Index

Lecture series are sponsored by

Prof. Devavrat Shah (MIT)

Abstract: We consider the task of computing choice (or preferences) over a collection of options based of partially revealed preferences between a subset of option. This task is central to a vast number of scenarios including data driven policy making, efficient business operations, ranking sports teams, recommendation systems, deciding conference programs and hiring faculty/admitting students in our department. In this talk, we shall address this question by modeling choice as a distribution over permutations of available options. The aim is to develop computationally and statistically efficient approaches that can scale gracefully with data and dimension. We shall discuss when it is feasible to achieve such solutions and when it is not. We shall also discuss how explicitly learning a choice model can be by-passed for making certain decisions.

Short Bio: Devavrat Shah is currently an Associate Professor with the department of Electrical Engineering and Computer Science at MIT. His research interests include algorithms for statistical inference and social networks. He has received Erlang Prize from INFORMS and Rising Star Award from ACM Sigmetrics. He is a distinguished young alumni of his alma mater IIT Bombay.