Students of this graduate seminar-style course will become junior researchers in reinforcement learning (RL) and adaptive agent design! This course guides those with a good understanding of the foundations of RL to gain a broad, yet deep and research-level, understanding of several aspects of modern RL research. This course explores wide variety of research topics in reinforcement learning (RL) and adaptive agent design. The topics include Imitation Learning, Advances in Policy Search, Empirical Evaluation in RL, Hierarchical RL, Advances in Exploration-Exploitation, Batch/Offline RL, Distributional RL, Multi-Agent RL, Safety and Robustness in RL, RL for Large Language Models, RL Acceleration, and Evolutionary Approach to Agent Design. The course emphasizes both algorithmic advances and theoretical foundations, and requires a solid understanding of the foundations of RL.
The course consists of reading papers each week, presenting and guiding discussions about the papers (about 2-3 papers in semester per students), active participation in the class, working on a research project as a team, peer reviewing the papers, and presenting them.