Dynamics on Modular Networks with Heterogeneous Correlations

We develop a new ensemble of modular random graphs in which degree-degree correlations can be different in each module and the inter-module connections are defined by the joint degree-degree distribution of nodes for each pair of modules. We present an analytical approach that allows one to analyze several types of binary dynamics operating on such networks, and we illustrate our approach using bond percolation, site percolation, and the Watts threshold model. The new network ensemble generalizes existing models (e.g., the well-known configuration model and LFR networks) by allowing a heterogeneous distribution of degree-degree correlations across modules, which is important for the consideration of nonidentical interacting networks.

💡 Research Summary

The paper introduces a novel ensemble of modular random graphs that allows each module to possess its own degree‑degree correlation structure, while inter‑module connections are governed by a joint degree‑degree distribution specific to each pair of modules. This construction generalizes the classic configuration model and the widely used LFR benchmark by incorporating heterogeneous assortativity or disassortativity across communities, thereby capturing a level of structural heterogeneity that is observed in many real‑world systems such as social, biological, and technological networks.

The authors develop an analytical framework based on message‑passing (or belief‑propagation) equations. For each module (r) they define an activation probability (\theta_r) that depends on (i) the intra‑module degree‑degree distribution (P_r(k,k’)), (ii) the inter‑module joint distribution (P_{rs}(k,k’)) for every other module (s), and (iii) the parameters of the binary dynamics under study. By iterating these equations they obtain closed‑form expressions for the percolation thresholds and the size of the giant component for three representative processes: bond percolation (edge occupation probability (p)), site percolation (node occupation probability (q)), and the Watts threshold model (node‑specific activation threshold (\phi)). The analysis shows that the critical point of the whole network is not a simple function of the global degree distribution; instead it is strongly modulated by the pattern of degree correlations within each community and by how high‑degree nodes are linked across communities.

To validate the theory, the authors generate synthetic networks with a prescribed number of modules, each having distinct degree distributions and correlation profiles, as well as specific inter‑module joint degree distributions. They also extract a real‑world example: a university collaboration network where different academic departments exhibit markedly different assortativity patterns. For each case they run extensive Monte‑Carlo simulations of the three dynamics and compare the empirical critical points and giant‑component growth curves with the predictions of their analytical model. The results demonstrate that the new ensemble reproduces the simulation outcomes with high accuracy, whereas the traditional configuration model (which assumes a single global (P(k,k’))) systematically misestimates the thresholds, especially when modules have opposing correlation tendencies (e.g., one assortative, another disassortative).

The paper’s contributions can be summarized as follows:

1. Model Innovation – A flexible random‑graph construction that encodes heterogeneous degree‑degree correlations at the module level and allows arbitrary joint degree distributions for inter‑module edges.
2. Unified Analytical Tool – A message‑passing formalism that yields exact (in the thermodynamic limit) predictions for a broad class of binary dynamics on these networks.
3. Insight into Critical Phenomena – Demonstration that inter‑module degree correlations can either raise or lower global percolation thresholds, thereby providing a mechanism to tune network robustness or vulnerability through community‑level wiring.
4. Empirical Relevance – Application to synthetic and real data shows that the model captures structural nuances missed by existing benchmarks, offering a more realistic platform for testing interventions such as immunization, information seeding, or cascade control.

Finally, the authors outline several promising extensions: incorporating multi‑state or continuous dynamics (e.g., SIS/SIR epidemic models, synchronization), allowing the modular structure and joint degree distributions to evolve over time, and developing statistical inference techniques to estimate (P_r(k,k’)) and (P_{rs}(k,k’)) from incomplete network observations. Such developments would further bridge the gap between abstract network theory and the practical design and management of complex, interacting systems.