Squeezing More from the Stream : Learning Representation Online for Streaming Reinforcement Learning
This work addresses the challenge of learning meaningful representations in resource-constrained, on-device streaming RL applications, representing an incremental improvement over existing methods.
The paper tackled the sample inefficiency problem in streaming reinforcement learning by extending Self-Predictive Representations to the streaming pipeline, resolving training instabilities with orthogonal gradient updates, and it systematically outperformed existing baselines across Atari, MinAtar, and Octax suites.
In streaming Reinforcement Learning (RL), transitions are observed and discarded immediately after a single update. While this minimizes resource usage for on-device applications, it makes agents notoriously sample-inefficient, since value-based losses alone struggle to extract meaningful representations from transient data. We propose extending Self-Predictive Representations (SPR) to the streaming pipeline to maximize the utility of every observed frame. However, due to the highly correlated samples induced by the streaming regime, naively applying this auxiliary loss results in training instabilities. Thus, we introduce orthogonal gradient updates relative to the momentum target and resolve gradient conflicts arising from streaming-specific optimizers. Validated across the Atari, MinAtar, and Octax suites, our approach systematically outperforms existing streaming baselines. Latent-space analysis, including t-SNE visualizations and effective-rank measurements, confirms that our method learns significantly richer representations, bridging the performance gap caused by the absence of a replay buffer, while remaining efficient enough to train on just a few CPU cores.