How can learning agents use prior knowledge to adapt safely and efficiently to new domains? In my research I try to answer this question by using concepts from reinforcement learning, constrained Markov decision processes, meta-learning and Bayesian inference.
