A0848
Title: Informative periphery detection and post-detection inference on weighted directed networks
Authors: Wenqin Du - University of Southern California (United States) [presenting]
Wen Zhou - New York University (United States)
Tianxi Li - University of Minnesota (United States)
Abstract: In network analysis, introduced by peripheral or non-essential components, noises and biases can mask pivotal structures and hinder the efficacy of many network modeling and inference procedures. Recognizing this, the identification of the core-periphery (CP) structure has emerged as a crucial data pre-processing step. Many existing efforts either fail to account for the directionality or lack the theoretical justification of the identification procedure. Answers to three pressing questions are sought: (i) How can informative and non-informative structures in weighted directed networks be distinguished? (ii) What approach offers computational efficiency in discerning these components? (iii) Upon detecting CP structure, can uncertainty be quantified to evaluate the detection? The signal-plus-noise model is adopted, categorizing uniform relational patterns as non-informative, by which the sender and receiver peripheries are defined. Furthermore, instead of confining the core component to a specific structure, it is considered complementary to either the sender or receiver peripheries. Based on the definitions of the sender and receiver peripheries, spectral algorithms are proposed to identify the CP structure in directed weighted networks. The algorithm stands out with statistical guarantees, ensuring the identification of sender and receiver peripheries with overwhelming probability. Additionally, the methods scale effectively for expansive directed networks.
