On the Convergence of Discounted Policy Gradient Methods

Chris Nota

Many popular policy gradient methods for reinforcement learning follow a biased approximation of the policy gradient known as the discounted approximation. While it has been shown…