Efficiently Identifying Clusters in Large Sparse Networks

A two-step procedure called Sieve can accurately identify clusters in large sparse networks by first efficiently separating the network into disjoint components and then optimizing a novel objective function to find communities within each component.