Policy Gradient in Partially Observable Environments: Approximation and Convergence
arXiv:1810.07900
Abstract
Policy gradient is a generic and flexible reinforcement learning approach that generally enjoys simplicity in analysis, implementation, and deployment. In the last few decades, this approach has been extensively advanced for fully observable environments. In this paper, we generalize a variety of these advances to partially observable settings, and similar to the fully observable case, we keep our focus on the class of Markovian policies. We propose a series of technical tools, including a novel notion of advantage function, to develop policy gradient algorithms and study their convergence properties in such environments. Deploying these tools, we generalize a variety of existing theoretical guarantees, such as policy gradient and convergence theorems, to partially observable domains, those which also could be carried to more settings of interest. This study also sheds light on the understanding of policy gradient approaches in real-world applications which tend to be partially observable.
References in corpus (9)
- Continuous control with deep reinforcement learning
- Trust Region Policy Optimization
- Global Convergence of Policy Gradient Methods for the Linear Quadratic Regulator
- Finite-Sample Analysis of Proximal Gradient TD Algorithms
- Neural Policy Gradient Methods: Global Optimality and Rates of Convergence
- Reinforcement Learning in POMDP's via Direct Gradient Ascent
- Combating Reinforcement Learning's Sisyphean Curse with Intrinsic Fear
- Geometry and Determinism of Optimal Stationary Control in Partially Observable Markov Decision Processes
- Experimental results : Reinforcement Learning of POMDPs using Spectral Methods