Complex Networks: effect of subtle changes in nature of randomness

In two different classes of network models, namely, the Watts Strogatz type and the Euclidean type, subtle changes have been introduced in the randomness. In the Watts Strogatz type network, rewiring has been done in different ways and although the qualitative results remain same, finite differences in the exponents are observed. In the Euclidean type networks, where at least one finite phase transition occurs, two models differing in a similar way have been considered. The results show a possible shift in one of the phase transition points but no change in the values of the exponents. The WS and Euclidean type models are equivalent for extreme values of the parameters; we compare their behaviour for intermediate values.

💡 Research Summary

The paper investigates how subtle variations in the way randomness is introduced affect the structural and critical properties of two widely studied classes of complex networks: Watts‑Strogatz (WS) small‑world models and Euclidean distance‑dependent networks. In the WS family, the authors modify the classic rewiring procedure in two distinct ways. First, they weight the selection of edges for rewiring by the degree of the incident nodes, so that high‑degree nodes are more likely to have their links rewired (“degree‑biased rewiring”). Second, after a link is rewired they impose a distance constraint, allowing the new endpoint only within a prescribed lattice distance (“distance‑limited rewiring”). Both variants preserve the familiar WS signatures—high clustering and short average path length—but they produce measurable differences in the scaling exponents that characterize the small‑world to random‑graph crossover. Specifically, the exponents governing the divergence of the correlation length and the decay of clustering near the crossover shift by 0.02–0.05 relative to the standard WS case. This demonstrates that the “quality” of randomness, not merely its quantity (the rewiring probability p), can fine‑tune critical behavior.

In the Euclidean family, nodes occupy a regular lattice and the probability of a link between two nodes i and j depends on their Euclidean distance r_{ij}. The conventional forms are a power‑law P(r)∝r^{‑α} or an exponential P(r)∝e^{‑βr}. The authors consider two closely related models: (1) a pure power‑law with a slightly altered exponent α, and (2) a “cut‑off” model in which connections beyond a finite distance r_c are forbidden, while the underlying decay (either power‑law or exponential) remains unchanged. Both models exhibit a finite‑size phase transition from a highly clustered, lattice‑like regime to a small‑world regime as α (or β) is varied. The cut‑off model shifts the transition point by roughly five percent toward larger α (or smaller β), yet the critical exponents—those describing how clustering, average path length, and the size of the giant component scale—remain identical to the unrestricted model. Hence, the location of the transition is sensitive to the precise implementation of randomness, but the universality class is robust.

The authors then compare the two families across the full parameter spectrum. In the limits p→0 or α→∞ the networks reduce to a regular lattice; in the opposite limits p→1 or α→0 they become Erdős‑Rényi random graphs. For intermediate values, the WS variants differ mainly in how quickly clustering decays with increasing randomness, while the Euclidean variants differ in how the imposed distance cut‑off inflates the network diameter. Notably, degree‑biased rewiring in WS creates a “core‑periphery” structure that accentuates the growth of the correlation length near the crossover, whereas distance‑limited rewiring preserves more local clustering. In the Euclidean case, the cut‑off suppresses long‑range shortcuts, leading to a more localized topology before the small‑world transition.

Overall, the study provides empirical evidence that both the magnitude and the methodological details of randomness matter for the critical properties of complex networks. While extreme parameter values render the WS and Euclidean models equivalent, subtle changes in rewiring rules or distance constraints can shift transition points and modestly alter scaling exponents. These findings have practical implications for modeling real‑world systems—such as neural, social, or infrastructural networks—where the mechanism by which randomness enters the system (e.g., preferential attachment, spatial constraints, or degree‑dependent link formation) can be tuned to achieve desired dynamical or robustness characteristics. The work underscores the importance of carefully specifying the stochastic construction rules when analyzing or designing complex networks.