文章基本信息

标题：Improved Automatic Discovery of Subgoals for Options in Hierarchical Reinforcement Learning
本地全文：下载
作者：R. Matthew Kretchmar ; Todd Feil ; Rohit Bansal 等
期刊名称：Journal of Computer Science and Technology
印刷版ISSN：1666-6046
电子版ISSN：1666-6038
出版年度：2003
卷号：3
期号：2
出版社：Iberoamerican Science & Technology Education Consortium
摘要：Options have been shown to be a key step in extendingreinforcement learning beyond low-level reactionary sys-tems to higher-level, planning systems. Most of the op-tions research involves hand-crafted options; there hasbeen only very limited work in the automated discoveryof options. We extend early work in automated optiondiscovery with a .exible and robust method
关键词：reinforcement learning; options; subgoal;discovery