Finite-Sample Analysis of Off-Policy Natural Actor-Critic Algorithm

               Sajad Khodadadian 1 Zaiwei Chen 2 Siva Theja Maguluri 1

             Abstract                 An AC algorithm can be thought as a generalized policy iter-
   In this paper, we provide finite-sample conver-       ation (Puterman, 1995), and consists of two phases, namely
   gence guarantees for an off-policy variant of the      actor and critic. The objective of the actor is to improve the
   natural actor-crit ...