Implementation of Language-Action Reward Network in Reinforcement Learning by Using Natural Language

S. Keerthi
{"title":"Implementation of Language-Action Reward Network in Reinforcement Learning by Using Natural Language","authors":"S. Keerthi","doi":"10.1109/INDISCON50162.2020.00026","DOIUrl":null,"url":null,"abstract":"Key issue in many RL approaches is deferred (long-term and postponed) rewards, this leads to difficulties in learning for an agent. Inspite of fact certain “reward shaping” deals with goal of making process of learning quick and easy (by starting process with an agent assigned with extra input), it leads to come complexity. Ongoing RL approaches have indicated complexity of implementation (example Atari games). So, to avoid such issues Potential-based reward shaping (PBRS) is used to increase performance of RL gents. It is an adaptable strategy to provide fundamental foundation data combining with temporal (time-based) distinction learning in a principled manner. In this PBRS approach, we propose a system LEARN (LanguagEAction Reward Network), certain maps common language (human understandable) to middle (intermediate) rewards based on activities of agent. These intermediate language based rewards canister be integrated (combined) into any standard RL algorithm. Experiments abide run on a grid world and a more complex LanguagE-Action Reward Network (LEARN) a framework certain show certain we canister learn tasks significantly faster when we specify intuitive priors on reward distribution.","PeriodicalId":371571,"journal":{"name":"2020 IEEE India Council International Subsections Conference (INDISCON)","volume":"76 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 IEEE India Council International Subsections Conference (INDISCON)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/INDISCON50162.2020.00026","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

Key issue in many RL approaches is deferred (long-term and postponed) rewards, this leads to difficulties in learning for an agent. Inspite of fact certain “reward shaping” deals with goal of making process of learning quick and easy (by starting process with an agent assigned with extra input), it leads to come complexity. Ongoing RL approaches have indicated complexity of implementation (example Atari games). So, to avoid such issues Potential-based reward shaping (PBRS) is used to increase performance of RL gents. It is an adaptable strategy to provide fundamental foundation data combining with temporal (time-based) distinction learning in a principled manner. In this PBRS approach, we propose a system LEARN (LanguagEAction Reward Network), certain maps common language (human understandable) to middle (intermediate) rewards based on activities of agent. These intermediate language based rewards canister be integrated (combined) into any standard RL algorithm. Experiments abide run on a grid world and a more complex LanguagE-Action Reward Network (LEARN) a framework certain show certain we canister learn tasks significantly faster when we specify intuitive priors on reward distribution.
语言-动作奖励网络在自然语言强化学习中的实现
许多强化学习方法中的关键问题是延迟(长期和延迟)奖励,这导致智能体学习困难。尽管某些“奖励塑造”的目标是使学习过程变得快速和简单(通过分配额外输入的代理开始学习过程),但它会导致复杂性。正在进行的强化学习方法表明了实现的复杂性(例如Atari游戏)。因此,为了避免这些问题,基于潜在的奖励塑造(PBRS)被用于提高RL代理的性能。有原则地结合时间(基于时间的)区分学习提供基础数据是一种适应性强的策略。在这种PBRS方法中,我们提出了一个系统LEARN (LanguagEAction Reward Network),它根据智能体的活动将公共语言(人类可理解的)映射到中间(中间)奖励。这些基于中间语言的奖励可以集成到任何标准的强化学习算法中。实验是在一个网格世界和一个更复杂的语言-行动奖励网络(LEARN)框架下进行的,这个框架表明,当我们指定奖励分配的直觉先验时,我们可以明显更快地学习任务。
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