摘要
本文概述了多层前向网络研究的发展历史，对其中有代表性的若干成就进行了较为系统的介绍和评论，分析了当前研究工作中存在的一些问题，提出了解决这些问题的几种可行方案。在对多层前向网络的有效逼近机理进行深入剖析的基础上，提出了合理的有限规模多层前向网络应当遵循的若干构造原则。
In this paper,①a brief introduction is given to the history of research on multilayer feedforward neural networks;②a systematic review is given to some important achievements in the field of multilayer feedforward networks;③some existing problems and tendencies in the study of multilayer feedforward networks are pointed out,and a few of feasible methods for solving those questions are suggested;④a systematic analysis of approximation mechanism of feedforward networks with a scale,limited discrete point set is made;⑤some construction rules,which govern a rational feedforward networks,are established.
出处
《力学进展》
EI
CSCD
北大核心
1995年第2期186-196,共11页
Advances in Mechanics
基金
中国博士后基金
关键词
多层前向网络
学习算法
神经网络
multilayer feedforward neural networks
mechanism analysis
least square approximation
learning algorithm
