Problems involving linearity of expectation and careful choice of random
variables
-
A bin contains a white balls and
b black balls. We take out the
balls at random from the bin without replacement. What is the probability
that first ball is white, second ball is white,
kth ball is white.
-
A bin contains m white balls and
n black balls. We take out the
balls uniformly randomly from the bin without replacing them. What is the
expected number of black balls left when all the white balls have been taken
out.
-
Given n balls and n bins. We throw each ball
uniformly randomly into the bins. What is the expected number of
(i) empty bins (ii) bins containing one ball (iii)
bins containing k balls
- Given a graphs G=(V,E) on
n vertices and m edges. Show that there is a bipartite sub-graph of the
graph G with atleast m/2 edges.
- Show that the expected number of comparisons made during randomized quick
sort on an input of size n is
O(nlog n) .
- Mulmuley Games (A slightly different formulation)
Consider a group consisting of
m men,
w women,
c children,
r rabbits.
Assume that no two individuals have same height. Moreover, each rabbit is
shorter in height than each child who in turn is shorter than each women and
each woman is shorter in height than each man.
There are five types of games. Each game starts with
a set of participants arranged in a line from left to right where each
permutation is equally likely. There is a prize at the left end of the line.
Assume that a participant at place i
would see the prize if none of the participants preceding him/her
is taller than him/her.
-
Game A : all rabbits and women are the participant. What would be the expected
number of women that would see the prize?
-
Game B : all rabbits, women and children participate in the game. What
would be the expected number of children who can see the prize.
-
Game C : all rabbits, women and children participate in the game. A woman
can receive the prize only if she can see the prize and there is atleast
at-least one of the participants preceding her in the line is a child.
What is the expected number of women that would win the prize.
-
Game D : all rabbits, women, children and men are the participants. A woman
can receive the prize only if she can see the prize and there is atleast
at-least one of the participants preceding her in the line is a child.
What is the expected number of women that would win the prize.
-
Game E :
all rabbits, women, children and men are the participants. A woman
can receive the prize only if she can see the prize and all children
appear before her. What is the expected number of women that would win the
prize?
-
A randomized binary search tree on n keys is
formed as follows.
Initially the tree is empty. We keep a pool of all the key that have to be
inserted in to the tree. At each step, we take out a key from the pool
uniformly randomly and insert it into the binary search tree. We continue until
the pool becomes empty, and we get a binary search tree.
Use Mulmuley's game A to show that the expected depth of
ith largest key in
a random binary search tree is
log i + log n-i.
-
A sick person has a bottle containing pills. There are two types of pills :
small pills and large pill. Each large pill is equivalent to
two small pills. The sick person has to take one small pill per day.
He wakes up daily in morning and takes out one pill at random from the bottle.
If it is a small pill, he consumes it and puts the bottle back, otherwise he
breaks the large pill into two halves, eat one half and put the other
half back into the bottle and this half piece of the large pill would be
treated as a small pill henceforth. What is the expected number of small pills
left in the bottle when all the large pills have been consumed.
More Ball-Bin Problems and Traversal Problems
-
Given an unlimited supply of balls and a set of n empty bins.
We take a ball from the supply and throw it randomly among the bins so that
its chances of landing into any of the n
bins is same. We continue throwing the balls until no bin is left empty.
What is the expected number of balls thrown before no bin is left empty ?
-
Consider the following experiment that proceeds in a sequence of rounds.
For the first round, we have n balls,
which are thrown uniformly randomly into n
bins. After round i , we discard every ball
that fell into a bin itself in round i .
The remaining balls are retained for i+1
round, in which they are again thrown independently and uniformly at
random into the n bins. We stop when there
is no ball left.
Prove that with probability 1- o(1) the
experiment will finish in loglog n number
of rounds.
-
There are n bins and
n players, and each player has an infinite
supply of balls. The bins are initially empty. We have a sequence of rounds :
in each round, each player throws a ball into an empty bin chosen
independently at random from all currently empty bins. Let the random variable
Z be the number of rounds before every bin
is non-empty. Determine the expected value of
Z.
What can you say about the distribution of
Z?
-
There are two bins A and
B, each containing one ball initially.
We have n balls with us that we shall put
one by one into these bins randomly in n
steps as explained below.
Let A(t) and
B(t) denote the number of balls in bin
A and B
respectively after t steps.
During (t+1)
step, we take out a ball from the box and put it into bin
A with probability
A(t)/(A(t)+B(t)), and into bin
B with probability
A(t)/(A(t)+B(t)).
What can you say about the distribution of the random variable
A(t). In other words, what is the
probability that bin A has
t balls
(for a given t less than or equal to
n).
-
Consider the following description of 1-dimensional random walk :
A particle is at the origin (mid-point) of interval
[-n,n] initially. In each minute, the
particle moves one unit to left or one unit to the right with equal
probability. What is the expected number of minutes until the particle
hits one of the boundaries of interval
[-n,n]? What can you say about the expected
distance of the particle from the origin?
-
***
Given a tree on n vertices, and
let u and
v be any two vertices in the tree. A monkey
is performing a random walk starting from vertex
u.
Use arguments analogous to the previous problem to show that the expected
number of steps required by the monkey to reach vertex
v is
O(n^2).
Problems on variance and application of Chernoff's bound.
-
There are m balls and
n bins. For each ball we select
a bin uniformly at random, and place it inside the bin. Show that
if m>=n\log n, the fullest bin
will have O(m/n) balls.
-
On Random Graphs
Definition : Let G(n,p) be the
random graph on n vertices which
is obtained by independently adding an edge between every pair of
vertices with probability p.
-
A triangle in G(n,p) is defined as
a set of three nodes such that every pair of these nodes is joined by an
edge. Let X be the random variable which
denotes the number of distinct triangles in
G(n,p). What is the expectation of
X? What is the variance of
X?
- Consider G(n,p) for
p=1/2. Show that the expected
diameter is at most a constant.
- Consider G(n,p) for
p=(c log n)/n , where
c is a large enough constant.
Show that the expected diameter of this graph is
O(log n).
- Given a connected graph G. Further
each node in the graph chooses c
nodes uniformly at random and adds edges to these, where
c is a large enough constant.
Show that this graph has expected diameter
O(log n).
Remark : Compare the two problems given above and note the
following observation : Unlike a random graph,
if the initial graph is connected, it requires only a constant number
of random edges per vertex to reduce the graph diameter to
O(log n).
-
On spreading of rumor
- Consider a group of n persons.
Initially one person has some information. This information is spread in
rounds as follows. In a round, each person who has the information
calls some person uniformly at random from the group and communicates
the information to him. Show that the expected number of rounds before
everyone knows the information is
O(log n) with very high probability
(i.e., with probability at least
1-1/O(n))
- The previous algorithm for spreading rumor is called 'push-algorithm'.
Now consider the 'pull-algorithm' for rumor spreading. In this algorithm,
we begin with one node having the information. In each round, every node
calls a node at random. If x calls a
node y and
y has the information, then
x also gets the information.
Show that the expected number of rounds before everyone knows the
information is
O(log n) with very high probability.
- ***
Show that the a combination of 'push-algorithm' and 'pull-algorithm'
would reduce the expected number of rounds to
O(loglog n).
Problems based on application of conditional probability
-
n persons, all of different heights are
standing in a queue facing the window of ticket collector in a cinema hall.
What is the expected number of persons visible to the person at the ticket
window if each permutation of n persons is equally likely.
-
We are given a random permutation of n real
numbers. Consider the following scheme P(k)
to estimate the largest of these numbers.
P(k) : Scan first
k numbers in the permutation and select the
largest of them. Let it be max(k).
Now continue scanning until we find a number larger than
max(k). If we find such a number, report
that number as the largest of all the numbers. If we do not succeed
(and hence end up scanning all the numbers), declare
max(k) to be the largest number.
- What is the probability (in terms of k
and n) that using scheme
P(k) succeeds in reporting the largest
number ?
- What value of k maximizes the success
of our scheme ?
-
Given an undirected weighted graph G(V,E)
on n vertices and
m edges.
For a subset S of edges, Let
T(V,S) be the minimum spanning tree for the
subgraph G(V,S) . An edge
e
of set E-S is said to be light with respect
to S if its addition can improve the mst.
Show that if set S is formed by picking
each edge independently with probability p ,
then expected number of light edges from the set
E-S is
n/p .
-
There is a family with two children. Assume that each child is equally likely
to be a girl or a boy independent of the other child. Given the additional
information that there is at-least one boy in the family, what is the
probability that the other child is also a boy.
How to sample with a coin ?
-
Given a biased coin with unknown bias, how can you use it to
simulate a fair coin?
-
You have to permute n
items uniformly. You have only a fair coin at your
disposal. Give two algorithms that output a uniformly picked permutation
of the items. (Hint : One algorithm will be based on the idea of quick sort
while the other algorithm will be based on throwing
n balls into
2n bins
with repating a throw in case of collosion). What is the expected
number of coin tosses required in each case.
-
We want n
random variables, each of them should take a value uniformly
in the range 1 to
n .
Furthermore we want that they should be pairwise
independent. How can you generate a sample of these n numbers with
only O(log n) random bits?
-
How can you use a fair coin to select a fruit uniformly from three
fruits : apple, banana, orange? What is the expected number of tosses
required to select one fruit?
-
On a day (exactly) one of n
possible events E1,E2,...,En can occur.
Given that event Ei
occurs with probability pi , with
\sum_{i\le n} pi = 1.
How can you simulate these events using a fair coin ?
What is the expected number of tosses required for predicting an event ?
-
You have a bag that can hold at-most k
apples. You enter a store having apples arranged in a row, and your objective
is to choose a uniform sample of k
apples from the store. However, the store is poorly lit and so you have no
idea of the number of apples in the store. Moreover, you are not allowed to
move backward. In other words, while moving along the row of apples,
you are given only one time access to each apple, and you have to decide in an
online manner whether to select the apple or not. How will you ensure that
you finally end up having k
uniformly picked appled from the store.