On Pattern Formation in a Class of Graph Neural Networks
Liviu Goraş, Paul Ungureanu, Monica Fira · 2021
In this communication we discuss several aspects regarding pattern formation in a class of networks consisting of identical cells placed in the nodes of a nondirected signed graph. The approach makes use of the spectral decomposition of nodal variables and mode decoupling as well as the root locus method. Connections to Turing patterns, consensus and oscillator synchronization are made.