李欣,朱景福,李啟勇
求解多右端對稱線性方程組的BMINBACK方法的理論分析與執(zhí)行
李欣,朱景福,李啟勇
(廣東石油化工學(xué)院 理學(xué)院,廣東 茂名 525000)
研究求解對稱多右端線性方程組的極小向后擾動塊方法.在塊Lanczos執(zhí)行的過程中采用極小向后擾動范數(shù)作為終止算法的條件,對向后擾動的格式及其范數(shù)極小值的求法做深入的理論分析,進(jìn)一步論證了塊方法與向后擾動相結(jié)合的可行性.通過多組數(shù)值實(shí)驗(yàn)驗(yàn)證新方法的有效性和優(yōu)越性.
Krylov子空間;多右端;線性方程組;塊Lanczos方法;病態(tài)矩陣
本文研究求解多右端對稱線性方程組
(2)
本文研究求解多右端對稱線性方程組(1)的極小向后擾動塊方法,把Lanczos塊方法和極小向后擾動相結(jié)合,在塊Lanczos的過程中采用極小向后擾動范數(shù)作為終止算法的條件,對極小向后擾動的形式及其范數(shù)極小值的求法和殘量進(jìn)一步深入論證,給出改進(jìn)的新算法.研究了新方法求解的殘量分析,通過多組數(shù)值實(shí)驗(yàn)驗(yàn)證新方法的有效性和優(yōu)越性.
考慮把多右端線性方程組(1)的近似解看作向后擾動方程
的精確解.
式(4)中的三項(xiàng)遞推公式可用矩陣
殘量矩陣表示為
證明根據(jù)式(9)及文獻(xiàn)[15]中第一章定理1.6,可以直接得到式(10),詳細(xì)證明方法與文獻(xiàn)[9]中定理一致. 證畢.
從而
基于推導(dǎo)過程,自然可得到式(9)中所述最小化問題的解的定理.
證明證明過程根據(jù)式(13)(15)(16)得到. 證畢.
給出求解對稱線性方程組的極小向后擾動的塊方法(BMINBACK).
算法(BMINBACK):
End
本文的數(shù)值實(shí)驗(yàn)均在個(gè)人計(jì)算機(jī)上實(shí)現(xiàn),具體配置:CPUintel(R)Core(TM)i7-8550u,主頻1.8 GHz,內(nèi)存8 GB,系統(tǒng)Win10 企業(yè)版,軟件Matlab R2014a.
應(yīng)用塊Lanczos方法(BLANCZOS)和極小向后擾動塊方法(BMINBACK)的求解結(jié)果見表1.
表1 例1的BLANCZOS和BMINBACK數(shù)值分析
由表1可以看出,在求解方程組(1)時(shí),BMINBACK計(jì)算速度接近BLANCZOS的計(jì)算速度.
表2 例2的BLANCZOS和BMINBACK數(shù)值分析
由表2可以看出,在求解方程組(1)時(shí),BMINBACK計(jì)算速度接近BLANCZOS的計(jì)算速度.BMINBACK的殘量精度遠(yuǎn)遠(yuǎn)高于BLANCZOS的殘量精度.
圖1 例2的BLANCZOS和BMINBACK的殘量對比
致謝:感謝南京航空航天大學(xué)理學(xué)院戴華教授給出寶貴的指導(dǎo)意見!
[1] O′Leary D P.The block conjugate gradient algorithm and related methods[J].Linear Algebra,1980,29:293-322.
[2] Underwood R R.An Iterative Block Lanczos Method for the Solution of Large Sparse Symmetric Eigenproblems[M].Stanford:Stanford University Press,1975.
[3] Saad Y.On the rates of convergence of the Lanczos and the block-Lanczos methods[J].SIAM J Numerical Analysis,1980,17:687-706.
[4] Simoncini V,Gallopoulos E.Convergence properties of block GREMS and matrix polynomials[J].Linear Algebra Appl,1996, 247:97-119.
[5] Simoncini V,Gallopoulos E.A hybrid block GMRES method for nonsymmetric systems with multiple right-hand sides[J].J Comput Appl Math,1996,66:457-469.
[6] Freund R W,Malhotra M.A block QMR algorithm for non-Hermitian linear systems with multiple right-hand sides[J].Linear Algebra Appl,1997,254:119-157.
[7] Gu Guiding,Wu Hebing.A block EN algorithm for nonsymmetric linear systems with multiple right-hand sides[J].Linear Algebra Appl,1999,299:1-20.
[8] GUTKNECHT M H.Block Krylov space methods for linear systems with multiple right-hand sides:an introduction[C]//Siddiqi A H,Duff S,Christensen O I.Modern Mathematical Models,Methods and Algorithms for Real World Systems.New Delhi:Anamaya Publishers,2007:420-447.
[9] Kasenally E M.A generalized minimum backward error algorithm for solving nonsymmetric linear systems[J].SIAM J Sci Comput,1995,16:698-719.
[10] Cao Z.Total generalized minimum backward error algorithm for solving nonsymmetric linear systems[J].Journal of Comp Math, 1998,16:539-550.
[11] 李欣,戴華.解對稱線性方程組的總體最小擾動方法[J].南京大學(xué)學(xué)報(bào)(數(shù)學(xué)半年刊),2005,22(2):315-322.
[12] 李欣,朱景福.循環(huán)收縮QMR方法[J].哈爾濱工業(yè)大學(xué)學(xué)報(bào),2009,41(9):225-227.
[13] Dai Hua.Two algorithms for symmetric linear system with multiple right-hand sides[J].Numerical Mathematics,2000,9(1): 91-110.
[14] Ben-Israel A,Greville T N E.Generalized Inverses:Theory and Applications[M].Berlin:Springer Science & Business Media,2003.
[15] WangG R.Matrix and Operator Generalized Inverse[M].Peking:Science Press,1998.
[16] Mirsky L.A trace inequality of John von Neumann[J].Monatshefte für mathematik,1975,79(4):303-306.
[17] 孫繼廣.矩陣擾動分析[M].北京:科學(xué)出版社,2001.
Theoretical analysis and implementation of the BMINBACK method for symmetric linear systems with multiple right-hand sides
LI Xin,ZHU Jingfu,LI Qiyong
(School of Science,Guangdong University of Petrochemical Technology,Maoming 525000,China)
The minimal backward perturbation block method for solving symmetric linear systems with multiple right-hand sides is studied.The minimal backward perturbation norm is used as the termination condition for the block Lanczos process,the theoretical analysis is discussed on the form of backward perturbation and the method of finding the minimum norm,the feasibility of combining the block process with backward perturbation is further demonstrated.Some numerical experiments are carried out to verify the effectiveness and superiority of the new method.
Krylov subspace;multiple right-hand sides;linear system;block Lanczos process;ill-conditioned matrix
1007-9831(2022)11-0001-07
O241.6
A
10.3969/j.issn.1007-9831.2022.11.001
2022-04-06
廣東石油化工學(xué)院人才引進(jìn)項(xiàng)目(2018rc44,2018rc45)
李欣(1968-),女,黑龍江訥河人,教授,碩士,從事數(shù)值代數(shù)研究.E-mail:2725696255@qq.com
朱景福(1970-),男,黑龍江克山人,教授,博士,從事計(jì)算機(jī)應(yīng)用研究.E-mail:v0085@126. com