Understanding Self-Predictive Learning for Reinforcement Learning
This addresses a fundamental issue in representation learning for reinforcement learning, though it appears incremental as it builds on existing self-predictive methods.
The paper tackled the problem of representation collapse in self-predictive learning for reinforcement learning, where trivial solutions minimize prediction error, and found that optimizing the predictor faster and using semi-gradient updates prevents this collapse, with experiments showing promise for a novel bidirectional algorithm.
We study the learning dynamics of self-predictive learning for reinforcement learning, a family of algorithms that learn representations by minimizing the prediction error of their own future latent representations. Despite its recent empirical success, such algorithms have an apparent defect: trivial representations (such as constants) minimize the prediction error, yet it is obviously undesirable to converge to such solutions. Our central insight is that careful designs of the optimization dynamics are critical to learning meaningful representations. We identify that a faster paced optimization of the predictor and semi-gradient updates on the representation, are crucial to preventing the representation collapse. Then in an idealized setup, we show self-predictive learning dynamics carries out spectral decomposition on the state transition matrix, effectively capturing information of the transition dynamics. Building on the theoretical insights, we propose bidirectional self-predictive learning, a novel self-predictive algorithm that learns two representations simultaneously. We examine the robustness of our theoretical insights with a number of small-scale experiments and showcase the promise of the novel representation learning algorithm with large-scale experiments.