摘要
本文对于无约束最优化问题提出了一类新的非单调信赖域算法。与通常的非单调信赖域算法不同,当试探步不成功时,并不重解信赖域子问题,而采用非单调线搜索,从而减小了计算量,在适当的条件下,证明了此算法的全局收敛性。
A new nonmonotonic trust algorithm for unconstrained optimization problems is presented. Different from usual nonmonotonic trust region method, it takes linesearch instead of resolving the subproblem when the trail step is not successful. This may allow a considerable computational saving. Global convergence is proved under certain conditions.
出处
《数学杂志》
CSCD
北大核心
2003年第3期290-294,共5页
Journal of Mathematics
关键词
无约束优化
非单调信赖域算法
线搜索
收敛性
unconstrained optimization
nonmonotonic trust region algorithm
line search
convergence
