Importance of using appropriate baselines for evaluation of data-efficiency in deep reinforcement learning for Atari

Open in new window