ALTERNATING MINIMIZATION AND BOLTZMANN MACHINE LEARNING

被引：31

作者：

BYRNE, W ^{[1
]}

机构：

[1] UNIV MARYLAND,SYST RES CTR,COLLEGE PK,MD 20742

来源：

IEEE TRANSACTIONS ON NEURAL NETWORKS | 1992年 / 3卷 / 04期

关键词：

D O I：

10.1109/72.143375

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Training a Boltzmann machine with hidden units is appropriately treated in information geometry using the information divergence and the technique of alternating minimization. The resulting algorithm is shown to be closely related to gradient descent Boltzmann machine learning rules, and the close relationship of both to the EM algorithm is described. An iterative proportional fitting procedure for training machines without hidden units is described and incorporated into the alternating minimization algorithm.

引用

页码：612 / 620

页数：9

共 11 条

[1]

Aarts E., 1989, SIMULATED ANNEALING

[2]

ACKLEY DH, 1985, COGNITIVE SCI, V9, P147

[3] INFORMATION GEOMETRY OF BOLTZMANN MACHINES [J].

AMARI, S ;

KURATA, K ;

NAGAOKA, H .

IEEE TRANSACTIONS ON NEURAL NETWORKS, 1992, 3 (02) :260-271

[4] I-DIVERGENCE GEOMETRY OF PROBABILITY DISTRIBUTIONS AND MINIMIZATION PROBLEMS [J].

CSISZAR, I .

ANNALS OF PROBABILITY, 1975, 3 (01) :146-158

[5]

CSISZAR I, 1984, STAT DECISIONS S, P205

[6]

CSISZAR I, 1981, INFORMATION THEORY C

[7] MAXIMUM LIKELIHOOD FROM INCOMPLETE DATA VIA EM ALGORITHM [J].

DEMPSTER, AP ;

LAIRD, NM ;

RUBIN, DB .

JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-METHODOLOGICAL, 1977, 39 (01) :1-38

[8] GEOMETRY OF EXPONENTIAL FAMILIES [J].

EFRON, B .

ANNALS OF STATISTICS, 1978, 6 (02) :362-376

[9]

HINTON G, 1983, 5TH P ANN C COGN SCI

[10]

Hinton G.E., 1986, PARALLEL DISTRIB PRO, V1, P2

← 1 2 →