CS 213 M 2023

Data Structures
Instructor: Milind Sohoni
2 quizzes, 1 midsem, 1 endsem.

Triple Engines: NPTEL, Teams, SAFE

TAs: Prateek, Jitesh, Abhishek, Meghna, Paolami, Gunjan
Slot 5: Wednesday and Friday, 9:30-11.

Event Dates

First Day of instructions Wednesday, 4th January 2023

Quiz 1 Wednesday, 1st February 2023

Mid-Semester Exam 18th-25th February 2023

Quiz 2 Friday, 24th March 2023

Last Date of Instructions Wednesday, 12th April 2023

End-Semester Exam 15th-25th April 2023

Microsoft Teams Link:

link
Team Code: 9nm4alc

Course Plan

We will be following a “flipped classroom” mode and use the lectures of Prof. Naveen Garg,
from IIT Delhi as the base. Besides this, we will also use 3-4 additional topics which expand on
the use of data structures in everyday applications. This thread will be called the ASD or
Applications of Data Structures thread. These will be developed in parallel to the standard
material. The main link for the standard material is:

Introduction to Data Structures and Algorithms

The topics list is as follows:

Modules / Lectures

Data Structures And Algorithms (link)

Our general strategy will be to introduce the material of the week in the Wednesday class, and
to solve previous week’s tutorial, to discuss any difficulties in the Friday class and assign a new
tutorial. Also, in every alternate week, we will introduce Applications of Data Structures (ADS)
Topics.

Week 1

ADS 1: What is data and how is it structured. module1short.pdf

● Introduction to Data Structures and Algorithms

Algorithms and data, size of data, running time, pseudo code, insertion sort and its
running time, best case, worst case, average case, order notation. Lecture - 1
Introduction to Data Structures and Algorithms

● Stacks
Abstract Data Types, Dynamic Set, The Stack and its rules, Array Implementation, An
application - Span of daily stock prices, an array implementation O(n^2) and a stack
implementation O(n).Lecture - 2 Stacks

● Queues and Linked Lists [0:40 minutes]

The Queue as an abstract data type, Its implementation as circular array, the one-sided
linked list operation, doubly linked list. Lecture - 3 Queues and Linked Lists

Questions.
1. What is the time complexity of binary addition and multiplication? How much time does it
take to do unary addition?
2. Recall the Big-O notation: f(n) is O(g(n)) if there is a constant M such that f(n)<= M.g(n).
To prove that f(n) is not O(g(n)) we need to show “for every constant M, there is an n
such that f(n)>M.g(n)”. Give f(n)=a_0 n^0 +...+ a_d n^d and g(n) =b_0 n^0 +...+ b_e n^e
with d>e, show that f(n) is not O(g(n)).
3. If f(n)<=F(n) and G(n)=>g(n) (in order sense) then show that h(n)=f(n) / G(n)<=F(n) /
g(n). Is f(n) the same order as g(n)=f(n)|sin(n)|?
4. There is a stack of dosas on a tava, of distinct radii. We want to serve the dosas of
increasing radii. Only two operations are allowed: (i) serve the top dosa, (ii) insert a
spatula (flat spoon) in the middle, say after the first k, hold up this partial stack and flip it
upside-down and put it back. Design a data structure to represent the tava, input a given
tava, and to produce an output in sorted order. What is the time complexity of your
algorithm?
 This is also related to the train-shunting problem.

5. One problem is to check if a string of opening and closing brackets of various types is a
valid (balanced) bracket sequence. For example {(){}[[]]} and ([[[]]]{}) are valid, that is
balanced, while (){) and (())) are invalid (unbalanced). Write a stack based algorithm to
check if the brackets are balanced or not.
6. The mess table queue problem: There is a common mess for K hostels. Each hostel has
some N1,...,Nk students. These students line up to pick up their trays in the common
mess. However, the queue is implemented as follows: If a student sees a person from
his/her hostel, she/he joins the queue behind this person. This is the “enqueue”
operation. The “dequeue” operation is as usual, at the front. Think about how you would
implement such a queue. What would be the time complexity of enqueue and dequeue?
Do you think the average waiting time in this queue would be higher or lower than a
normal queue? Would there be any difference in any statistic? If so, what?
7. Let h[0:N] be an array which stores the heights of N+1 equidistant points in a straight
line. We would like to know the list of points i which are visible from the top of h(0). Give
an algorithm for printing this list. How much time does this take?

Week 2

ADS 2: Dictionaries and Simple Text Processing dictionary.pdf

 ● Dictionaries
The dictionary as an ADT. Key-Element pairs. Ordering of Keys. The basic operations of
Find, Insert and Delete. Array implementation and binary search. Its drawbacks. Sparse
keys and the hash table. Collisions, efficiency and load factor alpha. Chaining.Lecture -
4 Dictionaries

● Hashing [0:40 minutes]

Hashing functions, Linear hash tables. Lecture - 5 Hashing

● Trees
Basic notation and examples. Binary trees, ordering of nodes. Recursive definition.
Height, number of nodes and basic statistics. Implementing trees. Unbounded
branching. Lecture - 6 Trees

Questions
1. Given that k elements have to be stored using a hash function with target space n. What
is the probability of the hash function having an inherent collision? What is an estimate
of the probability of a collision in the insertion of N elements?
2. Review linear chaining (where collisions are stored in the same array by locating the
next available spot) and its correctness. Do this in the following steps (i) specify what
does the chain C(h) of array indices which are to be queries to look for a single hash
value, (ii) how does this chain extend by an insertion, and how does it change by a
deletion. Is there a way of not using tombstones?
3. In Naveen’s version, a search for an index value k ends when an empty cell is
encountered. What if we mark the last element for an index as specialand terminate the
search when this last element is encountered? Would this work? Would this be efficient?
4. Suppose you want to store a large set of pairs of numbers (a_i ,b_i ), for example,
(name,address). You have operations which are addition, deletion and inspection of
elements in this set. You also have queries whether a particular name or address is
there in the set, and if so then count them and delete all such entries.How would you
design your hash tables?
5. Suppose a binary tree has n nodes and has the property that either a node has 0
children or 2 children. What can be its maximum height?
6. Given a tree with a maximum number of children as k. We give a label between 0 and k-
1 to each node with the following simple rules. (i) the root is labeled 0. (ii) For any vertex
v, suppose that it has r children, then arbitrarily label the children as 0,..,r-1. This
completes the labeling. For such a labeled tree T, and a vertex v, let seq(v) be the labels
of the vertices of the path from the root to v. Let Seq(T)={seq(w)| w in T} be the set of
label sequences. What properties does Seq(T) have? If a word w appears what words
 are guaranteed to appear in Seq(T). How many times does a word w appear as a prefix
of some words in Seq(T).
7. Suppose that you are given a binary tree, where, for any node v, the number of children
is no more than 2. We want to compute the mean of ht(v), i.e., the mean height of nodes
in T. Write a program to compute the mean height.
8. Suppose that we are given the heights of all leafs in the tree. Can we compute the mean
height? Given a sequence (n1,n2,...,nk) of the heights of k leaves, is there a binary tree
with exactly k leaves at the given heights?
9. For a binary tree T and a node v, let len(v)= length of the longest path from v in the
subtree rooted at v. Write a program assignlen(T) to assign length to every vertex in the
tree with root T.
10. For a binary tree pointed to by T with usual node structure, let us define a new attribute
node.parent which points to the parent. The parent of the root is null. Write a program to
assign parents to all nodes in a tree pointed to by T.
11. For a tree and two nodes v and w, the lowest common ancestor is the node x such that
both v and w are descendents of x, but belong to different subtrees of x. Assume that a
tree T has parent pointers set as above. Write a function x=lca(v,w,T) which returns the
lowest common ancestor if v and w. What is the time complexity of your program?
12. Given a rooted tree T, define the vector f(T) where f(T)(i) is the number of nodes at
depth i from the root. Write a recursive definition of f(T) in terms of its children. Now write
pseudo code to compute the vector f(T). How much time does this take?
Use the type node, where T.nos is the number of children, and T.child[i] is a pointer to
the i-th child.
13. Here is an application of the above. Trees and Water

Week 3

Tutorial 2.pptx
Binary Search Trees

● Tree Walks / Traversals

Preorder and postorder traversals, evaluation of expressions, In-order traversal, Euler
tour of a tree and expression with parentheses, building a tree from two traversals,
limitations. Special case. Lecture - 7 Tree Walks / Traversals

● Ordered Dictionaries
The simple dictionary - min, max,pred, succ, search. List and array implementation.
Insert and delete. Binary search trees. Implementing search, min and succ. The simple
insert at the leaf level. Time complexity. Order of insertion and heights.. Lecture - 8
Ordered Dictionaries
 ● Deletion[0:25]
The deletion - 3 cases, i.e., 0 leaves, 1 leaf, 2 leaves. Maintaining the inorder invariant.
Inorder traversals. Sorting using a BST. Best and worst case. Lecture - 9 Deletion

[0:25 onwards]: Argument on why, on the average, insertion of a random permutation

takes O(nlogn) time.

Questions. Tutorial 3. (Also see problems from TimeTable lecture)

1. Given the elements [1,2,...,7] and the complete binary tree T with 7 nodes, label the
nodes so that the preorder, inorder and postorder traversals produce the sequence
1,2,...,7 in that order.
2. Construct the BST tree T whose post-order traversal is 1 3 5 4 2 7 8 6. For this tree
delete the element 4 in two ways - by using its predecessor and its successor. Display
these trees.
3. For a given binary tree let prevloc(T,a) give the location of the position p such that
p.value will appear just before a.value in the inorder printing of T. Give a program to
compute prevloc. Assume that parent pointers exist.
4. Consider a binary tree with labels such that the postorder traversal of the tree lists the
elements in increasing order. Let us call such a tree a post-order search tree. Describe
how you will do search, min, max, insert and delete on this tree. Please write pseudo-
code. Do not see the solution below.
5. Given a BST T and an element a, the task is to delete all elements b<a from T. Write
pseudocode to do this. How much time does your algorithm take? What is the structure
of the tree left behind? What is its root?
6. Given a BST tree T, and a value v, write a program to locate the leftmost and rightmost
occurrence of the value v.
7. Let H(n) be the expected height of the tree obtained by inserting a random permutation
of [n]. Write the recurrence relation for H(n).
8. Let [a(1),...,a(n)] be a random permutation of n. Let p(i) be the probability that a(a(1))=i.
Compute p(i).

Week 4
Binary Search Trees
● Quick Sort
The algorithm and its average case analysis. The average permutation argument re-
visited and the O(nlogn) time complexity. Lecture - 10 Quick Sort

● AVL Trees - Basics

 The basic definition - height balance. The number of nodes in an AVL tree of height h.
Two methods of estimating the number of nodes. Insertion and the loss of balance.
Lecture - 11 AVL Trees

● AVL Trees
Insertion with the LL rotation. The LR case and double rotation. Deletion - reduction to
the case of leaf deletion. Why does it percolate up? Running time. See
https://www.cse.iitb.ac.in/cs213d/bst.pdf for a summary. Lecture - 12 AVL Trees

Questions. Tutorial 4.
Tutorial 4
1. In the ordinary BST, let us have an additional variable to be stored at the node, which is
the height of the subtree rooted there. Thus every node x has the following attributes:
left, right, value, height. Write pseudo-code for insert and delete in BST with the updation
of height. Be careful with delete.
2. This is to ensure that you understand the analysis of the “random permutation”. In
Quicksort, we had said that if [a(1),a(2),...,a(n)] is a random permutation, then the
probability that a(1)=i is precisely 1/n. Now, for a permutation as above, let b=a(a(1)). In
other words, if a=[2,3,1,4,5] then b=3, but if a=[5,4,3,2,1] then b=1 and so on. If I were to
select a random permutation, then what is the probability p(i) that b=i. Warning: p(i) is
not the same for every i.
3. Consider the structure below. Is the tree structure AVL? If we wish to store the set
1, . . . , 22, label each node with the correct number.

4. Now add 23 to the set and then delete 1. Also do the same in the reverse order. Are the
answers the same? When will the answers be the same?
5. Why is it that in insert, a single or double rotation is sufficient to balance the tree, but not
in deletion?
 6. Experiment with other definitions of balance. We say that a tree is balanced iff every
node is balanced. We say that a node T is balanced iff either (A) |TL|, |TR|<B for some
pre-fixed constant B, or (B) there is an alpha such that alpha <= |TL|/|TR| <= 1/alpha.
Show that if a tree is balanced then there is a constant beta >1 such that the number of
nodes in T are greater than betan^h, where h is the height.

Week 5

Strings and string processing. Lexicographic order and the next permutation. Words and
word processing. State machines.
strings
● Case Study: Searching for Patterns
The exact substring problem. The longest possible prefix which is a suffix at position i.
The suffix short-cut and the KMP algorithm. Successful and unsuccessful comparisons.
The running time O(m+n) assuming the h matrix. Lecture - 17 Case Study: Searching
for Patterns

● Tries
The Trie data structure. Motivation: Processing strings, computation of h matrix. Trie for
a subset of words. Basic operations. Optimization. Size of the trie. Tries and the internet.
Suffix trie and its construction. Lecture - 18 Tries

Questions. Tutorial 6.

1. This is concerning Prof. Garg’s argument that the total number of unsuccessful
comparisons = shifts in KMP algorithm is no more than |T|. Is this strictly true?
2. Please review the KMP algorithm to see how it detects two overlapping occurrences of
the pattern.
3. For a pattern P[1...n], h(i) is defined to be the smallest k>0 such that P[1]=P[k+1],..., P[i-
k]=P[i], but P[i-k+1]!=P[i+1]. If there is no such match, define h(i) to be i. Fill in the table
below.

1 2 3 4 5 6 7 8 9

P b a b b a a b b a

h(i)
 4. Suppose that there is a letter z in P of length n such that it occurs in only one place, say
k, which is given in advance. Can you optimize on the computation of h?
5. Compute the Suffix Trie for abracadabra$. Compress degree 1 nodes. Use substrings as
edge labels. Put a square around nodes where a word ends. Use it to locate the
occurrences of abr.
6. Review the argument that for a given text T, consisting of k words, the ordinary Trie
occupies space which is a constant multiple of |T|. How is it that the suffix tree for a text
T is of size O(|T|^2)?

Week 6-7

● Data Compression - Huffman Codes

Coding texts for compression. Weighted length as an optimization problem. The
Huffman coding tree. Lecture - 19 Data Compression

● Priority Queues
Examples of the need for priority queues. The basic delmin and insert operations. The
array implementations, the binary heap. The heap condition and insert. Lecture - 20
Priority Queues

● Binary Heaps
Delmin, bubble-up and bubble-down and its analysis. The heap sort and its analysis.
The delicate time analysis of heapify and heap-sort. Lecture - 21 Binary Heaps

● Merge Sort.
The Divide and Conquer Approach. Analysing recurrence relations. Lecture - 22 Why
Sorting
Tutorial 7

1. In an Huffman code instance, show that if there is a character with frequency greater
than ⅖ then there is a codeword of length 1. Show that if all frequencies are less than ⅓
then there is no codeword of length 1.
2. Suppose that there is a source which has three characters {a,b,c}. The output of the
source cycles in the order of a,b,c followed by a again, and so on. In other words, if the
last output was a b, then the next output will either be a b or a c. Each letter is equally
probable. Is the Huffman code the best possible encoding? Are there any other
possibilities? What would be the pros and cons of this?
3. Consider the following table of letters and frequency. Design a Huffman code tree.

a 20 d 7 g 8 j 4
 b 6 e 25 h 8 k 2

c 6 f 1 i 12 l 1

4. Can a Priority Queue be implemented as an AVL tree? What advantages does a Heap
implementation have over an AVL tree implementation?
5. The Heap implementation needs us to find the “last’ element in the heap. Write a code
snippet to maintain the last element. Suppose we maintain a pointer to the last element,
write a code snippet to go to the previous one. This will be useful if there are a sequence
of deletes.What is the worst and the average time complexity of locating the previous?
What happens if we do not maintain the last element? How do we locate the last
element?
6. Suppose we have a 2D array where we maintain the following conditions: for every (i,j),
we have A(i,j)<=A(i+1,j) and A(i,j)<=A(i,j+1). Can this be used to implement a priority
queue?

1 1 2 3 Q

2 2 4 Q Q

3 4 6 Q Q

5 5 8 Q Q

7 8 Q Q Q

7. Here is an interesting problem from Prof. Abhiram Ranade. A piecewise linear function
on [0,1] may be represented by a sequence of special x and y values such as the table
below:

x 0 0.3 0.5 1

f(x) 1.3 1.2 4.1 0.3

Thus, a good representation is f.NumberOfSegments, f.x (array of x values), f.y (array of

y values). Now given two functions f and g, let h be the minimum of f and g. Clearly, it is
also a piecewise linear function. Compute the representation of h.
Week 9
ADS 3: The South-Western Railway Timetable. SWRTimeTable.pdf

● Discussion on Timetables - data structures and problems.

Timetables
Sample input file.
https://drive.google.com/file/d/1uJzobM1ChLpJ_HW7x4gHscDEyjv1z_-L/view?
usp=sharing

● Graphs
Graphs - the basic definitions. Directed and undirected graphs. Paths and cycles.
Subgraphs. Connectedness, connected components. Trees and forests. Number of
edges and connectivity. Spanning trees. Nice (Eulerian) cycles. Degree of a node. The
graph ADT. Various methods.Lecture - 24 Graphs

● Data Structures for Graphs

The Edge-Vertex adjacency list. The adjacency matrix and the traditional adjacency list.
Various extensions of the adjacency list. The Breadth First Search. Examples and the
queue implementation. The O(E+V) time analysis. Predecessors and the BFS tree.
Tree and non-tree edges. Lecture - 25 Data Structures for Graphslec25.ppt.pdf

● Two Applications of Breadth First Search

Connected components and shortest paths. Bipartite graphs, odd cycles and detection
using BFS. Diameter of a graph. A crude algorithm. The adjacency matrix.Lecture - 26
Two Applications of Breadth First Search

Tutorial 8
1. The graph is an extremely useful modeling tool. Here is how a Covid tracing tool might
work. Let V be the set of all persons. We say (p,q) is an edge (i) in E1 if their names
appear on the same webpage, and (ii) in E2 if they have been together in a common
location for more than 20 minutes. What significance does the connected components in
these graphs and what does the BFS do? Does the second graph have epidemiological
significance? If so, what? If not, how would you improve the graph structure to get a
sharper epidemiological meaning?
2. Let us take a plane paper and draw circles and infinite lines to divide the plane into
various pieces. There is an edge (p,q) between two pieces if they share a common
boundary of intersection (which is more than a point). Is this graph bipartite? Under what
conditions is it bipartite?
3. There are three containers A, B and C, with capacities 5,3 and 2 liters respectively. We
begin with A having 5 liters of milk and B and C being empty. There are no other
measuring instruments. A buyer wants 4 liters of milk. Can you dispense this? Model this
as a graph problem with the vertex set V as the set of configurations c=(c1,c2,c3) and an
edge from c to d if d is reachable from c. Begin with (5,0,0). Is this graph directed or
undirected? Is it adequate to model the question: How to dispense 4 liters?
4. Suppose that there are M workers in a call center for a travel service which gives travel
directions within a city. It provides services for N cities - C1,...,CN. Not all workers are
familiar with all cities. The number of requests from each city per hour are R1,...,RN. A
worker can handle K calls per hour. Is the number of workers sufficient to address the
demand? How would you model this problem? Assume that R1,...,RN and K are small
numbers.
5. Study the BFS code from Prof. Naveen’s slides. Argue that at any time during the
running of the algorithm the d-values of vertices in the queue are non-decreasing and
can only take 1 or 2 consecutive values.
6. Look at the code below. List the properties of the white, grey and black vertices. In line
10, while processing u, can we encounter a vertex v which is gray? Or black? In such
cases what are the values possible for d[v]?
7. There are many variations of BFS to solve various needs. For example, suppose that
every edge e=(u,v) also has a weight w(e) (say the width of the road from u to v).
Assume that the set of values that w(e) can take is small. For a path p= (v1,v2,...,vk), let
the weight w(p) be the minimum of the weights of the edges in the path. We would like to
find a shortest path from a vertex s to all vertices v. If there are multiple such paths, we
would like to find a path whose weight is maximum. For example, in the graph below,
we would prefer path s->w->v. Can we adapt BFS to detect this path?

BFS(G,s)
01 for each vertex u ∈ V[G]-{s}

02 color[u] ← white
 03 d[u] ← ∞

04 π[u] ← NIL

05 color[s] ← gray

06 d[s] ← 0

07 π[u] ← NIL

08 Q ← {s}

09 while Q ≠ ∅ do

10 u ← head[Q]

11 for each v ∈ Adj[u] do

12 if color[v] = white then
13 color[v] ← gray

14 d[v] ← d[u] + 1

15 π[v] ← u
16 Enqueue(Q,v)
17 Dequeue(Q)
18 color[u] ← black

Week 10

GIS: Basic introduction to a GIS system. Maps and Applications. The stereographic projection
and X-Y plane. Raster and vector data. Representation of vector data - points, poly-lines and
polygons. Orientation of polygons. Membership within polygons. A running introduction to QGIS.
Some elementary operations. gis.pdf

Anthrokrishi. An introduction to the the project. The IT stack for agriculture. Polygon
segmentation. Analysing recognition. The variance of a polygon. The two-image analysis. The
cadastral-farm plot alignment problem. The excess function and its minimization. BFS as the
basic framework. Anthrokrishi CS213 Presentation.pdf

Tutorial 10.

1. The basic operations are seg-segment intersections and angle computations. These are
specified as follows: [flag, pt]=intersect(p1,p2,q1,q2) computes the intersection of the
segment [p1,p2] and [q1,q2]. flag=0 denotes no intersection. fla=1 denote an intersection
and pt is the point of intersection. theta=angle(v): where v is a 2D vector, angle returns
theta in [0, 2*pi) such that v.x=r*cos(theta) and v.v=r*sin(theta), where r is the length of
v.
 2. Let v1 and v2 be two non-collinear vectors. Let theta=rotation(v1,v2) be the angle
through which v1 is rotated to align with v2. Theta is in (-pi, pi).
3. Given a list of points P=(p1,...,pn) determine if they determine the vertices of a polygon
of anti-clockwise orientation with vertices in the order given.
4. Given a polygon P and a point pt, write a program to determine if pt is in the interior of P,
or is on it, or is in the exterior of P.
5. Given a polygon P and a segment S=(s1,s2), determine if S divides P into two parts. If
so, express the two parts as polygons P1 and P2.
6. Let P and Q be two polygons. Assume that no three points of P or Q together are
collinear. Outline an algorithm to compute the union of P and Q.

Week 11

● Review of Graph Theory.

Paths and vertex connectivity as an equivalence class. Implementing transitivity. Uses of
connectivity - game graphs. Cycles. Acyclic graphs. How to detect if a graph is acyclic.
Trees as the interface between acyclicity and connectedness. Spanning trees. Non-tree
edges and the exchange. The BFS tree. Other modifications of BFS. The DFS and 2-
edge-connectedness. Graphs-1 simple connectivity, bfs, dfs (elementary).pptx

● Depth First Search.

The recursive program. The global and the local variables. DFS numbering. DFS tree.
Tree and non-tree edges. Back edges and order of visiting. Arrival and departure times.

global counter
function DFS(G, v) is
label v as discovered
ar[u]=counter; counter=counter+1;
for all directed edges from v to w that are in G.adjacentEdges(v) do
if vertex w is not labeled as discovered then
//inlabel[(v,w)]=counter;
recursively call DFS(G, w) // outlabel[(v,w)]=counter
Endif
Endfor
dep[u]=counter; counter=counter+1;
Endfunction

Lecture - 27 Depth First Search

● Applications of DFS.
 Adjacency list implementation and a non-recursive implementation. Timing analysis. 2-
edge-connectedness, i.e., deletion of any edge keeps the graph connected and the
question of checking that every edge is in some cycle. Use of dfs to check this. Checking
if a tree-edge is essential.
Deepest back edge. Planar graphs. 2-vertex connected.
Lecture - 28 Applications of DFS

Tutorial 11

1. Please study the undirected graph below (on the left) and the tree edges (in black), non-
tree edges in green (and some forbidden edges in red, which should not be considered).
We have seen that this gives a tree, in fact, a shortest-path tree for the vertex s. In other
words, it is a tree T for which the path from s to another vertex u is the shortest path in
the original graph. Thus bfs produces a shortest path tree. Note that there may be many
shortest path trees and the particular tree produced by bfs is one of them, which
depends on the vertex ordering, i.e., how the vertices are ordered among themselves.
Now, given a shortest path tree T, is there always an ordering of the vertices which will
produce this tree T as outputs?

2. For the same graph above, and on the left, execute dfs and produce a trace of the
algorithm. List the vertices in order of push and pop events and mark the arrival and
departure numbers. Modify the DFS to label tree edges twice with the counter, once
while going forward, i.e., push (i.e., the recursive call) and the second time when
popping, i.e., when exiting.

3. Suppose that there is an undirected graph G(V,E) where the edges are colored either
red or blue. Given two vertices u and v. It is desired to (i) find the shortest path
irrespective of colour, (ii) find the shortest path, and of these paths, the one with the
fewest red edges, (iii) a path with the fewest red edges. Draw an example where the
above three paths are distinct. Clearly, to solve (i), BFS is the answer. How will you
design algorithms for (ii) and (iii)?
4. Suppose that we have a graph and we have run dfs starting at a vertex s and obtained a
dfs tree. Next, suppose that we add an edge (u,v). Under what conditions will the dfs
tree change? Give examples.

5. We have a new search algorithm which uses a set S for which we have two functions (i)
add(x,S) which adds x to S, and (ii) y=select(S) which returns an element of S following
a certain rule.

Function mysearch
Global visited;
For all u visited[u]=false;
for all edges e, found[e]=false;
S=empty;
add(s,S); nos=1; record[nos]=s;
While nonempty(S)
y=select(S)
nos=nos+1; record[nos]=y;
For all v adjacent to y
If visited[v]==false
visited[v]=true;
found[(u,v)]=true;
add(v,S);
Endif;
Endfor;
Endwhile
Endfunction;

(i) Compare the bfs and dfs algorithms with the above code. Take special care in
understanding visited.
(ii) Let us look at the sequence record[1],record[2],...,record[n]. Show that there is a path
from record[i] to record[i+1] using only edges which have been found at that point.
(iii) Compare BFS and DFS in terms of the above path lengths.

6. This is the theoretical basis of edge-2-connectedness. Let G(V,E) be a graph. We define

a relation on edges as follows: two edges e and f are related (denoted by e~f iff there is
a cycle containing both. Show that this is an equivalence relation. The equivalence class
[e] of an edge e is called its connected component.

7. Given a dfs tree for an undirected graph and a vertex v, we define C(v) as the edges
(x,y) where y is a descendent of v while x is a parent of v, with x!=v and y!=v. Modify dfs
 to list C(v) and to compute |C(v)| . In the example below C(6)=1, C(5)=0 and C(1) =3.
How much time does your algorithm take? Modify the code counterdfs.c

Week 12.
● DFS in Directed Graphs
The non-tree edges and the 4 types. The theorem of cycles and back-edges. Ordering
by departure times. An acyclic graph and its topological sort. Strongly connected graphs
and components. The O(mn) time algorithm to check strong-connectedness. Lecture -
29 DFS in Directed Graphs

● Applications of DFS in Directed Graphs

Checking strong connectedness - the 2 dfs algorithm by graph reversal. The 1-dfs
connected component algorithm of back-edges. Ordering by arrival times.
Lecture - 30 Applications of DFS in Directed Graphs

Tutorial 12.

1. One question about undirected graphs. Recall that an articulation vertex v is a vertex
whose removal disconnects the graph. Use dfs to label all articulation vertices in a
graph.
2. Run the smallest arrival time algorithm for the example graphs for DFS starting at A and
proceeding lexicographically. If you have not exhausted all vertices, start at the next
unvisited vertex in lexicographic order. Record this time in a table.

3. Compute the strongly connected components of graphs (i) and (ii) below.

4. Use the graph reversal algorithm to detect if (i) and (ii) are strongly connected.

5. We wish to detect if an edge e in a directed graph is contained in a cycle. Give an

algorithm. Can dfs be adapted to label all edges e which are contained in some cycle in
the graph?

6. Recall the definition of strong connectedness. Let us define v~w iff there is a path from v
to w and a path from w to v. This is an equivalence relation on vertices. Let [v1]…[vk] be
the equivalence classes of ~. We now define a new component graph G’(V’,E’), where
V’={ [v1],...,[vk]}. We say ([vi],[vj]) is an edge in the new graph iff there is a path from vi
to vj in the original graph.
(A) Show that this new graph G’ has no cycles.
(B) Show that the edge ([vi],[vj]) does not depend on the representative vi of [vi].
(C) Compute G’ for the two example graphs.

7. Suppose that a graph G is strongly connected. An edge e is such that G’=G-e does not
remain strongly connected. How would the component graph of G’ look like and how
would the edge e=(u,v) be located in it?

8. A city is given by a directed graph G(V,E). Its public transport runs along directed paths
pi(1),...,pi(k) in this directed graph. For each path pi(i), there is a timetable given for the
buses on this path. Given a source u and destination v find if it is possible to go from u to
v using public transport. Find a plan using the minimum number of hops. Also, how
would you find the time table from u to v?

9. A good way to study a function f:X->X is to study its graph, G(f), whose vertex set is X
and edge set is E={(x,y)|y=f(x)}. Let X={0,1,..,8} and let g(x)=x^2+1 mod 7. Draw the
graph of G(g). Can you describe the structure of G(f) when f is a bijection? Or for a
general f?
Week 13.

● Single Source Shortest Paths

The s-t shortest path problem. Enumerating paths and its complexity. Property of
shortest paths. The incremental algorithm. The specification of the Dijkstra’s algorithm.
Lecture - 34 Single Source Shortest Paths

● Correctness of Dijkstra's Algorithm

The inductive argument. Running time using priority queues. The problem with negative
edges. A counter example.
 Lecture - 35 Correctness of Dijkstras Algorithm

Tutorial 13.
1. Modify Dijkstra’s algorithm to compute the number of shortest paths from s to every
vertex t.
2. Compute the shortest path for all vertices starting from A. Do this in tabular form.

3. Show an example of a graph with negative edge weights and show how Dijkstra’s
algorithm may fail. Suppose that the minimum negative edge weight is -d. Suppose that
we create a new graph G’ with weights w’, where G’ has the same edges and vertices as
G, but w’(e)=w(e)+d. In other words, we have added d to every edge weight so that all
edges in the new graph have edge weights non-negative. Let us run Dijkstra on this
graph. Will it return the shortest paths for G?

4. Given a graph G(V,E) and a distinguished vertex s and a vertex v, there may be many
shortest paths from s to v. What shortest path is identified by Dijkstra?

5. Look at the following graph from Tutorial 9 with red edges and blue edges. Our task was
to find the path from s to every vertex t, with the fewest red edges. Run any modified bfs
of your choice and Dijkstra and compare the sequence of vertices visited by BFS and by
Dijkstra.
6. Let G(V,E) be a representation of a geography with V as cities and (u,v) an edge iff there
is a road between the cities u and v. Let d(u,v) be the length of this road. Suppose that
there is a bus plying on these roads with fare f(u,v)=d(u,v). Next suppose that you have
a free coupon which allows you one free bus ride. Find the least fare paths from s to
another city v using the coupon for this travel.

7. Same as above. Suppose w(u,v) is the width of the road between the cities u and w.
Given a path pi, the width w(pi) is the minimum of widths of all edges in pi. Given a pair
of cities s and u, is it possible to use Dijkstra to determine d such that it is the largest
width of all paths pi from s to v?

8. Same as above. For a path pi, define hop(pi)=max(d(e)) for all e in pi. Thus if one is
traveling on a motorcycle and if fuel is available only in cities, then hop(pi) determines
the fuel capacity of the tank of your motorcycle needed to undertake the trip. Now, for
any s and u, we want to determine the minimum of hop(pi) for all paths pi from s to u.
Again, can Dijkstra be used?

9. You are given a time table for a city. The city consists of n stops V={v1,v2,...,vn}. It runs
m services s1,s2,...,sm. Each service is a sequence of vertices and timings. For
example, the schedule for service K7 is given below. Now, you are at stop A at 8:00am
and you would like to reach stop B at the earliest possible time. Assume that buses may
be delayed by at most 45 seconds. Model the above problem as a shortest path
problem. The answer should be a travel plan.

Service : K7

H15 Convocation Hall Market Gate H15

7:15am 7:20am 7:30 7:40