资源描述：
《cognitive studies in medicinenew》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、BeliefMaintenanceinBayesianNetworksMarcoRamoniAlbertoRivaCognitiveStudiesinMedicineLaboratoriodiInformaticaMedicaMcGillCognitiveScienceCentreDipartimentodiInformaticaeSistemisticaMcGillUniversity,Montreal,CanadaUniversitadiPavia,Pavia,ItalyReprintedf
2、rom:TenthAnnualConferenceonUncertaintyinArti cialIn-telligence,July29-31,Seattle,Washington.Address:MarcoRamoni.McGillCognitiveScienceCentre.McGillUniver-sity.1110PineAvenueWest,Room201.Phone:(514)398-7331,Fax:(514)398-7246,Email:marco@psych.mcgill.ca
3、.BeliefMaintenanceinBayesianNetworksMarcoRamoniAlbertoRivaCognitiveStudiesinMedicineLaboratoriodiInformaticaMedicaMcGillCognitiveScienceCentreDipartimentodiInformaticaeSistemisticaMcGillUniversity,Montreal,CanadaUniversitadiPavia,Pavia,ItalyAbstracta
4、ndindependenceassumptionsamongvariables,thusmakingeasiertherepresentationandtheacquisitionofknowledge.BayesianBeliefNetworks(bbns)areapow-erfulformalismforreasoningunderuncer-Despiteaconsiderablesuccess,bbnsbearsomeseveretaintybutbearsomeseverelimitat
5、ions:theylimitations:requirealargeamountofinformationbe-foreanyreasoningprocesscanstart,theyIgnorance:Theyrequirealargeamountofinfor-havelimitedcontradictionhandlingcapabil-mation.Thenumberofconditionalprobabili-ities,andtheirabilitytoprovideexplana-t
6、iesneededtospecifyaconditionaldependencytionsfortheirconclusionisstillcontroversial.growsexponentiallywiththenumberofparentThereexistsaclassofreasoningsystems,variables.Thise ectleadstoseriousdicultiescalledTruthMaintenanceSystems(tmss),sincealltheco
7、nditionalsprobabilitiesde ningawhichareabletodealwithpartiallyspeci-conditionaldependencyamongvariables,aswell edknowledge,toprovidewell-foundedex-asallthepriorprobabilitiesfortherootvariables,planationfortheirconclusions,andtodetecthavetobeknownbefor
8、eanyreasoningprocessandhandlecontradictions.tmssincorporat-canstart.Moreover,eachprobabilityassignmentingmeasureofuncertaintyarecalledBeliefexpressesinasinglenumberboththebeliefaboutMaintenanceSystems(bmss).Thispaperde-aneventandthereliability
