Improved Regret Bound and Experience Replay in Regularized Policy Iteration | Read Paper on Bytez