标题:
一种非单调自适应信赖域法A Nonmonotonic Self-Adaptive Trust Region Algorithm
作者:
杭丹, 王晓燕, 郝建忠, 王娅
关键字:
无约束最优化, 信赖域, 固定步长, 非单调技术, 全局收敛性 Unconstrained Optimization; Trust Region; Fixed Stepsize; Nonmonotonic Technique; Global
Convergence
期刊名称:
《Pure Mathematics》, Vol.3 No.5, 2013-09-03
摘要:
求解无约束优化问题，本文给出了一种改进的非单调信赖域算法。采用了非单调技术，当试探步不被接受时，下一步的迭代步长由一个固定公式给出。并直接给出调整信赖域半径公式。数值试验结果表明新算法的有效性。在适当的条件下，证明了算法的全局收敛性。We propose an improved nonmonotonic trust region algorithm. Our method is to use the nonmonotone technique and if the trial step is rejected, the stepsize is computed by a fixed formula. The trust region radius is updated at a variable rate. Numerical experiment results show that the new algorithm is effective. Under mild conditions, we prove that the algorithm is global convergence.