Multi-objective Optimization with Nonnegative Matrix Factorization for Identifying Overlapping Communities in Networks
Community structure is one of the most important properties existing in complex networks, and community detection in complex networks is an intensively investigated problem in recent years. In real-world networks, a node is usually shared by several overlapping communities. The problem of detecting overlapping communities is much more complicated than the hard-partition problem. In this paper, a multi-objective immune algorithm with nonnegative matrix factorization as local search module (MOIA-Net) is proposed to uncover overlapping communities in networks. The proposed algorithm simultaneously optimizes two criteria, negative ratio association and ratio cut, to achieve a preferable soft-partition in networks. It adopts a nonnegative matrix factorization strategy as local search procedure to enhance the search ability. Experiments on synthetic networks show the efficiency of the proposed algorithm.
KeywordsComplex network Overlapping community detection Multi-objective optimization Nonnegative matrix factorization
- 9.Pizzuti, C.: A multi-objective genetic algorithm for community detection in networks. In: Proceedings of IEEE International Conference on Tools with Artificial Intelligence, pp. 379–386 (2009)Google Scholar
- 12.Tan, V.Y., Févotte, C.: Automatic relevance determination in nonnegative matrix factorization. In: Proceedings of Signal Processing with Adaptive Sparse Structured Representations (2009)Google Scholar