Accelerating Reinforcement Learning by Composing Solutions of Automatically Identified Subtasks

C. Drummond

This paper discusses a system that accelerates reinforcement learning by using transfer from related tasks. Without such transfer, even if two tasks are very similar at some abstra…