EE6340 - Information Theory

Problem Set 1 Solution

February 21, 2013

1. a) Random variable X= No. of coin tosses till the first head appears. If P(head)=p and P(tail) = q,

P(x = n) = pq n−1
X∞
=⇒ H(X) = − pq n−1 log(pq n−1 )
n=1
p log p pq log q
=− −
1−q p2
= H(p)/p
For p = 21 , H(X) = H(0.5)/0.5 = 2 bits.
b) Best questions are those which have equal chances of being answered as Yes or as No.
P(x = 1) = 12 and P(x=2 or 3 or 4...)= 12 (equally probable).
So first question is, "Is it 1 or not?’".
Similarly,P(x = 2) = 14 and P(x=3 or 4 or 5...)= 14 (equally probable)
So second question is, "Is it 2 or not?’".
The questions proceedP as above.
∞
E(No.of questions) = n=1 n 21n = 2 = H(X).
In general,E(No.of questions) ≥ H(X)
This problem can be interpreted as a source coding problem with 0=no, 1=yes, X=Source,
Y=Encoded sequence.
2. a) H(X, g(X)) = H(X) + H(g(X)|X) (Chain rule)
P P
b) H(g(X)|X) = x p(x)H(g(X)|X = x) = x p(x)0 = 0 (For a given x,g(x) is fixed).
=⇒ H(X, g(X)) = H(X)
c) H(X, g(X)) = H(g(X)) + H(X|g(X)) (Chain rule)
d) H(X|g(X)) ≥ 0 with equality if g(.) is one-to-one.
(a),(b) and (c) =⇒ H(X, g(X)) ≥ H(g(X)

3. Let there be 2 yi ’s y1 and y2 such that for x = x0 , p(x0 , y1 ) > 0 and p(x0 , y2 ) > 0.
=⇒ p(y1 |x0 ) > 0 and p(y2 |x0 ) > 0, neither of them being 0 or 1.
X
H(Y |X) = − p(x, y) log2 p(y|x)
x,y

≥ −p(x0 )p(y1 |x0 ) log2 p(y1 |x0 ) − p(x0 )p(y2 |x0 ) log2 p(y2 |x0 )
>0
since −t log t > 0 for 0 < t < 1. So, H(Y|X)= 0 iff Y is a function of X. Else H(Y|X)>0.

1
4. X=Outcome of world series
Y=No.of games played  {4,5,6,7}
1
For Y = 4, there are 2 outcomes {AAAA,BBBB} each with probability 24 .
For Y = 5,there are 2 × 43 = 8 outcomes each with probability 215 .

For Y = 6,there are 2 × 53 = 20 outcomes each with probability 216 .

for Y = 7,there are 2 × 63 = 40 outcomes each with probability 217 .

Thus,
1
P(Y = 4) =
8
1
P(Y = 5) =
4
5
P(Y = 6) =
16
5
P(Y = 7) =
16
X={AAAA,BBBB,....,BABABAB,....BBBAAAA}
P(BAAAA)= 215 , P(BABABAB)= 217 ,...
There are 2 sequences of length 4, 8 sequences of length 5, 20 sequences of length 6 and 40 sequences
of length 7.
X
H(X) = − p(x) log2 p(x)
x
       
1 1 1 1
=2 log2 16 + 8 log2 32 + 20 log2 64 + 40 log2 128
16 32 64 128
=5.8125
X
H(Y ) = − p(y) log2 p(y)
y
1 1 5 16 5 16
= log2 8 + log2 4 + log2 + log2
8 4 16 5 16 5
=1.924

Y=length(X), i.e, Y is a deterministic function of X.

H(Y |X) =0
H(X) + H(Y |X) =H(X, Y ) = H(Y ) + H(X|Y )
=⇒ H(X|Y ) =H(X) − H(Y )
=3.889
2 3 1
5. a) H(X) = 3 log2 2 + 3 log2 3 = 0.918 = H(Y )
1
b) H(X|Y ) = 3 H(X|Y = 0) + 32 H(X|Y = 1) = 0.667 = H(Y |X)
c) H(X, Y ) = H(X) + H(Y |X) = 1.585
d) H(Y ) − H(Y |X) = 0.251
e) I(X; Y ) = H(Y ) − H(Y |X) = 0.251
f)
6. Identically distributed =⇒ H(X1 ) = H(X2 )
H(X2 |X1 ) H(X1 )−H(X2 |X1 ) H(X2 )−H(X2 |X1 ) I(X1 ;X2 )
a) ρ = 1 - H(X1 ) = H(X1 ) = H(X1 ) = H(X1 )

2
 H(X) H(Y)

H(X|Y) I(X;Y) H(Y|X)

b) I(X1 ; X2 ) = H(X1 ) − H(X1 |X2 )