Architecture and Equilibra结构和平衡 刘瑞华 罗雪梅 导师:曾.ppt

Architecture and Equilibra结构和平衡 刘瑞华 罗雪梅 导师:曾.ppt

  1. 1、本文档共35页,可阅读全部内容。
  2. 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
  3. 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载
  4. 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多

ArchitectureandEquilibra

结构和平衡

刘瑞华罗雪梅

导师:曾平Chapter6

2004.11.101

Chapter6ArchitectureandEquilibriaPerfacelyaoynovstabletheorem2004.11.102

Chapter6ArchitectureandEquilibria

6.1NeutralNetworkAsStochasticGradientsystemClassifyNeutralnetworkmodelBytheirsynapticconnectiontopolgiesandbyhowlearningmodifiestheirconnectiontopologiessynapticconnectiontopolgieshowlearningmodifiestheirconnectiontopologies2004.11.103

Chapter6ArchitectureandEquilibria

6.1NeutralNetworkAsStochasticGradientsystem2004.11.104

Chapter6ArchitectureandEquilibria

6.1NeutralNetworkAsStochasticGradientsystemThreestochasticgradientsystemsrepresentthethreemaincategories:1)Feedforwardsupervisedneuralnetworkstrainedwiththebackpropagation(BP)algorithm.2)Feedforwardunsupervisedcompetitivelearningoradaptivevectorquantization(AVQ)networks.3)Feedbackunsupervisedrandomadaptivebidirectionalassociativememory(RABAM)networks.2004.11.105

Chapter6ArchitectureandEquilibria

6.2GlobalEquilibra:convergenceandstabilityNeuralnetwork:synapses,neuronsthreedynamicalsystems:synapsesdynamicalsystemsneuonsdynamicalsystemsjointsynapses-neuronsdynamicalsystemsHistorically,Neuralengineersstudythefirstorsecondneuralnetwork.Theyusuallystudylearninginfeedforwardneuralnetworksandneuralstabilityinnonadaptivefeedbackneuralnetworks.RABAMandARTnetworkdependonjointequilibrationofthesynapticandneuronaldynamicalsystems.2004.11.106

Chapter6ArchitectureandEquilibria

6.2GlobalEquilibra:convergenceandstabilityEquilibriumissteadystate.Convergenceissynapticequilibrium.Stabilityisneuronalequilibrium.Moregenerallyneuralsignalsreachsteadystateeventhoughtheactivationsstillchange.WedenotesteadystateintheneuronalfieldNeuronfluctuatefasterthansynapsesfluctuate.Stability-Convergencedilemma:Thesynapsedslowlyencodetheseneuralpat

文档评论(0)

***** + 关注
实名认证
内容提供者

该用户很懒,什么也没介绍

版权声明书
用户编号:8010045112000002

1亿VIP精品文档

相关文档