Incorporating Latent Constraints to Enhance Inference of Network Structure

A complex network is a model representation of interactions within technological, social, information, and biological networks. Oftentimes, we are interested in identifying the underlying network structure from limited and noisy observational data, which is a challenging problem. Here, to address this problem, we propose a novel and effective technique that incorporates latent structural constraints into binary compressed sensing. We show high accuracy and robust effectiveness of our proposed method by analyzing artificial small-world and scale-free networks, as well as two empirical networks. Our method requires a relatively small number of observations and it is robust against strong measurement noise. These results suggest that incorporating latent structural constraints into an algorithm for identifying the underlying network structure improves the inference of connections in complex networks.