Graphlet decomposition of a weighted network

We introduce the graphlet decomposition of a weighted network, which encodes a notion of social information based on social structure. We develop a scalable algorithm, which combines EM with Bron-Kerbosch in a novel fashion, for estimating the parameters of the model underlying graphlets using one network sample. We explore theoretical properties of graphlets, including computational complexity, redundancy and expected accuracy. We test graphlets on synthetic data, and we analyze messaging on Facebook and crime associations in the 19th century.

• Type: Journal
• Year: 2012
• Where: JMLR
• Authors: Hossein Azari Soufiani, Edo Airoldi
• Keywords: Computational Complexity, Facebook, JMLR 2012, Programming Languages, Scalable Algorithm