PHY 441/4101: Quantum Information Theory, Varsha 2014

Problem Set 1
(Dated: due: 27 August 2014)
1. The Monty Hall problem: This problem is based on the american television show Lets
Make a Deal. In the show, the host, Monty Hall, would show a contestant three doors. Behind
one door was the grand prize; behind the other two was nothing. The contestant chose one of
the doors. Monty then opened one of the remaining doors, revealing that it hid nothing. The
contestant was then oered the opportunity to stick with his original choice or to switch to the
one remaining door. Should he switch? Consider a generalization of this problem: there are
N doors, the contestant guesses one, Monty opens n of the other doors, showing that none of
them hides the prize, and the contestant is given the chance to switch to any of the N n 1
remaining doors.
(a) What is the probability p(R) that the contestants initial guess is right and the probability
p(W) that the initial guess is wrong?
(b) Now consider the situation after Montys revelation. Let g denote the door of the contes-
tants initial guess, and let r denote any one of the remaining doors. Given that the initial
guess is right, what is the conditional probability p(r|R) that the prize lies behind door r
and the conditional probability p(g|R) that the prize lies behind door g?
(c) What are the corresponding conditional probabilities, p(r|W) and p(g|W), given the that
the initial guess is wrong?
(d) Since the contestant doesnt know whether his initial guess is right or wrong, what is
relevant to his decision are the unconditioned probabilities p(r) and p(g). What are p(r)
and p(g)? For what values of n should the contestant change his initial guess to one of
the remaining doors?
2. A binary experiment is conducted such that p(0) = q and p(1) = p = 1 q. Also n independent
trials of this experiment are performed, and we are interested in the random variable K that
counts the number of 1s that occur. K has n + 1 possible values 0, 1, ..., n.
(a) Show that the probability distribution for K is given by
p(k) =

n
k

p
k
q
nk
.
(b) Find the mean and the variance of the binomial distribution.
(c) Stirling?sformula provides an approximation of n! for large n:
n!

2nn
n
e
n
.
Use this to nd an approximate formula for the binomial coecient. Once you have done
this, estimate the value of k that maximizes the probability p(k) for large n. What do you
nd?
2
3. Prove Jensens inequality:

i
p
X
(x
i
)f(x
i
) f

i
p
X
(x
i
)x
i

,
where X is a random variable and f(x) is a convex function (look up the denition of a convex
function if your dont know what it is).
4. A sequence of 16 binary symbols are sent out from a source which produces two independent
binary symbols with p(0) = p and p(1) = 1 p with p > 0.5. A typical sequence is dened to
have two or fewer symbols of 1.
(a) What is the most probable sequence that can be generated by this source and what is its
probability?
(b) What is the number of typical sequences that can be generated by this source? Assume that
one assigns a unique binary codeword for each typical sequence and neglect the non-typical
sequences.
(c) If the assigned codewords are all of the same length, nd the minimum codeword length
required to provide the above set with distinct codewords.
(d) Determine the probability that a sequence is not assigned with a codeword.