DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Hwang, Sung Ju | - |
dc.contributor.advisor | 황성주 | - |
dc.contributor.author | Nam, Taewook | - |
dc.date.accessioned | 2021-05-13T19:38:14Z | - |
dc.date.available | 2021-05-13T19:38:14Z | - |
dc.date.issued | 2020 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=925156&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/284995 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전산학부, 2020.8,[ii, 15 p. :] | - |
dc.description.abstract | As active research of deep reinforcement learning makes it possible to apply reinforcement learning to many high-dimensional environments, the sample efficiency of reinforcement learning has been more important. Learning strategy that utilizes background knowledge from previous tasks to new tasks, such as transfer learning and meta-learning, is one common approach for enhancing sample efficiency. In this work, we propose a meta-learning framework Meta-Distillation for Reinforcement Learning (MDRL) that efficiently transfers expert policies from previous environments to a new policy in an unseen environment. A weighted sum of discrepancies between current policy and expert policies is added to policy update loss, and the weights are determined by a weight network that is meta-trained to help training by considering tasks, training sample, and policy training progress. MDRL succeed to data-efficiently adapt new task when given distribution of environment is scarce and diverse. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Reinforcement Learning▼aMeta-Learning▼aMeta-RL▼aTransfer Learning▼aDistillation | - |
dc.subject | 강화 학습▼a메타 학습▼a메타 강화 학습▼a전이 학습▼a디스틸레이션 | - |
dc.title | Meta distillation for reinforcement learning | - |
dc.title.alternative | 강화 학습을 위한 메타 디스틸레이션 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전산학부, | - |
dc.contributor.alternativeauthor | 남태욱 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.