NEW APPROACH TO SELECTION OF INITIAL VALUES OF WEIGHTS IN NEURAL FUNCTION APPROXIMATION

被引：22

作者：

OSOWSKI, S

机构：

[1] Institute of the Theory of Electrical Engineering and Electrical Measurements, Technical University, Pl. Politechniki 1

来源：

ELECTRONICS LETTERS | 1993年 / 29卷 / 03期

关键词：

NEURAL NETWORKS;

D O I：

10.1049/el:19930214

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

It is proven that the weights and biases generated with certain constraints based on the piecewise linear principle result in an initial neutral network which is better able to form a function approximation of an arbitrary function. Use of these initial constraints greatly shortens the training time and avoids the local minima usually associated with an arbitrary random choice of initial weight.

引用

页码：313 / 315

页数：3