë¿©ü…¿‘ÔÛ>–°u"0TÎX‘d²I5g•ÀGÃÀœ•L•æj#¥à©H8[‚åϾÇ6þ“GͲ®sOŸ!ÞvT´W8GlÃÙàTŽ C×GçÄ@(Ò®±Ç~–SrU_ƒD;‚™é“o“ò –*ÎÓ!¼×›Ê¦¬Ž´QÑÎjÃÁpË@q,{dm±ÞŸ“5z$B#8PÉxƒZ–ŠMqS¦Á´RXÉþ’€ðÏ=Α„’¤fiô£Ó ãçeô<ÀBBÒ¡ºòD«³LhöÞvÒ=U”üÛ:†×Sò[ô=Æir«!XEª0µØòUvÍç…wñquQ»_Vñê©÷*ü ¤ ŠúLa’zãiðð9')agƒˆcp0Xê0RÀŽf=¼K[ŸâÒËB|´|(Ö2C2i€<‚1T1AêjÚ‹00ù' YåFú)ÚÝ$ð*_°úÑÏESH9º-Iç˜l˜U¥Í‹œ×JTò‘'²d½1Œj9ÈéÙæ‚ü&vWÒK*-Úæí`+޹¸x‰æƒÇÆŽNÁ;ÝŒÊLtÃG²¾_v<˜Lq}îz×¾² jáØñ.[ØåÅÈ‘ôƒ'šù#\ÝR9!}ðw¶¼ò§=Ë5½bIŒÑ¯»&£®9rÉÀîsÙ pn‘¤7ðs-â»D,· ¬™»&5ê7ö \ØaW6תV&˜÷Ô’?ˆ ÚT$ü;'PX©„©aüô!ÔúþÙjØÜ×é 2QåãºôÆ݇°÷ÇÐ=Ø'ó™;׿¬,ûÔ n˜èœ*LìÁï[¼ø¯äv >½këTY[Fúž©\¦¼/›.6$[)â[fÅq¤½jy ojОô6pŽïªº*1¹ÎR'MB@ú¡]óÎÉéB=¡ìnM&†¢g.ã®_Å)ƒõ©»,¦ÐŽ>À.$\¥ðýìð‹×»?ôëUê-uA-6®7cp6"vžö:MlkûR§:mç››œOÙ·oó)æœf§ç¯Ì/ì·ÝÇÜó´nê™lg÷)·ÒN/–FAøüÉ8wq§¡BïÍ‹<Ù±Lí?‹ü¤b!ÿŠ¡3Ï!ü„©uÌ)do'ÿÄ×èå܃瑼—v½-wys"—ȹė6öEø7ÃQÛÛ4ÖÔŒ=†á ã$U¿Â~@¼¹b MÒ Y'5îCì=CQÃ:y•w©©8¨àóÔäK’=¾õC‘H¦ÄÅâ|¡NE¡Ál…³¸bVÏÜ7–Þ».·1ª×<‰ å±»',À‘ñB1(+Bösƒ 6¦¯ftDN¥U륺nÌ/®DÁ6 m¶Í:Äã–:ùRœoÞäwšªº”ƒˆFvP–œ¥, cèƒëXV)Ñ3Q:^g†¹ÇÿÅãí®M¢7üy.³“&—–Ÿ^$éÕôíý×ÚŒuµÔI‹¹yc+î$ 3—MÇ{Œ q0³ÏÌ“<ƒé 69]<ŒÊ@Í÷\U3„©r'ËïµBŽÃ“WÂ9»öµÖÊúU3Y±W°ÓEpÉðŽ›‚n3øØËzØî¾ÏBõ¥÷ ™"ñç|¨ôèXjy!ÐÄaS•\iøÑþœì§‡söú£Ö)!Î Vñ0Ó–Ð3îó™™üE««ÞÑaäX=©ÙÝ„ý{µ±¼{ü¶xÚTV+f¡¸dÆë.ä{&”_TÒckÎÂÉ$gÌ>·xXÜ ð›l…ÁŒîÔ iQpç1ä³ÞÑ(xÞ-2ïJ´{WÙ‚ƒóJô†k«9ÍF]æÜÃÕu¦§¯¬$œ|[°!!xJÕ’—e=º}ç–7¼‹Œ³üÒUô˜êy®Gd”7‰!Pc´3ÔÖãËyÞ€%°{ à×I¨Où¢Ùæ'«÷òÅ㤬nè“ì4ÒÙÕV%éK(y¯]ý˜«·5Iù÷ã(¤Jš$¶´¦°uGD[ËZIà{kXÄ­ëDïCΈXüÅäo_2>¦¼£œž¨H+‘N{ËõéRð™9Üîø¡CˆVÉ¥©çXøQ,’Ï åëÐ@8\:Œà•Ì>–ƒ`¬–/4DÅK°³S+Ô.EXGrŠ×+ˆÏ'°)â¬÷ù€‰£:ŠÌ‡‘Ťû’Yjä¤ǼݦÒM'««*y¾¶xzÞ:ŒÕnþlêx©ÜµêϰU²<¯»(¨µfÄsy²Ê²hnõ<¶R¦dŒ3/·ð206í´äKÕŒ_Ý<ªò#»*ÓÉcî]kîëà ªw˜ž¶ˆš -T)ÿ¨8°¦PÄå$#Üc&ûÁ±‡É9©R j>JlFI‰«QOm¡’S=ŸãÁ¢Wß`õë=–6à= }¨„VÅwr´¨ù¦‚”ýøræ>qOKj£Ë‡LÛ] ËB¢bÈ”nç°5·i[ç§~Ô‘0ѹøoÂOõ‚Ìï € ™ìG‚WïS¨µ*y¾UÃçq‰ü(<‚çßÏ!ò ¼‡¯ë~ÁÀèÍ!¼’½³ÿ¹Óý‹ˆ^¼rW=µ±æ‰ 摉È]Zd “Ú± ¤Ð©l™'êŒÏÚ˜bn.ŸÝéˆ]d„²–éþ –)h·Åc쳂!ý=¥M#c<æ£jÙîå"mgE–2Êļ½½„{(TûË™Ž±E*Ë(·õ!yck©<5˜s–¤xã?óÚìÁ>wZ ˜»ñQØ+¡Ûºëa~4þõ)ç,?è863~~õ xéÚ"dÀnd}¢U¤F"ÿ0vƒp|ô–´FyÂ0Ar…Â5Á§té =Ž{ÑWF/ ­åÀ=;܋ߧp‚bTTªt}ë‡Ltc˜¦ØJaî¿áX†:§Í››Å %2ÿ€’÷ˆ#î/«¢½ïÎ`0úØlnB“”|(‰ ¦¿¶4ÉýK¶M²'¾‘ƒ+ñÖýµÇÜ´%º¥tãåÕ¤Ì ™ŌĠ¶U+M™Õ\’§_äû ÃØ°¶<‡(×î?P572ap—¡¯×ïvϗб£|·»ª©c–S«¬vHÜì²¾¿e¡ ?¯9éœÏ†Û—eyW+H¨7é©+ÔAØ%tY: „Ð|½É¹hq„[E¬fü®™š{Dh+7ÏeÉPKO³Øy(%£ƒï.!ô³Ôkµ6€Aå´÷\÷‘¯{ w¤ò©?˺9Z~"P‡.SöP«Nù¯.ÍãâÕ³÷ULpê}®‡º‹ÉtÂ*4hÆ_Ps«öôÛšÌb{ Ù©"•rÚf¹î&ºŒáÑ ²­i v`zæé¯hfä=Ïé<™a¢ÁòçK¹ÚýQ > stream xÚöPœÛÒ ã î2¸»»»»; ÎàîîÁÝCp 'ww—`Á% \¶œ³÷ùþ¿êÞšª™÷i}ºW÷š—ŠLUƒIÌÒÉ(íäèÆÄÆÌÊPÒPdc°²r0³²²#RQi‚Üìÿ‘#RiÁ® 'GþYH€fno2I3·7C%'G€¼»=€ÀÆÍÏÆÃÏÊ `geåû¡˜ iæ²(1䮈TNÎÞ`µÛ[žÿ<h-èl||<ŒºÄ€`…™#@ÉÌÍèð–ÑÂÌ ádºyÿOZA77g~OOOf3Wf'°µ0#ÀäfPºÁ@KÀ%”Í€—ÆŒHд¹þ¥Ðp²ró4o{ÐÑõÍÅÝѼehÈ)TœŽ+þeÀø»96f¶ÿ†ûÛû@ Ç?Í,,œœÍ½AŽÖ+= "­ÈìæåÆ0s´üÃÐÌÞÕéÍßÌà dofþfð'u3€´˜Àì­Â¿ësµƒœÝ\™]AöÔÈòG˜·6K9ZJ898Ý\ÿà' -ÞúîÍò÷áÚ9:y:úþY-­þ(ÃÒÝ™EËä┓üÛæM„øÌèàbeeåáã]@/ –?hz;ÿT²ý!~«Áß×ÙÉ`õVÐd|ûAôu5óÜÀî@ß+þ!²±,Ans 5ÈñŸèob Õ_øíüÁ /€ëÛø±Xÿøü÷ÉèmÂ,í½ÿ1ÿóˆYT•t$þ.ù¿Jqq'/€/7;€‰‹ÀÆÆÎày{ðÿß8ªf ¿yüËWÎÑÊ À÷Ý·>ý‡²Çß3@û÷‚Ðþ7–²ÓÛä´ÿ º!+«ÅÛÛÿçqÿÓåÿß”ÿåÿuÐÿ/#iw{û?õ´üÿèÍ@öÞ[¼M®»ÛÛ(9½í‚ãÿ5ÕþµºJ@K»ÃÿÕʹ™½mƒ˜£õÛD3ñ1srÿ%¹Jƒ¼€–ª 7 ›¿†æ/¹Öëfrª:¹‚þ¸`Ll¬¬ÿG÷¶cvo—ˆëÛdþ©¾­Ðÿ¦•r´p²ücר¹¸f`°™7âÛQ¿!.€/ÛÛRZ½þœe ³£“Û› à­D€•ñsåæ°ˆý!ú qXÄÿA<‰/€EòÄ`‘ú/ây³Tú½Y*ÿñ²XTÿAoù´þAoù´ÿ‹øÞ¢˜ýƒÞ2Xüq½Å´p²kÓ$t‹Åò_ Àü¾Í:Ë[·Ì\mþeÂ`±ú—ÉýKûæaý/È `ùÇ—ó··³ ðßÞdÿö£o÷/øÆØþ_ð­‡àÛž²ü+ÔÛÍÀâô/øFÍùŸÜo]r~[§UûöÄâò/øFü/øÆÌõ_ð-€Û¿àQ÷Á7¢ÿêÊ3¯Á7fÞÂÿ™; w0øíþó~xkóðŸ>è´@\Yt²³m ë|¨#ôd:˜dçD¼Ž¹}§³/…ïq¨"!øxÃQ»,Ö;ÿ=öqã”ýÓk«.«ñ–_V,?ÎJs'¬•ù£ß4†Ð¯SWEwöï ==ú¦Ž–Ö8_Ô‡=oEçèp]®re©¿Ôþ¬±¹€§iùÁŒ 5 ÇËË Ö%–ÛU+Þ 67ß-Géz-:y–B,}ü˜50ÆÎµ;¸ZÉ ÚÉ"<9AqV¸Í=Î<êã&P0ÌQí÷Iá=%цÜv†Óš»ŽkˆPäÉíO_Ë^Ÿçé6£½Î$F>*2¢!5óÑ;.vèª(`õrì+ $+ÙÅ^üÄ’ålsP7]€y§r¯¥×'ºÁo{uãfÔ íÇ1‚tÑÞ/‚«¯ZÇF£Ò2»V­:9;)¦Tó ”òÈB^ζœ«'X3RaxÚæ8ä”T{ï èj ä½$61"~ýj­‚þ.ÔBPJŒ’U®Ü’‚ÎEi'=xkE'Nöu ¬i\Xòni2êAfÚ*°FÎV4·“‘Gy¡`ôt„˜´Î€SÝ\ìYÅúAæõíÅÛk¶ê¯}ïO>í u ù›=õÜúîñÕ[­½ÜÎ×éÆÊ÷Æn´ç­Ï€bEV¡ïÅûr‹T~Vvó**òÄd««â·ÛsœÆªU™rŠÎ¶;£pó8—YósjØÙvÍ÷ä|‹Í^ÐÉ‘=$VB‹ÖØ5fg£…"ïÀlFµæw= C­ß1þ’±ŒÃ5{'e%ã¯y'|ÐtDĵ0º¾¶‰|¨)Íææ‘ã[s8@#BÏ?ç³qVóJòbÙä[ç=€3'÷lÉɽ†š¬xÝ$¿ß|œÐÇ6Ø ëc¬£ÊIDÊa™Ë8JxQ·99ì®PèÓF›‡{§&xÆE&ÿ¨—8Ò57ýV1 i—žxˆ°ÓœÀ{ùº!ܾAÈ÷”øÉ“–¤¬ë–„(iñÚ‰'ßM°œ þꀗç#å‰$úyèÕx%Û$Elõ{g…^ŒFÕñFg¡ÞÐÏšC1d4ó–-Ò†²¤óÎU¡KO£>I¯;¶äãI~˜ÿj¦&{ܹ;þy“ë7­ùITi_kÏÞ|\ègKÎ2àjØÇH½ §¯r‘çÑÙß“‘ K*¡HzT8E 6Cég—¾Mm“Y 0óël‡(ät?þn·ÈºÈ¨@ªœÙ¸ÂoKàæ9ÜŸÝ óÓ™wX ýñ€ißìÿ¾k÷Õ<ÎNYSï'6P< é¹õéý÷Kröçî a/•4ÓN7¶øH»šò4P*CçÐûÕöŒ[R‹ßrÊæš”Yë#ƒÞ3ü2=êù5†‹éRêðM/ßÒ:!m1mw½õ7Y`Òìi¨ÞÝ”ì›7«9j/[¨¥1N͉ÕÂfêðäB GóÇ¡…Ýe2”ì¿‘‚}´—;Ñ¿\õˆ(´‹sÀ·öb³¼ÒüF¢¹þ™H¹¥qÉæpç~)!ˆ~Õd¶Š#Á¨(€ãÃÁfÑé°¥vÂ­Î‡Žºyš [åëŠì:­˜e£À„£j}_W(?ïr2wŸ¦x"cï:gH>%.é šk¼²¨Åüú\ å! å Qú„‡©Ð€Òû™¾zyL&Жo~WzÖc¨]ÓOÞ¨¹³IgŽ #å³×8Ïgý=¯&Ø|„mTþjè®Ü žMÉO®fq’øOOH—J2qb^´Npðö홚M‹]PGê5-A˜ÚËŒý²¯)‚=90®ŽÛ(ÙjØæÌWW/l 0z ¦ú¤7tÜ8}+Ž&³SÚÝ/L43)>|ÓvœÙO3†O3øzÅgUHdÓ(òÐB:¨/n®‡N`üïÍ5ú¤NcòG\áËÏÔ—Ðúgb3¦ê³üi眞+¹6WµL¦¹KL3ç>:g&µFXªvcÓ$­ë‚¯Ð 8PUñ¤f ©¶“eñÁ1žÕä¢íLNl¨«Õæ 4?S—Î=%¼’t/ïç®|á–*gf­€’¤½ÎD¹=Ȫc!Õ %êèU§ÊЩÂÐôt¡õ>‰îÄÅÞ>CƒÄèĦp~[”åºJwuí|?ÍPFfZ'?´VˆGëüR-$.˜! ÿ‹¢Vîq“ßEðXª Æ[¥ñü†ÉèËô”°G‚ý ts„‚…ØöÞV\OAÙUXáó½ %6T&$á²`»J³óT’V<ÊBv´ù^ùã±ëÝﵜt6x?D¹K{Í!fTÉ$‡¾„u£Ô ßÉ*|C†ø(µÜÒÁ˸ûwÃt ™XÑåÝ0~}¬^§¤ËãHNµôm[´ø@Œ¸z…UÅFÞ\…¹¥/yôìª}¥+ w<ðX®U‘ŒPD;D5‰â#¼`ÐveƒwÓ÷Ñ'ï½T—^.‹ÙE¬Úü¶N½~zÓ›S=_2(†§ùëVlæÚ¨Úˆ(<zx%qÎ~ÜçÿP ãf°íl\&[’¬}IŸx¸ h›Ù!JÐ <…U-8ÒŽ$¥nèìÌ¿ÖþÔ»~9<—ÍÞ¼í§úÍñfÿ}(Õ`=7†ÝùŒ“½h_°ûWѰ%ЍDÓî·¬à*Kc›Ì´>K^nš5Ÿ"èÌÇë·×ò5ÜçIc«ò¥PþºÙ³ºðãì^—-óQ‹ç0ü–òõ0\vÂ’ïJ“#ˆpíç&Õ•ÀQ›Òmü*üÞ~U~`_¶ªçÍñ‡ª1¯µÎ÷Oè^û‘î*—¸¹®?!0pFš”ßçþ¸2« =í®Ê4˜—¿_¦ÅƬÄ^áèùÉ»#+`!»xãÅî»ûvêÁÿ°\C´xI+\ÒÙY»Z7 æ¡MP[?±ÿ>Íø@u#ÅÇÈè(KPr`oš¤¼BtW¬ þÁ+6nÉȰìÖ pòóþ@@!†.”Xø‘<Õ&³äJŒèÒÕ×{þ^õ‹åt¿%«y Ñ&g ÊÝæOÍnsóóC‘$mݵ–ðñ6L ¾V®hÇg¾ïïÊwË ±ø;Fú½8¬~L¨hNr†u¡tÃw§¯Œi|Q~îAµ™ã6ÌóÁó²ƒê©'üñÞJK:$ËÆXFªˆ ¿ŒI2Œ4_s;Ku‹º±7¤6GaâuYÏÒeô݉=É Ïâ“p@j\ø+!{>†¹L¨®R¬M»! T¡â¶ØÞ²ÏT0‹ÙÖýå,zý>(Ò¶Ð&UÒA ¤êùYi`Aˆ=Ÿ™û¨ËÀäVicÕyÓÉ@ô!HwŸ¡‚¢h;ÔÆãå»_yybf scá:dÿqò‘¯½žm;a¤‘E‚/ØI  x<Éî~‹c¦£ðNÊ‘VKp̲!l›Ê»ÁŸEŽ=÷öh€LÜùb¬F6ü—ž¼Ïiê4ùW‘5,CW,n±÷LaKˆ'\jiu~~+óxÉî-‡!ì:ϯ³*ÌuWƒáƒÎ¿^ÿ¹¼‡Î ‰i x P…»*xŒÄV¦5ÕÓÎSE[‚ë'ýÖûÚ”À«VÎ"H5FIØ0i'„vpFÊí ÑpŽ\ ¯Ïr(ﶸ‚)ïaeQ Ö—HùtÙ\¡9ó ÐlyÎ9áÑ2½&½>…ï‚“ŒŠ/lõ{™ÙºAø8óB ™ù¹ad¦€Ø6eŸ: ÕV+&?¨qÝWñ~¸OâÅJÓ¨™¸´ß‡º'Íb}Jðâ6|ˆW Æ‚»ƒë¡»¸Ðú Ú¤€¦•Ìô¤EWƒµiiäÉéßàRò©Pdi_}ƭ̲ŠÝo¾ÍçMbOxÝTv‘^€©h~¹|§3èÜ•Rì<Œž¹MÇø5,:9,CÚÎåöMÂŒ?Ón©©/m’«åŸÇÂ=f¾¦ˆ$:ôÑ>o*ƒ"¹;ëç«'Î/4(_ïJf3ôº¯‹év8ü%GãéeÌ ¨÷˺ŒHQ*?¹Ÿñ^O3iö©E·B/G¾ò©¸{••DØ2âøßÔú:Ù­æïÜüðºòøŠ/ˆŸUåRƒ›ô’3šëÖ\Æ~4û1¨~óscÉäDÙ"Äýµ¬×ýþ&™Þ Õ ‡«ÊJE<夬‹-ÿž®W"‹T nå^ÀgdGö™¨ÇÂo":µ‰e+‰¶æ.¤vâ3VìžKˆmn-aÏIaÅÈb^.CÆRqO&Æ„f.¿­;ˆÊé=O䲩ú}Í';ý Õ&f–§xþ”ûËèz¦D‘¾9iïìLŽ÷Nhö6n]è:¾nØ¡püËRCLÿ¼ ½ÄªãÇü| •×>ß^Á§KlÓ£™åœ~®ØWÉŸåàw¦2òìAȰcÄ ú3Ű•öLÑNßýÓ^3nS9BœJá3ÀUØÜv›^a¯uk]ÛÞ>æJ—KÞØ{ÈScßro¹PóäŒ!£^д3>áxa¦Õ‰­ÒÍZ~°¡·f\sèuÆ"ÍÃÞ5bðu0>Ò¦’‡+k1. qèt‰¬e/B#«]arí¤Ã˜l˜ÀËîï.?ˆH]9¤-¸¹Ïç+=QLýÛƒbѱI<>žÂZ >- gá$<ÀÏ(y?M\Üšº–ŠAÄî—ÊÍÎÌNªRæ|5uâ÷íߟ4K¹Y¦?R½ÓöfÅHÀuaán±V Ðõ­eüž²¤¦N-âþ×ÌË”z©–,ç^oä×d¶iSN*î›å0b2ç]âÀüݼ†y0ÇXòkô¦›3¸«é*3Z ËÚÚF>YÉ—x*7Y5¦X>7ü=xš7z1ÄÖ…ÆûŽê<¢íC‰Þdè芗¬ìoˆ?Bäâ]•h ÞW‡?uã$Nòœoï†ÍðÄꢑ˜ÞhñYººy Fñ¬ªáÊ\ ý0'5J* -È¥`ÈêÀFïU·9 X™™t¥h¦9º·-‚‘â2?eÊ3õ‚yÆs/žLß<¸èwȽbzÿ©ÔökBzb«vö°ü@{^½ÌèSÉ¥Ù´Ø4ëG‡ÒIáš\‡þ±xµ¤Ý·þ>·BÆ™­ßÆùا…ùåwI|=ÐÄÒŠÂ$x¨æpËoEÅ·˜q´k뉮Ğ€S5Z¡®¶ØÑ-0B°Ñâ!Èg¥Ú¹­ÆÕÎÏ]ž}½¢¾œ¼ÌdÃØħþfÚœ>­û®pþ›‡qýæ¬%Ô õ!-GÏ7&ULDÝéUœvLáea¸ÍN\„sª1ãX2µ9'æ’$–Ž”,ÌÿÁä³õ Ì:9…'ר—¸¹ï×cHÖ|$·¡ܸ®MÁ˜ñ¾"Œ7b^À׃òοÊ1éÀ%‡VtmÊÔv;ú$ëò`g¬Qèž#ªü¼×q1)Sõ à>Xîeåã:ãWŒ8÷‘­îThàÍI›¡žÕ ¼Œlþ#©‡Tµ©êdÚ ¾ OMYe7½\+²&ƒ€g’³«šE¼Q—pGNbÒ÷,¶VÈi£pÉu5ÆÊru• y5»(›e#a>…!É3U!©Z¼'¤Íe곬#Åœ°í³Þ\ÓîòžVkSÒ• 枺bޝ4¼DÑ#>}@X×ÜCPƒìm¼‹lû¨$pÊ|AÛäù„ö%–ËöáD(¤ýˆi—kc:[ž¼%ðuQ ÂÅD ìø:<£]_A Ë~¬ÏçU:ºê*|SŠ»ßÒzlá‘ëνýΈo+×ü̈Ýò6Ù®ïãl³Ü º-ùéüÚ+“œq9\Q7®—*ЧýýrܦÂk¹Rðøó;ÏÅð§HdIçþaœt<$20(-WÐÏçDS+fCö†ÝÊ@çRhoÊ]É4‹›f•·l |ÀÕ×®µQYbE‰ú´úôzÎÆ!ø¢¶Öà µÿY¶S%Äž—º?ªG¹^BÎçK 5âŒxÔ¼^ç£Sâóé$]±ŒiÕ„„bÓmþíñ¡ô6çw*…?s°E§ÛÉ>ü®Ac~(æÀ¾Ï1Ló¤ g“/VÖwO/æ©´c¨ÔÏŠàBÜØÕiºHÌ¿ïçÂà3ºóídÆ=A}„sÕL³ó‘+õâµp9`">5eëöžmAᔭιƒjž¤bÎÒG­ýªÞ“5ò ²õ yâBŠ@iu°\í_bþÃ>‚øAU‚a-âÔØ`F6ÝÖ'ÉÁÆwF¶’F¤‘\5Øä–ßÖéÙõØKO:¢×Ï^•¡¹BËX¤›3ÜhŒ3»"Å6\5 Á··I•,íÒ¥…~ æôÖý÷ øÚ÷âi‹}eE¡ÛÞ£G6”×"¶ºµHùÏ¡~YÍäÏ韇ñã6ϧí7“ןÑ?Ò;9,Íl§*ŒW‹uÁó\ÐÅââºÐø÷¾ÌLdGxöûYžÏO)#i%_ÎpôIŸ*o3eÚ‘Ê!æÒÏ€…£žön¥ø®v<Ì|¬oœ+Õ J•òûß *£R¹JÑ}Œ‚Ñ­¢ÆÊQø*ø~ÍàDÊQûR¨é­9i1à¹Ä¨,…ˆ9X¥WGH)!üšýBö°¹ÔÂÎz_ÙfýµƒÇv• ªé$öÓRðã4†±”ºç[;ãÆo™vAÏÉ(åÜ& ¦„¼“è6¸ PtŒÖ+LÑÙÇÒ4|ï~ýÉuåS¯ø„!dëÏA]`ÞŒ”ým)v©"¼ÎÚ‘;n†êíyÙfLî§ýB¼ Î\‚+°ZµSûàa{8õ×pjW+†f^à»›²]_UT~÷)>àøµ„®ëuÒ§'»‘Lzr#M©}Õd)_K«Ñ”ÀÁǾ.¢ý2·Ôi¢\<8ï40Fæ³Ú$(úÇ&H058Èà™™T‘ hí%íP S;Ãäòÿº¯OMÿúR’Ñž°+Jä­ óRl²î;auŸ|cNBÔÔwArZ#´ÄÆfã.é£*wì-–mL]C¢ñ·¦g%ÔhUáËbêƒÙ4‘‚„$ϯ^ïÓ”îi…[¥{óé\3‰œZu`ÅɲŠñÞøWÜ¿v·ü®Ù©!S ïÐ쪥Ø/>]F+³ÍTÛm¥Ïàá»xBþ>~½^)ÇÂ0îN‚ ™Ën~À—èJLcPcaë; Ý”cšœóëRŠ7¸/x@±Ìr›†ï¨lÉ:‘ÝæöÄŸ rèCá&•— i4V[vïÊ60ä†b]Ë"šSVP'Ñ¿øÂdœ|ÈÚ-»»($ÜY' „tÖp‡”Ê~7íš~£~«¶î°=ˆZ @O'y×û½-BÉ’qÑ8GhÂï¹'À¯Û8 ó âÁš"¼ELOÂ=Böû !Ýg‘jáé½wIÖF&Åg©X-Üh‹]ÈȃוNÓprûç4Û}›U÷¢®Ã†¨±î³·…Ù}=½vÚ¡ï¦Ô]Ú­;©”Ôz¾¾÷´ÛH¥oç(÷ÃUi¶;M-:ÆŒ(M(^À?ÌòþêvµÅk*•Ó6ãÄ۲Ǻk_´eYï9Ü<Ù±ÙAº›Svç)¬ípš¥ž{A¶¥Jì{CT‚IŸ!–K'Jä.ñ´˜áŽ‚qœQáoõƒ½à4é\d[Qõ2zÏ÷™ê•¾è2ÖÜ~×)‚’95þFPÝ$í¶I„OCæhdÏU¨”¾«$Uz”Z‘ØmüÆPuMã)ŒAh÷%/TZ¾œ†1bÝ/ÍK²%ÕrÙuDö™ÔüJÞì‹èMwªèóª’‘´±sB ø2gèn«¨NhYÑ¡æIûæöŠSëÙÂö(¨†WY­ÌNTf}¼„ó.ø ,±›xÂK%ÿΟP;û+„ápôñ¡Q º =gí2ììÍÊtÎe•bQmµ¦…‘(ü+%<`$‡4AJ}/tå<fé«hÍâãf©ûÇœHó õ)ÅÈÒV#OQ,H ÑÁé}ïwù P1À6Ê—”1ÿäŸtÙÕÚ!9DâÐŽÆâ% QgÁ´¢?ǫ҄ÛMG‚n9²m,òýî,> _ÖyDèé*½s±¡aÙ·§ âðM49ÁFNa<¯ËÞ0ОZf' ôõòÜà¡-ÜŸ ‰¿ôYžäfæà^(ux´:QGTƒwärEÍ©Už‰œÅœƒkEò¸Ç9&…þ§ãi¿6w%4)èÝï³çBòSÃá¦]ä°îÍJö¹Ð“B²“Æ”Á©¢¯Ï‘ßü¶éfMÑøðóÛLd F€ØE½…d‰ÈÐñÝ9½KËÜÕr}“i^{ïŽ6›æÝ¤剺üú é¤ô¨µh|sòðwb!¹Þ>ÚJVɧÀb°tÞŽí®|¥×ºz)ó)a›Ù¦I'1 ÷0Ú@æ+Û©ÎmZ;”tv$ŒG|«K.¾;¡n £!}$öA€…å^«à÷nSåHWåFðyÄé †Äžàqª¯5×øMjÔŒ Ið°f*†À ”ûïA’A"gr ¥1ÌooÐ&-†ŸÿB3*ƒé„Á¹žNe÷¢œÀ;š‡•\çJI9Õ5¥WÚ¨K— #ñu¢:gFlÇúUïdsÕ™ü‡ å艒!üU¦lý¦)ÙÅ‹|¿ÏêŠìN%3(ºï"èPɤ´`ߕ忞l]£îðÕ[·è^¥zêòTxo«ÛÕác-CÝüÈâÍ!À<;VAþê˼ßRDAî>ŠÎ=ê©Sad ï‘2ª2Ï6–rúX›‰—Ô£¼f@ ÎÙd}ÇËšÁDâDz|-Ô?lçÝ=ýåÔI! É.Ú5ÈôyFÅy=ÕN Ãñ–ÀEy¥Ç°æÔ¯Ë«V©@[îHq}Ke°©¯vCÿö^B‘¦~Œ|1Ø –U£øÈøÓT  ¦J9ð×+(•Zwb@% ¼\Tú­Âws(3:·¾\vš-½ÂSÍG ï'ÿ4ávÔ5]ƒ(Èg|XæqÖ«@ßp8 SÄk¦Á`S:…K ?¨ÞO¸î ßWIÞýÜÿ¾þ×RXf/ d’‚ÙoÂø$”ñ³ÜsBøý€ ø¢.ù×;¢­c šfþ™ “N®ŦÄçŸ,±WçÎ,¢,…ë2ó;_N*Éå}Ø4Ò·a@.‘ªAw$¸Ù¡PO-9§õûiŒÀ‹Ÿ<Ž„\ìåÿÊÌÛýOXðû«¨Â”Â;g‹¦pî}´óö&Ï4 fBS=úÍ–n¼ŠUJ—ÂØiëæ0.vðV`ò÷ëˆxüQ›ÂâÞ¢Uó8žu í*ÿÁª™GáÌsy¦%¡¤iÕ¤-=º¡êå“Ö–]PÆ\×I)ö%îA“C1eóÑÉ'¡í#ßÎje“KÏ£pç`j ߇Ñ÷Pi!Ü·¦Ztú½kãÑŸ~äÆ36¼Þþ¼­ÖråíœÅè#Ë|¶àvÛǯâOÃÓðïx™:ìI<pƒÌº]ãð)†uqPLf -Ù>ÃõÚ4fIBEA”œ´|u¥óÑà¨yê@C;NRcR£›z^rvê9åý´©¤‰oj«ø¡¼ßÝÀh‚öF¡A£AFsÞUx~hò‹6ýí“S _iˆÐîip<§Ûˆ!܆ÿ÷µúBä„ÙÑÓ#‚Ÿíj^Á.¬Ù>z2iìI^aÆÎÍ÷BÇZ&^`®F¸R„ ^Ùʈ0YXÖ°“¼t^¨‘$'Òôxk~©ÂøeÄ£\Œ3ð£UÃRÎÞ¼ì-s('–‡IËô…ÿ-½ €'϶üTÀJïÖB‡G3GgGßsA[#ï§'šìï¿!û—Ù€E>c%œ!Æ)$}ŚЂc•Pp&JíðÀOrîð÷¼rI51„Ø|‰*÷Ž) šó2X§NYqü=~‡Aøk~É2Ý©nÝÓ‹>#b„_H}p!­ªVVÑ,®š}Zr[IÆÅû¥Æ¨þ CFRI±Ï­D.½²VʤäqÇsn¿ÿ|ÅPŸBþáY!|ZüGíâ.Vm80"+‰·/¬5U}nÁ·:>VY4­~(tE¶%·&ºïôÃ­í †_+8ƒ=_dB圥¿™:ôË—Ï®vëùƹ5"ŒêšÅÞM|r?Q/ÑásJc›jÍGB¼)͹ª Ýsó‰ŠgF"ø€:áÌΫ*tä¼c‘*¡ÛˆO:y¹U¦iÁ’¼#ÏõÅRåŠkÍLiÖŒ¾ôÔ]Ô–QSðDŸë1Ìšó†ÃÝ ¹àÊ‚“NðBóíæ ®HAhWŽ^—SÏ5BåÞ“œ´ôr8«HrL*Õûü¦Ç#gÕ  ûöÅ'££s»ÏÞ2Æmf"ïZÞŽÌQN~k&ôeYºe ýº ÙÀ’íÃʼ¬1ÂÔG«´d¥àR;ç qo/ü=ZK±ìžÊ+v*ë]vƒûùØ-&gqï`=,TyÁbH<ˆóz.:u;12Ü__BÝRÕPM¼Q' hÍ1u‡ ÑüñÁ4\$ïýZ9Ê®WpÏù$q^&2<©5N-8ÏÛÚÕXŒ´–<˜TÄ{•y×ÜçœîùÙ÷¯VÃÄb³>úÇdád+\ùo s¼øÁÊëô5`^¨Rpų3u_æå6¿À‘9Û Âz62 8´ú•3K_31[¶f¡±o™!?…ñ×cƒÕXùJ.ܸ%2·€å×x«™® ÅmíÑvV¶Oßm B”°v­¾[¯²Û|ÒDe`õ”ïò-Ç5í6œÖƒ°kÏÇ„¹¯u[|t÷œ5S¥'°¹Q`w ý9ÙpŸëAñ)>.„ “#gfHÓߌ¡I¡Ë­ŽNaµuMIì7Jw2#ªÆGÑ6A|ùÄ-aêvÂÞ¸Òˆ*%Î9ÑH¹î—¢½u#f…c-y¸ëž~+«.\(ûBîóôÑGr…_ÉÚë~5Ú8¶u?XîÈ,Ç_¥<¸vrÝéÖ ³¿×£5©†£ª³v"°ã_Uý`GÚ nö<2£â’“ÉüJ4X­Àœ×¸z| @u ªÐƒc˜`ŒHÎD骘†±_R9lo*¦˜/±3jó3¤5F\«çl¨c0£o“ŠÌ…šöùåYàÉ­bœbã›'1à4y^?"jˆ 3šÑ¹¡©'e0k…ô” %+#ý _}ð0.²¤âD{s7ðBCÍìÇñ’PB”uL^ž6í…õb+³ÎL‡n¸yÄN2tª¯Ç³àÎÎgõüQ ò‹rx×yÿúùŽ ú]Éè—ÛÕ”))~ýþDZ…5ýapåñ‚³#ˆkŠ>»Éx¢«ùËý»©dÝżYá‘^÷ÖçðJ0>ékÇë0 ”ÛÌ)ŽÊCžÍÜbڗΞ­PÚ^ò’*N^íÎæUáL†q–ðczÄ‘ ²¤½(Ù¢ Æ6ò¯ ”Å#•)Ë4tÅ%ø˜Ès̨‡J·¬uða8ÑЧ¢f…(7Ú>ë{I/ƒ5ÚÝsÌh£µèâ\hÚR‰ˆˆÄžáÌv~à)'ShS`„—ÄMBéÝä·yP°ö’|/6f2iÈø|m Ú:"%ÛÈ1Åû(Ùd@èº ibG qìÑŒW”™9ÇKÖp!Éåöæ _P¦6*~˒ɤ§&[>/ô½ìúT‡öTzV{ƒrÍ™úºwËŸd.Fù&§0pK<°\Žè\7”L­XD‡XCg̯~‘¯`óºŒ ÐžÊ=±4;)] ;ÛôLØ; ÚHF屋ŸÁåàöB¡ÔvtT‚P ”dHؾâm‹ÿF¸ßë¨z “%7f«¼dÞ`¬5WÜG˜ Mlœ Ç5òàŸºöÛÑž”Šuò/ŸQ"[÷¥¿ÂåµãÕÊTç ¨o†;›I¤>sg™²Z#Ò]“±•]o¢â¤€>è6× zÞ½ï‰V{Éß·Ìo€Áõ]û½ sщ͗W¦ÔšÅF¯oºsj‰§Âl]ø­&w^1Yè,ú”æ0£rc,á-”½Š†h½…¦° ª'Š8‹\ald£gŸé±nÄs(uã>­q2›nî¡ç—`”@MYßÿ dTY­bwã,j;MÏtMÊ(h¥Ð´ÌQ¢8¼ÌÚVãô0v£å6×¹´Ýß¶¤!ÁczP…Û5(Wåa$û ;žŽÚ˜Xrôæ4Üæ\Ôï:~p7øYË%å=üULÆö'ê6¹[¡s°9jÍͬj¬êú«Ôe“—¢›˜»—ïL‹¡Êxu¸‚ôÀÚ-§¦~è´Ì2nN†¹ G–?\õE,6¶¶›§e¾~þ<&Qh*è6Œy±À¶šFœ.]1ïCïVý­åoœýõôbtÜÜ–ò­-½WGŸ}Ëd˜Ÿ·–¾x•rv³×;ºŸp¨\ çÅ2KSr)&:žþ•“ 4(1–ú ë v=%Má‘î^ïo/Æ­wcEÓ2¼É‡¯Ì8ŠPÉF y.¬Ð*#B“¡¾ÈÝ8ˆë÷BXµ='Xž({9ñGÎòípXÿó+¿€²€G7cL=1ª5•žÿ¨ç…â6¯« #¥û =ó;95²û/°Â«9ˆ­“Dl [ýáp+V™Ø6«­@^")€€;¬=š'Á쵕aZæ÷­‹Eò i!}Âܤèj~Ô)±šô(»}ÛB¬Va¶“.Å›óG´:?1_&‹t»µžT¨ÑsÞ;Q£Q^ñûÏyH¾B¨úâER[2G8&±Óò ³ú¶Ã‹$PÍ«d†ÙBtϳý¶ß?^sÕ»¤2˜1›åßÏm ‹º>‚ˆ‡Yµ¡hS—ÄäðêAÚ«†ø\º2"÷ÔrèÉ0ÊÑØ2Ó<ìMô¼tìãCóºÄÄ\ùœ„L<þ©‡q ±OŸhkÎÁéÞfÃl,ÞøwßœT"Ây“a¥,çÃd™"¯qÆ$x×bÝ¢iÉÒ š %NAødšŠr@q¨ÆDp¿ü¡—Á–ªHeðN¡ïš¾õ(=ŽÉ‡ ÛQ¹K‡!Îϸîzli°öÍ]žÉ_!J­Qº…)B){·ÄKé`s•ä‚ɼ€ d$!¬È°ª.~Éc£ûýý¯ïºÍgâ-:':ýгOlühö *fŒ`ß³Ôç¸èTT«Ø>ÉðÄÖ²L¹E*@&’³(ØðíŽÓ0÷麃_ÏR¸´ "¦R˜ƒºfÏ åäOçå6«)Màm-èZÛqÏÄ·©[ò³t ´Ÿ°éBAÛº O¼‰…‘f .²) œcf'ãÜÅïùÄ1U6W‰¿‚Ûá»äû‰÷̳ÍHâÂWËá¤2Š)Þ¶ JÐk’c7Èv¹éí(’]”W\2dñ™½:A@ªÍ`–ÓýP./õå. æÍ\Àø¡¨ýG§ËžâOñßçQôЙ­ ÙJæ6Ero‡U©“I«qO';l'Xó4¢j*ë@Yl„j¿ÇQ>}i`Vßb Ñ|À‚[Üß“Zô!¯yªvî1ÿ þîÙ”+Ð&ÏZy¢ƒ¼z‘µÖ9š-ƒM^: /dü^›Uï17§1:uÁI½Ç¨Ç_%Å—À€¶§îðÙäþ…4W½kD((ÍYJC¬Å$’!ùÕ‘ ¹šØ VŸ ¬`'6ÚlV.gôÚÃb|Üklݹ óÝ}.yN+Mù*uIÅê¬H,H‹‘^ä`,òZ£Lf¯LDGò>[)Jšˆr˜@ ƒM%_ WöB¤žú^K­ˆMj5@^¼‰4[ZN9哟 Í7;ÿÅ•Õ䡃9>é™,ºð¬¼R_tŸ!.òIdj$¶z”1fš^Pðcö§|ð;Ç¥£âØHȘ‡îuÎÑBb}þDʶ® ™¶eˆËÉk]ÎÍ×wQú§Òëâ¸Åm®&$Ö3Æ÷YÚψ1F!Ζ“é{•L>è|Pò"*òù¸0N&KŸTÀÕ²F„£ö5µT'Œ´IGuËO_è×U?ÏÛbŠåeÖ§ñg²Ûâò ôMáiîMÈßÝ"ïéÚƒ i¢ü-Œ±oaÿ¡~µr¶umÄ+9Tyç ð±œrrÿŽ,Ø]„ºfë0*b+¡Ñ­íY‘ª]Á™Ê6<ˆ˜NÒ–âÒÒÅ=æ&O„4l`¬Ò$¼h•QˆnU ßœ"ì! D©»OE'Þ|Ûꦜ‘sÆKQ Ö&L‚2ˆ{ ¤x­žýˆ‡Êc¡Hw6NDC–iØyÂô¸î]Ñ¡k/#’´\.“÷ÁÑïnö‚~8É4‹\xé÷’  3?äÁs°}ÄÔÆ¡ÛÔaûºŒ3NÙØ@Zì¹îý>²f£{þËGãd—íº¶—Š=dþ4ùjÄßµºZù¹ª»ŽÓ·:•®‰³Q2Ü }ß~IpLÞ*;u9+#Ècu)'}Æ«_žgpP…@‚ƒ"@µ¼›deu6:9;â«æ+<ÆÝVmý’¨´rÈî€bº°d°h¥ßÈ×cÞ§C =}tG…’ÇDé°:¹Jü%`"*¿Uqÿ.xµ ¾„Äœ—rxËX·P‚†ù#¤ÜZ <0¸dDŸåÔ“çi•ó‘3‰­·6%‰Q2€%­”º^GÉÕ¦%C÷5ŒwO™ gp»¨ç•˜8?Ò™ßfäQF¬œŒÐ1­0×rõ¡ ñ8&;oå3¿ŽÀ䱫„@½è’â2¯ú‡l…¡¤fëÎéöŠcfIâŸKË;É“‰¦æ%œ¹‘vÝ 5µïi$Ÿ¤‡U­*ŽR€_4a!¾¤×Vi:0)ŒàI ¾ÌÄô-ÆØû\;ÃyÎ@ùXó ßœøt7b”€9¡…•ñ‘VR.ÇïÌpöÞà©ÙMi¿Þl!¹k_ƒy^Ä“½F’žõ÷a€àÁ±ªç‘%t=ŸLû½v/³p{ZMlÇ79^J(––<²ÅV£\·L4רèªñàïY*…°žóÝG±uñâöiÛ²|n`6±ÙFz—_YTŠ¢ê-{!*kéì#¹(+˜ën§aNÆŽø{:ö÷ö‰µæœØ÷^åtX×^á~ÈoŒðÙ‡sbÚý.Ó¬?ŠˆÖ+(Âñ~ ¡·+ÌpX°gA–_5¦_V­ž’!/ÎhÌtª3!¬&ïå>¬û¾xGËës‘%ÏRÛ‡åfˆÚ³£DB#´–¢4€ÁZê ŒŠ¥¼a+õž¬¡u5°?,´MøW£t“1÷Tûƒ¼Â™ôh²Y­<Ï9BÄÚ÷SÁKâï¸1"œœ‘½+}˜tnƹž»YƒçŽwün\ÑýZßoB>T%i³§±½c"öev(=Í{¬YäÆ7VúÊ.ºpÔ~†¶$u‰üb"‚ýˆßb‹‚楞’…>¼â¨t¢š/¸óÁ#!ž˜ ‘L±þShÉ15ÅÐÌ©Äî«°ÔãÄQøÇdy-‰2;&›LV”¯Üçq­(õÓø2V¶MAhË7܇¾’âM3ëÂÉLòßíò’@,_N*@@»bט»Ït±Øòt‹b¤b\ÿè›{ endstream endobj 276 0 obj << /Length1 2255 /Length2 14081 /Length3 0 /Length 15425 /Filter /FlateDecode >> stream xÚ¶eTØÖˆ;w) ¸. îîîRXpww‚;Á5¸Kp\ÁÝ%¸™J÷{~ßÌY¬UÔ>¾Ï=çÞ¢"WRe6³7IØÛ¹0±2ù¢òªrjj¬@ÈÎ ²!QQ©Y¹Ø€þh¨4@NÎVöv|ÿ²u»€ebÆ.`Sy{;€Œ« €•ÀÊÅÇÊÍØ€@ÞÿÚ;ñČݬÌòÌ{;3•¨½ƒ§“•…¥ 8Ó¿hM鬼¼ÜŒ¹„mANV¦ÆvycK-8£©± @ÕÞÔ äâù?!hß[º¸8ð±°¸»»3Û:3Û;YÐ1Ü­\,* g“È ð›6@ÁØô9f$*€š¥•óß*U{swc',°±2Ù9ƒ\íÌ@Np~€ª´@Ñd÷·±Üߌ€ÿ´ÀÊÌúO¸ÿxÿde÷—³±©©½­ƒ±§•ÀÜÊP”cvñpaÛ™ý64¶q¶û»[Ù›€ þ*Þ !¬ 0süCgS'+gfg+›ß,Y~‡7ZÜÎLÔÞÖdçâŒô»>1+')¸óž,ÿñ;{w;ï?ØÜÊÎÌü73Wu;+GW´Ø¬À"¤?2 €ò€äy˜Z²üN¢æéúKÉú[ æáëí`ï0SùZ™ƒÀÿ¼Ý@'W¯÷¿ÿ‹XYfV¦.…•ÒŸè`1Èüo ž'+€.<„¬à￾éƒçÌÌÞÎÆóù_Í"®¥,­¡ÁðéÔ""öo&6 €‰ÀÅÊ àâåøþo %c«ÿò/Wi;s{ïßõ‚õßšÝþ3´ÿÙ:ÀÿÆR°0@ûgÞõ€œ@Sðëÿï©ÿËåÿkØGùÿ1ïÿ·& W›¿,hÿkòÿ²0¶µ²ñü x„]]Àë o^ »ÿkª ú{‹åAfV®¶ÿW+íb ^ a; ðh3ñ2spý-¶r–°ò™)Y¹˜Zþ5&‹Õ¯•HÉÞÙê÷U`bÿ¼k¦À׉3x:ÿRÀ«ô¿YÅíLíÍ~ï'ÀØÉÉØ *6NN€7+x9Í@Í3€…ÙÎÞì3ô˜Û;!ý>Z.N‹ðoÑ߈ À"òqXDÿ ‹ØÄ ½7À"ñ±X¤ÿ pL¹?SþÇTøñ€£(ýA¬å?ˆ À¢ò3¨þAµ?ÌHýg×øñ‚³ÿAàì&˜‘é?ˆ¬3µ·Ë%¿%¶¶üŸ‹Ù¿ ¸bПýFŽ®àÑþc&aþ'˜„¹•è_z°ÄâO°µÅïç<LÀUXþ© ÌÕÒÓÁd÷/ °Ìê_Lùÿ ˜×ŸŠ¸Àl~×=¸ ¶ øBaùWlð%Æbÿ';ØüØýK ®Øáìë~mìl@æ.¤¬ÿ‘þ}Cü#Sw°qýW%àÇ–ÅñÈæíèjï23ùWCYÁÒ5‡LýßÀgÿ'øbq±tý©¼S,.îöÿr÷ÊõÏ@€£ýõ8:›Ú;ý›%¸…n ø‚eñø§ñüC lërú;Çÿ,®©«¸ .ݱà­þ/þëå<@¦H‹óö¦ïB¬ëC:îk…‰Ü™v'Ø¿r©ß¾º’q ÅØ@n3Cõ)à(#âÈ¿áIÉSµ¨Ö…–‡† ‹'ávÿÀ+ÒÉ ‰Š %IxÂÜð^åvêø•0á#~Zç ÔyÅ­6•€è8¦j˜=Ž?ɽÎþ£"iEa¢âÁ</îcÐOE)?e>¸ïߺĀÖAáiÈ2ÝÍÓ42:ÈÍ'wXU˜Fsóý"¨œþßÃ$‘\êÏMˆ´Ë¸ùºÆ»Ù-;e…º {‹£òhö¯‹¿Ý-ÊA'úQ¾åwâVÆm¸ö%—ˆIÌŸŽ³µÝHÇ(ê;¸ L~§á›GI·F4w<.¶”úë•Ã]vü-€kVŠö,s¦Ûjf̈r&W!ú=nX¾ôÛpÑͨI‹C©Pá/-'·¯6 »\ †ˆÚOsÉC.hHÃj2û§ FdX'hE±ÞY#YcÝBâS†)ÚÌÑÙÍyl?q 5]žÖ/ÆÉÛ4‹…s|¦ý3nú:罕{Õža “òÓ¬ eZ^'Óþ¤n“ìúAÂU/íoùÑ…õ§ ÷ ¾>^ÌÊЪºÉÑÐq¬"š¶¾ª±Ÿž¡ ÌJAú~Ñé…x-£õ쀨F6M«DbÜ{˜1[¦¡6]ÐÅÖ#4[}ŽbuËÉ‹ä@ ŸŽÚ£²¾¨<Zž7\VZ1Œâ§,n«üørçQÚ64¸­Â„¤tòÌhú÷½ *Ú³ÔÖëÕ˜wú©¿¾+ˆÞá¸mw|Öq.íúñµ€±màŠ˜P°úʀ䒆“®vÂõ5È à–²?ÍåÅ,<— õÙ¬2&R¿Ó­ž„ $jxbªÒ¨¸Í/,þÔhuv ã[ñ•QÀÈþE¬_ë‹Îë©ĺ\+*Ò1ó±Ay¨+™Ù*®9C5ðl¯Ùú´´ìõJ¡óKq–`óKy/'BIÄGgÊt'~–|H=óNÛ1Ëh4'ÝñY³ÕvOoÆ"°®êÕ€1!¼Œô§¡dD2JÆgüiKøPÎTZŒpæ‡IÇ·Ê!—Ó·%„ö-AÝÌû6Usà½âzÙ2ŽXýNEƵ¯w꣔ß$³cT>Å^˜I³¾‰=øhá®m¸¨$¬üEY8C@ƒÝ¼ó0£j(‰ò^‚;¦tÝXú­ º¨'ø›Þf˜Kd4^SI9VÍk ƒ‚R8ø[VÊ#½7‚èxA¤´]k]'—KZ¼/ë÷ƒ›éŠ+í¥Yo™ —8èF§µïŽ™½„é ç°¡ò¡üãÜLµâ”05Ôc'™IÝÔt{öÎÊc6¤• .Öi c”,È—ž?o­ZÊHpœÔ7jÀ8ú¢§)ß`d+ˆ´ð}úñ˃tšôåWÓ®Üó¯¼H£¦Àûopã£^¢©Ï› ú·œÙ<º4§5}k·þ¿`ê€Î(W}?}lºˆ#ñ=‚–Sýz3½0R:ÝÄÐ^FM­%ÈèYFëV2$žtuƒhô šãiY1©LJžÇ ÅRëŽD)Ü„”1o2=‡ KnWÒiF¤ÉöF|{ðÔMð*Êž¢•R»î)ë,xùåjº‡¦èf j¦̓bÂIs½˜+r ¤…¤a ¦ë» Û?»4rÔIv²9“u-­úy`Z$Üû–w:W%Á-*aU¼³× aX 30¸7·°¥Ír–ë=ɈŸñÐqÜÚ0” Iàá jÀ×A˜-ŠŠPÏ.C9L‚f_“#ž³]É”ÿ"—ô‘0K.r|çy¹ÆÏ›¢jQ»ô2w8”©ûb¹Ÿˆ;hè,è-dá%°úéTA4’vÖ¶höa ¢/³èuxV4m›5Ý«+bU÷„¤VÄ\ßÚ·cX,xãþéAª‹ìP\>‹ûcëUUãÙÜ ˆÍsZïC GÌ85Öùk§[[ÏÜ{¤&âeJ“Ñ'MãóùŠ­“‡-¨rßð|õÅÓ|’ŠÆ|Ò3ÅŸ¾¨)°ÍÓgLîÍ¢ˆv8PáÔšM¿à¾¥ipóxVy`G­aĆå1TGoÙ±ªúGÌ=¹Ú‰£‡]Ö{N;lxôJsR0•˜è/yöòÒ9诘Ä.ÈB/axÉ+šèÙbq Eԯȱ]ŒE¿¨~a|;SóÉ2<³š)ú§cn¬š¶À¤ÿÀÝú‘={¨.„w&ÖÙË&aÞã¦Ù íimÂ3MuÁ¾l=Ÿzäe¯yq €ÚQkèUNZʼFñ[û•HÅ˶& Hò‡f j†h‡@ñ«ÑÉÞª(¤œ™Ò‚ÞÀÊH`ÑÜRS¿Ó…²þžV“×´WV>ä‚â~:Ù}ÎxWVvì•Eòº”5[÷ ¯:ã‚x ùfbœÊ…1ïÊA"VKÇÝp‰ÿ%¹s›dž.pý‰!•%Ì™¿.rñH|³ùM>ºÆç€QÒ‚$óÕȨ¬¨CÊ…Áš\ŸØg >ÇäŸÊ÷ƒƒË`ì&BcðÑõñU)ËÑ»á&˜Ùi ®ˆÖà6ßœ ù|9Ã0Zy×lDØ""0›avÉ¢{=y¦^‚ìû¬wòÜþM#Ѷùä„l$M¡üZ)㦙Öé¹0¸1Æ-ØQ¹õnn…ÅåKƒjíê‰=ÓÐÁò޾ª £Y×BüŠÛ¶hü'Ša7 ‹‘ëuË ßÍm™œC’¢wª“ì_›õhá[ÌÃ}g+õDvpB¾ŸÇí{£ÙñìLR™y\²ù0B“JžôÒmµáJ- x”í«#÷p07ùèp¾Õ5²â}&´aÆA^‡ñÆÃÊ–šëNQA{@>¥‹9ñ›bʾA²õ· nÛ9BcH”2ä¯è„ªg´\À'õ7Æ}Öµ40Ôs¤1©(êNý"‘|½J¿óm’#C^’uÉ©l§B)äL…í&èÝqvEok6"YÎb~Ôܪ.k ¾õ]…’݆­Ǚe-Syä´5s é·“ÛR’/QÈ(ýõÁÆ¡²k ñ|å¸G'+ĉ;¼ÖBJßëE›dÒÁp5­—„M'g¥hôæxBA:‘ʲ§è ÷n^]¥àZ'>R´Š›­…AÛÐ#˜ÍÔêÄÞAÕ!ò“¯Ôd¤ø~—k¯" O:&ŒÆNÙf y÷Ù×eˆe}Å ãgŒ³”¢ŠC>­ƒ¶;g°é¢·ñýžY÷Å{"Q󕵤}G,Ü’Òc»p3f*Èä3ÂðªxÆ#ÈÇå>F˜` Ipú2© ¹{ÃnÄê{ß"Ä—ú5'ÅŸ/¸ÊdÆ$/ “Y¿¼7f»ÛžˆÒágèˆ1 •MÄ¥Q`˜ÃØ}²9¶Õ*ÃF¿q¿}û¸í½ªNrSí$ÛP¶ÌOñóxN?Å8HÒEûAè) ›l‹ª’˜õiŠÊ?(È=–»I)óKQ&\›×L#Ê ß‹K²˓aF¯G§ô.r2OÒp@I@²*ÞU°{ w‡ëNÝ‘àHh`µôöÊŽôzž÷ Õ,òé9Òa64Oâ‹„„ gLdÂó7g¨(dû¬|X=«ñrVoZ‰‘Ïý÷¬[œ@·©¤ö sÔ#£8 <é­CÑùíw¡Á÷|'z ¢…—^ƒîË$9Ë{-­Ø/ï1ÐίãÂÒ*U˜Ocã§5,^/ÛjPõ`øu;]Ñ‘o؇‹Ã5sÔçNtn~ÙÁ©Sæ(øÑcøê–·^8¬Mõß [âiƒZ’J@Ù‚rqû›Õo­ë ´[Ë0ñq-'/àQø!½~ „ŒòçjXZH’žüÑ¿¿¨úííl ÕYªÒEr%Qúø³Î%dð‚W¡ šY×óÑ‚´‹2¶aÎ ßŸŠž 1ª|þN»¦e¥‚R¯WØáLXßaˆ"b!ÙaòYo§?Ãh['œîDŒHó4\%Ìç«SY"ÑÍÆýGê6o{-ÂØØEÙEz:ðÓqæ}É4–¥æjCïšWwÇ„1VˆíN˜ã«W-…‰WÔ-öm÷®VK“5-ØÜìŸÖ;+渢´6Cî&²³ƒˆaû;ÃH+‰7,u‹Çq¿²q­¶®%OÈÍ_@i›>|d¾®`Òc3Û¼˜Ý*uOß ã@‚sMX fz"ÐU1|iæÆñ˜Ï¹ªß×mK¸ú•³‚c›_ù†/íïñn–)q]!–YóØ0 °ã<”÷öä%"NÀ•Ímfä½ÚKÞ÷F›— ѲÑÐ }€Ò­jô`'Ý£Ñ_Ì¿@Ç>½Ú ¢id§¹K°äúüF4Êõ‚âx›lié=íF¹øgs”ëïj,×S´A‡Å±ß\7výTbS[… ¿(Ÿ\œÚsÿ°½êX†Š™w¶Ç£M\ï†\¡ö*üA±‡¾ç­G¯uÁëx®Ä0€ã~Dé(üƒD/kÁsǶæ•~žB?ÿ¸WƒE–Ô&@ÄiìE/+ÖÀóV€¢‘À鵡!ßÜqNqG<^Ö“KÛ¹+ñdcqÄ[ÏñüûD¢“(“N[ÏÚE‘ü-”›l̽­à•°kFùzbi]SŒ“#­¼rÄð!`‡?‡fËeæb8=“t¬‚ög¤ÄŒ0•F0»”«†—C_¿¼ÖÐãÏýNg[˜^ŸU”ãÏP§„Y˜€‰8Å4ãmmˆ{ÿÕHÈEg35ª¥[ûã!AýÞ wêCÄð§öÎÓÐi2‘¨,˜Ðs7sý½ ‚ «ôgxÍWé)ñ„jÅ"#¨÷Q7'w.”8ßXY‰g+©¿Nè]eà±®k=Ž‚¤Xû-Ï·Ï¿+>³>eªÓ •y« PÐyìóL¸¯²âðc©θ¼aÎö†H¨@@‡´UÜa,*.k›)™Ù%ѯ÷û…2ïÆsü®Îãz†wÌaâ†å]]Q›S­É;)úrÆ’}®–¬Éè džæÏ[Ï~/<és³$½ý} k2-Möá+¡|¯úÙ™Ìps™:'¼-…¥ùkF ÞÉKSéò2 Ø\ȸÍJ5åÖ ¯³@ìÁ( iÎéá}rŒ1Ÿ›kÑxØ\¡i¢.ZŠy³4Ãòýµ•Í8{Y3×x_M"‘­"\ü’IB$ókѹÎ%ãÔ¯vxìO[ŒÕ_º8›O¾äËÀÌ,·–'­ÚÕNÏœ‹ïÎßCâÖ wIPƒÄâ.ZS“ÖVœVj¤Œ;/´gC6A!¶hCöìù¡aOH9d¸W¯uD›7®“ïª^‰~qPȬóV@ÑUfVö.xEˆj·?@8R¸LeÛy .”ÕÈB¢´$@ŒÌ3FÛóÀÀæ"‘¥S0‰'1X‹Kßr– õ NŽy6Eîðßå¹ó<ŒC¬U…”mZ¯:((ŽÈÒ§<ÙT›¶÷Ëgšß0¼Sõ,ܹ+2ú½v'%í¥%ê& è'™‘9‘æ‡5Öö"ߨۨ¥ˆ’»³ÇÅ—ªß­=ìõZàÉaML…J# —«zù £GÇíäjüÙHëùÖ%vÚ™Ç^êt+&ð{püˆ#qâ|Ï3Q®pâoOû^ ýgB? Ëé…þæU³?-K¨•8n`ç'¼w¼n&òF¡JìÜúÊè–òvFžY'Kĵ Zèš™nvË݆‚ÅÙt/ÊÒ·ZÄsøCÁÊÉñ:{Í bw$(ÇÁô‰h?ÃBWv¯þ÷/wY ØAxZPî£ñªEÈU%èªÅaº=ìRµÁI’w Ðîȱ5päìZ_õµX»z#Sö–ªÂy½y$ò|VâV¹©PXƒ8 ¾Eº nÉ9ªå€]V+ïóg»Ÿ+C¬W3íwcáGÂw¹q7ùyðÐ=SŸPrr¡…ã™á’ ~¡0¬ ¡pÇÙf„î4ª‘®»^ÒØ!’~˜†&—°H»— U–A¶ïz6ߤF,I¹x'§/é_Ū¤Oöx9?†jmKuù±H³Èhf,GŒ]ƒA,Ÿ|èÛ¶XU'ÎÐ^ƒÜ9ƒ:Wá1ÀA“? âb£JÕÂÓigçLe_ŽMšN¿îò¬Õuq9Plňu>XrÊØîÚŸwˆ… í켙Ƿ™S©t"yÕ5ž–Ù)˯º§HÈAÑ2™)ß²1fÇ}ß.0òá‚)·Ù%ØE9RK¼ÈÖ$jN³Åð2_ñš|~JÚzydÂ5§Œuf²z¸üè¯ä¯oÍ|¢WˆªÚ+øqÿ"~Šç8öL^â³z÷6Öϲc'¥È~ª~⸋÷Ýg!«Iu×:®‹¾)Miå0)«S£rj/Wߪv½—`éCOÙƇ€ŒÉC3ë­­õÀ-·)Õ:fÛ?®àØÛaßiC!û­,¥™ÝØõ&]¶ZÏ+0@.Gf$“•o™jK§¡i6GdOÃß-®hòÌJŠA`_¡–¾·AmKjÐù8Õ n¾¶AQi(§™y‚™ñ©­ÊO%C~µ|›u}.‘Š/ÖÙX‡-Whñ)¥¶—ðâ;y15ÏŽÜf~lÃ×2\2Ë'`Ô8öü&ò¯^ÉfŠ­|õ‡·Ù(ÞÁ±Zf Ãa[y(×ë#»Ê5ñï*x<ÑÂéüúSŸ£sŽ0•.#W¾[8¤ ap»P¿æ>RIÆ…O/!^jv³ á¿NE积Rþ þìÌÖnï#n>¯“ C´×‰½[¤H–Z÷¾0ÍO?+u6¨1 Ž"CáŒõg–çÎ…ÒÊŠ±ƒ©)€ÆËÅTÖ@ÐAæøy×1ƒè« ²‹ùDÕ]¤YÂê.W»@`eX™EñöhÔcü ƒîØ÷;l,“B&9ßå°Szý®yØ‘ç8D^c-Ø&4ˆyiù—³r‡¦vq8¥ItW¸Ò»—ªí×{ĵá,uXv|ûÏ\× ñv‹r÷+”—ŽÆ†½ËÛzŽ_¨½²# Gï 2˜d™ô¥Q˜²Ù¾Ä¨MÖ95î°HlÆ 8훇Ò.“§$gõÓ%ðé ÜsšËl`IHsKô×6`tÐêÆ8Š.p(X(|§OW]}gRðÅÏ™ÆÚ,ô6k½b´{t,¹‡hWÇûu ânüxjC¸†Î&Z ݲi\¯ïÏN f'i }óÕ?ãëÍslSèiQo[’Ús$ I®"Z~òj‹ ‹J"‘ýîvò|yË—z’¦×T5cB7ÅéÁ0'BNV×°›gÅDŽ…0Ót¹Ÿ÷ñ}CWï‡8å/.‹Wep1l•Ë©„Ô÷ê>Uµä¨×,ým¡wnqÔ€*QÄ»P |ä³îl†K·«\–EÐm7)Mʾd®b)†“óÎ,ñPiƯêÁå"ó0=ˆüNZۋЮµï9ˆ® 77TzÈVlÃú$Í8êǿƢH Gâ¶Ÿ-rCæ:÷Fb™_l®m†ˆo'1n©×JÊÇAÓ>3ë¾Bk;{j¥šü±¬œC,€»o£¿ï•ñ«Ç}° ç˜’è¼²ßøüó]ç¦æìŸøéBtÿŠ0ËiP89¬•Ï  2¶Kê‘%Nè|R–Ö3>i5.ƒ¡a¸óòKîrñç¸h’P¨ÖîéÖqëx(˜ÞÙØyXí¨Ð“ 1ζk’dPr-gC[Ö§¤­`Å’PŠÄò"`Ô€ëþÙ͵BYnNλ‚±½õ³"ïH9´RuRS]Ÿ¾’Ì,FɈW\¹Ž}úÁŒÿ6ûå7]檫YɯFÝ¢’eÔ¥G©cÅ÷aRóø1¯,æñF`–>áVûÚOq›÷1¸k¡zµLIÔ›Ô‡ªÔÏ´Ù¸Ä*è'H(]ƒ¡jŒ ô´ÜØj)©A;‰ëÁkÖBµƒ»U •²o?¯†ü0±)+Ï£r;óÑ®ð…¼íá'v(ŠˆK)S Bƒ{ð¦r6¾rÿ‚]q-eÜAz½ËÞ4§Ä¡ñüÁü‡Ç.¹£„Ú"c·lxhráa€˜Iõûyü·|ÿ±³÷h×?ÕuCÆFüHÇbJûˆfïºZñ ÈWÒ…†Ç…:8¶Þ3¶--0øLk›ÜíéÇ©Ô)ðDˆ¿Šo*¾± ½tè™ìß]eµÑvO©h84©õÇGüzýMÏ »r5—q/ð} ^´¢øMŸÔ ”ÈŠT™E3}Bd´\wûœØp¥b åÛbœÔk‹£‘RÈ„Oó ñ äa߬Ý,³šØÒ´©¡GZÛÏt¶täÙ©p*S–Ï!æÅmr7Oî%ð÷ŸÃ]ªÒp¥Òò×0îEÃJŽPÃLéüà tdpõxíS6𮝖Äy±xF/ÜP©^µ#S$8M,yëšúŽ™W¿HGô ¢àøè›~Zv3ï“yã²·!VŒb^d½ÜÊSثϕ£mN ¤Qï[s ¾2~X5½ÄW.J}«ž37§SpD槪G¨`ª÷Ði–}ãk/Sl A»oŠx’hûÊp¼XÔ`¼úqÅÙÛ%¶ô†…Ö39ËÛ—M¶™ÐÔSСÐd€ÕSmKŠ’Tå-@§œ¤†Ú‡¦qÆê«nâ/|D•9`Õ+§zÁ™4ă#ŽÌB%zÚesÃ1˜ÃôûœÒ &ª¶lo>ëtÓ~>^:"ë¢Æ~‚ ’8f.«CN®Ù·÷½°ÏYBžúopÖtÖPÚ“/¨aP '‰eÌ;' ,aHœnÏ„ªTÚQÆJE–¾Û.Å RáU1£E™E«mQ¾my[õ|‰¯éƒ ÅúœTñMÑ“.=”–…¬·3OÖ.È­&”×/0F{h{ûóËV—P;NH×¢9rÆ=‹Îa…%cj¡ÎZ¹¼i<ŒO¾žTiW7‘T‰àV•–Õ@´Kyh¡ ZÁό߮â7>oäí ¾ ‡pHWõüðÊw´EM ì–Awîøé¾OüQD(À½µýÉÍ?ù溽ï;ðK¤×·ˆ^^3})Ħ۔QFÃÅI¦R6IzƒŸ j!ñ 3Z³Ê‡† çŒ"K@2…6š`"ìÆéžÌ‰Ž¡»™¥.VšªŸ7Ef}K¤¡é-¬HüsÞX<¬;ž°W4Pü"ÆUmŸˆb’ "áj¹ŽDȪy‹{ u¼´³ç?íÓaÀ:)î4G«²ZIÉçvjA­ òTXÅÑšš¢—æ‰@¦vµPèu"ض5dpJ«Ç\±SÖÜFDlSüò‹ÎI6*î£%ÚŸØ„ø•„oÛfIÓÚÍØqï(Ú Œ¦òöT]ËÙ}„bñ{ƫц Ø@tÞè›RÓÇ6r«åkux'nå°Bö¾CpR¾Š£<È)j×è÷7I½aèŠÌÇ5h¢f›x»'Êìç烶 [·È&ã¬Wð1/᥺ÊTZŸ=Mؒ¹‹r䕯dàësó¢~<$Cór”ãdc]štÍ%øFKe rB? ¿‰UIïLCd‹lë@©ôŸÓkÈépaI?þnY†ôÁ«œ ðXýPñJi0{N6…•Öp¥üV_6zëZÙ'^ch³N’Ì«B4çåš „9V:jÅ [Eþ)ñVš9T¦w/"W™ô<Û'£g[àmWîÙ’­p;²G‡ÔqH•Z]ß@­ªâOSù”~Ýrfï¿^ }ž ßGj<žI«fÌd¼”;´[§§'$K&tOåàZ¤©yíݯ¸P‡Ê’°ú‘¼Ã–æ%pp\éB:a*¼Uo+dœËl§ã"ÊíØ+wÓWGûžËE5›(³§Ue«3x³Ž°ÇËZU~R_q›úî} {Û¬CéçG@;—ú +)òSØ~=™Q`¼Õ1·Ù'Í­JrÍN-‹>iwÝŒÞWšYI*.<Œ{éâˆ`ÎL&Ä€ü›oçÇtæ0øC*ƒç_3b3mM–Ø#VvØur9ûj)#R_>_ö~É==¿Ü+A{wËÆ¦lZ/ ¸¶ÒÒ½týBT£ìïÝ ~ríæ¶4§úÂ8Øäôô«£>JÑÍáÈŠaéÁ¢–ȪÄ…±íúÃè}°Ây„’¦ªòàfJLÌRøŒÜ#(ß©éêømj]ê'©Ó¦™ÂÄ÷æúo?ç/±êEÇüÆü¬gÍO×T¦X“.iFdß/Ñ¿¡¦Ï¸¿uq»Cë3ÿàNl–~Ž‚ywðÆCÓ3¨ƒ¶."+·îgșӖâ”èya®}ÝfAƒáj³˜ãðX;%g?J1#[ p†¡BíX‚ÚÒ">½¿ ,O5ÅLó:Ðífê·!ÏÉA'Ä\1õß8æ·Sƒß k†,ÛàtỂ‰“auý*ç×É&;GÞ)¸…;å@f/mâ—’/X'ŒdÇ;³"ùâÈÔ9§äÃÁ{œpÈ*}ñ N…´@¶'Ešê¦fÕª͎¾70w£„ÔNý:œJ¬qÓ”q·ê+\#XB ¥N%rš“WýÍ!à¼òá½¹OØk„>Íj™-¥‚‡Ù3~¡öØ7°Þùê]­ ‹Ñý²;–¼„Tý¬Ô Ë”µÛ1„8æfMjs”Úl a=9u6*}ÄÎÉݶ¤ªÀ#Ù"ãŽÆzJ­3ûAÄEÕ¬F™*»Æ<,c<ñãëÑ,ªÆó(ã!¦¡¼žãíãß>V½Ç•E3ƒÊ+,ÒNåúTòÑ~¾®§#竆o1ü`Ç‘YŸ¿k tr,½åWµ§\½%tFÚ›J ( öá/Áhà;ÿ¢„zÉZôÚ9éï‰@ ‡¤¾icÞÁ;YôåýBIƒÅ®ÃÕûÇDz„9Šø\šÐ+Ím;¸Ñ®8=¦{8ƒ˜‰Ur<µõ£xù$u¹€p¤’Ý..{Uáe ÌñÝ"Ñ-uÓL^¨­£¢j½øè™ó\r$¼›µÆŠZy‰·6ƃ»ïŽ4³`ަ †.MŸû¦°«´'¬"9:ú\jÃiÉÑ*lÈ“RaI¨N¿P…¹ø“ÉÎG¢9P•@HD—`ñ‡jgP½ïûÊ…œÊ6Ê–/09ö„߸ÄSýRr2ê¢zÿ#™WèÓAóOåÈä°ù—:b‡ŽüÇ'Ûdݧ‡Š™üñ̪¬â¬’V<­Òž£™Ù(IÌÍyüëúÍ;áF[7.ûòŽÔà\“™àøÁ`îþ§ž,öÜ>½X©’˜wÐaî¶8lúÏÖ Lbçœ) Ç$– ±ÎÎY¶Gßw0*¾´£³T<¯mH©Ô0PMûí7¤9wW‹ÄßÅàèæe'EiUë[rë[qâhmEJôJý‘×Á½ÌͬÈ&ÁLÚä#-+9™/ðƒ‚{•ä™s×Ä·žz.€ »1Hø¿çnG/ܧóåìUK™‡Ï€W ¨¼GE.'£»ë‹Ó¨ÄùÄn¸–s©‹*?¥Þl±Ê Ì‹8X›b¹`Ò-c”L÷WËf­©kaÛŒøZ†ÿÌ “^ç‘S %2iwQƒ‰¦#ÑÓoyŠ%”¨GLºyk-+zp7dé'F÷zéÙîÆ”úJ:2ªT1zÇAK2‘Ïû 9x»£´­ïˆ¬Û§kÜÞˆ¡†–GÍ¥|ÐD Âg¯bÔ>fRW&̦&½R—ÚRqztáAšs#Nßaȶ™Kª&§ª HC1p-̼ÕL‡7~¾ÔÝvU+wÃÛïO~LÏúF˘`£¬1P Vè!¶9®[úS Æÿçà§=ã(”?ó²Õ!']#Ú°dq`á·ÎùC­¤W°äÛU-[¸Ç=üÔ>'ÑÒ¼}ŰëÝ…ýnHÌ$„3Œ*5¸_"Õx1tNE‘«g{^GãÔÛ¸]ï¾÷Ö‰96"Љþ\8b©27«¤¬m.1t½c¼q<·C9d$¾w=p}3ϵ801ãE;Q-„RL3áD8=𮻤õ«H]‚ý ²µVYU_[Pî#Ên¼~éùwùePUÌçÔ+ð–Z¬ñb㡳µ ))t@ Ø} ÍýnO™•Ɔ?¢¨_-Û³~”i ‹Ž²d…r sã´, ,ÕT`=6¼|»ƒÊûªëÙºg–+ÜRÖzµõg¶9Ž=™K^ébeiŽÆ¯æëKaFÒ›¦€QVEØÃS®ç×ÉhXÚÈS¯¤Þà*¼y»öZ}5qa±å«ËXˆH¤D=©êHs‡ÒP™oõAÏé:8 KF™+¶êÉÆë0®;fF·Ëå ÙØ¢è¢< £¼œxuûq LSËùj—ÜÛ×”ZbXƒ…«gKÇÉëȯÑ-+4$*d7„a{_cM{¨Öv“ÌU’=ÜåK_TEÄ;üpΜŒi‡„L-£Š>F¶1Â(\Ìäˆ<ZèègpO0Τc?¸d‘“*+YP~¾£î"i¨ÍK^;«¬ö—ù¡ÅùJ(SÄNÔoz…™$þj~ÏÐoý¾Ûš“q’$I¬DyŸÙΤ€Òrb'#ÿòi܇7øƒèKá+˜æØ˜æPŸ—Ê„|4 XèÈd Ôár<úéY•2ó_)¶ß¯Ê/Ù(³hÜF–CÖ˜<„ÃB%µø[Ï]%ëk;¹í8_šHu;Eú/ÝtÜ ¨#8Áx¾OüTîÁ/Ë|ÑÛ…`OùV8c:D¥9·è±Hß  Ý7ÏS™o°ËßD ­;RÓ,IŠ–x·NmÅH‡KÒÞ•‹4·•;! â«*ý‡ŠñG#žYªˆò¢•¶©#ŒÓ÷;ˆ©">ÁÂﮓ¼¬ÇÊ—ÆDCÅû­K$KkŠ´Ã͹°T%¶;L‘Ê:Žåûdút³EÇÙý7S£]9¤¦pgXolççÑH 1æP‘Û‘š¼‹yÅŸ;Sìñ­» aöà<½÷–h "úY¥i=ààä®±³“[ˆßS¢súú§dß å”õ¬ãI{ÂÑØ¯¨´£ËºÖúÝ& ½„+?7CûÕékËìµ|è5*çõ¼ïB>(I­àg5 æX,ΜXN¤*|·Bÿ]æF÷+Ä5ξø9»- jÏ¢.%zÖÍGt¦vŽUvgÆ`œ;nHy¯H‹¾‚ W:«’úµ¢h,/znzÑñZg.¬+@ÎeBЯ8×m‰Ué†2¢+ùбÔé”À-âB±–Cãì®´¸¤E•èT|7R㨽e>«¨¯É8íHLaÅ Ö–Ê‹Èá»<¹‘3OÒâà"vž}¬äf》uƒ‡ÑTÕäS4Þ²¹Ô\Éd‰ŸTAȵÛWÉçïÚ}“\%;~Âà»Z{O9Þ%Cf{­v#Õ¼æ¹Dà±ÚýáÍ›[ÞŠÉø0d¸éäK]„ñYdïQyR :äÃežxÂôfI½blÁà‰æQùó½FY¬Z%f“‘„–E“6¤óÒé—(jüV][c2 ;öªûZ£“¢Žü¢‘_‚4Ól.Æö6Ûë¶Õ*&}¦Åêþ°¿¨½(¡ìdfh‡%Qª Ð HsˆÕHýBÿØtxÎ?©”©djNƒQ†ˆ€÷ôËÌþeˆ6§Gû£³2ù¢W'††©G,R jŽ% ¦î£­++ì§/]Q!žÚòT‚ªmÇŠ—žáý_Sóù¢ä~1>ƽg‹nœ =&ÌVœlŸÝ:·–NVÒ럜C(?“;Ú› „™ÓL'V8Kc¦üRoôÅÜ ðù·”*–ÅÀìû~úÌ=‹´Øœ¨›¡Oårã]ÐY™×ìRL¸WG´Ë@@¬æAý±OÔ–&_Ç¥•“õþ‚ÎÔ³Ç :t—¨Ëˆa2Vq7“iÚCïÎU`ªiaQã~'žŒ3¢òtf¶KöQ¹ ƒýF½g¨áƒõ!°p íª÷F­Läu¸WqF¯jïPŸ.Éû‰€=›Å ³° ±“qËð¡5£Ð2nP‹·–-–èMõÚãê Ö — ƒn–tLCÌ•‰S-ÿ%;R1Å7R¾Ì[/š ÁÚ l@+GW\±Á‘õ%D‹ýŽ nä’¸a-•Êd Kfðie1ô¬•lÿûæ­I*?øµÏ§oIVº×ä‰{ð4q'ד´­ä©)ça F±þŠ |qL¢ËnF/3 C5TÑ©0Ç©¼}«{¶Ä=J#ç2«ëéDõå®Ã,—_GÞ §ïò7û>¦ÃB)AsæðLvåšUqtüŒmiöVÈãÍMÌÕ.7²§q… %¦ˆsJÛ½÷á=nºöµÐÇ©—“sNŒ|ܸ¡¼p™nõ¥&±xlØÌPú¾ÑÙddÔ`wˆ~-)¹2%Ыè•\ui¸mÔ¿IÞÇíRµ²Îw1Ù1w_plیͱD{I¬ø6ÜOêÏTª–vÕóKò›Uþ€ÝÇÜ…ùзkuÌð3æ$› ÃtvogWnbÇ”d¿w˜Ÿ³ˆIÖ8Kë°ÜT˜¤¾Škˆ½¡Ô÷wcx>C_¹¿Äµð5_álæã;Wf€f<Ë̹Ý;ø\†ˆw熗å_L¸®Ö™Ñ©o`sƒR´ ÛZ1”îc-|#¼ª‚ÝT’ƒ¨Q-®‘“ðŽéµJ\±›tc’ØòÑ!¾ïJ„‹åÊÇ/kõÍ©_^‘fþܤ4÷á9ß‘§³‰T»›˶®RÇùù²U¿Ê[iÝÞhØ¢nô¨#²Fj”‡4u³#L¼¦šç0dÒ·¯ßÈ[Û™ÚÊåô×µ%s;æ í6 ñ@Ùæ{Ÿ¡-'oN¿Ï>ý5ü|WÒ±QvÎüöJ—(e{æn' ¥¯LSPÐCb ÌŒ.c2¯ÍFºu80{kØ_[š¤™8Ÿ)Ú„¤K{Šõpó2v\=ØN‘}ò$†¹N’y ˆ‡Eá}ºcã|ãŠ`S©ÎG­þ àW!¥/Ü)p¸­_$ÞKPy’‡èŒ'£ŸFìÀ¿-Þðã>C®ïÚ†É YŽ8wõøBw—×0²ÁûG…±Ðu{ßlÂó'2½€–…ÎÞKœð"w£CÀ°a"}‡’…,[ƒ4‘¯ü6á1 ɦN»-¬hHá+óÔ®ªÌû y^¡ÌŸñÉЊe*št{ÌDyÎc‹ewÞü®áü,ëþI•Õ‘óKH Ùá­°#YJ:I`Þåv|Ø[6ä¨yt¢ÃIÃÑYüáyx6wÚ'”TF“> stream xÚxX“mß>"R’Òé@:7îîncÀ`lÈFw‡tI ‚tHH£¤€´”(HI‡€€úMŸ÷}Ÿ÷ÿ?Žï;vÛ}¿óW×uþî{;‹¾¿‚ÒªŠD ùA@I€’Ž‘…€@!Bvvcý'd7…z `H„ä0”< `4S£1D$ é €„ QI˜$%þ&"=$Ê`/˜@G ‰D@Q„ìJHw_˜“3“çïK„’ãûíPpƒzÀ `@Œv†ºa2BÀp€ƒ¢}ÿ‚KÚv—ôöö»¡N²Ü|oÚ`EA=¼ €_-tÁnÐ?­ ²Œa¨¿ FHG´7Ø ÀpŠ@a\<P&;ÀHC çEüEÖþ‹Àø³9è_áþxÿ CüvC H7w0†p8ÂàP€žª¶ÚÍ#~ÁpãöÃà`{ áwé`€ª‚ŒéðO(ˆÌ@Áà¿zü³Í*%¤›FþªOæ…`öÝWðÏáº"Þÿ¿WŽ0„ƒã¯6<ÝM°‡žP å? DøoÌ Šˆ%DEE@èCÔâ,ø+±¯;ô·ñ7Œé!ÐßépÄ´ „9B1„þ(°€öð„úÿ§áŸ+Bàƒ öP'‚ðßÑ10Ôñ¯5æü=`>+ F~ ð×ë_WÖ…9 pßÓ± ‘ž‘’¢)ÿeTTDúüù…$ü¢@ˆ‰‰ÿG ûSÇøj ‘‰¿ÊÅìÓß%{ýÑןáü3–.£\(€ëßBB0o ÿ³Ü»üÿTþ+Êÿ*ôÿ®HÕÿmçú‹ðÿØÁn0¸ïF¹žhÌè 1³€øoªô¯ÑÕ:À<ÝþÛªc¦Aá„Q4?è¾ðþ_8 ¥ ó:èÃÐç¿TónòkÞà0T‰‚ýºÃ`¼€Àÿ²a† ⊹‹ 0Òüm‚bfèŸyU¤Ã¯a€=<À¾„˜³Æ¬Dþ ÌT:@}~‹ (€@¢1.LG¤ᯃÝ`OÔ/ô7 „<áh˜;&Á/ìÙ ž˜Ñû- L)¯Ï9ê….Ì"!R./":.ê¼ù7Fq—Wºb“-úcDÐsOýµñrÔ¦*Ú9ÔÐŽ<ÖŸ¯ˆšœ `à9¸óyð:/t)S½¥Ò¿¦xÄ_8ôíäµ”ÙIï4.ö8vü0Å€‰ÔÎ ÇŠ3Ílö*R¬Gxóùçâë.Ô¦¥Ä.À PzFÆ’É€NPÈïÚÅ~dÁ¶‚¶x@–xÇÃ*ËÓXÃdüû,ñ\±#žî›ïšUböðÇoòú iõÖŠÖDQÐQúi²ä¤ÒQ¥äi¾Daq(Šœ)Ud$³Ò•¸Eq,Ý´[ú>]4О2(ìÙ䤫z[/èm³gRdV&¿gé˜Ñv=×ø› 1mKj¬3_Ýc6EöZ¬³á¬ÃZaxA/53·¡kñ§Î4²¬üÜBû}s<^P¸°Íç Ú »„œu† \|ðÓ3EQ}8³ûˆ£Ùû+é =;…¾ÆM]—Ï‘{|_ŸõêÌï¦!N·ZÞt ÷D!Ù˜¨¦B“Í;svûvR^jˆœ½#7åQE×kÍ!»m¹“ò“¦ƒ òqôÛ»Ô¤,å%æ'X“L…I É猵7±²t ,$#B¦ÉH(\*˜ž¿¯}ûT&¬E&C<´8qtCÙ`/‰¬~í¶AÏ·[6ÉöG‚Bø6Z]–ÎveºœšèûɰæÝä«-E‹©Ð#±Ù¥Å&Á¶åA ¾'~Ý{ËÔi9Șõjb¯·“3cÓG½éò çˆÂ~XMr5¢^Õœ$Ò¦ôq€*ÖÄj¬ÝqÚ‚ïÑóÖ±ùôÈê\êþz¨»ÍV“î¡û÷ʇŠÊ€;¤œó‘Âk5:!þ8D1¦ÁÊÄ_Åïƒ'šÂ„fXÎŽÍ„ëp9­÷ÔE»VlZâÈ.ÕÜÎz?R´Õ3 júAÎo–„©Š÷ÒþðIÐ\Oô¯²»þ hwÓÚì!ƒr˜N+nŒÅý岓¨ò"‹¤™·Eæú*GÚ×Bk¶B­·kZ+êþ„ Ÿ):Ê¡ÎîOº€v$ñÅÙ¥™î]&ÕôãÒýìÖ¤iìÓïékYð»ª?©k“c:ƒž3«æÛ)­ä†+]XJK{¸ð#½gἎã´y:»ûwÏárŒž¥Öƒ›`ÃêÏÖ•nQëdï`eZ•zCS‰,·ã_L¿ø¾IûÕü‡-zÚL*»ûí—¬AOy®{•hž·eÓ·Mžæù¬}ýjÞP-‡ßÌè:÷Á©fÀ+žB§ª›)”÷Ô¬J5°¼7]šøA¬Ô6LnŠlx¼èÕ–¿—*×Ó~W*=¯^fäƒ,_ÄÀZYU×ÖªoÍ94“†û·¼ •2æKpò ” éH¦åmùò©GòH;]ïP'Ø;Â74.ôm°+²½r(J4†Ôý\¹ çÂá>ß“™ã†B÷Œ†Mô5ýµÆ*FQô‘D£µ…û‚{^¯“Æ­ [¿Ï3·h³ëÊïf«‡G¥Ý‘^ÒçLÔ|6²ì•2Ã7ÀÖÈêlR¦#wÚ°+·’æÎ3!æE2G#³¨B®©‘¦É릊Ø gÄo4ˆ;¶Eè} J£Ã[õ¸#‰«bÓŸ§ØI" "ÔˆŠ/S ýžQœœ8nûGOØ+;8•^/òÏÎ^UÌÙ™ ¼|5°räT:ñ]Z­³|>ý\ÞÂ,žV {ÈN÷ù†VÌÕ}W=šýUò³&€Ôòp¾úH•GK4Ÿ]±ßeóˆá?¼iõžwïüòåØüÖôd{˜ÍÛKŽø“ckrqxFÒúKT+GDÄkÁäÕUîã¾0ñ öJõÁÂDà uh§k¡TÒVÆ•§ –Q,¡‚Èq¶¨¢ ùb‘b1Ûx ¿&;©~¼ƒ5rò¼‚ÂZw”?8Ð?fr`H¾`:“÷cŒäÆu<æŽZ\kœþ4ã_I¬[†uèÌŽúV¹}”Àj[Ñ-¦[ߨ6}¿>6Þ ®È±>Ý’›á#ÂÞ?¶GÂÓ²œÉDS~êÅxõØ–án¿mH6ØÜÔßg1p?'xÌpÞê!zÞñZª¤Á2s«{`ÝkžY•ȗ<‘ºI=Õ×]Š^yhn æB·Ê›<2Œ°®þ˜k罹$ÇÊýXâͲUŸ9*š0Rb¨¿Åöùe«äÏÎwÜC‰*÷æ·J8«{7ºø[›‰Þa 2ôØ¿$F/t3ÄÏ !r‰¬µ¼¢ï+Ò&uº9iGR.`WÇåj«—¼kÓúpzÒœÏõïa>ïÜ"ÄÂ7£´½¤ ÷!…­ÎÛøk;~ù<àSÜkÎÓ¼ñú‰³ ´QùP1©ƒÜÇzs* ®p‹ðȾsW]³&‹WûªóÒ[Š ™Û® ‹½xˆû\ïgÆm„æÙÉ€Ø W×{níXñc¤Ûð·päô›½ŠûtmT5…¾O(Ò’Þ{ݽsÈ”\\ÀÜ—®ÒI|÷báfªÎÀîœ`Íõ.z>òæ•:4'jA_©qÛK÷`“UM4¨tø¯ ߺ T¨@“ü*>7`³Žk½ÕtæËrœZy5\,LË‘ÁŒÒoÉ£Àv_滫?gåý‘(s@)W7moF}Ô!Ÿà‡†æ¿)í^ˈ€ÖXåÌ÷ùI¿­UöÝ›ãè-.[NW¢«ñcÞ¬ÐÍ·(¬§·¾îDûêÇÔ¯ ­°ŒC}Pc:EL‘ú±tP[JVÔ­2$Ïþc‰øñG…H—Z“}­3 —Â6>ÜÑ ^{PÝì‹/D½¹peíC˛¡çÍúüäiÇq'§ô„Eç2.”$xA_ Ï£"rï±Cô—bdb¦¿µR¥,Û›ª~N,1<ÏmXÓ·\âÞñë~Ý-­A‹'õky8}LpÏãÙP„ ÌÆÅ<ëÙ­'Ò6 :}óóê˜#¶f§Ü`-µKn¯°}nðÆL¿îþê¬?‘y•Š-â+`žt“qjô9÷—ÏŒEÆ…ûX¥g ­‚"תª·q.ƒ«Šð@?šJµçL+wƾLÚW˜l³ûELÓj¾h7/¦/êFkøûºž àLß‹í·J»WÌ®|Oå]Xª»{­*OUù×ú¥,¯C½i‡1„âv1ùÔMG‰pžZ†=sɯ]›²ƒ9’<£KÚLŽØ£ T½©àT)^&ªv`4÷Ä{…T§Þ:-žPs­ìEÓfN¿Ë“· Lï¹ gã·fŒ;ÒÉ{³“µÇ–I4w·m”aå×V¥G’(é'Z„zX¾P.™D_äbeÞÃý&b±ë8i·óP?4ðüj*@æg|æûLˆ^€~/î*GâKsbàúF!¯“»B_ßk™—*!r¥¡0±Ü¯3D©‡™µMçÉ‚¶UˆZG€E™hü•¶8lVFo(‘3=]7‰~ÈSèú¬›ûñf4²Q¾®~53ô§íÑŽÂT°pÙªÕv·"³éiÛ‘„JQu²‹ á#k9@ÕÖï[µ¶¤"Âç%ªNÓÝ ÷Ÿ(fo­çî)”WÝYé|øbüh„ôîò«¥ÉûMxqÌÑÅÉ çúŽ´ïªOøfú DO×:n1>õ(.³JùøÚËîQGSý÷šÓî`Ÿs5¶Ÿ %;ã…­c·ß¦¿,§ù0Þ¥:š©«ÏI “¡­yíÈBsã.ÏHg¥®à¦7n_¬¡x—ü£?bHK§|¯.‡†ÛãüÏdÍYÞ¸e@¿ ËÁ©IÌ~²N™*©o]<çJ3“rKÚ÷"OדœŽ·SRÞÙ/P:pˆÖm,é‹ ½ kqUÐ°Žæ¦Ñ¹ n6*úüÓ"sÑD–€‹›4ÃQÎc“È4˜Ây‘Ç"]¥½âx9 ½<Ö_üø¼‹y’~¿îÎÏ…JÞtIMH³ta/§ÛJ:Ûd;TÂŶ »Ì ãý×uÎRÝ\µÆÃ•[Y²$Øý壯\¬ï N„Íe5 ýä”÷Cdt/"otÚ#ˆ™üÒØ[pɲ.ˆól=êu-÷¾@®Æ]`}„œÄª¯õ'){¯¶^X‡u2$Sí:<õWI¥.é}äº@·’M.ÀFeº²éLÎ;ósesue³¾¿[K…sql)¼¬%kWÂ1Ë”?9ù.üäì±û]¸AíÓ4ƒÏÙ±8ö+åT‚î'7šNž2§=ìRDZý1£@žÚØ¥0x‡÷³ÈM+ÑŸÃOJz¦;7ÔíxÍ &övœºq7ïÙNêë‡æ!Æ7Çhi†™x×cA|¹ !7:îü‘qÇÓÖÚýŒ•Õ¢ ß{\kµvìïý+ÈÊX: ˜ó¨eŸÖSÇMÅ*ÞêG‡*ˇVèÉ÷NzZŒ ÏØ.Ù^ÈPÒ ›Ë“Ûq†p“Y WÒg’ìÌ0Î`±óèª3öÒgo—äK)9<[ɺ5äáÞIW¢t¹.P%§'ëŽD¢˜Ðûjœç«J2Ûܹqj¸ÏpäÙÞÍ!/⮟áj½Ì-iº­£ÓFØÔAüpÀ‹£¹vø\iyCX«YÛÕàm¬ïéŸBêh¯»ÐÏÄËÿ¤¦¢ÐäÑãzÕ¯&ú¬íZcQQ¼US]P1 9`WÆ>Ì!G»QÊÉOâõ. þäx¸Ýó¯Rv3T›Ÿ&ðÕ%Ÿ?]±Pg²¥5­¶«ú·x‡ÚXLæÐ‘ð>eS-—©®¼plJ/ç~Œç;ù¸•R‰]U¦=¬gp@1÷lá¬PMk^ß¹!H·ÑªGcØv'&D ÿȘÚoiÊ‘`“@z:3=þ*ô†Ý}ÌSY4«ÉÙq½ò‘Í.З4 ï}ðƒòòU]!WÇ"ð ÒSSLøI0ÙNã±×ÊN÷i°Ö¢™žÎô’°Ü\RB…αõ7I3:)Jr¯»É‘dY·ÌW¥nE(ÞM%ˆ-3§HÐ_Ñ’+`1.ŽÜ~‰¢Þ»yfb½þ<³VŒS4NEÿVqOÅ&)…||žqu:`†|ÆÊ> `g ø©èÍÚ±§rT^Z,Ià GÈuŸ¸„ÀõlYû¦=½—"„7X¥›WŽhH¤4†.qjÞª/;p:%ÐPÝ””ëA§Ï'ã™èAH\„”ܬJ>Ué‰É†ºš]º–Utb˜×r†¿8`*¥Î#¥ê‘6){1§»(®tpYDÍ‘“EËeßc ÿ’9‰æ4Ïp ’àU«?Ò]ÇÅÏvˆ÷ï·>Žwk}"Mw{à Y~û\"U‰½Ý# ì(\‹Ì³w½‘î§ÛÙ,Ù*}úOÄTå´ã}.ú¤«Ç5¬¹¯ÒãžÄõ/è$žJ)Ú$éU¤ ëæQ¼-Üî‰h¿ÉörYc€dçg­.ô‡Ý;»Yl¯–íÂù%¨ ûJ6t !ùró¯‡v¦êšn0“í9¯ß´Ž+µ?ê&¸Þnß©²Ý’ž¼ÍàÅ!ÛC3öTÖYçÕÇÐiþ¨Ð“/f×IÑ[Ú&F—6Í{á†GÍ< o´_›ì=Ö1U¥ŠüãuÕØ2“uÌ ¬x ›¸èk>FG7Wýl+t©Œ*í `-kü|z-:Ii¡öÉò&øÔUʩڲ”{‹ƒ8ˆ5b]%Mv4* Yʇ´ŸÈë;ûÅÝ÷µÆVûY\ëÃ0(v«Úcz—„å*Ü߸#´—ÏŒ¤§ƒA µ§¥¯-å cÖ´Êgö¼›¯õ¢¬çáµ>Ÿ8E½×DÍ_(‘u+k1ëáѽŽvÌår’÷¶v,5Íé±Ú Ú©~Ûf¶!Uöª&\ÕËõšªelºáEDèw¹aܲt9\øHiï,÷†U®Ÿïxê T½è:#òØÁPaÏŽ«?RÞ?Yæ–¯™l÷Q̺ìeJí¦¯üçþŸ­&+—ø¿ÿRèU—„øgIû*Ú‹UÀY8|’ÚŽnÉWD¿;ÌÀçw(ÄYìô\ßy€Õ3Šý™5+øœÓó`8€ŠEvQø;°1¦·ã™ÍÌótZî™l~ÿ—zI­uz³—Bd†å=¤>o¶vñïÆôãŸz+:M¤¬v\]·çç²ÎUÂsíÎü¤Zë³4|g¨ÌÝz=²ëîQîªCôRW¾Êñê©pF¼w8mtþ¹)tŸ{KÐÝpðbiáU%+­P“'a– 4Uä‚·o¡UfyÌIOYßZö5ëÓ¦}ŒÁ úÓ£:#ø—þ3@2½žÙRÈj»…–ß]ʸˆÍ«µsj•<¾5¨©Ð¹Œ ’‚›’?T2hú©XrŠÞdAh²Ev3÷&•´ÛÀ¬íi¸Þõ×É”ûï¶7XÉ‚ Kìå-¦U¤m°þã.®bÔ“4%ß×g¤B®{y‰![Íß< hcºB? GË3þâ4¥\ endstream endobj 157 0 obj << /Type /ObjStm /N 100 /First 901 /Length 4184 /Filter /FlateDecode >> stream xÚí\YoDz~篘ÇžÞ·‹à–d;v¢X–o‚(r$1¦H…‹ãœ_¿ª’CRTÈH÷'bö4»«{º¶¯ª†£(› Y(k ë åM¡µÆ5ÆádaCÂUÎÓw]¸ˆ«‹E”t²H–è±VßRÞJi…Ì*oÑÁ´–؃֙„ábg@ññ¸¥Š[[Át =hR¢;†B+q ­q?m¡ O¡c æqh‹Ý5ºQ¸â¤*¢ãöN  ÄWIÔ8Ρ“S-ܤ0Ò`^J,– b*ùÂâ#aÊ*œ3tÀ»–òqØ^âvÞᆠQΉ6¶´ÂªDgQØ01±/,1¤’*¬æ{t<±ùp§q\k!3­"4Á´J…õ4š”aeKC¾6016L$EˆÃI¾—A'xÛ‘ú ³BÁ†Œ3ØÞ9\q°ç,Î9ëtK9€„ ¬8Gº+ÎÚÛ{R‰ÁHÀN7õÆãΘFç5¦ðçЪO>´´qEp–Vy2¢ Eô‡60$¯©“Љ' “"[ÑVIóÎêÈ2­A'RÇ)ÙÔÒd³Ò€ MÖ#“¢Øî-Ù`T8%Œº…EkèAá+õ P@£Á‡2f ³Aõà°Â€¡,½ˆói2U+aÔšìÚÁZ(OÙ¤°xWN¼Nw Xá¾k2hI>dž£&ëYŸ,gF޳ê¼wÞœÿZuò_]ìžÏ†“jRœâèàE!N«o“b~àÓ?n*L´/«–ØÇá«Ád\>`yKWãátÔ©Æ fŽ‹—£öÍU¯3.ÞL'7ÓÉæ»;íT#š5¥/Õ“ú^Cz²,Ò[KœLÏ'ü•UKüܾ¦9Ý{íqź?Tý¯Õ¤×i·ÄóAgØí .~‰[ly&^íïu£¬ ÏÛÌVä çëÅûÞàÙ`Ü[Lô..*H’D¼×½Át h¿M‡“ª_]L(6ˆîüÇ=q9j­D»3T¢Óu¦×ýꛘôúÝJ\·;£á@œ*Ð`I»Ó’D·‡[Œ{cQBGÝêBŒpoÑ‚ûýö|ðj:¸l¦×ýöt"†—ÃAõEtÚ´ßø¦Ý©V,‰õÝäO½Áp;u«[Œú,~¯Äþ™²üL6Ã=3I•ŽÐé2Fà]‰à­tIƒî[÷I!^O‡Üò»‹ÞåtT•î ÙùƒœBG_J@ŸÒ©dœH¦”À K Ë»ë¾y  o˜¡mÚ‚3e‚£/aÆ*±–º¤(ãm,Õݤ$µ)Œ¥R"ömÀ¢æ4¡h#Ð P„[p©WaW\"¤M÷Â¥hÖp)ª{âRÈØNñ?_Óߟb¼ Ÿbzħ­‘ÁIà’A%Ùƒ‘‘”9¨Õ žmîB†ð€ø¤u‰T9®*U@–œ²ða¤À©p×)âžÂù’“cÔ¡MY!ò¦2Ýu†´Š‘Q71Ò”QêÍÙ$žŸw±4HZï$&ô¥mÄù¶&FšGÓVÄÆÄ’2ØÝðw‘"®c%U`ê^X™ÜV&óW°òTB68ÚbKsë–O÷†ýîíû6ðqüŸ„¯äÉîó,UzK\UØÉ‘Ö‡žZøû8–kÎCþÎΣš–þ©}sqc8oOÆê€êVGQÿ4G¡çT[¾eâD• ?X+-²éû9 =Õ‹÷rµî(êÑQîï(úVGÑÿ8GQ;D”%bíI|ä+=ËÞà(‚µá5ëžBÏÉå½É³R®}¢Œ˜±Ö––¿ÔŒÐ¯¢ïIjô”T0“zX7ZÞ‚o´è5©LŒeª[Þ¾9×8àmGÛôÑk§È»Ú@GÏm ˆ š[ž¸ðæƒZ  ÁtÆ£H†,øôðÁ¤|=_ñÚzwãi‹¼ÞŸèðA˜™ß¯h룫½†x³Z~C<º¤U½Ø£V…[¦ÎT#eN¼_VHc63û¹ÛÄgXè5®Jò’„)WKºilë­²LªjyµR,».é–URÊÙvë#MÝ7™h¶Þ“.s›×®6õ¶ ñ3xüü¿} aoÈs›èMÝ•ÊüôÊÒÏ”kŸèiޱŸØÉ=#Û8Q(iÉè•bÊ<>Û~ÉØ–ïEԹפ²–vÍ-ß±¹C4ó3ÞvºMŸzõüõi­!.¸Í#™»¨è±~ÍWTN5àrDiq²ãµŒxe’$1ÏþN¿¦âB$‘Ñ/D"ÌýèukÁG¢`4Ì7VŠ1¢>·å•OîýU®œ›i3ë¦îà UC›y´©c/ý|$ë¢^9§ÎˆÐ„ŒŽ[Ò y©Q¤ÅÇö±]mçÖ£#Ûkô£öh5RhUX„~]·5McÖ8ò´Ü¿ÿ¹ò>Í6ßËIrÖÜæ‘€YÊ$4h‚ŒäS¯‘Ä$Æ „ZWSšHãÎ:ŒDM<¢œ!¤@hL5M½3ïP÷,X•$¥J|—ú¾Ü†, x›æL4Zg®fš7®£s£o¨D[¢©…ǽ,ãL Œ”$8Ls)¬@t –Ë déܧô5p*`(`ú@D4”Vb–Úeؼ‰J Dš,IÜ$öIKuງŽâq,ø§!äÓðIãTM™\6ÂmÀžæw3Hk9 LrÈj-6„@>°›5ÇføN«åÆÖKÀE‰ÉÕöOuIŠÓÈ£†Ùä|‹R"*XèȧñüÐøÉ0—K¾D?ð!Ib¤õšA X.ÕÎNïU’Ø}$ Çbâ:*j:Ë¥'ct³4Îí¢±\v:͵¹Îòá5©Âz® #ÔÒ{äkr!ÇÏt¨SdsÃYéY V° %ñl-I×"ï¢'¤Äº”fJ‚¸ÄÚpuШÊ6óWŸæ­?>Ëy|–s׳ky8·Ž ÀÈõõ̼—gîÓÖ;æÖr®I°Qu~­ù;a,安8fÚÈA‰ð™*NNÊ Xx@`¼SC(!¤ÔŽâᦧˆ” ¢µ³#8‡NCy!Y@ˆš‹YËttcEoªænn} óSÓ}šÒhrä,Vò[膯9Ù³†SW.¬õ‚MÔæ‘ÄòçusÚSñUü.¾‰?–ªwèË/^~:†@Ÿo4(m!ÐD@f%³^HÔʦ@eCœ(6ˆÓ††4jJsÔîö:íþ^ïrÖ=ŸÌz“›Yïk5šTßÄxz}Ío¢t{ã›~ûÅfWlÌî"’×GÇÏN!’ÃW›DbtícÒÆ?±1s—=%×Êb z£XžÃÀ—ÌëæõÌk–aR—ô¢L5‚Eý ›Zñ·q¿=¾‚á|[–‰ßI&ŸÞ½zÁ2q·ŠÄ„Z"Ʀ’H’%òë2+qVNÞ~8zý’Y ·kWÍ4¨b%ª¬¼„ZO¡Ê«¹òÆkºò;aäÇÇÇoÞƒÁãMækåÌ|å½”5÷iŸ6ò—òFü¼R¼x-~Z®4!æ{õGÆÍóQ»ó¥š0Öý †æ«N9âd' œåˆjÐ%ƒ¯~›¶û½ô¯'øpÿþµFâú­)ü×ÿ*óUµ›Ýô§c@4¿ªÕ=ï3õìK^@è=®®{ùÐãê+î<î}›»ç-¸.þ-þ]†Ë¶°¼Ÿ>ûåíá/l Â¥±9\¦ÆÜŸ˜B6‚CV>©þ}(j¼¸=Hn¥ÒE'Y’,Å5î~~óæõk–áÀ0&ûS²(xmÚZ†nƒ iÊP6ex©rºE’"9}a´_M«üNðþñèäíé ±è7ä3Ä ?Qx3‰f£™‚çΰÌÔN@ÿüäÓûg¬· jÓ³˜E½þL…Í0˜Ñh3ka'ˆ?>ùeïÝsb-n°ÈY ƒ|Öün½_åyÃ,{ÍÔÿ®´önÄœ.Kj'üñÇ£Ã÷T0ü´)z=7ƒ{ÙöÓ´…,Â!À£ºPhOuøº˜UQ· 7«5EŒÕ: ìrÏ?¼}õîËéôtc%@(@ÿ—èPДRK‚¢¯sAi­7 j˼áU/,g ³°±’ÌÂ>ûEf¼"Gzƒz‹Êk9¨ßR€MÝj4î GUÖÀz¬ ;ñÉ›“ý=VÃÇJH Uc(ðQ¹…`»-=µ›3m~ ]\Oû“ÞMÅÀæhüüž· endstream endobj 280 0 obj << /Length1 1409 /Length2 6124 /Length3 0 /Length 7090 /Filter /FlateDecode >> stream xÚtT”ïö.Ý(Ýà €„0C‡„tHƒ´à0 0Ä Cw#% )¥R"‚HJwKw#]"(ðãœóÿ{׺wÍZß|ﳟ½ß½ß÷y>vVm=^Y+„%T Gñòó$òzÆ¢H bgׇ¡ a"v(Ò†€Kü/‚< F¡10 ÍÓ@Àj®~A¿ˆ¿¨ÄÿED % `7˜@ƒ †€C]ˆØåNžH˜- ½Í¿^œ.¿¸¸èƒßéYG(Ã`”-Ô½#ìÐC@`P”ç?JpJÚ¢PN@ »»;ØÑ…´‘æzp‡¡lºP(Ò jø50@ìý3;@ßæò×CX£ÜÁH( 8À P¸ :ÃnEЛôTÕZNPø²úÂÀß³ðóñÿ»Üßì_…`ðßÉ`áè†{Âà6k˜ ¥¤Î‡ò@=€áV¿ˆ`:ì†9€-ф߃J²:0zÀ¿ã¹@0'” Ÿ Ìá׈À_eЧ¬·’G8:Bá(¢_ý)ÀPúØ=nÖŽp‡{ÿ]XÃàVÖ¿†°ru>Ü]¡ª )hˆè?˜ ‰‹ˆ‰ Î¨Äø«¼¾§ôwÿŒžÀ×Û á°Fõ…YCÑDÞ.`7(…t…úzÿïÀ?WDüü+°„ÚÀàDÿ©Ž†¡ÖÖèËGÂ<¦ ´öø _¿¿=EËË wðüý÷ýMLdÕôžðü™øß199„À›—_À+.Ìà@?Džÿ,£ †ýmôŸ\U¸5 þ§[ô1ý«c·¿àük.À?ki"Ъ…8ÿ#r30‚~ðÿKýwÊÿMῪü¿Dþß )¹:8üsþŽÿa°#ÌÁó/-ZWÚ´ àÿM5„þ1­Ô æêøßQUmY¸ Z̼üB| ¡?8ÌE æµÒ†¡ ¶$óòËj08TáûõmAg@ÿCû bþ~¸ uù;EÛçŸû*Â!«_>€‘H°'-'aa€7?ÚVPßJùà:€žÑ`@ýºVt] ÔúËK¿¿14䃻ºü@ûè„„¡Ïèòý!®H$:÷·LÐÍýkýÛôP¨B4=€< ±« iø^.Ëèλ>€·¸Üô<θ#BÅ1ùÚÛVÿ•ò¨³Ü3«º¾í©Â°‘ FîýïƒfŸ3’U²QŠ«r‡¼oz.N>Û@Y¬ 6C¢)ÏWÉâu˜o?3Å1½ÿÒpâg¨h«àféFîU“˦‰øÀÿŒ5™#›Õ´ƒeÌʶŒê56#‹¥@š¦ºê«>ºÂŠ‚ÅžË5DÓ_x®šÆ¦õÎÏwgtv¤¿}ÜV&RFIO奯šÕ—ÐEOŸ¡Vë‚Á!'üM¾0)î.}žcǶcå8?[l;=ÈPÀd×' ‹ÀL@jÝVª®oéàè–dýéå‰÷š]jàVR?I Œî…•pB¶= zã›6šŽž+W¯/N&¹ ì€Ù2#õºéúöZ9峌¢œUnij]F.m~N‹sª—‚ >½£S*¬0b2£7«^Iâ£#2Îýb{²a&Oÿ‚‚‚Nþ®Ã‘ÑÆä+å£MÌÀ 9y">ÓŽa:nÙ} ÅMÓe ‹K㈾6‘Bå²åŒSÖ¢‰eâ0y³&îAŸ9bNh­z{°÷5I´r¹ÞhÎìÚ8w2|¬ÄÈðV=ÃÄ{(Àù$!Ê·±'gÙ¤Uá½°þ>@³’Xª¯ZœÇŸ¢wƒïþñ˜1ByD’á^ÍvÐt,Mhª¾ÀΫÑËyi„å*^R^M_ÐL××2åkBLËTÞš*/(h_ÒôP€ dÈå¹ó¢Èæý1~N¼ƒaUO$=¸¾ÿµÜ5õ8ýF¦€s'Ç“‚2þcÕâØÍi74Ó¥c9qGOþÁñ‘z¬s \Τ¥3/rI­a•Ñ ïì6—JOÕ÷3gpcªL vëÃIÂ;±qùv±*ß–°Š=O#o­½Ïé…”¡üÂaɉŽ^eU—4H—÷¢p‹KûŒÜ¤&ê3ûQÜ©‚Úvò£®•e¼ê¦ë·¶KBÓ†<ã;Ùo[¾½¨s_þ>l ð'’E 笯9;ÞÓ’P³Ã›OÈ o‡œ™ ú×Ìœ² ©4h»åð˜rÒoE»„­¡Éêã1³õ6›0ówê_|,Eå »üoÛï;^ÄÒu|!,e6$zS…X>Ön.œµîe‘Õš‹&ëÁˆ#à5Öç@òyNäõ\ˆ‡¯æ\EWÖÎL2ÄÍtÓ<ì(”ïø.ÖÅ5Ç ò‘má^ÚMúÄÁÃW`r6ÁËÆ³Ÿ ¡¥³ÅF’›>Œ÷D!ä…œù1©‰àG=ÝþY¸ˆ“íÔ+zk/¯c™ÈǽŒÄTÓô—lÆ5j`¤óqo?YsâANôÆÁñ>ß ¿õQóù§ÒÌ5*9ò1÷ûB?Ê\‘ù*Ó8üIBºM¤ R’“òWÞ¯ŠÖ*úKB½}L?ñCÛ.Ý+Z#:+ ùÌÞG„á½ÌI­RjŸÚâ¡Äx—OL§8ÁÔy~|Ìi§o¾;¿Ò%~ É/?LÓâÌ&íÚ~òÉ7®î ÷-QçÇÀö÷1‰q²È9ãy©ÂÄ%0E$%!íi­Æw².®%*iqªvàgJ(‰{¸½aâu™-Ù r#=î$Dú9u‘†±{ ¸…øŠÂ¢êKv+¯,G_›û¬Âd¬G¶ÙYKøÇÖ¾/ß?TÈ¥¹O/3¾r«FøƒB/¢s8 ÿÇ §ƒ N¤ÿ¾G­‘µ(|fI9hè΀Õ2„ŸÖ€¬Rl:Ò1s±…oÁÔWò|UôŠt¦>F[}AÔÏ ×%;»^Pñº¼'&ºÌ7`¡7ƒöÛ”Âíç={Ùß6qQçŒ7k]ïwf B·Û›ûwY'ŒÄEïKn³$}mu$ààÚUk"Ðs§b¬ÞÝ+^øÄT´”U5ˆÓ?T–bP÷U-yÝùApì““¶yÓtµ/:é £³¸ÕSÄÊVB«ÆrèbÊ“»©\ÏÌÚàût—+U<Ó_^·?H¢|‘,š9š^?Ó~Xr/òN?Kf϶‡edz`ãÀG„E=Yp„wÖøMçøÜí‡,®"Ôž-³›¢†+…_>qù޶âžËíLé{¯4úDJûYÉãUN… ¶9ä <ßù˜\ø8?%0TD芓h›¼Ó[T‘ Æ©W¡>à ÷ö|0²VùŠnîÖvxji}SúØ»ÂÑ­×A©[‡c±úó4”%.Ù8â¶g¦Ì—à{+?ÀT0²É.¶¸ë$Zlv"¨sË=næ Vëy¡" ½´õ»Å”zÇùÊ}VŒ9§AŸy"&n%~´bà‹!Ó5 èäflu¼†›°Šžmìa Òïb°T~ÞB\zœº²=]À‰¾_x"LâV” #yxaÖF]Þ,%rhÜO»§´Çî?cug+½ªX‰uèÑ⾯ï»HgEšÛŒIÅž–¹a`Ã%VØ¥—KÓãorÀ2.mÚëz¦îš Nh×ã·CeýW©ÌÉF(^_¶-ÁZuÏÜ›ü/cF9ØcÔBóCHå Ó¨™ƒÕ?ãÞqLA+É6²Òp…-ÊÌ"¬<ïÆ/¸X˜ߟ»+;ÿ•¿˜3ÿÀ@`¦A“@XZáTý»mŸë+MÑP ?ö®>B×RºÀ/‹F \l˜[ª_RVx‰´& ö~™ìФñ?dÉ(aå¯ ±³rìf”%‡& Z‹eO8\n[ÿÔÐMçÝ~*Tº€c1Sg`Ü7þ¦ì>²ãºß†gÁsc¹ÆRe„[ê^"%kì,}eöP.gº£Ÿ¬:Ž”{HÛeSâü+áçäÚ¦"Oû¸Í×FÏ p:S"˜*s¹¼Sœ-]­.é7ø6"Ô£4Ë¡\ƃÒ"ì;F¤kO} z¿áãZ{ò› HÝŽÛ*.ju Žf(zt:A·!‡“Rl¸ãdpGxŽ]¾ÂðUsÛð­#2ýBxDõ ·3‰€P½:G‚onš°A?àŠ ‹¿”¾®Œ‹ÉB}j¾c2žü`­ºÀ3Îõ¼!áë“oyãòJ yëá·ê!˜>š^„Yê»,Qv,¼UmîÍÒ›ZÏîÞët!•MC¹ºVçéÅúzì &‰tÓÂþ›½>TvÀŒwÿà E,-ž5+¹¼HªeReÝò‡þ:ð9&n¸Ó'ú2Òæ¥bfÖʧ‘g©ÀŸ°:û¯Î­ORnÖò†Þì}lnwþØMQ¥ü.(M¹I˜ WR¶îlgRµHÓqÁϰr5è­Y0R÷.³ì›Ò매4Åf?áóõPrVŽãî-ŠˆšçìÃf0…ú«€“¨£¾oéÞcH¶˜ãÛ%³QÂ2}[ jÆ)pÙ\l‚<žøÕÑË8*"üŸØð˜¼ÞF]½$ìëx¿ÑUÒˆH[T×ô„³BE™n²Híâ×î‹L½}%EƒTä—Aõ´~Y¨ã=¿-”‰ÉMÆsͬ.«³ÿÄ(€í3›[Ïø¾æ©*ÑrÐlÑæ—(’… nÂÄ ÝÅç9ÆìXÕšÂ=+á"»c»MùÖIØÅáHSz'=þ¹žÁþB—²¢€öÜ–ó’-`‡'Ô)x<‚g `2'2©] :]r±Ç˧Æô毞ˆ˜I ±ˆwÂè¬Îö:&üHkûõˆé]»&IIÖKÊåÒ6ûqùÔy–bµ¦Šmb’-u¦%Í­+¯fU ™ô„+6,}<’ ¾í2«v>ùW2´¼î0µT›'Úvt®´D¥åÅNÕ³‰¤ŠE*ùÙÈXP‹}»ô>=sµò`Š_ñ hÆg‘›ß?Å[–…}¶û8Æ®#fÚqyS3 JRi³£ï=*íä÷Â¥L5ë>Ay, U$HxLŸ·eM—`DÿáƒÆÙââ¼xE­\ÌâòØwUÄá ÑÚ³¸/‹‚¢T0-ÞfþrxVöiŒroœD¸UVãú^Ûâ0VÙ°¦žmërjNeB1À-ªÞ™W>Äd’óp½j¦&`õûÐ{/8, ¯,®È*UG*ó´œ^å ·¼¿# “Ûaïœ÷#çéWÖê©™‡·+ÿ×Yä8壄èµñ-G±†iØgó;¼ñ¸˜ ]ªSVœëTKΰ£Þ< ¶Kt¡ëñÞ?ß]ôžÝâ ·çQÑ";À-_¼2È¿‰<Öê—P 9ðŒ–.u.æÄ_8gÂÃeæ´o•;¹Œ4¯EèŽ~Û[è7úU0Éû Ì“Toaï±ÛjÔ/yÇÖ8OaÐ1…ñß~u¯f’¹`n8OŸÖòùXùp-°)¹¥+lÉSiv÷+ÿá{ÒCMà$ã±m¾‘]ÐŽƒ­ûK}š•ÚVsjæÌ´^ó®“cG(–Êë–T¬-LpèÔ#‡'w‡XM¦ºíÛB¤yi¤ñ_«öPªmðêÊrk?ªeum~~X³ÇÕ;Û6á†óú2çÃ՛׋ÐÔæ¸Ò‡†?óÅ×áU’ß©ûI;$Ô¥\iö7TÏ@MI?·åïtf¬yÍt5oMˆûGÏ´ÜVuLŸÕU³Dÿ<ŸéàŠá©—ð¹|Qg LFT2ºÄzaóø¹ùèe½ynÜM{¶=í…Tjó¶SV47T‰z/È™kã§1:=NcÀ[YÆýýÂÊ™j°Âôkd›ùiÄdÙZpÚ“–¶=žtö¾~B¬\Òøh>Ãøl †I!¤p‡§»õ@Õ#ürjõeÂó¦hN­ƒE6ìXÙ0#5!ýMãô÷w9ø˜‚™½¿9#ÃìËFägî!9šVn„~RGì¨36(hdœé5ÜÍx?a.œÁ”¼â^2ƒ!§ßÁ¦OQêZÁÑ1©ø™¾¦•JIð‘1ŸshÀ±PãÍ!Ï'áÁïD#cš¢~òßÎgá ñ²H¥ûñYú]c•ÄOÏurØW¨ê†y¥'³¤50 ÍwG›×ׯ¨Žûm «V)'H²\“5¾½ùú¤½D ɼµÎÉûýà-Š( ÒÉþ"Ð[WFT[‡"¸a]è*íƒ.n¡Æ…HPnã4˸F(]pÕ¸9§Ì畚ôj0J÷°ƒŸe±£Àzg‹kÖ¸™Ç“dïJ‰vÉý6âÏîgSù,Õ&i ߥ<ð#éÐS'|8V £æž¾x³»N$ñÞXvJµóìø¥ý¨ðlF8ûžÓõõ›ç=ƒº¹°Ð°¡Ž¤G3ߞ͟?„†ÌŽóušýäíŨP8"±oaåÆv#Û:É" ú4ÝL©™›«ÝaRùØþ§¹;ö7.p~Yn¿–õòbšz’È š|e//.¢ÑZ2\úù? Ô‹ž¯µõ2êÙ™üeöu«†½µŽ8=Y¥¯:‡tn®oôA G°9fCU e98¶g {$ ´k ¥"Ukm§öæs$I³¨ýˆ{|r˜AЕ‚ÔÄEÁRÌóVÕ¾÷>ƒ!ídµ=“!rí>GáŸ~t`æ%’½$m;Í/K~U›Û¬ã¡ÍÜ¡ rˆ3y£1z’7S%*+ôYÿ¥v†SyF–Æ`áÝ–òävüëÝüYžú¤ø¦hþõ¤æ.îz÷Á¢ >Ï[î“Ö÷p *ø–I|õá©¥YlêF«]Úþ˜MÆ/±³ÄóF°¹h¡ázш‚ô:c‡·é®¦>·@8vfMs’% ?2#$M+q1Òv‘ºsµ‹µ"2šçÈgš9¹öT¸l¤¶>{/~ù]÷TFÑOß¿¦˜óìúhÚ(‚â~uä3ýËæ•åÞ”w9wÛ»eуÅôØô6—yWŸä6´±í<‡±®µ\æhx¦˜PQ‘oãºLP&¸’ë.·Dú^\LsR2£â>ÑÍ—IŽ€YlÍé«Ê|"CJˆ¿ØÖІÎzÏÔ•ž©Ý:©äªÚ}öÒ£Â@ág1VÚ3§b&žÀÉ­öŠ;-^Máâ \Lƒ %å¦ÕÌÆ¤æÝ낤Hnÿdz?ÚsÂͺì×øÊ÷ÎH—™rªô «>Èë¦OÉR÷èp„G±Î€ßv•¢J2ÎL‘¯3éÖYÕLwèŽ{ç¼d½3Ì.H1OÃGì»H4É+×ËÒß¶v³šø£6sÈâßæ÷%¸NRCôCÀtX·CbëÞø¼ Ë†‹D¿o–oc»`?—'ï…þ{1¡zJìâ5?Á}²w·¾eŒÝ@ÔÔ±fó”uZ8Gs·Õj‘Ù¤t|ÙÖ¯&¼fb$ Éf‘3(Îø,¾d—@sëòeL”¡pO˜E,€*)·Öžd2+éY]]¿žD!tm;ðß|¼ÉýàжªJÛn£É{íC0q³‚ü©»¥-íéG`¹“ T¨š!û¾‡ vØjïÇÇ'ÄO±Ÿ•p è9ûí¥ñÀã Ѹ©,»ßG¤ž=Æi$Zî >ª‘ö©ÂRæYЗ‹ÚÎÔÝ8œ &Ôš8’]•ûèÐ:â½áâp—rðˆÿ\°:hß;Ê`]᥸‰Ó Ó,VÒIøB™N”¸Ÿ|à kj—ڮŠžSÛå®|CµÄ.¥QÂèˆÖbZ0ðæ)–h­JP0p{¬¯ˆ šXLª…Ä¿T\Eb™ ¡¥'©UêÅæ«ôI­§&NÂÕ‰Ïå_Þ±}Eä.wBp»£È* 94åó½Ú;£[]øb^ i¯i'SF»`Ó1’5ËÇÛ7T¼þÔ¡²³¹ç¸DìÍÐ+ jªkÉA5¬^å‚‘öG^ÍG ƒ}ŇgæHÖ£• D¹Équuõü®'rûó–Í tz9§µØrŒ­‡«€B «8O§Ql6“E—!jýà¤[ÁJÞh4¨Z©ã«Š×@!UîþÁ@Î*~¸ÿ™uD»¶šyñl0Óˬ]&³˜Ï/H@ ^s² Ç€dB÷ŒÛ= D†ä²÷ÞXoV˜­™KçB$ÚâõNÇ Šu±8·¯pö¶£÷)ØbQ¾G’,D*¸—€ˆüA<ÏZcik¾äÃ.²„£BÝ®•ÐbªOwì }ßSóÅi—Ù)òåäWŒ7úHÿº¸ÏFì±2•Å>¨Š5pqê•oü“[è©G+ónûûMÌÏU„;Ó‚ŽG"oÏÈ-»7u©›®nï¥Ô Ýó˜´53וb\N ¯÷oœÎÇyµ“¨VÙæåJíª¶¨U3ÍŸèþÌhÉØ0Â¥å©a¤ÁÈ,e!WjŸÐå?õ*&Â#Uè–À¢]±©Œ:ö.7>ö º¡¾Ü9˜ß$!î|œr¨c{æ@R0Õqó2Ÿ%Ó´ ¾=ŒþùV6Ÿ¹9Ä—š‰Ï¥ (yÉ|"CYúFæ¼âz¿ÏP­õÌzJ<’<ŠíÍ}¼ïô õÚçÖ’L[1¸ñV¨Ûs¶¥\:¾ñx¶ßl/ôbk:«ÇÍù±ÅJfË…$ïRɬýGï €œ‡¼;íØù…Ö3Ù‡±†Ù´ŽígÝïú««é…†áZÜzä,äññÞ\Rs¥k¦)‰l„­ò:ý’uÆ>¾‹Í°Žú€ôПâtØò4z\Eyç¯_ð¢ˆ"0dpõwsNßòg~ºÁÅ ±‹ÆÞjžïî­¹M/% sq¥Ÿä Øe cž%H—X·ïä”&$¼d/ÜÔ–¯'Œ%¾JO…{å›uO‚¯~~ Ứj^M®ûr±.þ)Ó Æ(¶®©ø"Ïν”Þ:ƒº£¨YáVÞ6Npe1çÕ ”;j¼ýA¡,ŠŽÇ·]ôh¼$âìÖ (µ[|4×]*Fb%ñÖÎ$Iø““¢å/‡­Æ ÚKæÛ}ûwÈœr²âGl S› Û°þÄi ù»v1k?=ž5Èh†cì>›.øÔ§Ô"Y+#Ëš ?˜,M&J¬Jð6J²tè3SŒ&+ù¥\ù °·Hvvëͳ¬Òšr:µ“ì$Å0È_‰?™IbÕºþ¸ìÍ q‡É‡ÕFŒ­)”ruZzЯ?šá<šP˜ç#ÙÞÿ>Å" endstream endobj 283 0 obj << /Length1 1478 /Length2 7916 /Length3 0 /Length 8908 /Filter /FlateDecode >> stream xÚ´TÔÛ6Lw‡„Ä€4ÈÀÀÐÝÝ)! 03ÄÐ H#‚Ò’’RÒHHJwHHI#¥€tó¢÷Þ}ßZï»f­ßœgïgï³÷9Ï>¬L:úemÖP%ù˜Ÿ—O ¯i ÊÏàãàåãá±²ÀNÐìx¬FP7w.ö y7(yoS€ 8@Íà À/àããã€øøDÿ!"ÜÄ O˜-@“ †€CÝñXå.>n0{äý>ÿ,6œ~QQaž?áYg¨ÌhBPçûm N}„ Šôù¯H¤‹èååÅ qvçE¸ÙKqò¼`H€Ôêæ µün q†þÝ/+ÀÀæþ—Ca‡ô‚¸A÷'˜ î~â·…ºîwè«j´] ð¿Èx€Ÿ—ÿ_éþŽþÿ ±±A8»@à>0¸=Àæh+ið"½‘<Üö7â䎸‡xB`Në{ŸÒ!%Y]ä¾Ã¿ûs·qƒ¹ ÝyÝaN¿{þNsÌŠp[y„³3ŽtÇû]ŸÌ jsî>À¿/׎ð‚ûýƒì`p[»ßmØz¸ á0W¨ªÂßœ{Þ¿möP$ÌÇÇ'@]Poàï |\ œü¿Í÷=ø¹ \v÷m@`vÐû???H |¿øï<:ØßuüG¬*Üý«Üûsú§dÏ¿5Àñ÷€pþ;—â^¹PÇ¿…nÎæ³¹ÿðÿ?ËýOÈÿŸÊgù¿ ý+Ròprúãçø‹ðÿñCœaN>3î•뼟MÄý,Àÿ—j ýkt5¡¶0çÿõª"!÷Ó ·¿Wôc~A^>Á¿ì0w%˜7ÔV†´qøK5Ù Ï› ÕA¸Ã~¿0÷Q||ÿã»2ÇûWÄý^š\Ðûúï}á6Ûßà  nn¼û»¾G`€ÿýTÚB½ÿˆä…#÷!€ûv7¼ß+,ªý6ýA"B Ñ¿¨0ù7­ÿ…øïE „þ¼ç:þÞKˆøx¿‹Ûø_ÅÛx¸¹ÝOò•Ýwöþól@¡ÞP¼Ùi„xè³êÐæóJÙ‡^7F@‚„=‡/qŒ×i<7uƒå%.Žò¢gdÛ'¿G_,“킊îêMð9Z.ù§F‹=˜­mÆ´³^ãñ#“¼Üu×ð}Wok{b%·µ°ß ×çu,3ÁIÅãúkc*c[Y‚­¡ü¬Ìa›½î/ÚHL46µˆˆZh‹lFK¹\•ìÄd«*‹û¡ÌÈÏz¥í‹ÔîAxµg®„¶’úpg‡ÐEý8cGêíV‡­ºÕ`‚u½C‘úê Gðqsè k÷!}éG♪ëc‡*‡ë1ªzö *½ËèÝ)èòËo˜–qeßy[³¯ZŠsµ)}z9›dPà+f/é(-U&—TÂÅïšÁUÿZÍûŽTÝ_£ª<ÞÒÀ=)s"µ€g0¼NGÿÈï+ÿ¸ î`tˆ[,ŒG®…~-P/%+ûÙW§§ß‚G‘U†ùÍÄ*ø¬LÄ/]1ì> 6ʧóÝR…åP¦ ]›Ù}dñÀ]”ª=šÞNÛËlÄ4ÿQO—ñ`ôSåÈ®i¼Ž¨¬*§³ ]ÿ2ùqQ!u®ýf]Dò°iEƆ-©xÌtâ$Ó*J/Ä>×tf7qÓ¸F7aqZ#Ií{]õv—žì”y`œrFoX5d÷¿4`4Ç5ÆÒ†'LæC’_dÍ PÇóý%é—júmÕ%]5ž',‰5ǰf–_ÕÚ¯ðrÓßÍêÆIï¢|Û¡–~*ñÈG-± «öxžWÝŒ²ÊP¶–—,zÒ¶ÑYõhXS²ێךìp¢¯t6°‰P™a‚ ~·øÇú‹œIfuƼ‡ds—ïJRÕTs­®Ö­ÝìùŸ\èâ]Ê‚´X$d4|O8XF¶§L„W¿ZÚéú|k$iA`HD%ò|hd}%†0x.ÁŒÞ§?zÐÖ(þlfM-ä–û|¦‘ɽYRÈéÏk)AŽ3¾óînÒÒ^ËÉHì×N¨ þÂ%G>Å%ãø3Þ7qäòóTýR_ÎmÝÖIÛMs|;‡h-™9®¾•øÄK—TÌ_º—¢„»+„R#ËÀtØêtçç“×®6¹…Foëw{»tæUÙoõ>õ¢_1Ð5¡Â™\öò`Mm¯œ\UØóã¥^¢.Oˆ’ÛpªçdµGBA¥€ŠüG —XfMÓÜÛ¹¸(Ÿ®‡WÛuuú ‘À±€Y’Q æ¬…ÝÏ&,“Y'»Ã¼Ç—J#þ3€NS¾,ÿ!u¸nOq%t+ЧÜNÛh*ÞZ¦LØÉà˜AúIóÕ|kTù^{d®SUé×Ï#^M­°Ùí9Á%Ûbæ•ÁÄïLÝ”[˜’}¢USn¨6™[µö¾%åKšÑ·,2‘"ø¹ÞémÖ)ë×þ2v¨íB™nñµP"¡TÉió¯Ùú®1âÏ,¯Xé ÷=£»¥`»œ¯\€ÍåÈ3LPWÌ£Wý­WË17\~ò¹gÛé4’U,–©á!ô f4Ǧ1¥ÙkѸoT>­xðœÊv°<¤9®Ãå½$|@KãAܺ¤T{àö¸sTê¬oŽø:—Œ6Ÿ€èËÏ Èü…˜âÈ.°ðP#+§ë€õ7‘ÙgÅÙ!BØñŒÎû”é]quË3»'«×OÞ)¼(;ühþ!XRO€ªŸòÒVŠÉt~$vW´I3R·k$ h4e[³RŽŠó­ƒûVns²â˜†-Dõ×~ðê‰K¦ìI!¹Wc1 “dz²s‹ôááéSâüJRCÇ"cд2ÛF Òñ=·ƒ‘KñSûBÝ™:Œc_Qé}ËÚ Ÿë*ñ9 ª…Ÿ;ÇCgΟ —Qu[œ¼Þq++dÓMúÊgÖš>.«i‘e7eò]ºÎÛ‚!Öxžúýn(u¹ÿ"wu á†c¼Ý@:Ž&$u˜—‚¯Á[ëU7ƒ„‡-ª”•ÞnŸ=’yle/IK…]n÷Ï ¡ZQ9<ð*‚ š®ê¡H‚Vzh?Nýì.Z÷º–ˆ)èxLÙÔ­3˜e:ç;`ÿÝ‘Æz{÷9E‹H‚ƒØî3zn†Â† ]„ˆ#ïùžÓÏŠb'RÒ8‚â`éùBš8{*HÇBw~6Q5yŽçdÞ%=´,–—´ò*ƒÝÿñ„{°Oå4º6¨P˜È%½RŒ®˜K!*xŸK—tÞÅ— N‚3~*ò,“zÍÅd ÄB½v©ã¡ =_ÞmsëçWoL\V}ÜùâcÁ˳†ÂÍ}Åη¹‰b^´C¦Ù‚£{ý¾}©§gK‰Û9xÇ-Q`úAu1i¢þÐC]ƒ1sšd7x°¨ [nRu#vHÇ™$³Ë! ê´~ÛfVf-:аx¡¥WìÚ„ÃʧN»¼ßË»ÖûÃZ¤Ë 7UÞï%¥ÇOQ| Vf¼Öžt~²á®™O¤Ç–Áꆤªó¾Eïf×í.VªÝ7 ´±§*Á¼”Ù@Ö4 &¯‰OÆTFÎÚÐ=''‚ñ¥LRévO²­0#-<[ û]s;Þ@—ŸÄ[3€.UžEñ¸Þ4•>Õ=ðÿùQÐÿ¿¾#¡¿ $¯¶.Õ"C§u]‡ÊΦ~Qv´9]ÙóËÏÄõ/Ôûtv£J‚¿EÕcAá)/ë°ÅÒCÌÎñ•Ù•AìóžçU–ôEEåèh÷ÈzJKG§$f/5M¡­5Æ9ºK:yZ/Žæ¡¢‰ôûW÷uš·o);yúSoŸ½ÏêmAñƒ‡JO®Ê¾FϨíæÀ %íD|Ú;t; üdŽx‚è ÆBQðeé‹‹®JS4¸\Ôò½£Nµî“Æ4´g‘ZpÃ& –Ý=— óÕ¯G…í,¤„ùœ„`jã>éÊV–ÎtFÿ Ñ»w·ùõ–“ŽsÅ Ò:¬ýù8O`àôƒa÷BÔ±no­ô™TÛ!ê&УK<œXt¤GÝõ U8¢›¼ºvm¨`ëë*‘TSÏ\8ù9Ò"üÈÇÂNå饸‘Ša<&ðÛ#fN†©;dÖ>oñ`Pq;Óh:Ì+Ê;+‘†Þ/p‡ÀGþø¼µôX+u>I}SªxÕZ’0¹ãç„ÄüõÌ;õ]}þÖ·ëjѾߌÛÝóGm*1qLT.É2Ìmç3㟵r x/°ÂoØ(­”êÚ…—ÎzæßTÁ.0Jíç`©1{Óyðz7íx‡èÍjÃ×ú¾þÚ G¢ë¨Š~‚°ÓJ"×dÙþŽ„4…ݨÆmA-¹œÃö5$ñ˜¯…ˇƾÂDæ‚Ød^鎡øñŠIà‚Ó‚™Ó7¶ŒM Z²(½ç¯½ä˜òŽãÏ®à="ܪ`/S Oò…êqCæÞ·ŒÇ!˯´h¨?/`’5¦‡~ãú†Õu"é÷$ó‡¡Ðk·ëð7É›Ês@[Sg©G@©«{ep{îj¥R…´^«²3$³í|¶L»‘%íCàs½kâ=fgH¥ñæµ1©îY¾ q¸áÂc¡rèüölã'nÕ5=IÈ&c;¯ž½zgY­¡æ •ãµFô›'S¦ãóÚ¿ÞçŸù¨}ÐMà¤k×AE¡°H˜ô6;°Ëãj3TbŒÑeì1ÞÕ<…›a‡‹µÇ½ìç;&ٓÿÃ]vS0#Kñä„ök,Ÿ5nÏÑy¿š²–®Vý¯&ˆ X<^ÕV+:8Ë?­¹ÝŒqK™œP@ ïapl¼IGP¤bìñ\ê¾7ÝÄUm¼Ú­‘SÜÆTkesk~bû` R[¿G|ò¹«’¨Mߣó÷¥66)4Ÿ¦S[m5zó­ùz’ÛH£ÓºŸ|=¨Ö ¡P+ìHqM*—à+½ž¥Yó—g‡:©Ú˜8Hy2f"„r§SÙVsSŽà41¼zÊkwòáÐê’P”ÞîümPª1½?žºxJvñ8 œP ÝûMíÓenl/!¾æ°ìIê¤xÅÙLïË"ë»5ÂÖ\ hþu¤;'§%yÑÆ)ì´8y1:ƒŸR¯{æÀ1JÝÜSÞK÷:®a‹Ñ^q¼ZÐ ø¹ÛD{Ì_ÕÎ:Ô/Ça„@1'2ØbçI÷±`¨Ïæ¡j–›B5êgæÒ .x2YßÛŸbŸX±ÈeU+}=ÿ;³@'ü¥<¾|hSÖá§Æ”¢ÂIÇ—}Z}ý-RÛ: ·Ä< Û@8{u|äÞ7Þ}ÒÝÎ}Z‡,•–_£c§DUxš6ˆ»À ù™Ì!Ï4³„ª`ýù‡n(K±Ø›vöëφqÓŸÝûêÆ7ñü‡é1‹É›9@SG¤Lk6¯h øðÂK¢S€‚ÿØF=Ðÿ658D€RáaøôõMQWgÍ#VNl•ÏGâÄvöåURiê> (4}È}ØáœpG4Êq…9©à×5WEó‚Úre=+TŠHebG©Õ¦kNKRþ`q Ƹíž0~°Z—D]‰6Zñ„LDV(1Qý‰¦ÃÃdgÅ3e“¬ÉìÓµì§âM´ëôT@äS¶6üæêróµà”ìL©Œ»Vêô„[»¯‡ÖRþVw×tiCýJî›Êgu s“™~ë¾cÙš%\^O=;S#l,i]^µ0.DŸ®~¤|¦Epbe½êB‡ÑóìeÈ/ãïízTWRÇ·ŠÙDÖavDýæœ~VM ì(Q»¿ž¾àpÂÈÂËB†ÈQna-â³Ç`|3É z—e:f€»»âÞðЮ‡B~%vw-ldQ»ûŠ¢:ÏÉÅ,ÇëoW}(òƒÄÝ¡‘z>«ôNÆXûϹ ÛžrDóÕì6¶x²YXã®ÿçâë©¡üÙuµµZu®m :Jý«C+›“#M5Ó5›¹ú #4p~Alb8%ûWêÌÔ$e®¹€ãòÁEzñ W<†V ¥)ž:”]CFŒ¦.¯X©GèR±FL¸úžs÷Ì49†É­ó½C‹>̬0gcDÉyrµ–0ÇÕž^’î£ÏüÔUÁˆ°©6zxnc¥º¿ÆØ­~cí-‡!§GûT™ø‹,€÷®³°Ý6¤ðVu›ôUâYÚ<‘ygDʉaœÅóÅ1ÏxAy©Ô‘¯ì¤ÔUHú´Î¼ÑƒgE¥³o° ù¸Gߌ»ÕÞ¸±:Ír|ßZ~3Ùàü IÑd»:.Ä. ›ÖßSþÂÀ^ž@‡^i‹Éƒ¸âÝ ]€‘„üK+°d™ƒèÇxïV~Á’§UÚŠlz,e”­¯iƒ&Þ= ë\ËÃŽ_»É<:à,tPÑ‘n½mð˜ßÃ9ºKÎ''=;ÚˆYkñnM¹N`ËôÑkã ÉÓFs]¤1Åf–akãÂlw¸nB» ¯ä nkžÅ3ƒ./gU¦¥È  žŠ”b.>°Á²&ÊÈz¤Î°X‹ùE(Ñ{= Ñ3Nš³ŽËª±¦r áŒ]©Å\°¢(`˜¦Xî½ê0z@Ú–žwý4{vT=¾\T³ôoí´ûN@-ë쯆egi¸Fƒ\Ȉ ¹Ú«¡<£Î~Œ×›%«~…i‰uS¹ôØ|ÜpÀ˜:·Z¾om‰lyGA•åª×ª`'ÁWg™àu±¼Ü>‰ï*ŽP^³†OØk®>êüNžÜRzoÖçæâ}ý²¨"õNz§]9Šœ ê`žï]Ì|ç}½ziKg¬%ÒI†IÕQT öf-ô#]g¿p3bÚØ<¶§3cHÃÕföž‚ï§yn_Eå²Ë t¹Õ`T ¤@  œ–­š‹lW7ëø1‚·¯YŽÏɱN6ìUÕ¼ ª©”0Žˆ=1ÉbÏûiòÄ븛ȔjºöY’reר9Ýa†‘ík©d²m®yÜWèC§rø«\¥ˆf¦O}I‡sc"s 9¯]ít¦Ï(÷=dŒÐ>vÔ—Câ´Y•—ébWí=çÑ]VY6ÍK¯–ÄO%0ȧÇ*Ën$*6¨$©’'KZ3rM|­Ò¬+Š@Ïâi—©f§Bº#Z úÎr&=°’CÞ'qŒô-_ÄnŒXÖšS$ÜP70½¢/“‰B¦ynâØ\|^ŠKº‰wN+ o¾mv,UWh¶•È“J[vTŠ¿\ä±²8XÚuÚ¹ºÈFî²ÑÛ”8K¾ÙçK]Ø<õ#Š¥4mËz]/‡Š|eÓUräÙÚÒNòH æ×Ók뾈¥9.¨OÑØS„œV“®áò4Ç «±ÉÊŽ+sn¡ð*bçèCàg(¡Å?§Ñ^4ÿzŽî¿E£˜.Ù]÷P=\Ç*œµø)Š'oÀN™'­t?UrÝ ×Î{gŽæ’8 8žQ ¡ößzCNn-IGå²Â÷>×ÁôùRášN}™§iízôvUQJ·Í„÷)%_f¸Þ[Ô=ñ…œñ·s5\‰r¥E$ÌæfÓ<:ËÆl{Ó_ÒÙ=.ÌRÀJ^+íµàE°k ñ÷`ÝÆ‚±VD æñÜüµ«?• t©Úgò83ô®²jEetÒz<È.–jfX¦p<ê  ;×ýöÁ­¿QðD¤´‰qDÅ;ÓNÈr3=“±ðpæòKáuT:e­ÿKô2zÖü+á°ëËÅë@«4÷&@g “wV{×…dçËý9TpoU1¡ÉKƒE3/ú7ûéç_ˆ,ñ[uQø™Bc%0ô+vìäI“®ó½]sà¢(œÖ7_g¤‹~<.<ûðIUí±òÜIZµ>+¦ÀdþG„W”riÛøõ„ˆ'qåôšùì2þ›Ýr æEýˆ©8‹üO…ï/"âË.¤ç³_ZQSoºš!4±2„,ö€AšÊ€ªL⣘ ó²S{l×ÝD‡D¡ˆdbld¾"fJ£*ìV94ݦ{øìË”WñᘷìGÎ2œyõi)Våm¿-aÄb ;˶'2¿cÐFÞÏ鑸­öˆZI¹RJ¨*·ŒYV?®ÎãîX„òRò¸¬£À„f­8 Nc%YàV®Äw®ºÜ1î‚ùÓݽwò•øÅ]„¼µ%<”w˜q½$ð{iûßl™ä¾¬*­e½2±þaö¬.3›ä{ÀAj€Yí›q³”rq´ uþF0ná³ë××ê§hZ?qu¤çÉ™¬žÚcúÓÓ··ÛæXÔ̯ ˆ,A,Š­3*w{C 2µu¬3mqÇâ_…{¦Éñ³y¨¼|ÍÐ÷p½Rr:÷}³°@ˤš31#†fOØ.'©,rú+dÓü:ßJæ(˽oT'ÆÖûÀ•²×"Â{f s•©±M{»X ²½4¨M9L^´*úÈtÄ1#QukÕÿÂÞI‘+¢oOõšéó?’Vºòè^öÏCÔ9¬ `xXî\»9«Ý—z:´K?ž‰‘»qú‹á䬤ç/<+îæeaඃŽÄGk‚d+Ǧ¨'ùjô­­8RÚ`ê²ÝËz¢ö­¡‰Nï‹Ô6±†œ—qÊ/@^U€±cNB€S{ö'­ýÀR‹dV¿ñEºÑ;9úù'&Y#Ãì=<;ƒ'‰Y.dÅü±r4b ƒF&ùS¿Ò(¿"¨ÊQVYŠa tVô!rJ¾ðjðÊñïJõj Õ5` xžZãxjû.÷à¥òÐôµ¬ÀlŠX®U°y´ïÔÚ¿h¸Œr•KhÚu{þêüZ endstream endobj 285 0 obj << /Length1 2048 /Length2 13919 /Length3 0 /Length 15157 /Filter /FlateDecode >> stream xÚõPÚÒ€‹bÁ%xðàîîœàœÜÝ‚wBpw‚»»;'ÁÝ.[ÎIÎÿ^Õ½5UÌ|í½V÷‚‚TI…AÄÄÆ(ivd`adæˆÉ«ª²0˜™Ù™™Y)(TAŽVÀÿÈ)Ôö 0ïbö@CÇW™¸¡ã«¡¼  ãd`a°pò²pñ23X™™yþchcÏ 7t™ä26` "…˜­›=ÈÌÜñ5Ï~¨i,<<\ô»D¬ö cC0@ÞÐÑhýšÑØÐ  bc :ºýOj~sGG[^&&FCkF{3Az€ ÈÑ  tÚ;Mµ P0´þÛ#"@ÕäðBÅÆÔÑÅÐxXŒ`‡W'° Ððš "-P´‚ÿ1–ûÇ€ðïáXYþî_ï¿À;ÛXÛ‚Ý@`3€)È P””cttu¤‚Mþ24´r°yõ7t6Y½ü]º!@Rä#ÀðµÃûs0¶Ù::0:€¬þê‘é¯0¯Ç,6³±¶‚ÿªOd4~=w7¦/×lãöø™‚À&¦µaâdˤÙ9¥Åÿµy!þ–™ÌÌÌÜl¬ èjlÎôWU7[àßJ–¿Ä¯=xyØÚØL_ÛzL¯_ˆ†Î@€£½ÐËãOÅÿ" Àdì0šÀˆ¿£¿Š¦ÿðëýÛƒ\:̯ãÇ`þëóß_º¯fb¶rûmþ÷3I(©}Q¢û·åÿ*EEm\ ìV66€“‡àõ¿Q” AÿVñ‡§4ØÔÀóO±¯§ôŸ‚ÿê׃ð¿±l^ç þ=柘9˜_ÿ°üö¿]þÿÍø_Qþ_ÇüÿV$édeõ·žúƒÿ½¡5ÈÊí_‹×¹ur|Ýy›×Mÿ_S à?‹+49Yÿ_­´£áë.ˆ€Í¬þ{ I+ÐD ählþ÷lü#VûkϬ@` ’诗ÀÀÂÌüt¯Ëelùúz8¼Žäß*àëîüoF °±É_KÆÊÁ 0´·7tCd~$V€Ëë6š]ÿb#ØÆñÕðÚÀÔÆñ¯+åä0‰ü%ú‡¸Lb¿‰À$þ›x^'ì¿ÄÅ `’üMl&éßÄ `’ûM¯1åÓkL…ÿ÷k¥ßÄ`úø›XLÊ¿é5ƒÊob0©þ—x^3þ¦× F¿éµjãÿÇ«ÎØÆêõ°ÿ#ùëÌ™LþÀ×€àk¦àkfàkæÿEö×£4w³5þýUú_+µü_˱ú_kµþ¯;ÌôG¨×WƒÉæ|­ÌöwîW[Û×§l4uü-eùWúÏþWüz?¶¯»gó»o¶×Nìœl&FÄò*µÿ_›qøåµ+Có? ^ÿNó:²LŽ.–üêáô¾vïü¾vïòY_kwû_Ûuÿÿgüì_Ûsüû…zÝÿðßÿp€@W 1ââœ1_ Euà÷ÛJ†1iŠdEûV§{T¸DšŠtÿuûk‘ÄÁ®·+ÛÔWÂK$OGMµp!Íñ[<õc•'wZ&púÆóDjz‰T…w=Ÿì<Õý,¡› Ûe(²íœ¸Q•r1o]z¤\kz‹—G‚çv>îVpÊ"=O1D©E~òû6C‘c”1ûŽ Ö‘žãÔmæêz#kü…D&–Ñë8Ší«‡ökôݬûj©*«C9žö;"è+Œ‘IJÑý/2¸ó…‘ëó™Ì4Œë#½$žÃí¤?yuä¤1p™­ìy¤FȽZj˜ü4»‘ÒwT li«mPçßÑ™º¸HJå*6ôèí„ Ü„ÚÓ™È;ÇÀ)¸yìn.Â\i]Éý¬×qÒ1ÝB@ÜÂ^i] ±óÂ’‡%¼ÜåîðÌ·H+"¯òè¶?¶vÝuUií+‘øå„Š—Ÿ˜‚ö3?µž‹b0öª “ýBO‹§NA¼%IµeF»³ÝgöŸ9#ÚÉܲî..„unV ãÒF $Ʊ?µ±K¡PSÂéã¹²òç‹e«xoß Ë,›” O¶HCŽÊ‡ìÝtêç<˜ ¿7£«¾?hØ< ‡™ÑÎ1ôilHÒ W¸Â ÿÀ¡Å´´Í¯]Ý—¿ŸŸÒ‰C 1úUgxŒ2;J©‹}öæ%òì´wæ¶–¦ ¡ :eé8¢ƒCç’]BfÊ)²E5ºa£ŸùU¦•ü;^ÐbJ¿ \oéŒvÇ/êV¾÷NÑžîÔ^pr¨ˆ¬©ê÷Ó ׉>€H£Ä¨Ûižå#ã!l Ròºój‹]oïž4¼š³`Ô‹Ê^þ@R· ÞÒ;¼ís‰œ–o˜¦uµnœÅÙÕ*OÂ<^Aa´ßÁ,ýlkêÁ,ÞáÉ)\²—uÁ§{ÈTÑfî9>ÁxûMYfm°RvÚ:ÝêZo…RcÅ7bµßƒ¸!÷ôÉZ²õ¤„(£6©`i6eí³¤¬·4NðW.aU.‘ùE%ÔpF¾èÕÏ‘†è@ÁšU™|^Õq·™X»½ZšWóN%$ßó8ú[l—?‹J/ØÌ Wà<esÙ~lbg6†ü23Ë~ê¤sŒâ¢ÕÚý%5s'ƒõ»ŽãÄÔàЗ¥¼ÈÚt!óË”ðF£ôÍgÚE"(r_D“¾¯áä¥ ä ´0^¿·dþÙgE4ÑEžj“'}EG‘¦¸°w%˜âuùŽ·ogL’º¾e ”¦¥sŒ^|÷¡#¨ÏÿeÁÕIq~IÂÙÂ@§Áƒœ Ve‚˜° ÅW[™<¨ºùÜÄ’à(1î?Õ°š+Œ2™ë‹Ë‚×W-©§,ô…!#1㼫À'N/ô!Š–ÅÌåfµ#‡© ;«‘­]¸dëS%—û²¡ûá$ áMXãFÊé…_„„«¿Uó@ñÍ!G¨¤±Ø¢u«uvÎkÿ³ rÔãíÈ>›¯Q–´nHzx§[c5oôÈÓ¹œ¢•ÈøÃ³‚‡BS³cŠvÛœ~Љ"15•GQÌlõ0ªKɳʮ 0Š[üålx’`«†u|¬c׋øF²ÌWt¶Ðf:ª9ö'¦L@gÒ/í‚,X¤£ÅªÄÏeÁ¨Š“’Éa݆›i~·ã|«éùJÖ"3tÙêóCî«CyqiÆ8$!ï ØôTQ³”œ)ñ++·;Kcuo”•ӊžŠÁèUp<ñ¼¿ù¹I'Ô/zD¾·ÒOs(WÍûZƒj—:âY´WªÁ>nD5'8qG ä ƒ(Jo™]¸GùÖg¬D‚,Íé¬eQkã¾ÈàD}RÏ€¢–:±lš Ö:åžI– Mvk…J0 =]bCIžW~ß#\…MŠUU˜óSÖ†Ã鬎f•ïòþ•$£yr÷›-‚¬ó+ÃâB—‘E…r5ÛÑoee—[©òe†´å£&{jHŸaàh¥ ÍíbçŸC‰i™¡³í~ù6WùVî*`÷«!ô²õ‘ Älp¡õøò(:Öë°yFrØ3C_4Ùö-ï ’a'B¾ ÂiuüJŸÃ ;P-0­µ‚Á\ŒâQ{’<”©"ÃбÕíBZz3d;f¹Ð™=ÖWx¬Åí;>ïž³·Ö&,Tã{ûÈK,ÁW~uŸ£ñB1¢™xC¾‹k4Èc»]+©g’*:F<'£L ɩàJû“4=jÒM €J¾î‰© ¾âçÓX_U2SÂH2!àMÒ³æ7Ê8嶉þ#$ÝBV9ßÃàNä3!snÊl&Q÷¦×_âIq›™CëNß6Èó÷F@k2Cº·²¬“s’c7x=øið¯³X쀶¸8÷žÈÛØ«€•µ5€½"¹zç±U§û6M¼«çg^DuµÃ…ÔÛ*Cã̌ñíÑ,k4H&8ï/†¤aGùZ|•tq1ZëΚp”!ÛŸ÷›¥¡éÈ„Ò2âý®·D³ÆÇç¹=W–¥Æ3×VO>ñ5@F÷m³†®W¸’hŽÁbe³ ñn`o¢7}Ùy‹ãøž¸{Gˆ¢iò&Ñ,0\¥¨ãžš…ò ­›Ó$Ý‹#C[0ÜÁö+r:_¹;å6 @ˆèE¹Ö÷­Ø/cPÚw좠JÂbÄàÅ,5é ¢Ãû¡°¨ª˜¼¹„x·£Eþ+È€&¿«ìj²çÝ<ï›pœ_í‰èºH·Êº’Ò媖®§ifF™âMà©y;·NJÒ #' ’(òmíÚ‹/q|JõÜ]é—<ôâFuíÓ¶8mʂ뚮rÖJ?Ä«®ˆ=Ïù&ì¶üt¹ïȱÜDCmäìQ|=L;êKÉ©’·½^´Ww´¨÷Å÷Ž&¹8ib îÆõ-©ßü9¤¼G}º3€CÖ¢“B Èðm]#wªöÞqC ‘ÚâAöwtˆOPNë2Īç÷eñB±!¹ µ6pÀ˜ãU ~hßoïá3_¦¼”{°s¿É¸Ö$rÏL+.o2â~ò2pZ}ÑÁ% ©ûvIqÐÄÖóÄÕõiÖì¬ CRwø1~žò”-YfYZTóæ¼MÑ¢=©¿|}i—®Nˆ?;áP$?«žAS|›üå¢s®XæLKýl8»ö¸ƒ¥ó Ö佌 ަ|í€&YošÑ·‘¨„'mM¦Æ ÓYöµ­HüfVÞQ'I±Z²ã~*;gºG&ˆÂ¡ù¾úp³6ü®V°û™Î·š#F,Õ WvÍ~ø$<åDÌýzëEÅn•8‘°#|WÑ®¯(½ÞÄ…—6ÎäñYÆÝÁÄ~ô+Š)ANëÉ=—ÌúäÜÙmàütÇM­JFÑ6U>»^Ó‘QÝ×icK%#xï>?ˆD未®îÛ‚Qw‚méé2Ò@Õ€Æåbòiú¯Äþ…ÇLÙû¢ùÇ[~”ɰÜ5)ýò.j7éÇìI K=ôé ‚-ƒõ&Zà‚®]”Ý¥|l>mC†w3A‡ûJ3ÿš›ŸÚ®Ø„pp’V^0A–¸nóðºµSFð烗➣ªQ6úl“w*º>[¸zní/¼lØZvå²"œüꉺÃÓMTT”"›`™¤hNÊvì&O mê»ÀÉrƒÂ»!c\&ŠÝy»CvÁ†±çx+e¸™¼¸pÎÄ´dòü#°ÄÉnó[< ÒŸÞ$÷qD™Ï®¢1|Š¿û!Gl4ƒÝI_a¦¦âÆ|(.+'ÅÅ—˜ ¦â#zbÁ„l§ùPá]j÷¬ö §K¯¹îÈrÆŒ|U§„¿}$Hp±³HÃŽŸ5×d¹©á£º*œxS«W`‡Æ?N}rásßÉ!Õä~«ÿÙpL‰€DÁ$‰ÎªD•̃ÄY©ïl¯®¸Œ“ ùÙöË^–±yÀºSŠkKN¥ÂäϾi5V¡s3±_XýJ°ÞÛ"ziIÃÉêøßõ†Ò·M߈$Ÿ(+À)Y‚øo4T!wâó?£l„:Ž)†Æ)õ¥ZÄ\«zÐ~bVE%¦]ÚÉVÈF®zÜd‹÷òóYî´6A³¶ eiíN<ÁPvGÐÐÃþ;û|ü¬0ÞI[qRgžsV Ôas:»GÜT™®ÁX@ õâP²¬ËMôœ:3am᛺Ì'QV!wØ£›þÁ³L%¾«²*¤ðþj ®ýÛéí0Qí·½ÊŽ–>´b«=Ô«j‚»”mÀÀMáy  '*Î]øî)0ÿWg;)¤/¼¨–ªÙÛdØ%Ø*/ÕUÃä±PÿЬ’q(Sì´³¬‚–™çSÍ}:Ç–V‘9TõÂÛ“mã€ùÊìØ’Bw%ÚÃEI*óÊ;&´${V ?qbŒ0-Cа \Zô{¿A¿#d^ÇöU¿Ýº[šiÝçãÙe“éåV.â•4‚LnË–Ð ¦EñY âPôQêt5]f}ŽFÿ$Ñ„}îVK'凉1ž9kû7žÝ\öÕRw¼!ç2öÏõükÔ%fRz]=µ5Þ®J°z¾²[k3x¸Z÷ZAâÃñd· äXB¦ \ÄIK©ë^èMqšÃ?löC÷ØÓ©¶èTƒ€ž\WÞY®·<’mqs°Cñ€D,'Ä‘“äÐ6¼9"‚#:Q\‹Èx×aô‡Œ¯ ¢±¶g}'Ör±IbPø<ð4iŠìâ¡z£­ŸÍS$Ö]rS‹ýZnA…ðÃP²Ìæ·Ìï,ŠôoF:íèŠBÝg½|ØYÊ.!ëMýeyBd™;ž¾|³åÒœ¿¢ 4¡¨>™„ïËÞ]:x±Fây|¡qÁ—9t„Ó܈Ʊ9E*c°A©¦ÙqÄy{ÝÇñ°>tpá‰xùe¥à•ÓœQäÜ{`zKeÀK2>sÖ UÛ^Ü ‰}Æm) I~%EŸø2"oàæ%uXŽ;$Dc·òfíâKê צ,)·èPk´ä ã 2”Íñ8NݬS¨3¢lC ŠÐ(³61Ñ÷sy½L ¨Í¹“~ÍPC ,ŠžwÿFƒ/)jÛ¿µT“s”›3à·Z2b–ƒ‚ŸŸÊ€c¥ÒõÈ*胢Ҫ ‘³3Îç‹ÀýºåKg+rÏ2ÐÍ1³íi+âGWˆI{›í÷CkRœÝÏöÒzN¡‹gì1…ý#À"£›äf¡uª C×Slbä.ÚÍQ„þÈïù1±Ëܰò„+æ)2ûT‡´1±ò](®)·!䙇=~R†òJ¹„5yttiÁgœÙé_,!&ïÓ1dx~Q¦HÇ@½¯Œ÷šE¯×Ó¶ˆ¢CÒ–a¯Hj€M‚彯•µœRŒ” üËœS÷Œ…ôÍä-9´mzlV±~fÕL™fC{ã®ÿ”êœy\¨; ¥Ïz›]š¼mf9¨ %ÁÙ œµ``ªÀd` û mol뀕]xï8Œ@ͤÓ#¼íÉ'V°¡khϹ=«y…¨ÑHõ],"Bç:ìãq˜–Aឌ¾hűù·»³LndR¾AÆóñ~‚vÚžv9ù›Òþ“½4£ö„yY¼$C#Ù4é]ð¤Z M#ˬÑ9[\çl”g¬r›‘úi Õ}Ú.¤Êýœôœú§¸e]МªÙ@D”Žþºvþ±E0ayÞsG~E!•· š|(¯s$bøY tãtñI‘Ïf¼a+߸r´A5òMÕX0êü䤈ÌÈad'H߃¯ "ê.LcÛíŠNŒP­2ë[°y(Àþ ?V¾kxɪ{^)½S‘E½%ÃklQÌVàoÊŠ¨WÇË&&ö™Oñí3¿Î@ˆ@ûLˆaÜô»¬û{ùá–ðúkJ/¹]°SFÏ»¯ìËYê:#tŒ¹úAF‡…äŽ×%ŸYë½çú§;'+<~”ðó§9߬4®ãŒ¸Ï]†ðß#Bœ}˜È[…ే]{WçG–näïXŸ2”Lîv0Q<—£~ ÐÕøò®knjF0ou›:ýʇD)_4®Aƒ–H[l°/-\1óPÅÏሹæÀwÖ0·Ùl®£vG%°Šc,ñ†Á‹+Ì~¦g‘¡3·QZ¬<ƃe )X$8ƒs›ë{L.h›cÝ]üIᦚ„˜ÊE® r‡£Ú‚,ñBùÆM~ÌÈÔ9­)ÿŽ’~6ø\ypÀÛÉúïR:þ¦A?{°)ã‹”:\k)¥1Þ¥¦¼åÇ€¨Ñ†“²÷Ýyýu¼©ä‡÷‚s0“ºµ9{ÒVÎånJ—‹&‘Ó:ÄÐþŽèÍî“.?ø&Í!Œ¿ò|©8ÐDVá>ÆðȈEÍÐ*•a·Ävh:¥"·§KsL} Ùå¢ ¤Y·¦ŒÓù8ƒ²-ÇOSVcvíÛà<;[OÆÚñýÅúV"í‰v ­W~èÚŸ—‹°âÊß:-¹˜´Žª¦‹­É«‡€_ÇðëG ̯°l¼#ðçU½‡æAÞ,½$yn‰øÞ¬IB¨îÉUÑmGþ%÷G5Ëlq^]oýv½uÇDž&]I š&œR„˜¨2| > dkt팄z´¥X«¥¢1jËWim¬Ò/ Ìð"h¼f›»|»+µ­·½ÕYÞPû8Д(PqVyÖ7Ï¥PêSûwTY,±öÛ¾ì„Åýɲ—â¥z%6Ÿ¢—!ž[žO?úZ"VÂ\L°RxùŽ ()ãlP®eÓyaËRûg³äÚ¿U¡¶aÒtbž$ÃÊ“ék§_¤D!4Ý‚WÏÑ…© ­´ßh©Ní£ôc5㨻N\•¾vÍËŒcÕïTL©¸W^H©üZ<¡aš´pWc‹<‡ï—(á\ÀuËìXÓÇ$Yì§„ï,ü8µ‘]¥³4Ãr¦ô¯àÜprÏæR£I'ðéM¡Ó~@˜eI¡ÌÇOµwÆÛ¼É÷…KžiBß•w|åÝ|{ÌuÆyko)ª1p]o´_(o#}4õé£Ý䎕<Õµ³ ´Žnð¯ñ!Ûvx_Gå®7Ù¹–ÃpE9dtæø;±÷ão†ã#âO})N*i”£uªq¼E>‘Jò“õH¤Æqɤ]t¸sýêšVpjûˆ<)€ïÈXÚVñ™ýûé]#öþvwt7Ý«óöFЕª8ôïpIϹ±mÓÛo–ò¼ŸÔ¿]µX1‡¾÷½wž*„h–6tè¡ZZs¥/$ƒ_™8JRê”>zØo™Bň {ÀÇP­a>ó 7’j{ÑC‘¨däˆp\ì‚6ìà½`04aiULlÕ·þzÆ*1ÔäòS7†«‰lRƒ²àÆáèY»ýq•àÉ1ϸ%®œûŠ (æÄJ ¬{³…NÜËj=ãä$øå"6åè|èn_I³Í| Y“9/úUÅ“ÊÔ"jG-Ýo$b¼ÿ«JÀÎ)R*r5Ì“¥¡è2‹Ë ÛZA5VËÆ•Ê3‘hnb5©šûÆÃU=òºd Q¢ äý‡SdwQæÉ‡—<.ç4lFTyXœ_¿(É7oÿóþ8'¤L~?m{`4AõŽ7xs1ÿð_mNÇzZœ _¸VfN£âqØq…ß õÝøAŒ‹dµ4¥òêôô^ŠùÓꊡŊwænãW£Èt6V³ô´[”·žK7‹<^Êb÷íž%“V'·ÂT¶™Ç?wÌ¡U©0²$ëÉ,¡‰îºh£ÌSÆÓüЕÅ<~TBíWrrɯÔÛ8_ô^%BõZ”Ãë¹YÝÖG6^ÓmÆ?¸:mË ò Á¯UÚFû…ƒóÔ`¨“!:t*kÙM®’FǺ †nj^áÄPÏ‹ãM޾ª/öÎULáI¬p:û¾_ßT 2Þx*æX>ÔdŠÆô“^’l–ÃîÀ¹$·ŽÕP••M4Y–®d–Û> sC¾jB̾’üñ> Ï,ÚÎMóSÍNô×N›Ñ¦Nhb Þ2ÎO™6+ÇF’ÞÉÇ»–Õo5ê“D º×;Á©§ä• Ä#½È«ú×üX.·˜’)0ï?"‘6'[?ÑóÅŒÒõ;ÇLœ4¸~Ë4îœ2/7®"ÂÎOÛ€ŸeböÊÕ-R|”PNS8fõâ÷¤Ýs¨E¶ºMm½Õ¨sü\~ƒ½Í ‡ê …Þ}l¢Ê›ŸØÂªƒ ê•w¬KãÎ5SV¼‡{m‘h!»5TÉ‹x‘°™ŠªÜÊÏÞwÎy!“þ([<辋i’¹yuk ¦µ!̤ÂôR0e),e¦6m‰DŽ ázìn÷`„¿±Wl¦&ŸÒÓ– þQ_ðÃ2#pQZ=Gøƒ¿Ì¿Ö{Øwü4$\ÃŒ™Öˆu&øi[SpÈ·°£&| wB{dP$ÈV?üùs£ïWÚqVa+6|r¾ˆØê|nÈRaÜ% ŽŠNO8MrORqbêoc¿”|j§Ù7=ð+ׄ¯:p]2 YµiÃè¾=e²ûÒ€Çnl÷E´º5øiÄwè&¦K·xÒ÷ž7‘˜ˆ‚‘à—‹îýäÁÀ­ÀlÛèÇ/ r6ŸŒMb²6<‘„Ìr׋;ŒošAéd·í?z1.*ÇžH—JùŸ¨a Lš;ðóíËÁfýΫ² ÷Ì)ìëïS;CÚ wçæ‡ð” BÛðL¨ a»Û¬…bH1Ë(,¥fðþæ=QŠ.^Þ=ÆL¥Ãp„úFÙaÁ´ÍÉ8dàñ8îïÕþ§y›åÚL‹WÇ^@ylN'HT5Ѽ°ÃeÌÚEÊû©3wådÈ™¾þÛ€{#(*~Xׯ œîƒ/Œ²\r/o¨póG¾ÿÈ—ëQ>1©Ëwõ„…øY~?&¯¼¨%ô°Ø?’o×§àg,ä¤|tŠ<¡X¼N3~«{Ç…2¤~ÃÏí­D«o• ”9Åà5Ë'Í>ø²ˆ°“ó8%gœÃ'ÏQœ¿N¨î:FµþHdåô?7C¨_º2ßGÎ[rîÉÊ™Klï6¢jfR½¬Ümïéâa­Q/ ýà§MZöÿ² Ñýiµ£Œœ ÈŽ^T+eÎIòê´Ý¿#àyšvèÍ»0åèR[ž¡[©g §šÍζ^W-U‰5ÉÆÏñFnˆÌ¦«<_mï. EMÐ;°k?†÷²¨ÎRýzVjL ý~dE,å¶Q»ºÄÃâ ÚŒ4ͬ$T×Â@â}šóV}”ÌGg0ˆU»ãzOÎÚ«tl¤TýN5ø Û»·ì™–ÎCÐV2ˆêªv 'DuªN·*Š´ŒØâE¼ k}LcU´­š³$ï±¼VÐh çqQ<`By\{‚ß‘tÆ« ûPÕßË=ÍNèuÌßøÊãÒ"?k%¨AQsE¡ãšF28ŒGáÄÙCçßê‰Êñ{÷+ÇoŠr¥]‹8–fû‹5Úø5 |þ‰Ù HÔÅ[ÍÜRXꬉ‘>ð iÐñ­qÈx G‚®L0­P†AÌ °éDùQh UµHH€=rõYH÷Âu=Ü‹^~b”vò¤ÌÆ”>«ñ ˜[טj–C†Õ ›é‘É’ÙàE‚øúv„¥\bDIw¶%¡â÷b¯!?±0¿ìé1¸cFR_2(.NÜ Çྂ`Z·+5Úœssb¶^Sÿ{"Mç…ó¤öüÑÑD¶ –Ë=¬vc ºlØì£R‹jAíÄm(¦aª†Gìì¢Úæî2<#\fëâÇžë²+'üml(xJ1%HØå£ÎA:Ct«rÌç@Õ±þ¾ÁY6`@߯ù̱Ñ{mH…ͪÔÛ°MCió$–vZi"Sþ,±ZÁOß¿ö÷Úæ E]âÀ¨j{[cøÜù} ¿H‚r0÷ŽH‡÷¦Ç»”@TÓ xMжÕΞäN.ÆtvˆhåºÉƒ1ð  ‹Â>FÈËû6$›€~ñƒ°s‘=Töf ºÕN ŒDÿ‰sÔ"UK5nu7_18CSÂ;lD¼árË‹©’ÆÈNÀ8ôñ ôPÓØ¯uˆeË…JÒ0æh =4í,9€3˜³òWÝ.ÂJ9ÆZÛa<ˆ„¢VºmÌL™`[Ö9þ³‹Ò ;|A›7~CÒ&ˆçñ^“ý4ÚíÀ4CÍÒ>d*K•n )pàŒ7w~:£n€¬z%¯øÇ@Mô²VÛa’›m³ ±syÄÇU*^O%P†Óðj¥°Jj-B?üÝ6·FÕpÿ-×>e®L›w2šü™8VÍ‘¼_Ý#—Cßú¥ ½¢D™Ub~ Ó¢³/Ç«e¦C…·+=aIõj}ÎÆ‹WA«ù?]všñèÉF8U+àäZþ¼þøŠ°EuzGåšÜàò˜µÝbF<\ž”ÅÚ’ÿU°:^½§úºÅáj¯#¾g‹v¯5ô;ÂÈIâ—”(9ÁI£ Xé6ÚIƒ{}Ýð"¾ÚDFû-Ãn•F6ëÝù6wLôfÞ‰¤Sè¼*ßåÅÉ\©(Œ+Q´¨àºo—¨L\ 3†KÅ“Ûߟ{%fa¬|À.Š,ÖȺ”b‡vUãFQ øÚ7ÒÚ§ÆžÁ’™Êñ ÷ý'ß—m·÷Ë1òÆ-±ä.Âoâƒlˆ‹DäZ¥ùé‚ Áa)wæUò—Ý«Ò_,´šÕrž»äØòWË"ìÈÎù™C÷vƒ#Ö'¯ÕŽú?m_ËG{ýÄÿ%‹À1\-Ô3~dnó\ò†BU°¾MÞX®3}GŒ3U{¶>Ƭ ©°þssK剭µ™8}ºÈQ4mÛÇhñ‡òÊÈ+%[gf¸ÄM»£Ñ{zåv¦ª(à´'1ÒìqËhàO©2sÃiE†Ø·û5Š?ÙlUA茫؉¢Ú”?’·±}fÖïè5`Pe”óYd;z‘Q«‹úAäR>D|D)ôΔ3Ò»DÒ|lt´{ƒïõòªï÷‰•)šIù#/$Õ5ШãµdH³–ý†÷ùn—™!¤¦}¡NÙˆ[i JŸÅ,診Üu©rº‰MØëoß1©øWÆxÀ¤­/ßÑ0 §Ýî:y‡æ¦¶/fUÁqœ àGÒ˜´‡¬Z¿»,Žp5â©ÁžZ¨à:SS¸hûÆý•_àngÞ1°6†#j|ñ~s{š´6øD4Zm"ꯑ¥“4×Ó…j™¶û¹f"1§¥‘HPNØ}QNú5¥%DR|$úÊ€ŸËy.y±ä§=í\¯¢©árA#ìjDæÕmÄøª1S*¸±êÀ$Bõ){mb\=ôĦ­Ñj‡|+•O'Eù<Ø{r+O_!«7X¶ñZçDÞûöÒ½}Zt"Dºü£’Œû݆XȾ©“ë¢ÃkÔùãYË}€$µrPé åY#þx,¦‰SÍl¾ÚÈ¢÷IÃÓ¨„+kÛŒ8 ^ž}:–Íw÷N¬»_/Z‡v¸lð±­r8 rh» (Q6ÜS-& J¤ðhÙ{9v7ç >Z´~b#ý"‰P÷Õ_äŽú\®wü!ð¡+ŸæÀª­ñFáþòÍ×ÐÕ>‡5üÌæ"õYç}«z|jîG팻àBØ{õ£8"¬j¹T¬ÅŽäéÝÂÙ£y-ަAJlê‰T?ÅHÿ#Ñ¡ÔúÊK»0®ìåâñJ9„ZD+•w1+S5Ê­.v/¹pó-øíÏã{6¤Ò &ÍŸóÆ-l=`sgq]ÎF±C–Öã_?T'õËy&.¯8’ÕÓÜÔÇU‘\ ëñÖiûÉGžã0'²TrnØXè³y²õ;ˆ^;[_‰BX³„ ² ×€¶Œ˜õö{›0½r­Í2g¶[D¹ó1ü¿Â|áSKã×uòñD+á¿ýúa¸–à¬éú‹Ë”ÊY˜Ôæ²¹»×2¿xUP×£kŽÜô5ëÂá'³Š.Æ©„YÚæ…»–Ÿ[RMõ¾©®"´K%nT\å^B`ɬ/؆æôUp”Hù÷áçͬ¹X+a¶:áS3\”ŸV ù‰èÈ·ê¹µ¼ë~ñ©Ø Oµ‡Ó(ÛÙ™T.Š+²uÓqÚG§èõjþÙ#±‹[•/DyÔ& ÞÉB‰9˜ÅvÂÓjä3£,z³ßËÅÕcC VkC¨¹îdº?FÌíøÜÇ`’ W•Ì›¢Nøº»‘R¨Ê-©å¹ 8CT‰añ7Å¿7¾U‹2ÍÔg¶^ªKË.Ñ)‡Ÿíj¶½¼;}ƒÝ€ºÃ‡¸[µ8À-É'Ó£åUû„ß»½é¤Xm G@…Wàs]î‘–;« ÷‡èQ}³˜v’•8šáµKm©®árQ+y@ ߊXLª9¼3 Ù:í"<Â{q:Xè»ËéÇJö)¯³¸!‹"Œ{±\ºÃ+C¯ÈæW;©¯ž ã(¡ÞD&Ñ–ÛREÕqö¯cOÕï;ó+ ñÝ·îó¾?¾Ïŵ½…ìܧäw. éÄt#œ*¶ 8?oÊcã»ÑômÛ²Sl×ÀáŸGžý>½çÅ_äã`ìô%5 ô¹RV øç ?9|€ñOi›’~~ ² š WûYt‹šÙ y·^g Z ¨Í£êS†lz¨úâ*ŒÉÞUÍy‹zQ¾°£2ò²š²÷éNŒçHSI ¶õUéëƒõ¯î¼')ò$¼5eï>gÿü’o½/HVZ#9»=t‡ãX¬»÷>üÆ’‘>ü›™„í.nÀ꥓ÄGk ˜ƒåÚ…}žyí™~¯UûçÏyŸÃéæ>ÁG;bHQ¾9ÈÌíBÜÄÕ@9]Ë…îñ“L,ŒòòkhSì7 kp(1|•üµ,Q:¾ýC5;¼z¯¨,ùÝyë¹ÕU! Ÿæ«bšCYhg7IäD:耚íCñ·L™c¸©w Í¥Îü‚I­jª*¤š¹*Š >ý­@¡îr ß"Q‡É»[C“MÃé=Î[E‰-£Öž‰û½‘úeŸßÞX9íi#yñŒÏ‰AÓx½”ˆVõƒÍ”Uì¢â`=OÁSêMopï†b¦¾wO vö¯¼z V3^Ë™¢·QK¿™6y’?aLUb'DO@Q/7<°QÜûÖ©R?nbpÚp6T4ÇÐ}í¥2VÐÞèŠ!~ŒºÅ»fb“â:~å’Þ|…£,¢ùElº½=YY^qB&m®ˆ*þÌ<‘/qþr{ÖÊ·pé«Ù¼sÀüAÞ¥6…¿²¥0„p¾/ßYØÞ©‚³~ôÀ¿X) àF›Š GY*³Öéù#àhÑQž—c«£æU.¸Ã-âðG¤&câA‘GqU™ÿ#Ñ55O©÷Uš·=‚T"-‹-uêý}å¤p¡gQ@êpEOÛÞ)„L·&Hp_jË×4ë&AV¬ÿÑñÃUIœ€pR܈Áùj:’ú NhÝÓ™¹Ú) GŸoø1‚˜O8ñ¾JMá)ä:Q,Öãzù¢j„Ž¢cî å4ÎŽÅ»¶µçŸõ!LafPÅl"¶ªdšýùÓEŸÞ§áÄõ×cÁÀ ¾»Õçœ"$|oÙç£9t¼šenº¾ý-zÙügÛ!1µˆY ƒ#Y®°Q¬ˆiÕ®èÔZÀùu+Q¤Ý ônŒy¬\,•Ÿñ¡ý²U4ë}BˆëšažÁ[®[ü¯šRíçØÓßm³ ’0ìÐ~ŸpvH¥f Ô¿Ÿ!óe§ó…¹3¬‹Þ£ü©ãmç NóCÛz@µFï6MÞÀyD,)-Â7µFÉ‚««N‘<œò™Ëc0S奦žŠ>zêWâÕ>ÞKÛq¢úÞFÒ¾ÀŒ’D8õJOùU;-UÀH‘ùŠþDâÙZšƒXHùÈQ#cQ„Kp;3G ¾³¶  Õ|^&³u dn@cÖ+6î5IÿÉÊÒ}ÑL¨*ާ¹ÄÒÒMÿÑÑ ªË9oWÇí0ôjÓë‹n|÷è¹BÜÂl2ö.ŠÐ>ÚrMp4Q·ÈÁžàYf4‚‚eÀË3hÉ¥px‡ŒÑI¦yXy<¯33ÉOÄ“u‹Ù9!ù*¼€ÚSI§e¯±c£ÜFWô^T[ Ï¢€Åò)9Õ§bT5½g~¤ µSrdž¾Þ*ù:½aJeäPõx|¸JSŸHZ¿F|pß.x^$Ia"ý R’×Aß `1ç=zŽÛâ»égQQ%ëypUiÇ¥¢òÕÕ ëEMü†Z“²j=ÁeR‚ìÚ¡FXÂì™QR‡mv]ÄãcwDZj Z‘_„‚ “@DkŸ]#½íVºî¦ħÀ5€ˆË}ž‰'Î’i"—®=ýDC<^ôȹÜvì5ò#ŸUÜX©»3Ímþ Ö €ªÏî™;®¿êÑ-ûÁä|ÃŒ,Ótä4ÓD¡)íDD‹›>ÀW~7fŠ¥`½½±%è©A¹D6WRE.oM{s”©ôaá×8 6†œñ0çוÇhWœwϲ·òʰš_ÀKËL×Ò‹P̱ɤßø©òß$)Òîé8kw»£8 ®A™œ¬1Z?,e~öRéÄž¦È}º B•Kúz5”zæÛ à‹Û¸aƒèqïoFÐm™øŒ Ž?S(í‰w«J‚Lk…ݹð~½ zCR©à¨|@yÈãh2㣒o§â¢¯Ù…·âœ …È `ºR¥ÓHl»Pþr=à ½ãóvõ¬ä¾#:k!‘)³Š',%rD:ÈnÜÓÙò5H.5|,_‹´ž™á‡Vò›îÙ ÉG•ÒÂÀ’Š;huÎÛø)ÄmlÚKÖ¢\x© .ݲ€‹À"Ä82½Ëë±pö&¬Pòä¨I‰ö­K?£Ó‹XEw•G u¬§»*̹i9|‚ ¦îüº”óoôÎÃéýtnx{ò»ðä™dTËgªNÓ2Ôb:ÒÒÌë)  Ê,ÊÕ‰_^'+eÁ|Ø3ј‘¹tÊqj DY™Y¹ýÄ^çëq$‘›ßh.?jx´ªX.pbCÓ&°® g§8HÖŒÞÃSéœ÷}&}›tÏãÖÞÎ/'^„î=,7ñ†šüZcêÍ÷õÔ”²“°Í«K»Ÿ¸q?‚]ÅböÔû¯¡¬Îyb w*˜žÙÊ-YÞÓh`ãU›ó˜]‰2°«mÛœH …ò$ûÉç寻#¿ñÿph·œçYÞsˆÔ´çÙ±€`Å„'´÷Î0$;½ÿdñ«dâ»’°ß‘—Ìñy =š£à›ÿrǃ• endstream endobj 287 0 obj << /Length1 1680 /Length2 3950 /Length3 0 /Length 4992 /Filter /FlateDecode >> stream xÚw 4”mØH˜’²¼ÙªGK–™±eß÷%{$ɘyÆLÆÌ˜†,!¥R$EÉRʲ“¥‰$K!ƒ¬5–7J}–7ïûÿŸó}gÎ1ó»®ßµÝ÷ïºÏ—°sT0À’½@S2‰®€RDjF6NN(%‰TVD"•`p¸NÛaðà •F “470Œ¨ šÙŒÑtˆhC&–þD¥  Ô4Q5‘H@ ‰ÔøM$S5ct Ø(–dHƒÁÈ” *ÁO‡êüþ È`d”†ÆAùá€/H%`Ð$ÀMǃ¾PE š8’1ô¯2Úx:¢‰@0 E´/M‘LõÖ••:pi 5Äë#‡Ð¾à¯ÑapÀ O ýt8’qtš HÀ€$âO‚Tª8ZX¶ô“lý“ ü:¥ˆú'ݯèõDÒ`4Cö¥ IA’7€#AÀÖÔZ‘H—Ð$ì:M¤‘¡xtš@D{A„­£S{ Møk>†J ÐiŠ4q}FÄzè˜MHX#²¯/H¢Ó`ëý¨ :÷ įËõ!‘¤àßG aqëc`ý)gÁÏ´0þÅL°?6o¨"‘HueeôÀ@ ±^À)ˆþp¢ÖÍÐ ¡Á2ÀAc€¡}Á‚iè SýÁÐàŽ# ` :àzH°?Ù!3ˆû‰¡û§£HH~(¹þùç×1HaX2‰ô‡þãŠFGœŒí ä~üÓÐ+ %e@UIPÓPBÿÅMøÕòO¤ G4~6 Òï†~)@æ×zÈÿÎuˆ édþÈÜ©ŠÄ@Pÿg±ÿùÿi|=Ëÿ*óÿvdêO$þðËü$ü?~´/ô‹éÖŸí€ ÚÒ©.àÏŵ±ßÿz-èhh HÞÄ’@3%‚X;ƒÿ¡Ÿfçõ=#H ™FXYù´\èõ A’üá¡ÝùwE†Œ]_2%U5M¥¢ƒ`HHIJªª@0 ÚF,øCÄB‘D¦C!4](€#SaëWªq@ ×M?‘:€ðúƒ4æ¤ ù0d"ÔèoËz¿ìˆà¨ p 2€ðÞUþ¨¢ ¡ ܘ²6@¨>q„šóýQP+b¡C7@¨ÊŸbj‚´GÞÐ; j‡ºBµi C¢Ñ4ü”ƒþþë:0þT*ô&ýØè®~ã ‚Ø›^2F+êDqTÕçB1†³CI…·™ua‘ÛeÌD$`Â>ÒH{eA93¦Ï ®gï㓃I‹éu­1“<¤Øeû0yº¦Ã‹µŠÛúE‹[N-î¨97Ï!a½³t5m´ð†ë±èNá¿è«ø“~n'²n¹?4ò/`·íÊxû/Øì>¼“ŒD<]YÔÂñp/Ü祟,=>w­÷[—ná~I+”ýuáLîz”ßÖðטsŽo5&ù4Ý–¥o<¾Û½½ óNg¢Æ†5þhmñîaêÒ ÓŸ»WÔú–ÆÑÿP±<©XjY¸u\ÄbI€ÖÚHVy(Z.‚Ĉpr„mÊèš“EÜ0ÿ*õÚ¤ÄTä[ôS¬—À-ùäŒZ~Í×–V-3w‚0†ÏLe¢œ‚ÙØ¾Rm¡‚•s¶×+Í”¦Ñô÷½-Lƒ¦¤Só×D¯´Õ¤Uâ¤ÞiZ«f5ÏV&›–=™j¸ ¿æ@\‹¬«Ð9ð›ÊD !†ø"µ%¤kÓšïÏ÷Ïj™´=ȤD¼v¸•hQ7Ymp7ÿt±tûRP«K¢å¾bÚÛƒèR‘[új CûiÆÌ$> Oâ|è†ÑUÞjñª^} ›±Ž¹Ûã3ïWù÷ÞÚ\6uaSqvý}ŽZˆ¾i‡elqí–ô0ú#)–í-¾­^róÍÛíqo2¥c➊ŸL˜¹õ¤òh¿šÈ‡Í.ìgè'ï;Ñ’<5Æ<Ô[V–¥ŠC‰îNr.Î ':«i„|asèHñT¨;¯®°bÝJ >«0)CÖ0R—zpÝ9 VèyÂTžV5›òѲüý$“*êWÀeÔߛ֔Ós75üÙÙoñ¡Ê<ó3{Ä–b"©Ò‰fƒ’&ÈJßÓCV’Òˆê)Vˆéç“ä´=A(yWRcpgÛ¢™ÊbµˆÖ=—]²Y¢Ì[n0É-i·Q:1©`Ú‘›³2ËsD݆‰¡ÙÕ;t&ÏùXËQ‚ôuN"n‹´Ð_ùl÷êÁàLÁÄé.Ê-ñ—³ª_L¾ \ž ÄëX‰?k=Ö]š…{óíç3ëM“u>H†Â¸ý¯dÕ\êíOynl®©Rƒ£Í¶Újë‚Çu{H‡œ éÖû»ä;œÒÚßû|æÐjø`'{pÌÀÜRðµòu_ø‹§E¢GûëCF,˜\¶1…Ek‰^ù·{Gf<²75#Z6¹é\:FÊ­OXû"ŒýÒæølî•‚Ä=J••Œ-ªðñèçm!çÌÖZˆ÷ñÔÎ »ô¶ìzÿÎSª‹í»$ ŸhpYO6ñÈ,â|€òæÇ®X~kàe°ø¤"“]kOLk~ü1žÜ›,J }À YõÑ}ãs<Ø Þ²(jXÏðëš“;+4_,ص¥&ä°ýÓhÆÔu]]¥‰Œô¹x‰ ’W›Lün4Uáu}÷ŸQW8O'å®)ˆ½½;t‡ÇêJø~Ô„Ž¾àüël÷s3öTó<ã»2hî,rµÃéáÍtÙœÝ ’6å\Õi‹QQYyZ¶[f<'=2ÍÒ¢Ø7 繜ÞqæÎ.K#¶3õàœôÃ(!j è‘eIÞœõ¬Òòöüû©»¾lš}V;no/H"ݨʗ¶[©éØ’®ç¡S¼¤:Ï(±(Ʀ —¦—;Çó½©ÌtãÍ—]婪¾;ŽQº©ú1~_d\m÷C WZ¨d×¥vï®Gîf=o‚…¾Ó^²8•*ÑS¿tÑX4HwF,¡ïšâb†å©¿§­¾žÑ77W· ¾ÈýWÇþƒÑ[ö[½Ù~ñQ^PØ\ØõÈ—¥)°216-‡’É/ÇûNwpU—ßa >Ò3﮳>¼—FbV$ÊTfÛWéM ^h†YG×åÙ˜kŠ?ûªi{LSÁ³-úH>´V'íp°é^‹)<f\0ìs‰nÌSn°UšçŒñ¾Ý66ž;÷ôÆ­X©F?uÌÐiÜ#ä¼Ô»YœÍæIÞú~ë¾´^¬#µ¹ÀÂò嬓øŠ5ò±uH[¨Q6>“Ïo,ÐÅLa& o°yæÞyCÞ¡èTöÊ0>Ž­sÌKý¾…{$š»Ï–’£E£HÁQŽåÖß»ÈS‘[Y‘:‚þ®Ü6©=ÜóekæßOÄrÕ8[BL/É\ÙA“Çrˆ]l+ülºÏöˆsbÙò¦Ø)>7Ê˶Óûì£ÅŸ® _ÆÍßy¯–‹|<ßå²ä±”Sêûzí«ì¼jg|ðà²/éÅQʋ⽮ÛÙu݆ä¬ÅJ¯ºË<[½M}Z"Oʱó¼?”WŒfîýu†?QbßJ>á1rÏTø..SĬRmÒðÀþhíhOYåÞçÎ'"ŠQ.왋ÂÔát-3ƒ‰%Ž¥ó™9]£á±¥Obn×íJÆMé»õ^ém_U\*³vmšUðˆ2 ã?‹T~oÔçJˆÉÖÎoúfxÁè E3l!.Õ¶"y­l9Ñ*þ•j澿ï¼â©X.ËG¤2ÜbÍÕ%-9ç٨‹¬iç€PSc¹QšñÄ[Cåû;Txò2L žá€¶è!Õ{µHYÂeJ ï²¢j4ºMpŠ$§tT .¶PÕ©y¨ßñ©æšö yëëäÀ'©dÿßêááR÷x*ÉÛô¹¶{ÿ¹qv}ÀÐÒ<†?Z‰þ(9æPW9ß\ëÞ—àxGæã+÷/33ž+)Ü'äÄÏrà|¿èêœ[°ăQ’ªÏ»f“Ôí0¡Œïsú-Aǃü ·#åV•™nF™CµSE³6Ñç~+§ŸdÐïˆqp”6 mYÏê¼Âr…/®ß›h';¬;íS¤û,ä«…4‚fèÝ ÃÞ¸²7Uc…d‰~å~¹*L~WPƒçpRNß”¹žÎR§:|®ó~Ât^ïLJ{Žõ¶z4”9^ ~ºÙD<âYJ’)òjϰ,Ρps= x+lÙs° ´óû‰ã wß–õs?d›öf y©YÒBGküß¼³¢z?÷°òÕ½ª€î7ÉçVoñÊ Kž fQZmwœ~Ó-±+ýìçö»‘±åvE"Ïxšn¤séù.Ö¾Ñr˜!Væ³Ì'…_É_bE¦¶pÎ÷|$Ì(•æ[9Â>ôå`ÌåɨΔi8²oÀGTCÆ“ïð¹ªó£^óå¢s%—x‹¶‡2"¯­lâöYæñ*ªIIÉro[¾}‚*I”ó7 a¹Ÿ«;ftÊNtW4XÁ²Ô‡ËŸ¤;ƒÂ‹÷(*U¢§6÷\i×Ýf•|„~­€xZk“BeîôÚ».Ô4JÎ(+ÆŸûœ)WE˜Ý’f‹_‰øZ› Š™k‹·WzÊä¸ (èâ¸',|[‹–gµ_ ÿ’ À½¤ýw¯¾0Ã×#ãòüz—íG=Us´Ç(Óf—ÆžæqÓï“7oÛ»f‚ñÞ LïS{k~ ³ìñs[„ÈÜŠ¾|AŽaÑ W%Îá%ˆ§±ýC“/õÚvÇÐ॔Ï{©"‘vÑ)Õã°Õ59gã=÷¨¥ Œ'‹E—é†Å¦ï”{bÃò¢2ܧô?¢îwU]éÔV›ág5ý Üè|1q©ï#{B¼µz²QõBvdò)ÍöØ{KµîTØ #•ç߯Jï¬q_k wb¬Ö<Ö3kÐ98}r³®ÚxîùºÔ}‰¥ÑTmé-Øå¢y¹üøÛ’;J·uÊïˆgeÝ ü¼¤ùæcgâp·íåä´|ƒ†Üú¶ùÈ!‡¥ÅŒy!³â×+W3¥k)G);†Ü%ljÉÁjú‚êÏG”='ã—¿4]GZ(n^Áž—=9¢{(KªkeæõWÎ.aäÇù箑yÎßù-*-¦Õ‹r„ƒÌL³®šg:²ª„u”Øýè‘Äñç¹Õ°esTÎv§÷+›“‡MJVÞìM8t$ªÿLµî¨ègb£ ½W­`¼…ÙØWº91AI»Ò=nB„Ñðy:‚“®Xu/D£¥™Çº±‰ß§Êþî ÷œoó—"÷²Dêcè#˜ëC[ÃMDÊsžÔ ð»í“òîÏ “ÅÇBJê'Dez™¯š¹˜Ñ¨xðük ¤¤ —r•ýûðHSÔ51T®-ƒ{—¹oµ«ÂiÒîô¼Lxp[òmê0j.êðƒ¬WçøÌ“\\@c%±5ì.˜!í—Ò_ò9ï|ßè{.yÈ4‰]],Iî~ä×Ñ<á·WÄÓž½XU˜hnÌdΰžW´ª†Žo÷Ͷ·9+c.xÏŽ½ˆmõ—M‰œ¼½U=6ËÏ+n”Sч d\^°ïöVĸæGfwVepYì)ãÞ¥T~Åum0;ùï¯ÚÕ/ll.vݰ¾™½’ÙË­òภœÜ„çY_±Ç²òaNœšT™ÎU¡†­m³»œþѼģ6¨Ô÷E–¸‘$ ¯sïœÓh<'ݬ²¶™Ïûpà C½Šî-®æ™]Ÿ>ƒ©ÿ¡ˆÚë endstream endobj 289 0 obj << /Length1 1778 /Length2 4836 /Length3 0 /Length 5935 /Filter /FlateDecode >> stream xÚu 8”{Ø~È’¥PG¢ò"[¶»ìdd/냱Ì03ö5;#kÈÙ³ï%k²eÏš²ÉZ¥o¨SÎùÿ¯ëû®¹.3÷³Üïýü~÷óâ`ÕÖå—³B["”Ñ(?X$(@õôÄHX¢äàÐCâ¿Â” ‰FIœ(PÀ `8BL†#ÔAÑ(@ÍÅ ` °¨@âÿ¢1€"Ìi@54 ¥äP@;y`6¶8Âcþù pÃy°¸¸(ßq; çˆÀ á0…álŽ„'Âa€.ŽDà<þEÁ-e‹Ã9I º¹¹ À±hŒÍM>À ‰³tXÆa hÂ?' äôl‘ØŸq]´5Î †A„€Ž@a .(+ <ÐUÕ´œ¨ŸÅ? ø€_g€À¿é~u!QÇÍ08íèCy Q6€5Òh)kàÜq| euTsÀ¢ ý0WÒfI(8V”ån0€¿ÆÃÂ1H'V‹t8Qðˆ†pÊJ(+´£#…ÃRéSDbp±{þ¼Y{Ú åõ X#QVÖGCX¹8 ê£Î.UÅ_%„埘 @@ ˜0@8w¸­à½ž‡â8 > &ðñrB;Ö„!>Hká‹Ò sE8Œ ÂÇëdâ߈ ¬p`‰°A¢(ÿ°Â럘pù¤;` "x €Ž>¿™ìe…F9xü)?¾_Au%»Fz¼?'þ“—G»^ü€_HX€Ü€"ââ€Ï¿I´aÈ_"@:UQÖh@ü§VÂ!ý£×õ×õsÿZ àß\šh‚g÷‹›€ 8áøÿlôã–ÿŸ¿Xþ7‹ÿW²‹ƒÃqšû8ÿÿ¤aŽH_˺àö‡¢ K€úo©!âçÊBVHÇÿfUq0ÂÈ¡l~#«ŒtGXi#qpÛccü ë­˜…ÐFc‘Gï€ ý'GØ+¸=á½%øñ8… ¬Í¿Ÿ¨„‚£­ŽöK"À0˜%ˆ`#!ðÑ á~ì`@P…ÆZÂt>€5Cyt¡"†ÒQ艋‚°ß,$B€X8’ ÏÁ ñ§J ´üƒ ðßBÈÁÑ„q~³<+ˆ8…AëP´9o‚¶¿!Á©‚¶N¶ˆ“t„ò$h´; ØŸ€9 ž "l¡ ú$(sú … LN á†û£Gä8†D[è"(Æœ€yØ?Çq„®'ôß Ðb`XÛ-VÜ HÐìzÎ×íþëâá. AÚñf\ñ>~Ë"î8åÄ.dWÔðµ\ŽÙ¡_èMÇVÄ Ã÷J—\?ÜPÚû,œ#×2²ˆß›¥_*øñÔÞˆdo6㌗`˜¨n µ¶œçó¤—Þ_Áj¸-ª77ßµFYÙ0¬?Óérû";Ìs‘Ïysa4mYEŠóYén‰í:9Wí'jâþûxrF11ajâ F¹´ÆRù ¹á‘&ÕkØ-ÙþÕø+ÊË{Éí=B¿;Þò!ç’™?~¤qRÿ’öñæÃ¥V&u‹A`˜ã}«ãA wÀ—† {Ü–Ø^í8â½Ë|§Ã‚«þ6Ä«7вè±Ô;ËÎ-– %F¤æàÉíUõ' ¬y 51qooì4R8oŒ2ʶÔYU•ôÞ×ÊŒjŸËSñ–‡YÚz,AÈ„xOߟ躘ÚìâôVÉÇÿEµ-œ¼jãq}Ta>¼t4çj—:5û^%.ܤŒãÌ›ˆ»&¼ÑÌ’«¢¸T|7lÆ1R k´ëY­ '>ºSFr€ÏËÆƒ_˜Èø)¢¢Ï”i»ÖÿH B®Ò€Ÿ¿(§P)‚Ÿ–Â¡Š˜øÚ!\ÌImÁA³åh·ÇézxêÙå®ÍÑ›ítùù®W'zQô²¨OæƒËô›Ÿ: T]òõ^•lD¢ÖGÅn& &XJ)K¿2š¸tÊô ]ryeø–B0•ù+CBQs+‰ ZGÖIb]H’-©QÉ{¹Ö·ÈÝ™¯Ó¬¾›ùñõv´å¬.o’¤s}ÿÉøï:d·]~Œ—È?|FÛ8AÓe½V·ªýmþå³"ÿYA:»7¨K¨O¾É'YáÍà±U\e¾L4õ‹€ÅªWpÚMÐ;MçâÞsDmŸ8J¢q󆡸…¶2¾•´ñÎM*<äß–]+õÛÛœÏÖ|4ÓÆ*·T†×f*%Œ«¥x±&1"ÎdÛ*L ûÎåÔ#§+8”|eïãíǪ¯n|}E6FÔ´kŒ¼£Þál:h»~ͶĜñíëœOî˜{F5ˆlܼÄsn¡>¨Éä  ]뎶ܕGRøŸG™°>3^ç–l`=k*”èÚu ¤öp†¢sïÁŸ;÷Óªëa¨Sˆ~¼€×^bi>•è˜Õ›æn:@䟷µòw€rÈ`pýê s_T)5‰Á'åœ.PɺJãº'<Ü"¬$fð"ФŒduõú¤ã˜ÑaÙ¢üÎgú°ˆ Š¥A©oöìÀEHáŽIçØw¹š˜³ý§<7K³ڦXë+,@ç×¢}Ë׺š:ôqÆ‘vïÆä°ÓíƒYÂF¹›Ï«cîÙâ33nŸÇK:HÄTV‡ØÊÍQx¼ÖÞtËŠ0×yxã‘’›k#Œ¶Ùhá½j꙾hzëìPóÙ@ù­Ü¯¥çgÛ{ú WV±ž:ªç~³Ýy‘Ôpí±QC'_Yéë•ÇëÃîÃ)Œá~üÛø=•8>GɯÆq]tçË^QFN(´5ºYˆg°PìEœ*Dg¸¾b%V×ÿæu²ÆH™Øøq³LÇâ|šª.¾j?ë—±k¸Pò—ðv|ºôñ”´M¯hUŽ‘qÔ©½ôÛ2ß13±5 czùOi¾›}}ö9ó© Ì,çìTƒ¢˜Ê莽ïÀœƒ=ë·ÒšÀ•­™%Ý‹Ì4 ôdÚlÜ7§ßOž«’ó ;”Ø14Ë÷ã ¼ÆtiWþË3ÂI×g~Eê^µ5¦&õ‘hÖÒ%\9Lë£Çù¾’X¢ÀÃÕ]šñºy ÅÃ.ŠÍ­j¾—bÁgô¹€!ï`7º¬½“R”Kzÿ¡¥À£¡{–¾7%êµÍQZ Ž—†ì?=e9ª5_cá\‚3ºòYŠÓ´Øc­˜åN,›‹_'»¥–J½´xY ƃl°xðÑ9§’ “múså1ño‡Bß{ ʿSÒÓqUsËÐÔ1A‚ºßMQî…m(¬*é‡Gµe˓DŽc•M1˜“UgaSÞcV†PhÑV¿ìÜ@õ’ßa§¾ëéœàþL¼…¾~¬f{—­¤~Äá_µNïê|ç$ë’Û%,ß-_ßÍ‘o[Ôê3y3!Kä¡2Lçu¶Œ®ÐËëJJ¢rZW¼#«F@lñ¦˜WŽŒWh‡+è-ÆXtžd{Žö?bŠ^òÎv{’ÅjÞøÆm`G:9ÞMÑæo©ºíGM?Y|(…5`ž˜)6 =¥YÄ ™ì6šç«;3Ç _äwï˜ÑhÕ”s+´§ö7AÅ |+¾_§ ù'µuÃç™bAfÑ×pÄcDÚž!L`¡.­÷AN_­ºÙžÛ*Ó\ÃEþQ‰*ìT¶äï^0•ù£ßi¤™¸»›–¡éÃ>ãÕ$…ËSgRä%Dé[ž&Õ~‚· (ê+èËôH¼¾Ú2~½A-9™v0ämà¿p k/`п’³Äðè;ý÷‚ ­J[Á¥9¾³ ~þxÛôÍו½ã^,Pš/‘mͨ瀷É#úZ¾$¹¼sƒj¡=ѰÄÀ?p6ÏöN(ÍÜ…Û2Ïa·SØ!ô5á/Þ±fJ=^&^XÂs»N$ ˆÓQÆô¼‡éy D­»ìÓÏ:•DÔtEšåAnrôý¶wéå 4Ä<8µ(¢¿„Æ0h“¥­jç³`A.r "iišù…Üœf}‰bæç/G§Yø¨¥/¥Û‰„ßð+¦GÇ_oPTÚß㋇ã^þRÑ×ÄGêͶ,“uU@oÍ0]a[Ö¬ÇqR¸Ïg¸ÿÐßzÄŽŸ…ÍÖê*î‘5 L kïwªCKˆÏ±†®$ß¹ÈYÕRZ'a´éæÞÖÒYÝV¨ÞÚd£Öùu¼ú]¡<­k cŒS¢sˆõ33lt“ŽÝÃsòg¯=–%ëaÞ&ðÞöQs” í]Ì΋vâW4*Îï]Vãïûi…SѾmŸ¼~ºãað`‡Aªz—lQ‡Ê]M‘Ü…Ö™ý9÷.£êÄ…"ùõ\ÀéZýLX;q#)•yošáZ%{åè˜Þ™$Cß·ò×öT®Ï3£BÝKç°ÝH–‹zÕ&Úä²9w‚x *°Äø'š MJWG€èë¾ìtdBÌáQéR/FÐ5lVD=q-ZQ#ñ£7€åmS½Å¸ænI\BuzÀ •7/ƒL£–Rr˜f%Îs¾Y³]$¬žT _b®Ò¸Jß¼)œmŸ]uÚ˜yMÚÅÙÅÓ±¶6]\¼lW²‘3ÃA1Ž ¨Rä{•ßÁ>n^—»( \¸ù[l/ j6»H3±ACg¦A•¢NŒx+uMJì!…8W3Kúy :Î[õ]xÔ…»*ˆºÂ/sß¾&ÃÜŠx…:'/m0ôÈõIØ·®ÝìÕóV•±ö¿³åËj¸e´y#1oOœÍ¨™#‡ý°ÈlVÓ€dÌ&Â=1ãs —sS^2´ì@ÿSŤeA]ÿ’éÇ4ÿ´ÂÊò>%«¼jA>†!¾RÄΗ%íDlϾ´Ç½õÙÈ@îtʇÐlŠ‚ÔPÖƒ¥¼Ê­ÞiÙ!”ª—:?Ä`cç OØ5÷dµø\Øl§Ýcg’ˆ±€ª>ÒÌ^ÛÄî åM¯Z å}Õ‡±ç RÊ6T·Êªî+´ô›QùmµŸJ ϸÙï>|Û©ø¼€«ÜhhÒÐ.¯ÙaÞûƒxØ&qÎêfó»[qj»p=U°(ý¨æcØ«>ÁôÂôi_ ú T´F_ ½˜%jitV§»h·;#vÃûãˆYä<÷ˆmºg¿Â‹­{ñ9Y Þz϶æ›ÊW vDeL·ýBHÞ}XkîŸÅÒ+¾þøµßÃRy+Š›b^¤¨~%Ì×ÍÊʼnʔД$‹›ó復ŒõA·GfÎehŸ†k°/3ÞR«¯òìA/iúÚÉ–Éá Ç*ÖÙ;º§²¸¼šo’ÄŠÎ1ÑGÔæ. ªë ú=¨+jÆÃi+ؤx…ï0JrbÈeO1êVc±ð²ªOÒ5ï"ËôÎ)*6§dª½Ú°§³Hä­±bÈQ|žå«d ˆ¼â}ôuJ–¢¢–výàBt7ÏõVG<Ä•˜¨…Å«õèsJ‹ˆj\U2J”4Ì[ü)i¶Õ.#ë~ï‹‚ |›£G1¨½•Ô …»˜ç®ÁÔZÿ <_qÏÔ özäÑ„÷§õˆ)¾Ý•B¢JûZg ¼5=/þú\ wZÜN’§!È%‹îÁNU‘9~{qV<Ù-f³OSq¸¯‰ÕˆáL0¬£þ5YŽŸÄv)6q÷7¹“kê|‹$—ÚɶùÉÒgÛZžÝ.©ìäÓlØNñIÏЗ3’mñGqTô&ÜHkfø¾øÄùÞzT¡(ƒXÉ%³ÚÅ>†çd[}ª¤ŒO#¥¬¿h/¦¾ µØÞi¡¬¹ ¬ÄÅatÒÀ¢½š‰z O(‚íT,Þ2 ©OÑ­R\6&o}ï©™*€ðs%¶ÀÑ[b(ø“ÈÂó˜D˜¤ù×*y*"§Y_G^ìóqJ0$U¤Ö3Ÿ­×ЏöáïÁ®OKëä#ê/ÉV_ߊ¤#fËûeK,ï.T`)£‰Îƒ„®r\)ý%ÀÙy±–Ô8ð€-ºGøHJWKuTM©ÔQ4´d ÒxÊN•ØSkÑlz¾þeCÛ¿ò¶È­+šŽ©i´~—CãÞ÷îe_T¦ôÚXQÏEZ¥[8) '²Ã¬ø©²X·”‰V›–%kù¤3$/#F ó×$äNiÖ¤7öN;6–ö9ÇI¦ØŒ3òõ§éÓÛ/7÷.iN×l¤wêŽu0Ã3‡e »Ò/¾°ìUþ¡À¥?­Ï¨T’cÙWýŽ}\ì?oEêÁ–TÕð ¾mÞ4èÌëlVÓΫaDÜëžÑîqþXÔ”Þ㵜`ØxÈQçU¥‚¯.Éi¬'Y»çÿ"¦{-e"í°µOÃR•)~%bÁtø¡°¸2Y·CþS‹©Ë~MhõU¶Ï½ÕoùÈ6ó*ÛÝ7Ê_ò¢³Ïxv_5¨þÎ+[ʰ2)zÁŽIá-­»PÇ] ÙêH.‹ 0“‘‘ wŠ$£ŒT#¿‹rDÐ×¶3°°ñÈ,0¨Ìº_¢pÚÓ=;ßÌinäþP¡¿¸Y?®:1=| ²oƒA4 rÊOìUÑ#¹u“öŒMUƒŸt-YÚ¸ŒŒx¯ütwÒðb6ʪÔÖðù³Í7Ö§-E³,™> endobj 282 0 obj << /Type /ObjStm /N 53 /First 445 /Length 1955 /Filter /FlateDecode >> stream xÚ¥YYsÛ6~ׯÀ£ÝŽ‹ f:‘¯ÄW¬Ëg'´LÛldQ©¤î¯ï‚¤h¤äJ}1ÍÅb¿ƒ ’X@A,ˆ1W…„ p Ö3QÂ9üшrÈ Q! ü •Z! š@†‚Œ€“Ždˆ©‘Ј‘€#Á¡41H@yÎ%ÒpŠ" qÀ=Àhg,\.e0blEœÒÊœQ4Aœó4JSÄ¥aXpM ’p ¼ ÑX<°Äþc(”Ø@U¬ÃáÊ`€ ¦ª¢ÀŽ(Ë)ÈABZY'µD¦@;d)ð†Ð¥0Î@ÄHÕaF_i#à §6G£ÜÄIÅÀUµæ Š iry) ¶rà¦ì8LÅH m:œH¤¤D!¥€#‡‡¡ Ôùý÷½Í"„“iv¥ãy<Ë’y'¿ÿ¾ÂÈý}÷txõëÁÅðNC|>§H ûûÉßèÏ=*Ñž%CüÑûÚÁÝtM3¤%éàƒpö9ŠŸ_2¤ÞÁÅŽíQeN²p»ÓçI ³èõ,éàÛr¸ E^Âù0ÊÐŽ&Ñ+Ìǯñt‘âÙ<~v :Ç1a`*AƒÎü'uŸ{Ëþ¨PÒ. Ûc¶K)À‡øã|Ž/ðÜÃ}<ÀC<ÊåŽñ8™$SüŸð3~Á/o³—hŠcða‚_ñܘáY8¦“è)+þ›[(<‹æqòˆ¿/’,z|˜à9Nq: Óœáìg‚øþ‰ßð?u3å&fÜûÝÂL¶bX3’Ð%ÊÈÿo¦\iæ†9vÎ8†Ô-ЛXpv4¸¿åíØ…û¢„-Tó¿Ðj¥GàA˜Žã8‹'‘cGÃŒ¿òþ©ì°˜5[¢¶ì蓺A¦Ý `¸xÈò[jûaÙØQö×`Óþm¾ìj~"»EÚ’ø8ž§™„l?‡Ë ¢oâÇì%E¶›6†?쟟Ý\ð¬žûð‚¸ðŒ¾ÃÃëjcøOwÇŸîÔ®^úðA žŽz¶9üioÐ]üÅI+¼öáePSOõ| øûËÞÉq/[ÐMãÑåÂ÷ÙÓÍá‡ýÛÞé§^7áU£ó4])~ ô»»ÁàòÐmÖ«FãÑ::wÐåæè£îUÿâ*Goi{%7hû`‹ÿåòòô4Gos¾ÑvZ¯Zób‹-ç®7ì†]µ€7ºÎ®*pé(çzsì£áýM7WÞ"\ÓµØÊulŽ=^í_Yì »Ñp\ÔL7Žð-vº³³ÞÅÝç‡çmý®åú}Þíw¾…ö£ÛþÉõu_žï|z½~·åù{Ýðrx°Ÿ¸k…o´»à˜óà·Øæ½Ï$t@×AKg›ãۼ߽O >¸h¬õzÛ ~‹Æ?Õûðꃧî:¿ÅñÂ?ûðAcÁË•;Ý»wõÐ iˆ¯í³Ì}¿¶¼czás”Â13YØCkîçöˆj?üçuÏâÇýYÌDÅ ÏÙïò+ÕÅ•åׯÛÐâý‡h Aƒ²¶)°)¯T|„AWa0öñ\¾L–¥FI \fZæ^.²I<µÓsûQ¹¬ûˆ•7EaYMŒ38e—›î¢r¬(—³zóè‡ýžÅ}^ùLf–3™jù%ú;CÞ9²˜)«™¬>Ó0Ôm¨%…BRYµ”´Ç¼útYŸkêSåðd¢É“ꪎXW‡¹uHKÊcJÖÔqm£ªY¦ò»nw ñÊ_ÚW@0×^×Ý f.]Ö¶D¡w“ñÞ0 çÙ.*NªhçóS’dÓ$‹~£»¨x÷»AAéù.*6,´3Žáþ)Ê¢)‘»ÐéÊ ‹_ó`a|¬=_ãÌãá¦~õùæcÉCHÑH§t:~I&ÿØêÆEL³pümG)Ñ»Hº²yòNÓoá[H¬PíŒýL&JQbŠhô}fq2ýXS„ñé»´êÉ Õ¨°VRU²}ŠŸóÂò ±IECW!nC²¦²F«ÀgZ‡Ïmð,G¬ ºNÒâQ…¢µ±¡b»Ý™Án˜GÜ{b3”Oð·Jkp+  #Ê­Š­EÀ*ÆÜ¤(éÜÛÎuÇ­‘Dø¤*ÀeR;¥¼·SͺÇÜ€õNh'`«²¾•¯Â4ÛgõKîl%nZƒ×û˜µ›Q¿0µQæG­‘L{Qkòb`Ÿ ¼X¨+¤ÆÃÍ]ͤiŸ/ø¼¬§DzAë+ñÙZo©­µ|%YÇÝ5ææk´Î4_£uNyk™zê³r‡MR¾ 4ÂÒ†—§¡Ù|9{ÙMÂî8L(¾©Uó2ü@U²<ý˜‚>[ž¨ÊM™•»°ý饸ë¶ü¶^–ÙjŠ+-¯ìƒ[ËÛþ¸S”ky­(geÈå¤åè{fá$yîGÀ÷ÓâòTç¼¹ó‡ÂIuxŽ.’Ç_¥ÑûðrM»9ôþ%æ¿-¾³( endstream endobj 309 0 obj << /Type /XRef /Index [0 310] /Size 310 /W [1 3 1] /Root 307 0 R /Info 308 0 R /ID [ ] /Length 785 /Filter /FlateDecode >> stream xÚ%ÔMlTUÆñ÷¹wl+m-m¡ô -PÚ)mùh¨¢´Ö¢P«¡-Zhv,Á„µ51¸q%¯1®Œ ã‚cX@B”˜ÃF&†ݹЅ:ÿ‡Í/Ï9sæž3÷>w""þ+"ŠP¼1¤¾ZîÞIp„¹~R ˜dn騃Yæ¶“ê¡Ö™ÛFz6ÔÈ-Ìõ1l„&æ:™{‚a3<Æ\s3lÐ mЊhñÁ7Áf¾Ñ M°…¹R4Bs>A´Ãh.ö`I7ô@¯bû€÷Ø >Ð t)š~àƒoã>†ïÁ†›Iý$~Q±“äß1@êd·§ŽÃnØû`/€ª¢sÉÁ~x†` ÃŒ*ª°8Ÿ§˜;þYzËçà büª¯ÌââY˜âh~ª‡¹sy„áó0£˜½ãoU,otz ¦“_yø¼Çà8×ó#;gaæÇþö7–àexNÀ¼âÄ›þô$¼ §à4,À¢bí§Zw¯üæu9©k{NqyÍsklN“–wû©ž‡X… Škw}ë|pÍ©^ñÁ Sƒâ³ÛNîi¡¸>á!}N7lj*'ïGò~$÷/ëŽz1­Ë6Å­9Û¿ÜsÚ$5û$GËfÅ'7=¤9é×€ò'¥ÉVÅëþ”†%Ï<)Cò"fâ뇞÷-{Æ&M*š}ŠK/¡*Ù¯øù]éPòÚçPï¬Â¤6/¡p9*5ÞðÆ&M›”5©mŽI=¯yÉ$äáA8Ä kÞ­œš˜Ó@ÃrŽEJŠ”)Ã,P¸œJ“”&)MÎ}Iú’§¤‘¼ùiiå/§鯰Ӣôå}§%é÷_Ψ¨Ë鬊÷*—U\_t:§â›÷Ϋøóu§•SN«*«—œ.¨\øÖé¢Ê÷¾wZSùÝ;Në*ÿ™vC•gjt\öPªL|îT¨reÕ©TåíE§ ôÂVð?È TavÁ0ŒÀ(ì†=°öÁŒÃ~U>ú·v‚O—âb(¨™ endstream endobj startxref 305752 %%EOF pcaMethods/man/0000755000175400017540000000000013556116437014455 5ustar00biocbuildbiocbuildpcaMethods/man/BPCA_dostep.Rd0000644000175400017540000000132413556116437017027 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/BPCA_dostep.R \name{BPCA_dostep} \alias{BPCA_dostep} \title{Do BPCA estimation step} \usage{ BPCA_dostep(M, y) } \arguments{ \item{M}{Data structure containing all needed information. See the source documentation of BPCA_initmodel for details} \item{y}{Numeric original data matrix} } \value{ Updated version of the data structure } \description{ The function contains the actual implementation of the BPCA component estimation. It performs one step of the BPCA EM algorithm. It is called 'maxStep' times from within the main loop in BPCAestimate. } \details{ This function is NOT intended to be run standalone. } \author{ Wolfram Stacklies } pcaMethods/man/BPCA_initmodel.Rd0000644000175400017540000000267613556116437017530 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/BPCA_initmodel.R \name{BPCA_initmodel} \alias{BPCA_initmodel} \title{Initialize BPCA model} \usage{ BPCA_initmodel(y, components) } \arguments{ \item{y}{numeric matrix containing missing values. Missing values are denoted as 'NA'} \item{components}{Number of components used for estimation} } \value{ List containing \item{rows}{Row number of input matrix} \item{cols}{Column number of input matrix} \item{comps}{Number of components to use} \item{yest}{(working variable) current estimate of complete data} \item{row_miss}{(Array) Indizes of rows containing missing values} \item{row_nomiss}{(Array) Indices of complete rows (such with no missing values)} \item{nans}{Matrix of same size as input data. TRUE if \code{input == NA}, false otherwise} \item{mean}{Column wise data mean} \item{PA}{ (d x k) Estimated principal axes (eigenvectors, loadings) The matrix ROWS are the vectors} \item{tau}{Estimated precision of the residual error} \item{scores}{ Estimated scores} Further elements are: galpha0, balpha0, alpha, gmu0, btau0, gtau0, SigW. These are working variables or constants. } \description{ Model initialization for Bayesian PCA. This function is NOT inteded to be run separately! } \details{ The function calculates the initial Eigenvectors by use of SVD from the complete rows. The data structure M is created and initial values are assigned. } \author{ Wolfram Stacklies } pcaMethods/man/DModX-pcaRes-method.Rd0000644000175400017540000000344413556116437020415 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{DModX,pcaRes-method} \alias{DModX,pcaRes-method} \alias{DModX} \title{DModX} \usage{ DModX(object, dat, newdata=FALSE, type=c("normalized","absolute"), ...) } \arguments{ \item{object}{a pcaRes object} \item{dat}{the original data, taken from \code{completeObs} if left missing.} \item{newdata}{logical indicating if this data was part of the training data or not. If it was, it is adjusted by a near one factor \eqn{v=(N/ (N-A-A0))^-1}} \item{type}{if absolute or normalized values should be given. Normalized values are adjusted to the the total RSD of the model.} \item{...}{Not used} } \value{ A vector with distances from observations to the PCA model } \description{ Distance to the model of X-space. } \details{ Measures how well described the observations are, i.e. how well they fit in the mode. High DModX indicate a poor fit. Defined as: \eqn{\frac{\sqrt{\frac{SSE_i}{K-A}}}{\sqrt{\frac{SSE}{(N-A-A_0)(K-A)}}}} For observation \eqn{i}, in a model with \eqn{A} components, \eqn{K} variables and \eqn{N} obserations. SSE is the squared sum of the residuals. \eqn{A_0} is 1 if model was centered and 0 otherwise. DModX is claimed to be approximately F-distributed and can therefore be used to check if an observation is significantly far away from the PCA model assuming normally distributed data. Pass original data as an argument if the model was calculated with \code{completeObs=FALSE}. } \examples{ data(iris) pcIr <- pca(iris[,1:4]) with(iris, plot(DModX(pcIr)~Species)) } \references{ Introduction to Multi- and Megavariate Data Analysis using Projection Methods (PCA and PLS), L. Eriksson, E. Johansson, N. Kettaneh-Wold and S. Wold, Umetrics 1999, p. 468 } \author{ Henning Redestig } pcaMethods/man/Q2.Rd0000644000175400017540000000654713556116437015242 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/xval.R \name{Q2} \alias{Q2} \title{Cross-validation for PCA} \usage{ Q2(object, originalData = completeObs(object), fold = 5, nruncv = 1, type = c("krzanowski", "impute"), verbose = interactive(), variables = 1:nVar(object), ...) } \arguments{ \item{object}{A \code{pcaRes} object (result from previous PCA analysis.)} \item{originalData}{The matrix (or ExpressionSet) that used to obtain the pcaRes object.} \item{fold}{The number of groups to divide the data in.} \item{nruncv}{The number of times to repeat the whole cross-validation} \item{type}{krzanowski or imputation type cross-validation} \item{verbose}{\code{boolean} If TRUE Q2 outputs a primitive progress bar.} \item{variables}{indices of the variables to use during cross-validation calculation. Other variables are kept as they are and do not contribute to the total sum-of-squares.} \item{...}{Further arguments passed to the \code{\link{pca}} function called within Q2.} } \value{ A matrix or vector with \eqn{Q^2} estimates. } \description{ Internal cross-validation can be used for estimating the level of structure in a data set and to optimise the choice of number of principal components. } \details{ This method calculates \eqn{Q^2} for a PCA model. This is the cross-validated version of \eqn{R^2} and can be interpreted as the ratio of variance that can be predicted independently by the PCA model. Poor (low) \eqn{Q^2} indicates that the PCA model only describes noise and that the model is unrelated to the true data structure. The definition of \eqn{Q^2} is: \deqn{Q^2=1 - \frac{\sum_{i}^{k}\sum_{j}^{n}(x - \hat{x})^2}{\sum_{i}^{k}\sum_{j}^{n}x^2}}{Q^2=1 - sum_i^k sum_j^n (x - \hat{x})^2 / \sum_i^k \sum_j^n(x^2)} for the matrix \eqn{x} which has \eqn{n} rows and \eqn{k} columns. For a given number of PC's x is estimated as \eqn{\hat{x}=TP'} (T are scores and P are loadings). Although this defines the leave-one-out cross-validation this is not what is performed if fold is less than the number of rows and/or columns. In 'impute' type CV, diagonal rows of elements in the matrix are deleted and the re-estimated. In 'krzanowski' type CV, rows are sequentially left out to build fold PCA models which give the loadings. Then, columns are sequentially left out to build fold models for scores. By combining scores and loadings from different models, we can estimate completely left out values. The two types may seem similar but can give very different results, krzanowski typically yields more stable and reliable result for estimating data structure whereas impute is better for evaluating missing value imputation performance. Note that since Krzanowski CV operates on a reduced matrix, it is not possible estimate Q2 for all components and the result vector may therefore be shorter than \code{nPcs(object)}. } \examples{ data(iris) x <- iris[,1:4] pcIr <- pca(x, nPcs=3) q2 <- Q2(pcIr, x) barplot(q2, main="Krzanowski CV", xlab="Number of PCs", ylab=expression(Q^2)) ## q2 for a single variable Q2(pcIr, x, variables=2) pcIr <- pca(x, nPcs=3, method="nipals") q2 <- Q2(pcIr, x, type="impute") barplot(q2, main="Imputation CV", xlab="Number of PCs", ylab=expression(Q^2)) } \references{ Krzanowski, WJ. Cross-validation in principal component analysis. Biometrics. 1987(43):3,575-584 } \author{ Henning Redestig, Ondrej Mikula } \keyword{multivariate} pcaMethods/man/R2VX-pcaRes-method.Rd0000644000175400017540000000144213556116437020177 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{R2VX,pcaRes-method} \alias{R2VX,pcaRes-method} \alias{R2VX} \title{R2 goodness of fit} \usage{ \S4method{R2VX}{pcaRes}(object, direction = c("variables", "observations", "complete"), data = completeObs(object), pcs = nP(object)) } \arguments{ \item{object}{a PCA model object} \item{direction}{choose between calculating R2 per variable, per observation or for the entire data with 'variables', 'observations' or 'complete'.} \item{data}{the data used to fit the model} \item{pcs}{the number of PCs to use to calculate R2} } \value{ A vector with R2 values } \description{ Flexible calculation of R2 goodness of fit. } \examples{ R2VX(pca(iris)) } \author{ Henning Redestig } pcaMethods/man/R2cum-pcaRes-method.Rd0000644000175400017540000000104713556116437020427 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{R2cum,pcaRes-method} \alias{R2cum,pcaRes-method} \alias{R2cum} \title{Cumulative R2 is the total ratio of variance that is being explained by the model} \usage{ \S4method{R2cum}{pcaRes}(object, ...) } \arguments{ \item{object}{a \code{pcaRes} model} \item{...}{Not used} } \value{ Get the cumulative R2 } \description{ Cumulative R2 is the total ratio of variance that is being explained by the model } \author{ Henning Redestig } pcaMethods/man/RnipalsPca.Rd0000644000175400017540000000423113556116437017000 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/nipalsPca.R \name{RnipalsPca} \alias{RnipalsPca} \title{NIPALS PCA implemented in R} \usage{ RnipalsPca(Matrix, nPcs = 2, varLimit = 1, maxSteps = 5000, threshold = 1e-06, verbose = interactive(), ...) } \arguments{ \item{Matrix}{Pre-processed (centered, scaled) numerical matrix samples in rows and variables as columns.} \item{nPcs}{Number of components that should be extracted.} \item{varLimit}{Optionally the ratio of variance that should be explained. \code{nPcs} is ignored if varLimit < 1} \item{maxSteps}{Defines how many iterations can be done before algorithm should abort (happens almost exclusively when there were some wrong in the input data).} \item{threshold}{The limit condition for judging if the algorithm has converged or not, specifically if a new iteration is done if \eqn{(T_{old} - T)^T(T_{old} - T) > \code{limit}}.} \item{verbose}{Show simple progress information.} \item{...}{Only used for passing through arguments.} } \value{ A \code{pcaRes} object. } \description{ PCA by non-linear iterative partial least squares } \details{ Can be used for computing PCA on a numeric matrix using either the NIPALS algorithm which is an iterative approach for estimating the principal components extracting them one at a time. NIPALS can handle a small amount of missing values. It is not recommended to use this function directely but rather to use the pca() wrapper function. There is a C++ implementation given as \code{nipalsPca} which is faster. } \examples{ data(metaboliteData) mat <- prep(t(metaboliteData)) ## c++ version is faster system.time(pc <- RnipalsPca(mat, method="rnipals", nPcs=2)) system.time(pc <- nipalsPca(mat, nPcs=2)) ## better use pca() pc <- pca(t(metaboliteData), method="rnipals", nPcs=2) \dontshow{stopifnot(sum((fitted(pc) - t(metaboliteData))^2, na.rm=TRUE) < 200)} } \references{ Wold, H. (1966) Estimation of principal components and related models by iterative least squares. In Multivariate Analysis (Ed., P.R. Krishnaiah), Academic Press, NY, 391-420. } \seealso{ \code{prcomp}, \code{princomp}, \code{pca} } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/asExprSet.Rd0000644000175400017540000000174113556116437016665 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-ExpressionSet.R \name{asExprSet} \alias{asExprSet} \title{Convert pcaRes object to an expression set} \usage{ asExprSet(object, exprSet) } \arguments{ \item{object}{\code{pcaRes} -- The object containing the completed data.} \item{exprSet}{\code{ExpressionSet} -- The object passed on to \code{pca} for missing value estimation.} } \value{ An object without missing values of class \code{ExpressionSet}. } \description{ This function can be used to conveniently replace the expression matrix in an \code{ExpressionSet} with the completed data from a \code{pcaRes} object. } \details{ This is not a standard \code{as} function as \code{pcaRes} object alone not can be converted to an \code{ExpressionSet} (the \code{pcaRes} object does not hold any \code{phenoData} for example). } \author{ Wolfram Stacklies \cr CAS-MPG Partner Institute for Computational Biology, Shanghai, China } \keyword{multivariate} pcaMethods/man/biplot-methods.Rd0000644000175400017540000000356613556116437017710 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{biplot-methods} \alias{biplot-methods} \alias{biplot.pcaRes} \alias{biplot,pcaRes-method} \title{Plot a overlaid scores and loadings plot} \usage{ \method{biplot}{pcaRes}(x, choices = 1:2, scale = 1, pc.biplot = FALSE, ...) \S4method{biplot}{pcaRes}(x, choices = 1:2, scale = 1, pc.biplot = FALSE, ...) } \arguments{ \item{x}{a pcaRes object} \item{choices}{which two pcs to plot} \item{scale}{The variables are scaled by \eqn{\lambda^{scale}}{lambda^scale} and the observations are scaled by \eqn{\lambda^{scale}}{lambda ^ (1-scale)} where \code{lambda} are the singular values as computed by \code{princomp}. Normally \eqn{0\le{}scale\le{}1}{0 <= scale <= 1}, and a warning will be issued if the specified 'scale' is outside this range.} \item{pc.biplot}{If true, use what Gabriel (1971) refers to as a "principal component biplot", with \eqn{\lambda=1}{lambda = 1} and observations scaled up by sqrt(n) and variables scaled down by sqrt(n). Then the inner products between variables approximate covariances and distances between observations approximate Mahalanobis distance.} \item{...}{optional arguments to be passed to \code{biplot.default}.} } \value{ a plot is produced on the current graphics device. } \description{ Visualize two-components simultaneously } \details{ This is a method for the generic function 'biplot'. There is considerable confusion over the precise definitions: those of the original paper, Gabriel (1971), are followed here. Gabriel and Odoroff (1990) use the same definitions, but their plots actually correspond to \code{pc.biplot = TRUE}. } \examples{ data(iris) pcIr <- pca(iris[,1:4]) biplot(pcIr) } \seealso{ \code{prcomp}, \code{pca}, \code{princomp} } \author{ Kevin Wright, Adapted from \code{biplot.prcomp} } \keyword{multivariate} pcaMethods/man/bpca.Rd0000644000175400017540000001332013556116437015650 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/bpca.R \name{bpca} \alias{bpca} \title{Bayesian PCA missing value estimation} \usage{ bpca(Matrix, nPcs = 2, maxSteps = 100, verbose = interactive(), threshold = 1e-04, ...) } \arguments{ \item{Matrix}{\code{matrix} -- Pre-processed matrix (centered, scaled) with variables in columns and observations in rows. The data may contain missing values, denoted as \code{NA}.} \item{nPcs}{\code{numeric} -- Number of components used for re-estimation. Choosing few components may decrease the estimation precision.} \item{maxSteps}{\code{numeric} -- Maximum number of estimation steps.} \item{verbose}{\code{boolean} -- BPCA prints the number of steps and the increase in precision if set to TRUE. Default is interactive().} \item{threshold}{convergence threshold} \item{...}{Reserved for future use. Currently no further parameters are used} } \value{ Standard PCA result object used by all PCA-based methods of this package. Contains scores, loadings, data mean and more. See \code{\link{pcaRes}} for details. } \description{ Implements a Bayesian PCA missing value estimator. The script is a port of the Matlab version provided by Shigeyuki OBA. See also \url{http://ishiilab.jp/member/oba/tools/BPCAFill.html}. BPCA combines an EM approach for PCA with a Bayesian model. In standard PCA data far from the training set but close to the principal subspace may have the same reconstruction error. BPCA defines a likelihood function such that the likelihood for data far from the training set is much lower, even if they are close to the principal subspace. } \details{ Scores and loadings obtained with Bayesian PCA slightly differ from those obtained with conventional PCA. This is because BPCA was developed especially for missing value estimation. The algorithm does not force orthogonality between factor loadings, as a result factor loadings are not necessarily orthogonal. However, the BPCA authors found that including an orthogonality criterion made the predictions worse. The authors also state that the difference between real and predicted Eigenvalues becomes larger when the number of observation is smaller, because it reflects the lack of information to accurately determine true factor loadings from the limited and noisy data. As a result, weights of factors to predict missing values are not the same as with conventional PCA, but the missing value estimation is improved. BPCA works iteratively, the complexity is growing with \eqn{O(n^3)}{O(n^3)} because several matrix inversions are required. The size of the matrices to invert depends on the number of components used for re-estimation. Finding the optimal number of components for estimation is not a trivial task; the best choice depends on the internal structure of the data. A method called \code{kEstimate} is provided to estimate the optimal number of components via cross validation. In general few components are sufficient for reasonable estimation accuracy. See also the package documentation for further discussion about on what data PCA-based missing value estimation makes sense. It is not recommended to use this function directely but rather to use the pca() wrapper function. There is a difference with respect the interpretation of rows (observations) and columns (variables) compared to matlab implementation. For estimation of missing values for microarray data, the suggestion in the original bpca is to intepret genes as observations and the samples as variables. In pcaMethods however, genes are interpreted as variables and samples as observations which arguably also is the more natural interpretation. For bpca behavior like in the matlab implementation, simply transpose your input matrix. Details about the probabilistic model underlying BPCA are found in Oba et. al 2003. The algorithm uses an expectation maximation approach together with a Bayesian model to approximate the principal axes (eigenvectors of the covariance matrix in PCA). The estimation is done iteratively, the algorithm terminates if either the maximum number of iterations was reached or if the estimated increase in precision falls below \eqn{1e^{-4}}{1e^-4}. \bold{Complexity:} The relatively high complexity of the method is a result of several matrix inversions required in each step. Considering the case that the maximum number of iteration steps is needed, the approximate complexity is given by the term \deqn{maxSteps \cdot row_{miss} \cdot O(n^3)}{maxSteps * row_miss * O(n^3)} Where \eqn{row_{miss}}{row_miss} is the number of rows containing missing values and \eqn{O(n^3)}{O(n^3)} is the complexity for inverting a matrix of size \eqn{components}{components}. Components is the number of components used for re-estimation. } \note{ Requires \code{MASS}. } \examples{ ## Load a sample metabolite dataset with 5\\\% missig values (metaboliteData)e data(metaboliteData) ## Perform Bayesian PCA with 2 components pc <- pca(t(metaboliteData), method="bpca", nPcs=2) ## Get the estimated principal axes (loadings) loadings <- loadings(pc) ## Get the estimated scores scores <- scores(pc) ## Get the estimated complete observations cObs <- completeObs(pc) ## Now make a scores and loadings plot slplot(pc) \dontshow{stopifnot(sum((fitted(pc) - t(metaboliteData))^2, na.rm=TRUE) < 200)} } \references{ Shigeyuki Oba, Masa-aki Sato, Ichiro Takemasa, Morito Monden, Ken-ichi Matsubara and Shin Ishii. A Bayesian missing value estimation method for gene expression profile data. \emph{Bioinformatics, 19(16):2088-2096, Nov 2003}. } \seealso{ \code{\link{ppca}}, \code{\link{svdImpute}}, \code{\link{prcomp}}, \code{\link{nipalsPca}}, \code{\link{pca}}, \code{\link{pcaRes}}. \code{\link{kEstimate}}. } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/center-pcaRes-method.Rd0000644000175400017540000000070213556116437020714 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{center,pcaRes-method} \alias{center,pcaRes-method} \alias{center} \title{Get the centers of the original variables} \usage{ center(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ Vector with the centers } \description{ Get the centers of the original variables } \author{ Henning Redestig } pcaMethods/man/centered-pcaRes-method.Rd0000644000175400017540000000070513556116437021230 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{centered,pcaRes-method} \alias{centered,pcaRes-method} \alias{centered} \title{Check centering was part of the model} \usage{ centered(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ TRUE if model was centered } \description{ Check centering was part of the model } \author{ Henning Redestig } pcaMethods/man/checkData.Rd0000644000175400017540000000242513556116437016616 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/checkData.R \name{checkData} \alias{checkData} \title{Do some basic checks on a given data matrix} \usage{ checkData(data, verbose = FALSE) } \arguments{ \item{data}{\code{matrix} -- Data to check.} \item{verbose}{\code{boolean} -- If TRUE, the function prints messages whenever an error in the data set is found.} } \value{ \item{isValid}{\code{boolean} -- TRUE if no errors were found, FALSE otherwise. isValid contains a set of attributes, these are: \itemize{ \item isNumeric - TRUE if data is numeric, false otherwise \item isInfinite - TRUE if data contains 'Inf' values, false otherwise \item isNaN - TRUE if data contains 'NaN' values, false otherwise \item isMatrix - TRUE if the data is in matrix format, FALSE otherwise \item naRows - TRUE if data contains rows in which all elements are 'NA', FALSE otherwise \item naCols - TRUE if data contains columns in which all elements are 'NA', FALSE otherwise }} } \description{ Check a given data matrix for consistency with the format required for further analysis. The data must be a numeric matrix and not contain: \itemize{ \item Inf values \item NaN values \item Rows or columns that consist of NA only } } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/completeObs-nniRes-method.Rd0000644000175400017540000000111513556116437021730 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{completeObs,nniRes-method} \alias{completeObs,nniRes-method} \alias{completeObs} \alias{completeObs,pcaRes-method} \title{Get the original data with missing values replaced with predicted values.} \usage{ completeObs(object, ...) } \arguments{ \item{object}{object to fetch complete data from} \item{...}{Not used} } \value{ Completed data (matrix) } \description{ Get the original data with missing values replaced with predicted values. } \author{ Henning Redestig } pcaMethods/man/cvseg.Rd0000644000175400017540000000170013556116437016051 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/xval.R \name{cvseg} \alias{cvseg} \title{Get CV segments} \usage{ cvseg(x, fold = 7, seed = NULL) } \arguments{ \item{x}{a factor, character or numeric vector that describes class membership of a set of items, or, a numeric vector indicating unique indices of items, or, a numeric of length 1 that describes the number of items to segment (without any classes)} \item{fold}{the desired number of segments} \item{seed}{randomization seed for reproducibility} } \value{ a list where each element is a set of indices that defines the CV segment. } \description{ Get cross-validation segments that have (as far as possible) the same ratio of all classes (if classes are present) } \examples{ seg <- cvseg(iris$Species, 10) sapply(seg, function(s) table(iris$Species[s])) cvseg(20, 10) } \seealso{ the \code{cvsegments} function in the \code{pls} package } \author{ Henning Redestig } pcaMethods/man/cvstat-pcaRes-method.Rd0000644000175400017540000000071713556116437020746 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{cvstat,pcaRes-method} \alias{cvstat,pcaRes-method} \alias{cvstat} \title{Get cross-validation statistics (e.g. \eqn{Q^2}).} \usage{ cvstat(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{not used} } \value{ vector CV statistics } \description{ Get cross-validation statistics (e.g. \eqn{Q^2}). } \author{ Henning Redestig } pcaMethods/man/deletediagonals.Rd0000644000175400017540000000134413556116437020072 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/xval.R \name{deletediagonals} \alias{deletediagonals} \title{Delete diagonals} \usage{ deletediagonals(x, diagonals = 1) } \arguments{ \item{x}{The matrix} \item{diagonals}{The diagonal to be replaced, i.e. the first, second and so on when looking at the fat version of the matrix (transposed or not) counting from the bottom. Can be a vector to delete more than one diagonal.} } \value{ The original matrix with some values missing } \description{ Replace a diagonal of elements of a matrix with NA } \details{ Used for creating artifical missing values in matrices without causing any full row or column to be completely missing } \author{ Henning Redestig } pcaMethods/man/derrorHierarchic.Rd0000644000175400017540000000061413556116437020224 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/derrorHierarchic.R \name{derrorHierarchic} \alias{derrorHierarchic} \title{Later} \usage{ derrorHierarchic(nlnet, trainIn, trainOut) } \arguments{ \item{nlnet}{the nlnet} \item{trainIn}{training data} \item{trainOut}{fitted data} } \value{ derror } \description{ Later } \author{ Henning Redestig, Matthias Scholz } pcaMethods/man/dim.pcaRes.Rd0000644000175400017540000000054513556116437016735 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \name{dim.pcaRes} \alias{dim.pcaRes} \title{Dimensions of a PCA model} \usage{ \method{dim}{pcaRes}(x) } \arguments{ \item{x}{a pcaRes object} } \value{ Get the dimensions of this PCA model } \description{ Dimensions of a PCA model } \author{ Henning Redestig } pcaMethods/man/errorHierarchic.Rd0000644000175400017540000000060713556116437020062 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/errorHierarchic.R \name{errorHierarchic} \alias{errorHierarchic} \title{Later} \usage{ errorHierarchic(nlnet, trainIn, trainOut) } \arguments{ \item{nlnet}{The nlnet} \item{trainIn}{training data} \item{trainOut}{fitted data} } \value{ error } \description{ Later } \author{ Henning Redestig, Matthias Scholz } pcaMethods/man/fitted-methods.Rd0000644000175400017540000000324213556116437017665 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{fitted-methods} \alias{fitted-methods} \alias{fitted.pcaRes} \alias{fitted,pcaRes-method} \title{Extract fitted values from PCA.} \usage{ \method{fitted}{pcaRes}(object, data = NULL, nPcs = nP(object), pre = TRUE, post = TRUE, ...) \S4method{fitted}{pcaRes}(object, data = NULL, nPcs = nP(object), pre = TRUE, post = TRUE, ...) } \arguments{ \item{object}{the \code{pcaRes} object of interest.} \item{data}{For standard PCA methods this can safely be left null to get scores x loadings but if set, then the scores are obtained by projecting provided data onto the loadings. If data contains missing values the result will be all NA. Non-linear PCA is an exception, here if data is NULL then data is set to the completeObs and propaged through the network.} \item{nPcs}{The number of PC's to consider} \item{pre}{pre-process \code{data} based on the pre-processing chosen for the PCA model} \item{post}{unpre-process the final data (add the center back etc to get the final estimate)} \item{...}{Not used} } \value{ A matrix representing the fitted data } \description{ Fitted values of a PCA model } \details{ This function extracts the fitted values from a pcaResobject. For PCA methods like SVD, Nipals, PPCA etc this is basically just the scores multipled by the loadings and adjusted for pre-processing. for non-linear PCA the original data is propagated through the network to obtain the approximated data. } \examples{ pc <- pca(iris[,1:4], nPcs=4, center=TRUE, scale="uv") sum( (fitted(pc) - iris[,1:4])^2 ) } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/forkNlpcaNet.Rd0000644000175400017540000000054413556116437017335 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/forkNlpcaNet.R \name{forkNlpcaNet} \alias{forkNlpcaNet} \title{Complete copy of nlpca net object} \usage{ forkNlpcaNet(nlnet) } \arguments{ \item{nlnet}{a nlnet} } \value{ A copy of the input nlnet } \description{ Complete copy of nlpca net object } \author{ Henning Redestig } pcaMethods/man/getHierarchicIdx.Rd0000644000175400017540000000052513556116437020154 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/nlpca.R \name{getHierarchicIdx} \alias{getHierarchicIdx} \title{Index in hiearchy} \usage{ getHierarchicIdx(hierarchicNum) } \arguments{ \item{hierarchicNum}{A number} } \value{ ... } \description{ Index in hiearchy } \author{ Henning Redestig, Matthias Scholz } pcaMethods/man/helix.Rd0000644000175400017540000000110213556116437016047 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pcaMethods-package.R \docType{data} \name{helix} \alias{helix} \title{A helix structured toy data set} \usage{ data(helix) } \description{ Simulated data set looking like a helix } \details{ A matrix containing 1000 observations (rows) and three variables (columns). } \references{ Matthias Scholz, Fatma Kaplan, Charles L. Guy, Joachim Kopka and Joachim Selbig. - Non-linear PCA: a missing data approach. \emph{Bioinformatics 2005 21(20):3887-3895} } \author{ Henning Redestig } \keyword{datasets} pcaMethods/man/kEstimate.Rd0000644000175400017540000001435513556116437016702 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/kEstimate.R \name{kEstimate} \alias{kEstimate} \title{Estimate best number of Components for missing value estimation} \usage{ kEstimate(Matrix, method = "ppca", evalPcs = 1:3, segs = 3, nruncv = 5, em = "q2", allVariables = FALSE, verbose = interactive(), ...) } \arguments{ \item{Matrix}{\code{matrix} -- numeric matrix containing observations in rows and variables in columns} \item{method}{\code{character} -- of the methods found with pcaMethods() The option llsImputeAll calls llsImpute with the allVariables = TRUE parameter.} \item{evalPcs}{\code{numeric} -- The principal components to use for cross validation or the number of neighbour variables if used with llsImpute. Should be an array containing integer values, eg. \code{evalPcs = 1:10} or \code{evalPcs = c(2,5,8)}. The NRMSEP or Q2 is calculated for each component.} \item{segs}{\code{numeric} -- number of segments for cross validation} \item{nruncv}{\code{numeric} -- Times the whole cross validation is repeated} \item{em}{\code{character} -- The error measure. This can be nrmsep or q2} \item{allVariables}{\code{boolean} -- If TRUE, the NRMSEP is calculated for all variables, If FALSE, only the incomplete ones are included. You maybe want to do this to compare several methods on a complete data set.} \item{verbose}{\code{boolean} -- If TRUE, some output like the variable indexes are printed to the console each iteration.} \item{...}{Further arguments to \code{pca} or \code{nni}} } \value{ A list with: \item{bestNPcs}{number of PCs or k for which the minimal average NRMSEP or the maximal Q2 was obtained.} \item{eError}{an array of of size length(evalPcs). Contains the average error of the cross validation runs for each number of components.} \item{variableWiseError}{Matrix of size \code{incomplete_variables} x length(evalPcs). Contains the NRMSEP or Q2 distance for each variable and each number of PCs. This allows to easily see for wich variables imputation makes sense and for which one it should not be done or mean imputation should be used.} \item{evalPcs}{The evaluated numbers of components or number of neighbours (the same as the evalPcs input parameter).} \item{variableIx}{Index of the incomplete variables. This can be used to map the variable wise error to the original data.} } \description{ Perform cross validation to estimate the optimal number of components for missing value estimation. Cross validation is done for the complete subset of a variable. } \details{ The assumption hereby is that variables that are highly correlated in a distinct region (here the non-missing observations) are also correlated in another (here the missing observations). This also implies that the complete subset must be large enough to be representative. For each incomplete variable, the available values are divided into a user defined number of cv-segments. The segments have equal size, but are chosen from a random equal distribution. The non-missing values of the variable are covered completely. PPCA, BPCA, SVDimpute, Nipals PCA, llsImpute an NLPCA may be used for imputation. The whole cross validation is repeated several times so, depending on the parameters, the calculations can take very long time. As error measure the NRMSEP (see Feten et. al, 2005) or the Q2 distance is used. The NRMSEP basically normalises the RMSD between original data and estimate by the variable-wise variance. The reason for this is that a higher variance will generally lead to a higher estimation error. If the number of samples is small, the variable - wise variance may become an unstable criterion and the Q2 distance should be used instead. Also if variance normalisation was applied previously. The method proceeds variable - wise, the NRMSEP / Q2 distance is calculated for each incomplete variable and averaged afterwards. This allows to easily see for wich set of variables missing value imputation makes senes and for wich set no imputation or something like mean-imputation should be used. Use \code{kEstimateFast} or \code{Q2} if you are not interested in variable wise CV performance estimates. Run time may be very high on large data sets. Especially when used with complex methods like BPCA or Nipals PCA. For PPCA, BPCA, Nipals PCA and NLPCA the estimation method is called \eqn{(v_{miss} \cdot segs \cdot nruncv \cdot)}{(v\_miss * segs * nruncv)} times as the error for all numbers of principal components can be calculated at once. For LLSimpute and SVDimpute this is not possible, and the method is called \eqn{(v_{miss} \cdot segs \cdot nruncv \cdot length(evalPcs))}{(v\_miss * segs * nruncv * length(evalPcs))} times. This should still be fast for LLSimpute because the method allows to choose to only do the estimation for one particular variable. This saves a lot of iterations. Here, \eqn{v_{miss}}{v\_miss} is the number of variables showing missing values. As cross validation is done variable-wise, in this function Q2 is defined on single variables, not on the entire data set. This is Q2 is calculated as as \eqn{\frac{\sum(x - xe)^2}{\sum(x^2)}}{sum(x - xe)^2 \ sum(x^2)}, where x is the currently used variable and xe it's estimate. The values are then averaged over all variables. The NRMSEP is already defined variable-wise. For a single variable it is then \eqn{\sqrt(\frac{\sum(x - xe)^2}{(n \cdot var(x))})}{sqrt(sum(x - xe)^2 \ (n * var(x)))}, where x is the variable and xe it's estimate, n is the length of x. The variable wise estimation errors are returned in parameter variableWiseError. } \examples{ ## Load a sample metabolite dataset with 5\\\% missing values (metaboliteData) data(metaboliteData) # Do cross validation with ppca for component 2:4 esti <- kEstimate(metaboliteData, method = "ppca", evalPcs = 2:4, nruncv=1, em="nrmsep") # Plot the average NRMSEP barplot(drop(esti$eError), xlab = "Components",ylab = "NRMSEP (1 iterations)") # The best result was obtained for this number of PCs: print(esti$bestNPcs) # Now have a look at the variable wise estimation error barplot(drop(esti$variableWiseError[, which(esti$evalPcs == esti$bestNPcs)]), xlab = "Incomplete variable Index", ylab = "NRMSEP") } \seealso{ \code{\link{kEstimateFast}, \link{Q2}, \link{pca}, \link{nni}}. } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/kEstimateFast.Rd0000644000175400017540000000553513556116437017520 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/kEstimateFast.R \name{kEstimateFast} \alias{kEstimateFast} \title{Estimate best number of Components for missing value estimation} \usage{ kEstimateFast(Matrix, method = "ppca", evalPcs = 1:3, em = "nrmsep", allVariables = FALSE, verbose = interactive(), ...) } \arguments{ \item{Matrix}{\code{matrix} -- numeric matrix containing observations in rows and variables in columns} \item{method}{\code{character} -- a valid pca method (see \code{\link{pca}}).} \item{evalPcs}{\code{numeric} -- The principal components to use for cross validation or cluster sizes if used with llsImpute. Should be an array containing integer values, eg. evalPcs = 1:10 or evalPcs = C(2,5,8).The NRMSEP is calculated for each component.} \item{em}{\code{character} -- The error measure. This can be nrmsep or q2} \item{allVariables}{\code{boolean} -- If TRUE, the NRMSEP is calculated for all variables, If FALSE, only the incomplete ones are included. You maybe want to do this to compare several methods on a complete data set.} \item{verbose}{\code{boolean} -- If TRUE, the NRMSEP and the variance are printed to the console each iteration.} \item{...}{Further arguments to \code{pca}} } \value{ \item{list}{Returns a list with the elements: \itemize{ \item minNPcs - number of PCs for which the minimal average NRMSEP was obtained \item eError - an array of of size length(evalPcs). Contains the estimation error for each number of components. \item evalPcs - The evaluated numbers of components or cluster sizes (the same as the evalPcs input parameter). }} } \description{ This is a simple estimator for the optimal number of componets when applying PCA or LLSimpute for missing value estimation. No cross validation is performed, instead the estimation quality is defined as Matrix[!missing] - Estimate[!missing]. This will give a relatively rough estimate, but the number of iterations equals the length of the parameter evalPcs.\cr Does not work with LLSimpute!! As error measure the NRMSEP (see Feten et. al, 2005) or the Q2 distance is used. The NRMSEP basically normalises the RMSD between original data and estimate by the variable-wise variance. The reason for this is that a higher variance will generally lead to a higher estimation error. If the number of samples is small, the gene - wise variance may become an unstable criterion and the Q2 distance should be used instead. Also if variance normalisation was applied previously. } \examples{ data(metaboliteData) # Estimate best number of PCs with ppca for component 2:4 esti <- kEstimateFast(t(metaboliteData), method = "ppca", evalPcs = 2:4, em="nrmsep") barplot(drop(esti$eError), xlab = "Components",ylab = "NRMSEP (1 iterations)") # The best k value is: print(esti$minNPcs) } \seealso{ \code{\link{kEstimate}}. } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/leverage-pcaRes-method.Rd0000644000175400017540000000213413556116437021227 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{leverage,pcaRes-method} \alias{leverage,pcaRes-method} \alias{leverage} \title{Extract leverages of a PCA model} \usage{ \S4method{leverage}{pcaRes}(object) } \arguments{ \item{object}{a \code{pcaRes} object} } \value{ The observation leverages as a numeric vector } \description{ The leverages of PCA model indicate how much influence each observation has on the PCA model. Observations with high leverage has caused the principal components to rotate towards them. It can be used to extract both "unimportant" observations as well as picking potential outliers. } \details{ Defined as \eqn{Tr(T(T'T)^{-1}T')}{Tr(T(T'T)^(-1)T')} } \examples{ data(iris) pcIr <- pca(iris[,1:4]) ## versicolor has the lowest leverage with(iris, plot(leverage(pcIr)~Species)) } \references{ Introduction to Multi- and Megavariate Data Analysis using Projection Methods (PCA and PLS), L. Eriksson, E. Johansson, N. Kettaneh-Wold and S. Wold, Umetrics 1999, p. 466 } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/lineSearch.Rd0000644000175400017540000000104013556116437017014 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/lineSearch.R \name{lineSearch} \alias{lineSearch} \title{Line search for conjugate gradient} \usage{ lineSearch(nlnet, dw, e0, ttGuess, trainIn, trainOut, verbose) } \arguments{ \item{nlnet}{The nlnet} \item{dw}{..} \item{e0}{..} \item{ttGuess}{..} \item{trainIn}{Training data} \item{trainOut}{Fitted data} \item{verbose}{logical, print messages} } \value{ ... } \description{ Line search for conjugate gradient } \author{ Henning Redestig, Matthias Scholz } pcaMethods/man/linr.Rd0000644000175400017540000000043313556116437015710 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/lineSearch.R \name{linr} \alias{linr} \title{Linear kernel} \usage{ linr(x) } \arguments{ \item{x}{datum} } \value{ Input value } \description{ Linear kernel } \author{ Henning Redestig, Matthias Scholz } pcaMethods/man/listPcaMethods.Rd0000644000175400017540000000105713556116437017672 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pca.R \name{listPcaMethods} \alias{listPcaMethods} \title{List PCA methods} \usage{ listPcaMethods(which = c("all", "linear", "nonlinear")) } \arguments{ \item{which}{the type of methods to get. E.g. only get the PCA methods based on the classical model where the fitted data is a direct multiplication of scores and loadings.} } \value{ A character vector with the current methods for doing PCA } \description{ Vector with current valid PCA methods } \author{ Henning Redestig } pcaMethods/man/llsImpute.Rd0000644000175400017540000000776613556116437016742 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/llsImpute.R \name{llsImpute} \alias{llsImpute} \title{LLSimpute algorithm} \usage{ llsImpute(Matrix, k = 10, center = FALSE, completeObs = TRUE, correlation = "pearson", allVariables = FALSE, maxSteps = 100, xval = NULL, verbose = FALSE, ...) } \arguments{ \item{Matrix}{\code{matrix} -- Data containing the variables (genes) in columns and observations (samples) in rows. The data may contain missing values, denoted as \code{NA}.} \item{k}{\code{numeric} -- Cluster size, this is the number of similar genes used for regression.} \item{center}{\code{boolean} -- Mean center the data if TRUE} \item{completeObs}{\code{boolean} -- Return the estimated complete observations if TRUE. This is the input data with NA values replaced by the estimated values.} \item{correlation}{\code{character} -- How to calculate the distance between genes. One out of pearson | kendall | spearman , see also help("cor").} \item{allVariables}{\code{boolean} -- Use only complete genes to do the regression if TRUE, all genes if FALSE.} \item{maxSteps}{\code{numeric} -- Maximum number of iteration steps if allGenes = TRUE.} \item{xval}{\code{numeric} Use LLSimpute for cross validation. xval is the index of the gene to estimate, all other incomplete genes will be ignored if this parameter is set. We do not consider them in the cross-validation.} \item{verbose}{\code{boolean} -- Print step number and relative change if TRUE and allVariables = TRUE} \item{...}{Reserved for parameters used in future version of the algorithm} } \value{ \item{nniRes}{Standard nni (nearest neighbour imputation) result object of this package. See \code{\link{nniRes}} for details.} } \description{ Missing value estimation using local least squares (LLS). First, k variables (for Microarrya data usually the genes) are selected by pearson, spearman or kendall correlation coefficients. Then missing values are imputed by a linear combination of the k selected variables. The optimal combination is found by LLS regression. The method was first described by Kim et al, Bioinformatics, 21(2),2005. } \details{ Missing values are denoted as \code{NA}\cr It is not recommended to use this function directely but rather to use the nni() wrapper function. The methods provides two ways for missing value estimation, selected by the \code{allVariables} option. The first one is to use only complete variables for the regression. This is preferable when the number of incomplete variables is relatively small. The second way is to consider all variables as candidates for the regression. Hereby missing values are initially replaced by the columns wise mean. The method then iterates, using the current estimate as input for the regression until the change between new and old estimate falls below a threshold (0.001). } \note{ Each step the generalized inverse of a \code{miss} x k matrix is calculated. Where \code{miss} is the number of missing values in variable j and \code{k} the number of neighbours. This may be slow for large values of k and / or many missing values. See also help("ginv"). } \examples{ ## Load a sample metabolite dataset (metaboliteData) with already 5\\\% of ## data missing data(metaboliteData) ## Perform llsImpute using k = 10 ## Set allVariables TRUE because there are very few complete variables result <- llsImpute(metaboliteData, k = 10, correlation="pearson", allVariables=TRUE) ## Get the estimated complete observations cObs <- completeObs(result) } \references{ Kim, H. and Golub, G.H. and Park, H. - Missing value estimation for DNA microarray gene expression data: local least squares imputation. \emph{Bioinformatics, 2005; 21(2):187-198.} Troyanskaya O. and Cantor M. and Sherlock G. and Brown P. and Hastie T. and Tibshirani R. and Botstein D. and Altman RB. - Missing value estimation methods for DNA microarrays. \emph{Bioinformatics. 2001 Jun;17(6):520-525.} } \seealso{ \code{\link{pca}, \link{nniRes}, \link{nni}}. } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/loadings-ANY-method.Rd0000644000175400017540000000077613556116437020461 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{loadings,ANY-method} \alias{loadings,ANY-method} \alias{loadings} \title{Crude way to unmask the function with the same name from \code{stats}} \usage{ \S4method{loadings}{ANY}(object, ...) } \arguments{ \item{object}{any object} \item{...}{not used} } \value{ The loadings } \description{ Crude way to unmask the function with the same name from \code{stats} } \author{ Henning Redestig } pcaMethods/man/loadings-pcaRes-method.Rd0000644000175400017540000000075113556116437021240 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{loadings,pcaRes-method} \alias{loadings,pcaRes-method} \title{Get loadings from a pcaRes object} \usage{ \S4method{loadings}{pcaRes}(object, ...) } \arguments{ \item{object}{a pcaRes object} \item{...}{not used} } \value{ The loadings as a matrix } \description{ Get loadings from a pcaRes object } \seealso{ \code{\link{loadings.pcaRes}} } \author{ Henning Redestig } pcaMethods/man/loadings.pcaRes.Rd0000644000175400017540000000063513556116437017764 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \name{loadings.pcaRes} \alias{loadings.pcaRes} \title{Get loadings from a pcaRes object} \usage{ \method{loadings}{pcaRes}(object, ...) } \arguments{ \item{object}{a pcaRes object} \item{...}{not used} } \value{ The loadings as a matrix } \description{ Get loadings from a pcaRes object } \author{ Henning Redestig } pcaMethods/man/metaboliteData.Rd0000644000175400017540000000174513556116437017672 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pcaMethods-package.R \docType{data} \name{metaboliteData} \alias{metaboliteData} \title{A incomplete metabolite data set from an Arabidopsis coldstress experiment} \description{ A incomplete subset from a larger metabolite data set. This is the original, complete data set and can be used to compare estimation results created with the also provided incomplete data (called metaboliteData). } \details{ A matrix containing 154 observations (rows) and 52 metabolites (columns). The data contains 5\% of artificially created uniformly distributed misssing values. The data was created during an in house Arabidopsis coldstress experiment. } \references{ Matthias Scholz, Fatma Kaplan, Charles L. Guy, Joachim Kopka and Joachim Selbig. - Non-linear PCA: a missing data approach.\emph{Bioinformatics 2005 21(20):3887-3895} } \seealso{ \code{\link{metaboliteDataComplete}} } \author{ Wolfram Stacklies } \keyword{datasets} pcaMethods/man/metaboliteDataComplete.Rd0000644000175400017540000000162513556116437021360 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pcaMethods-package.R \docType{data} \name{metaboliteDataComplete} \alias{metaboliteDataComplete} \title{A complete metabolite data set from an Arabidopsis coldstress experiment} \description{ A complete subset from a larger metabolite data set. This is the original, complete data set and can be used to compare estimation results created with the also provided incomplete data (called metaboliteData). The data was created during an in house Arabidopsis coldstress experiment. } \details{ A matrix containing 154 observations (rows) and 52 metabolites (columns). } \references{ Matthias Scholz, Fatma Kaplan, Charles L. Guy, Joachim Kopka and Joachim Selbig. - Non-linear PCA: a missing data approach.\emph{Bioinformatics 2005 21(20):3887-3895} } \seealso{ \code{\link{metaboliteData}} } \author{ Wolfram Stacklies } \keyword{datasets} pcaMethods/man/method-pcaRes-method.Rd0000644000175400017540000000063213556116437020716 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{method,pcaRes-method} \alias{method,pcaRes-method} \alias{method} \title{Get the used PCA method} \usage{ method(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ The used pca method } \description{ Get the used PCA method } \author{ Henning Redestig } pcaMethods/man/nObs-pcaRes-method.Rd0000644000175400017540000000073513556116437020343 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{nObs,pcaRes-method} \alias{nObs,pcaRes-method} \alias{nObs} \title{Get the number of observations used to build the PCA model.} \usage{ nObs(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ Number of observations } \description{ Get the number of observations used to build the PCA model. } \author{ Henning Redestig } pcaMethods/man/nP-pcaRes-method.Rd0000644000175400017540000000057013556116437020014 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{nP,pcaRes-method} \alias{nP,pcaRes-method} \alias{nP} \title{Get number of PCs} \usage{ nP(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{not used} } \value{ Number of PCs } \description{ Get number of PCs } \author{ Henning Redestig } pcaMethods/man/nPcs-pcaRes-method.Rd0000644000175400017540000000074313556116437020344 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{nPcs,pcaRes-method} \alias{nPcs,pcaRes-method} \alias{nPcs} \title{Get number of PCs.} \usage{ nPcs(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{not used} } \value{ Number of PCs } \description{ Get number of PCs. } \note{ Try to use \code{link{nP}} instead since \code{nPcs} tend to clash with argument names. } \author{ Henning Redestig } pcaMethods/man/nVar-pcaRes-method.Rd0000644000175400017540000000072413556116437020346 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{nVar,pcaRes-method} \alias{nVar,pcaRes-method} \alias{nVar} \title{Get the number of variables used to build the PCA model.} \usage{ nVar(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ Number of variables } \description{ Get the number of variables used to build the PCA model. } \author{ Henning Redestig } pcaMethods/man/nipalsPca.Rd0000644000175400017540000000361413556116437016662 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/nipalsPca.R \name{nipalsPca} \alias{nipalsPca} \title{NIPALS PCA} \usage{ nipalsPca(Matrix, nPcs = 2, varLimit = 1, maxSteps = 5000, threshold = 1e-06, ...) } \arguments{ \item{Matrix}{Pre-processed (centered, scaled) numerical matrix samples in rows and variables as columns.} \item{nPcs}{Number of components that should be extracted.} \item{varLimit}{Optionally the ratio of variance that should be explained. \code{nPcs} is ignored if varLimit < 1} \item{maxSteps}{Defines how many iterations can be done before algorithm should abort (happens almost exclusively when there were some wrong in the input data).} \item{threshold}{The limit condition for judging if the algorithm has converged or not, specifically if a new iteration is done if \eqn{(T_{old} - T)^T(T_{old} - T) > \code{limit}}.} \item{...}{Only used for passing through arguments.} } \value{ A \code{pcaRes} object. } \description{ PCA by non-linear iterative partial least squares } \details{ Can be used for computing PCA on a numeric matrix using either the NIPALS algorithm which is an iterative approach for estimating the principal components extracting them one at a time. NIPALS can handle a small amount of missing values. It is not recommended to use this function directely but rather to use the pca() wrapper function. } \examples{ data(metaboliteData) mat <- prep(t(metaboliteData)) pc <- nipalsPca(mat, nPcs=2) ## better use pca() pc <- pca(t(metaboliteData), method="nipals", nPcs=2) \dontshow{stopifnot(sum((fitted(pc) - t(metaboliteData))^2, na.rm=TRUE) < 200)} } \references{ Wold, H. (1966) Estimation of principal components and related models by iterative least squares. In Multivariate Analysis (Ed., P.R. Krishnaiah), Academic Press, NY, 391-420. } \seealso{ \code{prcomp}, \code{princomp}, \code{pca} } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/nlpca.Rd0000644000175400017540000000674113556116437016051 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/nlpca.R \name{nlpca} \alias{nlpca} \title{Non-linear PCA} \usage{ nlpca(Matrix, nPcs = 2, maxSteps = 2 * prod(dim(Matrix)), unitsPerLayer = NULL, functionsPerLayer = NULL, weightDecay = 0.001, weights = NULL, verbose = interactive(), ...) } \arguments{ \item{Matrix}{\code{matrix} --- Preprocessed data with the variables in columns and observations in rows. The data may contain missing values, denoted as \code{NA}} \item{nPcs}{\code{numeric} -- Number of components to estimate. The preciseness of the missing value estimation depends on thenumber of components, which should resemble the internal structure of the data.} \item{maxSteps}{\code{numeric} -- Number of estimation steps. Default is based on a generous rule of thumb.} \item{unitsPerLayer}{The network units, example: c(2,4,6) for two input units 2feature units (principal components), one hidden layer fornon-linearity and three output units (original amount ofvariables).} \item{functionsPerLayer}{The function to apply at each layer eg. c("linr", "tanh", "linr")} \item{weightDecay}{Value between 0 and 1.} \item{weights}{Starting weights for the network. Defaults to uniform random values but can be set specifically to make algorithm deterministic.} \item{verbose}{\code{boolean} -- nlpca prints the number of steps and warning messages if set to TRUE. Default is interactive().} \item{...}{Reserved for future use. Not passed on anywhere.} } \value{ Standard PCA result object used by all PCA-basedmethods of this package. Contains scores, loadings, data meanand more. See \code{\link{pcaRes}} for details. } \description{ Neural network based non-linear PCA } \details{ Artificial Neural Network (MLP) for performing non-linear PCA. Non-linear PCA is conceptually similar to classical PCA but theoretically quite different. Instead of simply decomposing our matrix (X) to scores (T) loadings (P) and an error (E) we train a neural network (our loadings) to find a curve through the multidimensional space of X that describes a much variance as possible. Classical ways of interpreting PCA results are thus not applicable to NLPCA since the loadings are hidden in the network. However, the scores of components that lead to low cross-validation errors can still be interpreted via the score plot. Unfortunately this method depend on slow iterations which currently are implemented in R only making this method extremely slow. Furthermore, the algorithm does not by itself decide when it has converged but simply does 'maxSteps' iterations. } \examples{ ## Data set with three variables where data points constitute a helix data(helix) helixNA <- helix ## not a single complete observation helixNA <- t(apply(helix, 1, function(x) { x[sample(1:3, 1)] <- NA; x})) ## 50 steps is not enough, for good estimation use 1000 helixNlPca <- pca(helixNA, nPcs=1, method="nlpca", maxSteps=50) fittedData <- fitted(helixNlPca, helixNA) plot(fittedData[which(is.na(helixNA))], helix[which(is.na(helixNA))]) ## compared to solution by Nipals PCA which cannot extract non-linear patterns helixNipPca <- pca(helixNA, nPcs=2) fittedData <- fitted(helixNipPca) plot(fittedData[which(is.na(helixNA))], helix[which(is.na(helixNA))]) } \references{ Matthias Scholz, Fatma Kaplan, Charles L Guy, Joachim Kopkaand Joachim Selbig. Non-linear PCA: a missing data approach. \emph{Bioinformatics, 21(20):3887-3895, Oct 2005} } \author{ Based on a matlab script by Matthias Scholz and ported to R by Henning Redestig } pcaMethods/man/nmissing-pcaRes-method.Rd0000644000175400017540000000067413556116437021273 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{nmissing,pcaRes-method} \alias{nmissing,pcaRes-method} \alias{nmissing} \alias{nmissing,nniRes-method} \title{Missing values} \usage{ nmissing(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ Get the number of missing values } \description{ Missing values } \author{ Henning Redestig } pcaMethods/man/nni.Rd0000644000175400017540000000237313556116437015535 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pca.R \name{nni} \alias{nni} \title{Nearest neighbour imputation} \usage{ nni(object, method = c("llsImpute"), subset = numeric(), ...) } \arguments{ \item{object}{Numerical matrix with (or an object coercible to such) with samples in rows and variables as columns. Also takes \code{ExpressionSet} in which case the transposed expression matrix is used.} \item{method}{For convenience one can pass a large matrix but only use the variable specified as subset. Can be colnames or indices.} \item{subset}{Currently "llsImpute" only.} \item{...}{Further arguments to the chosen method.} } \value{ A \code{clusterRes} object. Or a list containing a clusterRes object as first and an ExpressionSet object as second entry if the input was of type ExpressionSet. } \description{ Wrapper function for imputation methods based on nearest neighbour clustering. Currently llsImpute only. } \details{ This method is wrapper function to llsImpute, See documentation for \code{link{llsImpute}}. } \examples{ data(metaboliteData) llsRes <- nni(metaboliteData, k=6, method="llsImpute", allGenes=TRUE) } \seealso{ \code{\link{llsImpute}}, \code{\link{pca}} } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/nniRes.Rd0000644000175400017540000000252713556116437016210 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/AllClasses.R \docType{class} \name{nniRes} \alias{nniRes} \alias{nniRes-class} \title{Class for representing a nearest neighbour imputation result} \description{ This is a class representation of nearest neighbour imputation (nni) result } \details{ \bold{Creating Objects}\cr \code{new("nniRes", completeObs=[the estimated complete observations], k=[cluster size], nObs=[amount of observations], nVar=[amount of variables], centered=[was the data centered befor running LLSimpute], center=[original means], method=[method used to perform clustering], missing=[amount of NAs])} \bold{Slots}\cr \describe{ \item{completeObs}{"matrix", the estimated complete observations} \item{nObs}{"numeric", amount of observations} \item{nVar}{"numeric", amount of variables} \item{correlation}{"character", the correlation method used (pearson, kendall or spearman)} \item{centered}{"logical", data was centered or not} \item{center}{"numeric", the original variable centers} \item{k}{"numeric", cluster size} \item{method}{"character", the method used to perform the clustering} \item{missing}{"numeric", the total amount of missing values in original data} } \bold{Methods}\cr \describe{ \item{print}{Print function} } } \author{ Wolfram Stacklies } \keyword{classes} pcaMethods/man/optiAlgCgd.Rd0000644000175400017540000000073713556116437016770 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/optiAlgCgd.R \name{optiAlgCgd} \alias{optiAlgCgd} \title{Conjugate gradient optimization} \usage{ optiAlgCgd(nlnet, trainIn, trainOut, verbose = FALSE) } \arguments{ \item{nlnet}{The nlnet} \item{trainIn}{Training data} \item{trainOut}{fitted data} \item{verbose}{logical, print messages} } \value{ ... } \description{ Conjugate gradient optimization } \author{ Henning Redestig, Matthias Scholz } pcaMethods/man/orth.Rd0000644000175400017540000000121213556116437015714 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/orth.R \name{orth} \alias{orth} \title{Calculate an orthonormal basis} \usage{ orth(mat, skipInac = FALSE) } \arguments{ \item{mat}{matrix to calculate orthonormal base} \item{skipInac}{do not include components with precision below .Machine$double.eps if TRUE} } \value{ orthonormal basis for the range of matrix } \description{ ONB = orth(mat) is an orthonormal basis for the range of matrix mat. That is, ONB' * ONB = I, the columns of ONB span the same space as the columns of mat, and the number of columns of ONB is the rank of mat. } \author{ Wolfram Stacklies } pcaMethods/man/pca.Rd0000644000175400017540000000752313556116437015516 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pca.R \name{pca} \alias{pca} \title{Perform principal component analysis} \usage{ pca(object, method, nPcs = 2, scale = c("none", "pareto", "vector", "uv"), center = TRUE, completeObs = TRUE, subset = NULL, cv = c("none", "q2"), ...) } \arguments{ \item{object}{Numerical matrix with (or an object coercible to such) with samples in rows and variables as columns. Also takes \code{ExpressionSet} in which case the transposed expression matrix is used. Can also be a data frame in which case all numberic variables are used to fit the PCA.} \item{method}{One of the methods reported by \code{listPcaMethods()}. Can be left missing in which case the \code{svd} PCA is chosen for data wihout missing values and \code{nipalsPca} for data with missing values} \item{nPcs}{Number of principal components to calculate.} \item{scale}{Scaling, see \code{\link{prep}}.} \item{center}{Centering, see \code{\link{prep}}.} \item{completeObs}{Sets the \code{completeObs} slot on the resulting \code{pcaRes} object containing the original data with but with all NAs replaced with the estimates.} \item{subset}{A subset of variables to use for calculating the model. Can be column names or indices.} \item{cv}{character naming a the type of cross-validation to be performed.} \item{...}{Arguments to \code{\link{prep}}, the chosen pca method and \code{\link{Q2}}.} } \value{ A \code{pcaRes} object. } \description{ Perform PCA on a numeric matrix for visualisation, information extraction and missing value imputation. } \details{ This method is wrapper function for the following set of pca methods: \describe{\item{svd:}{Uses classical \code{prcomp}. See documentation for \code{\link{svdPca}}.} \item{nipals:}{An iterative method capable of handling small amounts of missing values. See documentation for \code{\link{nipalsPca}}.} \item{rnipals:}{Same as nipals but implemented in R.} \item{bpca:}{An iterative method using a Bayesian model to handle missing values. See documentation for \code{\link{bpca}}.} \item{ppca:}{An iterative method using a probabilistic model to handle missing values. See documentation for \code{\link{ppca}}.} \item{svdImpute:}{Uses expectation maximation to perform SVD PCA on incomplete data. See documentation for \code{\link{svdImpute}}.}} Scaling and centering is part of the PCA model and handled by \code{\link{prep}}. } \examples{ data(iris) ## Usually some kind of scaling is appropriate pcIr <- pca(iris, method="svd", nPcs=2) pcIr <- pca(iris, method="nipals", nPcs=3, cv="q2") ## Get a short summary on the calculated model summary(pcIr) plot(pcIr) ## Scores and loadings plot slplot(pcIr, sl=as.character(iris[,5])) ## use an expressionset and ggplot data(sample.ExpressionSet) pc <- pca(sample.ExpressionSet) df <- merge(scores(pc), pData(sample.ExpressionSet), by=0) library(ggplot2) ggplot(df, aes(PC1, PC2, shape=sex, color=type)) + geom_point() + xlab(paste("PC1", pc@R2[1] * 100, "\% of the variance")) + ylab(paste("PC2", pc@R2[2] * 100, "\% of the variance")) } \references{ Wold, H. (1966) Estimation of principal components and related models by iterative least squares. In Multivariate Analysis (Ed., P.R. Krishnaiah), Academic Press, NY, 391-420. Shigeyuki Oba, Masa-aki Sato, Ichiro Takemasa, Morito Monden, Ken-ichi Matsubara and Shin Ishii. A Bayesian missing value estimation method for gene expression profile data. \emph{Bioinformatics, 19(16):2088-2096, Nov 2003}. Troyanskaya O. and Cantor M. and Sherlock G. and Brown P. and Hastie T. and Tibshirani R. and Botstein D. and Altman RB. - Missing value estimation methods for DNA microarrays. \emph{Bioinformatics. 2001 Jun;17(6):520-5}. } \seealso{ \code{\link{prcomp}}, \code{\link{princomp}}, \code{\link{nipalsPca}}, \code{\link{svdPca}} } \author{ Wolfram Stacklies, Henning Redestig } \keyword{multivariate} pcaMethods/man/pcaMethods-deprecated.Rd0000644000175400017540000000105413556116437021131 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pcaMethods-package.R \name{pcaMethods-deprecated} \alias{pcaMethods-deprecated} \title{Deprecated methods for pcaMethods} \description{ \describe{ \item{plotR2}{Lack of relevance for this plot and the fact that it can not show cross-validation based diagnostics in the same plot makes it redundant with the introduction of a dedicated \code{plot} function for \code{pcaRes}. The new plot only shows R2cum but the result is pretty much the same.}} } \author{ Henning Redestig } pcaMethods/man/pcaMethods.Rd0000644000175400017540000000221713556116437017035 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pcaMethods-package.R \docType{package} \name{pcaMethods} \alias{pcaMethods} \alias{pcaMethods-package} \title{pcaMethods} \description{ Principal Component Analysis in R } \details{ \tabular{ll}{ Package: \tab pcaMethods \cr Type: \tab Package \cr Developed since: \tab 2006 \cr License: \tab GPL (>=3) \cr LazyLoad: \tab yes \cr } Provides Bayesian PCA, Probabilistic PCA, Nipals PCA, Inverse Non-Linear PCA and the conventional SVD PCA. A cluster based method for missing value estimation is included for comparison. BPCA, PPCA and NipalsPCA may be used to perform PCA on incomplete data as well as for accurate missing value estimation. A set of methods for printing and plotting the results is also provided. All PCA methods make use of the same data structure (pcaRes) to provide a unique interface to the PCA results. Developed at the Max-Planck Institute for Molecular Plant Physiology, Golm, Germany, RIKEN Plant Science Center Yokohama, Japan, and CAS-MPG Partner Institute for Computational Biology (PICB) Shanghai, P.R. China } \author{ Wolfram Stacklies, Henning Redestig } pcaMethods/man/pcaNet.Rd0000644000175400017540000000657213556116437016170 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/AllClasses.R \docType{class} \name{pcaNet} \alias{pcaNet} \alias{nlpcaNet} \alias{nlpcaNet-class} \title{Class representation of the NLPCA neural net} \description{ This is a class representation of a non-linear PCA neural network. The \code{nlpcaNet} class is not meant for user-level usage. } \details{ Creating Objects \code{new("nlpcaNet", net=[the network structure], hierarchic=[hierarchic design], fct=[the functions at each layer], fkt=[the functions used for forward propagation], weightDecay=[incremental decrease of weight changes over iterations (between 0 and 1)], featureSorting=[sort features or not], dataDist=[represents the present values], inverse=[net is inverse mode or not], fCount=[amount of times features were sorted], componentLayer=[which layer is the 'bottleneck' (principal components)], erro=[the used error function], gradient=[the used gradient method], weights=[the present weights], maxIter=[the amount of iterations that was done], scalingFactor=[the scale of the original matrix])} Slots \describe{ \item{net}{"matrix", matrix showing the representation of the neural network, e.g. (2,4,6) for a network with two features, a hidden layer and six output neurons (original variables).} \item{hierarchic}{"list", the hierarchic design of the network, holds 'idx' (), 'var' () and layer (which layer is the principal component layer).} \item{fct}{"character", a vector naming the functions that will be applied on each layer. "linr" is linear (i.e.) standard matrix products and "tanh" means that the arcus tangens is applied on the result of the matrix product (for non-linearity).} \item{fkt}{"character", same as fct but the functions used during back propagation.} \item{weightDecay}{"numeric", the value that is used to incrementally decrease the weight changes to ensure convergence.} \item{featureSorting}{"logical", indicates if features will be sorted or not. This is used to make the NLPCA assume properties closer to those of standard PCA were the first component is more important for reconstructing the data than the second component.} \item{dataDist}{"matrix", a matrix of ones and zeroes indicating which values will add to the errror.} \item{inverse}{"logical", network is inverse mode (currently only inverse is supported) or not. Eg. the case when we have truly missing values and wish to impute them.} \item{fCount}{"integer", Counter for the amount of times features were really sorted.} \item{componentLayer}{"numeric", the index of 'net' that is the component layer.} \item{error}{"function", the used error function. Currently only one is provided \code{errorHierarchic}.} \item{gradient}{"function", the used gradient function. Currently only one is provided \code{derrorHierarchic}} \item{weights}{"list", A list holding managements of the weights. The list has two functions, weights$current() and weights$set() which access a matrix in the local environment of this object.} \item{maxIter}{"integer", the amount of iterations used to train this network.} \item{scalingFactor}{"numeric", training the network is best made with 'small' values so the original data is scaled down to a suitable range by division with this number.}} Methods \describe{ \item{vector2matrices}{Returns the weights in a matrix representation.} } } \seealso{ \code{\link{nlpca}} } \author{ Henning Redestig } \keyword{classes} pcaMethods/man/pcaRes.Rd0000644000175400017540000000577213556116437016174 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/AllClasses.R \docType{class} \name{pcaRes} \alias{pcaRes} \alias{pcaRes-class} \title{Class for representing a PCA result} \description{ This is a class representation of a PCA result } \details{ \bold{Creating Objects}\cr \code{new("pcaRes", scores=[the scores], loadings=[the loadings], nPcs=[amount of PCs], R2cum=[cumulative R2], nObs=[amount of observations], nVar=[amount of variables], R2=[R2 for each individual PC], sDev=[stdev for each individual PC], centered=[was data centered], center=[original means], varLimit=[what variance limit was exceeded], method=[method used to calculate PCA], missing=[amount of NAs], completeObs=[estimated complete observations])} \bold{Slots}\cr \describe{ \item{scores}{"matrix", the calculated scores} \item{loadings}{"matrix", the calculated loadings} \item{R2cum}{"numeric", the cumulative R2 values} \item{sDev}{"numeric", the individual standard deviations of the score vectors} \item{R2}{"numeric", the individual R2 values} \item{cvstat}{"numeric", cross-validation statistics} \item{nObs}{"numeric", number of observations} \item{nVar}{"numeric", number of variables} \item{centered}{"logical", data was centered or not} \item{center}{"numeric", the original variable centers} \item{scaled}{"logical", data was scaled or not} \item{scl}{"numeric", the original variable scales} \item{varLimit}{"numeric", the exceeded variance limit} \item{nPcs,nP}{"numeric", the number of calculated PCs} \item{method}{"character", the method used to perform PCA} \item{missing}{"numeric", the total amount of missing values in original data} \item{completeObs}{"matrix", the estimated complete observations} \item{network}{"nlpcaNet", the network used by non-linear PCA} } \bold{Methods (not necessarily exhaustive)}\cr \describe{ \item{print}{Print function} \item{summary}{Extract information about PC relevance} \item{screeplot}{Plot a barplot of standard deviations for PCs} \item{slplot}{Make a side by side score and loadings plot} \item{nPcs}{Get the number of PCs} \item{nObs}{Get the number of observations} \item{cvstat}{Cross-validation statistics} \item{nVar}{Get the number of variables} \item{loadings}{Get the loadings} \item{scores}{Get the scores} \item{dim}{Get the dimensions (number of observations, number of features)} \item{centered}{Get a logical indicating if centering was done as part of the model} \item{center}{Get the averages of the original variables.} \item{completeObs}{Get the imputed data set} \item{method}{Get a string naming the used PCA method} \item{sDev}{Get the standard deviations of the PCs} \item{scaled}{Get a logical indicating if scaling was done as part of the model} \item{scl}{Get the scales of the original variablesb} \item{R2cum}{Get the cumulative R2} } } \author{ Henning Redestig } \keyword{classes} pcaMethods/man/plot.pcaRes.Rd0000644000175400017540000000237413556116437017144 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \name{plot.pcaRes} \alias{plot.pcaRes} \alias{plot,pcaRes-method} \title{Plot diagnostics (screeplot)} \usage{ \method{plot}{pcaRes}(x, y = NULL, main = deparse(substitute(object)), col = gray(c(0.9, 0.5)), ...) } \arguments{ \item{x}{\code{pcaRes} The pcaRes object.} \item{y}{not used} \item{main}{title of the plot} \item{col}{Colors of the bars} \item{...}{further arguments to barplot} } \value{ None, used for side effect. } \description{ Plot the computed diagnostics of PCA model to get an idea of their importance. Note though that the standard screeplot shows the standard deviations for the PCs this method shows the R2 values which empirically shows the importance of the P's and is thus applicable for any PCA method rather than just SVD based PCA. } \details{ If cross-validation was done for the PCA the plot will also show the CV based statistics. A common rule-of-thumb for determining the optimal number of PCs is the PC where the CV diagnostic is at its maximum but not very far from \eqn{R^2}. } \examples{ data(metaboliteData) pc <- pca(t(metaboliteData), nPcs=5, cv="q2", scale="uv") plot(pc) } \seealso{ \link{screeplot} } \author{ Henning Redestig } pcaMethods/man/plotPcs.Rd0000644000175400017540000000246213556116437016374 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pca.R \name{plotPcs} \alias{plotPcs} \title{Plot many side by side scores XOR loadings plots} \usage{ plotPcs(object, pcs = 1:nP(object), type = c("scores", "loadings"), sl = NULL, hotelling = 0.95, ...) } \arguments{ \item{object}{\code{pcaRes} a pcaRes object} \item{pcs}{\code{numeric} which pcs to plot} \item{type}{\code{character} Either "scores" or "loadings" for scores or loadings plot respectively} \item{sl}{\code{character} Text labels to plot instead of a point, if NULL points are plotted instead of text} \item{hotelling}{\code{numeric} Significance level for the confidence ellipse. NULL means that no ellipse is drawn.} \item{...}{Further arguments to \code{\link{pairs}} on which this function is based.} } \value{ None, used for side effect. } \description{ A function that can be used to visualise many PCs plotted against each other } \details{ Uses \code{\link{pairs}} to provide side-by-side plots. Note that this function only plots scores or loadings but not both in the same plot. } \examples{ data(iris) pcIr <- pca(iris[,1:4], nPcs=3, method="svd") plotPcs(pcIr, col=as.integer(iris[,4]) + 1) } \seealso{ \code{prcomp}, \code{pca}, \code{princomp}, \code{slplot} } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/ppca.Rd0000644000175400017540000000727413556116437015701 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/ppca.R \name{ppca} \alias{ppca} \title{Probabilistic PCA} \usage{ ppca(Matrix, nPcs = 2, seed = NA, threshold = 1e-05, maxIterations = 1000, ...) } \arguments{ \item{Matrix}{\code{matrix} -- Data containing the variables in columns and observations in rows. The data may contain missing values, denoted as \code{NA}.} \item{nPcs}{\code{numeric} -- Number of components to estimate. The preciseness of the missing value estimation depends on the number of components, which should resemble the internal structure of the data.} \item{seed}{\code{numeric} Set the seed for the random number generator. PPCA creates fills the initial loading matrix with random numbers chosen from a normal distribution. Thus results may vary slightly. Set the seed for exact reproduction of your results.} \item{threshold}{Convergence threshold.} \item{maxIterations}{the maximum number of allowed iterations} \item{...}{Reserved for future use. Currently no further parameters are used.} } \value{ Standard PCA result object used by all PCA-based methods of this package. Contains scores, loadings, data mean and more. See \code{\link{pcaRes}} for details. } \description{ Implementation of probabilistic PCA (PPCA). PPCA allows to perform PCA on incomplete data and may be used for missing value estimation. This script was implemented after the Matlab version provided by Jakob Verbeek ( see \url{http://lear.inrialpes.fr/~verbeek/}) and the draft \emph{``EM Algorithms for PCA and Sensible PCA''} written by Sam Roweis. } \details{ Probabilistic PCA combines an EM approach for PCA with a probabilistic model. The EM approach is based on the assumption that the latent variables as well as the noise are normal distributed. In standard PCA data which is far from the training set but close to the principal subspace may have the same reconstruction error. PPCA defines a likelihood function such that the likelihood for data far from the training set is much lower, even if they are close to the principal subspace. This allows to improve the estimation accuracy. A method called \code{kEstimate} is provided to estimate the optimal number of components via cross validation. In general few components are sufficient for reasonable estimation accuracy. See also the package documentation for further discussion on what kind of data PCA-based missing value estimation is advisable. \bold{Complexity:}\cr Runtime is linear in the number of data, number of data dimensions and number of principal components. \bold{Convergence:} The threshold indicating convergence was changed from 1e-3 in 1.2.x to 1e-5 in the current version leading to more stable results. For reproducability you can set the seed (parameter seed) of the random number generator. If used for missing value estimation, results may be checked by simply running the algorithm several times with changing seed, if the estimated values show little variance the algorithm converged well. } \note{ Requires \code{MASS}. It is not recommended to use this function directely but rather to use the pca() wrapper function. } \examples{ ## Load a sample metabolite dataset with 5\\\% missing values (metaboliteData) data(metaboliteData) ## Perform probabilistic PCA using the 3 largest components result <- pca(t(metaboliteData), method="ppca", nPcs=3, seed=123) ## Get the estimated complete observations cObs <- completeObs(result) ## Plot the scores plotPcs(result, type = "scores") \dontshow{ stopifnot(sum((fitted(result) - t(metaboliteData))^2, na.rm=TRUE) < 200) } } \seealso{ \code{\link{bpca}, \link{svdImpute}, \link{prcomp}, \link{nipalsPca}, \link{pca}, \link{pcaRes}}. } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/predict-methods.Rd0000644000175400017540000000335313556116437020043 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{predict-methods} \alias{predict-methods} \alias{predict.pcaRes} \alias{predict,pcaRes-method} \title{Predict values from PCA.} \usage{ \method{predict}{pcaRes}(object, newdata, pcs = nP(object), pre = TRUE, post = TRUE, ...) \S4method{predict}{pcaRes}(object, newdata, pcs = nP(object), pre = TRUE, post = TRUE, ...) } \arguments{ \item{object}{\code{pcaRes} the \code{pcaRes} object of interest.} \item{newdata}{\code{matrix} new data with same number of columns as the used to compute \code{object}.} \item{pcs}{\code{numeric} The number of PC's to consider} \item{pre}{pre-process \code{newdata} based on the pre-processing chosen for the PCA model} \item{post}{unpre-process the final data (add the center back etc)} \item{...}{Not passed on anywhere, included for S3 consistency.} } \value{ A list with the following components: \item{scores}{The predicted scores} \item{x}{The predicted data} } \description{ Predict data using PCA model } \details{ This function extracts the predict values from a pcaRes object for the PCA methods SVD, Nipals, PPCA and BPCA. Newdata is first centered if the PCA model was and then scores (\eqn{T}) and data (\eqn{X}) is 'predicted' according to : \eqn{\hat{T}=X_{new}P}{That=XnewP} \eqn{\hat{X}_{new}=\hat{T}P'}{Xhat=ThatP'}. Missing values are set to zero before matrix multiplication to achieve NIPALS like treatment of missing values. } \examples{ data(iris) hidden <- sample(nrow(iris), 50) pcIr <- pca(iris[-hidden,1:4]) pcFull <- pca(iris[,1:4]) irisHat <- predict(pcIr, iris[hidden,1:4]) cor(irisHat$scores[,1], scores(pcFull)[hidden,1]) } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/prep.Rd0000644000175400017540000000445613556116437015723 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/prep.R \name{prep} \alias{prep} \title{Pre-process a matrix for PCA} \usage{ prep(object, scale = c("none", "pareto", "vector", "uv"), center = TRUE, eps = 1e-12, simple = TRUE, reverse = FALSE, ...) } \arguments{ \item{object}{Numerical matrix (or an object coercible to such) with samples in rows and variables as columns. Also takes \code{ExpressionSet} in which case the transposed expression matrix is used.} \item{scale}{One of "UV" (unit variance \eqn{a=a/\sigma_{a}}) "vector" (vector normalisation \eqn{b=b/||b||}), "pareto" (sqrt UV) or "none" to indicate which scaling should be used to scale the matrix with \eqn{a} variables and \eqn{b} samples. Can also be a vector of scales which should be used to scale the matrix. \code{NULL} value is interpreted as \code{"none"}.} \item{center}{Either a logical which indicates if the matrix should be mean centred or not, or a vector with averages which should be suntracted from the matrix. \code{NULL} value is interpreted as \code{FALSE}} \item{eps}{Minimum variance, variable with lower variance are not scaled and warning is issued instead.} \item{simple}{Logical indicating if only the data should be returned or a list with the pre-processing statistics as well.} \item{reverse}{Logical indicating if matrix should be 'post-processed' instead by multiplying each column with its scale and adding the center. In this case, center and scale should be vectors with the statistics (no warning is issued if not, instead output becomes the same as input).} \item{...}{Only used for passing through arguments.} } \value{ A pre-processed matrix or a list with \item{center}{a vector with the estimated centers} \item{scale}{a vector with the estimated scales} \item{data}{the pre (or post) processed data} } \description{ Scaling and centering a matrix. } \details{ Does basically the same as \code{\link{scale}} but adds some alternative scaling options and functionality for treating pre-processing as part of a model. } \examples{ object <- matrix(rnorm(50), nrow=10) res <- prep(object, scale="uv", center=TRUE, simple=FALSE) obj <- prep(object, scale=res$scale, center=res$center) ## same as original sum((object - prep(obj, scale=res$scale, center=res$center, rev=TRUE))^2) } \author{ Henning Redestig } pcaMethods/man/rediduals-methods.Rd0000644000175400017540000000216513556116437020365 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{rediduals-methods} \alias{rediduals-methods} \alias{residuals.pcaRes} \alias{residuals,pcaRes-method} \alias{resid,pcaRes-method} \title{Residuals values from a PCA model.} \usage{ \method{residuals}{pcaRes}(object, data = completeObs(object), ...) \S4method{residuals}{pcaRes}(object, data = completeObs(object), ...) \S4method{resid}{pcaRes}(object, data = completeObs(object), ...) } \arguments{ \item{object}{\code{pcaRes} the \code{pcaRes} object of interest.} \item{data}{\code{matrix} The data that was used to calculate the PCA model (or a different dataset to e.g. adress its proximity to the model).} \item{...}{Passed on to \code{\link{predict.pcaRes}}. E.g. setting the number of used components.} } \value{ A \code{matrix} with the residuals } \description{ This function extracts the residuals values from a pcaRes object for the PCA methods SVD, Nipals, PPCA and BPCA } \examples{ data(iris) pcIr <- pca(iris[,1:4]) head(residuals(pcIr, iris[,1:4])) } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/repmat.Rd0000644000175400017540000000077313556116437016243 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/repmat.R \name{repmat} \alias{repmat} \title{Replicate and tile an array.} \usage{ repmat(mat, M, N) } \arguments{ \item{mat}{numeric matrix} \item{M}{number of copies in vertical direction} \item{N}{number of copies in horizontal direction} } \value{ Matrix consiting of M-by-N tiling copies of input matrix } \description{ Creates a large matrix B consisting of an M-by-N tiling of copies of A } \author{ Wolfram Stacklies } pcaMethods/man/robustPca.Rd0000644000175400017540000000632013556116437016707 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/robustPca.R \name{robustPca} \alias{robustPca} \title{PCA implementation based on robustSvd} \usage{ robustPca(Matrix, nPcs = 2, verbose = interactive(), ...) } \arguments{ \item{Matrix}{\code{matrix} -- Data containing the variables in columns and observations in rows. The data may contain missing values, denoted as \code{NA}.} \item{nPcs}{\code{numeric} -- Number of components to estimate. The preciseness of the missing value estimation depends on the number of components, which should resemble the internal structure of the data.} \item{verbose}{\code{boolean} Print some output to the command line if TRUE} \item{...}{Reserved for future use. Currently no further parameters are used} } \value{ Standard PCA result object used by all PCA-based methods of this package. Contains scores, loadings, data mean and more. See \code{\link{pcaRes}} for details. are used. } \description{ This is a PCA implementation robust to outliers in a data set. It can also handle missing values, it is however NOT intended to be used for missing value estimation. As it is based on robustSVD we will get an accurate estimation for the loadings also for incomplete data or for data with outliers. The returned scores are, however, affected by the outliers as they are calculated inputData X loadings. This also implies that you should look at the returned R2/R2cum values with caution. If the data show missing values, scores are caluclated by just setting all NA - values to zero. This is not expected to produce accurate results. Please have also a look at the manual page for \code{robustSvd}. Thus this method should mainly be seen as an attempt to integrate \code{robustSvd()} into the framework of this package. Use one of the other methods coming with this package (like PPCA or BPCA) if you want to do missing value estimation. It is not recommended to use this function directely but rather to use the pca() wrapper function. } \details{ The method is very similar to the standard \code{prcomp()} function. The main difference is that \code{robustSvd()} is used instead of the conventional \code{svd()} method. } \examples{ ## Load a complete sample metabolite data set and mean center the data data(metaboliteDataComplete) mdc <- scale(metaboliteDataComplete, center=TRUE, scale=FALSE) ## Now create 5\\\% of outliers. cond <- runif(length(mdc)) < 0.05; mdcOut <- mdc mdcOut[cond] <- 10 ## Now we do a conventional PCA and robustPca on the original and the data ## with outliers. ## We use center=FALSE here because the large artificial outliers would ## affect the means and not allow to objectively compare the results. resSvd <- pca(mdc, method="svd", nPcs=10, center=FALSE) resSvdOut <- pca(mdcOut, method="svd", nPcs=10, center=FALSE) resRobPca <- pca(mdcOut, method="robustPca", nPcs=10, center=FALSE) ## Now we plot the results for the original data against those with outliers ## We can see that robustPca is hardly effected by the outliers. plot(loadings(resSvd)[,1], loadings(resSvdOut)[,1]) plot(loadings(resSvd)[,1], loadings(resRobPca)[,1]) } \seealso{ \code{\link{robustSvd}, \link{svd}, \link{prcomp}, \link{pcaRes}}. } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/robustSvd.Rd0000644000175400017540000000620413556116437016741 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/robustPca.R \name{robustSvd} \alias{robustSvd} \title{Alternating L1 Singular Value Decomposition} \usage{ robustSvd(x) } \arguments{ \item{x}{A matrix whose SVD decomposition is to be computed. Missing values are allowed.} } \value{ The robust SVD of the matrix is x=u d v'. \item{d}{A vector containing the singular values of \code{x}.} \item{u}{A matrix whose columns are the left singular vectors of \code{x}.} \item{v}{A matrix whose columns are the right singular vectors of \code{x}.} } \description{ A robust approximation to the singular value decomposition of a rectangular matrix is computed using an alternating L1 norm (instead of the more usual least squares L2 norm). As the SVD is a least-squares procedure, it is highly susceptible to outliers and in the extreme case, an individual cell (if sufficiently outlying) can draw even the leading principal component toward itself. } \details{ See Hawkins et al (2001) for details on the robust SVD algorithm. Briefly, the idea is to sequentially estimate the left and right eigenvectors using an L1 (absolute value) norm minimization. Note that the robust SVD is able to accomodate missing values in the matrix \code{x}, unlike the usual \code{svd} function. Also note that the eigenvectors returned by the robust SVD algorithm are NOT (in general) orthogonal and the eigenvalues need not be descending in order. } \note{ Two differences from the usual SVD may be noted. One relates to orthogonality. In the conventional SVD, all the eigenvectors are orthogonal even if not explicitly imposed. Those returned by the AL1 algorithm (used here) are (in general) not orthogonal. Another difference is that, in the L2 analysis of the conventional SVD, the successive eigen triples (eigenvalue, left eigenvector, right eigenvector) are found in descending order of eigenvalue. This is not necessarily the case with the AL1 algorithm. Hawkins et al (2001) note that a larger eigen value may follow a smaller one. } \examples{ ## Load a complete sample metabolite data set and mean center the data data(metaboliteDataComplete) mdc <- prep(metaboliteDataComplete, center=TRUE, scale="none") ## Now create 5\% of outliers. cond <- runif(length(mdc)) < 0.05; mdcOut <- mdc mdcOut[cond] <- 10 ## Now we do a conventional SVD and a robustSvd on both, the original and the ## data with outliers. resSvd <- svd(mdc) resSvdOut <- svd(mdcOut) resRobSvd <- robustSvd(mdc) resRobSvdOut <- robustSvd(mdcOut) ## Now we plot the results for the original data against those with outliers ## We can see that robustSvd is hardly affected by the outliers. plot(resSvd$v[,1], resSvdOut$v[,1]) plot(resRobSvd$v[,1], resRobSvdOut$v[,1]) } \references{ Hawkins, Douglas M, Li Liu, and S Stanley Young (2001) Robust Singular Value Decomposition, National Institute of Statistical Sciences, Technical Report Number 122. \url{http://www.niss.org/technicalreports/tr122.pdf} } \seealso{ \code{\link{svd}}, \code{\link[ade4:nipals]{nipals}} for an alternating L2 norm method that also accommodates missing data. } \author{ Kevin Wright, modifications by Wolfram Stacklies } \keyword{algebra} pcaMethods/man/sDev-pcaRes-method.Rd0000644000175400017540000000077313556116437020345 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{sDev,pcaRes-method} \alias{sDev,pcaRes-method} \alias{sDev} \title{Get the standard deviations of the scores (indicates their relevance)} \usage{ sDev(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ Standard devations of the scores } \description{ Get the standard deviations of the scores (indicates their relevance) } \author{ Henning Redestig } pcaMethods/man/scaled-pcaRes-method.Rd0000644000175400017540000000072613556116437020675 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{scaled,pcaRes-method} \alias{scaled,pcaRes-method} \alias{scaled} \title{Check if scaling was part of the PCA model} \usage{ scaled(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ TRUE if scaling was part of the PCA model } \description{ Check if scaling was part of the PCA model } \author{ Henning Redestig } pcaMethods/man/scl-pcaRes-method.Rd0000644000175400017540000000101013556116437020206 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{scl,pcaRes-method} \alias{scl,pcaRes-method} \alias{scl} \title{Get the scales (e.g. standard deviations) of the original variables} \usage{ scl(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ Vector with the scales } \description{ Get the scales (e.g. standard deviations) of the original variables } \seealso{ \code{\link{prep}} } \author{ Henning Redestig } pcaMethods/man/scores-pcaRes-method.Rd0000644000175400017540000000075213556116437020737 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{scores,pcaRes-method} \alias{scores,pcaRes-method} \alias{scores} \title{Get scores from a pcaRes object} \usage{ \S4method{scores}{pcaRes}(object, ...) } \arguments{ \item{object}{a pcaRes object} \item{...}{not used} } \value{ The scores as a matrix } \description{ Get scores from a pcaRes object } \seealso{ \code{\link{scores.pcaRes}} } \author{ Henning Redestig } pcaMethods/man/scores.pcaRes.Rd0000644000175400017540000000062113556116437017455 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \name{scores.pcaRes} \alias{scores.pcaRes} \title{Get scores from a pcaRes object} \usage{ \method{scores}{pcaRes}(object, ...) } \arguments{ \item{object}{a pcaRes object} \item{...}{not used} } \value{ The scores as a matrix } \description{ Get scores from a pcaRes object } \author{ Henning Redestig } pcaMethods/man/show-methods.Rd0000644000175400017540000000122113556116437017361 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{show-methods} \alias{show-methods} \alias{showPcaRes} \alias{print,pcaRes-method} \alias{print,nniRes-method} \alias{show,pcaRes-method} \alias{show,nniRes-method} \title{Print/Show for pcaRes} \usage{ showPcaRes(x, ...) \S4method{print}{pcaRes}(x, ...) \S4method{show}{pcaRes}(object) } \arguments{ \item{x}{a pcaRes object} \item{...}{not used} \item{object}{the object to print information about} } \value{ nothing, used for its side effect } \description{ Print basic information about pcaRes object } \author{ Henning Redestig } pcaMethods/man/showNniRes.Rd0000644000175400017540000000060213556116437017041 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-nniRes.R \name{showNniRes} \alias{showNniRes} \title{Print a nniRes model} \usage{ showNniRes(x, ...) } \arguments{ \item{x}{An \code{nniRes} object} \item{...}{Not used} } \value{ Nothing, used for side-effect } \description{ Print a brief description of nniRes model } \author{ Henning Redestig } pcaMethods/man/simpleEllipse.Rd0000644000175400017540000000156313556116437017560 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pca.R \name{simpleEllipse} \alias{simpleEllipse} \title{Hotelling's T^2 Ellipse} \usage{ simpleEllipse(x, y, alfa = 0.95, len = 200) } \arguments{ \item{x}{first variable} \item{y}{second variable} \item{alfa}{confidence level of the circle} \item{len}{Number of points in the circle} } \value{ A matrix with X and Y coordinates for the circle } \description{ Get a confidence ellipse for uncorrelated bivariate data } \details{ As described in 'Introduction to multi and megavariate data analysis using PCA and PLS' by Eriksson et al. This produces very similar ellipse as compared to the ellipse function the ellipse package except that this function assumes that and y are uncorrelated (which they of are if they are scores or loadings from a PCA). } \seealso{ ellipse } \author{ Henning Redestig } pcaMethods/man/slplot-pcaRes-method.Rd0000644000175400017540000000340613556116437020755 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{slplot,pcaRes-method} \alias{slplot,pcaRes-method} \alias{slplot} \title{Side by side scores and loadings plot} \usage{ slplot(object, pcs=c(1,2), scoresLoadings=c(TRUE, TRUE), sl="def", ll="def", hotelling=0.95, rug=TRUE, sub=NULL,...) } \arguments{ \item{object}{a pcaRes object} \item{pcs}{which two pcs to plot} \item{scoresLoadings}{Which should be shown scores and or loadings} \item{sl}{labels to plot in the scores plot} \item{ll}{labels to plot in the loadings plot} \item{hotelling}{confidence interval for ellipse in the score plot} \item{rug}{logical, rug x axis in score plot or not} \item{sub}{Subtitle, defaults to annotate with amount of explained variance.} \item{...}{Further arguments to plot functions. Prefix arguments to \code{par()} with 's' for the scores plot and 'l' for the loadings plot. I.e. cex become scex for setting character expansion in the score plot and lcex for the loadings plot.} } \value{ None, used for side effect. } \description{ A common way of visualizing two principal components } \details{ This method is meant to be used as a quick way to visualize results, if you want a more specific plot you probably want to get the scores, loadings with \code{scores(object)}, \code{loadings(object)} and then design your own plotting method. } \note{ Uses layout instead of par to provide side-by-side so it works with Sweave (but can not be combined with \code{par(mfrow=..))} } \examples{ data(iris) pcIr <- pca(iris[,1:4], scale="uv") slplot(pcIr, sl=NULL, spch=5) slplot(pcIr, sl=NULL, lcex=1.3, scol=as.integer(iris[,5])) } \seealso{ \code{\link{pca}}, \code{\link{biplot}} } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/sortFeatures.Rd0000644000175400017540000000070513556116437017434 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/sortFeatures.R \name{sortFeatures} \alias{sortFeatures} \title{Sort the features of NLPCA object} \usage{ sortFeatures(nlnet, trainIn, trainOut) } \arguments{ \item{nlnet}{The nlnet} \item{trainIn}{Training data in} \item{trainOut}{Training data after it passed through the net} } \value{ ... } \description{ Sort the features of NLPCA object } \author{ Henning Redestig } pcaMethods/man/summary.Rd0000644000175400017540000000070213556116437016440 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \name{summary} \alias{summary} \alias{summary.pcaRes} \alias{summary,pcaRes-method} \title{Summary of PCA model} \usage{ \method{summary}{pcaRes}(object, ...) } \arguments{ \item{object}{a pcaRes object} \item{...}{Not used} } \value{ Nothing, used for side-effect } \description{ Print a brief description of the PCA model } \author{ Henning Redestig } pcaMethods/man/svdImpute.Rd0000644000175400017540000000602413556116437016726 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/svdImpute.R \name{svdImpute} \alias{svdImpute} \title{SVDimpute algorithm} \usage{ svdImpute(Matrix, nPcs = 2, threshold = 0.01, maxSteps = 100, verbose = interactive(), ...) } \arguments{ \item{Matrix}{\code{matrix} -- Pre-processed (centered, scaled) data with variables in columns and observations in rows. The data may contain missing values, denoted as \code{NA}.} \item{nPcs}{\code{numeric} -- Number of components to estimate. The preciseness of the missing value estimation depends on the number of components, which should resemble the internal structure of the data.} \item{threshold}{The iteration stops if the change in the matrix falls below this threshold.} \item{maxSteps}{Maximum number of iteration steps.} \item{verbose}{Print some output if TRUE.} \item{...}{Reserved for parameters used in future version of the algorithm} } \value{ Standard PCA result object used by all PCA-based methods of this package. Contains scores, loadings, data mean and more. See \code{\link{pcaRes}} for details. } \description{ This implements the SVDimpute algorithm as proposed by Troyanskaya et al, 2001. The idea behind the algorithm is to estimate the missing values as a linear combination of the \code{k} most significant eigengenes. } \details{ Missing values are denoted as \code{NA}. It is not recommended to use this function directely but rather to use the pca() wrapper function. As SVD can only be performed on complete matrices, all missing values are initially replaced by 0 (what is in fact the mean on centred data). The algorithm works iteratively until the change in the estimated solution falls below a certain threshold. Each step the eigengenes of the current estimate are calculated and used to determine a new estimate. Eigengenes denote the loadings if pca is performed considering variable (for Microarray data genes) as observations. An optimal linear combination is found by regressing the incomplete variable against the \code{k} most significant eigengenes. If the value at position \code{j} is missing, the \eqn{j^th}{j^th} value of the eigengenes is not used when determining the regression coefficients. } \note{ Each iteration, standard PCA (\code{prcomp}) needs to be done for each incomplete variable to get the eigengenes. This is usually fast for small data sets, but complexity may rise if the data sets become very large. } \examples{ ## Load a sample metabolite dataset with 5\\\% missing values data(metaboliteData) ## Perform svdImpute using the 3 largest components result <- pca(metaboliteData, method="svdImpute", nPcs=3, center = TRUE) ## Get the estimated complete observations cObs <- completeObs(result) ## Now plot the scores plotPcs(result, type = "scores") } \references{ Troyanskaya O. and Cantor M. and Sherlock G. and Brown P. and Hastie T. and Tibshirani R. and Botstein D. and Altman RB. - Missing value estimation methods for DNA microarrays. \emph{Bioinformatics. 2001 Jun;17(6):520-5.} } \author{ Wolfram Stacklies } \keyword{multivariate} pcaMethods/man/svdPca.Rd0000644000175400017540000000260513556116437016167 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/pca.R \name{svdPca} \alias{svdPca} \title{Perform principal component analysis using singular value decomposition} \usage{ svdPca(Matrix, nPcs = 2, varLimit = 1, verbose = interactive(), ...) } \arguments{ \item{Matrix}{Pre-processed (centered and possibly scaled) numerical matrix samples in rows and variables as columns. No missing values allowed.} \item{nPcs}{Number of components that should be extracted.} \item{varLimit}{Optionally the ratio of variance that should be explained. \code{nPcs} is ignored if varLimit < 1} \item{verbose}{Verbose complaints to matrix structure} \item{...}{Only used for passing through arguments.} } \value{ A \code{pcaRes} object. } \description{ A wrapper function for \code{prcomp} to deliver the result as a \code{pcaRes} method. Supplied for compatibility with the rest of the pcaMethods package. It is not recommended to use this function directely but rather to use the \code{pca()} wrapper function. } \examples{ data(metaboliteDataComplete) mat <- prep(t(metaboliteDataComplete)) pc <- svdPca(mat, nPcs=2) ## better use pca() pc <- pca(t(metaboliteDataComplete), method="svd", nPcs=2) \dontshow{stopifnot(sum((fitted(pc) - t(metaboliteDataComplete))^2, na.rm=TRUE) < 200)} } \seealso{ \code{prcomp}, \code{princomp}, \code{pca} } \author{ Henning Redestig } \keyword{multivariate} pcaMethods/man/tempFixNas.Rd0000644000175400017540000000063413556116437017025 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/xval.R \name{tempFixNas} \alias{tempFixNas} \title{Temporary fix for missing values} \usage{ tempFixNas(mat) } \arguments{ \item{mat}{a matrix} } \value{ The original matrix with completely missing rows/cols filled with zeroes. } \description{ Simply replace completely missing rows or cols with zeroes. } \author{ Henning Redestig } pcaMethods/man/vector2matrices-matrix-method.Rd0000644000175400017540000000100213556116437022631 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/vector2matrices.R \docType{methods} \name{vector2matrices,matrix-method} \alias{vector2matrices,matrix-method} \title{Tranform the vectors of weights to matrix structure} \usage{ \S4method{vector2matrices}{matrix}(object, net) } \arguments{ \item{object}{an nlpcaNet} \item{net}{the neural network} } \value{ weights in matrix structure } \description{ Tranform the vectors of weights to matrix structure } \author{ Henning Redestig } pcaMethods/man/vector2matrices-nlpcaNet-method.Rd0000644000175400017540000000074313556116437023104 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/vector2matrices.R \docType{methods} \name{vector2matrices,nlpcaNet-method} \alias{vector2matrices,nlpcaNet-method} \title{Tranform the vectors of weights to matrix structure} \usage{ \S4method{vector2matrices}{nlpcaNet}(object) } \arguments{ \item{object}{an nlpcaNet} } \value{ weights in matrix structure } \description{ Tranform the vectors of weights to matrix structure } \author{ Henning Redestig } pcaMethods/man/wasna-pcaRes-method.Rd0000644000175400017540000000144013556116437020545 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/methods-pcaRes.R \docType{methods} \name{wasna,pcaRes-method} \alias{wasna,pcaRes-method} \alias{wasna} \title{Get a matrix with indicating the elements that were missing in the input data. Convenient for estimating imputation performance.} \usage{ wasna(object, ...) } \arguments{ \item{object}{pcaRes object} \item{...}{Not used} } \value{ A matrix with logicals } \description{ Get a matrix with indicating the elements that were missing in the input data. Convenient for estimating imputation performance. } \examples{ data(metaboliteData) data(metaboliteDataComplete) result <- pca(metaboliteData, nPcs=2) plot(completeObs(result)[wasna(result)], metaboliteDataComplete[wasna(result)]) } \author{ Henning Redestig } pcaMethods/man/weightsAccount.Rd0000644000175400017540000000106013556116437017730 0ustar00biocbuildbiocbuild% Generated by roxygen2: do not edit by hand % Please edit documentation in R/AllClasses.R \name{weightsAccount} \alias{weightsAccount} \title{Create an object that holds the weights for nlpcaNet. Holds and sets weights in using an environment object.} \usage{ weightsAccount(w) } \arguments{ \item{w}{\code{matrix} -- New weights} } \value{ A weightsAccound with \code{set} and \code{current} functions. } \description{ Create an object that holds the weights for nlpcaNet. Holds and sets weights in using an environment object. } \author{ Henning Redestig } pcaMethods/src/0000755000175400017540000000000013556147214014466 5ustar00biocbuildbiocbuildpcaMethods/src/RcppExports.cpp0000644000175400017540000000102413556116437017463 0ustar00biocbuildbiocbuild#include using namespace Rcpp; // Nipals List Nipals(SEXP Mat, SEXP params); RcppExport SEXP pcaMethods_Nipals(SEXP MatSEXP, SEXP paramsSEXP) { BEGIN_RCPP SEXP __sexp_result; { Rcpp::RNGScope __rngScope; Rcpp::traits::input_parameter< SEXP >::type Mat(MatSEXP ); Rcpp::traits::input_parameter< SEXP >::type params(paramsSEXP ); List __result = Nipals(Mat, params); PROTECT(__sexp_result = Rcpp::wrap(__result)); } UNPROTECT(1); return __sexp_result; END_RCPP } pcaMethods/src/nipals.cpp0000644000175400017540000000731113556116437016465 0ustar00biocbuildbiocbuild#include #include #include using namespace std; using namespace Rcpp; double difference(vector& vec1, vector& vec2) { double diff = 0; double a; int len = vec1.size(); for(int i = 0; i < len; i++) { a = vec1[i] - vec2[i]; diff += a * a; } return(diff); } void norm(vector& vec) { double siz = 0; int len = vec.size(); for(int i = 0; i < len; i++) { siz += vec[i] * vec[i]; } siz = sqrt(siz); for(int i = 0; i < len; i++) { vec[i] = vec[i] / siz; } } // [[Rcpp::export]] List Nipals(SEXP Mat, SEXP params) { try{ bool cnt; int count = 0; double tsize; Rcpp::List rl = R_NilValue; Rcpp::List rparams(params); int maxSteps = Rcpp::as(rparams["maxSteps"]); double eps = Rcpp::as(rparams["threshold"]); int nPcs = Rcpp::as(rparams["nPcs"]); double varLimit = Rcpp::as(rparams["varLimit"]); Rcpp::NumericMatrix mat(Mat); Rcpp::NumericMatrix omat = Rcpp::clone( Mat ); int nr = mat.nrow(); int nc = mat.ncol(); Rcpp::NumericMatrix est_mat(nr, nc); Rcpp::NumericMatrix tt(nr, nPcs); Rcpp::NumericMatrix pp(nc, nPcs); vector r2cum; vector thold(nr); vector th(nr); vector phold(nc); vector ph(nc); double tss = 0; double sse = 0; int np = 0; double anotherPc = true; for (int r = 0; r < nr; r++) { for (int c = 0; c < nc; c++) { if(!ISNAN(mat(r,c))) { tss += mat(r,c) * mat(r,c); } } } while(anotherPc) { for(int r = 0; r < nr; r++) { th[r] = 0; if(!ISNAN(mat(r,0))) { th[r] = mat(r,0); } } cnt = true; count = 0; while(cnt) { count++; for(int c = 0; c < nc; c++) { ph[c] = 0; } tsize = 0; for(int r = 0; r < nr; r++) { tsize += th[r] * th[r]; } for(int r = 0; r < nr; r++) { double ti = th[r] / tsize; for(int c = 0; c < nc; c++) { if(!ISNAN(mat(r,c))) { ph[c] += mat(r,c) * ti; } } } norm(ph); thold = th; for(int r = 0; r < nr; r++) { th[r] = 0; for(int c = 0; c < nc; c++) { if(!ISNAN(mat(r,c))) { th[r] += mat(r,c) * ph[c]; } } } if(count > maxSteps) { throw 1; } if(difference(thold, th) <= eps) { cnt = false; } } //deflate mat sse = 0; double mathat = 0; double err = 0; for(int r = 0; r < nr; r++) { for(int c = 0; c < nc; c++) { if(!ISNAN(mat(r,c))) { mathat = th[r] * ph[c]; est_mat(r, c) += mathat; err = omat(r,c) - est_mat(r, c); sse += err * err; mat(r,c) -= mathat; } } } r2cum.push_back(1 - (sse / tss)); for(int r = 0; r < nr; r++) { tt(r,np) = th[r]; } for(int c = 0; c < nc; c++) { pp(c,np) = ph[c]; } if(fabs(varLimit - 1) > 1e-4) { if(r2cum[np] >= varLimit) { anotherPc = false; } } if (np + 1 >= nPcs){ anotherPc = false; } np++; } if(np != nPcs) { Rcpp::NumericMatrix ttt(nr, np); Rcpp::NumericMatrix ppp(nc, np); for(int r = 0; r < nr; r++) { for(int p = 0; p < np; p++) { ttt(r,p) = tt(r,p); } } for(int c = 0; c < nc; c++) { for(int p = 0; p < np; p++) { ppp(c,p) = pp(c,p); } } rl["scores"] = ttt; rl["loadings"] = ppp; } else { rl["scores"] = tt; rl["loadings"] = pp; } rl["R2cum"] = r2cum; return rl; }catch(int e) { if(e == 1) { ::Rf_error("Too many iterations, quitting"); }else { ::Rf_error("unknown error"); } } catch(std::exception& ex) { forward_exception_to_r(ex); } catch(...) { ::Rf_error("unknown error"); } return R_NilValue; } pcaMethods/vignettes/0000755000175400017540000000000013556147214015707 5ustar00biocbuildbiocbuildpcaMethods/vignettes/missingValues.Rnw0000644000175400017540000000565413556116437021245 0ustar00biocbuildbiocbuild\documentclass[a4paper]{article} %\VignetteIndexEntry{Missing value imputation} \usepackage{hyperref} \title{Imputing missing values using the pcaMethods package} \author{Wolfram Stacklies and Henning Redestig\\ CAS-MPG Partner Institute for Computational Biology (PICB)\\ Shanghai, P.R. China \\ and\\ Max Planck Institute for Molecular Plant Physiology\\ Potsdam, Germany\\ \url{http://bioinformatics.mpimp-golm.mpg.de/} } \date{\today} \begin{document} \setkeys{Gin}{width=1.0\textwidth} @ \maketitle \section{Missing value imputation} One application for missing value robust principal component analysis is that it effectively can be used to impute the missing values and thus obtain an estimated complete data set. The pcaMethods package was partly written with this application in mind. PCA is a way of creating a model of a matrix, $X$, by defining two parameter matrices, the scores, $T$, and the loadings, $P$, which together have less values than the original matrix but when multiplied with each other well reconstruct the original matrix. I.e.: $$X=1\times{}\bar{x} + TP' + E$$ where $E$ is the error matrix and $1\times{}\bar{x}$ denotes the original variable averages. Now if $X$ contains missing values but we still are able to get complete estimates of $P$ and $T$ than we can use: $$\hat{X}=1\times{}\bar{x} + TP'$$ as an estimate for $x_{i,j}$ if $x_{i,j}$ is missing. This is can be done as the following example illustrates. First we attach the metabolite data set with missing values. <>= library(pcaMethods) @ <<>>= data(metaboliteData) mD <- metaboliteData sum(is.na(mD)) @ Now we get the estimated data set by using PPCA and three principal components. <<>>= pc <- pca(mD, nPcs=3, method="ppca") imputed <- completeObs(pc) @ If we compare with the original values we see that the error is rather low. <<>>= data(metaboliteDataComplete) mdComp <- metaboliteDataComplete sum((mdComp[is.na(mD)] - imputed[is.na(mD)])^2) / sum(mdComp[is.na(mD)]^2) @ When using a different PCA algorithm, we get different performance. <<>>= imputedNipals <- completeObs(pca(mD, nPcs=3, method="nipals")) sum((mdComp[is.na(mD)] - imputedNipals[is.na(mD)])^2) / sum(mdComp[is.na(mD)]^2) @ If the data we are interested in was gene expression set of class 'ExpressionSet' we could simply do <<>>= library(Biobase) data(sample.ExpressionSet) exSet <- sample.ExpressionSet exSetNa <- exSet exprs(exSetNa)[sample(13000, 200)] <- NA lost <- is.na(exprs(exSetNa)) pc <- pca(exSetNa, nPcs=2, method="ppca") impExSet <- asExprSet(pc, exSetNa) sum((exprs(exSet)[lost] - exprs(impExSet)[lost])^2) / sum(exprs(exSet)[lost]^2) @ Different results will be obtained with different PCA algorithms. Which one to use depends on the general structure of the data set and the imputation performance can be estimated by cross-validation. Please see the 'introduction' vignette on further details on how to use the cross-validation capabilities of this package. \end{document} pcaMethods/vignettes/outliers.Rnw0000644000175400017540000001203113556116437020245 0ustar00biocbuildbiocbuild\documentclass[a4paper]{article} %\VignetteIndexEntry{Data with outliers} \usepackage{hyperref} \title{Handling of data containing outliers} \author{Wolfram Stacklies and Henning Redestig\\ CAS-MPG Partner Institute for Computational Biology (PICB)\\ Shanghai, P.R. China \\ and\\ Max Planck Institute for Molecular Plant Physiology\\ Potsdam, Germany\\ \url{http://bioinformatics.mpimp-golm.mpg.de/} } \date{\today} \begin{document} \setkeys{Gin}{width=1.0\textwidth} @ \maketitle \section{PCA robust to outliers} Away from often showing missing values, Microarray or Metabolite data are often corrupted with extreme values (outliers). Standard SVD is highly susceptible to outliers. In the extreme case, an individual data point, if sufficiently outlying, can draw even the leading principal component toward itself. This problem can be addressed by using a robust analysis method. Hereto we provide \texttt{robustSvd}, a singular value decomposition robust to outliers. \texttt{robustPca} is a PCA implementation that resembles the original \texttt{R} \texttt{prcomp} method, with the difference that it uses \texttt{robustSvd} instead of the standard \texttt{svd} function.\\ Robust SVD and its application to microarray data were proposed in \cite{hawkins01} and \cite{liu03}. The algorithm is based on the idea to use a sequential estimation of the eigenvalues and left and right eigenvectors that ignores missing values and is resistant to outliers. \\ The \texttt{robustSvd} script included here was contributed by Kevin Wright. Thanks a lot to him! \section{Outliers and missing value imputation} The problem of outliers is similar to the missing data problem in the sense that extreme values provide no or wrong information. They are generally artifacts of the experiment and provide no information about the underlying biological processes. \\ Most of the PCA methods coming with the package were not designed to be robust to outliers in the sense that they will converge to the standard PCA solution on a complete data set. Yet, an applicable solution is to remove obvious outliers from the data first (by setting them NA) and to then estimate the PCA solution on the incomplete data. This is likely to produce accurate results if the number of missing data does not exceed a certain amount, less than 10\% should be a good number. The following example illustrates the effect of outliers and the use of robust methods. First, we attach the complete metabolite data set and create 5\% outliers. We mean center the data before we create outliers because these large artificial outliers will strongly shift the original means. This would not allow for objective comparison between the differnt results obtained, e.g. when doing scatterplots. <>= library(pcaMethods) @ <<>>= data(metaboliteDataComplete) mdc <- scale(metaboliteDataComplete, center=TRUE, scale=FALSE) cond <- runif(length(mdc)) < 0.05 mdcOut <- mdc mdcOut[cond] <- 10 @ Then we calculate a PCA solution using standard SVD and robust SVD. <>= resSvd <- pca(mdc, method="svd", nPcs=5, center=FALSE) resSvdOut <- pca(mdcOut, method="svd", nPcs=5, center=FALSE) resRobSvd <- pca(mdcOut, method="robustPca", nPcs=5, center=FALSE) @ Now we use \texttt{PPCA} to estimate the PCA solution, but set the outliers NA before. <>= mdcNa <- mdc mdcNa[cond] <- NA resPPCA <- pca(mdcNa, method="ppca", nPcs=5, center=FALSE) @ To check the robustness to outliers we can just do a scatterplot comparing the results to the optimal PCA solution for the complete data set (which is \texttt{resSvd}). In Figure \ref{fig:svdPlot} we plot the estimated and original loadings against each other. \begin{figure}[!ht] \centering <>= par(mfrow=c(2,2)) plot(loadings(resSvd)[,1], loadings(resSvdOut)[,1], xlab="Loading 1 SVD", ylab="Loading 1 SVD with outliers") plot(loadings(resSvd)[,1], loadings(resRobSvd)[,1], xlab="Loading 1 SVD", ylab="Loading 1 robustSVD with outliers") plot(loadings(resSvd)[,1], loadings(resPPCA)[,1], xlab="Loading 1 SVD", ylab="Loading 1 PPCA with outliers=NA") plot(loadings(resRobSvd)[,1], loadings(resPPCA)[,1], xlab="Loading 1 robust SVD with outliers", ylab="Loading 1 svdImpute with outliers=NA") @ \caption{Figures show (from left to right): \newline Original PCA solution vs. solution on data with outliers; \newline Original PCA solution vs. robust PCA solution on data with outliers; \newline Original PCA solution vs. PPCA solution on data where outliers=NA; \newline Robust PCA solution vs. PPCA solution on data with outliers / outliers=NA. \label{fig:svdPlot} } \end{figure} \begin{thebibliography}{2006} \bibitem{hawkins01} Hawkins, D.M., Liu, L. and Young, S.S. {\sl Robust Singular Value Decomposition.} National Institute of Statistical Sciences, 2001, Tech Report 122. \bibitem{liu03} Liu, L., Hawkins, D.M., Ghosh, S. and Young, S.S. {\sl Robust singular value decomposition analysis of microarray data.} PNAS, 2003;100:13167--13172. \end{thebibliography} \end{document} pcaMethods/vignettes/pcaMethods.Rnw0000644000175400017540000005554213556116437020504 0ustar00biocbuildbiocbuild\documentclass[a4paper]{article} %\VignetteIndexEntry{Introduction} \usepackage{hyperref} \title{The pcaMethods Package} \author{Wolfram Stacklies and Henning Redestig\\ CAS-MPG Partner Institute for Computational Biology (PICB)\\ Shanghai, P.R. China \\ and\\ Max Planck Institute for Molecular Plant Physiology\\ Potsdam, Germany\\ \url{http://bioinformatics.mpimp-golm.mpg.de/} } \date{\today} \begin{document} \setkeys{Gin}{width=1.0\textwidth} @ \maketitle \section*{Overview} The \texttt{pcaMethods} package \cite{stacklies07} provides a set of different PCA implementations, together with tools for cross validation and visualisation of the results. The methods basically allow to perform PCA on incomplete data and thus may also be used for missing value estimation. When doing PCA one assumes that the data is restricted to a subspace of lower dimensionality, e.g. correlation patterns between jointly regulated genes. PCA aims to extract these structures thereby filtering noise out. If only the most significant loadings (eigenvectors, also referred to as principal components) are used for projection this can be written as: \begin{equation} X = 1\times{}\bar{x}^T + TP^T + V \end{equation} Where the term $1\times{}\bar{x}^T$ represents the original variable averages, $X$ denotes the observations, $T={t_1, t_2,\ldots,t_k}$ the latent variables or scores, $P={p_1, p_2,\ldots,p_k}$ the transformation matrix (consisting of the most significant eigenvectors of the covariance matrix) and $V$ are the residuals. Missing values may be estimated by projecting the scores back into the original space using $\hat{X} = 1\times{}\bar{x}^T + TP^T$. Optimally, this produces an estimate of the missing data based on the underlying correlation structure, thereby ignoring noise. This will only produce reasonable results if the residuals $V$ are sufficiently small, implying that most of the important information is captured by the first $k$ components. In order to calculate the transformation matrix $P$ one needs to determine the covariance matrix between variables or alternatively calculate $P$ directly via SVD. In both cases, this can only be done on complete matrices. However, an approximation may be obtained by use of different regression methods. The PCA methods provided in this package implement algorithms to accurately estimate the PCA solution on incomplete data. Although the focus of this package is clearly to provide a collection of PCA methods we also provide a cluster based method for missing value imputation. This allows to better rate and compare the results. \section{Algorithms} All methods return a common class called \texttt{pcaRes} as a container for the results. This guarantees maximum flexibility for the user. A wrapper function called \texttt{pca()} is provided that receives the desired type of pca as a string. \subsection*{svdPca} This is a wrapper function for $R's$ standard \texttt{prcomp} function. It delivers the results as a \texttt{pcaRes} object for compatibility with the rest of the package. \subsection*{svdImpute} This implements the SVDimpute algorithm as proposed by Troyanskaya et~al \cite{troyanskaya01}. The idea behind the algorithm is to estimate the missing values as a linear combination of the $k$ most significant eigengenes\footnote{The term ``eigengenes'' denotes the loadings when PCA was applied considering variables (here the genes) as observations.}. The algorithm works iteratively until the change in the estimated solution falls below a certain threshold. Each step the eigengenes of the current estimate are calculated and used to determine a new estimate. An optimal linear combination is found by regressing an incomplete variable against the $k$ most significant eigengenes. If the value at position $j$ is missing, the $j^{th}$ value of the eigengenes is not used when determining the regression coefficients.\\ SVDimpute seems to be tolerant to relatively high amount of missing data (> 10\%). \subsection*{Probabilistic PCA (ppca)} Probabilistic PCA combines an EM approach for PCA with a probabilistic model. The EM approach is based on the assumption that the latent variables as well as the noise are normal distributed. In standard PCA data which is far from the training set but close to the principal subspace may have the same reconstruction error, see Figure \ref{fig:pcaSubspace} for explanation. <>= library(pcaMethods) x <- c(-4,7); y <- c(-3,4) distX <- rnorm(100, sd=0.3)*3 distY <- rnorm(100, sd=0.3) + distX * 0.3 mat <- cbind(distX, distY) res <- pca(mat, nPcs=2, method="svd", center=F) loading <- loadings(res)[1,] grad <- loading[2] / loading[1] if (grad < 0) grad <- grad * -1 lx <- c(-4,7) ly <- c(grad * -4, grad * 7) @ \begin{figure} \centering <>= par(mar=c(2, 3, 2, 2)) plot(x,y, type="n", xlab="", ylab="") abline(v=0, col="dark gray", lwd = 2); abline(h=0, col = "dark gray", lwd = 2) points(distX, distY, type = 'p', col = "blue") lines(lx,ly, lwd = 2) points(-1, -1 * grad + 0.5, pch = 19, col = "red", lwd=4) points(6, 6 * grad + 0.5, pch = 19, col = "red", lwd=4) @ \caption{Normal distributed data with the first loading plotted in black. The two red points have the same reconstruction error because PCA does not define a density model. Thus the only measure of how well new data fits the model is the distance from the principal subspace. Data points far from the bulk of data but still close to the principal subspace will have a low reconstruction error. \label{fig:pcaSubspace}} \end{figure} PPCA defines a likelihood function such that the likelihood for data far from the training set is much lower, even if they are close to the principal subspace. This allows to improve the estimation accuracy.\\ PPCA is tolerant to amounts of missing values between 10\% to 15\%. If more data is missing the algorithm is likely not to converge to a reasonable solution. The method was implemented after the draft ``\texttt{EM Algorithms for PCA and Sensible PCA}'' written by Sam Roweis and after the Matlab \texttt{ppca} script implemented by \emph{Jakob Verbeek}\footnote{\url{http://lear.inrialpes.fr/~verbeek/}}. Please check also the PPCA help file. \subsection*{Bayesian PCA (bpca)} Similar to probabilistic PCA, Bayesian PCA uses an EM approach together with a Bayesian model to calculate the likelihood for a reconstructed value.\\ The algorithm seems to be tolerant to relatively high amounts of missing data (> 10\%). Scores and loadings obtained with Bayesian PCA slightly differ from those obtained with conventional PCA. This is because BPCA was developed especially for missing value estimation and is based on a variational Bayesian framework (VBF), with automatic relevance determination (ARD). In BPCA, ARD leads to a different scaling of the scores, loadings and eigenvalues when compared to standard PCA or PPCA. The algorithm does not force orthogonality between loadings. However, the authors of the BPCA paper found that including an orthogonality criterion made the predictions worse. They also state that the difference between ``real'' and predicted Eigenvalues becomes larger when the number of observation is smaller, because it reflects the lack of information to accurately determine true loadings from the limited and noisy data. As a result, weights of factors to predict missing values are not the same as with conventional PCA, but the missing value estimation is improved. BPCA was proposed by Oba et~al \cite{oba03}. The method available in this package is a port of the \texttt{bpca} Matlab script also provided by the authors\footnote{ \url{http://hawaii.aist-nara.ac.jp/\%7Eshige-o/tools/}}. \subsection*{Inverse non-linear PCA (NLPCA)} NLPCA \cite{scholz05} is especially suitable for data from experiments where the studied response is non-linear. Examples of such experiments are ubiquitous in biology -- enzyme kinetics are inherently non-linear as are gene expression responses influenced by the cell cycle or diurnal oscillations. NLPCA is based on training an auto-associative neural network composed of a component layer which serves as the ``bottle-neck'', a hidden non-linear layer and an output layer corresponding to the reconstructed data. The loadings can be seen as hidden in the network. Missing values in the training data are simply ignored when calculating the error during back-propagation. Thus NLPCA can be used to impute missing values in the same way as for conventional PCA. The only difference is that the loadings $P$ are now represented by a neural network.\\ A shortcoming of the current implementation is that there is no reasonable stop criterion. The quality of the estimated solution depends on the number of iterations. This should in most cases be somewhat between 500 and 1500. We recommend to use \texttt{kEstimate} or \texttt{kEstimateFast} to determine this parameter. \subsection*{Nipals PCA} Nipals (Nonlinear Estimation by Iterative Partial Least Squares) \cite{wold66} is an algorithm at the root of PLS regression which can execute PCA with missing values by simply leaving those out from the appropriate inner products. It is tolerant to small amounts (generally not more than 5\%) of missing data. \subsection{Local least squares (LLS) imputation} The package provides an algorithm called \texttt{llsImpute} for missing value estimation based on a linear combination of the $k$ nearest neighbours of an incomplete variable (in Microarray experiments normally a gene). The distance between variables is defined as the absolute value of the Pearson, Spearman or Kendall correlation coefficient. The optimal linear combination is found by solving a local least squares problem as described in \cite{kim05}. In tests performed in the cited paper the llsImpute algorithm is able to outperform knnImpute\cite{troyanskaya01} and competes well with BPCA. In the current implementation two slightly different ways for missing value estimation are provided. The first one is to restrict the neighbour searching to the subset of complete variables. This is preferable when the number of incomplete variables is relatively small. The second way is to consider all variables as candidates. Here, missing values are initially replaced by the columns wise mean. The method then iterates, using the current estimate as input for the LLS regression until the change between new and old estimate falls below a certain threshold (0.001). \section{Getting started} \paragraph{Installing the package.} To install the package first download the appropriate file for your platform from the Bioconductor website (\url{http://www.bioconductor.org/}). For Windows, start \texttt{R} and select the \texttt{Packages} menu, then \texttt{Install package from local zip file}. Find and highlight the location of the zip file and click on \texttt{open}. For Linux/Unix, use the usual command \texttt{R CMD INSTALL} or set the option \texttt{CRAN} to your nearest mirror site and use the command \texttt{install.packages} from within an \texttt{R} session. \paragraph{Loading the package:} To load the \texttt{pcaMethods} package in your \texttt{R} session, type \texttt{library(pcaMethods)}. \paragraph{Help files:} Detailed information on \texttt{pcaMethods} package functions can be obtained from the help files. For example, to get a description of \texttt{bpca} type \texttt{help("bpca")}. \paragraph{Sample data:} Two sample data sets are coming with the package. \texttt{metaboliteDataComplete} contains a complete subset from a larger metabolite data set. \texttt{metaboliteData} is the same data set but with 10 \% values removed from an equal distribution. \section{Some examples} <>= library(lattice) library(pcaMethods) @ To load the package and the two sample data sets type: <>= library(pcaMethods) data(metaboliteData) data(metaboliteDataComplete) @ Now centre the data <<>>= md <- prep(metaboliteData, scale="none", center=TRUE) mdC <- prep(metaboliteDataComplete, scale="none", center=TRUE) @ Run SVD pca, PPCA, BPCA, SVDimpute and nipalsPCA on the data, using the \texttt{pca()} wrapper function. The result is always a \texttt{pcaRes} object. <>= resPCA <- pca(mdC, method="svd", center=FALSE, nPcs=5) resPPCA <- pca(md, method="ppca", center=FALSE, nPcs=5) resBPCA <- pca(md, method="bpca", center=FALSE, nPcs=5) resSVDI <- pca(md, method="svdImpute", center=FALSE, nPcs=5) resNipals <- pca(md, method="nipals", center=FALSE, nPcs=5) resNLPCA <- pca(md, method="nlpca", center=FALSE, nPcs=5, maxSteps=300) @ Figure \ref{fig:eigenvalues} shows a plot of the eigenvalue structure (\texttt{sDev(pcaRes)}). If most of the variance is captured with few loadings PCA is likely to produce good missing value estimation results. For the sample data all methods show similar eigenvalues. One can also see that most of the variance is already captured by the first loading, thus estimation is likely to work fine on this data. For BPCA, the eigenvalues are scaled differently for reasons discussed above, see Figure \ref{fig:loadingBPCA}. The order of the loadings remains the same. \begin{figure} \centering <>= sDevs <- cbind(sDev(resPCA), sDev(resPPCA), sDev(resBPCA), sDev(resSVDI), sDev(resNipals), sDev(resNLPCA)) matplot(sDevs, type = 'l', xlab="Eigenvalues", ylab="Standard deviation of PC", lwd=3) legend(x="topright", legend=c("PCA", "PPCA", "BPCA", "SVDimpute","Nipals PCA","NLPCA"), lty=1:6, col=1:6, lwd=3) @ \caption{Eigenvalue structure as obtained with different methods\label{fig:eigenvalues}} \end{figure} To get an impression of the correctness of the estimation it is a good idea to plot the scores / loadings obtained with classical PCA and one of the probabilistic methods against each other. This of course requires a complete data set from which data is randomly removed. Figure \ref{fig:loadingBPCA} shows this for BPCA on the sample data. \begin{figure} \centering <>= par(mfrow=c(1,2)) plot(loadings(resBPCA)[,1], loadings(resPCA)[,1], xlab="BPCA", ylab="classic PCA", main = "Loading 1") plot(loadings(resBPCA)[,2], loadings(resPCA)[,2], xlab="BPCA", ylab="classic PCA", main = "Loading 2") @ \caption{Loading 1 and 2 calculated with BPCA plotted against those calculated with standard PCA. \label{fig:loadingBPCA}} \end{figure} \section{Cross validation} \texttt{Q2} is the goodness measure used for internal cross validation. This allows to estimate the level of structure in a data set and to optimise the choice of number of loadings. Cross validation is performed by removing random elements of the data matrix, then estimating these using the PCA algorithm of choice and then calculating $Q^2$ accordingly. At the moment, cross-validation can only be performed with algorithms that allow missing values (i.e. not SVD). Missing value independent cross-validation is scheduled for implementation in later versions. $Q^2$ is defined as following for the mean centered data (and possibly scaled) matrix $X$. $$\mathrm{SSX}=\sum (x_{ij})^2$$ $$\mathrm{PRESS}=\sum (x_{ij} - \hat{x}_{ij})^2$$ $$Q^2=1 - \mathrm{PRESS}/\mathrm{SSX}$$ The maximum value for $Q^2$ is thus 1 which means that all variance in $X$ is represented in the predictions; $X=\hat{X}$. <>= q2SVDI <- Q2(resSVDI, mdC, fold=10) q2PPCA <- Q2(resPPCA, mdC, fold=10) @ <>= # PPCA does not converge / misestimate a value in very rare cases. # This is a workaround to avoid that such a case will break the # diagram displayed in the vignette. # From the 2.0 release of bioconductor on, the convergence threshold # for PPCA was lowert to 1e-5, this should make the method much more # stable. So this workaround might be obsolete now... # [nope it is not, ppca is unstable] while( sum((abs(q2PPCA)) > 1) >= 1 ) { q2PPCA <- Q2(resPPCA, mdC, fold=10) } @ \begin{figure}[!ht] \centering <>= q2 <- data.frame(Q2=c(drop(q2PPCA), drop(q2SVDI)), method=c("PPCA", "SVD-Impute")[gl(2, 5)], PC=rep(1:5, 2)) print(xyplot(Q2~PC|method, q2, ylab=expression(Q^2), type="h", lwd=4)) @ \caption{Boxplot of the \texttt{Q2} results for BPCA, Nipals PCA, SVDimpute and PPCA. PPCA and SVDimpute both deliver better results than BPCA and Nipals in this example.\label{fig:Q2}} \end{figure} The second method called \texttt{kEstimate} uses cross validation to estimate the optimal number of loadings for missing value estimation. The \texttt{NRMSEP} (normalised root mean square error of prediction) \cite{feten05} or Q2 can be used to define the average error of prediction. The NRMSEP normalises the square difference between real and estimated values for a certain variable by the variance within this variable. The idea behind this normalisation is that the error of prediction will automatically be higher if the variance is higher. The \texttt{NRMSEP} for mean imputation is $\sqrt{\frac{nObs}{nObs - 1}}$ when cross validation is used, where $nObs$ is the number of observations. The exact definition is: \begin{equation} NRMSEP_k = \sqrt{\frac{1}{g} \sum_{j \in G} \frac{\sum_{i \in O_j} (x_{ij} - \hat{x}_{ijk})^2}{o_j s_{x_j}^2}} \end{equation} where $s^2_{x_j} = \sum_{i=1}^n (x_{ij} - \overline{x}_j)^2 / (n - 1)$, this is the variance within a certain variable. Further, $G$ denotes the set of incomplete variables, $g$ is the number of incomplete varialbes. $O_j$ is the set of missing observations in variable $j$ and $o_j$ is the number of missing observations in variable $j$. $\hat{x}_{ijk}$ stands for the estimate of value $i$ of variable $j$ using $k$ loadings. See Figure \ref{fig:kEstimate} for an example. The NRMSEP should be the error measure of choice. But if the number of observations is small, the variance within a certain variable may become and unstable criterion. If so or if variance scaling was applied we recommend to use Q2 instead. <>= errEsti <- kEstimate(md, method = "ppca", evalPcs=1:5, nruncv=1, em="nrmsep") @ \begin{figure}[!ht] \centering \begin{minipage}[c]{0.6\textwidth} \centering <>= barplot(drop(errEsti$eError), xlab="Loadings", ylab="NRMSEP (Single iteration)") @ \end{minipage} \begin{minipage}[c]{0.3\textwidth} \caption{Boxplot showing the \texttt{NRMSEP} versus the number of loadings. In this example only 1 iteration of the whole cross validation were performed. It is normally advisable to do more than just one iteration. \label{fig:kEstimate}} \end{minipage} \end{figure} \texttt{kEstimate} also provides information about the estimation error for individual variables. The $Q^2$ distance or the NRMSEP are calculated separately for each variable. See the manpage for \texttt{kEstimate} and \texttt{kEstimateFast} for details. Plotting the variable - wise results gives information about for which variables missing value estimation makes sense, and for which no imputation or mean imputation is preferable, see Figure \ref{fig:variableWiseError}. If you are not interested in variable - wise information we recommend to use the faster \texttt{kEstimateFast} instead. \begin{figure}[!ht] \centering \begin{minipage}[c]{0.6\textwidth} \centering <>= barplot(drop(errEsti$variableWiseError[, which(errEsti$evalPcs == errEsti$bestNPcs)]), xlab="Incomplete variable Index", ylab="NRMSEP") @ \end{minipage} \begin{minipage}[c]{0.3\textwidth} \caption{Boxplot showing the \texttt{NRMSEP} for all incomplete variables in the data set. For the first 7 variables missing value imputation does not seem to make too much sense. \label{fig:variableWiseError}} \end{minipage} \end{figure} \newpage \section{Visualisation of the results} \subsection{Quick scores and loadings plot} Some methods for display of scores and loadings are also provided. The function \texttt{slplot()} aims to be a simple way to quickly visualise scores and loadings in an intuitive way, see Figure \ref{fig:slplot}. Barplots are provided when plotting only one PC and colours can be specified differently for the scores and loadings plots. For a more specific scatter plot it is however recommended to access scores and loadings slots and define own plot functions. \begin{figure}[!h] \centering <>= slplot(resPCA) @ \caption{\texttt{slplot} for scores and loadings obtained with classical SVD based PCA. \label{fig:slplot}} \end{figure} \noindent Another method called \texttt{plotPcs()} allows to visualise many PCs plotted against each other, see Figure \ref{fig:plotPcs}. \begin{figure}[!ht] \centering <>= plotPcs(resPPCA, pc=1:3, type="score") @ \caption{A plot of score 1:3 for PPCA created with \texttt{plotPcs()} \label{fig:plotPcs}} \end{figure} \subsection{Using ggplot2} For using ggplot, the scores and loadings should best be added to a data frame that add other relevant descriptive factors. For example, after doing PCA on the Iris dataset, we may add the scores back to the original data frame and use ggplot to visualise, see Figure \ref{fig:ggplot}. \begin{figure}[!ht] \centering <>= pc <- pca(iris) irdf <- merge(iris, scores(pc), by=0) library(ggplot2) ggplot(irdf, aes(PC1, PC2, colour=Species)) + geom_point() + stat_ellipse() @ \caption{Score plot using ggplot2} \label{fig:ggplot} \end{figure} \cleardoublepage \begin{thebibliography}{2006} \bibitem{stacklies07} Stacklies W., Redestig H., Scholz M., and Walther D., and Selbig J. {\sl pcaMethods -- a Bioconductor package providing PCA methods for incomplete data} Bioinformatics. 2007, 23, 1164-1167. {\sl Non-linear PCA: a missing data approach.} Bioinformatics. 2005, 21, 3887-3895. \bibitem{scholz05} Scholz, M. , Kaplan, F., Guy, C.L., Kopka, J. and Selbig, J. {\sl Non-linear pca: a missing data approach.} Bioinformatics. 2005, 21, 3887-3895. \bibitem{troyanskaya01} Troyanskaya O. and Cantor M. and Sherlock G. and Brown P. and Hastie T. and Tibshirani R. and Botstein D. and Altman RB. {\sl Missing value estimation methods for DNA microarrays.} Bioinformatics. 2001 Jun;17(6):520-525. \bibitem{feten05} Feten G. and Almoy T. and Aastveit A.H. {\sl Prediction of Missing Values in Microarray and Use of Mixed Models to Evaluate the Predictors.}, Stat. Appl. Genet. Mol. Biol. 2005;4(1):Article 10 \bibitem{oba03} Oba S. and Sato MA. and Takemasa I. and Monden M. and Matsubara K. and Ishii S. {\sl A Bayesian missing value estimation method for gene expression profile data.} Bioinformatics. 2003 Nov 1;19(16):2088-96. \bibitem{wold66} Wold H. {Estimation of principal components and related models by iterative least squares.} In Multivariate Analysis (Ed. P.R. Krishnaiah), Academic Press, NY, 391-420. \bibitem{kim05} Kim H. and Golub G.H. and Park H. {\sl Missing value estimation for DNA microarray gene expression data: local least squares imputation} Bioinformatics. 2005 21(2) :187-198 \end{thebibliography} \end{document}