Empirical Algorithms for General Stochastic Systems with Continuous States and Actions
Empirical Algorithms for General Stochastic Systems with Continuous States and Actions
复制标题
DOI：
10.1109/cdc40024.2019.9029308
复制
发表时间：
2019-12
期刊：
影响因子：
--
通讯作者：
Hiteshi Sharma;R. Jain;W. Haskell
中科院分区：
文献类型：
--
作者：
Hiteshi Sharma;R. Jain;W. Haskell
在本文中，我们提出了具有连续状态和动作空间的 MDP 的随机经验值学习（RAEVL）算法。该算法结合了动作空间上的随机搜索思想和随机函数逼近方法来推广状态空间上的值函数。我们的理论分析是在随机算子框架下结合随机优势论证进行的。这提供了所提出算法的有限时间分析并给出了样本复杂性。
In this paper, we present Randomized Empirical Value Learning (RAEVL) algorithm for MDPs with continuous state and action spaces. This algorithm combines the ideas of random search over action space with randomized function approximation method to generalize the value functions over state space . Our theoretical analysis is done under a random operator framework combined with stochastic dominance argument. This provides finite-time analysis of the proposed algorithm as well as give the sample complexity.