Finding Statistically Significant Communities in Networks
The levels of grey of the squares represent different loop levels. One can provide an initial partition/cover as input, from which the algorithm starts operating, or no input, in which case the algorithm will build the clusters about individual vertices, chosen at random. OSLOM performs first a cleaning procedure of the clusters, followed by a check of their internal structure and by a decision on possible cluster unions. This is repeated with different choices of random numbers in order to obtain better statistics and a more reliable information. The final step is to generate a super-network for the next level of the hierarchical analysis.