8.4 Hyperplanes (超平面)

本文为《Linear algebra and its applications》的读书笔记

Hyperplanes

Hyperplanes play a special role in the geometry of ${\mathbb{R}}^n$ because they divide the space into two disjoint pieces, just as a plane separates ${\mathbb{R}}^3$ into two parts and a line cuts through ${\mathbb{R}}^2$. The key to working with hyperplanes is to use simple $implicit$ descriptions, rather than the $explicit$ or parametric representations of lines and planes used in the earlier work with affine sets.

An implicit equation of a line in ${\mathbb{R}}^2$ has the form $ax+by=d$. An implicit equation of a plane in ${\mathbb{R}}^3$ has the form $ax+by+cz=d$. Both equations describe the line or plane as the set of all points at which a linear expression (also called a $linear$ $functional$ (线性函数)) has a fixed value, $d$.

If $f$ is a linear functional on ${\mathbb{R}}^n$, then the standard matrix of this linear transformation $f$ is a $1\times n$ matrix $A$, say $A=\left[\begin{array}{cccc}{a}_1& a_2& ...& a_n\end{array}\right]$. So

If $f$ is a nonzero functional, then $rankA=1$, and $dimNulA=n-1$. Thus, the subspace $[f:0]$ has dimension $n-1$ and so is a hyperplane. Also, if $d$ is any number in $R$, then

Recall that the set of solutions of $A\mathbf{x}=\mathbf{b}$ is obtained by translating the solution set of $A\mathbf{x}=0$. Then

Thus the sets $[f:d]$ are hyperplanes parallel to $[f:0]$. See Figure 1.

When $A$ is a $1\times n$ matrix, the equation $A\mathbf{x}=d$ may be written with an inner product $\mathbf{n}\cdot \mathbf{x}$, using $\mathbf{n}$ in ${\mathbb{R}}^n$ with the same entries as $A$. Thus, from (2),

Then [ f : 0 ] = { x ∈ R n : n ⋅ x = 0 } [f: 0]=\{\boldsymbol x\in R^n:\boldsymbol n\cdot \boldsymbol x=0\} [f:0]={x∈Rn:n⋅x=0}, which shows that $[f,0]$ is the orthogonal complement of the subspace spanned by $\mathbf{n}$. In the terminology of calculus and geometry for ${\mathbb{R}}^3$, $\mathbf{n}$ is called a normal vector (法向量) to $[f:0]$. (A “normal” vector in this sense need not have unit length.) Also, $\mathbf{n}$ is said to be normal to each parallel hyperplane $[f:d]$, even though $\mathbf{n}\cdot x$ is not zero when $d\ne 0$.

Another name for $[f:d]$ is a level set (水平集) of $f$ , and $n$ is sometimes called the gradient (梯度) of $f$ when $f(\mathbf{x})=\mathbf{n}\cdot \mathbf{x}$ for each $\mathbf{x}$.

---

The next three examples show connections between implicit and explicit descriptions of hyperplanes.

EXAMPLE 4
In ${\mathbb{R}}^2$, give an explicit description of the line $x-4y=13$ in parametric vector form.
SOLUTION

EXAMPLE 5
Let

, and let $L_1$ be the line through ${\mathbf{v}}_1$ and ${\mathbf{v}}_2$. Find a linear functional $f$ and a constant $d$ such that $L_1=[f:d]$.
SOLUTION
The line $L_1$ is parallel to the translated line $L_0$ through ${\mathbf{v}}_2-{\mathbf{v}}_1$ and the origin. The defining equation for $L_0$ has the form

Since $\mathbf{n}$ is orthogonal to the subspace $L_0$, which contains ${\mathbf{v}}_2-{\mathbf{v}}_1$, then

By inspection, a solution is $[ab]=[25]$. Let $f(x,y)=2x+5y$. From (5), $L_0=[f:0]$, and $L_1=[f:d]$ for some $d$. Since ${\mathbf{v}}_1$ is on line $L_1$, $d=f({\mathbf{v}}_1)=2(1)+5(2)=12$. Thus, the equation for $L_1$ is $2x+5y=12$.

EXAMPLE 6
Let

Find an implicit description $[f:d]$ of the plane $H_1$ that passes through ${\mathbf{v}}_1,{\mathbf{v}}_2,$ and ${\mathbf{v}}_3$.
SOLUTION
$H_1$ is parallel to a plane $H_0$ through the origin that contains the translated points

Since these two points are linearly independent, H 0 = S p a n { v 2 − v 1 , v 3 − v 1 } H_0= Span\{\boldsymbol v_2-\boldsymbol v_1, \boldsymbol v_3-\boldsymbol v_1\} H0​=Span{v2​−v1​,v3​−v1​}. Let $\mathbf{n}=\left[\begin{array}{c}a\\ b\\ c\end{array}\right]$ be the normal to $H_0$. Then ${\mathbf{v}}_2-{\mathbf{v}}_1$ and ${\mathbf{v}}_3-{\mathbf{v}}_1$ are each orthogonal to $\mathbf{n}$:

These two equations form a system whose augmented matrix can be row reduced:

Row operations yield

Set $c=4$, for instance. Then $\mathbf{n}=\left[\begin{array}{c}-2\\ 5\\ 4\end{array}\right]$ and $H_0=[f:0]$, where $f(\mathbf{x})=-2x_1+5x_2+4x_3$.

The parallel hyperplane $H_1$ is $[f:d]$. To find $d$, use the fact that ${\mathbf{v}}_1$ is in $H_1$, and compute $d=f({\mathbf{v}}_1)=f(1,1,1)=7$.

The procedure in Example 6 generalizes to higher dimensions. However, for the
special case of ${\mathbb{R}}^3$, one can also use the cross-product formula (叉积公式) to compute $n$, using a symbolic determinant as a mnemonic device:

If only the formula for $f$ is needed, the cross-product calculation may be written as an ordinary determinant:

---

PROOF
Suppose that $H$ is a hyperplane, take $\mathbf{p}\in H$, and let $H_0=H-\mathbf{p}$. Then $H_0$ is an $(n-1)$-dimensional subspace. Next, take any point $\mathbf{y}$ that is not in $H_0$. By the Orthogonal Decomposition Theorem,

where ${\mathbf{y}}_1$ is a vector in $H_0$ and $\mathbf{n}$ is orthogonal to every vector in $H_0$. The function $f$ defined by

is a linear functional, by properties of the inner product. Now, $[f:0]$ is a hyperplane that contains $H_0$, by construction of $\mathbf{n}$. It follows that

Finally, let $d=f(\mathbf{p})=\mathbf{n}\cdot \mathbf{p}$. Then, as in (3) shown earlier,

The converse statement that $[f:d]$ is a hyperplane follows from (1) and (3) above.

---

Many important applications of hyperplanes depend on the possibility of “separating” two sets by a hyperplane. The following terminology and notation will help to make this idea more precise.

Topology: 拓扑
open ball: 开球
A set is open: 开集
A set is closed: 闭集
A set is bounded: 有界集
A set is compact: 紧致集

EXERCISE 27
Give an example of a closed subset $S$ of ${\mathbb{R}}^2$ such that $convS$ is not closed.
SOLUTION
S = { p ∣ p = ( x , y ) , y = 1 / x , x ≥ 1 / 2 } S=\{\boldsymbol p|\boldsymbol p=(x,y),y=1/x,x\geq1/2\} S={p∣p=(x,y),y=1/x,x≥1/2}

EXERCISE 29
Prove that the open ball B ( p , δ ) = { x : ∥ x − p ∥ < δ } B(\boldsymbol p,\delta)=\{\boldsymbol x:\left\|\boldsymbol x-\boldsymbol p\right\|<\delta\} B(p,δ)={x:∥x−p∥<δ} is a convex set.
SOLUTION
[Hint: Use the Triangle Inequality.] (三角不等式)

EXAMPLE 7
Let

as shown in Figure 3. Then the set $S$ is closed since it contains all its boundary points. The set $S$ is bounded since $S\subset B(0,3)$. Thus $S$ is also compact.

$Notation$: If $f$ is a linear functional, then $f(A)\le d$ means $f(\mathbf{x})\le d$ for each $\mathbf{x}\in A$.

strictly seperate: 严格分割

Notice that strict separation requires that the two sets be disjoint, while mere separation does not.

PROOF
Suppose that $(convA)\cap (convB)=\varphi$. Since the convex hull of a compact set is compact, Theorem 12 ensures that there is a hyperplane $H$ that strictly separates $convA$ and $convB$. Clearly, $H$ also strictly separates the smaller sets $A$ and $B$.

Conversely, suppose the hyperplane $H=[f:d]$ strictly separates $A$ and $B$. Without loss of generality, assume that $f(A)<d$ and $f(B)>d$. Let $\mathbf{x}=c_1{\mathbf{x}}_1+...+c_k{\mathbf{x}}_k$ be any convex combination of elements of $A$. Then

Thus $f(convA)<d$. Likewise, $f(convB)>d$, so $H=[f:d]$ strictly separates $convA$ and $convB$. By Theorem 12, $convA$ and $convB$ must be disjoint.

---

EXERCISE 14
Let $F_1$ and $F_2$ be 4-dimensional flats in ${\mathbb{R}}^6$, and suppose that $F_1\cap F_2\ne \varphi$. What are the possible dimensions of $F_1\cap F_2$?
SOLUTION

下面的答案是我自己写的，感觉论证啰嗦且不太严谨，仅供参考
如果有好的解答，欢迎一起交流~

Let $F_1=W_1+{\mathbf{p}}_1,F_2=W_2+{\mathbf{p}}_2$, where $W_1,W_2$ are two 4-dimensional subspaces. Suppose ${\mathbf{a}}_1,...,{\mathbf{a}}_4$ and ${\mathbf{b}}_1,...,{\mathbf{b}}_4$ be the basis of $W_1$ and $W_2$ respectively. Let $F_1\cap F_2=W+\mathbf{p}$ and $\mathbf{x}\in W$, then there exist $m_i,n_i\in R$ ($1\le i\le 4$) such that
$$\begin{gathered} {\mathbf{p}}_1+m_1{\mathbf{a}}_1+...+m_4{\mathbf{a}}_4={\mathbf{p}}_2+n_1{\mathbf{b}}_1+...+n_4{\mathbf{b}}_4=\mathbf{x}+\mathbf{p} \\ {\mathbf{p}}_1-\mathbf{p}+m_1{\mathbf{a}}_1+...+m_4{\mathbf{a}}_4={\mathbf{p}}_2-\mathbf{p}+n_1{\mathbf{b}}_1+...+n_4{\mathbf{b}}_4=\mathbf{x}(1) \end{gathered}$$

Notice that $dim{F}_1\cap F_2=dimW$ ($dimW$ is the dimension of the solution set of $\mathbf{x}$).

Since $\mathbf{x}$ can be $0$, there exist $m_i^{\prime },n_i^{\prime }\in R$ ($1\le i\le 4$) such that

$${\mathbf{p}}_1+m_1^{\prime }{\mathbf{a}}_1+...+m_4^{\prime }{\mathbf{a}}_4={\mathbf{p}}_2+n_1^{\prime }{\mathbf{b}}_1+...+n_4^{\prime }{\mathbf{b}}_4=\mathbf{p}(2)$$

From $(2)$, we know that ${\mathbf{p}}_1-\mathbf{p}$ is a linear combination of { a 1 , . . . , a 4 } \{\boldsymbol a_1,...,\boldsymbol a_4\} {a1​,...,a4​} and ${\mathbf{p}}_2-\mathbf{p}$ is a linear combination of { b 1 , . . . , b 4 } \{\boldsymbol b_1,...,\boldsymbol b_4\} {b1​,...,b4​}.Thus by $(1)$, there exist $t_i,s_i\in R$ ($1\le i\le 4$) such that
$$\begin{gathered} t_1{\mathbf{a}}_1+...+t_4{\mathbf{a}}_4=s_1{\mathbf{b}}_1+...+s_4{\mathbf{b}}_4=\mathbf{x}(3) \\ {t}_1{\mathbf{a}}_1+...+t_4{\mathbf{a}}_4-s_1{\mathbf{b}}_1-...-s_4{\mathbf{b}}_4=0(4) \end{gathered}$$

Let $A=\left[\begin{array}{cccccc}{\mathbf{a}}_1& ...& {\mathbf{a}}_4& {\mathbf{b}}_1& ...& {\mathbf{b}}_4\end{array}\right]$, then according to $(4)$:
$$\begin{gathered} A\left[\begin{array}{c}{t}_1\\ ...\\ t_4\\ -s_1\\ ...\\ -s_4\end{array}\right]=0 \\ \because 4\le rankA\le 6,dimNulA=8-rankA \\ \therefore 2\le dimNulA\le 4 \end{gathered}$$

According to $(3)$, it can be shown that the dimension of the solution set of $\mathbf{x}$ and $\left[\begin{array}{c}{t}_1\\ ...\\ t_4\end{array}\right]$ are equal. We can also observe that $\left[\begin{array}{c}{t}_1\\ ...\\ t_4\end{array}\right]\mapsto \left[\begin{array}{c}{s}_1\\ ...\\ s_4\end{array}\right]$ is a linear transformation, thus $dimNulA=$ the dimension of the solution set of $\left[\begin{array}{c}{t}_1\\ ...\\ t_4\end{array}\right]$. So $dim{F}_1\cap F_2=dimNulA$ and $2\le dim{F}_1\cap F_2\le 4$.