节点文献
一个等式约束问题的拟Newton-信赖域型方法及其收敛性
A Quasi-Newton-Trust Region Methodfor Equality Constrained Optimizationand Its Convergence Properties
【摘要】 在[1]中,Vardi提出一个信赖域方法,而收敛性证明却是在精确λ-搜索下给出的.本文在[1]的基础上提出一个新的算法-拟Newton-信赖域型算法,并证明该算法是全局收敛的.通过利用二阶修正技术去修正该算法,我们证明了该算法是局部超线性收敛的.
【Abstract】 In [1], Vardi proposed a trust region algorithm for equality constrained optimization, but the global convergence was obtained under exact A-search. In this paper, we propose a new method-quasi-Newton-trust region algorithm, which is based on the method in [1] and prove that it is convergent. By using twice order correction technique to modify the algorithm, we prove that the algorithm is convergent superlinearly.
【关键词】 拟Newton-信赖域型方法;
全局收敛性;
超线性收敛速度;
精确罚函数;
等式约束最优化;
【Key words】 Quasi-Newton-trust region method; global convergence; super-linear convergence; exact penalty function; equality constrained optimization.;
【Key words】 Quasi-Newton-trust region method; global convergence; super-linear convergence; exact penalty function; equality constrained optimization.;
【基金】 国家自然科学基金资助项目(GrantNo.39830070,10171055)
- 【文献出处】 运筹学学报 ,Or Transactions , 编辑部邮箱 ,2001年04期
- 【分类号】O242.23
- 【被引频次】7
- 【下载频次】105