Varieties of Helmholtz machine

被引：71

作者：

Dayan, P ^{[1
]}

Hinton, GE ^{[1
]}

机构：

[1] UNIV TORONTO, TORONTO, ON M5S 1A1, CANADA

来源：

NEURAL NETWORKS | 1996年 / 9卷 / 08期

基金：

加拿大自然科学与工程研究理事会;

关键词：

expectation-maximization; unsupervised learning; feedback connections;

D O I：

10.1016/S0893-6080(96)00009-3

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

The Helmholtz machine is a new unsupervised learning architecture that uses top-down connections to build probability density models of input and bottom-up connections to build inverses to those models. The wake-sleep learning algorithm for the machine involves just the purely local delta rule. This paper suggests a number of different varieties of Helmholtz machines, each with its own strengths and weaknesses. and relates them to cortical information processing. Copyright (C) 1996 Elsevier Science Ltd.

引用

页码：1385 / 1403

页数：19

共 65 条

[41] CONNECTIONIST LEARNING OF BELIEF NETWORKS [J].

NEAL, RM .

ARTIFICIAL INTELLIGENCE, 1992, 56 (01) :71-113

[42]

NEAL RM, UNPUB FACTOR ANAL US

[43]

NEAL RM, 1994, UNPUB BIOMETRIKA

[44]

Pearl, 1989, MORGAN KAUFMANN SERI, DOI DOI 10.1016/C2009-0-27609-4

[45]

PECE AEC, 1992, ARTIFICIAL NEURAL NETWORKS, 2, VOLS 1 AND 2, P865

[46]

Peterson C., 1987, Complex Systems, V1, P995

[47]

Rescorla R. A., 1972, Classical conditioning: current research and theory, P64, DOI [DOI 10.1101/GR.110528.110, 10.1101/gr.110528.110]

[48]

Rissanen J., 1989, STOCHASTIC COMPLEXIT

[49] EM ALGORITHMS FOR ML FACTOR-ANALYSIS [J].

RUBIN, DB ;

THAYER, DT .

PSYCHOMETRIKA, 1982, 47 (01) :69-76

[50]

Rumelhart David E., 1995, Backpropagation: Theory, architectures, and applications, P1

← 1 2 3 4 5 6 7 →