A Deep Reinforcement Learning Approach to Marginalized Importance
          Sampling with the Successor Representation

                   Scott Fujimoto 1 David Meger 1 Doina Precup 1

             Abstract               approach can have significantly lower variance than tradi-
                                tional importance sampling methods (Precup et al., 2001),
    Marginalized importance sampling (MIS), which
                                which consider a product of ratios over traject ...