Learn a Bayesian Posterior distribution over reward functions using human feedback of different kinds