Contrastive Learning as Goal-Conditioned Reinforcement Learning
This work addresses the challenge of unstable representation learning in RL for researchers and practitioners, offering a novel integration that improves performance, though it is incremental as it builds upon prior contrastive and RL methods.
The paper tackles the problem of learning good representations in reinforcement learning by showing that contrastive representation learning methods can be reinterpreted as goal-conditioned RL algorithms, leading to a simpler method that achieves higher success rates across tasks, including in offline and image-based settings without data augmentation.
In reinforcement learning (RL), it is easier to solve a task if given a good representation. While deep RL should automatically acquire such good representations, prior work often finds that learning representations in an end-to-end fashion is unstable and instead equip RL algorithms with additional representation learning parts (e.g., auxiliary losses, data augmentation). How can we design RL algorithms that directly acquire good representations? In this paper, instead of adding representation learning parts to an existing RL algorithm, we show (contrastive) representation learning methods can be cast as RL algorithms in their own right. To do this, we build upon prior work and apply contrastive representation learning to action-labeled trajectories, in such a way that the (inner product of) learned representations exactly corresponds to a goal-conditioned value function. We use this idea to reinterpret a prior RL method as performing contrastive learning, and then use the idea to propose a much simpler method that achieves similar performance. Across a range of goal-conditioned RL tasks, we demonstrate that contrastive RL methods achieve higher success rates than prior non-contrastive methods, including in the offline RL setting. We also show that contrastive RL outperforms prior methods on image-based tasks, without using data augmentation or auxiliary objectives.