Homework #3
Introduction to Algorithms
601.433/633
Format: Please start each problem on a new page.
Where to submit: On Gradescope, please mark the pages for each question
1 Problem 1 (24 points)
Recall that when using the QuickSort algorithm to sort an array A of length n, we
picked an element x ∈ A which we called the pivot and split the array A into two
arrays AS, AL such that ∀y ∈ AS, y ≤ x and ∀y ∈ AL, y > x.
We will say that a pivot from an array A provides tn − t separation if t elements
in A are smaller than or equal to the pivot, and n − t elements are strictly larger
than the pivot.
Suppose Bob knows a secret way to find a good pivot with n
3

2n
3
separation in
constant time. But at the same time Alice knows her own secret technique, which
provides separation n
4

3n
4
, her technique also works in constant time.
Recall that in the QuickSort algorithm, as per Section 7.1 in CLRS, the PARTITION subroutine picks a pivot x from A by simply picking the first element in the
array. Alice and Bob’s subroutines are subroutines for picking the pivot x in the
PARTITION subroutine for QuickSort.
Alice and Bob applied their secret techniques as subroutines in the QuickSort algorithm to pick pivots. Whose algorithm works asymptotically faster? Or are the
runtimes asymptotically the same? Prove your statement.
1
2 Problem 2 (13 points)
Resolve the asymptotic complexity of the following recurrences, i.e., solve them
and give your answer in BigΘ notation. Use Master theorem, if applicable. In all
examples assume that T(1) = 1. To simplify your analysis, you can assume that
n = a
k
for some a, k.
Your final answer should be as simple as possible, i.e., it should not contain any
sums, recurrences, etc.
1. T(n) = 2T(n/8) + n
1
5 log n log log n
2. T(n) = 8T(n/2) + n
3 − 8n log n
3. T(n) = T(n/2) + log n
4. T(n) = T(n − 1) + T(n − 2)
5. T(n) = 3T(n
2
3 ) + log n
3 Problem 3 (13 points)
Let A and B be two sorted arrays of n elements each. We can easily find the
median element in A – it is just the element in the middle – and similarly we can
easily find the median element in B. (Let us define the median of 2k elements as
the element that is greater than k −1 elements and less than k elements.) However,
suppose we want to find the median element overall – i.e., the nth smallest in the
union of A and B.
Give an O(log n) time algorithm to compute the median of A∪B. You may assume
there are no duplicate elements.
As usual, prove correctness and the runtime of your algorithm.
2