Policy gradient methods are extensively used in reinforcement learning as a way to optimize
expected return. In this thesis, we build upon the idea that in general, the evolution of a …