I am currently a postdoc at UC Berkeley, working with Peter Bartlett. I finished my PhD at the Center for Biological and Computational Learning, MIT.

I will be moving to UPenn in December '08.


Research interests:
Machine Learning
Online Learning / Expert Advice
Optimization
Statistical Learning Theory
Multitask Learning
Generalization Bounds
Algorithmic Stability
Empirical Process Theory


Publications
Curriculum Vitae
Talks
Transfer Learning Toolkit (Matlab)
I coordinate the UC Berkeley CIS seminar
Co-Teaching in Spring'08:
    Statistical Learning Theory and Applications at MIT
    and Statistical Learning Theory at UC Berkeley


Personal page (really outdated...)


News:
"Conference on Online Learning Theory" seems to be the appropriate name for COLT'08, judging by the number of online-learning and, in particular, bandit papers.
We found an efficient sqrt{T}-regret algorithm for the adversarial continuum-armed bandit problem over arbitrary convex sets.
I finished a new teapot.
Next conference: Foundations of Computational Mathematics, 16-18 June 2008, Hong Kong, China.

UC Berkeley
527 Soda Hall
Berkeley, CA 94720


E-mail: rakhlin at cs dot berkeley dot edu