Knowledge Commons of Institute of Automation,CAS
PiCor: Multi-Task Deep Reinforcement Learning with Policy Correction | |
Bai FS(白丰硕)1,2![]() ![]() ![]() ![]() | |
2023-06 | |
会议名称 | the AAAI Conference on Artificial Intelligence |
会议录名称 | Proceedings of the AAAI Conference on Artificial Intelligence |
卷号 | 37 |
期号 | 6 |
页码 | 6728-6736 |
会议日期 | 2023.02.07 - 2023.02.14 |
会议地点 | 美国 华盛顿 |
会议举办国 | 美国 |
会议录编者/会议主办者 | Brian Williams ; Sara Bernardini ; Yiling Chen ; Jennifer Neville |
出版地 | 美国 |
产权排序 | 1 |
摘要 | Multi-task deep reinforcement learning (DRL) ambitiously aims to train a general agent that masters multiple tasks simultaneously. However, varying learning speeds of different tasks compounding with negative gradient interference makes policy learning inefficient. In this work, we propose PiCor, an efficient multi-task DRL framework that splits learning into policy optimization and policy correction phases. The policy optimization phase improves the policy by any DRL algothrim on the sampled single task without considering other tasks. The policy correction phase first constructs a performance constraint set with adaptive weight adjusting. Then the intermediate policy learned by the first phase is constrained to the set, which controls the negative interference and balances the learning speeds across tasks. Empirically, we demonstrate that PiCor outperforms previous methods and significantly improves sample efficiency on simulated robotic manipulation and continuous control tasks. We additionally show that adaptive weight adjusting can further improve data efficiency and performance. |
关键词 | Reinforcement Learning Algorithms Transfer Domain Adaptation Multi-Task Learning |
学科领域 | 计算机科学技术 ; 人工智能 ; 人工智能其他学科 |
学科门类 | 工学 |
DOI | https://doi.org/10.1609/aaai.v37i6.25825 |
URL | 查看原文 |
收录类别 | EI |
语种 | 英语 |
WOS研究方向 | 机器学习 |
WOS类目 | 人工智能 |
是否为代表性论文 | 是 |
七大方向——子方向分类 | 人工智能基础理论 |
国重实验室规划方向分类 | 人工智能基础前沿理论 |
是否有论文关联数据集需要存交 | 否 |
引用统计 | |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/52322 |
专题 | 复杂系统认知与决策实验室_听觉模型与认知计算 |
通讯作者 | Bai FS(白丰硕) |
作者单位 | 1.School of Artificial Intelligence, University of Chinese Academy of Sciences 2.Institute of Automation, Chinese Academy of Sciences (CASIA) 3.University of Alberta 4.Université Paris-Saclay |
第一作者单位 | 中国科学院自动化研究所 |
通讯作者单位 | 中国科学院自动化研究所 |
推荐引用方式 GB/T 7714 | Bai FS,Zhang HM,Tao TY,et al. PiCor: Multi-Task Deep Reinforcement Learning with Policy Correction[C]//Brian Williams, Sara Bernardini, Yiling Chen, Jennifer Neville. 美国,2023:6728-6736. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
PiCor final.pdf(1663KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论