Spring 2006 Seminar Series

DATE: Thursday, April 13th, 2006
LOCATION: E40-298
TIME: 4:15pm
Reception immediately following in the Philip M. Morse Reading Room, E40-106

SPEAKER:
David Parkes

John L. Loeb Associate Professor of the Natural Sciences and Associate Professor of Computer Science

Harvard University

TITLE
Adaptive Online Mechanism Design for Sequential Environments

ABSTRACT
In online mechanism design (MD), the problem is to implement a desirable sequential decision policy despite the private information of self-interested agents. Each agent has an arrival time, departure time, and local information about its value for different policies. If the designer has a model of the environment then an online-VCG mechanism implements the expected value-maximizing policy in a Bayes-Nash equilibrium (BNE), or an epsilon-BNE if the optimal policy is computed via an epsilon-approximation algorithm. But, what if the mechanism must learn a model of the environment, while maintaining incentive-compatibility? We present positive results for adaptive online MD in a sub-case of the general online MD problem.

 

Joint work with Satinder Singh, University of Michigan.