Random spherical graphs

TL;DR

This paper introduces random spherical graphs (RSGs), analyzing their topological properties and comparing them to traditional random geometric and rectangular graphs, supported by analytical models and simulations.

Contribution

It provides the first analytical expressions for RSG properties and highlights their similarities to RGGs, especially in clustering, with validation through simulations.

Findings

Analytical expressions for RSG degree, connectivity, and clustering.

RSGs exhibit higher clustering coefficients similar to RGGs.

Simulations confirm the accuracy of the theoretical models.

Abstract

This work addresses a modification of the random geometric graph (RGG) model by considering a set of points uniformly and independently distributed on the surface of a $(d-1)$-sphere with radius $r$ in a $d-$dimensional Euclidean space, instead of on an unit hypercube $[0,1]^d$ . Then, two vertices are connected by a link if their great circle distance is at most $s$. In the case of $d=3$, the topological properties of the random spherical graphs (RSGs) generated by this model are studied as a function of $s$. We obtain analytical expressions for the average degree, degree distribution, connectivity, average path length, diameter and clustering coefficient for RSGs. By setting $r=\sqrt{\pi}/(2\pi)$, we also show the differences between the topological properties of RSGs and those of two-dimensional RGGs and random rectangular graphs (RRGs). Surprisingly, in terms of the average clustering coefficient, RSGs look more similar to the analytical estimation for RGGs than RGGs themselves, when their boundary effects are considered. We support all our findings by computer simulations that corroborate the accuracy of the theoretical models proposed for RSGs.