矩阵论（三）

矩阵分解

矩阵的三角分解

$$\begin{array}{rl}A\in C_n^{n\times n},{\Delta }_k\ne 0& ⟺A=LR\\ & ⟺有唯一解\{\begin{array}{l}\text{Doolittle , L is unit lower triangle}\\ \text{Crout , R is unit upper triangle }\\ \text{LDR }\end{array}\end{array}$$

$$A\in C_r^{n\times n},{\Delta }_k\ne 0\Rightarrow A=LR$$

$$A\in C^{n\times n},A是Hermite正定矩阵\Rightarrow A=G{G}^H\text{\hspace{1em}(Cholesky)}$$

QR分解

Houseloder Matrix

$$H=I-2u{u}^H,u^{H}u=1\text{\hspace{1em}(Householder Matrix)}$$
$$y=Hx,\text{H is Householder Matrix}\text{\hspace{1em}(Householder change)}$$
$$\begin{array}{rl}& H^H=H,\text{ Hermite Matrix}\\ & H^{H}H=I,\text{ Uintary Matrix}\\ & H^2=I\\ & H^{-1}=H\\ & \left[\begin{array}{cc}{I}_r& O\\ O& H\end{array}\right]\text{ is n+r Householder Matrix}\\ & detH=-1\end{array}$$

$$\forall x\in C^n,\exists H\text{ is Householder Matrix},Hx=\alpha z,\alpha =\parallel x{\parallel }_2,z^{H}z=1$$
$$H=I-2u{u}^H,u=\frac{x-\alpha z}{\parallel x-\alpha z{\parallel }_2}$$
$\text{In particular}$
$$\forall x\in C^n,\exists H\text{ is Householder Matrix},Hx=\alpha e_1,\alpha =\parallel x{\parallel }_2$$
$$\forall x\in R^n,\exists H\text{ is Householder Matrix},Hx=\alpha e_1,\alpha =\pm \parallel x{\parallel }_2$$

Givens Matrix

$$T_{pq}=\left[\begin{array}{cccccccccccc}1& & & & & & & & & & & \\ & \ddots & & & & & & & & & & \\ & & 1& & & & & & & & & \\ & & & \overline{c}& & & & \overline{s}& & & & \\ & & & & 1& & & & & & & \\ & & & & & \ddots & & & & & & \\ & & & & & & 1& & & & & \\ & & & -s& & & & c& & & & \\ & & & & & & & & 1& & & \\ & & & & & & & & & \ddots & & \\ & & & & & & & & & & 1& \end{array}\right]$$
$$y=T_{pq}x\text{\hspace{1em}(Givens change)}$$
$$\begin{array}{r}{T}_{pq}^H{T}_{pq}=I\\ det{T}_{pq}=1\end{array}$$
$$\forall x=({\xi }_1,{\xi }_2,\cdots ,{\xi }_n{)}^T,\exists T_{pq},y=T_{pq}x,y_{[p]}>0,y_{[q]}=0$$
$$c=\frac{{\xi }_p}{\sqrt{\mid {\xi }_p{\mid }^2+\mid {\xi }_q{\mid }^2}},s=\frac{{\xi }_q}{\sqrt{\mid {\xi }_p{\mid }^2+\mid {\xi }_q{\mid }^2}}$$
$$\forall x\in C^n,\exists T_{12},T_{13},\cdots ,T_{1n},T_{1n}\cdots T_{13}{T}_{12}x=\parallel x{\parallel }_2{e}_1$$

矩阵的QR分解

$$\forall A\in C^{n\times n},A=QR,Q^{H}Q=I,\text{R is upper triangle matrix}$$

• $\text{Householder trans}$
  $$\begin{array}{rl}& A=(a_1,a_2,\cdots ,a_n)\\ & H_1{a}_1=\alpha e_1\\ & H_{1}A=\left(\begin{array}{cccc}{\alpha }_1& \ast & \cdots & \ast \\ 0& & & \\ ⋮& & {\mathbf{B}}_{n-1}& \\ 0& & & \end{array}\right)\\ & {\tilde{H}}_2{b}_2={\alpha }_2{\tilde{e}}_1\\ & H_2=\left[\begin{array}{cc}1& 0^T\\ 0& {\tilde{H}}_2\end{array}\right]\\ & H_2(H_{1}A)=\cdots \\ & \cdots \\ & H_{n-1}\cdots H_2{H}_{1}A=R\\ & A=H_1{H}_2\cdots H_{n-1}R=QR\end{array}$$
• $\text{Givens trans}$
  $$\begin{array}{rl}& A=(a_1,a_2,\cdots ,a_n)\\ & T_{1n}\cdots T_{12}{a}_1=\parallel a_1{\parallel }_2{e}_1\\ & \cdots \\ & T_{n-1,n}\cdots T_{2n}\cdots T_{23}{T}_{1n}\cdots T_{12}A=R\\ & A=T_{12}^H\cdots T_{1n}^H{T}_{23}^H\cdots T_{2n}^H\cdots T_{n-1,n}^{H}R=QR\end{array}$$
• $\text{Schmidt trans}$
  $A\in C_n^{n\times n}$，$A$可唯一地分解为
  $$A=QR$$
  $$\begin{array}{rl}A=& (a_1,a_2,\cdots ,a_n)\\ \Rightarrow & (p_1,p_2,\cdots ,p_n)\text{ Schmidt}\\ \Rightarrow & (q_1,q_2,\cdots ,q_n)\text{ Unit}\\ A=& (q_1,q_2,\cdots ,q_n)\left[\begin{array}{cccc}\parallel p_1{\parallel }_2& {\lambda }_{21}\parallel p_1{\parallel }_2& \cdots & {\lambda }_{n1}\parallel p_1{\parallel }_2\\ & \parallel p_2{\parallel }_2& \cdots & {\lambda }_{n2}\parallel p_2{\parallel }_2\\ & & \ddots & ⋮\\ & & & \parallel p_n{\parallel }_2\end{array}\right]=QR\\ & {\lambda }_{ij}=\frac{(a_i,p_j)}{(p_j,p_j)}\end{array}$$

矩阵酉相似于Hessenberg矩阵

$$A=(a_{ij}{)}_{n\times n}\in C^{n\times n},a_{ij}=0(i>j+1)\text{\hspace{1em}(Hessenberg matrix)}$$
使用$\text{Householder,Givens}$变换，将矩阵逐步化简成$\text{Hessenberg}$型，例如$H_{1}A{H}_1=A_1,\cdots ,\text{ or }{T}_{12}A{T}_{12}^T=A_1$
最后得到
$$Q^{H}AQ=H$$

• $A\in C^{n\times n},\exists Q,Q^{H}AQ\text{ is Hessenberg matrix}$
• $A\in R^{n\times n},\exists Q,Q{Q}^T=I,Q^{T}AQ\text{ is Hessenberg matrix}$
• $A\in C^{n\times n},A^H=A,\exists Q,Q^{H}AQ\text{ is tridoginal }$

矩阵的满秩分解

Hermite标准形

如果$A$经过有限次初等变换变为矩阵$B$，则$A$与$B$等价，其充要条件是
$$\begin{array}{rl}& rankA=rankB\\ & \exists S\in C_m^{m\times m},T\in C_n^{n\times n},SAT=B\end{array}$$
设$A\in C_r^{m\times n}$，则$A$可通过初等行变换化为如下条件的矩阵$H$，即$\exists S\in C_m^{m\times m},SA=H$
$$\mathbf{H}=\left(\begin{array}{cccccccccccccc}0& \cdots & 0& 1& \ast & \cdots & \ast & 0& \ast & \cdots & 0& \ast & \cdots & \ast \\ 0& \cdots & 0& 0& 0& \cdots & 0& 1& \ast & \cdots & ⋮& ⋮& & ⋮\\ ⋮& & ⋮& ⋮& ⋮& & ⋮& ⋮& ⋮& & 0& \ast & \cdots & \ast \\ 0& \cdots & 0& 0& 0& \cdots & 0& 0& 0& \cdots & 1& \ast & \cdots & \ast \\ 0& \cdots & 0& 0& 0& \cdots & 0& 0& 0& \cdots & 0& 0& \cdots & 0\\ ⋮& & ⋮& ⋮& ⋮& & ⋮& ⋮& ⋮& & ⋮& ⋮& & ⋮\\ 0& \cdots & 0& 0& 0& \cdots & 0& 0& 0& \cdots & 0& 0& \cdots & 0\end{array}\right\}$$

• $H$的前$r$行中每一行至少含一个非零元素，且第一个非零元素是$1$， 其所在列元素均为$0$（除去自身）

• 若$H$中第$i$行的第一个非零元素$1$位于$j_i$列，则
  $$j_1<j_2<\cdots j_r$$

• $H$的第$j_1,j_2,\cdots ,j_r$为$I_m$的前$r$列

这样的矩阵为$\text{Hermite}$标准形
可通过
$$S(A,I_m)=(SA,S)=(H,S)$$
的方法求出
以$n$阶单位矩阵$I_n$的$n$个列向量$e_1,e_2,\cdots ,e_n$为列构成的$n$阶方阵
$$P=(e_{i_1},e_{i_2},\cdots ,e_{i_n})$$
称为$n$阶置换矩阵，$i_1,i_2,\cdots ,i_n$是$n$的一个全排列

• $P$是正交矩阵
• $AP$是将$A$的列按照全排列顺序（$Hermiter$标准形每一行第一个非零元素列的顺序）重新排列的矩阵
  设$A\in C_r^{m\times n},\exists S\in S_m^{m\times m}$和$n$阶置换矩阵$P$，使得
  $$SAP=\left[\begin{array}{cc}{I}_r& K\\ O& O\end{array}\right]$$
  如果继续进行初等列变换，可得到
  $$SAT=\left[\begin{array}{cc}{I}_r& O\\ O& O\end{array}\right]$$
  求$T$的方法类似于$S$的求法

矩阵的满秩分解

$$\begin{array}{r}A\in C_r^{m\times n},\exists F\in C_r^{m\times r},G\in C_r^{r\times n},\\ A=FG\\ SAT=\left[\begin{array}{cc}{I}_r& O\\ O& O\end{array}\right]\\ F=S^{-1}[前r行]，G=T^{-1}[前r列]\end{array}$$
简便求$FG$的方法如下
$$\begin{array}{r}SA=H\\ F=A[前j_1,j_2,\cdots ,j_r列]\\ G=H[前r行]\end{array}$$

矩阵的奇异值分解

设$A,B\in C^{m\times n}$,若存在$m$阶酉矩阵$U$，$n$阶酉矩阵$V$，使得$U^{H}AV=B$，则称$A$与$B$酉等价。
设$A\in C_r^{m\times n}$,$A^{H}A$的特征值为
$${\lambda }_1\ge {\lambda }_2\ge \cdots \ge {\lambda }_r>{\lambda }_{r+1}=\cdots =0$$
则称${\sigma }_i=\sqrt{{\lambda }_i}$为$A$的奇异值
酉等价矩阵有相同的奇异值
设$A\in C_r^{m\times n}$，则存在$m$阶酉矩阵$U$和$n$阶酉矩阵$V$，使
$$U^{H}AV=\left[\begin{array}{cc}\Sigma & O\\ O& O\end{array}\right],\Sigma =diag({\sigma }_1,{\sigma }_2,\cdots ,{\sigma }_r)$$
$$A=U\left[\begin{array}{cc}\Sigma & O\\ O& O\end{array}\right]V^H$$
称为$A$的奇异值分解
$$\left[\begin{array}{c}{V}^H{A}^{H}AV=diag({\lambda }_1,{\lambda }_2,\cdots ,{\lambda }_n)\\ V=(V_1,V_2)(V_1\in C^{n\times r})\\ U_1=A{V}_1{\Sigma }^{-1}\\ U=(U_1,U_2),U^{H}U=I\end{array}\right]$$
设$A\in C_r^{m\times n},b\in C^m$,对$A$进行奇异分解，
$$x^{(0)}=V\left[\begin{array}{cc}{\Sigma }^{-1}& O\\ O& O\end{array}\right]U^{H}b$$
是矛盾方程组$Ax=b$的最小二乘解，如果最小二乘解不唯一，则$x^{(0)}$是其中具有最小$2$范数的向量，称其为极小范数最小二乘解。

特征值的估计与表示

特征值的包含区域

Gerschgorin定理

设$A=(a_{ij}{)}_{n\times n}$，记
$$R_i=\sum _{j=1,j\ne i}^n\mid a_{ij}\mid$$
称复平面上的圆域
G i = { z ∣ ∣ z − a i i ∣ ≤ R i , z ∈ C } ( i = 1 , 2 , ⋯   , n ) G_i = \{z \mid \mid z - a_{ii} \mid \leq R_i,z\in C \} (i=1,2,\cdots,n) Gi​={z∣∣z−aii​∣≤Ri​,z∈C}(i=1,2,⋯,n)
为矩阵$A$的第$i$个$\text{Gerschgorin}$圆，称$R_i$为盖尔圆$G_i$的半径
$${\lambda }_1,{\lambda }_2,\cdots ,{\lambda }_n\in \bigcup _{i=1}^n{G}_i\text{\hspace{1em}(Gerschgorin1)}$$
$A$的特征值与$A^T$的特征值相同，$A^T$的盖尔圆叫做$A$的列盖尔圆

• $\text{Gerschgorin2}$：若矩阵$A$的某一连通部分由$A$的$k$个盖尔圆构成，则其中有且仅有$A$的$k$个特征值

• 若$A$按行（列）严格对角占优，则$detA\ne 0$

• 实矩阵的复特征值成对共轭出现，如果其盖尔圆各自独立且均关于实轴对称，则特征值均为实数

特征值的隔离

选取正数，令$d_i$
$$D=diag(d_1,d_2,\cdots ,d_n)$$
$$B=DA{D}^{-1}=(a_{ij}\frac{d_i}{d_j}{)}_{n\times n}$$
$A$与$B$有相同的特征值，通过设某一个$d_i>1\text{ or }{d}_i<1$,其他$d$设为$1$，可放大或缩小$G_i$，同时缩小放大其它盖尔圆。

$D$中不为$1$的位置为$d_i$，$DA{D}^{-1}=B$，$B$等于对$A$的第$i$行元素乘以$d_i$，第$i$列元素乘以$\frac{1}{d_i}$，不对$a_{ii}$操作。

Rayleign商

$A$是$n$阶实对称矩阵，$x\in R^n$,
$$R(x)=\frac{x^{T}Ax}{x^{T}x}=\frac{(Ax,x)}{(x,x)}$$
为$A$的$\text{Rayleign}$商

• $R(\lambda x)=R(x)$
• ${\lambda }_1={\max }_{x\ne 0}R(x)$
• ${\lambda }_n={\min }_{x\ne 0}R(x)$
• $R(x)=\frac{x^{T}Ax}{x^{T}x}=y^{T}Ay,\parallel y{\parallel }_2=1$