Robust stability of a class of neural networks with time delays (Q1363324)

Neural networks are modelled by nonlinear ODE's with one or two constant delays. It is assumed that the linear part, centred on a static equilibrium, is dominant. The stability analysis is carried out by examining the eigenvalues (roots of quasipolynomials), and their dependence on parameters. The permissible change of the latter determines ``robustness''. This notion does not coincide with an analoguos notion introduced by Andronov and extended by Smale. The stability so determined is local. The possibility of small ``hard-type'' self-oscillations is not considered.