WEKO3
-
RootNode
アイテム
How do we define the reward in reinforcement learning?
https://ipsj.ixsq.nii.ac.jp/records/142439
https://ipsj.ixsq.nii.ac.jp/records/1424398413d7cd-4471-41f9-8859-06834eb48db7
名前 / ファイル | ライセンス | アクション |
---|---|---|
![]() |
Copyright (c) 2015 by the Information Processing Society of Japan
|
|
オープンアクセス |
Item type | SIG Technical Reports(1) | |||||||
---|---|---|---|---|---|---|---|---|
公開日 | 2015-06-16 | |||||||
タイトル | ||||||||
タイトル | How do we define the reward in reinforcement learning? | |||||||
タイトル | ||||||||
言語 | en | |||||||
タイトル | How do we define the reward in reinforcement learning? | |||||||
言語 | ||||||||
言語 | jpn | |||||||
キーワード | ||||||||
主題Scheme | Other | |||||||
主題 | 招待講演 | |||||||
資源タイプ | ||||||||
資源タイプ識別子 | http://purl.org/coar/resource_type/c_18gh | |||||||
資源タイプ | technical report | |||||||
著者所属 | ||||||||
沖縄科学技術大学院大学神経計算ユニット | ||||||||
著者名 |
内部, 英治
× 内部, 英治
|
|||||||
論文抄録 | ||||||||
内容記述タイプ | Other | |||||||
内容記述 | In application of reinforcement learning algorithms to real world problems, the design of reward functions is critical for successful achievement of a task. We may have only a very rough idea of the reward function whose optimization would generate desirable behavior, so straightforward reinforcement learning may not be usable. To find a good reward function, two approaches are considered. One is inverse reinforcement learning which infer the reward function from observed behaviors which are usually assumed to be optimal. The other approach is so-called intrinsically motivated reinforcement learning, in which the agent learns behaviors from extrinsic rewards from the environment and intrinsic rewards calculated by the agent based on information theory, emotion, task complexity, and so on. This talk briefly introduces inverse reinforcement learning and intrinsically motivated reinforcement learning. Next, we will explain our methods for those problems: inverse reinforcement learning with density ratio estimation and constrained policy gradient for intrinsic and extrinsic rewards. | |||||||
書誌レコードID | ||||||||
収録物識別子タイプ | NCID | |||||||
収録物識別子 | AN10505667 | |||||||
書誌情報 |
研究報告数理モデル化と問題解決(MPS) 巻 2015-MPS-103, 号 38, p. 1-1, 発行日 2015-06-16 |
|||||||
ISSN | ||||||||
収録物識別子タイプ | ISSN | |||||||
収録物識別子 | 2188-8833 | |||||||
Notice | ||||||||
SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc. | ||||||||
出版者 | ||||||||
言語 | ja | |||||||
出版者 | 情報処理学会 |