USE OF SOME SENSITIVITY CRITERIA FOR CHOOSING NETWORKS WITH GOOD GENERALIZATION ABILITY

被引：204

作者：

DIMOPOULOS, Y

BOURRET, P

LEK, S

机构：

[1] UNIV TOULOUSE 3,UMR 9964,F-31062 TOULOUSE,FRANCE

[2] CERT,ONERA,DEPT ETUD & RECH INFORMAT,F-31055 TOULOUSE,FRANCE

来源：

NEURAL PROCESSING LETTERS | 1995年 / 2卷 / 06期

关键词：

D O I：

10.1007/BF02309007

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

In most applications of the multilayer perceptron (MLP) the main objective is to maximize the generalization ability of the network. We show that this ability is related to the sensitivity of the output of the MLP to small input changes. Several criteria have been proposed for the evaluation of the sensitivity. We propose a new index and present a way for improving these sensitivity criteria. Some numerical experiments allow a first comparison of the efficiencies of these criteria.

引用

页码：1 / 4

页数：4

共 7 条

[1] Bishop C. M, 1990, P SEATTL INT JOINT C P SEATTL INT JOINT C, P749
[2] Cybenko G., 1989, Mathematics of Control, Signals, and Systems, V2, P303, DOI 10.1007/BF02551274
[3] FRIEDMAN JH, 1993, STATISTICS NEURAL NE, P1
[4] Fu L, 1993, ICNN, DOI 10.1109/ICNN.1993.298559
[5] MATSUOKA K, 1990, P INNC 90 PARIS, P765
[6] FUNCTIONAL APPROXIMATION BY FEEDFORWARD NETWORKS - A LEAST-SQUARES APPROACH TO GENERALIZATION
WEBB, AR
[J]. IEEE TRANSACTIONS ON NEURAL NETWORKS, 1994, 5 (03): : 363 - 379
[7] WEIGEND AS, 1992, NONLINEAR MODELING F, P395

← 1 →