Approximately Optimal Approximate Reinforcement Learning

Abstract

In order to solve realistic reinforcement learning problems, it is critical that approximate algorithms be used. In this paper,