skip to content

Multiresolution network models

Presented by: 
Tyler McCormick University of Washington
Tuesday 26th July 2016 - 13:30 to 14:00
INI Seminar Room 1
Social networks exhibit two key topological features: global sparsity and local density.  That is, overall the propensity for interaction between any two randomly selected actors is infinitesimal, but for any given individual there is massive variability in the propensity to interact with others in the network.  Further, the relevant scientific questions typically differ depending on the scale of analysis.  In this talk, we propose a class of multiresolution statistical models that model network structures on multiple scales to enable inference about relevant population-level parameters.  We capture global graph structure using a mixture over projective models that capture local graph structures. This approach is advantageous as it allows us to differentially invest modeling effort within subgraphs of high density, while maintaining a parsimonious structure between such subgraphs. We illustrate the utility of our method using simulation and data on household relations from Karnataka, India.  This is joint work with Bailey Fosdick (CSU) and Ted Westling (UW).  
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