Global asymptotic stability of a class of feedback neural networks with an application to optimization problems (Q1126905)

The chaos of feedback neural networks containing Hopfield like networks as a special case is treated from the point of view of global asymptotic stability of a unique equilibrium. But one should mention that the class considered here has a form similar to that of Cohen-Grossberg neural networks. The authors then use some results from homotopy theory to establish criteria for the existence and uniqueness of an equilibrium for such networks. This problem is reformulated by means of differential inequalities for which Lur'e type Lyapunov functions are employed. Then criteria for global asymptotic stability of an equilibrium are found out. These are more general than several of those given in the literature. The effects of perturbations on the stability of the equilibrium, or the robustness of such solutions, is analyzed, too. This allows for some intervals of variations of parameters describing the feedback neural networks to keep the global asymptotic stability of a unique equilibrium. The results obtained are then applied to an optimization problem in the form of a programming problem. In a concrete example of a linear programming problem, the methods used here are compared with results obtained in the literature. It seems that the results of the paper are more flexible.