Figure 2. Comparison of our NMF method and Ball’s method in the three sets of synthetic networks, measured by the fraction of vertices classified correctly (FVCC).

Each data point in the figure is an average over 50 network instances. (a) FVCC accuracy as a function of the expected degree <k> of all nodes. (b) FVCC accuracy as a function of the size of the larger community. (c) FVCC accuracy as a function of the amount of overlap between the two communities.