Learning temporally-consistent representations for data-efficient reinforcement learning

摘要: Deep reinforcement learning (RL) agents that exist in high-dimensional state spaces, such as those composed of images, have interconnected learning burdens. Agents must learn an …