THE EM ALGORITHM AND INFORMATION GEOMETRY IN NEURAL-NETWORK LEARNING

被引：10

作者：

AMARI, S

机构：

来源：

NEURAL COMPUTATION | 1995年 / 7卷 / 01期

关键词：

D O I：

10.1162/neco.1995.7.1.13

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Hidden units play an important role in neural networks, although their activation Values are unknown in many learning situations. The EM algorithm (statistical algorithm) and the em algorithm (information-geometric one) have been proposed so far in this connection, and the effectiveness of such algorithms is recognized in many areas of research. The present note points out that these two algorithms are equivalent under a certain condition, although they are different in general.

引用

页码：13 / 18

页数：6

共 10 条

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

AMARI, S ;

KURATA, K ;

NAGAOKA, H .

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

[2] DUALISTIC GEOMETRY OF THE MANIFOLD OF HIGHER-ORDER NEURONS [J].

AMARI, S .

NEURAL NETWORKS, 1991, 4 (04) :443-451

[3]

AMARI S, 1994, METR944 U TOK

[4]

AMARI S, 1985, SPRINGER LECTURE NOT, V28

[5] STATISTICAL-INFERENCE UNDER MULTITERMINAL RATE RESTRICTIONS - A DIFFERENTIAL GEOMETRIC APPROACH [J].

AMARI, SI ;

HAN, TS .

IEEE TRANSACTIONS ON INFORMATION THEORY, 1989, 35 (02) :217-227

[6] ALTERNATING MINIMIZATION AND BOLTZMANN MACHINE LEARNING [J].

BYRNE, W .

IEEE TRANSACTIONS ON NEURAL NETWORKS, 1992, 3 (04) :612-620

[7]

CSISZAR I, 1984, STAT DECISIONS S, P205

[8] HIERARCHICAL MIXTURES OF EXPERTS AND THE EM ALGORITHM [J].

JORDAN, MI ;

JACOBS, RA .

NEURAL COMPUTATION, 1994, 6 (02) :181-214

[9]

Murray M. K., 1993, MONOGRAPHS STAT APPL, V48

[10]

NEAL RN, 1994, IN PRESS NEW VERSION

← 1 →