A new perspective on network clustering
https://cs.stanford.edu/people/jure/pubs/motifs-wsdm19.pdf
This is a paper co-authored by our instructor Prof. Austin Benson on a new method to cluster edges in a network. On the high level, current edge clustering methods has no measurement for clusters attribute to the head node. While on the other hand, the head node analysis is common in network analysis. Hence, the paper proposed local closure coefficient: “the fraction of length-2 paths emanating from the head node that induce a triangle”, for a better head-node-based clustering criterion. This paper is very interesting in defining a novel metric to evaluate clustering of edges, with well intuition. It’s still a very new paper but this might be an important work that motivates further research directions.
This paper is highly relevant to the course in that we have discussed about clustering coefficient in our lectures. However, we have spent little time thinking about what clustering coefficient really refers to and potential problem that is might cause in specially crafted network. By proposing local closure coefficient, the paper takes local information into account rather than just looking into the global structure.