Stochastic Recursive Momentum for Policy Gradient Methods
This work addresses efficiency and usability issues in reinforcement learning for practitioners, though it is incremental as it builds on existing variance-reduced methods.
The authors tackled the problem of high sample complexity and complex parameter tuning in policy gradient methods by proposing STORM-PG, which achieves a provably sharp O(1/ε^3) sample complexity bound and simplifies tuning by avoiding batch alternations.
In this paper, we propose a novel algorithm named STOchastic Recursive Momentum for Policy Gradient (STORM-PG), which operates a SARAH-type stochastic recursive variance-reduced policy gradient in an exponential moving average fashion. STORM-PG enjoys a provably sharp $O(1/ε^3)$ sample complexity bound for STORM-PG, matching the best-known convergence rate for policy gradient algorithm. In the mean time, STORM-PG avoids the alternations between large batches and small batches which persists in comparable variance-reduced policy gradient methods, allowing considerably simpler parameter tuning. Numerical experiments depicts the superiority of our algorithm over comparative policy gradient algorithms.