摘要
对与约束最优化相关的多面凸锥理论进行了讨论 ,证明了几个重要性质 .利用正基 ,该文对线性约束的非线性规划问题设计了一种新算法 .在该算法中 ,每次迭代时无需求解一个线性规划子问题 ,而且算法实现也比较简单 .该文还证明了只要当目标函数连续时 ,算法或有限步终止于一个K_T点 ,或产生一个无穷点列 ,其每一个聚点皆为K_T点 .
In this paper, some important properties of convex cone are discussed, which are relevant to nonlinear constrained programming. A new algorithm for nonlinear programming with linear constraints is proposed by using positive basis. Comparing with other algorithms, the new algorithm does not need LP problem to be solved in every iterative step and it is easy to implement. The global convergence is also proved.
出处
《曲阜师范大学学报（自然科学版）》
CAS
2000年第4期8-10,共3页
Journal of Qufu Normal University(Natural Science)
基金
山东省自然科学基金资助!(Q97A0 4115 )
关键词
凸锥
非线性规划
约束最优化
convex cone
nonlinear programming
positive basis, K_T point
