资源描述:
《Feature Selection for Nonlinear Regression and its Application to SDM_0533-000057》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、FeatureSelectionforNonlinearRegressionanditsApplicationtoCancerResearchYijunSunJinYaoySteveGoodisonzAbstracttheissueistoperformfeatureselectiontoextracttheFeatureselectionisafundamentalprobleminmachinemostrelevantinformationabouteachobserveddatumlearning.Withtheadventofh
2、igh-throughputtechnolo-fromapotentiallyoverwhelmingquantityofitsfeaturesgies,itbecomesincreasinglyimportantinawiderange[7].Anexamplewherefeatureselectionplaysacriticalofscienticdisciplines.Inthispaper,weconsidertheroleistheuseofoligonucleotidemicroarrayfortheiden-problem
3、offeatureselectionforhigh-dimensionalnon-ticationofcancer-associatedgeneexpressionprolesoflinearregression.Thisproblemhasnotyetbeenwellad-prognosticvalue.Typically,thenumberofsamplesisdressedinthecommunity,andexistingmethodssueraroundonehundred,whilethenumberofgenesass
4、oci-fromissuessuchaslocalminima,simpliedmodelas-atedwithrawdataisontheorderofthousandsorevensumptions,highcomputationalcomplexityandselectedtensofthousands.Theidenticationofasmallfrac-featuresnotdirectlyrelatedtolearningaccuracy.Wetionofgenesthatdrivecanceroustumorgrowt
5、hand/orproposeanewwrappermethodthataddressessomeofspreadcansignicantlyimprovetheaccuracyofcancertheseissues.Westartbydevelopinganewapproachprognosis.Inadditiontodefyingthecurseofdimen-toestimatingsampleresponsesandpredictionerrors,sionality,eliminatingirrelevantfeaturesc
6、analsoreduceandthendeployafeatureweightingstrategytondaprocessingtimeofdataanalysisandthecostofcollect-featuresubspacewhereapredictionerrorfunctionisingirrelevantfeatures.Inmanycases,featureselectionminimized.Weformulateitasanoptimizationprob-canalsoprovidesignicantinsi
7、ghtsintothenatureoflemwithintheSVMframeworkandsolveitusingantheproblemunderinvestigation.iterativeapproach.Ineachiteration,agradientdescentTheproblemoffeatureselectionhasbeenex-basedapproachisderivedtoecientlyndasolution.tensivelystudiedinthemachinelearningcommunityAlar
8、ge-scalesimulationstudyisperformedonfoursyn-[11,7,23,24,25].However,themajorityoftheworkthetican