arXiv++ Combinatorics

We define an analogue of the shortest-path distance for graphons. The proposed method is rooted on the extension to graphons of Varadhan's formula, a result that links the solution of the heat equation on a Riemannian manifold to its geodesic distance. The resulting metric is integer-valued, and for step graphons obtained from finite graphs it is essentially equivalent to the usual shortest-path distance. We further draw a link between the Varadhan distance and the communicability distance, that contains information from all paths, not just shortest-paths, and thus provides a finer distance on graphons along with a natural isometric embedding into a Hilbert space.

Graph-limit theory focuses on the convergence of sequences of graphs when the number of nodes becomes arbitrarily large. This framework defines a continuous version of graphs allowing for the study of dynamical systems on very large graphs, where classical methods would become computationally intractable. Through an approximation procedure, the standard system of coupled ordinary differential equations is replaced by a nonlocal evolution equation on the unit interval. In this work, we adopt this methodology to explore the continuum limit of random walks, a popular model for diffusion on graphs. We focus on two classes of processes on dense weighted graph, in discrete and in continuous time, whose dynamics are encoded in the transition matrix and the random-walk Laplacian. We also show that previous works on the discrete heat equation, associated to the combinatorial Laplacian, fall within the scope of our approach. Finally, we apply the spectral theory of operators to characterize the relaxation time of the process in the continuum limit.