Learning Markov State Abstractions for Deep Reinforcement Learning

Learning a Markov abstract state representation is crucial for improving deep reinforcement learning efficiency.