Title: Robust (MD + ML) = Learned Mechanisms

Abstract: Machine learning has developed a variety of tools for learning and representing high-dimensional distributions with structure. Recent years have also seen big advances in designing multi-item mechanisms. Akin to overfitting, however, these mechanisms can be extremely sensitive to the Bayesian prior that they are designed for, which becomes problematic when that prior is only approximately known. We present a modular robustification framework for designing multi-dimensional mechanisms using learned or approximate priors. Our framework disentangles the statistical challenge of estimating a multi-dimensional prior from the task of designing a good mechanism for it, robustifying the performance of the latter against the estimation error of the former. As applications of our framework, we show how to combine structured models such as Markov random fields, Bayesian networks, and topic models with mechanism design, exploiting our framework and the expressive power of these models to reduce the effective dimensionality of the mechanism design problem at hand. 

(The talk will be based on works with Johannes Brustle and Costis Daskalakis.)

Bio: Yang Cai is an Associate Professor of Computer Science and Economics (secondary appointment) at Yale University. He finished his PhD at MIT in Computer Science and received his BSc in EECS at Peking University. His research interests lie in theoretical computer science and its interface with economics, probability, learning and statistics. He has been honored with the Sloan Research Fellowship, the NSF CAREER Award, and the William Dawson Scholarship.