Taylor Expansions of Discount Factors

               Yunhao Tang 1 Mark Rowland 2 Remi Munos 3 Michal Valko 3

             Abstract                example, T could be the first time the MDP gets into a termi-
   In practical reinforcement learning (RL), the dis-     nal state (e.g., a robot falls); when the MDP does not have a
   count factor used for estimating value functions      natural terminal state, T could be enforced as a deterministic
   often differs from that used for defining the ...