My research focuses on developing algorithms for reinforcement learning (RL) in complex domains. I have worked on algorithms for transfer learning in RL domains, as well as the application of RL to domains with continuous and high dimensional state spaces. My recent work has looked at developing RL algorithms that are stable when planning over long time horizons. I have also done work on the problem of computers learning from human teachers. Specifically, I have looked at ways that computers can learn more effectively from positive and negative feedback.
Refereed Conference Papers
Refereed Workshop Papers
- Robert Loftin, James MacGlashan, Michael L. Littman, Matthew E. Taylor and David L. Roberts. A Strategy-Aware Technique for Learning Behaviors from Discrete Human Feedback. Technical report TR-2014-3, Computer Science Department, North Carolina State University. 2014.