Pretraining Representations for Data-Efficient Reinforcement Learning

Max Schwarzer, Nitarshan Rajkumar, Michael Noukhovitch, Ankesh Anand, Laurent Charlin, R Devon Hjelm, Philip Bachman, Aaron Courville

Keywords: Pretraining, Sample Efficiency, Reinforcement Learning, Self-Supervised Learning

TL;DR: We show that pre-training encoders with a set of self-supervised learning tasks greatly improves performance in data-efficient RL.

Abstract: Data efficiency is a key challenge for deep reinforcement learning. We address this problem by using unlabeled data to pretrain an encoder which is then finetuned on a small amount of task-specific data. To encourage learning representations which capture diverse aspects of the underlying MDP, we employ a combination of latent dynamics modelling and unsupervised goal-conditioned RL. When limited to 100k steps of interaction on Atari games (equivalent to two hours of human experience), our approach significantly surpasses prior work combining offline representation pretraining with task-specific finetuning, and compares favourably with other pretraining methods that require orders of magnitude more data. Our approach shows particular promise when combined with larger models as well as more diverse, task-aligned observational data -- approaching human-level performance and data-efficiency on Atari in our best setting.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

Code: https://github.com/mila-iqia/SGI

Community Implementations: [ 3 code implementations](https://www.catalyzex.com/paper/arxiv:2106.04799/code)