文档详情

《人工智能与数据挖掘教学课件》l.pptx

发布:2025-05-12约7.04千字共10页下载文档
文本预览下载声明

2025/5/13AIDM1Chapter8

NeuralNetworksPartIII:AdvanceDataMiningTechniques

Content2025/5/13AIDM2No.3WhatWhyANN(8.1FeedforwardNeuralNetwork)HowANNworks-workingprinciple(8.2.1SupervisedLearning)MostpopularANN-BackpropagationNetwork(8.5.1TheBackpropagationAlgorithm:Anexample)No.2No.1

2025/5/13AIDM3WhatWhyANN:

ArtificialNeuralNetworks(ANN)ANNisaninformationprocessingtechnologythatemulatesabiologicalneuralnetwork.Neuron(神经元)vsNode(Transformation)Dendrite(树突)vsInputAxon(轴突)vsOutputSynapse(神经键)vsWeightStartsin1970s,becomeverypopularin1990s,becauseoftheadvancementofcomputertechnology.

2025/5/13AIDM6WhatisANN:BasicsTypesofANNNetworkstructure,e.g.Figure17.917.10(Turban,2000,version5,p663)NumberofhiddenlayersNumberofhiddennodesFeedforwardandfeedbackward(timedependentproblems)Linksbetweennodes(existorabsentoflinks)Theultimateobjectivesoftraining:obtainasetofweightsthatmakesalltheinstancesinthetrainingdatapredictedascorrectlyaspossible.Back-propagationisonetypeofANNwhichcanbeusedforclassificationandestimationmulti-layer:Inputlayer,Hiddenlayer(s),OutputlayerFullyconnectedFeedforwardErrorback-propagation

Content2025/5/13AIDM7WhatWhyANN(8.1FeedforwardNeuralNetwork)01HowANNworks-workingprinciple(8.2.1SupervisedLearning)01MostpopularANN-BackpropagationNetwork(8.5.1TheBackpropagationAlgorithm:Anexample)01

2025/5/13AIDM82.HowANN:workingprinciple(I)Step1:CollectdataStep2:SeparatedataintotrainingandtestsetsfornetworktrainingandvalidationrespectivelyStep3:Selectnetworkstructure,learningalgorithm,andparametersSettheinitialweightseitherbyrulesorrandomlyRateoflearning(pacetoadjustweights)Selectlearningalgorithm(Morethanahundredlearningalgorithmsavailableforvarioussituationsandconfigurations)

2.ANNworking

显示全部
相似文档