Presentation 2006-03-17
Stochastic Complexity of Complete Bibpartite Graph-type Boltzmann Machines in Mean Field Approximation
Yu NISHIYAMA, Sumio WATANABE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In the learning of singular learning machines, the superiority of Bayesian learning is shown. However, it requires huge computational costs to realize the Bayesian a posteriori distribution. To overcome this problem, the mean field approximation, which is originally known in statistical physics, is used in the practical information systems. Recently, the theoretical properties such as generalization error or free energy in the mean field approximation has been studied. The theoretical results give us the comparison with the regular statistical model and the foundation of a model selection. In this paper, we treat the complete bibpartite Boltzmann machines and derive the upper bound of asymptotic free energy of the mean field approximation.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Singular Learning Machines / Stochastic Complexity / Mean Field Approximation / Boltzmann Machines
Paper # NC2005-172
Date of Issue

Conference Information
Committee NC
Conference Date 2006/3/10(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) Stochastic Complexity of Complete Bibpartite Graph-type Boltzmann Machines in Mean Field Approximation
Sub Title (in English)
Keyword(1) Singular Learning Machines
Keyword(2) Stochastic Complexity
Keyword(3) Mean Field Approximation
Keyword(4) Boltzmann Machines
1st Author's Name Yu NISHIYAMA
1st Author's Affiliation Department of Computational Intelligence and System Science, Tokyo Institute of Technology()
2nd Author's Name Sumio WATANABE
2nd Author's Affiliation Precision and Intelligence Laboratory. Tokyo Institute of Technology
Date 2006-03-17
Paper # NC2005-172
Volume (vol) vol.105
Number (no) 659
Page pp.pp.-
#Pages 6
Date of Issue