skip to content

Random walk models of networks: modeling and inferring complex dependence

Presented by: 
Benjamin Bloem-Reddy Columbia University
Wednesday 27th July 2016 - 11:30 to 12:00
INI Seminar Room 1
A signature of many network datasets is strong local dependence, a phenomenon that gives rise to frequently observed properties such as transitive triples, pendants, and structural heterogeneity. One difficulty in modeling such dependence is that the notion of locality may not be well-defined, and it is likely to be heterogeneous throughout the network. Furthermore, models that do not assume some form of conditional independence on the edges typically are intractable or too simplistic to serve as useful statistical models. We introduce a class of models, based on random walks, that allows the scale of dependence to vary; it is able to generate a range of network structures faithful to those observed in real data, and it admits tractable inference procedures.

This is joint work with Peter Orbanz.
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