| Presentation | 2018-09-21 A Study on MDL Learning of Mixture Families Kohei Miyamoto, Jun'ichi Takeuchi, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | MDL estimators based on two stage codes are studied. The Barron and Cover's theory shows that redundancy ofa two stage code bounds the risk of the MDL estimatordefined by the two stage code. It is known thatcoding regret of two stage codes for exponential familiescan be close to the minimax regret, which is related to stochastic complexity, while it is not for non-exponential families. I this paper, we propose a new method of two stage codes for two componentsmixture families, whose regretcan be close to the minimax regret. The method is based on local exponential family bundles, which are used to evaluate the minimax regret for non-exponential families. Further, we obtain a tight risk bound of MDL estimators for mixture families. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | MDL principle / two stage codes / mixture families |
| Paper # | PRMU2018-51,IBISML2018-28 |
| Date of Issue | 2018-09-13 (PRMU, IBISML) |
| Conference Information | |
| Committee | PRMU / IBISML / IPSJ-CVIM |
|---|---|
| Conference Date | 2018/9/20(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Shinichi Sato(NII) / Hisashi Kashima(Kyoto Univ.) |
| Vice Chair | Yoshihisa Ijiri(Omron) / Toru Tamaki(Hiroshima Univ.) / Masashi Sugiyama(Univ. of Tokyo) / Koji Tsuda(Univ. of Tokyo) |
| Secretary | Yoshihisa Ijiri(NEC) / Toru Tamaki(Osaka Univ.) / Masashi Sugiyama(Nagoya Inst. of Tech.) / Koji Tsuda(AIST) |
| Assistant | Go Irie(NTT) / Yoshitaka Ushiku(Univ. of Tokyo) / Tomoharu Iwata(NTT) / Shigeyuki Oba(Kyoto Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding / Technical Committee on Infomation-Based Induction Sciences and Machine Learning / Special Interest Group on Computer Vision and Image Media |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A Study on MDL Learning of Mixture Families |
| Sub Title (in English) | |
| Keyword(1) | MDL principle |
| Keyword(2) | two stage codes |
| Keyword(3) | mixture families |
| 1st Author's Name | Kohei Miyamoto |
| 1st Author's Affiliation | Kyushu University(Kyushu Univ.) |
| 2nd Author's Name | Jun'ichi Takeuchi |
| 2nd Author's Affiliation | Kyushu University(Kyushu Univ.) |
| Date | 2018-09-21 |
| Paper # | PRMU2018-51,IBISML2018-28 |
| Volume (vol) | vol.118 |
| Number (no) | PRMU-219,IBISML-220 |
| Page | pp.pp.105-112(PRMU), pp.105-112(IBISML), |
| #Pages | 8 |
| Date of Issue | 2018-09-13 (PRMU, IBISML) |