Presentation 2001/2/1
Function approximation by on-line variational Bayes learning
Shin Ishii, Masa-aki Sato,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We formerly proposed an on-line EM algorithm for the normalized Gaussian network model. Athough the algorithm conducts on-line model selection schemes based on probabilistic interpretation, ambiguity has remained in their criterion. This study intends to remove the ambiguity by using a Bayes learning method. We propose an on-line variational Bayes learning method, in which the Bayes learning for the NGnet is implemented as a similar algorithm to the on-line EM algorithm. When applied to a simple two-dimensional function approximation problem and eight-dimensional benchmark problems, our method exhibits good performance.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Variational Bayes learning / EM algorithm / On-line learning / Normalized Gaussian network / Model selection
Paper # NC2000-90
Date of Issue

Conference Information
Committee NC
Conference Date 2001/2/1(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) Function approximation by on-line variational Bayes learning
Sub Title (in English)
Keyword(1) Variational Bayes learning
Keyword(2) EM algorithm
Keyword(3) On-line learning
Keyword(4) Normalized Gaussian network
Keyword(5) Model selection
1st Author's Name Shin Ishii
1st Author's Affiliation Nara Institute of Science and Technology:CREST Doya Project, Japan Science and Technology Corporation()
2nd Author's Name Masa-aki Sato
2nd Author's Affiliation Advanced Telecommunication Research Institute International(ATR):CREST Doya Project, Japan Science and Technology Corporation
Date 2001/2/1
Paper # NC2000-90
Volume (vol) vol.100
Number (no) 617
Page pp.pp.-
#Pages 8
Date of Issue