From Santa Fe Institute Events Wiki
Statistical Inference for Complex Networks Workshop, December 3-5, 2008, Santa Fe NM
Jure Leskovec (homepage)
Kronecker graphs and connections between local and global models of network structure
We have many network models of different flavors and characteristics. At one side of the spectrum we have mechanistic models (like Preferential attachment) that specify mechanisms or principles by which new edges are formed and global network characteristics (like scale-free structure) emerge. On the other hand we also have more statistical models like exponential random graphs that have been very successful at modeling local network interactions. There has been a disconnect between accurately modeling individual edges while still preserving the global network structure. I will present our work on extending Kronecker graphs where we attempt to consider node attribute data to accurately predict individual edge attachments while using the recursive nature of Kronecker graphs to accurately model the global structure of large networks of hundreds of thousands of nodes.