The shades of reinforcement learning

Abstract:
We review the scope of reinforcement learning and argue that it is about several different problems, each one bringing about different challenges: offline learning based on a model, a simulator, historical data, or experiments with a physical system, as well as online learning. We also review the main types of reinforcement learnign algoirithms (value function approximation, policy learning, and actor-critic methods), and conclude with a discussion of research directions.

Short-Bio:
John N. Tsitsiklis was born in Thessaloniki, Greece, in 1958. He received the B.S. degree in Mathematics (1980), and the B.S. (1980), M.S. (1981), and Ph.D. (1984) degrees in Electrical Engineering, all from the Massachusetts Institute of Technology, Cambridge, Massachusetts, U.S.A.During the academic year 1983-84, he was an acting assistant professor of Electrical Engineering at Stanford University, Stanford, California. Since 1984, he has been with the department of Electrical Engineering and Computer Science (EECS) at the Massachusetts Institute of Technology (MIT), where he is currently a Clarence J Lebel Professor of Electrical Engineering.

more info