Other Information Optimization and Transferable State Abstractions in Deep Reinforcement Learning. August 23, 2022