Speaker: Roger Guimera, ICREA and Rovira i Virgili University
Location: Warren Weaver Hall 1302
Date: December 11, 2015, 11:30 a.m.
Host: Dennis Shasha
Suppose someone gives you a network representing harmful interactions between pairs of drugs, and asks you "Which interactions here are likely to be wrong, and which are likely to be missing?" This talk will explain how to answer such questions without any knowledge of pharmacology (and similar questions about networks in many other domains) using non-parametric Bayesian methods and some mild assumptions about the inherent structure of networks. The goal is to go beyond description, and to predict the relationships that a noisy network data set has incorrectly included or is missing. The talk will show applications to problems as diverse as the prediction of conflict within work teams, the prediction of user ratings on movies and books, and the interactions among proteins.
Refreshments will be offered starting 15 minutes prior to the scheduled start of the talk.