skip to content

Bayesian Methods for Networks

Presented by: 
Peter Hoff University of Washington
Monday 25th July 2016 - 10:00 to 11:00
INI Seminar Room 1
Statistical analysis of social network data presents many challenges: Realistic models often require a large number of parameters, yet maximum likelihood estimates for even the simplest models may be unstable. Furthermore, network data often exhibit non-standard statistical dependencies, and most network datasets lack any sort of replication.

Statistical methods to address these issues have included random effects and latent variable models, and penalized likelihood methods. In this talk I will discuss how these approaches fit naturally within a Bayesian framework for network modeling. Additionally, we will discuss how standard Bayesian concepts such as exchangeability play a role in the development and interpretation of probability models for networks. Finally, some thoughts on the use of Bayesian methods for large-scale dynamic networks will be presented.

[ The video of this talk is temporarily unavailable. Please try later. ]

University of Cambridge Research Councils UK
    Clay Mathematics Institute London Mathematical Society NM Rothschild and Sons