Bandit and Limited-Feedback Problems
Investigators: Alexander (Sasha) Rakhlin
In this project we aim to understand theoretical guarantees of sequential decision-making with limited feedback. We also aim to apply our results in various domains, such as quantitative analysis of embedded systems.
Recent papers on the subject:
J. Abernethy and A. Rakhlin. Beating the Adaptive Bandit with High Probability, COLT 2009.
J. Abernethy, E. Hazan, A. Rakhlin. Competing in the Dark: An Efficient Algorithm for Bandit Linear Optimization, COLT 2008.
P. Bartlett, V. Dani, T. Hayes, S. Kakade, A. Rakhlin, and A. Tewari. High probability regret bounds for online optimization, COLT 2008.