Only Relevant Information Matters: Filtering Out Noisy Samples to Boost RL
This addresses sample efficiency and performance issues in RL for continuous control domains, though it appears incremental as it builds on existing policy gradient methods.
The paper tackles the problem of noisy samples degrading policy gradient reinforcement learning by introducing SAUNA, a method that filters out non-informative transitions based on variance explained by the value function, which significantly improves performance on benchmark continuous control problems.
In reinforcement learning, policy gradient algorithms optimize the policy directly and rely on sampling efficiently an environment. Nevertheless, while most sampling procedures are based on direct policy sampling, self-performance measures could be used to improve such sampling prior to each policy update. Following this line of thought, we introduce SAUNA, a method where non-informative transitions are rejected from the gradient update. The level of information is estimated according to the fraction of variance explained by the value function: a measure of the discrepancy between V and the empirical returns. In this work, we use this metric to select samples that are useful to learn from, and we demonstrate that this selection can significantly improve the performance of policy gradient methods. In this paper: (a) We define SAUNA's metric and introduce its method to filter transitions. (b) We conduct experiments on a set of benchmark continuous control problems. SAUNA significantly improves performance. (c) We investigate how SAUNA reliably selects samples with the most positive impact on learning and study its improvement on both performance and sample efficiency.