Steering Macro-Scale Network Community Structure by Micro-Scale Features

Abstract

Network science plays an increasingly important role to model complex data in many scientific disciplines. One notable feature of network organization is community structure, which refers to clusters of tightly interconnected nodes. A prominent problem is how to investigate the relationship between macro-scale modules that are retrieved by optimizing global network measures, and micro-scale structure that are defined by specific queries of the analysis (e.g., nodal features). By generalizing fundamental concepts of joint space-frequency localization to network theory, here we propose a flexible framework to study interactions between micro- and macro-structure. Similar to pointing and focusing a magnifying glass, the analysis can be directed to specific micro-scale structure, while the degree of interaction with the macro-scale community structure can be seamlessly controlled. In addition, the method is computationally efficient as a result of the underlying low-dimensional optimization problem.