Presentation 1998/3/19
On Boltzmann Machine Learning Algorithms Using Fisher Information Matrices
Tetsuya Kojima,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Boltzmann machine learning is an algorithm which approximates the given probability distribution with its stochastically equilibrium distribution, but it requires much computational time in computer simulations. From the information-theoretical point of view, it has been known that Fisher information matrices make natural parameters of the Boltzmann distribution converge straight the most likelihood estimator of the given probability distribution. In this paper, numerical studies show some properties of the Boltzmann machine learning using Fisher information matrices in both cases where they have all of the elements and where only the diagonal elements are contained in them. Introduction to associative memory model is also presented.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Boltzmann machines / learning algorithms / Fisher information matrix / exponential family / associative memory model
Paper #
Date of Issue

Conference Information
Committee NC
Conference Date 1998/3/19(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) On Boltzmann Machine Learning Algorithms Using Fisher Information Matrices
Sub Title (in English)
Keyword(1) Boltzmann machines
Keyword(2) learning algorithms
Keyword(3) Fisher information matrix
Keyword(4) exponential family
Keyword(5) associative memory model
1st Author's Name Tetsuya Kojima
1st Author's Affiliation Graduate School of Information Systems, University of Electro-Communications()
Date 1998/3/19
Paper #
Volume (vol) vol.97
Number (no) 623
Page pp.pp.-
#Pages 8
Date of Issue