Reconstructing Community-Bridge Networks from Heterogeneous and Outlier-Contaminated Datasets

Abstract

Dear Editor, Reconstructing structures of community-bridge networks from nodal states has been a challenge in cyber-physical systems, where observed data are often short in duration, collected under varying conditions, and subject to different noise levels and outliers. This paper proposes a sparse Bayesian learning method that jointly analyzes multiple small datasets by incorporating a hyperblock-coupling prior. A parameter estimation strategy is thereby designed to suppress outliers and accommodate heterogeneous noise. The proposed method is theoretically proved to be convergent. Experiments are conducted to verify its effectiveness and superiority.