求解非线性优化问题的记忆梯度摄动投影算法  

Memory gradient perturbed rosen projection method for nonlinear programming with nonlinear inequality constraints

作　　者：桑兆阳 

机构地区：[1]中国石油大学理学院,山东青岛266580

出　　处：《中国石油大学学报（自然科学版）》2012年第4期186-190,共5页Journal of China University of Petroleum(Edition of Natural Science)

基　　金：国家自然科学基金项目(10971118);中央高校基本科研业务费专项资金资助项目(09CX04020A)

摘　　要：利用摄动投影矩阵建立求解非线性约束优化问题的记忆梯度摄动投影下降算法,并证明算法的收敛性,同时给出结合FR、PR、HS参数和拟牛顿方程的记忆梯度摄动投影算法,从而将经典的共轭梯度法推广用于求解约束优化问题。数值结果表明算法是有效的。By using perturbed projection matrix, a new descent memory gradient perturbed projection method for nonlinear constrained optimization problems was developed. The global convergence properties of the new method were discussed. Combining FR, PR, HS parameters and the quasi-Newton equation with the new method, four new classes of memory gradient perturbed projection methods were presented. The classical conjugate gradient method is generalized for constrained optimiza- tion problems. The numerical results show that the new method is effective.

关 键 词：非线性规划 摄动投影 记忆梯度 收敛 

分 类 号：O224[理学—运筹学与控制论]