A regularization term to avoid the saturation of the sigmoids in multilayer neural networks

被引：4

作者：

Garrido, L

Gomez, S

Gaitan, V

SerraRicart, M

机构：

[1] UNIV BARCELONA,FAC FIS,DEPT ESTRUCTURA & CONSTITUENTS MAT,E-08028 BARCELONA,SPAIN

[2] UNIV AUTONOMA BARCELONA,INST FIS ALTES ENERGIES,E-08193 BARCELONA,SPAIN

[3] INST ASTROFIS CANARIAS,E-38200 LA LAGUNA,SPAIN

来源：

INTERNATIONAL JOURNAL OF NEURAL SYSTEMS | 1996年 / 7卷 / 03期

关键词：

D O I：

10.1142/S0129065796000233

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

In this paper are propose a new method to prevent the saturation of any set of hidden units of a multilayer neural network. This method is implemented by adding a regularization term to the standard quadratic error function, which is based on a repulsive action between pairs of patterns.

引用

页码：257 / 262

页数：6

共 9 条

[1]

COTTRELL GW, 1987, P 9 ANN C COGN SCI S, P462

[2]

Hertz J., 1991, Introduction to the Theory of Neural Computation

[3] A PRACTICAL BAYESIAN FRAMEWORK FOR BACKPROPAGATION NETWORKS [J].