- 1、本文档共75页,可阅读全部内容。
- 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、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、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
连续感知器学习算法的有限收敛性及连续距离转换神经网络论文
Abstract
Feedforwardstructure
trainedthe hasbeenone
by backpropagation(BP)algorithm
usedartificialneural
ofthemost networkmodels.Aremarkableofthe
widely aspect
current onneuralnetworksconcernsthetheoretical
debate foundationsof
commonly
used algorithms.
learning
Asthe unitofthefeedforwardneural
basic havethe
networks,perceptrons
ability
to
classify separable
linearly trainingpatterns
correctly.Especially,some
algorithms,
role
suchas anddeltarulebasedonLMS
pereeptron method,have
provedconvergent
for forthe
continuous
linearlyseparabletrainingpatterns[4].But perceptrons,despite
theexcellent havenotfound resultsforthe
application,we satisfactorily
proved linearly
havealso toobtainthe ofthe
separableproblem.Researchersattempted convergence
which to a resultforonlineBP
ny
&Maggini【7】inthey proveconvergence multilayer
with
neuralnetworks undersome
linearly
separablepatterns
contains
amathematicalmistakethat
文档评论(0)