Markov Abstraction

Learning Markov State Abstractions for Deep Reinforcement Learning

We introduce a method of learning an abstract state representation for Markov Decision Processes (MDPs) with rich observations.