模糊联想记忆FUZZY ASSOCIATIVE MEMMORIESⅡ.ppt

《模糊联想记忆FUZZY ASSOCIATIVE MEMMORIESⅡ.ppt》由会员上传分享，免费在线阅读，更多相关内容在PPT专区-天天文库。

1、模糊联想记忆FUZZYASSOCIATIVEMEMMORIESⅡPresentedbyYangBaishengE.E.Dept.XidianUniversityOUTLINEFuzzyHebbFAMs（续）6.BinaryInput-OutputFAMs7.MultiantecedentFAMRules8.AdaptiveDecompositionalInferenceAdaptiveFAMs(Product-SpaceClusteringinFAMCells)1.AdaptiveFAM-RuleGene

2、ration2.AdaptiveBIOFAMClustering3.AdaptiveBIOFAMExample:InvertedPendulumBinaryInput-OutputFAMsBIOFAMsmapsystem-variabletocontrol,classification,orotheroutputdata.Forexample:ABIOFAMmapstrafficdensitiestoscreen(andred)lightdurations.Ininverted-pendulumexamp

3、le,thesystemmapsthesystem-variable()tocontroldata().MultiantecedentFAMRules(多前提FAM规则)1.ConsidertheFAMrule:“IFXisA,THENCisZ,”orforshort.2.Theruleis“IFXisAANDYisB,THENCisZ,”orforshort.Whattodo?MultiantecedentFAMRules(多条件FAM规则)2Single-antecedentFAMs:Multia

4、ntecedentFAMRules:Defuzzifyittoyieldtheexactoutput.MultiantecedentFAMRulesSupposewepresenttheexactinputs,tothesingle-FAM-rulesystemthatstores(A,B;C).Wepresenttheunitbitvectorsandtoasnonfuzzysetinputs.ThenPropertyofHebbMatrixMultiantecedentFAMRulesRepresen

5、tingwithitsmembershipfunctionForallin:BIOFAMprescriptionMultiantecedentFAMRulesAlso,WecangettheFAMrules:IFweencode:andwithcorrelation-productencoding,decompositionalinferencegivestheBIOFAMversionofcorrelation-productinference:Correlation-ProductEncodingAd

6、aptiveDecompositionalInferenceLetdefineanarbitraryneural-networksystemthatmapsfuzzysubsetoftofuzzysubsetsof.candefineadifferentneural-network.Theneural-networkchangewithtime.AdaptiveFAMs(Product-SpaceClusteringinFAMCells)AdaptiveFAM-RuleGenerationAdaptive

7、BIOFAMClusteringAdaptiveBIOFAMExample:InvertedPendulumAdaptiveFAM-RuleGenerationLetdenotequantizationvectorsintheinput-outputproductspace.WecountthenumberofquantizingvectorsineachFAMcell.AdaptiveBIOFAMClusteringThroughneural-networklearningalgorithm,learn

8、todistributeinput-outputdataintheinput-outputproductspace.DataclustersreflectFAMrules,suchasthesteady-stateFAMrule”IFisZEANDisZE,THENisZE.”确定状态变量（条件变量）和控制变量（结论变量）收集相应的训练样本(大量的有代表性的)根据训练样本的分布区间，划分为相应的模糊数，并赋于模糊语言量。自适应