周 婷, 郭文彬, 崔 燕
(1.衡水學(xué)院 數(shù)學(xué)與計算機學(xué)院 河北 衡水 053000;2.聊城大學(xué) 數(shù)學(xué)科學(xué)學(xué)院 山東 聊城 252059;3.南京理工大學(xué) 計算機科學(xué)與技術(shù)學(xué)院 江蘇 南京 210094)
研究線性方程組
Ax=b,
(1)
其中A是一個n階實矩陣,x和b是n維實向量.不失一般性,令A(yù)=I-L-U,其中I是單位矩陣,-L和-U分別是A的嚴格下三角和嚴格上三角矩陣.w和r是實參數(shù),w≠0,那么基本的AOR迭代法的迭代矩陣[1]為
Tr,w=(I-rL)-1[(1-w)I+(w-r)L+wU],
(2)
眾所周知,當參數(shù)w和r取特定的值時,可得到SOR,Gauss-Seidel,JOR和Jacobi迭代法.當P是非奇異矩陣時,把線性方程組(1)轉(zhuǎn)化為等價的預(yù)條件形式為
PAx=Pb.
(3)
本文給出兩類新的預(yù)條件矩陣Pα=I+Sα和Pβ=I+Sβ,這里,
定義1[15]設(shè)A=(aij)∈Rn×n.若對?i≠j有aij≤0,稱A為Z-矩陣; 若A=sI-B,B≥0,且s>ρ(B),其中ρ(B)表示矩陣B的譜半徑,則稱A為非奇異M-矩陣; 如果對?i,j滿足aij≥0(aij>0),則稱A為非負矩陣(正矩陣),記為A≥0(A>0).類似的可定義非負(正)向量.
引理1[16]設(shè)A是Z-矩陣,A是M-矩陣當且僅當存在向量u=(u1,…,un)T>0使得Au>0.
引理2[3]令A(yù)是一個H-矩陣,如果0≤r≤w≤1,w≠0,則ρ(Tr,w)<1.
考慮預(yù)條件矩陣Pα=I+Sα,令A(yù)α=(I+Sα)A=Dα-Lα-Uα,其中Dα,-Lα,-Uα分別是Aα的對角、嚴格下三角和嚴格上三角部分,則對應(yīng)的預(yù)條件AOR迭代法的迭代矩陣為
(4)
類似的,考慮預(yù)條件矩陣Pβ=I+Sβ.令A(yù)β=(I+Sβ)A=Dβ-Lβ-Uβ,其中Dβ,-Lβ,-Uβ分別是Aβ的對角、嚴格下三角和嚴格上三角部分.則對應(yīng)的預(yù)條件AOR迭代法的迭代矩陣為
(5)
>0.
證明令(〈Aα〉u)i是向量〈Aα〉u的第i個元素.則有
(6)
(7)
當0≤αi≤1(i=1,…,n-1)時,有
>0.
(8)
>0.
(9)
>0.
證明令(〈Aβ〉v)i是向量〈Aβ〉v的第i個元素.則有
(10)
(11)
當0≤βi≤1(i=2,…,n)時,有
>0.
(12)
>0.
(13)
考慮線性方程組(1)的系數(shù)矩陣A[13],這里,
[1] Hadjimos A.Accelerated over-relaxation method[J].Math Comp,1978,32(141): 149-157.
[2] Liu Qingbing,Chen Guoliang,Cai Jing.Convergence analysis of the preconditioned Gauss-Seidel method forH-matrices[J].Comput Math Appl,2008,56(8): 2048-2053.
[3] Li Yaotang,Yang Shunfeng.A multi-parameters preconditioned AOR iterative method for linear systems[J].Appl Math Comput,2008,206(1): 465-473.
[4] Wu Meijun,Wang Li,Song Yongzhong.Preconditioned AOR iterative method for linear systems[J].Appl Numer Math,2007,57(5/6/7): 672-685.
[5] Kotakemori H,Harada K,Morimoto M,et al.A comparison theorem for the iterative method with the preconditioner (I+Smax)[J].J Compute Appl Math,2002,145(2): 373-378.
[6] Wang Xuezhong,Huang Tingzhu,Fu Yingding.Comparison results on preconditioned SOR-type iterative method forZ-matrices linear systems[J].J Comput Appl Math,2007,206(2): 726-732.
[7] Wang Hongjuan,Li Yaotang.A new preconditioned AOR iterative method forL-matrices[J].J Compute Appl Math,2009,229(1): 47-53.
[8] 劉慶兵,陳果良.預(yù)條件AOR和2PPJ迭代法收斂性的注記[J].華東師范大學(xué)學(xué)報: 自然科學(xué)版,2009 (4): 26-34.
[9] Kotakemori H,Niki H,Okamoto N.Convergence of a preconditioned iterative method forH-matrices[J].J Comput Appl Math,1997,83(1): 115-118.
[10] Wang Li,Song Yongzhong.Preconditioned AOR iterative methods forM-matrices[J].J Comput Appl Math,2009,226(1): 114-124.
[11] Huang Tingzhu,Wang Xuezhong,Fu Yingding.Improving Jacobi methods for nonnegativeH-matrices linear systems[J].Appl Math Comput,2007,186(2): 1542-1550.
[12] Zheng Bing,Miao Shuxin.Two new modified Gauss-Seidel methods for linear system withM-matrices[J].J Compute Appl Math,2009,233(4): 922-930.
[13] Kohno T,Kotakemori H,Niki H.Improving the modified Gauss-Seidel method forZ-matrices[J].Linear Algebra Appl,1997,267: 113-123.
[14] 李世存.曲譜集構(gòu)造Jacobi矩陣[J].鄭州大學(xué)學(xué)報:自然科學(xué)版,1986,4(2):44-47.
[15] Varga R S.Matrix Iterative Analysis[M].2th Edition.Berlin:Springer,2000.
[16] Fan K Y.Topological proofs for certain theorems on matrices with non-negative elements[J].Monatsh Math,1958,62(3): 219-237.