Presentation | 1999/12/16 Variational Bayesian Learning with Split and Merge Operations : From Model Selection to Model Search Naonori UEDA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | When learning a nonlinear model, we are confronted by two difficulties in practice: (1) the local optimal, and (2) appropriate model complexity determination problems. As for (1), I recently proposed the split and merge EM algorithm within the framework of the maximum likelihood by simulataneously spliting and merging model components, but the model complexity was fixed there. As for (2), it can be thought that the conventional information criteria such as AIC are available. In the case of nonlinear models, however, since the asymptotic normality assumption of the maximum likelihood estimate, in general, does not hold, they do not work well in practice. In this report, I propose a new learning method to simultaneously solve both (1) and (2) problems by introducing the split and merge operations to variational Bayes learning framework. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Bayes learning / split and merge / model search. |
Paper # | PRMU99-174 |
Date of Issue |
Conference Information | |
Committee | PRMU |
---|---|
Conference Date | 1999/12/16(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 | Pattern Recognition and Media Understanding (PRMU) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Variational Bayesian Learning with Split and Merge Operations : From Model Selection to Model Search |
Sub Title (in English) | |
Keyword(1) | Bayes learning |
Keyword(2) | split and merge |
Keyword(3) | model search. |
1st Author's Name | Naonori UEDA |
1st Author's Affiliation | NTT Communication Science Laboratories() |
Date | 1999/12/16 |
Paper # | PRMU99-174 |
Volume (vol) | vol.99 |
Number (no) | 514 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |