MA&ALA4.1&4.2_空间和子空间 (Space and Subspace)

注：本文是对Matrix Analysis and Applied Linear Algebra一书4.1节Space and Subspace和4.2节Four Fundamental Subspaces的学习笔记

向量空间(Vector Space)

定义一个向量空间涉及4样东西：非空向量集$\mathcal{V}$，标量场$\mathcal{F}$，代数运算向量加(vector addition)和标量乘(salar multiplication)

The set $\mathcal{V}$ is called a vector space over $\mathcal{F}$ when the vector addition and scalar multiplication operations satisfy the following properties:

(A1) $\mathbf{x}+\mathbf{y}\in \mathcal{V}$ for all $\mathbf{x},\mathbf{y}\in \mathcal{V}$ (closure property for vector addition)

(A2) $(\mathbf{x}+\mathbf{y})+\mathbf{z}=\mathbf{x}+(\mathbf{y}+\mathbf{z})$ for all $\mathbf{x},\mathbf{y},\mathbf{z}\in \mathcal{V}$

(A3) $\mathbf{x}+\mathbf{y}=\mathbf{y}+\mathbf{x}$ for all $\mathbf{x},\mathbf{y}\in \mathcal{V}$

(A4) There is an element $0\in \mathcal{V}$ such that $\mathbf{x}+0=\mathbf{x}$ for all $\mathbf{x}\in \mathcal{V}$

(A5) For each $\mathbf{x}\in \mathcal{V}$, there is an element $(-\mathbf{x})\in \mathcal{V}$ such that $\mathbf{x}+(-\mathbf{x})=0$

(M1) $\alpha \mathbf{x}\in \mathcal{V}$ for all $\alpha \in \mathcal{F}$ and $\mathbf{x}\in \mathcal{V}$. (closure property for scalar multiplication)

(M2) $(\alpha \beta )\mathbf{x}=\alpha (\beta \mathbf{x})$ for all $\alpha ,\beta \in \mathcal{F}$ and every $\mathbf{x}\in \mathcal{V}$

(M3) $\alpha (\mathbf{x}+\mathbf{y})=\alpha \mathbf{x}+\alpha \mathbf{y}$ for every $\alpha \in \mathcal{F}$ and all $\mathbf{x},\mathbf{y}\in \mathcal{V}$

(M4) $(\alpha +\beta )\mathbf{x}=\alpha \mathbf{x}+\beta \mathbf{x}$ for all $\alpha ,\beta \in \mathcal{F}$ and every $\mathbf{x}\in \mathcal{V}$

(M5) $1\mathbf{x}=\mathbf{x}$ for all $\mathbf{x}\in \mathcal{V}$

一些向量空间的例子：

• The set ${\Re }^{m\times n}$ of $m\times n$ real matrices is a vector space over $\Re$.

• The set ${\mathcal{C}}^{m\times n}$ of $m\times n$ complex matrices is a vector space over $\mathcal{C}$.

• If $n=1$(or $m=1$), we have coordinate spaces
  $${\Re }^{m\times 1}=\left\{\left(\begin{array}{c}{x}_1\\ x_2\\ ⋮\\ x_m\end{array}\right),x_i\in \Re \right\}\text{ or }{\mathcal{C}}^{m\times 1}=\left\{\left(\begin{array}{c}{x}_1\\ x_2\\ ⋮\\ x_m\end{array}\right),x_i\in \mathcal{C}\right\}$$

• With function addition and multiplication defined by
  $$(f+g)(x)=f(x)+g(x)\text{ and }(\alpha f)(x)=\alpha f(x)$$
  the following sets are vector spaces over $\Re$:

  • The set of functions mapping the interval $[0,1]$ into $\Re$
  • The set of all real-valued continuous functions defined on $[0,1]$
  • The set of real-valued functions that are differentiable on $[0,1]$
  • The set of all polynomials with real coefficients

• Consider the vector space ${\Re }^2$, and let L = { ( x , y ) ∣ y = α x } \mathcal L=\{(x,y)|y=\alpha x\} L={(x,y)∣y=αx} be a line through the origin. $\mathcal{L}$ is a subset of ${\Re }^2$, but $\mathcal{L}$ is a special kind of subset because $\mathcal{L}$ also satisfies the properties (A1-A5) and (M1-M5) that define a vector space.

最后一个例子表明，一个向量空间可能包含一个“更小”的向量空间，这就引出了子空间(subspace)的定义。

子空间(Subspace)

Let $\mathcal{S}$ be a nonempty subset of a vector space $\mathcal{V}$ over $\mathcal{F}$, i.e., $\mathcal{S}\subseteq \mathcal{V}$.

If $\mathcal{S}$ is also a vector space over $\mathcal{F}$ using the same addition and scalar multiplication operations, then $\mathcal{S}$ is said to be a subspace of $\mathcal{V}$.

It’s not necessary to check all 10 of the defining conditions in order to determine if a subset is also a subspace—only the closure conditions (A1) and (M1) need to be considered. That is, a nonempty subset $\mathcal{S}$ of a vector space $\mathcal{V}$ is a subspace of $\mathcal{V}$ if and only if

(A1) $\mathbf{x},y\in \mathcal{S}⟹\mathbf{x}+\mathbf{y}\in \mathcal{S}$

(M1) $\mathbf{x}\in \mathcal{S}⟹\alpha \mathbf{x}\in \mathcal{S}$ for all $\alpha \in \mathcal{F}$

为什么满足两条就够了？

首先，因为$\mathcal{S}$是$\mathcal{V}$的子集，它自动满足除了(A1), (A4), (A5), (M1)之外的所有条件，而(A1)+(M1)是可以推出(A4), (A5)的：由(M1)有$(-\mathbf{x})=(-1)\mathbf{x}\in \mathcal{S}$，可以推出(A5)成立；由(A1)又有$\mathbf{x}+(-\mathbf{x})\in \mathcal{S}$, 结合(M1)有$0\in \mathcal{S}$, (A4)成立。

一些子空间的例子：

• Given a vector space $\mathcal{V}$, the set Z = { 0 } \mathcal Z=\{\mathbf 0\} Z={0} containing only the zero vector is a subspace of $\mathcal{V}$. This subspace is called the trivial subspace.

• We have already observed that straight lines through the origin in ${\Re }^2$ are subspaces, but what about straight lines not through the origin? No—they cannot be subspaces because subspaces must contain the zero vector. Consequently, the only proper subspaces of ${\Re }^2$ are the trivial subspace and lines through the origin.

• Similarly, the only proper subspaces of ${\Re }^3$ are the trivial subspace and lines through the origin, and planes through the origin.

• Generalize to higher dimension: For a set of vectors S = { v 1 , v 2 , ⋯   , v r } \mathcal S=\{ \mathbf v_1,\mathbf v_2,\cdots,\mathbf v_r\} S={v1​,v2​,⋯,vr​} from a vector space $\mathcal{V}$, the set of all possible linear combinations of the ${\mathbf{v}}_i$'s is denoted by
  s p a n ( S ) = { α 1 v 1 + α 2 v 2 + ⋯ + α r v r ∣ α i ∈ F } span(\mathcal S)=\{\alpha_1 \mathbf v_1+\alpha_2 \mathbf v_2+\cdots+\alpha_r \mathbf v_r|\alpha _i \in \mathcal F \} span(S)={α1​v1​+α2​v2​+⋯+αr​vr​∣αi​∈F}
  $span(\mathcal{S})$ is a subspace of $\mathcal{V}$.

事实上，所有的子空间都可以用$span(\mathcal{S})$的形式表示出来，因此我们引入了下面的定义

For a set of vectors S = { v 1 , v 2 , ⋯   , v r } \mathcal S=\{ \mathbf v_1,\mathbf v_2,\cdots,\mathbf v_r\} S={v1​,v2​,⋯,vr​}, the subspace
s p a n ( S ) = { α 1 v 1 + α 2 v 2 + ⋯ + α r v r } span(\mathcal S)=\{\alpha_1 \mathbf v_1+\alpha_2 \mathbf v_2+\cdots+\alpha_r \mathbf v_r\} span(S)={α1​v1​+α2​v2​+⋯+αr​vr​}
generated by forming all linear combinations of vectors from $\mathcal{S}$ is called the space spanned by $\mathcal{S}$.

If $\mathcal{V}$ is a vector space such that $\mathcal{V}=span(\mathcal{S})$, we say $\mathcal{S}$ is a spanning set for $\mathcal{V}$. In other words, $\mathcal{S}$ spans $\mathcal{V}$ whenever each vector in $\mathcal{V}$ is a linear combination of vectors from $\mathcal{S}$.

一些例子：

• $\mathcal{S}=\left\{\left(\begin{array}{c}1\\ 1\end{array}\right),\left(\begin{array}{c}2\\ 2\end{array}\right)\right\}$ spans the line $y=x$ in ${\Re }^2$
• The unit vectors { e 1 , e 2 , ⋯   , e n } \{\mathbf e_1,\mathbf e_2,\cdots,\mathbf e_n\} {e1​,e2​,⋯,en​} in ${\Re }^n$ form a spanning set for ${\Re }^n$.
• For a set of vectors S = { a 1 , ⋯   , a n } \mathcal S=\{\mathbf a_1,\cdots,\mathbf a_n\} S={a1​,⋯,an​} from a subspace $\mathcal{V}\subseteq {\Re }^{m\times 1}$, let $\mathbf{A}$ be a matrix containing the ${\mathbf{a}}_i$'s as its columns. $\mathcal{S}$ spans $\mathcal{V}$ iff for each $\mathbf{b}\in \mathcal{V}$ there corresponds a column $\mathbf{x}$ such that $\mathbf{A}\mathbf{x}=\mathbf{b}$. (i.e., iff $\mathbf{A}\mathbf{x}=\mathbf{b}$ is a consistent system for every $\mathbf{b}\in \mathcal{V}⟺rank[\mathbf{A}|\mathbf{b}]=rank(\mathbf{A})$)
• The finite set { 1 , x , x 2 , ⋯   , x n } \{1,x,x^2,\cdots,x^n\} {1,x,x2,⋯,xn} spans the space of all polynomials such that $\mathrm{d}\mathrm{e}\mathrm{g}p(x)\le n$, and the infinite set { 1 , x , x 2 , ⋯   } \{1,x,x^2,\cdots\} {1,x,x2,⋯} spans the space of all polynomials.

我们也可以定义两个子空间的“加法”，相加后得到的依然是一个子空间，而且此时的spanning set就是原先两个spanning sets的并：

If $\mathcal{X}$ and $\mathcal{Y}$ are subspaces of a vector space $\mathcal{V}$, then the sum of $\mathcal{X}$ and $\mathcal{Y}$ is defined to be the set of all possible sums of vectors from $\mathcal{X}$ with vectors from $\mathcal{Y}$. That is
X + Y = { x + y ∣ x ∈ X  and  y ∈ Y } \mathcal X+\mathcal Y=\{\mathbf x+\mathbf y|\mathbf x\in \mathcal X\text{ and } \mathbf y\in \mathcal Y\} X+Y={x+y∣x∈X and y∈Y}

• The sum $\mathcal{X}+\mathcal{Y}$ is again a subspace of $\mathcal{V}$
• If ${\mathcal{S}}_X,{\mathcal{S}}_Y$ span $\mathcal{X},\mathcal{Y}$, then ${\mathcal{S}}_X\cup {\mathcal{S}}_Y$ spans $\mathcal{X}+\mathcal{Y}$

一个例子：若$\mathcal{X}\subseteq {\Re }^2$和$\mathcal{Y}\subseteq {\Re }^2$是两个子空间，在几何上相当于两条过原点的线，那么$\mathcal{X}+\mathcal{Y}={\Re }^2$. （平行四边形法则）

四个基本子空间

列空间(Column Space)和行空间(Row Space)

子空间的概念和线性方程其实是紧密联系在一起的。

For a linear function $f$ mapping ${\Re }^n$ into ${\Re }^m$, let $\mathcal{R}(f)$ denote the range of $f$. That is, R ( f ) = { f ( x ) ∣ x ∈ ℜ n } ⊆ ℜ m \mathcal R(f)=\{f(\mathbf x)|\mathbf x\in \Re^n\}\subseteq \Re^m R(f)={f(x)∣x∈ℜn}⊆ℜm is the set of all “images” as $\mathbf{x}$ varies freely over ${\Re }^n$.

• The range of every linear function $f:{\Re }^n\to {\Re }^m$ is a subspace of ${\Re }^m$, and every subspace of ${\Re }^m$ is the range of some linear function.

For this reason, subspaces of ${\Re }^m$ are sometimes called linear spaces.

这个结果表明对于任意矩阵$\mathbf{A}\in {\Re }^{m\times n}$，都可以由线性方程$f(\mathbf{x})=\mathbf{A}\mathbf{x}$构造一个${\Re }^m$的子空间。类似地，${\mathbf{A}}^T\in {\Re }^{n\times m}$也可以由$f(\mathbf{y})={\mathbf{A}}^T\mathbf{y}$定义一个${\Re }^n$的子空间。从这里我们可以得到四个基本子空间中的两种：

The range of a matrix $\mathbf{A}\in {\Re }^{m\times n}$ is defined to be the subspace $\mathcal{R}(\mathbf{A})$ of ${\Re }^m$ that is generated by the range of $f(\mathbf{x})=\mathbf{A}\mathbf{x}$. That is
R ( A ) = { A x ∣ x ∈ ℜ n } ⊆ ℜ m b ∈ R ( A )    ⟺    b = A x  for some  x \mathcal R(\mathbf A)=\{\mathbf A \mathbf x|\mathbf x\in \Re^n\}\subseteq \Re^m\\ \mathbf b \in \mathcal R(\mathbf A)\iff \mathbf b=\mathbf A \mathbf x\text{ for some }\mathbf x R(A)={Ax∣x∈ℜn}⊆ℜmb∈R(A)⟺b=Ax for some x
Similarly, the range of matrix ${\mathbf{A}}^T$ is the subspace of ${\Re }^n$ defined by
R ( A T ) = { A T y ∣ y ∈ ℜ m } ⊆ ℜ n a ∈ R ( A T )    ⟺    a = A T y  for some  y \mathcal R(\mathbf A^T)=\{\mathbf A^T \mathbf y|\mathbf y\in \Re^m\}\subseteq \Re^n\\ \mathbf a\in \mathcal R(\mathbf A^T)\iff \mathbf a=\mathbf A^T \mathbf y\text{ for some }\mathbf y R(AT)={ATy∣y∈ℜm}⊆ℜna∈R(AT)⟺a=ATy for some y
Some people also call $\mathcal{R}(\mathbf{A})$ the image space of $\mathbf{A}$.

如果我们把$\mathbf{A}$和$\mathbf{x}$拆开来写
$$\mathbf{A}\mathbf{x}=({\mathbf{A}}_{\ast 1}|{\mathbf{A}}_{\ast 2}|\cdots |{\mathbf{A}}_{\ast n})\left(\begin{array}{c}{\xi }_1\\ {\xi }_2\\ ⋮\\ {\xi }_n\end{array}\right)=\sum _{j=1}^n{\xi }_j{\mathbf{A}}_{\ast j}$$
可以看到$\mathcal{R}(\mathbf{A})$其实就是$\mathbf{A}$的列向量的线性组合，也就是由$\mathbf{A}$的列向量张成的空间，这就是为什么$\mathcal{R}(\mathbf{A})$经常被称作**列空间(column space)**.

类似地，$\mathcal{R}({\mathbf{A}}^T)$是$\mathbf{A}$的行向量张成的空间，也被称为**行空间(row space)**.

---

有时候我们想知道两个矩阵是否有一样的列空间/行空间，这时我们可以看它们是否列等价/行等价：

For two matrices $\mathbf{A}$ and $\mathbf{B}$ of the same shape:

• $\mathcal{R}(\mathbf{A})=\mathcal{R}(\mathbf{B})$ iff $\mathbf{A}\stackrel{\text{col}}{\sim }\mathbf{B}$
• $\mathcal{R}({\mathbf{A}}^T)=\mathcal{R}({\mathbf{B}}^T)$ iff $\mathbf{A}\stackrel{\text{row}}{\sim }\mathbf{B}$

证明：证明行等价的情况，列等价时类似。

$⟸:$ 首先根据$\mathbf{A}\stackrel{\text{row}}{\sim }\mathbf{B}$可知存在可逆矩阵$\mathbf{P}$使得$\mathbf{P}\mathbf{A}=\mathbf{B}$. 假设向量$\mathbf{a}$属于行空间$\mathcal{R}({\mathbf{A}}^T)$
$$\begin{array}{rl}\mathbf{a}\in \mathcal{R}({\mathbf{A}}^T)& ⟺{\mathbf{a}}^T={\mathbf{y}}^T\mathbf{A}={\mathbf{y}}^T{\mathbf{P}}^{-1}\mathbf{P}\mathbf{A}\text{ for some }{\mathbf{y}}^T\\ & ⟺{\mathbf{a}}^T={\mathbf{z}}^T\mathbf{B}\text{ for }{\mathbf{z}}^T={\mathbf{y}}^T{\mathbf{P}}^{-1}\\ & ⟺\mathbf{a}\in \mathcal{R}({\mathbf{B}}^T)\end{array}$$
$⟹:$ 若$\mathcal{R}({\mathbf{A}}^T)=\mathcal{R}({\mathbf{B}}^T)$, 则
s p a n { A 1 ∗ , A 2 ∗ , ⋯   , A m ∗ } = s p a n { B 1 ∗ , B 2 ∗ , ⋯   , B m ∗ } span\{\mathbf A_{1*},\mathbf A_{2*},\cdots,\mathbf A_{m*} \}=span\{\mathbf B_{1*},\mathbf B_{2*},\cdots,\mathbf B_{m*}\} span{A1∗​,A2∗​,⋯,Am∗​}=span{B1∗​,B2∗​,⋯,Bm∗​}
$\mathbf{B}$的每一行都能用矩阵$\mathbf{A}$行向量的线性组合表示出来，反之也成立。因此可以通过行变换由$\mathbf{A}$变到$\mathbf{B}$（略去细节），也即$\mathbf{A}\stackrel{\text{row}}{\sim }\mathbf{B}$.

---

我们已经知道$\mathbf{A}$的行向量可以张成$\mathcal{R}({\mathbf{A}}^T)$，$\mathbf{A}$的列向量可以张成$\mathcal{R}(\mathbf{A})$. 但有些时候张成这些空间并不需要用到所有的行向量/列向量：

Let $\mathbf{A}$ be an $m\times n$ matrix, and let $\mathbf{U}$ be any row echelon form derived from $\mathbf{A}$. Spanning sets for the row and column spaces are as follows:

• The nonzero rows of $\mathbf{U}$ span $\mathcal{R}({\mathbf{A}}^T)$
• The basic columns in $\mathbf{A}$ span $\mathcal{R}(\mathbf{A})$

这条性质可以通过证明列等价/行等价并运用上一条性质得到。

---

零空间(Nullspace)和左零空间(Left-Hand Nullspace)

另外两个基本子空间也可以从线性函数的角度考虑。记$f$为一个从${\Re }^n$映射到${\Re }^m$的线性函数，定义
N ( f ) = { x ∣ f ( x ) = 0 } \mathcal N(f)=\{\mathbf x|f(\mathbf x)=\mathbf 0\} N(f)={x∣f(x)=0}
容易证明$\mathcal{N}(f)$是一个子空间，因为它满足**(A1)和(M1)**:

若${\mathbf{x}}_1,{\mathbf{x}}_2\in \mathcal{N}(f)$, 则由$f$的线性性，
$$f({\mathbf{x}}_1+{\mathbf{x}}_2)=f({\mathbf{x}}_1)+f({\mathbf{x}}_2)=0+0=0⟹{\mathbf{x}}_1+{\mathbf{x}}_2\in \mathcal{N}(f)$$
类似地，若$\alpha \in \Re$，且$\mathbf{x}\in \mathcal{N}(f)$,
$$f(\alpha \mathbf{x})=\alpha f(\mathbf{x})=\alpha 0=0⟹\alpha \mathbf{x}\in \mathcal{N}(f)$$
接下来我们给出**零空间(Nullspace)和左零空间(Left-Hand Nullspace)**的正式定义：

• For an $m\times n$ matrix $\mathbf{A}$, the set N ( A ) = { x n × 1 ∣ A x = 0 } ⊆ ℜ n \mathcal N(\mathbf A)=\{\mathbf x_{n\times 1}|\mathbf A \mathbf x=\mathbf 0 \}\subseteq \Re^n N(A)={xn×1​∣Ax=0}⊆ℜn is called the nullspace of $\mathbf{A}$. In other words, $\mathcal{N}(\mathbf{A})$ is simply the set of all solutions to the homogeneous system $\mathbf{A}\mathbf{x}=0$.
• The set N ( A T ) = { y m × 1 ∣ A T y = 0 } ⊆ ℜ m \mathcal N(\mathbf A^T)=\{\mathbf y_{m\times 1}|\mathbf A^T\mathbf y=\mathbf0 \}\subseteq \Re^m N(AT)={ym×1​∣ATy=0}⊆ℜm is called the left-hand nullspace of $\mathbf{A}$ because $\mathcal{N}({\mathbf{A}}^T)$ is the set of all solutions to the left-hand homogeneous system ${\mathbf{y}}^T\mathbf{A}=0^T$.

既然零空间是子空间，它也可以被一组向量张成。由齐次方程组解的相关结论（MA&ALA2_行阶梯矩阵和秩）

To determine a spanning set for $\mathcal{N}(\mathbf{A})$, where $rank({\mathbf{A}}_{m\times n})=r$, row reduce $\mathbf{A}$ to a row echelon form $\mathbf{U}$, and solve $\mathbf{U}\mathbf{x}=0$ for the basic variables in terms of the free variables to produce the general solution of $\mathbf{A}\mathbf{x}=0$ in the form
$$\mathbf{x}=x_{f_1}{\mathbf{h}}_1+x_{f_2}{\mathbf{h}}_2+\cdots +x_{f_{n-r}}{\mathbf{h}}_{n-r}$$

• By definition, the set H = { h 1 , h 2 , ⋯   , h n − r } \mathcal H=\{\mathbf h_1,\mathbf h_2,\cdots,\mathbf h_{n-r} \} H={h1​,h2​,⋯,hn−r​} spans $\mathcal{N}(\mathbf{A})$.
• N ( A ) = { 0 }    ⟺    r a n k ( A ) = n \mathcal N(\mathbf A)=\{\mathbf 0\}\iff rank(\mathbf A)=n N(A)={0}⟺rank(A)=n
• N ( A T ) = { 0 }    ⟺    r a n k ( A ) = m \mathcal N(\mathbf A^T)=\{\mathbf 0\}\iff rank(\mathbf A)=m N(AT)={0}⟺rank(A)=m

接下来我们考虑左零空间的生成集合(spanning set). 当然我们可以把${\mathbf{A}}^T$看作${\mathbf{A}}^{\prime }$求最简阶梯型${\mathbf{E}}_{{\mathbf{A}}^T}$然后得到齐次方程组的解以及生成集合，但是考虑到其它三个基本子空间都可以通过${\mathbf{E}}_{\mathbf{A}}$求得，我们希望最好也能从${\mathbf{E}}_{\mathbf{A}}$出发得到左零空间的生成集合。但这不是一个显然的过程，因为${\mathbf{E}}_{\mathbf{A}}\ne {\mathbf{E}}_{{\mathbf{A}}^T}$.

If $rank\left({\mathbf{A}}_{m\times n}\right)=r$, and if $\mathbf{P}\mathbf{A}=\mathbf{U}$, where $\mathbf{P}$ is nonsingular and $\mathbf{U}$ is in row echelon form, then the last $m-r$ rows in $\mathbf{P}$ span the left-hand nullspace of $\mathbf{A}$. In other words, if $\mathbf{P}=\left(\begin{array}{c}{\mathbf{P}}_1\\ {\mathbf{P}}_2\end{array}\right)$, where ${\mathbf{P}}_2$ is $(m-r)\times m$, then
$$\mathcal{N}\left({\mathbf{A}}^T\right)=\mathcal{R}\left({\mathbf{P}}_2^T\right)$$

证明：记$\mathbf{U}=\left(\begin{array}{c}\mathbf{C}\\ 0\end{array}\right)$, 其中${\mathbf{C}}_{r\times n}$，由$\mathbf{P}\mathbf{A}=\mathbf{U}$有${\mathbf{P}}_2\mathbf{A}=0$.

首先证明$\mathcal{R}({\mathbf{P}}_2^T)\subseteq \mathcal{N}({\mathbf{A}}^T)$: 若$\mathbf{b}\in \mathcal{R}({\mathbf{P}}_2^T)$，则存在某个$\mathbf{x}$使得${\mathbf{P}}_2^T\mathbf{x}=\mathbf{b}$. 因为${\mathbf{A}}^T\mathbf{b}={\mathbf{A}}^T{\mathbf{P}}_2^T\mathbf{x}=({\mathbf{P}}_2\mathbf{A}{)}^T\mathbf{x}=0$, 则$\mathbf{b}\in \mathcal{N}({\mathbf{A}}^T)$.

然后证明$\mathcal{N}({\mathbf{A}}^T)\subseteq \mathcal{R}({\mathbf{P}}_2^T)$. 假设${\mathbf{y}}^T\in \mathcal{N}({\mathbf{A}}^T)$, 记${\mathbf{P}}^{-1}=({\mathbf{Q}}_1{\mathbf{Q}}_2)$
$$0={\mathbf{y}}^T\mathbf{A}={\mathbf{y}}^T{\mathbf{P}}^{-1}\mathbf{U}={\mathbf{y}}^T{\mathbf{Q}}_1\mathbf{C}⟹0={\mathbf{y}}^T{\mathbf{Q}}_1$$
这是因为 r a n k ( C ) = r    ⟺    N ( C T ) = { 0 } rank(\mathbf C)=r\iff\mathcal N(\mathbf C^T)=\{\mathbf 0\} rank(C)=r⟺N(CT)={0}（见上一条结论）

根据$\mathbf{P}{\mathbf{P}}^{-1}=\mathbf{I}={\mathbf{P}}^{-1}\mathbf{P}$，有${\mathbf{P}}_1{\mathbf{Q}}_1={\mathbf{I}}_r$和${\mathbf{Q}}_1{\mathbf{P}}_1={\mathbf{I}}_m-{\mathbf{Q}}_2{\mathbf{P}}_2$, 那么
$$\begin{array}{rl}0={\mathbf{y}}^T{\mathbf{Q}}_1& ⟹0={\mathbf{y}}^T{\mathbf{Q}}_1{\mathbf{P}}_1={\mathbf{y}}^T(\mathbf{I}-{\mathbf{Q}}_2{\mathbf{P}}_2)\\ & ⟹{\mathbf{y}}^T={\mathbf{y}}^T{\mathbf{Q}}_2{\mathbf{P}}_2=({\mathbf{y}}^T{\mathbf{Q}}_2){\mathbf{P}}_2\\ & ⟹\mathbf{y}\in \mathcal{R}({\mathbf{P}}_2^T)\end{array}$$
类似地，我们还可以证明

If $rank\left({\mathbf{A}}_{m\times n}\right)=r$, and if $\mathbf{P}\mathbf{A}=\mathbf{U}=\left(\begin{array}{c}\mathbf{C}\\ 0\end{array}\right)$, where $\mathbf{P}=\left(\begin{array}{c}{\mathbf{P}}_1\\ {\mathbf{P}}_2\end{array}\right)$ is nonsingular and $\mathbf{U}$ is in row echelon form, then
$$\mathcal{R}(\mathbf{A})=\mathcal{N}({\mathbf{P}}_2)$$

证明：

首先证明$\mathcal{R}(\mathbf{A})\subseteq \mathcal{N}({\mathbf{P}}_2)$: 若$\mathbf{b}\in \mathcal{R}(\mathbf{A})$，则存在某个$\mathbf{x}$使得$\mathbf{A}\mathbf{x}=\mathbf{b}$. 因为${\mathbf{P}}_2\mathbf{b}={\mathbf{P}}_2\mathbf{A}\mathbf{x}=({\mathbf{P}}_2\mathbf{A})\mathbf{x}=0$, 则$\mathbf{b}\in \mathcal{N}({\mathbf{P}}_2)$.

然后证明$\mathcal{N}({\mathbf{P}}_2)\subseteq \mathcal{R}(\mathbf{A})$. 若$\mathbf{b}\in \mathcal{N}({\mathbf{P}}_2)$,
$$\mathbf{P}\mathbf{b}=\left(\begin{array}{c}{\mathbf{P}}_1\\ {\mathbf{P}}_2\end{array}\right)\mathbf{b}=\left(\begin{array}{c}{\mathbf{P}}_1\mathbf{b}\\ {\mathbf{P}}_2\mathbf{b}\end{array}\right)=\left(\begin{array}{c}{\mathbf{d}}_{r\times 1}\\ 0\end{array}\right)$$
因此$\mathbf{P}(\mathbf{A}|\mathbf{b})=(\mathbf{P}\mathbf{A}|\mathbf{P}\mathbf{b})=\left(\begin{array}{cc}\mathbf{C}& \mathbf{d}\\ 0& 0\end{array}\right)$， 这意味着
$$rank[\mathbf{A}|\mathbf{b}]=r=rank(\mathbf{A})$$
也就是说$\mathbf{A}\mathbf{x}=\mathbf{b}$有唯一解，因此$\mathbf{b}\in \mathcal{R}(\mathbf{A})$.

有时候我们需要知道两个矩阵是否有同样的零空间或者左零空间，可以用下面这条性质：

For two matrices $\mathbf{A}$ and $\mathbf{B}$ of the same shape:

• $\mathcal{N}(\mathbf{A})=\mathcal{N}(\mathbf{B})⟺\mathbf{A}\stackrel{\text{row}}{\sim }\mathbf{B}$
• $\mathcal{N}({\mathbf{A}}^T)=\mathcal{N}({\mathbf{B}}^T)⟺\mathbf{A}\stackrel{\text{col}}{\sim }\mathbf{B}$

---

总结

The four fundamental subspaces associated with ${\mathbf{A}}_{m\times n}$ are as follows.

• The range or column space: R ( A ) = { A x } ⊆ ℜ m R(\mathbf{A})=\{\mathbf{A} \mathbf{x}\} \subseteq \Re^{m} R(A)={Ax}⊆ℜm

• The row space or left-hand range: $R\left({\mathbf{A}}^T\right)=\left\{{\mathbf{A}}^T\mathbf{y}\right\}\subseteq {\Re }^n$

• The nullspace: N ( A ) = { x ∣ A x = 0 } ⊆ ℜ n N(\mathbf{A})=\{\mathrm{x} | \mathbf{A} \mathbf{x}=\mathbf{0}\} \subseteq \Re^{n} N(A)={x∣Ax=0}⊆ℜn

• The left-hand nullspace: $N\left({\mathbf{A}}^T\right)=\left\{\mathbf{y}|{\mathbf{A}}^T\mathbf{y}=0\right\}\subseteq {\Re }^m$

Let $\mathbf{P}$ be a nonsingular matrix such that $\mathbf{P}\mathbf{A}=\mathbf{U}$, where $\mathbf{U}$ is in row echelon form, and suppose $rank(\mathbf{A})=r$.

• Spanning set for $R(\mathbf{A})=$ the basic columns in $\mathbf{A}$

• Spanning set for $R\left({\mathbf{A}}^T\right)=$ the nonzero rows in $\mathbf{U}$

• Spanning set for $N(\mathbf{A})=$ the ${\mathbf{h}}_i$'s in the general solution of $\mathbf{A}\mathbf{x}=0$

• Spanning set for $N\left({\mathbf{A}}^T\right)=$ the last $m-r$ rows of $\mathbf{P}$

If $\mathbf{A}$ and $\mathbf{B}$ have the same shape, then

• $\mathbf{A}\stackrel{\text{ row }}{\sim }\mathbf{B}⟺N(\mathbf{A})=N(\mathbf{B})⟺R\left({\mathbf{A}}^T\right)=R\left({\mathbf{B}}^T\right)$
• $\mathbf{A}\stackrel{\text{ col }}{\sim }\mathbf{B}⟺R(\mathbf{A})=R(\mathbf{B})⟺N\left({\mathbf{A}}^T\right)=N\left({\mathbf{B}}^T\right)$