指示函数是将分段函数表达成一个式子的技术
指示函数及应用
集合
A
A
A的指示函数定义为
I
(
x
∈
A
)
=
I
A
(
x
)
=
{
1
,
当
x
∈
A
时
0
,
当
x
∉
A
时
\begin{align} \mathbb{I} (\boldsymbol{x} \in A) & =\mathbb{I}_A (\boldsymbol{x})\notag \\ & = \begin{cases} \, 1\, ,\qquad \text{当$\boldsymbol{x} \in A$时} \\ \, 0\, ,\qquad \text{当$\boldsymbol{x} \notin A$时} \end{cases} \tag{B1} \end{align}
I(x∈A)=IA(x)={1,当x∈A时0,当x∈/A时(B1)
显然有
I
A
(
x
)
+
I
A
‾
(
x
)
=
1
\mathbb{I}_A (\boldsymbol{x})+\mathbb{I}_{\overline{A}} (\boldsymbol{x})=1
IA(x)+IA(x)=1,其中,
A
‾
\overline{A}
A为集
A
A
A的补集。
我们看分段函数的指示函数表示,若
f
(
x
)
=
{
1
,
当
x
∈
A
时
−
1
,
当
x
∉
A
时
\begin{align*} f(\boldsymbol{x})= \begin{cases} \, 1\, ,\qquad \text{当$\boldsymbol{x} \in A$时} \\ \, -1\, ,\qquad \text{当$\boldsymbol{x} \notin A$时} \end{cases} \end{align*}
f(x)={1,当x∈A时−1,当x∈/A时
则
f
(
x
)
=
I
A
(
x
)
−
I
A
‾
=
I
A
(
x
)
−
(
1
−
I
A
)
=
2
I
A
(
x
)
−
1
\begin{align} f(\boldsymbol{x}) & =\mathbb{I}_A (\boldsymbol{x})-\mathbb{I}_{\overline{A}} \notag \\ & =\mathbb{I}_A (\boldsymbol{x})-(1-\mathbb{I}_A) \notag \\ & =2\mathbb{I}_A (\boldsymbol{x})-1 \tag{B2} \end{align}
f(x)=IA(x)−IA=IA(x)−(1−IA)=2IA(x)−1(B2)
同样地,若
f
(
x
)
=
{
a
(
x
)
,
当
x
∈
A
时
b
(
x
)
,
当
x
∉
A
时
\begin{align} f(\boldsymbol{x})= \begin{cases} \, a(\boldsymbol{x})\, ,\qquad \text{当$\boldsymbol{x} \in A$时} \\ \, b(\boldsymbol{x})\, ,\qquad \text{当$\boldsymbol{x} \notin A$时} \end{cases} \tag{B3} \end{align}
f(x)={a(x),当x∈A时b(x),当x∈/A时(B3)
则
f
(
x
)
=
a
(
x
)
I
A
(
x
)
+
b
(
x
)
I
A
‾
(
x
)
\begin{align} f(\boldsymbol{x}) & =a(\boldsymbol{x})\mathbb{I}_A (\boldsymbol{x})+b(\boldsymbol{x})\mathbb{I}_{\overline{A}} (\boldsymbol{x}) \tag{B4} \end{align}
f(x)=a(x)IA(x)+b(x)IA(x)(B4)
特别地,当
b
(
x
)
≡
0
b(\boldsymbol{x})\equiv 0
b(x)≡0时,有
f
(
x
)
=
a
(
x
)
I
A
(
x
)
\begin{align} f(\boldsymbol{x}) & =a(\boldsymbol{x})\mathbb{I}_A (\boldsymbol{x}) \tag{B5} \end{align}
f(x)=a(x)IA(x)(B5)
式(B4)是以和来表示,同样可以用积来表示
f
(
x
)
=
a
(
x
)
I
A
(
x
)
b
(
x
)
I
A
‾
(
x
)
\begin{align} f(\boldsymbol{x}) & =a(\boldsymbol{x})^{\mathbb{I}_A (\boldsymbol{x})}b(\boldsymbol{x})^{\mathbb{I}_{\overline{A}} (\boldsymbol{x})} \tag{B6} \end{align}
f(x)=a(x)IA(x)b(x)IA(x)(B6)
扩展到多段,第
i
i
i段的变量范围为
A
i
(
x
)
A_i(\boldsymbol{x})
Ai(x)、函数为
a
i
(
x
)
a_i(\boldsymbol{x})
ai(x),则
f
(
x
)
=
∑
i
=
1
n
a
i
(
x
)
I
[
A
i
(
x
)
]
f
(
x
)
=
∏
i
=
1
n
a
i
(
x
)
I
[
A
i
(
x
)
]
\begin{align} f(\boldsymbol{x}) & =\sum_{i=1}^na_i(\boldsymbol{x})\mathbb{I}[A_i(\boldsymbol{x})] \tag{B7} \\ f(\boldsymbol{x}) & =\mathop{\prod}\limits_{i=1}^na_i(\boldsymbol{x})^{\mathbb{I}[A_i(\boldsymbol{x})]} \tag{B8} \end{align}
f(x)f(x)=i=1∑nai(x)I[Ai(x)]=i=1∏nai(x)I[Ai(x)](B7)(B8)
特别地,集合
{
a
i
}
i
=
1
n
\{a_i\}_{i=1}^n
{ai}i=1n中任一元素
a
k
a_k
ak可以表示为全体元素的“和”(“积”)
a
k
=
∑
i
=
1
n
a
i
I
(
i
=
k
)
a
k
=
∏
i
=
1
n
a
i
I
(
i
=
k
)
\begin{align} a_k & =\sum_{i=1}^na_i\mathbb{I}(i=k) \tag{B9} \\ a_k & =\mathop{\prod}\limits_{i=1}^n{a_i}^{\mathbb{I}(i=k)} \tag{B10} \end{align}
akak=i=1∑naiI(i=k)=i=1∏naiI(i=k)(B9)(B10)
指示函数的数学期望
E
x
∈
D
I
A
(
x
)
=
∫
D
I
A
(
x
)
p
(
x
)
d
x
=
∫
A
I
A
(
x
)
p
(
x
)
d
x
+
∫
D
∖
A
I
A
(
x
)
p
(
x
)
d
x
=
∫
A
1
⋅
p
(
x
)
d
x
+
∫
D
∖
A
0
⋅
p
(
x
)
d
x
=
∫
A
p
(
x
)
d
x
=
P
(
x
∈
A
)
\begin{align} \mathop{\mathbb{E} }\limits_{x\in D}\mathbb{I}_A (x) & =\int_D \mathbb{I}_A (x)p(x)\,\mathrm{d}x\notag \\ & =\int_A \mathbb{I}_A (x)p(x)\,\mathrm{d}x+\int_{D\setminus A} \mathbb{I}_A (x)p(x)\,\mathrm{d}x\notag \\ & =\int_A 1\cdot p(x)\,\mathrm{d}x+\int_{D\setminus A} 0\cdot p(x)\,\mathrm{d}x\notag \\ & =\int_A p(x)\,\mathrm{d}x\notag \\ & =P(x\in A) \tag{B11} \end{align}
x∈DEIA(x)=∫DIA(x)p(x)dx=∫AIA(x)p(x)dx+∫D∖AIA(x)p(x)dx=∫A1⋅p(x)dx+∫D∖A0⋅p(x)dx=∫Ap(x)dx=P(x∈A)(B11)
由上述若干式子,即可理解【西瓜书式(8.9)】的推导过程。 如:
e
−
f
(
x
)
α
t
h
t
(
x
)
=
{
e
−
α
t
,
当
f
(
x
)
=
h
t
(
x
)
时
e
α
t
,
当
f
(
x
)
≠
h
t
(
x
)
时
=
e
−
α
t
I
(
f
(
x
)
=
h
t
(
x
)
)
+
e
α
t
I
(
f
(
x
)
≠
h
t
(
x
)
)
(由式(B4))
\begin{align*} \mathrm{e}^{-f(\boldsymbol{x}){\alpha }_th_t(\boldsymbol{x})} & = \begin{cases} \, \mathrm{e}^{-{\alpha }_t}\, ,\qquad \text{当$f(\boldsymbol{x})=h_t(\boldsymbol{x})$时}\notag \\ \, \mathrm{e}^{{\alpha }_t}\, ,\qquad \text{当$f(\boldsymbol{x})\neq h_t(\boldsymbol{x})$时}\notag \\ \end{cases} \\ & =\mathrm{e}^{-{\alpha }_t}\mathbb{I} (f(\boldsymbol{x})=h_t(\boldsymbol{x}))+\mathrm{e}^{{\alpha }_t}\mathbb{I} (f(\boldsymbol{x})\neq h_t(\boldsymbol{x}))\qquad \text{(由式(B4))} \end{align*}
e−f(x)αtht(x)={e−αt,当f(x)=ht(x)时eαt,当f(x)=ht(x)时=e−αtI(f(x)=ht(x))+eαtI(f(x)=ht(x))(由式(B4))
这即是【西瓜书式(8.9)】推导过程的第2个等号,由式(B11)即得第3个等号。
本文为原创,您可以:
- 点赞(支持博主)
- 收藏(待以后看)
- 转发(他考研或学习,正需要)
- 评论(或讨论)
- 引用(支持原创)
- 不侵权
上一篇:5、含矩阵的偏导数
下一篇:7、有趣的距离与范数