DSpace@MIT

This paper describes a probabilistic framework for incorporating structured inductive biases into reinforcement learning. These inductive biases arise from policy priors, probability distributions over optimal policies. ...