arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2007.06184cs.LGstat.ML
收藏

具有弱线性函数逼近的大型MDP中的有效规划

Efficient Planning in Large MDPs with Weak Linear Function Approximation

University of Alberta(阿尔伯塔大学);Amii(阿尔伯塔机器智能研究所);DeepMind

Roshan Shariff, Csaba Szepesvári

摘要:

大规模马尔可夫决策过程(MDP)需要规划算法的运行时间独立的MDP的状态数。我们认为在MDP的规划问题,使用线性价值函数近似只有弱的要求:低近似误差的最佳价值函数,和一个小的“核心”状态,其功能跨越其他国家。特别是,我们不做任何假设的政策或非最优政策的价值函数的可代表性。我们的算法产生几乎最优的行动,为任何状态使用生成的预言机(模拟器)的MDP,而其计算时间尺度多项式的功能,核心状态,行动和有效的地平线的数量。

英文摘要:

Large-scale Markov decision processes (MDPs) require planning algorithms with runtime independent of the number of states of the MDP. We consider the planning problem in MDPs using linear value function approximation with only weak requirements: low approximation error for the optimal value function, and a small set of "core" states whose features span those of other states. In particular, we make no assumptions about the representability of policies or value functions of non-optimal policies. Our algorithm produces almost-optimal actions for any state using a generative oracle (simulator) for the MDP, while its computation time scales polynomially with the number of features, core states, and actions and the effective horizon.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑