Գրեք սա: Adaptive representations for reinforcement learning