Optimally frequency synchronized networks of non identical Kuramoto oscillators

TL;DR

This paper develops a local greedy algorithm to design optimal, symmetric, and economical networks of non-identical Kuramoto oscillators that achieve global frequency synchronization without imposing prior constraints.

Contribution

It introduces a novel unconstrained optimization approach for network topology that ensures synchronization and reveals unexpected correlations between natural frequencies and connection weights.

Findings

Optimal networks are highly symmetric and economical.

Strong correlation between natural frequencies and incoming connection weights.

Unexpected correlation between natural frequencies and outgoing connection weights.

Abstract

Based on a local greedy numerical algorithm, we compute the topology of weighted, directed, and of unlimited extension networks of non identical Kuramoto oscillators which simultaneously satisfy 2 criteria: i) global frequency synchronisation and ii) minimal total mass of the connection weights. This problem has been the subject of many previous interesting studies, but from our best knowledge, this is the first time that no a priori constraint is imposed, either on the form or on the dynamics of the connections. The results are surprising: the optimal networks turn out to be strongly symmetric, very economical, to display a strong rich club structure, and next to the already reported strong correlation between natural frequencies and the weight of incoming connections, we also observe a correlation, even more marked, between these same natural frequencies and the weight of outgoing connections. The latter result is at odds with theoretical predictions.