摘要
设A是一个n阶的任意复矩阵且E是A的Hermite秩1扰动,即E=xx',其中x是n维的复列向量,x'是x的共轭转置向量.则A+E为矩阵A的Hermite秩1修正矩阵.基于矩阵分析理论中Hermite矩阵特征值分布的性质,研究得到了矩阵A特征值的任意Hermite秩1修正扰动的上下界限,即给出了矩阵A+E特征值的上下界限:λ_i(H(A))+l_i(x)+δ_i≤R(λ_i(A+xx'))≤λ_i(H(A))+u_i(x)+δ'_i(i=1,n),λ_i(H(A))+l_i(x)+δ_i≤R(λ_i(A+xx'))≤min{λ_i(H(A))+u_i(x),λ_(i-1)(H(A))}+δ'_i(2≤i≤n-1),且λ_(min)(-SH(A)τ)≤S(λ_i(A+xx'))≤λ_(max)(-SH(A)τ)(1≤i≤n),其中δ_i=sgn(‖SH(A)‖_2)[λ_(min)(H(A))-λ_(i-1)(H(A))-u_i(x)],δ'_i=sgn(‖SH(A)‖_2)[λ_(max)(H(A))-λ_i(H(A))-l_i(x)+‖x‖_2~2],gap_i=λ_(i-1)(A)-λ_i(A),i=2,…,n,H(A)和SH(A)分别代表矩阵A的Hermite部分和反Hermite部分,τ=(-1)^(1/2),sgn(·)代表符号函数.当A为Hermite矩阵时,上述结果退化为已有的结果λ_i(A)-‖x‖_2~2≤R(λ_i(A+xx'))≤λ_i(A)+‖x‖_2~2.
Assume that matrix A is an arbitrary complex matrix of order n and E is a Hermitian rank-one matrix,i. e.,E = xx',where x is a complex column vector of order n and x' is the conjugate transpose vector of x. So,A + E is called Hermitian rank-one update of matrix A. Based on the properties of matrix analysis theory in Hermitian matrix eigenvalue distribution,Hermitian rank-one perturbation bounds of an arbitrary matrix is given and two-side bounds for eigenvalues of A + E are presented as follows:
λi(H(A))+li(x)+δi≤R(λi(A+xx'))≤λi(H(A))+ui(x)+δ'i(i=1,n),λi(H(A))+li(x)+δi≤R(λi(A+xx'))≤min{λi(H(A))+ui(x),λ(i-1)(H(A))}+δ'i(2≤i≤n-1),且λ(min)(-SH(A)τ)≤S(λi(A+xx'))≤λ(max)(-SH(A)τ)(1≤i≤n),其中δi=sgn(‖SH(A)‖2)[λ(min)(H(A))-λ(i-1)(H(A))-ui(x)],δ'i=sgn(‖SH(A)‖2)[λ(max)(H(A))-λi(H(A))-li(x)+‖x‖2^2],gapi=λ(i-1)(A)-λi(A),i=2,…,n
H(A) and SH(A) stand for Hermitian part and anti-Hermitian part of matrix A,τ and sgn( · ) denotes sign function. When A is Hermitian, the above results reduce to λi(A) -||x||2^2≤R(λi(A +xx')≤ λi(A) + ||x||2^2.
出处
《华南师范大学学报(自然科学版)》
CAS
北大核心
2015年第2期158-160,共3页
Journal of South China Normal University(Natural Science Edition)
基金
江苏省自然科学青年基金项目(BK2013098)
江苏省高校自然科学基金项目(13KJB110020)
关键词
特征值
上下界
秩1修正
eigenvalue
two-side bounds
rank-one update