Conceitos Básicos
A degree-corrected distribution-free model is proposed to detect communities in weighted networks, which extends previous distribution-free models by considering node degree heterogeneity. An efficient spectral clustering algorithm is designed to fit the model, with theoretical guarantees on consistent estimation.
Resumo
The content presents a Degree-Corrected Distribution-Free Model (DCDFM) for community detection in weighted networks. The key highlights are:
DCDFM extends previous distribution-free models by allowing nodes within the same community to have different expected degrees, making it more suitable for real-world weighted networks.
An algorithm called nDFA is designed based on spectral clustering to fit the DCDFM model. Theoretical analysis shows nDFA enjoys consistent estimation under the proposed model.
A general modularity is proposed as an extension of Newman's modularity to measure the performance of community detection methods on weighted networks, including those with negative edge weights.
Experiments on simulated and real-world networks demonstrate the advantages of the degree-corrected model and the effectiveness of the general modularity.
Estatísticas
The content does not contain any explicit numerical data or statistics. The focus is on the theoretical model and algorithm development.