| Presentation | 2016-09-06 Hyper-parameter Optimization with Derivative-free Method Yoshihiko Ozaki, Masaki Yano, Masaki Onishi, Takahito Kuno, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In machine learning methods, an appropriate hyper-parameter tuning is really important for classifiers to perform its best. However, in general, the relationship between the performance of classifiers and their hyper-parameters is an unknown function. So the features with the objective function such as gradient are not available to optimize hyper-parameters. In this paper, we apply derivative-free optimization methods without the gradient information of the objective function in the hyper-parameter tuning of classifiers and evaluated their performance by computational experiments. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Derivative-Free Optimization / Machine Learning / Hyper-parameter Optimization / Coordinate-Search Method / Nelder-Mead Method / Support Vector Machine / Convolutional Neural Network |
| Paper # | PRMU2016-84,IBISML2016-39 |
| Date of Issue | 2016-08-29 (PRMU, IBISML) |
| Conference Information | |
| Committee | PRMU / IPSJ-CVIM / IBISML |
|---|---|
| Conference Date | 2016/9/5(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Eisaku Maeda(NTT) / / Kenji Fukumizu(ISM) |
| Vice Chair | Seiichi Uchida(Kyushu Univ.) / Hironobu Fujiyoshi(Chubu Univ.) / / Masashi Sugiyama(Univ. of Tokyo) / Hisashi Kashima(Kyoto Univ.) |
| Secretary | Seiichi Uchida(Kyoto Univ.) / Hironobu Fujiyoshi(NTT) / / Masashi Sugiyama(Univ. of Tokyo) / Hisashi Kashima(Nagoya Inst. of Tech.) |
| Assistant | Masaki Oonishi(AIST) / Takuya Funatomi(NAIST) / / Toshihiro Kamishima(AIST) / Tomoharu Iwata(NTT) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding / Special Interest Group on Computer Vision and Image Media / Technical Committee on Infomation-Based Induction Sciences and Machine Learning |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Hyper-parameter Optimization with Derivative-free Method |
| Sub Title (in English) | |
| Keyword(1) | Derivative-Free Optimization |
| Keyword(2) | Machine Learning |
| Keyword(3) | Hyper-parameter Optimization |
| Keyword(4) | Coordinate-Search Method |
| Keyword(5) | Nelder-Mead Method |
| Keyword(6) | Support Vector Machine |
| Keyword(7) | Convolutional Neural Network |
| 1st Author's Name | Yoshihiko Ozaki |
| 1st Author's Affiliation | University of Tsukuba/National Institute of Advanced Industrial Science and Technology(Univ. Tsukuba/AIST) |
| 2nd Author's Name | Masaki Yano |
| 2nd Author's Affiliation | University of Tsukuba/National Institute of Advanced Industrial Science and Technology(Univ. Tsukuba/AIST) |
| 3rd Author's Name | Masaki Onishi |
| 3rd Author's Affiliation | National Institute of Advanced Industrial Science and Technology(AIST) |
| 4th Author's Name | Takahito Kuno |
| 4th Author's Affiliation | University of Tsukuba(Univ. Tsukuba) |
| Date | 2016-09-06 |
| Paper # | PRMU2016-84,IBISML2016-39 |
| Volume (vol) | vol.116 |
| Number (no) | PRMU-208,IBISML-209 |
| Page | pp.pp.227-232(PRMU), pp.227-232(IBISML), |
| #Pages | 6 |
| Date of Issue | 2016-08-29 (PRMU, IBISML) |