Name: Rollno:

CS345: Design and Analysis of Algorithms (Midsem Exam)

18th September 2024

Total Number of Pages: 7 Time: 2 hr Total Points 80

Instructions Question Points Score

1. All questions are compulsory. 1 12
2. Answer all the questions in the space provided 2 18
in the question paper booklet. The space pro-
vided is much more than what is necessary to 3 12
write down the solution. 4 12

S
3. Use the space provided in the paper for rough 5 12
work.
6 14

N
4. You may cite and use algorithms and their com-
plexity as done in the class. Total: 80
5. Cheating or resorting to any unfair means will
be severely penalized.
IO
6. Superfluous and irrelevant writing will result in
negative marking.
7. Using pens (blue/black ink) and not pencils.
Do not use red pens. for answering.
8. Please bring your ID cards.
T

9. For questions involving algorithm design, it is

sufficient to give a clear and concise description
of your algorithm. You dont need to give the
LU

formal pseudocode.
10. If you design an algorithm, that is less efficient
than what has been asked, you will get partial
credit, provided you clearly mention what is the
time complexity of your algorithm.
11. It is advisable to solve a problem first before
SO

writing down the solution.

12. The questions are not arranged according to the
increasing order of difficulty.

Page 1 of 7
Name: Rollno:
Question 1. Consider the following directed weighted graph G.

3 -3
d e f

-2
3 -2 -1 2 -5

a b c
4 3

(a) (1 point) Does G have any negative weight cycle? No

(b) (5 points) Give the matrix obtained by running Bellman Ford algorithm on G starting at vertex a.

Solution:
1 2 3 4 5
a 0 0 0 0 0
b 4 4 4 4 4
c ∞ 7 4 3 3
d 3 2 2 2 2
e ∞ 6 5 5 5
f ∞ ∞ 2 -1 -2

(c) (3 points) What is the corresponding shortest path tree? (If there are multiple such trees, you can
give any one)

Solution:
4 -2 3 -2 -5
a b d e c f

(d) (3 points) What is the distance of each vertex from a?

a b c d e f

0 4 3 2 5 -2

Question 2. Consider the following flow network, where each edge is labeled with the current flow and
capacity (that is, a/b means that the capacity of the edge is b and the current flow value is a). The
source node is s, and the sink is t.

Page 2 of 7
Name: Rollno:
5/5 3/3 4/4
s u v w

5/5 3/9 2/2 8/8 3/3 4/11 4/6

5/7 2/4 5/5

x y z t

(a) (1 point) Is this a valid flow? Yes

(b) (1 point) What is the value of the flow? 13
(c) (4 points) Draw the residual network corresponding to the given flow.

Solution:
5 3 4
s u v w

6 7
5 2 8 3 2 4
3 4

2 y 2 5
x z t
5 2

(d) (3 points) Suppose we perform one more iteration of the Ford Fulkerson algorithm, starting with
the above flow. Give any one corresponding augmenting path of that iteration.

Solution:
6 y 2 3 7
s z v t

(e) (3 points) Let f 0 be the flow obtained by augmenting along the path in the previous part. Draw
the residual network corresponding to f 0 .

Solution:

Page 3 of 7
Name: Rollno:

5 3 4
s u v w

4 5
5 2 8 2 1 2 4
5 6

2 y 4 5
x z t
5

(f) (3 points) What is the value of the maximum flow in this network? 15
(g) (3 points) List the vertices in the source side of the st-cut corresponding to this max flow? What
is the capacity of this cut?

Solution: {s, u, x, y}
Capacity of the cut = 15

Question 3. Let G be a DAG. A sink path from a vertex u is a path that ends in a sink. v is a terminal
vertex for a vertex u, if v is a sink and there is a path from u to v.
(a) (6 points) Design an O(n + m) time algorithm NumSink(G) that given a DAG G computes the
number of sink paths from every vertex in G. The number of sink paths from a sink is 1.

Solution: We will store the number of sinks from a vertex v in the array S[v].
Algorithm 1: NumSink(G)
Data: Given a DAG G
foreach sink vertex u in G do
Set S[u] = 1;
end
Perform a topological ordering of G, say v1 , v2 , . . . vn ;
for i = n − 1 to 1 do
if vi is not a sink
P then
Set S[vi ] = (vi ,vj )∈G S[vj ];
end
end

(b) (6 points) Design an O(n + m) time algorithm NumTerminus(G, u) that computes the number of
terminals for the vertex u in the DAG G.

Page 4 of 7
Name: Rollno:

Solution:
Algorithm 2: NumTerminus(G,u)
Data: Given a DAG G and a vertex u
Set c = 0;
Perform DFS(u) on G;
Whenever a sink is reached that is unvisited, increment c by 1;
return c;

Question 4. Prove or disprove the following statements.

(a) (6 points) If G is a directed graph containing a path from u to v, and D(u) < D(v) in a DFS of G,
then v is a descendent of u in the corresponding DFS forest.

Solution: The statement is false. Consider the following graph.

w

u v

Suppose DFS visits w first, and among the neighbours of w, it visits u followed by v. In this
case D(u) < D(v). Also there is a directed path from u to v. However the DFS forest looks as
follows.
w

u v

(b) (6 points) A directed graph G is acyclic if and only if DFS(G) produces no back edges.

Solution: If DFS(G) produces a back edge, say (v, u), then by definition of back edge v is a
descendent of u in the DFS forest of G. Therefore, we get a cycle by traversing from u to v in
G via the tree edges and then going from v to u via the back edge (v, u).
Suppose G has a cycle, say C. Let u be the first vertex to be discovered by the DFS in C. Since
none of the other vertices in C have been discovered till this point, they would be descendants
of u in the DFS tree. Since C is a cycle, there is a vertex v in C such that (v, u) is an edge.
Now since v is also a descendant of u in the DFS tree, this edge will become a back edge in the
DFS.

Question 5. Consider an n node complete binary tree T , where n = 2d − 1 for some d. Each node v ∈ T is
labeled with a real number xv . You may assume that all the values are distinct. A node v ∈ T is said
to be a local minima if the label xv is less than the label xw for all nodes w that are joined to v by an
edge.
You are given such a complete binary tree T , but the labelling is only specified in the following implicit
way: For each node v, you can determine the value xv by querying the node v.
(a) (6 points) Give an algorithm to find a local minimum of T using only O(log n) queries to the nodes
of T .

Page 5 of 7
Name: Rollno:

Solution:
Algorithm 3: TreeLocalMinima(T)
Data: Given a complete binary tree T
Set u to be the root of T ;
while u is not a local minima do
Move to a child v of u such that xv < xu ;
Set u = v;
end
return u;
Since the depth of a complete binary tree is log n the time complexity is O(log n) as well.

(b) (6 points) State the proof of correctness of your algorithm.

Solution: There always exists a local minima in T , since the global minima is also a local
minima. In every iteration of the while loop, the value of the current node decreases. Since
the number of nodes in T is finite, therefore the algorithm terminates. Also by the condition
of the while loop, when the algorithm terminates, it must be at a local minima.
The depth of a complete binary tree in log n. By design, the algorithm traces a path from the
root to a leaf in the worst case. Hence the time is O(log n).

Question 6. Given a string s, an operation on s is defined as follows: Either (i) inserting a single character
into s, (ii) deleting a single character from s, or (iii) replacing a character in s with another character.
For two strings s1 and s2 , the edit distance of s1 and s2 , ED(s1 , s2 ) is defined to be the minimum number
of operations to convert s1 to s2 .
For example, ED(kitten, sitting) = 3. We can (i) Replace ‘k’ with ‘s’, (ii) Replace ‘e’ with ‘i’, (iii) Insert
‘g’ at the end.
(a) (3 points) Prove that ED(s1 , s2 ) = ED(s2 , s1 ).

Solution: This is true because, the inverse of the insert operation is delete, and vice versa.
And the inverse of the replace operation is itself. Therefore any sequence of operation that
converts s1 to s2 , can be replaced with the corresponding inverses, to convert s2 to s1 .

(b) (6 points) Give a suitable recursive formulation of ED(s1 , s2 ). (Do not forget the base case)

Solution: Let s1 be a string of length n and s2 be a string of length m. Let s1 [1 . . . i] denote

the prefix of s1 consisting of the first i characters. Similarly s2 [1 . . . i] denote the prefix of s2
consisting of the first i characters.
Let f (i, j) be the minimum number of operations required to convert s1 [1 . . . i] to s2 [1 . . . j].
Then in this notation, ED(s1 , s2 ) = f (n, m).
For i, j ≥ 1 we have:


 f (i − 1, j − 1), if s1 [i] = s2 [j]
1 + f (i, j − 1), if s1 [i] 6= s2 [j] (corresponds to inserting a character)

f (i.j) =

 1 + f (i − 1, j), if s1 [i] 6= s2 [j] (corresponds to deleting a character)
1 + f (i − 1, j − 1), if s1 [i] 6= s2 [j] (corresponds to replacing a character)


Base Case: If i = 0, then f (i, j) = j and if j = 0, then f (i, j) = i.

Page 6 of 7
Name: Rollno:
(c) (5 points) Given s1 and s2 , provide a dynamic programming solution to compute ED(s1 , s2 ).

Solution: We create a matrix A of size (n + 1) × (m + 1) such that A[i, j] = f (i, j). We then
fill the matrix and output the entry A[n, m].
Algorithm 4: EditDistance(s1 , s2 )
Data: Given two strings s1 and s2
for i = 0 to n do
A[i, 0] = i;
end
for j = 0 to m do
A[0, j] = j;
end
for i = 1 to n do
for j = 1 to m do
if s1 [i] == s2 [j] then
A[i, j] = (f (i − 1, j − 1);
else
A[i, j] = 1 + min(f (i − 1, j − 1), f (i − 1, j), f (i, j − 1));
end
end
end
return A[n, m];

Page 7 of 7