Linearization learning method of BP neural networks |
| |
Authors: | Zhou Shaoqian Ding Lixin Zhang Jian Tang Xinhua |
| |
Affiliation: | (1) State Key Laboratory of Software Engineering, Wuhan University, 430072 Wuhan, China;(2) The Software Institute of Wuhan University, 430072 Wuhan |
| |
Abstract: | Feedforward multi-layer neural networks have very strong mapping capability that is based on the non-linearity of the activation function, however, the non-linearity of the activation function can cause the multiple local minima on the learning error-surfaces, which affect the learning rate and solving optimal weights. This paper proposes a learning method linearizing non-linearity of the activation function and discusses its merits and demerits theoretically. Supported by Wuhan Shenguang Project, the National Natural Science Foundation of China and the 863 High Techology Project Zhou Shaoqian: born in Mar. 1962,Ph. D graduate student |
| |
Keywords: | BP neural networks activation function linearization method |
本文献已被 SpringerLink 等数据库收录! |
|