Observability transitions in clustered networks

TL;DR

This paper analyzes how clustering in networks influences the observability transition, revealing that higher clustering slightly lowers the critical node fraction needed for large observable and unobservable components, especially in networks with low average degree.

Contribution

It provides analytical derivations of observability transition thresholds in clustered networks and evaluates the impact of clustering, confirming findings with simulations.

Findings

Clustering slightly lowers critical node fractions for observability.

Effect of clustering is negligible at higher average degrees.

Analytical results are validated by Monte Carlo simulations.

Abstract

We investigate the effect of clustering on network observability transitions. In the observability model introduced by Yang, Wang, and Motter [Phys. Rev. Lett. 109, 258701 (2012)], a given fraction of nodes are chosen randomly, and they and those neighbors are considered to be observable, while the other nodes are unobservable. For the observability model on random clustered networks, we derive the normalized sizes of the largest observable component (LOC) and largest unobservable component (LUC). Considering the case where the numbers of edges and triangles of each node are given by the Poisson distribution, we find that both LOC and LUC are affected by the network's clustering: more highly-clustered networks have lower critical node fractions for forming macroscopic LOC and LUC, but this effect is small, becoming almost negligible unless the average degree is small. We also evaluate bounds for these critical points to confirm clustering's weak or negligible effect on the network observability transition. The accuracy of our analytical treatment is confirmed by Monte Carlo simulations.