Loss is its own Reward: Self-Supervision for Reinforcement Learning

作者用action, reward, state等当做lalbel，进行有监督训练。