Transformability reveals the interplay of dynamics across different network orders

Ming Xie, Shibo He, Aming Li, Zi‐Ke Zhang, Youxian Sun, Jiming ChenarXiv 2025 · DOI · arXiv

TL;DR

This paper introduces a "transformability" framework that bridges the gap between complex higher-order network dynamics (like group interactions) and simpler pairwise interactions. By decomposing system instability into specific dynamical and structural factors, the authors demonstrate that these seemingly distinct processes are intrinsically coupled and can be systematically mapped to one another across different network representations.

What Problem Does This Paper Address?

Traditional network science primarily focuses on pairwise interactions—think of a simple graph where an edge connects exactly two people. However, real-world systems often involve group interactions, such as a team working on a project or a virus spreading through a household. These are modeled using "higher-order" structures like hypergraphs and simplicial complexes.

The challenge is that dynamics (like contagion or opinion spread) on these complex structures are often studied in isolation. Researchers haven't had a unified way to understand how moving from a simple graph to a complex hypergraph actually changes the system's behavior. This paper seeks to find the "interplay" between these orders, asking: Can we transform a complex higher-order process into a simpler one, and what exactly is lost or gained in that translation?

Methods

The authors develop a holistic analysis framework based on two primary pillars:

  1. Transformability Conditions: They derive mathematical conditions under which contagion dynamics on a hypergraph can be made equivalent to those on a simplicial complex or a dyadic (pairwise) graph. This involves "tuning" infection rates to account for the overlapping nature of group interactions.
  2. Information Entropy (System Instability): To compare different network orders globally, they use information entropy ($\Phi$) to quantify "system instability"—essentially measuring the diversity of node states (e.g., Susceptible vs. Infected) over time.
  3. Factor Decomposition: They identify that the difference in instability between a higher-order network and its pairwise counterpart ($\Delta\Phi^*$) can be broken down into two additive components:
  4. Dynamical Factor ($\Delta\Phi_1$): Stemming from the nonlinearity of the contagion process itself.
  5. Structural Factor ($\Delta\Phi_2$): Stemming from the physical arrangement of group interactions (multiple contacts between nodes).

They validate this using mean-field theory to derive master equations and use Support Vector Machines (SVM) to test if the dynamics of different network orders can be statistically distinguished.

Datasets

The study utilizes a mix of empirical and synthetic datasets:

  • Empirical Social Networks: Real-world interaction data from a primary school, a high school, the Email-Enron network, and NDC-classes (national drug code).
  • Synthetic Networks: Generated using the HyperCL configuration model, including scale-free hypergraphs (SFH), Barabási–Albert hypergraphs (BAH), and Erdős–Rényi hypergraphs (ERH) to test the framework across varying topologies.

Key Insights and Ideas

  • The Additivity Principle: The most significant "aha moment" is that the gap between higher-order and pairwise dynamics is simply the sum of the structural gap and the dynamical gap. These two factors are decoupled and additive, providing a clear "recipe" for understanding system complexity.
  • Dynamics are Interconnected: Higher-order dynamics are not independent "islands." By satisfying specific transformation conditions, a researcher can make a pairwise network behave exactly like a hypergraph, revealing that higher-order interactions primarily add layers of nonlinearity and structural redundancy rather than entirely new physical laws.
  • Entropy as a Universal Metric: While local infection probabilities are hard to compare across different network types, global information entropy provides a stable, unified "language" to describe the evolution of any dynamical system.

Key Results

  • Successful Mapping: The authors successfully derived conditions (C1, C2, C3) that allow for the consistent transformation of SI (Susceptible-Infected) contagion across hypergraphs, simplicial complexes, and dyadic graphs.
  • Theoretical Alignment: Quantitative predictions of entropy differences ($\Delta\Phi^T$) showed strong agreement with simulation results ($\Delta\Phi^S$) across all tested datasets.
  • Generalizability: The framework was successfully extended from contagion models to opinion dynamics (using the Deffuant-Weisbuch and Voter models). It accurately captured how individuals in groups reach consensus or maintain polarized views.
  • Indistinguishability: SVM classifiers showed that when the transformation conditions are met, the F1-score (a measure of classification accuracy) drops significantly, proving that the dynamics on different network orders become statistically indistinguishable.

Limitations and Future Work

  • Network Heterogeneity: The basic mean-field approach sometimes struggles with highly heterogeneous networks (those with extreme "hubs" and many "leaf" nodes). While the authors introduced an "effective degree" concept to mitigate this, more refined models for extreme topologies are needed.
  • Static vs. Temporal: The current work focuses on static higher-order networks. In reality, group interactions are often temporal (groups form and dissolve over time). Extending "transformability" to time-varying networks is a logical next step.
  • Broader Dynamics: While contagion and opinion models were covered, other complex processes like evolutionary games or synchronization in power grids remain to be explored within this framework.

Summary updated

Abstract

Recent studies have investigated various dynamic processes characterizing collective behaviors in real-world systems. However, these dynamics have been studied individually in specific contexts. In this article, we present a holistic analysis framework that bridges the interplays between dynamics across networks of different orders, demonstrating that these processes are not independent but can undergo systematic transformations. Focusing on contagion dynamics, we identify and quantify dynamical and structural factors that explains the interplay between dynamics on higher-order and pairwise networks, uncovering a universal model for system instability governed by these factors. Furthermore, we validate the findings from contagion dynamics to opinion dynamics, highlighting its broader applicability across diverse dynamical processes. Our findings reveal the intrinsic coupling between diverse dynamical processes, providing fresh insights into the distinct role of complex dynamics governed by higher-order interactions.
BibTeX
@misc{xie2025transformability,
  author = {Ming Xie and Shibo He and Aming Li and Zi‐Ke Zhang and Youxian Sun and Jiming Chen},
  title = {{Transformability reveals the interplay of dynamics across different network orders}},
  year = {2025},
  eprint = {2501.16016},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2501.16016},
  howpublished = {arXiv},
}

Loading related content...

Figure