A2028
Title: Dynamic networks with node heterogeneity and homophily
Authors: Binyan Jiang - The Hong Kong Polytechnic University (Hong Kong) [presenting]
Abstract: A framework is proposed for modeling node heterogeneity and link homophily in dynamic networks. The framework provides new insights into how networks evolve over time and supplies sophisticated tools for predicting future networks with statistical guarantees. The model accounts for link homophily associated with both observed and latent traits. Joint modeling of node heterogeneity and both observed and latent homophily effects presents a significant challenge in statistical inference due to the large number of confounding parameters. A novel normalized squared loss is introduced to enable stable parameter estimation in high-dimensional settings. Rigorous theoretical analysis of the estimation method is provided, and effectiveness is demonstrated through extensive simulations and illustration with real-world network data.
