Aran Komatsuzaki   @arankomatsuzaki   6/10/2021       

Pretraining Representations for Data-Efficient Reinforcement Learning Proposes SGI, which significantly surpasses prior work on Atari with the steps limited to 100k with an improved unsupervised goal-conditioned RL. abs: https://t.co/27XPu6NajO code: https://t.co/758gWsD2yK

 Reply  0     Retweet   15      Like   56





Posted by Aran Komatsuzaki