資源描述：
《多類支持向量機(jī)算法的研究和改進(jìn)》由會員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、上海交通大學(xué)碩士學(xué)位論文多類支持向量機(jī)算法的研究和改進(jìn)姓名：嚴(yán)慧敏申請學(xué)位級別：碩士專業(yè)：計算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：謝康林20060101上海交通大學(xué)碩士論文1)首先轉(zhuǎn)換這兩種方案對應(yīng)的最優(yōu)化問題到它們的對偶形式從而使用更為方便的數(shù)值算法來進(jìn)行求解并且我們在實(shí)現(xiàn)這兩種方案的時候使用了適合大規(guī)模數(shù)據(jù)的算法因此可以支持大規(guī)模的數(shù)據(jù)運(yùn)行2)接著通過實(shí)驗(yàn)對這兩種方案的有效性進(jìn)行了驗(yàn)證實(shí)驗(yàn)結(jié)果顯示我們提出的改進(jìn)方案很有效關(guān)鍵詞支持向量機(jī)多分類問題多類支持向量機(jī)大規(guī)模優(yōu)化算法II上海交通大學(xué)碩士論文MULTI-CLASSSUPPORTVE
2、CTORMACHINES’STUDYANDIMPROVEMENTABSTRACTAsthemostsuccessfulmachinelearningmethod,SupportVectorMachinehasmadealotofgoodapplications,includingtextclassification,hand-writtencharactersrecognition,facerecognitionetc.Thebiggestdiscriminationfromothermachineslearningmeth
3、odsisthatSVMiscorrespondingtoseveralprinciplesinstatisticallearningtheory,suchasstructureriskminimization.AnditcouldtheoreticallyprovedthattheexpectedriskofSVMhasanupperbound.SVMisreallysuccessfulexceptforonepoint:itisbinaryinnature.Butinreallife,multi-classproblem
4、isprevailing.AndSVM'sapplicationinmulti-classproblemstillhasalongwaytogo..TheexistingmethodsforSVMtosolvemulti-classproblemgoesintwodirections:firstistoconvertmulti-classtoseveralbinary;secondistherealmulti-classSVM,thatis,consideringallthedataatonce.Inthisthesis,w
5、ehavestudiedtheprinciplesandimplementationmethods,andalsoimprovedthealgorithm.Wehaveproposedtwoimprovementmethods,respectivelyfordifferenttargetsinmulti-classSVM:costfactorandsubproblem.Theformerone'smainideaistoconsidertherelationshipbetweenclasses,forexample,thed
6、istance,andcorporatetherelationshipintooriginalalgorithm.Thelatter'smainideatobalanceeverysubprobleminmulti-classSVM,insteadofjustaddingallthesubproblemstofindasolution.Twomethodscomefromtwopointsofview,buttheyallshowoneimportantidea:togetaglobaloptimalsolutions..A
7、ftertheproposal,wehavedonethefollowingwork:1)Firstweconvertthesetwoproblemstotheirdualform,soastoutilizeeasiernumericalmethodtosolvetheproblem,andwealsousealgorithmwhichissuitableforlarge-scaledata;thereforetheimplementationcouldrunlarge-scaledataset,whichisveryimp
8、ortantnowadayswithsomuchdata.III上海交通大學(xué)碩士論文2)Secondwevalidatethesetwomethodsbyexperiments.Andtheexperimentresultsshowourmethodsareveryeffective.Ke
