Building Recurrent Neural Networks to Implement Multiple Attractor Dynamics Using the Gradient Descent Method英文文献资料.docVIP

下载本文档

8
0
约3.57万字
约 12页
2017-05-10 发布于上海
举报
版权申诉

Building Recurrent Neural Networks to Implement Multiple Attractor Dynamics Using the Gradient Descent Method英文文献资料.doc

1、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。。
2、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。
3、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)。
4、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
5、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
6、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
7、VIP文档为合作方或网友上传，每下载1次，网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档

Building Recurrent Neural Networks to Implement Multiple Attractor Dynamics Using the Gradient Descent Method英文文献资料

HindawiPublishingCorporation AdvancesinArti?cialNeuralSystems Volume2009,ArticleID846040,11pages doi:10.1155/2009/846040 ResearchArticle BuildingRecurrentNeuralNetworkstoImplementMultiple AttractorDynamicsUsingtheGradientDescentMethod JunNamikawaandJunTani BrainScienceInstitute,RIKEN,2-1Hirosawa,WakoCity,Saitama351-0198,Japan CorrespondenceshouldbeaddressedtoJunNamikawa,jnamika@bdc.brain.riken.jp Received31March2008;Accepted22August2008 RecommendedbyAkiraImada Thepresentpaperproposesarecurrentneuralnetworkmodelandlearningalgorithmthatcanacquiretheabilitytogenerate desiredmultiplesequences.Thenetworkmodelisadynamicalsysteminwhichthetransitionfunctionisacontractionmapping, andthelearningalgorithmisbasedonthegradientdescentmethod.Weshowanumericalsimulationinwhicharecurrentneural networkobtainsamultipleperiodicattractorconsistingof?veLissajouscurves,oraVanderPoloscillatorwithtwelvedi?erent parameters.Thepresentanalysisclari?esthatthemodelcontainsmanystableregionsasattractors,andmultipletimeseriescan beembeddedintotheseregionsbyusingthepresentlearningmethod. Copyright?2009J.NamikawaandJ.Tani.ThisisanopenaccessarticledistributedundertheCreativeCommonsAttribution License,whichpermitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalworkisproperly cited. 1.Introduction At the end of successful learning, each expert implements attractor dynamics as locally represented knowledge, and a gating mechanism chooses only one expert at any time. The system can acquire many attractor patterns although there is a disadvantage in that the system does not have the generalization ability on the attractor patterns. As the other approach to implement multiple patterns, the parametric bias (PB) method has been developed to improve the learning capability of RNNs [9, 10]. In an RNN that employs the PB method (RNNPB), PB values provide the information needed in order to individualize each sequence. It has been reported that the number of time series that RNNPBs can