Using Spectral Radius Ratio for Node Degree to Analyze the Evolution of Scale Free Networks and Small World Networks

In this paper, we show the evaluation of the spectral radius for node degree as the basis to analyze the variation in the node degrees during the evolution of scale-free networks and small-world networks. Spectral radius is the principal eigenvalue of the adjacency matrix of a network graph and spectral radius ratio for node degree is the ratio of the spectral radius and the average node degree. We observe a very high positive correlation between the spectral radius ratio for node degree and the coefficient of variation of node degree (ratio of the standard deviation of node degree and average node degree). We show how the spectral radius ratio for node degree can be used as the basis to tune the operating parameters of the evolution models for scale-free networks and small-world networks as well as evaluate the impact of the number of links added per node introduced during the evolution of a scale-free network and evaluate the impact of the probability of rewiring during the evolution of a small-world network from a regular network.

💡 Research Summary

The paper introduces a novel metric – the spectral radius ratio for node degree (R = λ₁/⟨k⟩) – and demonstrates its usefulness for monitoring and controlling the evolution of two canonical network models: the Barabási‑Albert (BA) scale‑free model and the Watts‑Strogatz (WS) small‑world model. The spectral radius λ₁ is the leading eigenvalue of the adjacency matrix and is known to lie between the maximum degree Δ and the average degree ⟨k⟩. By normalizing λ₁ with ⟨k⟩, the authors obtain a dimensionless quantity that captures the heterogeneity of the degree distribution.

A large‑scale simulation study is conducted. For the BA model, networks of sizes N = 1 000, 5 000, and 10 000 are grown while varying the number of edges added per new node (m = 1…10). For each configuration, λ₁, ⟨k⟩, the standard deviation of degrees σ_k, and the coefficient of variation CV = σ_k/⟨k⟩ are recorded. The results show a very strong linear correlation (Pearson r > 0.95) between R and CV across all N and m. As m increases, both R and CV decrease, reflecting a shift from a highly hub‑dominated topology toward a more homogeneous degree distribution. The correlation persists when N grows, although a slight upward drift of R and CV is observed for fixed m, indicating that larger networks naturally develop a few extreme hubs even under the same attachment rule.

For the WS model, the authors start from a regular ring lattice (degree k = 2K) and rewire each edge with probability p (p = 0, 0.01, 0.05, 0.1, 0.2, 0.5, 1). While λ₁ and ⟨k⟩ remain essentially constant (because rewiring does not change the total number of edges), σ_k and CV increase sharply as p grows. A pronounced “transition zone” appears around p ≈ 0.1–0.2, where R and CV jump from values near 1 (almost regular) to significantly larger values, indicating that a modest amount of random shortcuts is sufficient to generate a highly heterogeneous degree profile. Again, the correlation between R and CV exceeds 0.95.

These empirical findings lead to a practical methodology: given a desired degree heterogeneity (expressed as a target CV), one can use the established R–CV relationship to compute the corresponding R_target, then select the model parameter (m for BA, p for WS) that yields an R closest to R_target. This inverse design approach allows researchers and engineers to generate networks with prescribed variability without repeatedly measuring the full degree distribution. Because λ₁ can be obtained with a single eigenvalue computation, the method scales well to very large graphs.

The paper also discusses limitations. While R captures degree heterogeneity, it does not directly encode other structural aspects such as clustering coefficient, average path length, or community structure. Consequently, R should be complemented with additional metrics when a comprehensive network quality assessment is required. Future work is suggested to integrate R into multi‑objective optimization frameworks and to validate the approach on empirical networks from biology, sociology, and technology.

In summary, the spectral radius ratio for node degree emerges as a robust, computationally inexpensive indicator of degree variability. Its strong correlation with the coefficient of variation enables precise tuning of the BA and WS model parameters, offering a valuable tool for both theoretical investigations of network growth dynamics and practical network design tasks.