Provably Efficient Learning of Transferable Rewards

       Alberto Maria Metelli * 1 Giorgia Ramponi * 1 Alessandro Concetti 1 Marcello Restelli 1

              Abstract                 theoretically, under the strong assumption of reward unique-
   The reward function is widely accepted as a suc-       ness (Abbeel & Ng, 2004; Pirotta & Restelli, 2016; Ramponi
   cinct, robust, and transferable representation of a      et al., 2020b). Nevertheless, as noted in Osa et al. (2018),
   task. Typi ...