基于條件互信息的集成學(xué)習(xí)的研究與應(yīng)用

1、中國海洋大學(xué)碩士學(xué)位論文基于條件互信息的集成學(xué)習(xí)的研究與應(yīng)用姓名：李濤申請學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：丁香乾20090601ResearchandAppIicationofEnsembIeLearningBasedonConditionaIMutuallnformationAbstractEnsemblelearningisthefirstofthefourmajorresearchdirectionsAlongwith

2、involvinginresearchingitbymoreandmoreresearchers，thetwomainalgorithmsBaggingandBoostinghavebeenstudiedinadeepwaySosomeothermethodsshouldbefoundtodevotetothedevelopmentofensemblelearningSelectiveensemblelearningisaspecial

3、paradigmofensemblelearningItensemblesdistinctandstrongestgeneralizationindividuallearnerswithoptimizedmethod，andgainsthebetterperformancethanthatofintegratingallindividuallearnersCurrentlyensemblelearninghasbeenusedinman

4、yfieldssuchasplanetaryexploration，seismicanalysis，Webinformationfiltering，biometricidentification，computeraidedmedicaldiagnosis，etcInthispaperadetailedanalysisoftheoryandalgorithmsaboutensemblelearningisgiven，andthenitpr

5、obesintothereasonwhyensemblelearningworksanditsadvantageswhicharesuperiortotraditionalmethodsAfterthat，itpointsouttheshortcomingsanddevelopmentdirectionsofensemblelearningAtlast，anintroductiontothetheoryandmainalgorithms

6、ofselectiveensemblelearningisfollowedNowtherearethemaintwoworksthatalefinishedinthispaperasfollowsFirstlyitintroducesthetheoryofconditionalmutualinformationininformationtheoryandthencombinesitwithensemblelearningtoconstr

7、uctanovelmethodcalledCMISEN(ConditionalMutualInformationBasedSelectiveEnsemble)toselecttheoptimizedindividuallearnersThismethodconsiderstheinfluenceamongindividuallearners，andCanpreventselectingtheredundantindividuallear

8、nertomakethedifferencesamongindividualleamersthathavebeenselectedmuchbiggerThisnovelmethodisimplementedinweakenvironmenti11JavalanguageInordertocomparethismethodwiththeexistingmethodssuchasBaggingandMISEN(MutualInformati