Description
Problem 1. (16 marks)
a. (2 marks) Consider the following program:
Smallest (A, f, l)
**Precondition: A[f..l] is an array of integers, f ≤ l and f, l ∈ N.
**Postcondition: Returns the smallest element of A[f..l].
if f = l then
return A[f]
else
m = b
f+l
2
c
return min(Smallest(A, f, m), Smallest(A, m + 1, l))
end if
Write a recurrence relation that describes the time complexity of this program in terms of the size of
array A and solve it.
b. (2 marks) Show the result after each operation below, starting on an NIL BST: Insert(16), Insert(4),
Insert(3), Insert(9), Insert(1), Insert(44), Insert(29), Delete(1), Delete(4), Insert(34).
c. (2 marks) First, draw an AVL tree of height 4 that contains the minimum number of nodes. Then, delete
a leaf node (any leaf node) from your AVL tree above so that the resulting tree violates the AVL-property.
Then, apply tree-rotation to turn it to an AVL tree.
d. (2 marks) Demonstrate what happens when the following keys are inserted to a hash table with
collisions resolved by chaining: 18, 22, 32, 8, 6, 10, 20, 16, 2, 29. Let the table have 9 slots and the hash
function be h(k) = k mod 9.
e. (2 marks) CLRS Exercise 6.5-9 (which can also be found in Problem Set #3) asks you to give
an O(n lg k)-time algorithm to merge k sorted lists into one sorted list, where n is the total number of
elements in all the input lists. Use the following example to explain how your algorithm works for three
sorted lists: L1 = [2,5], L2 = [4,8], and L3 = [3,7].
f. (3 marks) Consider Problem 7 of Problem Set #3
Suppose the root of a tree T with 7 nodes is nearly balanced (note that the subtrees rooted at other
nodes of the tree may not be nearly balanced). What is the maximum height of such a T?
To calculate the maximum possible height of a nearly balanced tree with n nodes, we start from
Hmax(n) = 1 + Hmax(
2
3
(n − 1))
1
Use n = 8 as an example to explain how this equation helps derive the maximum height of a nearly
balanced tree. To support your answer, draw a nearly balanced tree that consists of 8 nodes. (Hint:
The above equation can be viewed as a recursive definition, with some base cases added.)
If a BST is nearly balanced, what would be the running time for searching an element in it? Briefly
justify your answer.
g. (3 marks) Consider Problem 9 of Problem Set #3. Explain, using n = 10, how your solution works.
In particular, explain how you identify which bottle is poisoned.
Problem 2. (8 marks)
A d-ary heap is like a binary heap except that non-leaf nodes have d children instead of 2 children
(with one possible exception for the parent of the last leaf). Also see CLRS Problem 6.1, page 167. In this
question, we consider 3-ary max-heaps.
a. Given an array A and index i, how do you compute its left child, middle child, and right child, and
how do you compute the parent of A[i]?
b. What is the height of a 3-ary heap of n elements? Justify your answer briefly.
c. Write a version of Max-Heapify for a 3-ary max-heap and analyze its running time.
d. Write a version of Build-Heap for a 3-ary max-heap and analyze its running time.
e. Let array A = [2, 8, 5, 4, 7, 10, 12, 6]. Invoke your Build-Heap(A) to build a 3-ary max-heap. Just
show the resulting array.
Problem 3. (8 marks) Given a max-heap A containing n keys and two element x and y such that x < y.
Propose an efficient algorithm for reporting (say printing) all the keys z in A satisfying x ≤ z ≤ y (note
that x and y may not necessarily be in A). The running time of your algorithm should be O(k) where k
is the number of elements that are greater than or equal to x and less than or equal to y.
Give your algorithm in pseudocode and analyse its running time. In particular, comment on why your
algorithm is more efficient than a linear search over A: check each element a in A to see whether it is the
case that x ≤ a ≤ y.
Problem 4. (8 marks) We have a set J of n jars and a set L of n lids such that each lid in L has a unique
matching jar in J. Unfortunately all the jars look the same (and all the lids look the same). We can only
try fitting a lid ` to a jar j for any choice of ` ∈ L and j ∈ J; the outcome of such an experiment is either
a) the lid is smaller, b) the lid is larger, or c) the lid fits the jar. We cannot compare two lids or two jars
directly. Describe an algorithm to match all the lids and jars. The expected number of tries (experiments)
of fitting a lid to a jar performed by your algorithm must be O(n log n).
We describe an algorithm and then you will be asked to answer some questions.
The idea of the algorithm is very similar to that of Quicksort.
1. If n = 1 (i.e. only one jar and one lid) then they must be matching; return this as the solution.
2. If n > 1 we do the following:
(a) Take anyone of the lids, say L[`] as the pivot (for example take one of them randomly)
2
(b) Compare this lid against all the jars (using n comparison) to partition the jars into three sets:
1) Js are all those jars that are smaller than L[`], 2) Jl are all those jars that are larger than
L[`], 3) the 3rd set contains the unique jar, say j, that fits into L[`].
(c) Use jar j (that is matching lid `) and compare against all the lids in L (except `) to partition L
into a set of smallers, denoted by Ls, and those that are largers, denoted by Ll
; this takes n − 1
comparisons.
(d) Note Js are all the jars smaller than ` and Ls are all the lids smaller than j, and therefore the
lids of jars in Js are exactly those in Ls; similarly the lids of jars in Jl are exactly those in Jl
.
Solve the following two subproblems recursively: jars and lids in Js and Ls as one subproblem,
and jars and lids in Jl and Ll as another subproblem.
Here are the questions that you are asked to answer.
Provide a running time analysis.
Suppose it is required that your algorithm always select the last lid as the pivot in the given line up.
Consider a problem with 7 lids l1, . . . , l7 and 7 jars j1, . . . , j7, in the increasing order of their sizes,
where all jars have distinct sizes and li uniquely fits ji for all 1 ≤ i ≤ 7. Given the sequence of lids,
hl5, l1, l6, l7, l2, l3, l4i, and an arbitrary sequence of jars, explain how your algorithm works. In your
answer you can assume that any subproblem of the given problem is solved correctly. That is, you
don’t need to show how subproblems are solved, just give the result.
Suppose your algorithm must take the last lid as the pivot in the given line up. What would be
the worst-case running time of your algorithm and when it is guaranteed to happen? What is the
best-case running time? You do not need to present your algorithm, but it must be an efficient one
under the requirement.
3
