Professional Documents
Culture Documents
• Entropy: ∑
H(X) = − p(x) log2 p(x) (bits)
∫x
H(X) = − f (x) log f (x)dx (bits)
H(X n)
H(X ) = lim = lim H(Xn|Xn−1, . . . , X1)
n→∞ n n→
∑
=− µiPij log Pij for 1st order Markov chain
ij
;4#47.(<"=%$>78%"(
!"#$%&'(( (
56.7896(!"#$%&'( <*+?(@,(-()*+,(2()*+0@,(
)*+,(-((
3*&00:,( (
.%/01023*&004,((
<*+?(@,(-(3*&*AB(',00&*A,&*',,(
H(X ) H(Y )
• Chain rules:
H(X, Y ) = H(X) + H(Y ∑n|X),
H(X1, X2, · · · , Xn) = ∑ i=1 H(Xi |Xi−1 , · · · , X1 )
n
I(X1, X2, · · · , Xn; Y ) = i=1 I(Xi; Y |Xi−1, · · · , X1)
• H(X) ≥ 0 (but differential entropy can be < 0), I(X; Y ) ≥ 0 (for both
discrete and continuous)
I(X; Z) ≤ I(X; Y )
I(X; Z) ≤ I(Y ; Z)
• Fano’s inequality:
H(X|Y ) − 1
Pe ≥
log |X |
Physical Channel
/0+1'+"#$,"-#''
10.'#23$).2")')' !"#$%""&'
()*'+"#$,"-#.'
^
min l (X; X ) max l (X; Y )
Lossless compression:
• I(X; X̂) = H(X) − H(X|X̂)
Root
10
110
111
^
min l (X; X ) max l (X; Y )
Channel capacity:
• given fixed channel with transition probability p(y|x)
• C = maxp(x) I(X; Y )
• water-filling
Power
P1
P2
N3
N1
N2
%&'()"!*" %&'()"!!"
$" #"
^
min l (X; X ) max l (X; Y )
Rate-distortion:
• given source with distribution p(x)
1 ∑
N
√ (xn − µ) → N (0, 1)
N σ2 n=1
• AEP
1 1
log → H(X)
n p(X1, X2, . . . , Xn)
p(X1, X2, . . . , Xn) ≈ 2−nH(X)
n:| |n elements
Non-typical set
Typical set
A(n) : 2n(H + )
∋
∋ elements
Discrete:
• H(X) ≤ log |X |, equality when X has uniform distribution
Continuous:
• H(X) ≤ 12 log(2πe)n|K|, EX 2 = K
equality when X ∼ N (0, K)
Codeword
Length Codeword X Probability
2 01 1 0.25 0.3 0.45 0.55 1
2 10 2 0.25 0.25 0.3 0.45
2 11 3 0.2 0.25 0.25
3 000 4 0.15 0.2
3 001 5 0.15
2%33"#4' 80&(*'
+*-$'
5$$-6
!9:.'
7*)*3*#'
?.2' :*)%&'
;'<=%&1%)>'-"%4&%3'
I
Un Encoder 1
Decoder (Û n , V̂ n)
J
Vn Encoder 2
R1 + R2 ≥ H(U, V )
R2
H(V )
H(V |U )
H(U |V ) H(U ) R1
X1n
W1 Encoder 1
Yn
p(y|x1 , x2 ) Decoder (Ŵ1 , Ŵ2 )
X2n
W2 Encoder 2
2 2 1
R2
I(X2 ; Y |X1)
I(X2 ; Y )
• Stein’s lemma...