Gate calibration with reinforcement learning