根据弱链对角占优M-矩阵A的逆矩阵元素,定义新的参数,结合不等式放缩技巧,给出
根据弱链对角占优M-矩阵A的逆矩阵元素,定义新的参数,结合不等式放缩技巧,给出 ‖ A − 1 ‖ ∞ 的新上界估计式。理论分析和数值例子说明,新估计式改进了现有文献的有关结果。
弱链对角占优M-矩阵,逆矩阵,无穷范数的上界
Huijun Li, Hongmin Mo*
College of Mathematics and Statistics, Jishou University, Jishou Hunan
Received: Jun. 15th, 2022; accepted: Jul. 12th, 2022; published: Jul. 19th, 2022
According to the inverse matrix elements of weakly chain diagonally dominant M-matrix A, the new parameters are defined and the inequality scaling technique is used to obtain ‖ A − 1 ‖ ∞ . Theoretical analysis and numerical examples show that the new estimate improves the relevant results in the existing literature.
Keywords:Weak Chain Diagonally Dominant Matrices, Inverse Matrices, Upper Bound on an Infinite Norm
Copyright © 2022 by author(s) and beplay安卓登录
This work is licensed under the Creative Commons Attribution International License (CC BY 4.0).
http://creativecommons.org/licenses/by/4.0/
弱链对角占优矩阵在现代经济学、网络、信息论和算法设计等领域都有着广泛的应用。从1974年开始,众多学者对弱链对角占优M-矩阵的逆矩阵的无穷范数的上界估计进行了广泛研究,得到了一些不同的估计结果并将其进行了应用(见文献 [
C n × n ( R n × n ) 表示n阶复(实)矩阵集,设 A = ( a i j ) ∈ C n × n , N = { 1 , 2 , ⋯ , n } ,为方便叙述给出下列符号:
R i = ∑ k ≠ i | a i k | ,
d i = R i | a i i | ,
r i j = | a i j | | a i i | − ∑ k ≠ i , j | a i k | j ≠ i ,
s i j = | a i j | + ∑ k ≠ i , j | a i k | r k j | a i i | j ≠ i ,
v i j = | a i j | + ∑ k ≠ i , j | a i k | s k j | a i i | j ≠ i ,
v i = max j ≠ i { v j i } ,
m k i = { | a k i | | a k k | − ∑ j = i + 1 , j ≠ k n | a k j | s j i | a k i | ≠ 0 , 0 | a k i | = 0.
m i = { max i + 1 ≤ k ≤ n { m k i } 1 ≤ i ≤ n − 1 , 0 i = n .
定义1 [
定义2 [
定义3 [
引理1 [
引理2 [
α 11 = 1 Δ ,
α i 1 = 1 Δ ∑ k = 2 n β i k ( − a k 1 ) ,
α 1 j = 1 Δ ∑ k = 2 n β k j ( − a 1 k ) ,
α i j = β i j + α 1 j ∑ k = 2 n β i k ( − a k 1 ) ,
Δ = a 11 − ∑ k = 2 n a 1 k ( ∑ k = 2 n β k i a i 1 ) > 0.
引理3 [
| α i j | ≤ d i | α j j | ≤ | α j j | .
引理4 [
g i = | a i j i j | − ∑ l = j + 1 n | a i j i l | > 0.
引理5 若 A = ( a i j ) ∈ R n × n 为弱链对角占优M-矩阵, n ≥ 2 , A − 1 = ( α i j ) ,满足 u 1 < 1 ,则
α 11 ≤ 1 a 11 ( 1 − m 1 d 1 ) .
证明
因A为M-矩阵,则 A − 1 ≥ 0 ,设 A m = x ,其中 m = ( 1 , m 1 , ⋯ , m 1 ) T , x = ( x 1 , x 2 , ⋯ , x n ) T ,因为 0 ≤ m 1 ≤ 1 , u 1 = ∑ k = 2 n | a 1 k | / | a 11 | < 1 ,则
x 1 = a 11 + ∑ k = 2 n a 1 k m 1 = | a 11 | − ∑ k = 2 n | a 1 k | m 1 ≥ | a 11 | − ∑ k = 2 n | a 1 k | > 0 ,
对 ∀ i ∈ N , 2 ≤ i ≤ n ,有
∑ k = 2 n a i k = | a i i | − ∑ k ∈ N , k ≠ 1 , i | a i k | ≥ | a i i | − ∑ k ∈ N , k ≠ i | a i k | ≥ 0.
当 a i 1 = 0 时, x i = ∑ k = 2 n a i k m 1 ≥ 0 ;当 a i 1 ≠ 0 时, x i = a i 1 + ∑ k = 2 n a i k m 1 ≥ a i 1 + ( ∑ k = 2 n a i k ) m i 1 = 0 ,即
x 1 > 0 , x i ≥ 0 ( i = 2 , ⋯ , n ) .
再由 A − 1 x = m , A − 1 ≥ 0 ,有
α 11 ≤ 1 x 1 = 1 a 11 + m 1 ∑ k = 2 n a 1 k = 1 a 11 ( 1 − m 1 d 1 ) .
引理6 设 A = ( a i j ) ∈ R n × n 为弱链对角占优M-矩阵,且 A − 1 = ( α i j ) ,对 ∀ i , j ∈ N , i ≠ j ,有
| α i j | ≤ | a i j | + ∑ k ≠ i , j | a i k | s k j | a i i | | α j j | ≤ v j | α j j | ≤ | α j j | .
证明
根据引理3,设
s j i ( ε , 1 ) = { | a i j | + ∑ k ≠ i , j | a i k | r k j + ε | a i i | i ∈ J ( A ) , 1 i ∈ N , i ∉ J ( A ) .
其中 ε > 0 足够小,使得 0 < s j i ( ε , 1 ) ≤ 1 ( i , j ∈ N , i ≠ j ) 。
设
S j ( ε , 1 ) = d i a g ( s j 1 ( ε , 1 ) , ⋯ , s j j − 1 ( ε , 1 ) , 1 , s j j + 1 ( ε , 1 ) , ⋯ , s j n ( ε , 1 ) ) , j ∈ N .
显然,当 S j ( ε , 1 ) = d i a g ( 1 , 1 , ⋯ , 1 ) = I 时, A S j ( ε , 1 ) 是弱链对角占优矩阵,且当 S j ( ε , 1 ) ≠ I 时, A S j ( ε , 1 ) 是严格对角占优矩阵,所以, A S j ( ε , 1 ) 一定是一个弱链对角占优矩阵,从而
| α i j | s j i ( ε , 1 ) ≤ | a i j | + ∑ k ≠ i , j | a i k | s k j ( ε , 1 ) | a i i | s j i ( ε , 1 ) | α j j | , i ≠ j
也就是
| α i j | ≤ | a i j | + ∑ k ≠ i , j | a i k | s k j ( ε , 1 ) | a i i | | α j j | , i ≠ j
当 ε → 0 时,得
| α i j | ≤ | a i j | + ∑ k ≠ i , j | a i k | s k j | a i i | | α j j | ≤ v j | α j j | ≤ | α j j | , i ≠ j .
1974年,P N Shivakumar在文献 [
‖ A − 1 ‖ ∞ ≤ ∑ i = 1 n [ a i i ∏ j = 1 i − 1 ( 1 − u j ) ] − 1 . (1)
2012年,潘淑珍在文献 [
‖ A − 1 ‖ ∞ ≤ 1 | a 11 | ( 1 − u 1 t 1 ) + ∑ i = 2 n [ 1 | a i i | ( 1 − u i t i ) ∏ j = 1 i − 1 ( 1 + u j 1 − u j t j ) ] . (2)
其中
t i = { max i + 1 ≤ k ≤ n { t k i } 1 ≤ i ≤ n − 1 , 0 i = n . 0 ≤ t i ≤ 1 , ∀ i ∈ N
t k i = { | a k i | | a k k | − ∑ j = i + 1 , j ≠ k n | a k j | | a k i | ≠ 0 , 0 | a k i | = 0. ∀ k ∈ N , i + 1 ≤ k ≤ n , 1 ≤ i ≤ ( n − 1 )
本文继续给出弱链对角占优M-矩阵A的 ‖ A − 1 ‖ ∞ 的新上界估计式。
设 A = ( a i j ) ∈ R n × n 为弱链对角占优M-矩阵, n ≥ 2 , A − 1 = ( α i j ) , B = A ( 2 , n ) , B − 1 = ( β i j ) ,满足 u 1 < 1 ,则
‖ A − 1 ‖ ∞ ≤ max { 1 a 11 ( 1 − m 1 d 1 ) + d 1 ‖ B − 1 ‖ ∞ 1 − m 1 d 1 , v 1 a 11 ( 1 − m 1 d 1 ) + ( v 1 d 1 1 − m 1 d 1 + 1 ) ‖ B − 1 ‖ ∞ } .
证明
记 r i = ∑ j = 1 n α i j , M A = ‖ A − 1 ‖ ∞ , M B = ‖ B − 1 ‖ ∞ ,即 M A = max i ∈ N { r i } , M B = max 2 ≤ i ≤ n { ∑ j = 2 n β i j } ,由引理2和引理5可得
r 1 = α 11 + ∑ j = 2 n α 1 j = 1 Δ + 1 Δ ∑ k = 2 n ( − a 1 k ) ∑ j = 2 n β k j ≤ 1 Δ + 1 Δ ∑ k = 2 n ( − a 1 k ) M B = 1 Δ + 1 Δ a 11 d 1 M B ≤ 1 a 11 ( 1 − m 1 d 1 ) + d 1 1 − m 1 d 1 M B , (3)
对 2 ≤ i ≤ n ,由引理3知,
α i 1 ≤ α 11 ,
对 2 ≤ j ≤ n ,由引理6知,
α i j = β i j + α 1 j ∑ k = 2 n β i k ( − a k 1 ) ≤ β i j + α 1 j v 1 ,
则
r i = α i 1 + ∑ j = 2 n α i j ≤ v 1 α 11 + ∑ j = 2 n ( β i j + α 1 j v 1 ) = r 1 v 1 + M B ≤ v 1 ( 1 a 11 ( 1 − m 1 d 1 ) + d 1 M B 1 − m 1 d 1 ) + M B = v 1 a 11 ( 1 − m 1 d 1 ) + ( v 1 d 1 1 − m 1 d 1 + 1 ) M B , (4)
由(3)式和(4)式可得
M A = max { r 1 , r i : 2 ≤ i ≤ n } ≤ max { 1 a 11 ( 1 − m 1 d 1 ) + d 1 ‖ B − 1 ‖ ∞ 1 − m 1 d 1 , v 1 a 11 ( 1 − m 1 d 1 ) + ( v 1 d 1 1 − m 1 d 1 + 1 ) ‖ B − 1 ‖ ∞ } .
设 A = ( a i j ) ∈ R n × n 为弱链对角占优M-矩阵,且对 ∀ k ∈ N ,有 u k < 1 ,则
M A ≤ max { 1 a 11 ( 1 − m 1 u 1 ) + ∑ i = 2 n [ 1 a i i ( 1 − m i u i ) ∏ j = 1 i − 1 ( u j 1 − m j u j ) ] , v 1 a 11 ( 1 − m 1 u 1 ) + ∑ i = 2 n [ v i a i i ( 1 − m i u i ) ∏ j = 1 i − 1 ( v j u j 1 − m j u j + 1 ) ] } . (5)
证明
设A为弱链对角占优M-矩阵,对 ∀ k ∈ N , 1 ≤ k ≤ n − 1 , A ( k , n ) 为弱链对角占优M-矩阵,由定理1关于k对 A ( k , n ) 做数学归纳法可得。
注
设 A = ( a i j ) ∈ R n × n 为弱链对角占优M-矩阵,且对 ∀ k ∈ N ,满足 u k < 1 ,由 v i , m i , t i 表达式可知 0 ≤ v i ≤ 1 , 0 ≤ m i ≤ t i ≤ 1 ,那么(5)式中 ‖ A − 1 ‖ ∞ 的上界小于等于(2)式中的上界,即
max { 1 a 11 ( 1 − m 1 u 1 ) + ∑ i = 2 n [ 1 a i i ( 1 − m i u i ) ∏ j = 1 i − 1 ( u j 1 − m j u j ) ] , v 1 a 11 ( 1 − m 1 u 1 ) + ∑ i = 2 n [ v i a i i ( 1 − m i u i ) ∏ j = 1 i − 1 ( v j u j 1 − m j u j + 1 ) ] } ≤ 1 | a 11 | ( 1 − u 1 t 1 ) + ∑ i = 2 n [ 1 | a i i | ( 1 − u i t i ) ∏ j = 1 i − 1 ( 1 + u j 1 − u j t j ) ] .
例1设
A = ( 4 − 1 − 1 − 2 5 − 3 − 1 − 2 4 ) .
J = { 1 , 3 } ,易得到A为弱链对角占优M-矩阵, ‖ A − 1 ‖ ∞ = 1.100 。
应用(1)式得
‖ A − 1 ‖ ∞ ≤ 2.7500 ;
应用(2)式得
‖ A − 1 ‖ ∞ ≤ 1.500 ;
应用定理2得
‖ A − 1 ‖ ∞ ≤ 1.1503.
例2 设
A = ( 4 − 2 − 1 − 1 3 − 2 − 1 0 2 ) .
易得到A为弱链对角占优M-矩阵, ‖ A − 1 ‖ ∞ = 1.54 。
应用(1)式得
‖ A − 1 ‖ ∞ ≤ 11 ;
应用(2)式得
‖ A − 1 ‖ ∞ ≤ 5.6667 ;
应用定理2得
‖ A − 1 ‖ ∞ ≤ 1.996.
理论证明本文所得弱链对角占优M-矩阵逆矩阵无穷范数的新上界估计式优于文献 [
感谢莫宏敏老师对本篇论文的悉心指导和帮助。
吉首大学研究生科研项目(JDY21012)。
李慧君,莫宏敏. 弱链对角占优M-矩阵逆的无穷范数新上界估计New Upper Bound Estimates of the Infinite Norm for the Inverse of Weakly Chain Diagonally Dominant M-Matrices[J]. 应用数学进展, 2022, 11(07): 4683-4689. https://doi.org/10.12677/AAM.2022.117494
https://doi.org/10.1137/S0895479894276370
https://doi.org/10.1016/j.laa.2009.09.012
https://doi.org/10.1080/03081087.2019.1649995
https://doi.org/10.1016/j.amc.2019.124788