Presentation 2006-03-15
A mean field algorithm for Bayesian learning in perceptron type probabilistic models
Shinsuke UDA, Yoshiyuki KABASHIMA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We develop a computationally tractable approximate algorithm for Bayesian learning in perceptron type probabilistic models. The algorithm is designed to efficiently solve the adaptive TAP equations, recently introduced by Opper and Winther, based on message passing scheme. Unlike the conventional method, the adaptive TAP scheme adaptively approximates the Bayesian inference without knowledge of the generative model of given data, which is preferred in analysis of real world data. The obtained algorithm is applied to the two tasks; (a) a classification problem using real world data and (b) multiuser detection for code division multiple access (CDMA) communication. Numerical experiments for the two tasks indicate the efficacy of the developed algorithm.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Adaptive TAP mean field method / Bayesian learning / Belief propagation / Classification / CDMA communication
Paper # NC2005-114
Date of Issue

Conference Information
Committee NC
Conference Date 2006/3/8(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) A mean field algorithm for Bayesian learning in perceptron type probabilistic models
Sub Title (in English)
Keyword(1) Adaptive TAP mean field method
Keyword(2) Bayesian learning
Keyword(3) Belief propagation
Keyword(4) Classification
Keyword(5) CDMA communication
1st Author's Name Shinsuke UDA
1st Author's Affiliation Graduate School of Science and Engineering, Tokyo Institute of Technology()
2nd Author's Name Yoshiyuki KABASHIMA
2nd Author's Affiliation Graduate School of Science and Engineering, Tokyo Institute of Technology
Date 2006-03-15
Paper # NC2005-114
Volume (vol) vol.105
Number (no) 657
Page pp.pp.-
#Pages 6
Date of Issue