A modified ANN for convex programming with linear constraints

Convex programming with linear constraints represents a large class of optimization problems which has wide applications, such as linear programming, quadratic programming and some network flow programming problems. In this paper an artificial neural network (ANN) approach based on the Lagrangian multiplier method (Lagrangian ANN) is discussed to solve this problem. The emphasis of the paper is on analyzing the defect of premature of the conventional Lagrangian ANN, and a modification to it in order to overcome the premature defect is presented. It is proved that the modified Lagrangian ANN can always give the optimal solution. Numerical simulations demonstrate the effectiveness of the proposed modification