Affect-based Intrinsic Rewards for Learning General Representations: https://arxiv.org/abs/1912.00403 - View it on GitHub