An Online Prediction Algorithm for Reinforcement Learning with Linear Function Approximation using Cross Entropy Method
This addresses the challenge of stable value function estimation in model-free reinforcement learning for practitioners, though it is incremental as it builds on existing cross entropy and stochastic approximation methods.
The paper tackles the problem of stable online prediction in reinforcement learning with linear function approximation by introducing two new algorithms based on the cross entropy method, which achieve good performance in computational efficiency, accuracy, and stability on benchmark problems.
In this paper, we provide two new stable online algorithms for the problem of prediction in reinforcement learning, \emph{i.e.}, estimating the value function of a model-free Markov reward process using the linear function approximation architecture and with memory and computation costs scaling quadratically in the size of the feature set. The algorithms employ the multi-timescale stochastic approximation variant of the very popular cross entropy (CE) optimization method which is a model based search method to find the global optimum of a real-valued function. A proof of convergence of the algorithms using the ODE method is provided. We supplement our theoretical results with experimental comparisons. The algorithms achieve good performance fairly consistently on many RL benchmark problems with regards to computational efficiency, accuracy and stability.