A weighted simpler GMRES algorithm for shifted linear systems
-
摘要: 结合加权策略和简化的广义最小残量算法(GMRES),提出可有效求解位移线性方程组的加权简化GMRES算法,并给出加权简化GMRES算法与简化GMRES算法之间的联系与性质,最后数值算例给出了新算法的有效性.Abstract: Combining the strategy of weighted and simpler GMRES methods, this paper presents a weighted simpler GMRES algorithm for solving shifted linear systems, and gives some properties of the proposed algorithm. Numerical results illustrate the performance and effectiveness of the algorithm.
-
Key words:
- shifted linear systems /
- simpler GMRES /
- weighted matrix
-
表 1 实验所用矩阵信息
Tab. 1 The content of the test matrices
矩阵名称 大小 非零元个数 条件数 应用领域 add20 2 395$\times$2 395 13 151 1.76$\times10^{4}$ 半导体系统设计技术 bfw782a 782$\times$782 7 516 4.6$\times10^{3}$ 电磁学 memplus 17 758$\times$17 758 126 150 2.67$\times10^{5}$ 电路仿真 sherman4 1 104$\times$1 104 3 786 7.16$\times10^{3}$ 计算流体力学 young1c 841$\times$841 4 089 1.01$\times10^{3}$ 声学 CONF6.0-00L4X4-2000 3 072$\times$3 072 119 808 8.35$\times10^{2}$ QCD 表 2 SGMRES-Sh与WSGMRES-Sh的数值结果比较
Tab. 2 Comparison of SGMRES-Sh and WSGMRES-Sh
矩阵名称 SGMRES-Sh 加权矩阵 WSGMRES-Sh $cpu$ $mv$ $cpu$ $mv$ add20 0.341 1 199 $D^1$ 0.352 1 531 $D^2$ 0.150 533 bfw782a 0.525 3 779 $D^1$ 0.383 2 837 $D^2$ 0.199 1 395 memplus 11.843 7 921 $D^1$ 10.391 6 077 $D^2$ 3.099 1 819 sherman4 0.130 663 $D^1$ 0.080 447 $D^2$ 0.100 600 young1c 1.330 5 524 $D^1$ 1.109 5 044 $D^2$ 0.989 4 489 $\rm A_{\rm QCD}$ 2.812 2 413 $D^1$ 0.660 509 $D^2$ 0.683 509 -
[1] SIMONCINI V, PEROTTI F. On the numerical solution of $(\lambda^2A + \lambda B + C)x = b$ and application to structual dynamics[J]. SIAM J Sci Comput, 2002, 23(6): 1876-1898. [2] DARNELL D, MORGAN R B, WILCOX W. Deflation of eigenvalues for iterative methods in lattice QCD[J]. Nucl Phys B—Proceedings Supplements, 2004, 129/130: 856-858. doi: 10.1016/S0920-5632(03)02734-8 [3] LANGVILLE A N, MEYER C D. Google's PageRank and Beyond: The Science of Search Engine Rankings[M]. Princeton: Princeton University Press, 2006. [4] AHMAD M I, SZYLD D B, VAN GIJZEN M B. Preconditioned multishift BiCG for $\mathcal{H}_2$-optimal model reduction[J]. SIAM J Matrix Anal Appl, 2017, 38(2): 401-424. doi: 10.1137/130914905 [5] WALKER H, ZHOU L. A simpler GMRES[J]. Numer Linear Algebra Appl, 1994, 1(6): 571-581. doi: 10.1002/nla.1680010605 [6] JING Y F, YUAN P, HUANG T Z. A simpler GMRES and its adaptive variant for shifted linear systems[J]. Numer, Linear Algebra Appl, 2017, 24(1): 1-7. http://cn.bing.com/academic/profile?id=dcf37798e116925d24332d3d5a758f34&encoded=0&v=paper_preview&mkt=zh-cn [7] 杨圣炜, 卢琳璋.一种加权的Simpler GMRES算法[J].厦门大学学报(自然科学版), 2008(4): 484-488. doi: 10.3321/j.issn:0438-0479.2008.04.006 [8] CAO Z H, YU X Y. A note on weighted FOM and GMRES for solving nonsymmetric linear systems[J]. Appl Math Comput, 2004, 151: 719-727. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=8502da35ae4222aa42c72336cb6e8f9c [9] ZHONG H X, WU G. Thick restarting the weighted harmonic Arnoldi algorithm for large interior eigenproblems[J]. Int J Comput Math, 2011, 88(5): 994-1012. doi: 10.1080/00207160.2010.489639