FDPP: A Fast Stochastic Competitive Learning Model for Dynamic Network Community Detection^*

To better tackle the problem of escalating algorithmic time complexity arising from high parameter volumes and computational intricacy, this paper presents a fast stochastic competitive learning mechanism for dynamic network community detection. It unifies the treatment of isolated nodes and nodes in small connected components in the existing stochastic competitive learning mechanisms, focusing computational power on particle competition within the main body of the network. This approach significantly reduces time complexity and computation time while minimizing the impact on accuracy. The paper introduces the superiority of the FDPP (Fast Dynamic Particle Propagation) model in terms of time complexity through theoretical analysis and experimental validation, with experiments relying on the SNAP (Stanford Network Analysis Project) open-source dataset.