Fig. 5

Performance of the community detection algorithm on the Lancichinetti-Fortunato-Radicchi (LFR) benchmark graph. We display results only for the reduced version of the algorithm but with both initial conditions. As a term of comparison, we display results obtained by Louvain and Infomap in the same set of benchmark graphs. In the various panels, performance is measured in terms of normalized mutual information (NMI). This quantity is evaluated as a function of the mixing parameter μ of the model. We consider the following experimental settings: a Small communities with N=1000 nodes; b Big communities with N=1000 nodes; c N=5000 nodes with Small communities; d N=5000 nodes with Big communities