摘要
强化学习是机器学习领域的一个重要分支,但在强化学习系统中,学习的数量会随着状态变量的个数成指数级增长,从而形成"维数灾"。为此提出了一种基于MAXQ的分层强化学习方法,通过引入抽象机制将强化学习任务分解到不同层次上来分别实现,使得每层上的学习任务仅需在较小的空间中进行,从而大大减少了学习的数量和规模。并给出具体算法——MAXQ-RLA。
Reinforcement learning is an important branch of machine learning. In the system of reinforcement learning,the learning stategies increase exponentially along with the number of state variables, which is called ＂dimensions disaster＂. Here a hierarchical reinforcement learning based on the MAXQ is proposed to solve this problem,which is realized by decomposing the task to different level,thus sub - tasks in every level can be solved in relatively smaller scale. This method turns out to be effective to decrease the stategies. Finally,offer the concerned algorithm-MAXQ- RLA.
出处
《计算机技术与发展》
2009年第4期154-156,169,共4页
Computer Technology and Development
基金
安徽省教育重点项目(KJ2008A142C)
安徽省自然科学基金项目(KJ2007B061)
