Message Board

Dear readers, authors and reviewers,you can add a message on this page. We will reply to you as soon as possible!

2018 Volume 40 Issue 8
Article Contents

Qin ZHONG, Yan WANG, Xin ZHOU, et al. Some Inequalities for the Upper Bound of the Spectral Radius of Hadamard Product of Nonnegative Matrices[J]. Journal of Southwest University Natural Science Edition, 2018, 40(8): 77-81. doi: 10.13718/j.cnki.xdzk.2018.08.010
Citation: Qin ZHONG, Yan WANG, Xin ZHOU, et al. Some Inequalities for the Upper Bound of the Spectral Radius of Hadamard Product of Nonnegative Matrices[J]. Journal of Southwest University Natural Science Edition, 2018, 40(8): 77-81. doi: 10.13718/j.cnki.xdzk.2018.08.010

Some Inequalities for the Upper Bound of the Spectral Radius of Hadamard Product of Nonnegative Matrices

More Information
  • Received Date: 27/06/2017
    Available Online: 20/08/2018
  • MSC: O151.21

  • The Hadamard product of non negative matrices is an important problem in the matrix analysis theories. Based on the Hölder inequality, taking into consideration the fact that similar matrices have the same eigenvalues, this paper gives the upper bounds of the spectral radius AB for the Hadamard product of the nonnegative matrices of order n A and B. The new bounds only depend on the entries of nonnegative matrices A and B, therefore, they are easy to calculate. Numerical examples are given to show that the new bounds have improved several existing results.
  • 加载中
  • [1] HORN R A, JOHNSON C R. Topics in Matrix Analysis[M].北京:人民邮电出版社, 2005.

    Google Scholar

    [2] 顾幸生, 刘漫丹, 张凌波.现代控制理论及应用[M].上海:华东理工大学出版社, 2008.

    Google Scholar

    [3] FANG M Z. Bounds on Eigenvalues of Hadamard Product and the Fan Product of Matrices[J]. Linear Algebra Appl, 2007, 425(1):7-15. doi: 10.1016/j.laa.2007.03.024

    CrossRef Google Scholar

    [4] LIU Q B, CHEN G L. On Two Inequalities for the Hadamard Product and the Fan Product of Matrices[J]. Linear Algebra Appl, 2009, 431(5):974-984.

    Google Scholar

    [5] HUANG R. Some Inequalities for the Hadamard Product and the Fan Product of Matrices[J]. Linear Algebra Appl, 2008, 428(7):1551-1559. doi: 10.1016/j.laa.2007.10.001

    CrossRef Google Scholar

    [6] LIU Q B, CHEN G L, ZHAO L L. Some New Bounds on the Spectral Radius of Matrices[J]. Linear Algebra Appl, 2010, 432(4):936-948. doi: 10.1016/j.laa.2009.10.006

    CrossRef Google Scholar

    [7] 陈付彬, 禹旺勋.非负矩阵Hadamard积谱半径的界[J].江南大学学报(自然科学版), 2014, 13(1):117-120.

    Google Scholar

    [8] LI Y T, LI Y Y, WANG R W, et al. Some New Bounds on Eigenvalues of the Hadamard Product and the Fan Product of Matrices[J]. Linear Algebra Appl, 2010, 432(2):536-545.

    Google Scholar

    [9] CHENG G H. New Bounds for Eigenvalues of the Hadamard Product and the Fan Product of Matrices[J]. Taiwan J Math, 2014, 18(1):305-312. doi: 10.11650/tjm.18.2014.2594

    CrossRef Google Scholar

    [10] ZHAO L L. Two Inequalities for the Hadamard Product of Matrices[J]. J Inequal Appl, 2012, 2012(1):122-127. doi: 10.1186/1029-242X-2012-122

    CrossRef Google Scholar

    [11] GUO Q P, LI H B, SONG M Y. New Inequalities on Eigenvalues of the Hadamard Product and the Fan Product of Matrices[J]. J Inequal Appl, 2013, 2013:433-443. doi: 10.1186/1029-242X-2013-433

    CrossRef Google Scholar

    [12] 孙德淑.非负矩阵Hadamard积的谱半径上界和M-矩阵Fan积的最小特征值下界的新估计[J].西南师范大学学报(自然科学版), 2016, 41(2):7-11.

    Google Scholar

    [13] 黄廷祝, 钟守铭, 李正良.矩阵理论[M].北京:高等教育出版社, 2003.

    Google Scholar

  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Article Metrics

Article views(985) PDF downloads(136) Cited by(0)

Access History

Other Articles By Authors

Some Inequalities for the Upper Bound of the Spectral Radius of Hadamard Product of Nonnegative Matrices

Abstract: The Hadamard product of non negative matrices is an important problem in the matrix analysis theories. Based on the Hölder inequality, taking into consideration the fact that similar matrices have the same eigenvalues, this paper gives the upper bounds of the spectral radius AB for the Hadamard product of the nonnegative matrices of order n A and B. The new bounds only depend on the entries of nonnegative matrices A and B, therefore, they are easy to calculate. Numerical examples are given to show that the new bounds have improved several existing results.

  • 非负矩阵的Hadamard积是特殊的矩阵乘积,在偏微分方程中的弱极小原理、概率论中的特征函数以及控制论等领域有重要的应用[1-2].许多专家和学者对非负矩阵Hadamard积的谱半径上界进行了广泛的研究,并取得了一系列的研究成果.

    Rn×n(Cn×n)为n阶实(复)矩阵的集合.设A=(aij)∈Rn×n,用A≥0表示满足所有元素aij≥0的矩阵A,此时称矩阵A为非负矩阵.用ρ(A)表示非负矩阵A的谱半径.

    A=(aij)∈Rn×nB=(bij)∈Rn×n,称AB=(aijbij)为AB的Hadamard积.关于非负矩阵Hadamard积谱半径的上界估计已经有很多的研究,并有一些经典的结论:

    结论1[1]   设A=(aij)n×n≥0,B=(bij)n×n≥0,则

    结论2[3]   设A=(aij)n×n≥0,B=(bij)n×n≥0,则

    结论3[4]   设A=(aij)n×n≥0,B=(bij)n×n≥0,则

    文献[5-12]也对非负矩阵Hadamard积谱半径的上界估计进行了深入的探讨,本文继续对这个问题进行研究,结合Hölder不等式给出非负矩阵Hadamard积谱半径的一组新上界.

1.   非负矩阵Hadamard积谱半径的上界
  • 引理1[1]   设ABRn×nDE是两个对角矩阵,则有

    引理2[1](Gerschgorin圆盘定理)   设A=(aij)∈Cn×n,则A的所有特征值包含在如下n个圆盘的并之中:

    引理3[13](Hölder不等式)   设

    为非负向量,0<α<1.则

    定理1   设:

    记:

    $ \rho \left( {\mathit{\boldsymbol{A}} \circ \mathit{\boldsymbol{B}}} \right) \le \mathop {\max }\limits_i \left\{ {{a_{ii}}{b_{ii}} + \tilde R_i^\alpha \tilde C_i^{1 - \alpha }} \right\}$.

      当α=0和α=1时,由Gerschgorin圆盘定理易得结论成立.故设0<α<1.令U=diag(u1,…,un),因为u=(u1,…,un)T>0,所以U可逆.记$ \mathit{\boldsymbol{\tilde A = UA}}{\mathit{\boldsymbol{U}}^{ - 1}} $,根据引理1有

    所以$ \rho \left( {\mathit{\boldsymbol{\tilde A}} \circ \mathit{\boldsymbol{B}}} \right) = \rho \left( {\mathit{\boldsymbol{A}} \circ \mathit{\boldsymbol{B}}} \right) $,这里

    $ \rho \left( {\mathit{\boldsymbol{\tilde A}} \circ \mathit{\boldsymbol{B}}} \right) = \lambda $λ对应的特征向量为x=(x1,…,xn)T≠0.

    以下分两种情况进行证明:

    情况1   若

    $ \left( {\mathit{\boldsymbol{\tilde A}} \circ \mathit{\boldsymbol{B}}} \right)\mathit{\boldsymbol{x}} = \lambda \mathit{\boldsymbol{x}} $知,对任意的自然数s,根据Hölder不等式得

    知,(4)式等价于

    所以

    将(5)式两边对s求和,得

    若对满足xs≠0的所有s都有$ {\left( {\frac{{\lambda - {a_{ss}}{b_{ss}}}}{{\tilde R_s^\alpha }}} \right)^{\frac{1}{{1 - \alpha }}}} > {\tilde C_s} $,与(6)式矛盾.因此,至少有满足xs≠0的s,使得

    也即$ \lambda \le {a_{ss}}{b_{ss}} + \tilde R_{_s}^{^\alpha }{\tilde C_s}^{1 - \alpha } $,从而

    情况2   若对某

    则在${\tilde R_s} = 0 $的任一行中添上一个小的正元素使$ \mathit{\boldsymbol{\tilde A}} \circ \mathit{\boldsymbol{B}} $产生摄动,所得新矩阵的特征值包含域大于$ \mathit{\boldsymbol{\tilde A}} \circ \mathit{\boldsymbol{B}} $的特征值包含域,并且在摄动趋于0的极限情形同样可推得结论成立.

    推论1   设A=(aij)n×n≥0,B=(bij)n×n≥0,u=(u1,…,un)T>0,则:

      在定理1中令α=1得(7)式,令α=0得(8)式.

    定理2   设A=(aij)n×n≥0具有非零行和,B=(bij)n×n≥0,0≤α≤1.记:

    $ \rho \left( {\mathit{\boldsymbol{A}} \circ \mathit{\boldsymbol{B}}} \right) \le \mathop {\max }\limits_i \left\{ {{a_{ii}}{b_{ii}} + R_{_i}^{^\alpha }\left( A \right)\tilde C_{_i}^{^{1 - \alpha }}\left( \mathit{\boldsymbol{A}} \right)} \right\} $.

      在定理1中取U=diag(R1(A),…,Rn(A))即得.

    同理,若B=(bij)n×n≥0具有非零行和,取U=diag(R1(B),…,Rn(B)),显然有:

    定理3   设A=(aij)n×n≥0,B=(bij)n×n≥0具有非零行和,0≤α≤1.记:

    $ \rho \left( {\mathit{\boldsymbol{A}} \circ \mathit{\boldsymbol{B}}} \right) \le \mathop {\max }\limits_i \left\{ {{a_{ii}}{b_{ii}} + \tilde R_{_i}^{^\alpha }\left( \mathit{\boldsymbol{B}} \right)\tilde C_{_i}^{^{1 - \alpha }}\left( \mathit{\boldsymbol{B}} \right)} \right\} $.

    定理4   设A=(aij)n×n≥0,B=(bij)n×n≥0,且AB都有非零行和,0≤α≤1,记:

2.   数值算例
  • 例1   估计矩阵AB的谱半径,其中:

    应用文献[1]中(1)式,文献[5]中定理6,文献[3]中(2)式,文献[4]中(3)式,文献[6]中定理3,文献[7]中定理1,文献[8]中定理4.1分别得到:

    $ \alpha = \frac{1}{2} $,利用(9)式得ρ(AB)≤21.732 1.实际上,ρ(AB)=20.743 9.

    例2   估计矩阵AB的谱半径,其中:

    应用文献[3]中定理4,文献[10]中定理2.1,文献[4]中定理4及文献[11]中定理2.1,文献[12]中定理1分别得到:

    $ \alpha = \frac{1}{2}$,应用(9)式得ρ(AB)≤6.449 5.实际上,ρ(AB)=5.733 9.

3.   结束语
  • 本文给出了非负矩阵Hadamard积谱半径的上界估计式.数值例子表明,在一定条件下新估计优于已有的相关结果,而且本文得到的ρ(AB)的界仅仅依赖于矩阵AB的元素,易于计算.

Reference (13)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return