神经网络在二次规划问题中的应用

Neural Network for Solving Quadratic Programming Problems

利用对偶神经网络解决了基于线性等式、不等式和有界约束的二次规划问题,表明所研究的对偶神经网络具有整体指数收敛性,与包含高次非线性条件的神经网络相比,所提出的网络使用了更少的神经元,并且网络的体系结构更简单.数值实验结果表明了该方法的有效性.

This paper presents a dual neural network which is globally exponential convergence for solving quadratic programming problems based on linear equation, in-equation and bounded constraint. Compared with other neural network models which contained high-order nonlinear condition, the dual neural network uses less neuron and has simpler structure. Finally, numerical tests show the validity of the model.