|
Abstract:
|
Community detection is one of the fundamental topics in the statistical network analysis. Many methods have been proposed in the literature to solve this problem, however, approaches are not feasible for large size network data. In this paper, we propose a model to shrink the distances among nodes inside a network using the lasso penalty, after projecting nodes, edges as well as related covariates into a prespecified latent space. We use the alternating direction method of multipliers (ADMM) algorithm in the estimation procedure to make the model be friendly to large size data. Simulation results and an example using open sourced data are provided in detail.
|
