Self-Attention based Temporal Intrinsic Reward for Reinforcement Learning | IEEE Conference Publication | IEEE Xplore