Synchronization of interconnected heterogeneous networks: The role of network sizes

Huixin Zhang, Weidong Zhang, Jianxi GaoScientific 2019 · DOI

TL;DR

This paper introduces a new analytical framework to study synchronization in interconnected networks (NONs) with varying sizes. They demonstrate that network size differences significantly impact synchronization, and that the optimal strategy for connecting networks is not always to link the highest-degree nodes, but rather depends on the relative sizes of the networks.

What Problem Does This Paper Address?

Real-world systems are increasingly understood as networks of networks (NONs), where individual networks (like power grids, social networks, or biological systems) are interconnected. Synchronization, a phenomenon where coupled oscillators align their behavior, is crucial in many of these systems. While synchronization in single networks has been extensively studied, understanding it in NONs, especially when the constituent networks have different sizes (heterogeneity), remains a significant challenge. Previous research often assumed networks to be of equal size, overlooking how size differences might affect synchronization dynamics and optimal connection strategies. This paper aims to fill this gap by providing an analytical method to understand how network sizes influence synchronization and to identify optimal connection strategies in heterogeneous NONs.

Methods

The authors employ a mechanical analog of the Kuramoto model to represent interconnected oscillators. They extend this model to two interconnected networks, each with its own internal coupling strength ($\lambda$) and a shared external coupling strength ($\beta$) that links them. To analyze synchronization, they derive critical values for these coupling strengths at which a phase transition from synchronization to incoherence occurs. A key methodological contribution is the development of an analytical framework that can handle nonlinear dynamics and varying network sizes, moving beyond linearized models and the computational burden of super-Laplacian matrices. They model the networks as star-like motifs to facilitate analytical derivations, which are then validated against simulations on more complex Scale-Free (SF) networks. They also explore various connection strategies, including hub-to-hub (H-H), hub-to-leaf (H-L), and leaf-to-leaf (L-L) connections, considering both single and double links between networks.

Datasets

Not discussed in the paper. The study relies on simulated data generated from the coupled Kuramoto model and Scale-Free networks.

Key Insights and Ideas

  • Network Size Matters (A Lot): The most striking finding is that the critical coupling strengths for synchronization are highly dependent on the sizes of the interconnected networks. This challenges previous assumptions of equal network sizes.
  • Reducing Size Difference Enhances Synchronization: The paper shows that reducing the disparity between network sizes generally improves the overall synchronization of the system. This suggests that balancing network sizes is crucial for robust synchronization.
  • Optimal Connection Strategy is Dynamic: Contrary to prior beliefs that connecting high-degree nodes is always best, this study reveals that the optimal connection strategy (e.g., H-H, H-L, L-L) depends on the relative sizes of the networks. This implies that connection strategies need to be adapted based on network characteristics.
  • "Unbalance" as a Universal Obstacle: The concept of "unbalance" in network sizes is identified as a universal factor that obstructs synchronization in interconnected networks. This holds true even when comparing star networks and more complex SF networks.
  • Analytical Tool for Heterogeneous NONs: The development of an analytical framework that accounts for nonlinear dynamics and heterogeneous network sizes is a significant contribution, offering a more efficient and accurate way to predict synchronization behavior.

Key Results

  • Analytical expressions for critical internal ($\lambda$) and external ($\beta$) coupling strengths are derived for various connection strategies and network sizes.
  • Phase diagrams are generated, illustrating the "operating space" (regions of synchronization) as a function of network sizes. These diagrams visually demonstrate how network size differences affect synchronization.
  • For H-L L-L connections, a unique finding is the existence of both a lower and an upper limit for the external coupling strength ($\beta$) required for synchronization, indicating a more constrained synchronization regime.
  • The study validates its theoretical predictions against numerical simulations on Scale-Free networks, showing good agreement and confirming the applicability of the derived results to more realistic network structures.
  • The paper quantifies the "acreage" of the operating space for different connection strategies on both star and SF networks, finding a consistent ratio between them, suggesting a potential universality.

Limitations and Future Work

  • Model Simplification: While the star-like motif simplifies analysis, real-world networks are more complex. The authors acknowledge that their findings on homogenous networks can be applied to heterogeneous ones, but further research on more intricate network structures is needed.
  • Specific Dynamics: The study focuses on the Kuramoto model. Extending the framework to other nonlinear dynamics or more complex oscillator behaviors would be valuable.
  • Dynamic Network Sizes: While the paper addresses varying network sizes, real-time, highly dynamic changes in network topology and size might require further investigation.
  • Practical Control: The authors suggest their work can inform time-demanding control strategies for systems like smart grids. However, implementing these insights in real-world, large-scale systems with diverse components presents a significant engineering challenge.
  • Universality Exploration: The observed universal line relating the operating space of star and SF networks warrants further theoretical exploration.

Summary updated

Abstract

Increasing evidence shows that real networks interact with each other, forming a network of networks (NONs). Synchronization, a ubiquitous process in natural and engineering systems, has fascinatingly gained rising attentions in the context of NONs. Despite efforts to study the synchronization of NONs, it is still a challenge to understand how do the network sizes affect the synchronization and its phase diagram of NONs coupled with nonlinear dynamics. Here, we model such NONs as star-like motifs to analytically derive the critical values of both the internal and the external coupling strengths, at which a phase transition from synchronization to incoherence occurs. Our results show that the critical values strongly depend on the network sizes. Reducing the difference between network sizes will enhance the synchronization of the whole system, which indicates the irrationality of previous studies that assume the network sizes to be the same. The optimal connection strategy also changes as the network sizes change, a discovery contradicting to the previous conclusion that connecting the high-degree nodes of each network is always the most effective strategy to achieve synchronization unchangeably. This finding emphasizes the crucial role of network sizes which has been neglected in the previous studies and could contribute to the design of a global synchronized system.
BibTeX
@inproceedings{zhang2019synchronization,
  author = {Huixin Zhang and Weidong Zhang and Jianxi Gao},
  title = {{Synchronization of interconnected heterogeneous networks: The role of network sizes}},
  booktitle = {Scientific},
  year = {2019},
  doi = {10.1038/s41598-019-42636-6},
}
Figure