|
Annual Meeting 2003 - Abstracts
|
|
Reinforcement learning, safety and robustness, and autonomous helicopter flight
|
Reinforcement learning gives a set of tools for automatically making sequential decisions under uncertainty. For example, given a helicopter, can we automatically learn how to fly it? Closely related are also the issues of safety and robustness: For safety-critical applications, can we further provide rigorous performance guarantees for our learnings algorithms? In this talk, I outline a number of recent developments in the theory and practice of policy search approaches to reinforecement learning, including our successful application of these ideas to autonomous flight.
|
Biography |
Andrew Ng is an assistant professor in the Computer Science Department. His research interests include machine learning and pattern recognition, reinforcement learning and adaptive control, and algorithms for text and web data processing.
|
|

2003 Schedule

Stanford Robotics Laboratory
|