ARBITRARY NONLINEARITY IS SUFFICIENT TO REPRESENT ALL FUNCTIONS BY NEURAL NETWORKS - A THEOREM

被引：97

作者：

KREINOVICH, VY

机构：

[1] Univ of Texas at El Paso, El Paso, United States

来源：

NEURAL NETWORKS | 1991年 / 4卷 / 03期

关键词：

BACK PROPAGATING NETWORK; MAPPING NETWORKS; UNIVERSAL APPROXIMATION; NETWORK REPRESENTATION CAPABILITY; NONIDEAL NEURONS;

D O I：

10.1016/0893-6080(91)90074-F

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

It is proved that if we have neurons implementing arbitrary linear functions and a neuron implementing one (arbitrary but smooth) nonlinear function g(x), then for every continuous function f(x1, ..., x(m)) of arbitrarily many variables and for arbitrary e > 0 we can construct a network that consists of g-neurons and linear neurons and computes f with precision e.

引用

页码：381 / 383

页数：3

共 14 条

[1]

Hecht-Nielsen R., 1987, IEEE First International Conference on Neural Networks, P19

[2]

HECHTNIELSEN R, 1988, 1988 INT NEUR NETW S

[3] MULTILAYER FEEDFORWARD NETWORKS ARE UNIVERSAL APPROXIMATORS [J].

HORNIK, K ;

STINCHCOMBE, M ;

WHITE, H .

NEURAL NETWORKS, 1989, 2 (05) :359-366

[4]

JUNG WG, 1974, IC OP AMP COOKBOOK

[5]

KOLMOGOROV AN, 1957, DOKL AKAD NAUK SSSR+, V114, P953

[6]

KREINOVICH V, 1988, P INT C COMPUTER LOG, P31

[7]

KREINOVICH V, 1989, CLASSIFICATION TRANS

[8]

KREINOVICH V, 1986, NOT AM MATH SOC, V33, P945

[9]

KREINOVICH VY, 1990, LECT NOTES COMPUT SC, V417, P112

[10] GENERAL-FORM OF MEASUREMENT TRANSFORMATIONS WHICH ADMIT THE COMPUTATIONAL METHODS OF METROLOGICAL ANALYSIS OF MEASURING-TESTING AND MEASURING-COMPUTING SYSTEMS [J].

KROTKOV, IN ;

KREINOVICH, VY ;

MAZIN, VD .

MEASUREMENT TECHNIQUES USSR, 1987, 30 (10) :936-939

← 1 2 →