| Presentation | 2019-03-04 Adjustment of exploratory behavior using mutual information in reinforcement learning Kaiji Koyama, Jun Ohkubo, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | One of the important problems in reinforcement learning is the exploration-exploitation trade-off. In this research, we propose a method to use mutual information as a exploration bonus in experimental settings with sudden environmental change; for example, we consider a maze problem in which walls will suddenly appear or disappear. As for the environmental changes, there are some previous researches such as the usage of entropy as exploration bonus and a meta-parameter control method in Boltzmann selection rule. Here, the proposed method using the mutual information is implemented in the Q learning, including the meta-parameter control method, and numerical experiments are performed. The numerical results show that the mutual information can work well as the exploration bonus. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | reinforcement learning / mutual information |
| Paper # | NC2018-51 |
| Date of Issue | 2019-02-25 (NC) |
| Conference Information | |
| Committee | NC / MBE |
|---|---|
| Conference Date | 2019/3/4(3days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | University of Electro Communications |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Yutaka Hirata(Chubu Univ.) / Masaki Kyoso(TCU) |
| Vice Chair | Hayaru Shouno(UEC) / Taishin Nomura(Osaka Univ.) |
| Secretary | Hayaru Shouno(Nagoya Univ.) / Taishin Nomura(NAIST) |
| Assistant | Keiichiro Inagaki(Chubu Univ.) / Takashi Shinozaki(NICT) / Takumi Kobayashi(YNU) / Ｙａｓｕｙｕｋｉ Ｓｕｚｕｋｉ(Osaka Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Neurocomputing / Technical Committee on ME and Bio Cybernetics |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Adjustment of exploratory behavior using mutual information in reinforcement learning |
| Sub Title (in English) | |
| Keyword(1) | reinforcement learning |
| Keyword(2) | mutual information |
| 1st Author's Name | Kaiji Koyama |
| 1st Author's Affiliation | Saitama University(Saitama Univ.) |
| 2nd Author's Name | Jun Ohkubo |
| 2nd Author's Affiliation | Saitama University(Saitama Univ.) |
| Date | 2019-03-04 |
| Paper # | NC2018-51 |
| Volume (vol) | vol.118 |
| Number (no) | NC-470 |
| Page | pp.pp.43-47(NC), |
| #Pages | 5 |
| Date of Issue | 2019-02-25 (NC) |