Fig. 1

Example of how noise from random edges may affect the detection of communities in social networks. a A network formed from the scientific collaborations of F. M. Petters from the American Physical Society. The vertices are researchers and there is an edge between two vertices if the respective researchers coauthored a paper in the past. b Communities detected by the Louvain algorithm [5] in the network depicted in a. c The same network after filtering random edges. d Communities detected by the Louvain algorithm in the filtered network, which are strikingly different from the ones depicted in b