摘要
外梯度法是一种可以用来解决非线性互补问题的常规算法,其计算简单,存储小,但是收敛速度比较慢.本文基于Anderson加速的思想对其进行改进,从理论上证明了新算法的收敛性,并在数值实验上表明该算法不仅比原始算法加速明显,也比投影收缩算法性能优越,而且在大规模问题上的加速效果稳定.
The extragradient method is a general tool to solve the nonlinear complementarity problems（NCP）, which requires less memory and is easy to be implemented, but has a slow convergence.This paper improves it by using Anderson Acceleration and proves the convergence of the new algorithm in theory. In the numerical experiment, it shows that the improved algorithm has better performance than the original method and the projection and contraction methods as well. Besides, it has a steady effect of acceleration in high-dimensional problems.
出处
《应用数学》
CSCD
北大核心
2018年第1期229-236,共8页
Mathematica Applicata
基金
国家自然科学基金(11601318)
