How Complex is a Complex Network? Insights from Linear Systems Theory

TL;DR

This paper introduces a new measure of network complexity based on linear systems theory, linking the structure and heterogeneity of network components to the overall system complexity.

Contribution

It proposes a complexity index derived from the McMillan degree, connecting network topology and node heterogeneity to system complexity.

Findings

Complexity index is invariant to interconnection weights for a given structure.

Index depends on the matching number of subgraphs related to nodal dynamics.

Network architecture and component diversity jointly influence complexity.

Abstract

This paper leverages linear systems theory to propose a principled measure of complexity for network systems. We focus on a network of first-order scalar linear systems interconnected through a directed graph. By locally filtering out the effect of nodal dynamics in the interconnected system, we propose a new quantitative index of network complexity rooted in the notion of McMillan degree of a linear system. First, we show that network systems with the same interconnection structure share the same complexity index for almost all choices of their interconnection weights. Then, we investigate the dependence of the proposed index on the topology of the network and the pattern of heterogeneity of the nodal dynamics. Specifically, we find that the index depends on the matching number of subgraphs identified by nodal dynamics of different nature, highlighting the joint impact of network architecture and component diversity on overall system complexity.