支持向量机源码,可在www.csie.ntu.edu.tw/~cjlin/libsvm/下载到最新版本,该版本是2013年4月更新的,3.17版。
压缩包里面有源代码和文档。
以下摘自前述网站:IntroductionLIBSVMisanintegratedsoftwareforsupportvectorclassification,(C-SVC,nu-SVC),regression(epsilon-SVR,nu-SVR)anddistributionestimation(one-classSVM).Itsupportsmulti-classclassification.Sinceversion2.8,itimplementsanSMO-typealgorithmproposedinthispaper:R.-E.Fan,P.-H.Chen,andC.-J.Lin.WorkingsetselectionusingsecondorderinformationfortrainingSVM.JournalofMachineLearningResearch6,1889-1918,2005.Youcanalsofindapseudocodethere.(howtociteLIBSVM)OurgoalistohelpusersfromotherfieldstoeasilyuseSVMasatool.LIBSVMprovidesasimpleinterfacewhereuserscaneasilylinkitwiththeirownprograms.MainfeaturesofLIBSVMincludeDifferentSVMformulationsEfficientmulti-classclassificationCrossvalidationformodelselectionProbabilityestimatesVariouskernels(includingprecomputedkernelmatrix)WeightedSVMforunbalanceddataBothC++andJavasourcesGUIdemonstratingSVMclassificationandregressionPython,R,MATLAB,Perl,Ruby,Weka,CommonLISP,CLISP,Haskell,OCaml,LabVIEW,andPHPinterfaces.C#.NETcodeandCUDAextensionisavailable.It'salsoincludedinsomedataminingenvironments:RapidMiner,PCP,andLIONsolver.Automaticmodelselectionwhichcangeneratecontourofcrossvaliationaccuracy.
1