| Presentation | 2012-03-12 Hyperparameter Selection of Infinite Gaussian Mixture Model via Widely Applicable Information Criterion Takushi MIKI, Masahiro KOHJIMA, Sumio WATANABE, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Recently, nonparametric Bayesian method is applied to wide range of research fields such as natural language processing, object tracking, and recommendation system. Major advantage of this method is that the complexity of the model need not to be explicitly decided and is implicitly controlled by hyperparameters. Although the choice of hyperparameters has a considerable effect to learning results and predictive distribution, the design of hyperparameters is not yet studied. In this paper, authors propose hyperparameter design method based on Widely Applicable Information Criterion(WAIC), which has a theoretical support in general parametric models. Experimental results show the effectiveness of proposed method. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | nonparametric Bayesian method / infinite gaussian mixture / hyperparameter selection / information criterion / widely applicable information criterion |
| Paper # | IBISML2011-91 |
| Date of Issue |
| Conference Information | |
| Committee | IBISML |
|---|---|
| Conference Date | 2012/3/5(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) | Hyperparameter Selection of Infinite Gaussian Mixture Model via Widely Applicable Information Criterion |
| Sub Title (in English) | |
| Keyword(1) | nonparametric Bayesian method |
| Keyword(2) | infinite gaussian mixture |
| Keyword(3) | hyperparameter selection |
| Keyword(4) | information criterion |
| Keyword(5) | widely applicable information criterion |
| 1st Author's Name | Takushi MIKI |
| 1st Author's Affiliation | Department of Computational Intelligence and Systems Science, Tokyo Institute of Technology() |
| 2nd Author's Name | Masahiro KOHJIMA |
| 2nd Author's Affiliation | Department of Computational Intelligence and Systems Science, Tokyo Institute of Technology |
| 3rd Author's Name | Sumio WATANABE |
| 3rd Author's Affiliation | Department of Computational Intelligence and Systems Science, Tokyo Institute of Technology |
| Date | 2012-03-12 |
| Paper # | IBISML2011-91 |
| Volume (vol) | vol.111 |
| Number (no) | 480 |
| Page | pp.pp.- |
| #Pages | 5 |
| Date of Issue | |