Adaptive Thresholding in Structure Learning of a Bayesian Network.pdf

Adaptive Thresholding in Structure Learning of a Bayesian Network.pdf

ID:34592857

大小:1.09 MB

页数:7页

时间:2019-03-08

Adaptive Thresholding in Structure Learning of a Bayesian Network.pdf_第1页
Adaptive Thresholding in Structure Learning of a Bayesian Network.pdf_第2页
Adaptive Thresholding in Structure Learning of a Bayesian Network.pdf_第3页
Adaptive Thresholding in Structure Learning of a Bayesian Network.pdf_第4页
Adaptive Thresholding in Structure Learning of a Bayesian Network.pdf_第5页
资源描述:

《Adaptive Thresholding in Structure Learning of a Bayesian Network.pdf》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、ProceedingsoftheTwenty-ThirdInternationalJointConferenceonArtificialIntelligenceAdaptiveThresholdinginStructureLearningofaBayesianNetworkBoazLerner,MichalAfek,RafiBojmelBen-GurionUniversityoftheNegev,Israelboaz@bgu.ac.il;{caspimic,rafibojmel}@gmail.comAbstractthresholdse

2、lection–awrongthresholdinfluencesstructureThresholdingameasureinconditionallearning[2,12,14,20].Commonly,athresholdisselectedindependence(CI)testsusingafixedvalueenablesaccordingtothreemainmethods:asauserdefault[10,29,learningandremovingedgesaspartoflearninga31,33],basedo

3、nlimitedtrialanderrorexperimentation[5],Bayesiannetworkstructure.However,thelearnedorautomaticallyasthethreshold,fromarangeofstructureissensitivetothethresholdthatiscandidates,thatmax(in)imizesaperformancemeasure[9,commonlyselected:1)arbitrarily;2)irrespective14,20,23,35]

4、.Thefirstmethodprovidessimplicityandofcharacteristicsofthedomain;and3)fixedforalllowruntime,butcannotguaranteesatisfactoryperformanceCItests.Weanalyzetheimpactonmutualformostproblems.Thesecondmethodovercomesinformation–aCImeasure–offactors,suchasexhaustiveexperimentationu

5、singalimitednumberofsamplesize,degreeofvariabledependence,andcandidates,butitismanualandincomplete.Thethirdvariables’cardinalities.Following,wesuggesttomethodmayfindaglobaloptimalthresholdbutmayalsoadaptivelythresholdindividualtestsbasedonthesufferfromalongruntime.factors

6、.WeshowthatadaptivethresholdsbetterAthresholdisusuallynotonlyselectedarbitrarilybutdistinguishbetweenpairsofdependentvariablesalsoirrespectivelyoftestfactors.Whilethecriticalvalueandpairsofindependentvariablesandenableincreaseswithdf,itisindifferenttosamplesize.TheCMIlear

7、ningstructuresmoreaccuratelyandquicklythresholdisindifferenttoallfactors.Toourbestknowledge,thanwhenusingfixedthresholds.noideaforsettingaCMIthresholdthatdependsonthefactorshaseverbeenoffered.Also,noproposalhaseverbeengiventoadaptathresholdtofactorsofeachCMItest.Wesuggest

8、adaptivethresholdingforindividualCItests1Introductionbasedontestfactors,andapplythisconceptinCMI

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。