Revisiting Resolution and Inter-Layer Coupling Factors in Modularity for Multilayer Networks

TL;DR

This paper introduces a new, parameter-free definition of multilayer network modularity that better captures structure and order, improving community detection in dynamic and complex multilayer networks.

Contribution

It revises the roles of resolution and inter-layer coupling in modularity, providing unsupervised, structure-aware computation methods and accommodating layer ordering.

Findings

Significant impact of different resolution and coupling functions on community detection.

Parameter-free approaches improve modularity's relevance to network structure.

Experimental results validate the effectiveness of the proposed modularity.

Abstract

Modularity for multilayer networks, also called multislice modularity, is parametric to a resolution factor and an inter-layer coupling factor. The former is useful to express layer-specific relevance and the latter quantifies the strength of node linkage across the layers of a network. However, such parameters can be set arbitrarily, thus discarding any structure information at graph or community level. Other issues are related to the inability of properly modeling order relations over the layers, which is required for dynamic networks. In this paper we propose a new definition of modularity for multilayer networks that aims to overcome major issues of existing multislice modularity. We revise the role and semantics of the layer-specific resolution and inter-layer coupling terms, and define parameter-free unsupervised approaches for their computation, by using information from the within-layer and inter-layer structures of the communities. Moreover, our formulation of multilayer modularity is general enough to account for an available ordering of the layers and relating constraints on layer coupling. Experimental evaluation was conducted using three state-of-the-art methods for multilayer community detection and nine real-world multilayer networks. Results have shown the significance of our modularity, disclosing the effects of different combinations of the resolution and inter-layer coupling functions. This work can pave the way for the development of new optimization methods for discovering community structures in multilayer networks.