通过引入新的加速变量,对解决鞍点问题的GSSOR-like算法进行了修正,得到了更一般的新的GSSOR-like算法(记作NGSSOR-like),并研究了新算法收敛的充分必要条件。最后,通过对数值例子的求解表明,选择合适的参数后,NGSSOR-like算法比GSSOR-like算法具有更快的迭代速度和更少的迭代次数。
By adding new accelerating parameters,the GSSOR-like method is modified for solving the saddle point problems,and the NGSSOR-like method is presented.Then the analysis of its convergence conditions is given.Finally,numerical results show that this method outperforms the old one.