CONVERGENCE OF ONLINE GRADIENT METHOD WITH A PENALTY TERM FOR FEEDFORWARD NEURAL NETWORKS WITH STOCHASTIC INPUTS
CONVERGENCE OF ONLINE GRADIENT METHOD WITH A PENALTY TERM FOR FEEDFORWARD NEURAL NETWORKS WITH STOCHASTIC INPUTS作者机构：Department of Applied Mathematics Dalian University of Technology Dalian 116024 PRC. Department of Applied Mathematics Dalian University of Technology Dalian 116024 PRC.
出 版 物：《Numerical Mathematics A Journal of Chinese Universities(English Series)》 (NUMERICAL MATHEMATICS A JOURNAL OF CHINESE UNIVERSITIES ENGLISH SERIES)
年 卷 期：2005年第14卷第1期
页 面：87-96页
学科分类：07[理学] 0701[理学-数学] 070101[理学-基础数学]
基 金：Partly supported by the National Natural Science Foundation of China,and the Basic Research Program of the Committee of Science Technology and Industry of National Defense of China
主 题：前馈神经网络系统 收敛 随机变量 单调性 有界性原理 在线梯度计算法
摘 要：Online gradient algorithm has been widely used as a learning algorithm for feedforward neural network training. In this paper, we prove a weak convergence theorem of an online gradient algorithm with a penalty term, assuming that the training examples are input in a stochastic way. The monotonicity of the error function in the iteration and the boundedness of the weight are both guaranteed. We also present a numerical experiment to support our results.
维普期刊数据库
同方期刊数据库