首页 | 本学科首页   官方微博 | 高级检索  
     

用在线梯度法训练积单元神经网络的收敛性分析
引用本文:张超,李正学,陈先华,熊焱. 用在线梯度法训练积单元神经网络的收敛性分析[J]. 高等学校计算数学学报, 2010, 32(3)
作者姓名:张超  李正学  陈先华  熊焱
作者单位:1. 大连理工大学数学科学学院,大连,116023
2. 辽宁科技大学理学院数学系,鞍山,114051
摘    要:<正>1引言仅由加和单元构成的传统前向神经网络已经广泛应用于模式识别及函数逼近等领域.但在处理比较复杂的问题时,这种网络往往需要补充大量的隐节点,这样就不可避免地增

关 键 词:梯度算法  收敛性分析  单元  在线方式  前向神经网络  误差函数  梯度法  训练周期  不等式  高等学校  

CONVERGENCE ANALYSIS OF AN ONLINE GRADIENT METHOD FOR PRODUCT UNIT NEURAL NETWORKS
Zhang Chao,Li Zhengxue,Chen Xianhua,Xiong Yan. CONVERGENCE ANALYSIS OF AN ONLINE GRADIENT METHOD FOR PRODUCT UNIT NEURAL NETWORKS[J]. Numerical Mathematics A Journal of Chinese Universities, 2010, 32(3)
Authors:Zhang Chao  Li Zhengxue  Chen Xianhua  Xiong Yan
Affiliation:Zhang Chao Li Zhengxue Chen Xianhua (School of Mathematical Sciences,Dalian University of Technology,Dalian 116024) Xiong Yan (Department of Mathematics,School of Science,University of Science and Technology LiaoNing,Anshan 114051)
Abstract:Product unit neural network is one kind of Higher-Order neural network with exponential weights.Due to the special structure with exponential weights,it can provide more powerful internal representation capability than traditional feed-forward neural networks.In this paper,a convergence result for an online gradient method to train product unit neural networks is presented.The monotonicity of the error function during the training iteration process is also guaranteed. The experiment results are given by the...
Keywords:product unit neural network  online gradient method  monotonicity  convergence  
本文献已被 CNKI 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号