Quasi-Newton Iteration in Deterministic Policy Gradient
This work addresses convergence speed issues in reinforcement learning for practitioners, though it is incremental as it builds on existing policy gradient frameworks.
The paper tackles the problem of slow convergence in deterministic policy gradient methods by introducing a model-free approximation of the Hessian using Quasi-Newton steps, which achieves superlinear convergence under rich policy parametrization and generalizes the natural policy gradient method.
This paper presents a model-free approximation for the Hessian of the performance of deterministic policies to use in the context of Reinforcement Learning based on Quasi-Newton steps in the policy parameters. We show that the approximate Hessian converges to the exact Hessian at the optimal policy, and allows for a superlinear convergence in the learning, provided that the policy parametrization is rich. The natural policy gradient method can be interpreted as a particular case of the proposed method. We analytically verify the formulation in a simple linear case and compare the convergence of the proposed method with the natural policy gradient in a nonlinear example.