CMU Campus
Center for                           Nonlinear Analysis
CNA Home People Seminars Publications Workshops and Conferences CNA Working Groups CNA Comments Form Summer Schools Summer Undergraduate Institute PIRE Cooperation Graduate Topics Courses SIAM Chapter Seminar Positions Contact
Seminar Abstracts

Francisco S. Melo, Computer Science Department, Carnegie Mellon University

Reinforcement Learning: Overview and Current Challenges

Abstract: Reinforcement learning (RL) is a framework used to address optimal control problems in which the decision-maker has minimum knowledge on the environment/task to be tackled. In theory, RL can be applied to address any optimal control task, yielding optimal solutions while requiring very little a priori information on the system itself. Over the last 25 years, RL has greatly evolved as a research field. It has benefited from and contributed to research in such disparate fields as artificial intelligence, optimal control, neuroscience, psychology, economics, operations research and others. In this talk I will briefly review this exciting topic of research. I will describe the standard framework used to address the "classical" RL problem and present several "classical" solution methods. I will provide a brief sketch of the main mathematical tools used in the design/analysis of such RL methods and conclude by describing some of the current challenges and main difficulties faced by the RL community in extending RL to more general problems.

TUESDAY, March 24, 2009
Time: 1:30 P.M.
Location: PPB 300