Presentation | 2012-11-08 Considering the multiplicity in mixture model brings better results Tetsuo FURUKAWA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | The purpose of this work is to re-examine the probabilistic model of the mixture model, and to derive the algorithm by the variational Bayesian method. In the mixture model with K components, there exist K! equivalent solutions with respect to the permutation of the components. Usually we only need to obtain one of those solutions, and the others can be ignored. However, those equivalent solutions occasionally interfere each other when we apply the variational Bayesian (VB), and it causes the local miminum problem. In this work, we derived the probabilistic model which considers the equivalent solutions. The obtained algorithm is more robust to the local optimum, while the calculation cost is as low as the conventional one. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Mixture Model / EM algorithm / variational Bayesian / local optimum solution / nonidentifiability problem / label identifiability |
Paper # | IBISML2012-89 |
Date of Issue |
Conference Information | |
Committee | IBISML |
---|---|
Conference Date | 2012/10/31(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 | Information-Based Induction Sciences and Machine Learning (IBISML) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Considering the multiplicity in mixture model brings better results |
Sub Title (in English) | |
Keyword(1) | Mixture Model |
Keyword(2) | EM algorithm |
Keyword(3) | variational Bayesian |
Keyword(4) | local optimum solution |
Keyword(5) | nonidentifiability problem |
Keyword(6) | label identifiability |
1st Author's Name | Tetsuo FURUKAWA |
1st Author's Affiliation | Department of Brain Science and Engineering Kyushu Institute of Technology() |
Date | 2012-11-08 |
Paper # | IBISML2012-89 |
Volume (vol) | vol.112 |
Number (no) | 279 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |