资源描述：
《Kernel Methods for Pattern Analysis%0A.pdf》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、ThispageintentionallyleftblankKernelMethodsforPatternAnalysisPatternAnalysisistheprocessoffindinggeneralrelationsinasetofdata,andformsthecoreofmanydisciplines,fromneuralnetworkstoso-calledsyn-tacticalpatternrecognition,fromstatisticalpatternrecognitiontomachi
2、nelearninganddatamining.Applicationsofpatternanalysisrangefrombioin-formaticstodocumentretrieval.Thekernelmethodologydescribedhereprovidesapowerfulandunifiedframeworkforallofthesedisciplines,motivatingalgorithmsthatcanactongeneraltypesofdata(e.g.strings,vecto
3、rs,text,etc.)andlookforgeneraltypesofrelations(e.g.rankings,classifications,regressions,clusters,etc.).Thisbookfulfilstwomajorroles.Firstlyitprovidespractitionerswithalargetoolkitofalgorithms,kernelsandsolutionsreadytobeimplemented,manygivenasMatlabcodesuitabl
4、eformanypatternanalysistasksinfieldssuchasbioinformatics,textanalysis,andimageanalysis.Secondlyitfurnishesstudentsandresearcherswithaneasyintroductiontotherapidlyexpandingfieldofkernel-basedpatternanalysis,demonstratingwithexampleshowtohandcraftanalgorithmorak
5、ernelforanewspecificapplication,whilecoveringtherequiredconceptualandmathematicaltoolsnecessarytodoso.Thebookisinthreeparts.Thefirstprovidestheconceptualfoundationsofthefield,bothbygivinganextendedintroductoryexampleandbycov-eringthemaintheoreticalunderpinnings
6、oftheapproach.Thesecondpartcontainsanumberofkernel-basedalgorithms,fromthesimplesttosophis-ticatedsystemssuchaskernelpartialleastsquares,canonicalcorrelationanalysis,supportvectormachines,principalcomponentsanalysis,etc.Thefinalpartdescribesanumberofkernelfun
7、ctions,frombasicexamplestoadvancedrecursivekernels,kernelsderivedfromgenerativemodelssuchasHMMsandstringmatchingkernelsbasedondynamicprogramming,aswellasspecialkernelsdesignedtohandletextdocuments.Allthoseinvolvedinpatternrecognition,machinelearning,neuralne
8、t-worksandtheirapplications,fromcomputationalbiologytotextanalysiswillwelcomethisaccount.KernelMethodsforPatternAnalysisJohnShawe-TaylorUniversityofSouthamptonNelloCristianiniUniversityofCalifor
