skip to content
 

From Graphon to Graphex: Models and Estimators for Sparse Networks using Exchangeable Random Measures

Presented by: 
Daniel Roy University of Toronto
Date: 
Tuesday 26th July 2016 - 14:30 to 15:00
Venue: 
INI Seminar Room 1
Abstract: 
A major challenge for Bayesian network analysis is that we lack general nonparametric models of sparse graphs. To meet this challenge, we introduce and study the general class of random graphs defined by the exchangeability of their real-valued vertex labels, an idea inspired by a model due to Caron and Fox. A straightforward adaptation of a result by Kallenberg yields a representation theorem: every such random graph is characterized by three (potentially random) components: a nonnegative real I, an integrable function S : R+ to R+, and a symmetric measurable function W: R+^2 to [0,1] that satisfies several weak integrability conditions. We call the triple (I,S,W) a graphex, in analogy to graphons, which characterize the (dense) exchangeable graphs on the naturals. I will present some results about the structure and consistent estimation of these random graphs. This is joint work with Victor Veitch.
The video for this talk should appear here if JavaScript is enabled.
If it doesn't, something may have gone wrong with our embedded player.
We'll get it fixed as soon as possible.
University of Cambridge Research Councils UK
    Clay Mathematics Institute London Mathematical Society NM Rothschild and Sons