A Novel Community Finding Method on Large Networks

November 21, 2016

On the 14th of November Dávid Deritei, a physicist by training, chose a technical, more mathematics and algorithm-centric topic, contrary to last week’s talk about a much more applied topic: forbidden triads in the jazz music community. He summarized a paper from Tiago et.al. [1] about a block model-like community detection method.

In their proposed model the independent parameter is the number of blocks within the nested model. For a given number of blocks, they used a maximum likelihood method with a certain entropy measure to optimize the communities. Their method resolves the resolution limit, scales quite well, gave promising results when comparing to benchmarks and show deeper insights about community structure.

[1] Hierarchical Block Structures and High Resolution Model Selection in Large Networks, Tiago P. Peixoto, Physical Review X, 2014 - APS

Written by Milan Janosov