Preference elicitation and inverse reinforcement learning

Constantin Rothkopf, Christos Dimitrakakis

We state the problem of inverse reinforcement learning in terms of preference elicitation, resulting in a principled (Bayesian) statistical formulation. This generalises previous w…