A time-varying recurrent neural system for convex programming

The asymptotic stability of a recurrent neural network with monotonically time-varying penalty parameter for optimization is theoretically justified. The conditions of feasibility of solutions generated by the recurrent neural networks are characterized. The conditions of optimality of solutions to convex programming problems generated by the recurrent neural networks are characterized. The design methodology of the operating characteristics of the recurrent neural networks are presented by illustrative examples.<>