Blind Inference of Eigenvector Centrality Rankings

We consider the problem of estimating a network's eigenvector centrality only from data on nodes, with no information about network topology. Leveraging versatility graph filters to model processes, supported nodes is modeled as signal obtained via output filter applied white noise. seek simplify downstream task ranking by bypassing topology inference methods and, instead, inferring structure directly signals. To this end, we propose two simple algorithms for set connected an unobserved edges. derive asymptotic and non-asymptotic guarantees these algorithms, revealing key features that determine complexity at hand. Finally, illustrate behavior proposed synthetic real-world datasets.