Community detection through vector-label propagation algorithms

Community detection is a fundamental and important problem in network science, as community structures often reveal both topological functional relationships between different components of the complex system. In this paper, we first propose gradient descent framework modularity optimization called vector-label propagation algorithm (VLPA), where node associated with vector continuous labels instead one label. Retaining weak structural information vector-label, VLPA outperforms some well-known methods, particularly improves performance networks structures. Further, incorporate stochastic strategies into to avoid stuck local optima, leading (sVLPA). We show that sVLPA performs better than Louvain Method, widely used algorithm, on artificial benchmarks real-world networks. Our theoretical scheme based can be directly applied high-dimensional each has multiple features, also for optimizing other partition measures such resolution parameters. • frame detect communities. (VLPA) retains information. obtains when structure weak. further proposed via equipping optima. classic Method real