Computer Science CSC373H September 21, 2022

St. George Campus University of Toronto

Homework Assignment #2

Due: October 5, 2022, by 11:00 am

• You must submit your assignment through the Crowdmark system. You will receive by
email an invitation through which you can submit your work. If you haven’t used Crowdmark
before, give yourself plenty of time to figure it out!

• You must submit a separate PDF document with for each question of the assignment.

• To work with one or two partners, you and your partner(s) must form a group on Crowdmark (one
submission only per group). We allow groups of up to three students. Submissions by groups of
more than three students will not be graded.

• The PDF file that you submit for each question must be typeset (not handwritten) and clearly
legible. To this end, we encourage you to learn and use the LATEX typesetting system, which is
designed to produce high-quality documents that contain mathematical notation. You can use
other typesetting systems if you prefer, but handwritten documents are not accepted.

• If this assignment is submitted by a group of two or three students, for each assignment question
the PDF file that you submit should contain:

1. The name(s) of the student(s) who wrote the solution to this question, and
2. The name(s) of the student(s) who read this solution to verify its clarity and correctness.

• By virtue of submitting this assignment you (and your partners, if you have any) acknowledge
that you are aware of the homework collaboration policy for this course, as stated here.

• For any question, you may use data structures and algorithms previously described in class, or in
prerequisites of this course, without describing them. You may also use any result that we covered
in class (in lectures or tutorials) by referring to it.

• Unless we explicitly state otherwise, you should justify your answers. Your paper will be marked
based on the correctness and efficiency of your answers, and the clarity, precision, and conciseness
of your presentation.

• The total length of your pdf submission should be no more than 4 pages long in a 11pt font.

1
Question 1. (10 marks) Consider Huffman’s algorithm.
a. Find a small set of symbols Γ and associated frequencies f , such that: (i) there is exactly one
symbol with frequency 31 , (ii) all the other symbols have frequency strictly less than 31 , and (iii) executing
Huffman’s algorithm on (Γ, f ) can produce a codeword of length 1. Justify your answer.
1
b. Prove that for all sets of symbols Γ with associated frequencies f such that f (x) < 3 for all x ∈ Γ,
executing Huffman’s algorithm on (Γ, f ) does not produce any codeword of length 1.

Question 2. (10 marks) You own a company of trucks and you want to use them to supply a set of cities.
Each pair of cities may be connected by a road that traverses no other city (some roads may be one-way).
Each road may go through a single tunnel. Each tunnel has a height that specifies the maximum height of
trucks that can traverse it: a tunnel of height h can be traversed only by trucks of height at most h.
An itinerary from a city s to a city t is a sequence of alternating cities and roads that start at s and
finishes at t (without repeating any city). The height H of an itinerary is the minimum height of all the
road tunnels that it traverses; note that a truck with height more than H can not take this itinerary. So for
your truck company, a best itinerary from s to t is one whose height H is maximum among all itineraries
from s to t. We want an efficient algorithm to find best itineraries starting from a given city s.
More precisely, the algorithm’s input is:

1. A set of n cities {1, 2, . . . , n}.

2. For each city i, a list L[i] containing pairs of the form (j, hij ) with j ̸= i and hij ≥ 0.
L[i] contains (j, hij ) iff: (a) there is a road from city i to city j, and (b) the height of the tunnel on
that road is hij ; if there is no tunnel on that road hij = +∞.

3. A start city s ∈ {1, 2, . . . , n}.

The algorithm must find for every city t: (1) a best itinerary from s to t, and (2) the height of this itinerary.
If no itinerary from s to t exists, the algorithm should report that the height of the best itinerary from s
to t is 0.
Your task is to modify Dijkstra’s algorithm to do so. First describe these modifications clearly and concisely
in English, and then give the full algorithm’s pseudocode. Do not prove its correctness.

Question 3. (20 marks) Isabella Rosselini has n assignments A1 , A2 , . . . , An due now, and she has not
started working on any of them (not surprisingly, she was making a movie). Assignment Ai takes her
di days to do, and has penalty pi for each day it is submitted late (assume that di and pi are positive
integers). For example, if Isabella starts by doing assignment Ai first, her penalty on this assignment will
be pi · di , and if Isabella does the assignments in the order S = A1 , A2 , . . . , An , the total penalty (sum of
penalties) is:

PS = p1 d1 + p2 (d1 + d2 ) + p3 (d1 + d2 + d3 ) + · · · + pn (d1 + d2 + · · · + dn )

because if she does them in this order, she will finish assignment Ai after d1 + d2 + · · · + di days (and
then she will immediately start working on assignment Ai+1 ). Note that Isabella can only work on one
assignment at a time.
There are n! different orders in which the assignments can be done. An ordering of the assignment is
optimal if it minimizes the total penalty. Isabella just learned about Greedy Algorithms, so she hopes
that this approach will help her find an optimal ordering.

2
For each of the following three greedy strategies for ordering the assignments, either show that it always
results in an optimal ordering of the assignments or provide a counterexample:
a. Do the assignments in order of increasing length (number of days that they take).
That is, schedule them in an order such that if di < dj then Ai occurs before Aj in that order.
b. Do the assignments in order of decreasing daily penalty.
That is, schedule them in an order such that if pi > pj then Ai occurs before Aj in that order.
c. Do the assignments in order of decreasing penalty-to-length ratio.
That is, schedule them in an order such that if pi /di > pj /dj then Ai occurs before Aj in that order.
In the three strategies above, do not assume that the pi ’s, di ’s, or pi /di ’s are necessarily distinct.