Stabilization and structure optimization of asymmetric neural networks

Abstract

A new approach for stabilizing asymmetric neural networks is suggested in the present paper. A nonlinear control system is used as the model of the given neural network, then by decomposing the interconnection matrix and by using the deremainder control technique, some asymptotically stable outputs can be attained as the network is driven by prescribed inputs. A structure optimization scheme for the network is also suggested following the same technique.<>