天天看點

矩陣分析筆記(二)極大線性無關組

兩個向量組之間的線性表示關系

設V是\mathbb{F}上的線性空間,\alpha_1,\alpha_2,...,\alpha_p和\beta_1,\beta_2,...,\beta_q是V中的兩個向量組,\beta\in V

  1. 如果存在p個數k_i\in \mathbb{F},i=1,2,...,p,使得\alpha_1k_1+\alpha_2k_2+···+\alpha_pk_p=\beta,稱向量\beta可由向量組\alpha_1,\alpha_2,...,\alpha_p線性表示
  2. 如果每個\beta_j都可以由向量組\alpha_1,\alpha_2,...,\alpha_p線性表示,j=1,2,...,q。為了友善,\beta_1,\beta_2,...,\beta_q可由\alpha_1,\alpha_2,...,\alpha_p線性表示,用符号記為\{\beta_1,\beta_2,...,\beta_q\}≤_{lin}\{\alpha_1,\alpha_2,...,\alpha_p\}

線性表示關系的傳遞性

設V是\mathbb{F}上的線性空間,\alpha_1,\alpha_2,...,\alpha_p;\beta_1,\beta_2,...,\beta_q;\gamma_1,\gamma_2,...,\gamma_t是V中三個向量組。若

$$ \{\alpha_1,\alpha_2,...,\alpha_p\}≤_{lin}\{\beta_1,\beta_2,...,\beta_q\}\\ \{\beta_1,\beta_2,...,\beta_q\}≤_{lin}\{\gamma_1,\gamma_2,...,\gamma_t\} $$

則\{\alpha_1,\alpha_2,...,\alpha_p\}≤_{lin}\{\gamma_1,\gamma_2,...,\gamma_t\}

證明:我們利用線性表示關系的矩陣表達。由存在T\in \mathbb{F}^{q\times p},S\in \mathbb{F}^{t\times q}滿足

$$ [\alpha_1,\alpha_2,...,\alpha_p]=[\beta_1,\beta_2,...,\beta_q]T\\ [\beta_1, \beta_2,...,\beta_q]=[\gamma_1,\gamma_2,...,\gamma_t]S $$

得[\alpha_1,\alpha_2,...,\alpha_p]=[\gamma_1,\gamma_2,...,\gamma_t](ST)。由于ST\in \mathbb{F}^{t\times p}得證

扁(列>行)的齊次線性方程組必有非零解

設A\in \mathbb{F}^{m\times n},1≤m<n,則齊次線性方程組Ax=0必有非零解

x=\begin{bmatrix} c_1 \\ c_2 \\ \vdots \\ c_n\end{bmatrix}\in \mathbb{F}^n,\begin{bmatrix} c_1 \\ c_2 \\ \vdots \\ c_n\end{bmatrix}\neq0

這裡,m<n

證明:對m用數學歸納法

  1. m=1時,n≥2。根據a_{11}是否為0分兩種情況若a_{11}=0,則取\begin{bmatrix} c_1 \\ c_2 \\ \vdots \\ c_n\end{bmatrix}=\begin{bmatrix}1\\ 0\\ \vdots \\0\end{bmatrix},易知其為一非零解若a_{11}\neq0,由n≥2,可取\begin{bmatrix} c_1 \\ c_2 \\ \vdots \\ c_n\end{bmatrix}=\begin{bmatrix}-\frac{a_{12}}{a_{11}}\\ 0\\ \vdots \\0\end{bmatrix},易知這樣取得c為一非零解
  2. 假設m≤p時命題成立,下面證明m=p+1的情形,此時n≥p+2,同樣根據x_1的系數,也就是系數矩陣A的第一列\alpha_1=\begin{bmatrix}a_{11}\\a_{21}\\ \vdots \\a_{p+1,1}\end{bmatrix}是否為0分兩種情況若\alpha_1=0,則取c=\begin{bmatrix}c_1\\ c_2 \\ \vdots \\ c_n\end{bmatrix}=\begin{bmatrix}1\\ 0\\ \vdots \\ 0\end{bmatrix}若\alpha_1\neq0,不妨設\alpha_{11}\neq0,則有與Ax=0同解的線性方程組\begin{bmatrix}a_{11} &a_{12}&\cdots &a_{1n} \\ 0 &0 & \cdots &0 \\ \vdots & \vdots & \vdots & \vdots \\ 0 & a_{p+1,2}-\frac{a_{p+1,1}}{a_{11}}·a_{12} & \cdots & a_{p+1, n}-\frac{a_{p+1,1}}{a_{11}}·a_{1n}\end{bmatrix}\begin{bmatrix}x_1\\ x_2 \\ \vdots \\ x_n\end{bmatrix}=0p個方程,n-1(≥p+1>p)\begin{bmatrix}a_{22}-\frac{a_{21}}{a_{11}}·a_{12} & \cdots & a_{2n}-\frac{a_{21}}{a_{11}}·a_{1n} \\ \vdots & \vdots & \vdots \\ a_{p+1,2}-\frac{a_{p+1,1}}{a_{11}}·a_{12} & \cdots & a_{p+1,n}-\frac{a_{p+1,1}}{a_{11}}·a_{1n}\end{bmatrix}\begin{bmatrix}x_2 \\ \vdots \\ x_n\end{bmatrix}=0\begin{bmatrix}x_2\\ \vdots \\ x_n\end{bmatrix}=\begin{bmatrix}c_2\\ \vdots \\ c_n\end{bmatrix}\neq 0,再取x_1=c_1=-\frac{1}{a_{11}}(a_{12}c_2+···+a_{1n}c_n),易知\begin{bmatrix}x_1 \\ x_2\\ \vdots \\ x_n\end{bmatrix}=\begin{bmatrix}c_1 \\ c_2\\ \vdots \\ c_n\end{bmatrix}\neq 0是方程組Ax=0的非零解。證畢

線性表示與線性無關性

設V是\mathbb{F}上的線性空間,\alpha_1,\alpha_2,...,\alpha_p和\beta_1, \beta_2,...,\beta_q是V中兩個向量組。若\alpha_1,\alpha_2,...,\alpha_p線性無關,且\{\alpha_1,\alpha_2,...,\alpha_p\}≤_{lin}\{\beta_1,\beta_2,...,\beta_q\},則p≤q

證明:用反證法,假設p<qT\in \mathbb{F}^{q\times p},使得

\begin{bmatrix}\alpha_1 &\alpha_2 & \cdots & \alpha_p\end{bmatrix}=\begin{bmatrix}\beta_1 &\beta_2 & \cdots & \beta_q\end{bmatrix}T

因為扁的齊次方程組必有非零解,是以存在c\in \mathbb{F}^p非零,使得Tc=0。上述等式兩邊右乘c得

\begin{bmatrix}\alpha_1 &\alpha_2 & \cdots & \alpha_p\end{bmatrix}c=\begin{bmatrix}\beta_1 &\beta_2 & \cdots & \beta_q\end{bmatrix}Tc=0

因為c\in \mathbb{F}^p非零,此結論與\alpha_1,\alpha_2,...,\alpha_p線性無關沖突,證畢

以少表多,多必相關

\alpha_1, \alpha_2,...,\alpha_r與\beta_1, \beta_2,...,\beta_s是線性空間V中的兩個向量組。若:

  1. \alpha_1, \alpha_2,...,\alpha_r可由\beta_1, \beta_2,...,\beta_s線性表示
  2. r>s

則\alpha_1, \alpha_2,...,\alpha_r線性相關

證明:因為\{\alpha_1,\alpha_2,...,\alpha_r\}≤_{lin}\{\beta_1,\beta_2,...,\beta_s\},是以\exists A\in \mathbb{F}^{s\times r},使得

\begin{bmatrix}\alpha_1 & \alpha_2 & \cdots & \alpha_r\end{bmatrix} = \begin{bmatrix}\beta_1 & \beta_2 & \cdots & \beta_s\end{bmatrix}A

于是有r(\begin{bmatrix}\alpha_1 & \alpha_2 & \cdots & \alpha_r\end{bmatrix})≤min\{r(\begin{bmatrix}\beta_1 & \beta_2 & \cdots & \beta_s\end{bmatrix}), r(A)\}≤s<r

即r(\begin{bmatrix}\alpha_1 & \alpha_2 & \cdots & \alpha_r\end{bmatrix})<r\alpha_1, \alpha_2,...,\alpha_r線性相關

推論:若\{\alpha_1,\alpha_2,...,\alpha_r\}≤_{lin}\{\beta_1,\beta_2,...,\beta_s\},且\alpha_1, \alpha_2,...,\alpha_r線性無關,則r≤s