Presentation | 2018-12-07 Evolutionary Multitask Deep Reinforcement Learning in 2D Maze Task Shota Imai, Yuichi Sei, Yasuyuki Tahara, Akihiko Ohsuga, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In Deep reinforcement learning, it is difficult to converge when the exploration is insufficient or a reward is sparse. Besides, in a specific task, the number of exploration may be limited. Therefore, it is considered effective to learn in source tasks previously to promote leaning in the target tasks. In this research, we propose a method to train a model that can work well on variety of target tasks with Evolutionary Computation in source task. In this method, agents explore multiple environments with diverse set of neural net works to train a general model. In the experiments, we assume multiple maze source tasks. After the model training with our method in the source tasks, we shows how effective the model is for the maze tasks of the target tasks |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Reinforcement Learning / Deep Learning / Deep Reinforcement Learning / Evolutionary Computation / Multitask Leaning / Neuroevolution |
Paper # | AI2018-29 |
Date of Issue | 2018-11-30 (AI) |
Conference Information | |
Committee | AI |
---|---|
Conference Date | 2018/12/7(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | Tsunenori Mine(Kyushu Univ.) |
Vice Chair | Daisuke Katagami(Tokyo Polytechnic Univ.) / Naoki Fukuta(Shizuoka Univ.) |
Secretary | Daisuke Katagami(Ritsumeikan Univ.) / Naoki Fukuta(Univ. of Electro-Comm.) |
Assistant | Yuko Sakurai(AIST) |
Paper Information | |
Registration To | Technical Committee on Artificial Intelligence and Knowledge-Based Processing |
---|---|
Language | JPN-ONLY |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Evolutionary Multitask Deep Reinforcement Learning in 2D Maze Task |
Sub Title (in English) | |
Keyword(1) | Reinforcement Learning |
Keyword(2) | Deep Learning |
Keyword(3) | Deep Reinforcement Learning |
Keyword(4) | Evolutionary Computation |
Keyword(5) | Multitask Leaning |
Keyword(6) | Neuroevolution |
1st Author's Name | Shota Imai |
1st Author's Affiliation | The University of Electro-Communications(UEC) |
2nd Author's Name | Yuichi Sei |
2nd Author's Affiliation | The University of Electro-Communications(UEC) |
3rd Author's Name | Yasuyuki Tahara |
3rd Author's Affiliation | The University of Electro-Communications(UEC) |
4th Author's Name | Akihiko Ohsuga |
4th Author's Affiliation | The University of Electro-Communications(UEC) |
Date | 2018-12-07 |
Paper # | AI2018-29 |
Volume (vol) | vol.118 |
Number (no) | AI-350 |
Page | pp.pp.19-24(AI), |
#Pages | 6 |
Date of Issue | 2018-11-30 (AI) |