Foundations of Reinforcement Learning

The course is a graduate level course, focusing on theoretical foundations of reinforcement learning. It covers basics of Markov Decision Process (MDP), dynamic programming based algorithms, policy optimization, planning, exploration, as well as information theoretical lower bounds. Various advanced topics are also discussed, including off-policy evaluation, function approximation, partial observable MDP and deep reinforcement learning. This course puts special emphases on the algorithms and their theoretical analyses. Prior knowledge on linear algebra, probability theory, and stochastic process is required.