2.1 信息速率是指平均每秒传输的信息量
点和划出现的信息量分别为
,
log
log,
3
3
2
一秒钟点和划出现的次数平均为
一秒钟点和划分别出现的次数平均为
1
22.0
3
10
4
15
4
14.0
3
5.
4
那么根据两者出现的次数,可以计算一秒钟其信息量平均为
10
4
log
3
2
5
4
log
3
15
4
log
53
2
2.3 解:
(a)骰子 A 和 B,掷出 7 点有以下 6 种可能:
A=1,B=6; A=2,B=5; A=3,B=4; A=4,B=3; A=5,B=2; A=6,B=1
概率为 6/36=1/6,所以信息量
-log(1/6)=1+log3≈2.58 bit
(b) 骰子 A 和 B,掷出 12 点只有 1 种可能:
A=6,B=6
概率为 1/36,所以信息量
-log(1/36)=2+log9≈5.17 bit
2.5 解:
出现各点数的概率和信息量:
1 点:1/21,log21≈4.39 bit; 2 点:2/21,log21-1≈3.39 bit; 3 点:1/7,log7≈2.81bit;
4 点:4/21,log21-2≈2.39bit;
6 点:2/7,log(7/2)≈1.81bit
平均信息量:
(1/21)×4.39+(2/21)×3.39+(1/7)×2.81+(4/21)×2.39+(5/21)×2.07+(2/7)×1.81≈2.4bit
5 点:5/21,log(21/5)≈2.07bit;
2.7 解:
X=1:考生被录取; X=0:考生未被录取;
Y=1:考生来自本市;Y=0:考生来自外地;
Z=1: 考生学过英语;Z=0:考生未学过英语
P(X=1)=1/4,
P(Z=1/ Y=1)=1, P(Z=1 / X=0, Y=0)=0.4, P(Z=1/ X=1, Y=0)=0.4,
P(Z=1/Y=0)=0.4
(a) P(X=0,Y=1)=P(Y=1/X=0)P(X=0)=0.075, P(X=1,Y=1)= P(Y=1/X=1)P(X=1)=0.125
P(Y=1/ X=1)=1/2; P(Y=1/ X=0)=1/10;
P(X=0)=3/4;
P(Y=1)= P(X=0,Y=1)+ P(X=1,Y=1)=0.2
P(X=0/Y=1)=P(X=0,Y=1)/P(Y=1)=0.375, P(X=1/Y=1)=P(X=1,Y=1)/P(Y=1)=0.625
I(X ;Y=1)=
)1Y(I)1Y/(P
log)1Y/(P
x;
x
x
)1Y/(P
x
)P(
x
x
=
log)1Y/0X(P
x
)1Y/0X(P
P(X
0)
log)1Y/1X(P
)1Y/1X(P
P(X
1)
=0.375log(0.375/0.75)+0.625log(0.625/0.25)=(5/8)log5-1≈0.45bit
(b) 由于 P(Z=1/ Y=1)=1, 所以 P(Y=1,Z=1/X=1)= P(Y=1/X=1)=0.5
P(Y=1,Z=1/X=0)= P(Y=1/X=0)=0.1
那么 P(Z=1/X=1)= P(Z=1,Y=1/X=1)+ P(Z=1,Y=0/X=1)=0.5+ P(Z=1/Y=0,X=1)P
(Y=0/X=1)=0.5+0.5*0.4=0.7
P(Z=1/X=0)= P ( Z=1,Y=1/X=0 ) + P ( Z=1,Y=0/X=0 ) =0.1+P(Z=1/Y=0 ,
X=0)P(Y=0/X=0)=0.1+0.9*0.4=0.46
P(Z=1,X=1)= P(Z=1/X=1)*P(X=1)=0.7*0.25=0.175
P(Z=1,X=0)= P(Z=1/X=0)*P(X=0)= 0.46*0.75=0.345
P(Z=1) = P(Z=1,X=1)+ P(Z=1,X=0) = 0.52
P(X=0/Z=1)=0.345/0.52=69/104
P(X=1/Z=1)=35/104
I(X ;Z=1)=
x
)1Z(I)1Z/(P
x;
x
log)1Z/(P
x
x
=
log)1Z/0X(P
)1Z/0X(P
P(X
0)
log)1Z/1X(P
)1Z/(P
x
)P(
x
P(X
)1Z/1X(P
1)
=(69/104)log(23/26)+( 35/104)log(35/26) ≈0.027bit
(c)H(X)=0.25*log(1/0.25)+0.75*log(1/0.75)=2-(3/4)log3=0.811bit
H(Y/X)=-P(X=1,Y=1)logP(Y=1/X=1) -P(X=1,Y=0)logP(Y=0/X=1)
-P(X=0,Y=1)logP(Y=1/X=0) -P(X=0,Y=0)logP(Y=0/X=0)
=-0.125*log0.5-0.125*log0.5-0.075*log0.1-0.675*log0.9
=1/4+(3/40)log10-(27/40)log(9/10)≈0.603bit
H(XY)=H(X)+H(Y/X)=9/4+(3/4)log10-(21/10)log3=1.414bit
P(X=0,Y=0,Z=0)= P(Z=0 / X=0, Y=0)* P( X=0, Y=0)=(1-0.4)*(0.75-0.075)=0.405
P(X=0,Y=0,Z=1)= P(Z=1 / X=0, Y=0)* P( X=0, Y=0)=0.4*0.675=0.27
P(X=1,Y=0,Z=1)= P(Z=1/ X=1,Y=0)* P(X=1,Y=0)=0.4*(0.25-0.125)=0.05
P(X=1,Y=0,Z=0)= P(Z=0/ X=1,Y=0)* P(X=1,Y=0)=0.6*0.125=0.075
P(X=1,Y=1,Z=1)=P(X=1,Z=1)- P(X=1,Y=0,Z=1)=0.175-0.05=0.125
P(X=1,Y=1,Z=0)=0
P(X=0,Y=1,Z=0)=0
P(X=0,Y=1,Z=1)= P(X=0,Z=1)- P(X=0,Y=0,Z=1)= 0.345-0.27=0.075
H(XYZ)=-0.405*log0.405-0.27*log0.27-0.05*log0.05-0.075*log0.075-0.125*log0.125-0.075*log
0.075=(113/100)+(31/20)log10-(129/50)log3
=0.528+0.51+0.216+0.28+0.375+0.28=2.189 bit
H(Z/XY)=H(XYZ)-H(XY)= -28/25+(4/5)log10-12/25log3 =0.775bit
2.9 解:
A,B,C 分别表示三个筛子掷的点数。
X=A, Y=A+B, Z=A+B+C
由于 P(A+B+C/ A+B)=P(C/A+B)=P(C)
所以 H(Z/Y)=H(A+B+C/ A+B)=H(C)=log6 =2.58bit
H(X/Y)= H(A/Y)
Y
12
11
10
9
8
7
6
5
4
3
2
组合数目
组合情况(A+B)
P(A=a/Y=y)
1
2
3
4
5
6
5
4
3
2
1
6+6
5+6,6+5
4+6,5+5,6+4
3+6,4+5,5+4,6+3
...
1+6,2+5,3+4,4+3,5+2,6+1
...
...
...
...
1+1
1
1/2
1/3
1/4
...
1/6
...
...
...
...
1
一共 36 种情况,每种情况的概率为 1/36,即 P(A=a,Y=y)=1/36
H(X/Y)=H(A/Y)=(1/36)[(-1*log1-2*log(1/2)-3*log(1/3)-4*log(1/4)-5*log(1/5) )*2-6*log(1/6)]=1.
89bit
由于 P(A+B+C/ A+B,A)=P(C/A+B,A)=P(C)
H(Z/XY)=H(C) =log6 =2.58bit
由于 P(A=x,A+B+C=z/A+B=y)=P(A=x,C=z-y/ A+B=y)=P(A=x/A+B=y)P(C=z-y/A+B=y)=
P(A= x / A+B=y)P(C=z-y)=P(A/Y)P(C)
P(A/Y)上面已经给出。
Y
12
11
10
9
8
7
6
5
4
3
2
组合数目
6
12
18
24
30
36
30
24
18
12
6
组合情况(A+B+C)
6+6+1, 6+6+2,...., 6+6+6
...
...
...
...
...
...
...
...
...
...
P(A=x,A+B+C=z/A+B=y)
1/6
1/12
1/18
1/24
...
1/36
...
...
...
...
1/6
一共 216 种情况,每种情况的概率为 1/216,即 P(XYZ)=1/216
H(XZ/Y)=
(1/216)[(-6*log(1/6)-12*log(1/12)-18*log(1/18)-24*log(1/24)-30*log(1/30))*2-36*log(1/36)]=
(1/36)*[(log6+2log12+3log18+4log24+5log30)*2+6log36]=4.48 bit
由于 P(Z/X)=P(B+C/A)=P(B+C)
B+C 的组合共 36 种:
B+C
12
组合数目
1
组合情况(B+C)
6+6
P(Z/X)
1/36
2
3
4
5
6
5
4
3
2
1
)log (
p z x
/
)
11
10
9
8
7
6
5
4
3
2
/
)
H Z X
xyz
(
( )
p a
abc
(
(
p xz
( )
p a p a b c a
/
)log (
p a b c a
/
)
(
p b c
)log (
p b c H B C
)
(
a
bc
5+6,6+5
4+6,5+5,6+4
3+6,4+5,5+4,6+3
...
1+6,2+5,3+4,4+3,5+2,6+1
...
...
...
...
1+1
)
2/36
3/36
4/36
...
5/36
...
...
...
...
1/36
= (1/36)*{[log36+2log(36/2)+ 3log(36/3)+ 4log(36/4)+ 5log(36/5)]*2+6log(36/6)}bit
2.11 解:P(0/0)=P(1/1)=1- p, P(1/0)=P(0/1)= p
(a) P(ul)=1/8
P(ul,0)=P(ul)×P(0/ul)=(1/8)×(1-p)
接收的第一个数字为 0 的概率:
P(0)=P(ul)×P(0/ul)+ P(u2)×P(0/u2)+……. P(u8)×P(0/u8)
=4×(1/8)×(1-p)+ 4×(1/8)×p=1/2
I(ul; 0)=log[ P(ul,0)/P(0)P(ul)]=1+log(1-p)
(b) P(ul,00)=P(ul)×P(00/ul)=(1/8)×(1-p)2
P(00)=P(ul)×P(00/ul)+ P(u2)×P(00/u2)+……. P(u8)×P(00/u8)
=2×(1/8)×(1-p)2 +4×(1/8)×p (1-p)+ 2×(1/8)×p2
=1/4
I(ul; 00)=log[ P(ul,00)/P(00)P(ul)]= 2+2log(1-p)
(c) P(ul,000)=P(ul)×P(000/ul)=(1/8)×(1-p)3
P(000)=P(ul)×P(000/ul)+ P(u2)×P(000/u2)+……. P(u8)×P(000/u8)
= (1/8)×(1-p)3 +3×(1/8)×p (1-p) 2+3×(1/8)×p 2 (1-p) +(1/8)×p3
=1/8
I(ul; 000)=log[ P(ul,000)/P(000)P(ul)]= 3+3log(1-p)
(d) P(ul,0000)=P(ul)×P(0000/ul)=(1/8)×(1-p)4
P(0000)=P(ul)×P(0000/ul)+ P(u2)×P(0000/u2)+……. P(u8)×P(0000/u8)
= (1/8)×(1-p)4 +6×(1/8)×p 2 (1-p) 2+ (1/8)×p4
I(ul; 0000)=log[ P(ul,0000)/P(0000)P(ul)]=
3log{2 2 } log{
p
(1
2
6
p
)
p
(1
}
2
p
)
4
p
(1
4
p
)
2.12 解:
Z
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
概率 1/63 3/63 6/63 10/6315/6321/6325/6327/6327/6325/6321/6315/6310/636/63 3/63 1/63
I(Y;Z)=H(Z)-H(Z/Y)
I(X;Z)= H(Z)-H(Z/X)
I(XY ;Z)=H(Z)-H(Z/XY)
I(Y;Z/X)=I(XY;Z)-I(X;Z)
I(X;Z/Y)= I(XZ;Y)-I(Y;Z)= H(XZ)-H(XZ/Y)
-I(Y;Z)
以上可以根据 2.9 的结果求出
-I(Y;Z) = H(X)+H(Z/X)
-H(XZ/Y)
2.27 解:考虑到约束条件
0
( ) 1,
q x
0
xq x m
( )
采用拉格朗日乘子法
( ( ))
H q x
c
(
1
2
m
)
0
0
( )log ( )
q x
q x dx
x
e
1
2
( )
q x
( )log
q x
[
1
0
xq x dx m
( )
]
[
2
0
dx
(
1
2
m
)
log
e
0
( )
q x dx
a
1]
1
2
x
( )
q x
( )[
q x
1]
dx
当且仅当
( )
q x
x
a
1
2
时,等式成立。
将
( )
q x
x
a
1
2
带入
0
( )
q x dx
1,
0
xq x dx m
( )
:
1
1
ln
m a
2
,
a
m
实现最大微分熵的分布
( )
q x
x
ln
m a
1
m
a
1
m
e
ln (
a
x
ln
m a
)
x
m
e
1
m
,相应的熵值 log(me)
2.29 证明:
( )
Q x
(a)
所以 Q(x)为概率分布。
( )
Q x
1
(1
)
( )
Q x
2
(1
) 1
(b) 即证明熵的凸性。
)
(1
)
1
(
H U
Q x
1
( )log
(
H U
(1
)
)
Q x
( )log
2
(
( )
Q x
1
(1
)
Q x
( ))log
2
1
( )
Q x
Q x
1
( )log
(1
)
Q x
( )log
2
log
e
Q x
1
( )[
1]
log
e
(1
)
Q x
( )[
( )
Q x
( )
Q x
2
1] 0
1
( )
Q x
2
( )
Q x
( )
Q x
2
2
)
1
(
H U
1
( )
Q x
1
( )
Q x
( )
Q x
1
( )
Q x
( )
Q x
1