この資料をSMS送信: Adaptive representations for reinforcement learning