Multiple Network Motif Clustering with Evolutionary Computation

The definition of community, usually, relies on the concept of edge density. Network motifs, however, have been recognized as fundamental building blocks of networks and, similarly to edges, may give insights for uncovering communities in complex networks. In this work, we propose a novel approach for identifying communities of network motifs. Differently from previous approaches, our method focuses on searching communities where nodes simultaneously participate in several types of motifs. Based on a genetic algorithm, the method finds a number of communities by minimizing the concept of multiple-motifs conductance. Simulations on a real-world network show that the proposed algorithm is able to better capture the real modular structure of the network, outperforming both motifs-based and classic community detection algorithms.