elements of statistical learning (2008)

《elements of statistical learning (2008)》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、ThisispagevPrinter:OpaquethisToourparents:ValerieandPatrickHastieVeraandSamiTibshiraniFlorenceandHarryFriedmanandtoourfamilies:Samantha,Timothy,andLyndaCharlie,Ryan,Julie,andCherylMelanie,Dora,Monika,andIldikoviThisispageviiPrinter:OpaquethisPrefacetotheSecondEdition

2、InGodwetrust,allothersbringdata.–WilliamEdwardsDeming(1900-1993)1WehavebeengratifiedbythepopularityofthefirsteditionofTheElementsofStatisticalLearning.This,alongwiththefastpaceofresearchinthestatisticallearningfield,motivatedustoupdateourbookwithasecondedition.Wehaveadd

3、edfournewchaptersandupdatedsomeoftheexistingchapters.Becausemanyreadersarefamiliarwiththelayoutofthefirstedition,wehavetriedtochangeitaslittleaspossible.Hereisasummaryofthemainchanges:1OntheWeb,thisquotehasbeenwidelyattributedtobothDemingandRobertW.Hayden;howeverProfe

4、ssorHaydentoldusthathecanclaimnocreditforthisquote,andironicallywecouldfindno“data”confirmingthatDemingactuallysaidthis.viiiPrefacetotheSecondEditionChapterWhat’snew1.Introduction2.OverviewofSupervisedLearning3.LinearMethodsforRegressionLARalgorithmandgeneralizationsof

5、thelasso4.LinearMethodsforClassificationLassopathforlogisticregression5.BasisExpansionsandRegulariza-AdditionalillustrationsofRKHStion6.KernelSmoothingMethods7.ModelAssessmentandSelectionStrengthsandpitfallsofcross-validation8.ModelInferenceandAveraging9.AdditiveModel

6、s,Trees,andRelatedMethods10.BoostingandAdditiveTreesNewexamplefromecology;somematerialsplitofftoChapter16.11.NeuralNetworksBayesianneuralnetsandtheNIPS2003challenge12.SupportVectorMachinesandPathalgorithmforSVMclassifierFlexibleDiscriminants13.PrototypeMethodsandNeares

7、t-Neighbors14.UnsupervisedLearningSpectralclustering,kernelPCA,sparsePCA,non-negativematrixfactorizationarchetypalanalysis,nonlineardimensionreduction,Googlepagerankalgorithm,adirectapproachtoICA15.RandomForestsNew16.EnsembleLearningNew17.UndirectedGraphicalModelsNew

8、18.High-DimensionalProblemsNewSomefurthernotes:•Ourfirsteditionwasunfriendlytocolorblindreaders;inparticular,wetendedtofavorred/gree