A global asymptotic stability result for a class of totally asynchronous discrete nonlinear systems (Q1293211)

The paper brings the proof of global stability for a class of nonlinear discrete-time dynamical systems where only a subset of variables is updated at any given instant. The authors claim that ``as far as they know, the main result of the paper is new in the area of convergence of asynchronous iterations''. But they go further to claim that ``in the context of neural networks, the main result is also the global version of a local stability'' obtained earlier by them in other cases. The idea exploited by the authors, namely the use of a Lyapunov function to prove the global stability of equilibrium points for Hopfield-like neural networks is more than 15 years old, now. The point is that the authors treat the problem only partly. In fact, this is done under very restrictive conditions, sufficient however for the stability of the asynchronous iterations considered, when they obtain the global stability of the equilibrium state. One should also mention that the authors do not take into account a learning process as it is almost the case. Besides their asynchronous iterations need not have much in common with standard neural networks. At the end an interesting question remains: to know if from the proof of global stability of equilibrium points for fully synchronized neural networks, i.e., parallel updating at any moment of all neurons, there follows a proof of global stability of such equilibrium points in the case of asynchronously updated activations of neurons.