资源描述：
《mining local data sources for learning global cluster models via local model exchange》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。
1、16FeatureArticle:Xiao-FengZhang,Chak-ManLamandWilliamK.CheungMiningLocalDataSourcesForLearningGlobalClusterModelsViaLocalModelExchangeXiao-FengZhang,Chak-ManLam,WilliamK.Cheung,Member,IEEEAbstract—Distributeddatamininghasrecentlycaughtalottheoverallresult.Theyhaveappli
2、edittolearningBayesianofattentionastherearemanycaseswherepoolingdistributedNetworksforWebloganalysis[12],[8].dataforminingisprobibited,duetoeitherhugedatavolumeorRegardingincorporationoflocaldataprivacycontrolindataprivacy.Inthispaper,weaddressedtheissueoflearningadist
3、ributeddatamining,Cliftonetal.[13],[14],[15]andDuglobalclustermodel,knownasthelatentclassmodel,byminingdistributeddatasources.Mostoftheexistingmodellearningetal.[16],[17],[18]haveproposedsolutionstodistributedalgorithms(e.g.,EM)requireaccesstoalltheavailabletrainingass
4、ociationrulesminingwithprivacypreservingcapability.data.Instead,westudiedamethodologybasedonperiodicmodelUnderthepremisethatpartiesprefertosharethelocaldataexchangeandmerge,andappliedittoWebstructuremodeling.miningresultsinsteadoftheoriginallocaldata,eachpartyInadditio
5、n,wehavetestedanumberofvariationsofthebasicsitelearnsanddiscloseonlytheirlocalpatterns,whichwillidea,includingconfiningtheexchangetosomeprivacyfriendlyparametersandvaryingthenumberofdistributedsources.Ex-eventuallybeaggregatedtogethertoformsomeglobalpat-perimentalresult
6、sshowthattheproposeddistributedlearningterns.Otherthantakingassociatedrulemining,Meruguetal.schemeiseffectivewithaccuracyclosetothecasewithallthe[19],[20],[21],[22],[23]worksonminingglobalclustersdataphysicallysharedforthelearning.Also,ourresultsshow(intheformofGaussia
7、nmixturemodel)ofhighdimensionempiricallythatsharinglessmodelparametersasafurtherfeaturevectorswhicharedistributedindifferentsites.Theirmechanismforprivacycontroldoesnotresultinsignificantperformancedegradationforourapplication.proposedmethodstartswithcreatinglocalcluste
8、rmodelsandthenresamplingfromthecombinedmodels“virtual”globalIndexTerms—Distributeddatamining,model-basedlearning,late
