A Robust Offline Reinforcement Learning Algorithm Based on Behavior Regularization Methods

Yan Zhang, Tianhan Gao, Qingwei Mi
{"title":"A Robust Offline Reinforcement Learning Algorithm Based on Behavior Regularization Methods","authors":"Yan Zhang, Tianhan Gao, Qingwei Mi","doi":"10.1109/IAICT55358.2022.9887435","DOIUrl":null,"url":null,"abstract":"Offline deep reinforcement learning algorithms are still in developing. Some existing algorithms have shown that it is feasible to learn directly without using environmental interaction under the condition of sufficient datasets. In this paper, we combine an offline reinforcement learning method through behavior regularization with a robust offline reinforcement learning algorithm. Moreover, the algorithm is verified and analyzed with a high-quality but limited dataset. The experimental results show that it is feasible to combine the behavior regularization method with the robust offline reinforcement learning algorithm, to gain better performance under the condition of limited data compared with the baseline algorithms.","PeriodicalId":154027,"journal":{"name":"2022 IEEE International Conference on Industry 4.0, Artificial Intelligence, and Communications Technology (IAICT)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-07-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE International Conference on Industry 4.0, Artificial Intelligence, and Communications Technology (IAICT)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/IAICT55358.2022.9887435","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

Abstract

Offline deep reinforcement learning algorithms are still in developing. Some existing algorithms have shown that it is feasible to learn directly without using environmental interaction under the condition of sufficient datasets. In this paper, we combine an offline reinforcement learning method through behavior regularization with a robust offline reinforcement learning algorithm. Moreover, the algorithm is verified and analyzed with a high-quality but limited dataset. The experimental results show that it is feasible to combine the behavior regularization method with the robust offline reinforcement learning algorithm, to gain better performance under the condition of limited data compared with the baseline algorithms.
基于行为正则化方法的鲁棒离线强化学习算法
离线深度强化学习算法仍在发展中。已有的一些算法表明,在数据集充足的情况下,不使用环境交互直接学习是可行的。本文将一种基于行为正则化的离线强化学习方法与一种鲁棒离线强化学习算法相结合。最后,用高质量但有限的数据集对算法进行了验证和分析。实验结果表明,将行为正则化方法与鲁棒离线强化学习算法相结合是可行的,在数据有限的情况下,与基线算法相比,可以获得更好的性能。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信