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

Evan Shelhamer, Parsa Mahmoudieh, Max Argus, Trevor Darrell

Browse the full ICLR paper archive.