1.設
X
1
,
⋯
,
X
n
{\displaystyle X_{1},\cdots ,X_{n}}
是數學期望值為0的獨立的隨機變數。若對所有
i
{\displaystyle i}
,
|
X
i
|
≤
M
{\displaystyle |X_{i}|\leq M}
幾乎必然成立,則對任意正數
t
{\displaystyle t}
P
(
∑
i
=
1
n
X
i
>
t
)
≤
exp
(
−
t
2
/
2
∑
E
X
j
2
+
M
t
/
3
)
{\displaystyle \mathbb {P} \left(\sum _{i=1}^{n}{X_{i}}>t\right)\leq \exp {\left(-{\frac {t^{2}/2}{\sum {\mathbb {E} X_{j}^{2}}+Mt/3}}\right)}}
2.設
X
1
,
⋯
,
X
n
{\displaystyle X_{1},\cdots ,X_{n}}
是獨立的隨機變數。若存在正實數
L
{\displaystyle L}
,使得對任意整數
k
>
1
{\displaystyle k>1}
,都有
E
|
X
i
k
|
≤
1
2
k
!
L
k
−
2
E
X
i
2
{\displaystyle \mathbb {E} |X_{i}^{k}|\leq {\frac {1}{2}}k!L^{k-2}\mathbb {E} X_{i}^{2}}
,則對
0
<
t
<
1
2
L
∑
E
X
j
2
{\displaystyle 0<t<{\frac {1}{2L}}{\sqrt {\sum {\mathbb {E} X_{j}^{2}}}}}
P
(
∑
i
=
1
n
X
i
≥
2
t
∑
E
X
i
2
)
<
exp
(
−
t
2
)
{\displaystyle \mathbb {P} \left(\sum _{i=1}^{n}{X_{i}}\geq 2t{\sqrt {\sum {\mathbb {E} X_{i}^{2}}}}\right)<\exp {(-t^{2})}}
3.設
X
1
,
⋯
,
X
n
{\displaystyle X_{1},\cdots ,X_{n}}
是獨立的隨機變數。若對任意整數
k
≥
4
{\displaystyle k\geq 4}
,都有
E
|
X
i
k
|
≤
k
!
4
!
(
L
5
)
k
−
4
{\displaystyle \mathbb {E} |X_{i}^{k}|\leq {\frac {k!}{4!}}\left({\frac {L}{5}}\right)^{k-4}}
,記
A
k
=
∑
i
=
1
n
E
X
i
k
{\displaystyle A_{k}=\sum _{i=1}^{n}{\mathbb {E} X_{i}^{k}}}
,則對於
0
<
t
≤
5
2
A
2
4
L
{\displaystyle 0<t\leq {\frac {5{\sqrt {2A_{2}}}}{4L}}}
P
(
|
∑
j
=
1
n
X
j
−
A
3
t
2
3
A
2
|
≥
2
A
2
t
(
1
+
A
4
t
2
6
A
2
2
)
)
<
2
exp
(
−
t
2
)
{\displaystyle \mathbb {P} \left(\left|\sum _{j=1}^{n}{X_{j}}-{\frac {A_{3}t^{2}}{3A_{2}}}\right|\geq {\sqrt {2A_{2}}}t\left(1+{\frac {A_{4}t^{2}}{6A_{2}^{2}}}\right)\right)<2\exp {(-t^{2})}}
4.伯恩施坦也把以上不等式推廣到弱相關隨機變數的情況。例如,不等式(2)可以推廣成以下形式。
X
1
,
⋯
,
X
n
{\displaystyle X_{1},\cdots ,X_{n}}
可以不是獨立隨機變數。若對任意正整數
i
{\displaystyle i}
,
E
(
X
i
|
X
1
,
⋯
,
X
i
−
1
)
=
0
,
{\displaystyle \mathbb {E} (X_{i}|X_{1},\cdots ,X_{i-1})=0,}
E
(
X
i
2
|
X
1
,
⋯
,
X
i
−
1
)
≤
R
i
E
X
i
2
,
{\displaystyle \mathbb {E} (X_{i}^{2}|X_{1},\cdots ,X_{i-1})\leq R_{i}\mathbb {E} X_{i}^{2},}
E
(
X
i
k
|
X
1
,
⋯
,
X
i
−
1
)
≤
k
!
L
k
−
2
2
E
(
X
i
2
|
X
1
,
⋯
,
X
i
−
1
)
,
{\displaystyle \mathbb {E} (X_{i}^{k}|X_{1},\cdots ,X_{i-1})\leq {\frac {k!L^{k-2}}{2}}\mathbb {E} (X_{i}^{2}|X_{1},\cdots ,X_{i-1}),}
則對於
0
<
t
<
1
2
L
∑
i
=
1
n
R
i
E
X
i
2
{\displaystyle 0<t<{\frac {1}{2L}}{\sqrt {\sum _{i=1}^{n}{R_{i}\mathbb {E} X_{i}^{2}}}}}
,
P
(
∑
i
=
1
n
X
i
≥
2
t
∑
i
=
1
n
R
i
E
X
i
2
)
<
exp
(
−
t
2
)
{\displaystyle \mathbb {P} \left(\sum _{i=1}^{n}{X_{i}}\geq 2t{\sqrt {\sum _{i=1}^{n}{R_{i}\mathbb {E} X_{i}^{2}}}}\right)<\exp {(-t^{2})}}