1.TheTasksofTestsandEstimationsMutualInformationandAverageLogLikelihoodChangeswithz’inTestsAStoryaboutLookingforz*:SimilartoCatchingaCricket5.TheResearchHistory6.Importantstep:UsingTruthFunctiontoProduceLikelihoodFunction7.SemanticInformationMeasureCompatiblewithShannon，Popper，Fisher，andZadeh’sThoughts8.SemanticKullback-LeiblerInformationandSemanticMutualInformation9.Channels’MatchingAlgorithm10.SemanticChannelMatchesShannon’sChannel11.Multi-labelLogicalClassificationandSelectiveClassificationwithMLCriteriony0:test-negativey1:test-positiveTooptimizeT(θj|X):T(θ1|x1)=T(θ0|x0)=1T(θ1|x0)=b1’*=P(y1|x0)/P(y1|x1)T(θ0|x1)=b0’*=P(y0|x1)/P(y0|x0)Tooptimizeclassifier：j=1,213.UsingR(G)functiontoProveIterativeConvergence14.UsingR(G)FunctiontoProveCMAlgorithm’sConvergenceforTestsandEstimations15.AnExampleforEstimationsShowstheConvergentReliability16.TheCMAlgorithmforMixtureModelsDifference:LookingfortheTrueShannonchannel17.IllustratingtheConvergenceoftheCMAlgorithm18.ACounterexamplewithR>R*orQ>Q*againsttheEM19.IllustratingFuzzyClassificationforMixtureModels20.TheNumbersofIterationsforConvergence21.MSIinComparisonwithMLEandMAP22.Summary