摘要
人工神经网络在解决优化问题方面具有速度和精度的优势.将一种新的Hopfield网络模型与信赖域技术融合起来,采用逐次二次规划方法将约束非线性规划问题转换成一系列的二次规划子问题,并采用信赖域技术协调计算速度与精度之间的矛盾.同时通过Hopfield网络求解各个二次规划子问题,得到原来规划问题的最优解.通过对大量函数进行仿真计算,取得了很好的仿真结果.
Artificial Neural Networks(ANNs)have the merits of their rapidness and accuracy.This paper describes a new algorithm.It can solve the constrained Nonlinear Programming(NLP)problems by a new Hopfield network and Trust Region(TR)strategy.Using the Successive Quadratic Programming(SQP)technology,the original NLP problems will be converted into a series of quadratic programming subproblems.Then the TR strategy can harmonize the conflict of the rapidness and accuracy;the Hopfield network can solve the subproblems.In the final,the algorithm will get the optimal solution of the orginal NLP problem.The effectiveness of the algorithm is demonstrated by simulation results.
出处
《天津大学学报（自然科学与工程技术版）》
EI
CAS
CSCD
北大核心
2002年第6期705-709,共5页
Journal of Tianjin University：Science and Technology
基金
国家自然科学基金资助项目(79970042)
