参数非线性规划问题的L_1精确罚函数神经网络方法分析

ANALYSING L_1 EXACT PENALTY FUNCTION NEURAL NETWORKS METHOD OF PARAMETRIC NONLINEAR PROGRAMMING PROBLEMS

针对约束条件中含有参数的非线性规划问题,提出一种基于L1精确罚函数神经网络的新型计算方法。该方法的罚因子为有限实数,并且取值小,便于硬件实现。在改进现有网络模型的基础上,利用最速下降原理构造了神经网络的动力学方程。给出所提神经网络模型在优化计算方面的具体应用步骤。最后,通过数值实例进行仿真验证,结果表明所提方法能够更加快速、精准地收敛于原规划问题的最优解。

In view of nonlinear programming problem with parameters in constraints, we propose a new computational method which is based on L1 exact penalty function neural networks. The method has the penalty factors in finite real number and valuing small, which is easy for hardware implementation. Based on improving the existing network model, we construct dynamics equation of neural network using gradi- ent descent method. We also give specific application steps of the proposed neural network model in computation optimisation. Finally, simu- lation verification is carried out through numerical examples, the results demonstrate that the proposed method is able to converge to the opti- ma~ solution of original programming problem.