解广义二次规划的神经网络

Neural network for the extended quadratic programming problem

考虑了广义二次规划问题 ,基于其鞍点的充要条件 ,提出了求解它的一个神经网络 .运用Lya punov稳定性理论与LaSalle不变原理证明了该网络是稳定的 ,并且收敛于一个精确解 .模拟实验表明新模型不仅是有效的 ,而且非常可靠 .

This paper considers the extended quadratic programming problem. Based on the necessary and sufficient conditions for a saddle point, a neural network for solving it is proposed. It is strictly proved to be stable and convergent to an exact solution by the Lyapunov stability theory and the LaSalle invariance principle. The feasibility and efficiency of the proposed network are confirmed by numerical experiments.