Detecting community structure of complex networks by affinity propagation

The question of finding the community structure of a complex network has been addressed in many different ways. Here we utilize a clustering method called affinity propagation, associating with some existent measures on graphs, such as the shortest path, the diffusion distance and the dissimilarity index, to solve the network partitioning problem. This method considers all nodes as potential exemplars, and transmits real valued messages between nodes until a high quality set of exemplars and corresponding communities gradually emerges. It is demonstrated by simulation experiments that the algorithms can not only identify the community structure of a network, but also determine the number of communities automatically during the model selection. Moreover, they are successfully applied to several real-world networks, including the karate club network and the dolphins network.