MIME：互信息最小化探索
MIME: Mutual Information Minimisation Exploration
摘要：
我们表明，强化学习代理人，学习惊喜（missal）陷入突然的环境过渡边界，因为这些过渡是很难学习。我们提出了一个反直觉的解决方案，我们称之为互信息最小化探索（MIME），其中代理学习环境的潜在表示，而不试图预测未来的状态。我们表明，我们的代理执行显着更好的尖锐的过渡边界，同时匹配的性能，在其他地方的驱动代理。特别是，我们在困难的学习游戏上展示了最先进的性能，如Gravitar，Montezuma的复仇和Doom。
英文摘要：
We show that reinforcement learning agents that learn by surprise (surprisal) get stuck at abrupt environmental transition boundaries because these transitions are difficult to learn. We propose a counter-intuitive solution that we call Mutual Information Minimising Exploration (MIME) where an agent learns a latent representation of the environment without trying to predict the future states. We show that our agent performs significantly better over sharp transition boundaries while matching the performance of surprisal driven agents elsewhere. In particular, we show state-of-the-art performance on difficult learning games such as Gravitar, Montezuma's Revenge and Doom.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
