摘要
本文提出一个新的求解非线性不等式约束优化问题的罚函数型序列二次约束二次规划(SQCQP)算法.算法每次迭代只需求解一个凸二次约束二次规划(QCQP)子问题,且通过引入新型积极识别集技术,QCQP子问题的规模显著减小,从而降低计算成本.在不需要函数凸性等较弱假设下,算法具有全局收敛性.初步的数值试验表明算法是稳定有效的.
In this paper, a new penalty-function type sequential quadratically constrained quadratic programming （SQCQP） algorithm for nonlinear inequality constrained optimiza- tion problems is presented. The algorithm solves at each iteration only a quadratically constrained quadratic programming （QCQP） subproblem, and by employing a new active identification set technique, the scale of the QCQP subproblem is greatly decreased, thus the computational cost is also reduced. Without assuming the convexity of the objection function or the constraints, the algorithm possesses global convergence under weaker con- ditions, and some preliminary numerical results show that the proposed algorithm is stable and promising.
出处
《应用数学学报》
CSCD
北大核心
2015年第2期222-234,共13页
Acta Mathematicae Applicatae Sinica
基金
国家自然科学基金(11271086)
广西自然科学基金(2013GXNSFAA019013
2014GXNSFFA118001)
玉林师范学院一般项目(2014YJYB04)资助项目
