BOUNDS ON THE NUMBER OF HIDDEN NEURONS IN MULTILAYER PERCEPTRONS

被引：220

作者：

HUANG, SC

HUANG, YF

机构：

[1] Department of Electrical Engineering, University of Notre Dame, Notre Dame

来源：

IEEE TRANSACTIONS ON NEURAL NETWORKS | 1991年 / 2卷 / 01期

关键词：

D O I：

10.1109/72.80290

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

This paper investigates some fundamental issues concerning the capability of multilayer perceptrons with one hidden layer. The studies are focused on realizations of functions which map from a finite subset of E(n) into E(d). Both real-valued and binary-valued functions are considered. In particular, a least upper bound is derived for the number of hidden neurons needed to realize an arbitrary function which maps from a finite subset of E(n) into E(d). A nontrivial lower bound is also obtained for realizations of injective functions. This result will be useful in studying pattern recognition and database retrieval. In addition, an upper bound is given for realizing binary-valued functions that are related to pattern classification problems.

引用

页码：47 / 55

页数：9

共 21 条

[11] KUNG SY, 1988, P IEEE INT C NEURAL, P363
[12] Luenberger D. G., 1984, INTRO LINEAR NONLINE
[13] ON HIDDEN NODES FOR NEURAL NETS
MIRCHANDANI, G
CAO, W
[J]. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS, 1989, 36 (05): : 661 - 664
[14] THE CAPACITY OF MULTILEVEL THRESHOLD FUNCTIONS
OLAFSSON, S
ABUMOSTAFA, YS
[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 1988, 10 (02) : 277 - 281
[15] Pao YH, 1989, ADAPTIVE PATTERN REC
[16] Rosenblatt F., 1961, PRINCIPLES NEURODYNA
[17] Rumelhart DE, 1986, PARALLEL DISTRIBUTED, V1-2
[18] SCHLAFLI L, 1950, GESAMMELTE MATH ABHA, V1, P209
[19] SOON VC, 1988, DEC P CONTR DEC C AU, P348
[20] SOON VC, 1988, 1988 P C INF SCI SYS, P308

← 1 2 3 →