摘要
分析了强化学习与遗传算法工作机制,在提出基因空间分割概念的基础上,提出了一种将强化学习与遗传算法内在结合起来的算法RLGA,在遗传算法的框架下实现强化学习机制.从理论上分析了RLGA的收敛性,讨论了RLGA的时间和空间效率及其与基因空间分割的关系,通过实验分析了RLGA中基因空间分割的指导范围.实验结果表明,RLGA具有良好的全局收敛性能.
RLGA,an algorithm which implements mechanism of reinforcement learning under the framework of genetic algorithm is described, by using gene space division the algorithm maps the gene space of genetic algorithm into the strategy spcaces of multi-agent. The convergence theorems for the algorithm are presented, and the lime and the space efficiency of the algorithm as well as the relation between them and the division granularity are discussed. The experimental results show that RLGA has well global convergence performance, and the further experiments provide the guide range of the size of gene space division in RLGA.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2006年第5期856-860,866,共6页
Acta Electronica Sinica
基金
国家自然科学基金(No.60475026)
江苏省自然科学基金(No.BK2004079)
安徽省教育厅自然科学研究重点项目(No.2006KJ027A)
关键词
强化学习
遗传算法
收敛性
reinforcement learning
genetic algorithm
convergence
