Presentation | 2021-02-22 A Study on the Application of Curriculum Learning in Deep Reinforcement Learning Ikumi Kodaka, Fumiaki Saito, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Deep reinforcement learning is attracting attention because it can be applied to higher-dimensional environments compared to conventional reinforcement learning. However, an important issue is to increase the number of trials required for action acquisition, particularly in high-dimensional and sparsely rewarded tasks. Therefore, in this study, we applied curriculum learning, which improves learning performance by gradually changing the difficulty level of tasks, in the action acquisition in a shooting game AI. Through experimental evaluation, we verified the speeding up of action acquisition and considered the transition of difficulty and its efficiency. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Deep Reinforcement Learning / Curriculum Learning / Deep Q-Network / Game AI |
Paper # | AI2020-47 |
Date of Issue | 2021-02-15 (AI) |
Conference Information | |
Committee | AI |
---|---|
Conference Date | 2021/2/22(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Online |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Web/IoT Intelligence, etc. |
Chair | Naoki Fukuta(Shizuoka Univ.) |
Vice Chair | Yuichi Sei(Univ. of Electro-Comm.) / Yuko Sakurai(AIST) |
Secretary | Yuichi Sei(Nagoya Inst. of Tech.) / Yuko Sakurai(Tokyo Univ. of Agriculture and Technology) |
Assistant |
Paper Information | |
Registration To | Technical Committee on Artificial Intelligence and Knowledge-Based Processing |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | A Study on the Application of Curriculum Learning in Deep Reinforcement Learning |
Sub Title (in English) | action acquisition in shooting game AI as an example |
Keyword(1) | Deep Reinforcement Learning |
Keyword(2) | Curriculum Learning |
Keyword(3) | Deep Q-Network |
Keyword(4) | Game AI |
1st Author's Name | Ikumi Kodaka |
1st Author's Affiliation | Chiba Institute of Technology(CIT) |
2nd Author's Name | Fumiaki Saito |
2nd Author's Affiliation | Chiba Institute of Technology(CIT) |
Date | 2021-02-22 |
Paper # | AI2020-47 |
Volume (vol) | vol.120 |
Number (no) | AI-379 |
Page | pp.pp.47-52(AI), |
#Pages | 6 |
Date of Issue | 2021-02-15 (AI) |