nonparametric bayesian networks

《nonparametric bayesian networks》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、NonparametricBayesianNetworks*UniversityPressScholarshipOnlineOxfordScholarshipOnlineBayesianStatistics9JoséM.Bernardo,M.J.Bayarri,JamesO.Berger,A.P.Dawid,DavidHeckerman,AdrianF.M.Smith,andMikeWestPrintpublicationdate:2011PrintISBN-13:9780199694587PublishedtoOxfordScholarshipOnline:Janua

2、ry2012DOI:10.1093/acprof:oso/9780199694587.001.0001NonparametricBayesianNetworks*KatjaIckstadtBjöornBornkampMarcoGrzegorczykJakobWieczorekMalikR.SheriffHernáanE.GreccoEliZamirDOI:10.1093/acprof:oso/9780199694587.003.0010AbstractandKeywordsAconvenientwayofmodellingcomplexinteractionsisbye

3、mployinggraphsornetworkswhichcorrespondtoconditionalindependencestructuresinanunderlyingstatisticalmodel.OnemainclassofmodelsinthisregardareBayesiannetworks,whichhavethedrawbackofmakingparametricassumptions.Bayesiannonparametricmixturemodelsofferapossibilitytoovercomethislimitation,butha

4、vehardlybeenusedincombinationwithnetworks.ThismanuscriptbridgesthisgapbyintroducingnonparametricBayesianPage1of40NonparametricBayesianNetworks*networkmodels.Wereview(parametric)Bayesiannetworks,inparticularGaussianBayesiannetworks,fromaBayesianperspectiveaswellasnonparametricBayesianmixt

5、uremodels.AfterwardsthesetwomodellingapproachesarecombinedintononparametricBayesiannetworks.ThenewmodelsarecomparedbothtoGaussianBayesiannetworksandtomixturemodelsinasimulationstudy,whereitturnsoutthatthenonparametricnetworkmodelsperformfavourablyinnon‐Gaussiansituations.Thenewmodelsarea

6、lsoappliedtoanexamplefromsystemsbiology,namelyfindingmoduleswithintheMAPKcascade.Keywords:GaussianBayesiannetworks,SystemsBiology,NonparametricMixtureModels,SpeciesSamplingModelsSummaryAconvenientwayofmodellingcomplexinteractionsisbyemployinggraphsornetworkswhichcorrespondtoconditionalin

7、dependencestructuresinanunderlyingstatisticalmodel.OnemainclassofmodelsinthisregardareBayesiannetworks,whichhavethedrawbackofmakingparametricassumptions.Bayesiannonparametricmixturemodelsofferapossibilitytoovercomethislimitation,buthavehardlybeenusedincombinationwithnetwo