一种改进的的非单调自适应信赖域算法  

A Modified Method of Nonmonotone Self-adaptive Trust-region for Unconstrained Optimization

作　　者：钱慧敏[1] 周新慧[1] 

机构地区：[1]西安电子科技大学理学院,西安710071

出　　处：《长春理工大学学报（自然科学版）》2013年第5期147-149,153,共4页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：中央高校基本科研业务费专项资助项目(K50513100007)

摘　　要：非单调信赖域算法是解决无约束优化问题的一种有效的算法。本文针对无约束优化问题提出了一种改进的非单调自适应信赖域算法。将非单调技术引入到一种新的自适应信赖域算法中,这样做避免了Maratos效应,减少了运算量。并且文中提出了一些假设,证明了这种改进的非单调自适应算法的全局收敛性,最后数据试验验证了这种算法的有效性。The nonmonotone trust-region methods are well-known techniques for solving unconstrained optimization problems.In this paper,a modified nonmonotone self-adaptive trust-region method is proposed for unconstrained optimi-zation. The nonmonotone strategies are introduced into a new self-adaptive trust-region. Maratos effects are avoided and the amount of calculation is reduced. Some reasonable assumptions are proposed. The global convergence properties of the algorithm are established. Lastly preliminary numerical results show that the new method is efficient for solving unconstrained optimization problem.

关 键 词：自适应 信赖域 非单调 全局收敛性 

分 类 号：O221[理学—运筹学与控制论]