Approximate Learning Algorithm for Restricted Boltzmann Machines

Author

Yasuda, Muneki ; Tanaka, Kazuyuki

Author_Institution

Grad. Sch. of Inf. Sci., Tohoku Univ., Sendai, Japan

fYear

2008

fDate

10-12 Dec. 2008

Firstpage

692

Lastpage

697

Abstract

A restricted Boltzmann machine consists of a layer of visible units and a layer of hidden units with no visible-visible or hidden-hidden connections. The restricted Boltzmann machine is the main component used in building up the deep belief network and has been studied by many researchers. However, the learning algorithm for the restricted Boltzmann machine is a NP-hard problem in general. In this paper we propose a new approximate learning algorithm for the restricted Boltzmann machines using the EM algorithm and the loopy belief propagation.

Keywords

Boltzmann machines; computational complexity; learning (artificial intelligence); NP-hard problem; approximate learning algorithm; belief network; hidden unit layer; learning algorithm; loopy belief propagation; restricted Boltzmann machines; visible unit layer; Bayesian methods; Belief propagation; Cost function; Inference algorithms; Machine learning; Machine learning algorithms; Markov random fields; NP-hard problem; Stochastic processes; Traveling salesman problems; Approximate learing algorithm; Belief propagation; EM algorithm; Machine learnings; Markov random field; Restricted Boltzmann machine;

fLanguage

English

Publisher

ieee

Conference_Titel

Computational Intelligence for Modelling Control & Automation, 2008 International Conference on

Conference_Location

Vienna

Print_ISBN

978-0-7695-3514-2

Type

conf

DOI

10.1109/CIMCA.2008.57

Filename

5172709