BIASED LEARNING IN BOOLEAN PERCEPTRONS

被引：4

作者：

KINOUCHI, O ^{[1
]}

CATICHA, N ^{[1
]}

机构：

[1] UNIV SAO PAULO,BR-05508 SAO PAULO,BRAZIL

来源：

PHYSICA A | 1992年 / 185卷 / 1-4期

关键词：

D O I：

10.1016/0378-4371(92)90482-6

中图分类号：

O4 [物理学];

学科分类号：

0702 ;

摘要：

The generalization ability of Hebbian Boolean perceptrons can be improved by a kind of feedback mechanism in which the student net judges the difficulty of a new example from its previous experience. It is shown that by giving a higher weight to the hard examples both generalization and learning abilities can be increased. Analytical as well as numerical results are presented for both cases where the examples are drawn at random or selected in an intelligent form.

引用

收藏

页码：411 / 416

页数：6

相关论文

共 8 条

[1] OPTIMAL LEARNING IN NEURAL NETWORK MEMORIES [J].

ABBOTT, LF ;

KEPLER, TB .

JOURNAL OF PHYSICS A-MATHEMATICAL AND GENERAL, 1989, 22 (14) :L711-L717

[2] LEARNING AND RETRIEVAL IN ATTRACTOR NEURAL NETWORKS ABOVE SATURATION [J].

GRINIASTY, M ;

GUTFREUND, H .

JOURNAL OF PHYSICS A-MATHEMATICAL AND GENERAL, 1991, 24 (03) :715-734

[3] 1ST-ORDER TRANSITION TO PERFECT GENERALIZATION IN A NEURAL NETWORK WITH BINARY SYNAPSES [J].

GYORGYI, G .

PHYSICAL REVIEW A, 1990, 41 (12) :7097-7100

[4]

GYORGYI G, 1990, 17 P STATPHYS WORKSH

[5] IMPROVING A NETWORK GENERALIZATION ABILITY BY SELECTING EXAMPLES [J].

KINZEL, W ;

RUJAN, P .

EUROPHYSICS LETTERS, 1990, 13 (05) :473-477

[6] A THEORY OF THE LEARNABLE [J].

VALIANT, LG .

COMMUNICATIONS OF THE ACM, 1984, 27 (11) :1134-1142

[7] THE HEBB RULE FOR LEARNING LINEARLY SEPARABLE BOOLEAN FUNCTIONS - LEARNING AND GENERALIZATION [J].

VALLET, F .

EUROPHYSICS LETTERS, 1989, 8 (08) :747-751

[8] LINEAR AND NONLINEAR EXTENSION OF THE PSEUDO-INVERSE SOLUTION FOR LEARNING BOOLEAN FUNCTIONS [J].

VALLET, F ;

CAILTON, JG ;

REFREGIER, P .

EUROPHYSICS LETTERS, 1989, 9 (04) :315-320