EVOLVING A LEARNING ALGORITHM FOR THE BINARY PERCEPTRON

被引：26

作者：

FONTANARI, JF ^{[1
]}

MEIR, R ^{[1
]}

机构：

[1] BELLCORE,MORRISTOWN,NJ 07960

来源：

NETWORK-COMPUTATION IN NEURAL SYSTEMS | 1991年 / 2卷 / 04期

关键词：

D O I：

10.1088/0954-898X/2/4/002

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

We employ a genetic algorithm to explore the space of learning algorithms for a single-layer, binary-weight neural network. We find that the directed drift algorithm, proposed recently by Venkatesh, emerges spontaneously from the genetic tournaments as the local and on-line algorithm that performs best in highly diverse environments.

引用

页码：353 / 359

页数：7

共 17 条

[1] BELEW RK, 1990, CSE CS90174 U CAL TE
[2] CHALMERS D, 1990, 1990 P CONN MOD SUMM
[3] DERRIDA B, 1991, FINITE SIZE EFFECTS
[4] Duda R. O., 1973, PATTERN CLASSIFICATI, V3
[5] LANDSCAPE STATISTICS OF THE BINARY PERCEPTRON
FONTANARI, JF
KOBERLE, R
[J]. JOURNAL DE PHYSIQUE, 1990, 51 (13): : 1403 - 1413
[6] THE SPACE OF INTERACTIONS IN NEURAL NETWORK MODELS
GARDNER, E
[J]. JOURNAL OF PHYSICS A-MATHEMATICAL AND GENERAL, 1988, 21 (01): : 257 - 270
[7] 3 UNFINISHED WORKS ON THE OPTIMAL STORAGE CAPACITY OF NETWORKS
GARDNER, E
DERRIDA, B
[J]. JOURNAL OF PHYSICS A-MATHEMATICAL AND GENERAL, 1989, 22 (12): : 1983 - 1994
[8] Garey M.R., 1979, COMPUTERS INTRACTABI, V174
[9] Goldberg DE, 1989, GENETIC ALGORITHMS S
[10] 1ST-ORDER TRANSITION TO PERFECT GENERALIZATION IN A NEURAL NETWORK WITH BINARY SYNAPSES
GYORGYI, G
[J]. PHYSICAL REVIEW A, 1990, 41 (12): : 7097 - 7100

← 1 2 →