摘要
讨论一般约束最优化问题,利用序列二次规划(SQP)技术和强收敛方法思想建立问题的一个新的拟可行下降算法,算法每次迭代只需解一个要求较弱的二次规划或用广义投影技术产生搜索方向。分析和论证了算法的全局收敛性、强收敛性、超线性收敛性和二次收敛率。
Optimization problems with general constraints are discussed. By using the SQP technique
and the idea of strongly convergent method, a new quasi-feasible descent algorithm is presented. In
order to yield the search direction at each iteration of the algorithm, only one weakly quadratic pro-
gramming or a generalized projection need to be solved or computed. The global convergence, strong
convergence, superlinear convergence and the rate of quadratic convergence are analyzed and proved.
出处
《工程数学学报》
CSCD
北大核心
2004年第4期525-530,共6页
Chinese Journal of Engineering Mathematics
基金
国家自然科学基金(10261001)
广西科学基金(0236001
0249003)
关键词
一般约束
最优化
SQP方法
超线性收敛
二次收敛
general constraints
optimization
SQP methods
superlinear convergence
quadratic convergence
