资源描述：
《Access Methods for Markovian Streams》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、1AccessMethodsforMarkovianStreamsJulieLetchner#1,ChristopherRe´#2,MagdalenaBalazinska#3,MatthaiPhilipose4#ComputerScience&EngineeringDepartment,UniversityofWashingtonSeattle,Washington,USAf1letchner,2chrisre,3magdag@cs.washington.eduIntelResearchSeattleSeattle,Washin
2、gton,USA4matthai.philipose@intel.comAbstract—Model-basedviewshaverecentlybeenproposedasarchived.Ourgoalistosupportarchive-basedapplicationsaneectivemethodforqueryingnoisysensordata.CommonlythatleveragethissmootheddatainordertoprovidethemostusedmodelsfromtheAIliterature
3、(e.g.,thehiddenMarkovaccuratepossibleanswerstohistoricalqueries(e.g.,“WasBobmodel)exposetoapplicationsastreamofprobabilisticandinhisoceyesterday?”,“DidMargottakehermedicationcorrelatedstateestimatescomputedfromthesensordata.Manyapplicationswanttodetectsophisticatedpat
4、ternsofstatesfrombeforebreakfasteverydaylastmonth?”,etc.).theseMarkovianstreams.Suchqueriesarecalledeventqueries.TheresultofanysmoothingtechniqueisaprobabilisticInthispaper,wepresentanewMarkovianstreamstoragestreaminwhicheachtimestepencodesnotasinglestate,manager,Calde
5、ra.WedevelopandevaluateCalderaasacompo-butadistributionoverpossiblestates.IntheRFIDtrackingnentofLahar,aMarkovianstreameventqueryprocessingsystemexample,suchastreammightindicate,foreachtimestep,thedevelopedinpreviouswork.AttheheartofCalderaisasetofaccessmethodsforMarko
6、vianstreamsthatcanimproveeventdistributionoverpossiblelocationsofatag:e.g.,attime7,Bobqueryperformancebyordersofmagnitudecomparedtoexistingwasinthehallwaywithprobability0.8andinhisocewithtechniques,whichmustscantheentirestream.Ouraccessprobability0.2.Additionally,stat
7、esatconsecutivetimestepsmethodsusenewadaptationsoftraditionalB+treeindexes,canbecorrelated:e.g.,Bob’slocationattime8iscorrelatedandanewindex,calledtheMarkov-chainindex.Theyecientlywithhislocationattime7.Wecalltheseprobabilistic,extractonlytherelevanttimestepsfromastre
8、am,whileretainingthestream’sMarkovianproperties.WehaveimplementedourcorrelatedstreamsMarkovianstreams.Theyareamateria
