Deep Reinforcement and InfoMax Learning