An Improved Neural Network for Solving Optimization of Quadratic Programming Problems

For the quadratic programming problems with both equality and inequality constraints, an improved neural network is proposed based on the Lagrange function reconstructed based on the saddle point theorem of optimization theory. The proposed neural network has less neuron quantity than the traditional method with slack variables does. The stability and convergency of the proposed neural network is investigated. The feasibility of the neural network is verified with computation examples are discussed. The simulation results illustrate the proposed neural network have an effective computational capability and performance for optimization of the quadratic programming problems