Average-Reward Off-Policy Policy Evaluation with Function Approximation

            Shangtong Zhang 1 * Yi Wan 2 * Richard S. Sutton 2 Shimon Whiteson 1

             Abstract                 which aim to generate a policy that maximizes the reward
                                  rate by iteratively improving the policy using its estimated
    We consider off-policy policy evaluation with        differential value function (see, e.g., Howard (1960); Konda
    function approximation (FA) in av ...