Timezone: »

 
Offline Q-learning on Diverse Multi-Task Data Both Scales And Generalizes
Aviral Kumar · Rishabh Agarwal · XINYANG GENG · George Tucker · Sergey Levine

@
in Foundation Models for Decision Making »
Event URL: https://openreview.net/forum?id=_K8EF3i1gdU »

The potential of offline reinforcement learning (RL) is that high-capacity models trained on large, heterogeneous datasets can lead to agents that generalize broadly, analogously to similar advances in vision and NLP. However, recent works argue that offline RL methods encounter unique challenges to scaling up model capacity. Drawing on the learnings from these works, we re-examine previous design choices and find that with appropriate choices: ResNets, cross-entropy based distributional backups, and feature normalization, offline Q-learning algorithms exhibit strong performance that scales with model capacity. Using multi-task Atari as a testbed for scaling and generalization, we train a single policy on 40 games with near-human performance using up-to 80 million parameter networks, finding that model performance scales favorably with capacity. In contrast to prior work, we extrapolate beyond dataset performance even when trained entirely on a large (400M transitions) but highly suboptimal dataset (51% human-level performance). Compared to return-conditioned supervised approaches, offline Q-learning scales similarly with model capacity and has better performance, especially when the dataset is suboptimal. Finally, we show that offline Q-learning with a diverse dataset is sufficient to learn powerful representations that facilitate rapid transfer to novel games and fast online learning on new variations of a training game, improving over existing state-of-the-art representation learning approaches.

Author Information

Aviral Kumar (UC Berkeley)
Rishabh Agarwal (Google Research, Brain Team)

My research work mainly revolves around deep reinforcement learning (RL), often with the goal of making RL methods suitable for real-world problems, and includes an outstanding paper award at NeurIPS.

XINYANG GENG (UC Berkeley)
George Tucker (Google Brain)
Sergey Levine (UC Berkeley)

More from the Same Authors