Offline Reinforcement Learning
I study how agents learn good sequential decision-making policies from fixed, previously collected data, without further interaction with the environment. I'm particularly interested in what shapes the quality of the policies learned this way.