1.1 Show that ∥ ⋅ ∥ 1 2 \|\cdot\|_{\frac{1}{2}} ∥⋅∥21 is not a norm.
解:
x
=
(
1
0
)
,
y
=
(
1
1
)
\mathbf{x}=\begin{pmatrix} 1\\ 0\\ \end{pmatrix},\mathbf{y}=\begin{pmatrix} 1\\ 1\\ \end{pmatrix}
x=(10),y=(11)
∥
x
+
y
∥
1
2
=
(
2
+
1
)
2
=
3
+
2
2
\|\mathbf{x}+\mathbf{y}\|_\frac{1}{2}=(\sqrt{2}+1)^2=3+2\sqrt{2}
∥x+y∥21=(2+1)2=3+22
∥
x
∥
1
2
+
∥
y
∥
1
2
=
1
+
2
<
∥
x
+
y
∥
1
2
\|\mathbf{x}\|_\frac{1}{2}+\|\mathbf{y}\|_\frac{1}{2}=1+\sqrt{2}<\|\mathbf{x}+\mathbf{y}\|_\frac{1}{2}
∥x∥21+∥y∥21=1+2<∥x+y∥21
所以不是
1.2 Prove that for any
x
∈
R
n
\mathbf{x}\in\mathbb{R}^n
x∈Rn one has
∥
x
∥
∞
=
lim
p
→
∞
∥
x
∥
p
\|\mathbf{x}\|_{\infty}=\lim\limits_{p\to\infty}\|\mathbf{x}\|_{p}
∥x∥∞=p→∞lim∥x∥p
解:
lim
p
→
∞
∥
x
∥
p
=
lim
p
→
∞
∑
i
=
1
n
∣
x
i
∣
p
p
=
max
∣
x
i
∣
=
∥
x
∥
∞
\begin{aligned} &\quad \lim\limits_{p\to\infty}\|\mathbf{x}\|_{p}\\ &=\lim\limits_{p\to\infty}\sqrt[p]{\sum_{i=1}^{n}\left|x_i\right|^p}\\ &=\max \left|x_i\right|\\ &=\|\mathbf{x}\|_{\infty} \end{aligned}
p→∞lim∥x∥p=p→∞limpi=1∑n∣xi∣p=max∣xi∣=∥x∥∞
1.3 Show that for any
x
,
y
,
z
∈
R
n
\mathbf{x},\mathbf{y},\mathbf{z}\in\mathbb{R}^n
x,y,z∈Rn
∥
x
−
z
∥
≤
∥
x
−
y
∥
+
∥
y
−
z
∥
\|\mathbf{x}-\mathbf{z}\|\le \|\mathbf{x}-\mathbf{y}\|+\|\mathbf{y}-\mathbf{z}\|
∥x−z∥≤∥x−y∥+∥y−z∥
解:
∥
x
−
z
∥
=
∥
x
−
y
+
y
−
z
∥
≤
∥
x
−
y
∥
+
∥
y
−
z
∥
\begin{aligned} & \quad \|\mathbf{x}-\mathbf{z}\|\\ &=\|\mathbf{x}-\mathbf{y}+\mathbf{y}-\mathbf{z}\|\\ &\le \|\mathbf{x}-\mathbf{y}\|+\|\mathbf{y}-\mathbf{z}\| \end{aligned}
∥x−z∥=∥x−y+y−z∥≤∥x−y∥+∥y−z∥
1.4 Prove the Cauchy–Schwarz inequality (Lemma 1.5).Show that equality holds if and only if the vectors
x
\mathbf{x}
x and
y
\mathbf{y}
y are linearly dependent.
解.
∑
i
=
1
n
(
x
i
z
−
y
i
)
2
=
∑
i
=
1
n
(
x
i
2
z
2
−
2
x
i
y
i
z
+
y
i
2
)
=
(
∑
i
=
1
n
x
i
2
)
z
2
−
2
(
∑
i
=
1
n
x
i
y
i
)
z
+
∑
i
=
1
n
y
i
2
≥
0
\begin{aligned} &\quad \sum_{i=1}^{n}(x_iz-y_i)^2\\ &=\sum_{i=1}^{n}(x_i^2z^2-2x_iy_iz+y_i^2)\\ &=\left(\sum_{i=1}^{n}x_i^2\right) z^2-2\left(\sum_{i=1}^{n}x_iy_i\right) z+\sum_{i=1}^{n}y_i^2\\ &\ge0 \end{aligned}
i=1∑n(xiz−yi)2=i=1∑n(xi2z2−2xiyiz+yi2)=(i=1∑nxi2)z2−2(i=1∑nxiyi)z+i=1∑nyi2≥0
⇒
Δ
=
4
(
∑
i
=
1
n
x
i
y
i
)
2
−
4
(
∑
i
=
1
n
x
i
2
)
(
∑
i
=
1
n
y
i
2
)
≤
0
⇒
∣
x
T
y
∣
≤
∥
x
∥
∥
y
∥
\Rightarrow \Delta=4\left(\sum_{i=1}^{n}x_iy_i\right)^2-4\left(\sum_{i=1}^{n}x_i^2\right)\left(\sum_{i=1}^{n}y_i^2\right)\le0\\ \Rightarrow \left|\mathbf{x}^T\mathbf{y}\right|\le\|\mathbf{x}\|\|\mathbf{y}\|
⇒Δ=4(i=1∑nxiyi)2−4(i=1∑nxi2)(i=1∑nyi2)≤0⇒∣∣xTy∣∣≤∥x∥∥y∥
当且仅当
z
x
=
y
z\mathbf{x}=\mathbf{y}
zx=y时,
Δ
=
0
\Delta=0
Δ=0,
∣
x
T
y
∣
=
∥
x
∥
∥
y
∥
\left|\mathbf{x}^T\mathbf{y}\right|=\|\mathbf{x}\|\|\mathbf{y}\|
∣∣xTy∣∣=∥x∥∥y∥
1.5 Suppose that
R
m
\mathbb{R}^m
Rm and
R
n
\mathbb{R}^n
Rn are equipped with norms
∥
⋅
∥
b
\|\cdot \|_b
∥⋅∥b and
∥
⋅
∥
a
\|\cdot \|_a
∥⋅∥a, respectively.Show that the induced matrix norm
∥
⋅
∥
a
,
b
\|\cdot \|_{a,b}
∥⋅∥a,bsatisfies the triangle inequality. That is,for any
A
,
B
∈
R
m
×
n
\mathbf{A},\mathbf{B}\in\mathbb{R}^{m\times n}
A,B∈Rm×n the inequality
∥
A
+
B
∥
a
,
b
≤
∥
A
∥
a
,
b
+
∥
B
∥
a
,
b
\|\mathbf{A}+\mathbf{B}\|_{a,b}\le\|\mathbf{A}\|_{a,b}+\|\mathbf{B}\|_{a,b}
∥A+B∥a,b≤∥A∥a,b+∥B∥a,b
holds.
解:
∥
A
+
B
∥
a
,
b
=
max
∥
x
∥
a
≤
1
∥
(
A
+
B
)
x
∥
b
∥
x
∥
a
≤
max
∥
x
∥
a
≤
1
∥
A
x
∥
b
+
∥
B
x
∥
b
∥
x
∥
a
≤
max
∥
x
∥
a
≤
1
∥
A
x
∥
b
∥
x
∥
a
+
max
∥
y
∥
a
≤
1
∥
B
y
∥
b
∥
y
∥
a
=
∥
A
∥
a
,
b
+
∥
B
∥
a
,
b
\begin{aligned} &\quad \|\mathbf{A}+\mathbf{B}\|_{a,b}\\ &=\max_{\|\mathbf{x}\|_a\le 1}\frac{\|\left(\mathbf{A}+\mathbf{B}\right)\mathbf{x}\|_b}{\|\mathbf{x}\|_a}\\ &\le \max_{\|\mathbf{x}\|_a\le 1}\frac{\|\mathbf{A}\mathbf{x}\|_b+\|\mathbf{B}\mathbf{x}\|_b}{\|\mathbf{x}\|_a}\\ &\le \max_{\|\mathbf{x}\|_a\le 1}\frac{\|\mathbf{A}\mathbf{x}\|_b}{\|\mathbf{x}\|_a}+\max_{\|\mathbf{y}\|_a\le 1}\frac{\|\mathbf{B}\mathbf{y}\|_b}{\|\mathbf{y}\|_a}\\ &=\|\mathbf{A}\|_{a,b}+\|\mathbf{B}\|_{a,b} \end{aligned}\\
∥A+B∥a,b=∥x∥a≤1max∥x∥a∥(A+B)x∥b≤∥x∥a≤1max∥x∥a∥Ax∥b+∥Bx∥b≤∥x∥a≤1max∥x∥a∥Ax∥b+∥y∥a≤1max∥y∥a∥By∥b=∥A∥a,b+∥B∥a,b
1.6 Let ∥ ⋅ ∥ \|\cdot \| ∥⋅∥ be a norm on R n \mathbb{R}^n Rn.Show that the norm function f ( x ) = ∥ x ∥ f\left(\mathbf{x}\right)=\|\mathbf{x}\| f(x)=∥x∥ is a continuous function over R n \mathbb{R}^n Rn.
解:
对于
∀
ϵ
>
0
,
∃
δ
>
0
\forall \epsilon>0,\exists \delta>0
∀ϵ>0,∃δ>0,
当
0
<
∥
x
−
y
∥
<
δ
0<\|\mathbf{x}-\mathbf{y}\|<\delta
0<∥x−y∥<δ时
∣
f
(
x
)
−
f
(
y
)
∣
=
∣
∥
x
∥
−
∥
y
∥
∣
≤
∥
x
−
y
∥
<
ϵ
\left|f\left(\mathbf{x}\right)-f\left(\mathbf{y}\right)\right|=\left|\|\mathbf{x}\|-\|\mathbf{y}\|\right|\le\|\mathbf{x}-\mathbf{y}\|<\epsilon
∣f(x)−f(y)∣=∣∥x∥−∥y∥∣≤∥x−y∥<ϵ
取 δ = ϵ \delta=\epsilon δ=ϵ即可
1.7 (attainment of the maximum in the induced norm definition). Suppose that R m \mathbb{R}^m Rm and R n \mathbb{R}^n Rn are equipped with norms ∥ ⋅ ∥ b \|\cdot\|_b ∥⋅∥b and ∥ ⋅ ∥ a \|\cdot\|_a ∥⋅∥a, respectively, and let A ∈ R m × n \mathbf{A}\in\mathbb{R}^{m\times n} A∈Rm×n. Show that there exists x ∈ R n \mathbf{x}\in\mathbb{R}^n x∈Rn such that ∥ x ∥ a ≤ 1 \|\mathbf{x}\|_a\le 1 ∥x∥a≤1 and ∥ A x ∥ b = ∥ A ∥ a , b \|\mathbf{Ax}\|_b=\|\mathbf{A}\|_{a,b} ∥Ax∥b=∥A∥a,b.
解:
连续函数在紧集中可以取到最大值,所以成立
1.8 Suppose that
R
m
\mathbb{R}^m
Rm and
R
n
\mathbb{R}^n
Rn are equipped with norms
∥
⋅
∥
b
\|\cdot \|_b
∥⋅∥b and
∥
⋅
∥
a
\|\cdot \|_a
∥⋅∥a,respectively.Show that the induced matrix norm
∥
⋅
∥
a
,
b
\|\cdot\|_{a,b}
∥⋅∥a,b can be computed by the formula
∥
A
∥
a
,
b
=
max
x
{
∥
A
x
∥
b
:
∥
x
∥
a
=
1
}
\|\mathbf{A}\|_{a,b}=\max_{\mathbf{x}}\left\{\|\mathbf{Ax}\|_b: \|\mathbf{x}\|_a=1\right\}
∥A∥a,b=xmax{∥Ax∥b:∥x∥a=1}
解:
∥
A
∥
a
,
b
=
max
∥
x
∥
a
≤
1
∥
A
x
∥
b
∥
x
∥
a
=
max
∥
x
∥
a
≤
1
∥
A
x
∥
x
∥
a
∥
b
=
max
∥
y
∥
a
=
1
∥
A
y
∥
b
\begin{aligned} &\quad \|\mathbf{A}\|_{a,b}\\ &=\max_{\|\mathbf{x}\|_a\le 1}\frac{\|\mathbf{Ax}\|_b}{\|\mathbf{x}\|_a}\\ &=\max_{\|\mathbf{x}\|_a\le 1}\|\mathbf{A}\frac{\mathbf{x}}{\|\mathbf{x}\|_a}\|_b\\ &=\max_{\|\mathbf{y}\|_a= 1}\|\mathbf{A}\mathbf{y}\|_b\\ \end{aligned}
∥A∥a,b=∥x∥a≤1max∥x∥a∥Ax∥b=∥x∥a≤1max∥A∥x∥ax∥b=∥y∥a=1max∥Ay∥b
1.9 Suppose that
R
m
\mathbb{R}^m
Rm and
R
n
\mathbb{R}^n
Rn are equipped with norms
∥
⋅
∥
b
\|\cdot \|_b
∥⋅∥b and
∥
⋅
∥
a
\|\cdot \|_a
∥⋅∥a,respectively.Show that the induced matrix norm
∥
⋅
∥
a
,
b
\|\cdot\|_{a,b}
∥⋅∥a,b can be computed by the formula
∥
A
∥
a
,
b
=
max
x
≠
0
∥
A
x
∥
b
∥
x
∥
a
\|\mathbf{A}\|_{a,b}=\max_{\mathbf{x}\neq 0}\frac{\|\mathbf{Ax}\|_b}{\|\mathbf{x}\|_a}
∥A∥a,b=x=0max∥x∥a∥Ax∥b
解:
max
x
≠
0
∥
A
x
∥
b
∥
x
∥
a
=
max
x
≠
0
∥
A
x
∥
x
∥
a
∥
b
=
max
∥
y
∥
a
=
1
∥
A
y
∥
b
=
∥
A
∥
a
,
b
\begin{aligned} &\quad \max_{\mathbf{x}\neq 0}\frac{\|\mathbf{Ax}\|_b}{\|\mathbf{x}\|_a}\\ &=\max_{\mathbf{x}\neq 0}\|\mathbf{A}\frac{\mathbf{x}}{\|\mathbf{x}\|_a}\|_b\\ &=\max_{\|\mathbf{y}\|_a= 1}\|\mathbf{A}\mathbf{y}\|_b\\ &=\|\mathbf{A}\|_{a,b} \end{aligned}
x=0max∥x∥a∥Ax∥b=x=0max∥A∥x∥ax∥b=∥y∥a=1max∥Ay∥b=∥A∥a,b
1.10 Let
A
∈
R
m
×
n
,
B
∈
R
n
×
k
\mathbf{A}\in\mathbb{R}^{m\times n},\mathbf{B}\in\mathbb{R}^{n\times k}
A∈Rm×n,B∈Rn×k and assume that
R
m
,
R
n
\mathbb{R}^m,\mathbb{R}^n
Rm,Rn, and
R
k
\mathbb{R}^k
Rk are equipped with the norms
∥
⋅
∥
c
,
∥
⋅
∥
b
\|\cdot\|_c,\|\cdot\|_b
∥⋅∥c,∥⋅∥b, and
∥
⋅
∥
a
\|\cdot\|_a
∥⋅∥a,respectively. Prove that
∥
A
B
∥
a
,
c
≤
∥
A
∥
b
,
c
∥
B
∥
a
,
b
\|\mathbf{AB}\|_{a,c}\le \|\mathbf{A}\|_{b,c}\|\mathbf{B}\|_{a,b}
∥AB∥a,c≤∥A∥b,c∥B∥a,b
解:
∥
A
B
∥
a
,
c
=
max
∥
x
∥
a
≤
1
∥
A
B
x
∥
c
∥
x
∥
a
≤
max
∥
x
∥
a
≤
1
∥
A
∥
b
,
c
∥
B
x
∥
b
∥
x
∥
a
=
∥
A
∥
b
,
c
∥
B
∥
a
,
b
\begin{aligned} &\quad \|\mathbf{AB}\|_{a,c}\\ &=\max_{\|\mathbf{x}\|_a\le 1}\frac{\|\mathbf{ABx}\|_{c}}{\|\mathbf{x}\|_{a}}\\ &\le\max_{\|\mathbf{x}\|_a\le 1}\frac{\|\mathbf{A}\|_{b,c}\|\mathbf{Bx}\|_{b}}{\|\mathbf{x}\|_{a}}\\ &=\|\mathbf{A}\|_{b,c}\|\mathbf{B}\|_{a,b} \end{aligned}
∥AB∥a,c=∥x∥a≤1max∥x∥a∥ABx∥c≤∥x∥a≤1max∥x∥a∥A∥b,c∥Bx∥b=∥A∥b,c∥B∥a,b
1.11 Prove the formula of the ∞ \infty ∞-matrix norm given in Example 1.9.
解:
∥
A
∥
∞
=
max
∥
x
∥
∞
=
1
∥
A
x
∥
∞
=
max
∥
x
∥
∞
=
1
max
i
∣
∑
j
=
1
n
a
i
j
x
j
∣
≤
max
∥
x
∥
∞
=
1
max
i
∑
j
=
1
n
∣
a
i
j
x
j
∣
≤
max
i
∑
j
=
1
n
∣
a
i
j
∣
\begin{aligned} &\quad \|\mathbf{A}\|_\infty\\ &=\max_{\|\mathbf{x}\|_\infty=1}\|\mathbf{Ax}\|_\infty\\ &=\max_{\|\mathbf{x}\|_\infty=1}\max_{i}\left|\sum_{j=1}^{n}a_{ij}x_{j}\right|\\ &\le\max_{\|\mathbf{x}\|_\infty=1}\max_{i}\sum_{j=1}^{n}\left|a_{ij}x_{j}\right|\\ &\le\max_{i}\sum_{j=1}^{n}\left|a_{ij}\right| \end{aligned}
∥A∥∞=∥x∥∞=1max∥Ax∥∞=∥x∥∞=1maximax∣∣∣∣∣j=1∑naijxj∣∣∣∣∣≤∥x∥∞=1maximaxj=1∑n∣aijxj∣≤imaxj=1∑n∣aij∣
设
x
0
=
(
x
1
,
⋯
,
x
n
)
\mathbf{x_0}=\left(x_1,\cdots,x_n\right)
x0=(x1,⋯,xn)
其中
x
j
=
{
∣
a
i
j
∣
a
i
j
,
a
i
j
≠
0
1
,
a
i
j
=
0
x_j=\begin{cases} \frac{\left|a_{ij}\right|}{a_{ij}},&a_{ij}\neq 0\\ 1,&a_{ij}=0 \end{cases}
xj={aij∣aij∣,1,aij=0aij=0
∥
A
x
0
∥
∞
=
max
i
∑
j
=
1
n
∣
a
i
j
∣
\|\mathbf{Ax_0}\|_\infty=\max_{i}\sum_{j=1}^{n}\left|a_{ij}\right|
∥Ax0∥∞=imaxj=1∑n∣aij∣
1.12 Let
A
∈
R
m
×
n
\mathbf{A}\in\mathbb{R}^{m\times n}
A∈Rm×n.Prove that
(i)
1
n
∥
A
∥
∞
≤
∥
A
∥
2
≤
m
∥
A
∥
∞
\frac{1}{\sqrt{n}}\|\mathbf{A}\|_\infty\le\|\mathbf{A}\|_2\le\sqrt{m}\|\mathbf{A}\|_\infty
n1∥A∥∞≤∥A∥2≤m∥A∥∞
(ii)
1
m
∥
A
∥
1
≤
∥
A
∥
2
≤
n
∥
A
∥
1
\frac{1}{\sqrt{m}}\|\mathbf{A}\|_1\le\|\mathbf{A}\|_2\le\sqrt{n}\|\mathbf{A}\|_1
m1∥A∥1≤∥A∥2≤n∥A∥1
解:
(i)
利用
∥
x
∥
∞
≤
∥
x
∥
2
≤
n
∥
x
∥
∞
\|\mathbf{x}\|_\infty\le\|\mathbf{x}\|_2\le \sqrt{n}\|\mathbf{x}\|_\infty
∥x∥∞≤∥x∥2≤n∥x∥∞
(可以用 最大<=求和<=n*最大来证明)
于是
1
n
∥
A
∥
∞
=
1
n
max
x
≠
0
∥
A
x
∥
∞
∥
x
∥
∞
≤
max
x
≠
0
∥
A
x
∥
2
∥
x
∥
2
=
∥
x
∥
2
≤
max
x
≠
0
m
∥
A
x
∥
∞
∥
x
∥
∞
=
m
∥
A
∥
∞
\begin{aligned} &\quad \frac{1}{\sqrt{n}}\|\mathbf{A}\|_\infty\\ &=\frac{1}{\sqrt{n}}\max_{\mathbf{x}\neq0}\frac{\|\mathbf{Ax}\|_\infty}{\|\mathbf{x}\|_\infty}\\ &\le \max_{\mathbf{x}\neq0}\frac{\|\mathbf{Ax}\|_2}{\|\mathbf{x}\|_2}=\|\mathbf{x}\|_2\\ &\le \max_{\mathbf{x}\neq0}\frac{\sqrt{m}\|\mathbf{Ax}\|_\infty}{\|\mathbf{x}\|_\infty}\\ &=\sqrt{m}\|\mathbf{A}\|_\infty \end{aligned}
n1∥A∥∞=n1x=0max∥x∥∞∥Ax∥∞≤x=0max∥x∥2∥Ax∥2=∥x∥2≤x=0max∥x∥∞m∥Ax∥∞=m∥A∥∞
(ii)
利用
1
m
∥
x
∥
1
≤
∥
x
∥
2
≤
∥
x
∥
1
\frac{1}{\sqrt{m}}\|\mathbf{x}\|_1\le\|\mathbf{x}\|_2\le \|\mathbf{x}\|_1
m1∥x∥1≤∥x∥2≤∥x∥1
(左边用算术平均<=平方平均,右边可以数学归纳)
于是
1
m
∥
A
∥
1
=
1
m
max
x
≠
0
∥
A
x
∥
1
∥
x
∥
1
≤
max
x
≠
0
∥
A
x
∥
2
∥
x
∥
2
=
∥
x
∥
2
≤
max
x
≠
0
∥
A
x
∥
1
∥
x
∥
1
n
=
n
∥
A
∥
1
\begin{aligned} &\quad \frac{1}{\sqrt{m}}\|\mathbf{A}\|_1\\ &=\frac{1}{\sqrt{m}}\max_{\mathbf{x}\neq0}\frac{\|\mathbf{Ax}\|_1}{\|\mathbf{x}\|_1}\\ &\le \max_{\mathbf{x}\neq0}\frac{\|\mathbf{Ax}\|_2}{\|\mathbf{x}\|_2}=\|\mathbf{x}\|_2\\ &\le \max_{\mathbf{x}\neq0}\frac{\|\mathbf{Ax}\|_1}{\frac{\|\mathbf{x}\|_1}{\sqrt{n}}}\\ &=\sqrt{n}\|\mathbf{A}\|_1 \end{aligned}
m1∥A∥1=m1x=0max∥x∥1∥Ax∥1≤x=0max∥x∥2∥Ax∥2=∥x∥2≤x=0maxn∥x∥1∥Ax∥1=n∥A∥1
1.13 Let
A
∈
R
m
×
n
\mathbf{A}\in\mathbb{R}^{m\times n}
A∈Rm×n Show that
(i)
∥
A
∥
=
∥
A
T
∥
\|\mathbf{A}\|=\|\mathbf{A}^T\|
∥A∥=∥AT∥(here
∥
⋅
∥
\|\cdot\|
∥⋅∥ is the spectral norm),
(ii)
∥
A
∥
F
2
=
∑
i
=
1
n
λ
i
(
A
T
A
)
\|\mathbf{A}\|_F^2=\sum_{i=1}^{n}\lambda_i\left(\mathbf{A}^T\mathbf{A}\right)
∥A∥F2=∑i=1nλi(ATA)
解:
(i)
(
I
B
0
I
)
(
0
0
A
A
B
)
=
(
B
A
0
A
0
)
(
I
B
0
I
)
\begin{pmatrix} I&B\\ 0&I \end{pmatrix}\begin{pmatrix} 0&0\\ A&AB \end{pmatrix}=\begin{pmatrix} BA&0\\ A&0 \end{pmatrix}\begin{pmatrix} I&B\\ 0&I \end{pmatrix}
(I0BI)(0A0AB)=(BAA00)(I0BI)
所以
(
0
0
A
A
B
)
∼
(
B
A
0
A
0
)
\begin{pmatrix} 0&0\\ A&AB \end{pmatrix}\sim \begin{pmatrix} BA&0\\ A&0 \end{pmatrix}
(0A0AB)∼(BAA00)
因为
(
0
0
A
A
B
)
\begin{pmatrix} 0&0\\ A&AB \end{pmatrix}
(0A0AB)的特征值为
A
B
AB
AB的特征值,以及几个0
(
B
A
0
A
0
)
\begin{pmatrix} BA&0\\ A&0 \end{pmatrix}
(BAA00)的特征值为
B
A
BA
BA的特征值,以及几个0
所以
A
B
AB
AB非零特征值与
B
A
BA
BA非零特征值相同
所以
A
A
T
\mathbf{A}\mathbf{A}^T
AAT的非零特征值与
A
T
A
\mathbf{A}^T\mathbf{A}
ATA的非零特征值相同
∥
A
∥
2
=
λ
1
(
A
T
A
)
=
λ
1
(
A
A
T
)
=
∥
A
T
∥
2
\|\mathbf{A}\|_2=\sqrt{\lambda_1\left(\mathbf{A}^T\mathbf{A}\right)}=\sqrt{\lambda_1\left(\mathbf{A}\mathbf{A}^T\right)}=\|\mathbf{A}^T\|_2
∥A∥2=λ1(ATA)=λ1(AAT)=∥AT∥2
(ii)
∥
A
∥
F
2
=
∑
i
=
1
m
∑
j
=
1
n
a
i
j
2
=
t
r
(
A
T
A
)
=
∑
i
=
1
n
λ
i
(
A
T
A
)
\|\mathbf{A}\|_F^2=\sum_{i=1}^{m}\sum_{j=1}^{n}a_{ij}^2=tr\left(\mathbf{A}^T\mathbf{A}\right)=\sum_{i=1}^{n}\lambda_i\left(\mathbf{A}^T\mathbf{A}\right)
∥A∥F2=i=1∑mj=1∑naij2=tr(ATA)=i=1∑nλi(ATA)
1.14 Let
A
∈
R
n
×
n
\mathbf{A}\in\mathbb{R}^{n\times n}
A∈Rn×n be a symmetric matrix.Show that
max
x
{
x
T
A
x
:
∥
x
∥
2
=
1
}
=
λ
m
a
x
(
A
)
\max_{\mathbf{x}}\left\{\mathbf{x}^T\mathbf{Ax}:\|\mathbf{x}\|^2=1\right\}=\lambda_{max}\left(\mathbf{A}\right)
xmax{xTAx:∥x∥2=1}=λmax(A)
解:
由瑞利商显然成立
1.15 Prove that a set U ⊆ R n U\subseteq \mathbb{R}^n U⊆Rn is closed if and only if its complement U c U^c Uc is open.
解:
⇐
\Leftarrow
⇐ 假设
U
U
U是闭的,取
x
∈
U
c
x\in U^c
x∈Uc,有
x
∉
U
x\notin U
x∈/U,所以
x
x
x不是
U
U
U的极限点。
因此存在
x
x
x的一个邻域
N
N
N,使得
U
∩
N
=
∅
U\cap N=\empty
U∩N=∅,
于是
N
⊂
U
c
N\subset U^c
N⊂Uc,即
x
x
x是
U
c
U^c
Uc的内点
于是
U
c
U^c
Uc是开的
⇒ \Rightarrow ⇒假设 U c U^c Uc是开的,假设 x x x是 U U U的一个极限点,那么 x x x的每一个邻域都包含 U U U的点,所以 x x x不是 U c U^c Uc的内点,所以 U U U是开的
1.16
(i)Let
{
A
i
}
i
∈
I
\left\{A_i\right\}_{i\in I}
{Ai}i∈I be a collection of open sets where
I
I
I is a given index set. Show that
∪
i
∈
I
A
i
\cup_{i\in I} A_i
∪i∈IAi is an open set. Show that if
I
I
I is finite,then
∪
i
∈
I
A
i
\cup_{i\in I} A_i
∪i∈IAi is open.
(ii)Let
{
A
i
}
i
∈
I
\left\{A_i\right\}_{i\in I}
{Ai}i∈I be a collection of closed sets where
I
I
I is a given index set. Show that
∩
i
∈
I
A
i
\cap_{i\in I} A_i
∩i∈IAi is an closed set. Show that if
I
I
I is finite,then
∩
i
∈
I
A
i
\cap_{i\in I} A_i
∩i∈IAi is closed.
解:
(i)
因为开集的所有点都是内点
取并集后,所有点依然是内点,所以成立
(ii)
取补集,然后由(i)和1.15题,成立
1.17 Give an example of open sets A i , i ∈ I A_i,i\in I Ai,i∈I for which ∩ i ∈ I A i \cap_{i\in I} A_i ∩i∈IAi is not open
解:
∩
i
=
1
∞
(
−
1
i
,
1
i
)
=
{
0
}
\cap_{i=1}^{\infty}\left(-\frac{1}{i},\frac{1}{i}\right)=\left\{0\right\}
∩i=1∞(−i1,i1)={0}
1.18 Let A , B ⊆ R n A,B\subseteq \mathbb{R}^n A,B⊆Rn.Prove that c l ( A ∩ B ) ⊆ c l ( A ) ∩ c l ( B ) cl(A\cap B)\subseteq cl(A)\cap cl(B) cl(A∩B)⊆cl(A)∩cl(B).Give an example in which the inclusion is proper.
解:
A
⊆
c
l
(
A
)
A\subseteq cl(A)
A⊆cl(A)
B
⊆
c
l
(
B
)
B\subseteq cl(B)
B⊆cl(B)
所以
A
∩
B
⊆
c
l
(
A
)
∩
c
l
(
B
)
A\cap B \subseteq cl(A)\cap cl(B)
A∩B⊆cl(A)∩cl(B)
因为
c
l
(
A
∩
B
)
cl(A\cap B)
cl(A∩B)是包含
A
∩
B
A\cap B
A∩B的所有闭包中最小的
且
c
l
(
A
)
∩
c
l
(
B
)
cl(A)\cap cl(B)
cl(A)∩cl(B)是闭集
所以
c
l
(
A
∩
B
)
⊆
c
l
(
A
)
∩
c
l
(
B
)
cl(A\cap B)\subseteq cl(A)\cap cl(B)
cl(A∩B)⊆cl(A)∩cl(B)
A
=
(
0
,
1
)
A=\left(0,1\right)
A=(0,1)
B
=
(
1
,
2
)
B=\left(1,2\right)
B=(1,2)
c
l
(
A
∩
B
)
=
∅
cl(A\cap B)=\empty
cl(A∩B)=∅
c
l
(
A
)
∩
c
l
(
B
)
=
{
1
}
cl(A)\cap cl(B)=\left\{1\right\}
cl(A)∩cl(B)={1}
c
l
(
A
∩
B
)
⊂
c
l
(
A
)
∩
c
l
(
B
)
cl(A\cap B)\subset cl(A)\cap cl(B)
cl(A∩B)⊂cl(A)∩cl(B)
1.19 Let A , B ⊆ R n A,B\subseteq \mathbb{R}^n A,B⊆Rn.Prove that i n t ( A ∩ B ) = i n t ( A ) ∩ i n t ( B ) int(A\cap B)=int(A)\cap int(B) int(A∩B)=int(A)∩int(B) and that i n t ( A ) ∪ i n t ( B ) ⊆ i n t ( A ∪ B ) int(A)\cup int(B) \subseteq int(A\cup B) int(A)∪int(B)⊆int(A∪B).Show an example in which the latter inclusion is proper.
解:
设
x
∈
i
n
t
(
A
∩
B
)
x \in int(A\cap B)
x∈int(A∩B),
于是
∃
r
>
0
,
B
(
x
,
r
)
⊂
A
∩
B
\exists r>0,B(x,r)\subset A\cap B
∃r>0,B(x,r)⊂A∩B
所以
B
(
x
,
r
)
⊂
A
,
B
(
x
,
r
)
⊂
B
⇒
x
∈
i
n
t
(
A
)
,
x
∈
i
n
t
(
B
)
⇒
x
∈
i
n
t
(
A
)
∩
i
n
t
(
B
)
B(x,r)\subset A,B(x,r)\subset B\Rightarrow x\in int(A),x\in int(B)\Rightarrow x\in int(A)\cap int(B)
B(x,r)⊂A,B(x,r)⊂B⇒x∈int(A),x∈int(B)⇒x∈int(A)∩int(B)
所以
i
n
t
(
A
∩
B
)
⊆
i
n
t
(
A
)
∩
i
n
t
(
B
)
int(A\cap B) \subseteq int(A)\cap int(B)
int(A∩B)⊆int(A)∩int(B)
设
x
∈
i
n
t
(
A
)
∩
i
n
t
(
B
)
x\in int(A)\cap int(B)
x∈int(A)∩int(B)
假设
∃
r
1
>
0
,
r
2
>
0
,
s
.
t
.
B
(
x
,
r
1
)
⊂
A
,
B
(
x
,
r
2
)
⊂
B
\exists r_1>0,r_2>0,\ s.t.\ B(x,r_1)\subset A,B(x,r_2)\subset B
∃r1>0,r2>0, s.t. B(x,r1)⊂A,B(x,r2)⊂B
取
r
=
min
{
r
1
,
r
2
}
,
B
(
x
,
r
)
⊂
A
,
B
(
x
,
r
)
⊂
B
r=\min\left\{r_1,r_2\right\},B(x,r)\subset A,B(x,r)\subset B
r=min{r1,r2},B(x,r)⊂A,B(x,r)⊂B
于是
B
(
x
,
r
)
⊂
A
∩
B
B(x,r)\subset A\cap B
B(x,r)⊂A∩B
所以
x
∈
i
n
t
(
A
∩
B
)
x\in int(A\cap B)
x∈int(A∩B)
所以
i
n
t
(
A
)
∩
i
n
t
(
B
)
⊆
i
n
t
(
A
∩
B
)
int(A)\cap int(B) \subseteq int(A\cap B)
int(A)∩int(B)⊆int(A∩B)
⇒ i n t ( A ∩ B ) = i n t ( A ) ∩ i n t ( B ) \Rightarrow int(A\cap B)=int(A)\cap int(B) ⇒int(A∩B)=int(A)∩int(B)
设
x
∈
i
n
t
(
A
)
x\in int(A)
x∈int(A)
则
∃
r
>
0
,
B
(
x
,
r
)
⊂
A
⊂
A
∪
B
\exists r>0,B(x,r)\subset A \subset A\cup B
∃r>0,B(x,r)⊂A⊂A∪B
所以
x
∈
i
n
t
(
A
∪
B
)
x\in int(A\cup B)
x∈int(A∪B)
所以
i
n
t
(
A
)
⊆
i
n
t
(
A
∪
B
)
int(A)\subseteq int(A\cup B)
int(A)⊆int(A∪B)
设
x
∈
i
n
t
(
B
)
x\in int(B)
x∈int(B)
同理
i
n
t
(
B
)
⊆
i
n
t
(
A
∪
B
)
int(B)\subseteq int(A\cup B)
int(B)⊆int(A∪B)
⇒
i
n
t
(
A
)
∪
i
n
t
(
B
)
⊆
i
n
t
(
A
∪
B
)
\Rightarrow int(A)\cup int(B) \subseteq int(A\cup B)
⇒int(A)∪int(B)⊆int(A∪B)
设
A
=
(
−
∞
,
0
]
,
B
=
[
0
,
+
∞
)
A=(-\infty,0],B=[0,+\infty)
A=(−∞,0],B=[0,+∞)
i
n
t
(
A
)
∪
i
n
t
(
B
)
=
R
\
{
0
}
int(A)\cup int(B)=\mathbb{R}\backslash\left\{0\right\}
int(A)∪int(B)=R\{0}
i
n
t
(
A
∪
B
)
=
R
int(A\cup B)=\mathbb{R}
int(A∪B)=R
所以
⇒
i
n
t
(
A
)
∪
i
n
t
(
B
)
⊂
i
n
t
(
A
∪
B
)
\Rightarrow int(A)\cup int(B) \subset int(A\cup B)
⇒int(A)∪int(B)⊂int(A∪B)