Speaker: François Caron (Oxford)
Title: Statistical models and algorithms for large networks
Abstract: Being able to define statistical network models which capture the salient features of large graphs as well as deriving algorithms for large scale inference of the model parameters are major challenges in network analysis. In this talk, I will present a novel class of graph models based on exchangeable point processes, with interpretable parameters and which can generate dense as well as sparse graphs with heavy-tailed degree distributions. I then present a Markov chain Monte Carlo algorithm for efficient exploration of the posterior distribution of the parameters. Network properties are explored for a range of real datasets, including networks with hundreds of thousands of nodes/edges.