SVM Tutorial.doc
上传人:sy****28 上传时间:2024-09-12 格式:DOC 页数:12 大小:1.5MB 金币:16 举报 版权申诉
预览加载中,请您耐心等待几秒...

SVM Tutorial.doc

SVMTutorial.doc

预览

免费试读已结束,剩余 2 页请下载文档后查看

16 金币

下载此文档

如果您无法下载资料,请参考说明:

1、部分资料下载需要金币,请确保您的账户上有足够的金币

2、已购买过的文档,再次下载不重复扣费

3、资料包下载后请先用软件解压,在使用对应软件打开

PAGE\*MERGEFORMAT12|SVMTutorialTutorialonSupportVectorMachine(SVM)VikramadityaJakkula,SchoolofEECS,WashingtonStateUniversity,Pullman99164.Abstract:InthistutorialwepresentabriefintroductiontoSVM,andwediscussaboutSVMfrompublishedpapers,workshopmaterials&materialcollectedfrombooksandmaterialavailableonlineontheWorldWideWeb.InthebeginningwetrytodefineSVMandtrytotalkaswhySVM,withabriefoverviewofstatisticallearningtheory.ThemathematicalformulationofSVMispresented,andtheoryfortheimplementationofSVMisbrieflydiscussed.FinallysomeconclusionsonSVMandapplicationareasareincluded.SupportVectorMachines(SVMs)arecompetingwithNeuralNetworksastoolsforsolvingpatternrecognitionproblems.ThistutorialassumesyouarefamiliarwithconceptsofLinearAlgebra,realanalysisandalsounderstandtheworkingofneuralnetworksandhavesomebackgroundinAI.IntroductionMachineLearningisconsideredasasubfieldofArtificialIntelligenceanditisconcernedwiththedevelopmentoftechniquesandmethodswhichenablethecomputertolearn.Insimpletermsdevelopmentofalgorithmswhichenablethemachinetolearnandperformtasksandactivities.Machinelearningoverlapswithstatisticsinmanyways.Overtheperiodoftimemanytechniquesandmethodologiesweredevelopedformachinelearningtasks[1].SupportVectorMachine(SVM)wasfirstheardin1992,introducedbyBoser,Guyon,andVapnikinCOLT-92.Supportvectormachines(SVMs)areasetofrelatedsupervisedlearningmethodsusedforclassificationandregression[1].Theybelongtoafamilyofgeneralizedlinearclassifiers.Inanotherterms,SupportVectorMachine(SVM)isaclassificationandregressionpredictiontoolthatusesmachinelearningtheorytomaximizepredictiveaccuracywhileautomaticallyavoidingover-fittothedata.SupportVectormachinescanbedefinedassystemswhichusehypothesisspaceofalinearfunctionsinahighdimensionalfeaturespace,trainedwithalearningalgorithmfromoptimizationtheorythatimplementsalearningbiasderivedfromstatisticallearningtheory.SupportvectormachinewasinitiallypopularwiththeNIPScommunityandnowisanactivepartofthemachinelearningresearcharoundthewor