So you want to be a Competitive Programmer?

Indian National Olympiad in Informatics 2018 — Training Guide

by Aulene De

This guide is meant to build a foundation in Algorithms and Data

Structures for the Indian National Olympiad in Informatics,
referred to as the INOI, the Indian National Test for entry to the
International Olympiad in Informatics. This guide is for you if you
are a beginner to competitive programming. Although this is
mainly aimed towards the INOI, I’d still give this a whirl if you’re
still brand new to programming contests.

This guide is definitely not a comprehensive document like Antti

Laaksonen’s Competitive Programming Handbook. He does a
better job explaining concepts than I ever could, and I highly
recommend reading his document over mine.

The objective is to ensure that a person learns through hands-on

problem-solving rather than giving a lot of theory. There will be a
little bit of theory here, but I recommend people consult an
Algorithms textbook like CLRS for the more theoretical parts.

I recommend that a person try every problem mentioned here for

45-60 minutes at the minimum. Try to come up with a logic,
implement it, find bugs, debug, figure out why that approach didn’t
work and make sure knowledge gained doesn’t go to waste.
This guide isn’t supposed to be the one-stop-shop. In fact, it is a
track with breadcrumbs. I recommend everyone use Google
throughout every point of the guide, looking for problems on
certain topics and alternative explanations on the more
theoretical parts. I expect everyone to not limit themselves to this
one document.

Without much talk, let’s begin. I try to embed humour into my

guides so that reading doesn’t become stale, but I find it hard to
convey humour through text. Hope that you have a great time
ahead, and good luck!


 A Brief Overview of the INOI Syllabi
Bold = Important

1. Complexity Analysis of Algorithms

2. Sorting
3. Searching
4. Basic Graph Algorithms
5. Dynamic Programming
6. Heaps
7. Permutations
8. Directed Acyclic Graphs (DAGs)
9. Prefix Sums
10. Sliding Window Algorithms

Here’s a trade secret. Traditionally, the INOI has two problems —

usually a Dynamic Programming problem and a Graph problem.
We’ll be covering a good portion of the syllabi in this guide.

Both problems are divided into subtasks (for a total score of 200,
100 for each problem). What this means is that, depending on the
efficiency of the Algorithm/Data Structure you use, your program
could work on a partial score, but not be efficient enough to get a
full 100.

In example, if a problem is to be graded out of 100, it could be

divided into 20-40-40, and you can use a program that might not
be as efficient in order to get a partial score.

But nobody wants a partial score.

“So Aulene-senpai, I really want to go to IOITC, this is super cool I

want to get a 200/200 PLEASE HELP ME D:<“. Fear not. I shall
take you as my padawan. With that, we shall move on to our first
encounter with Algorithms.
 Getting familiar with Competitive Programming

The best way to start is by solving problems. I’ll be linking a few

problems here to introduce you to the online judge system.
Basically, since there is no way to test out your code on every
single case, there is a certain amount of ‘test data’ made against
which your code is run. Then, your output is compared with the
output of another program which is guaranteed to be correct. If
outputs match, we can confidently say that your code/algorithm is
correct.

The first step is to make an account on CodeChef, SPOJ and the

IARCS Online Judge. A majority of the problems I’ll be giving you
will be on these websites.

Without much ado, I encourage you to familiarise yourself with

CodeChef. Also, most problems I’ll tell you to solve will have
editorials linked below the problem. Look at them if you feel stuck,
but don’t look at them before giving ample time to the problem.
Whenever you feel that there’s nothing more you can do, look at
the editorial and get the accepted verdict. Try the following
problems. -

• TEST - Do read the Input/Output tab.

• INTEST
• HS08TEST
• FLOW001
• FCTRL2

Soon enough, you could make a nice little folder with all the
problems you massacre and brag about it later. :D
Searching and Algorithm Analysis

Task: Given an a array of numbers, check if a number is present in

the array. If it is, return the index (location) of the number in that
array.

Example
Let our array be {3, 4, 1, 5, 2}. If I ask you the location of the
number ‘3’, you return 1*. If I ask you the index of the number ‘6’,
you return 0 i.e false **.

* - 1-based indexing, meaning that locations of the numbers start

from 1 (Location of ‘3’ is 1, location of ‘4’ is 2 .. location of ‘2’ is 5)
** - In Boolean, 1 corresponds to being ‘True’ and 0 corresponds to
a number being false.

Seems simple enough right? Mhm, it is! Let’s take our first look at
two Algorithms — Linear Search and Binary Search — see what
makes them so different and when to use them.
Linear Search
The idea of Linear Search is simple, we start at the first index and
check the number at that index. If we find our number, we say
“Found it Aulene-san” and return the index, and if we don’t, we say
“It’s not here ;c” and return false (0).

Note: As we go further along this guide, I will introduce to

somewhat-formal notations, so please do read every section 2-3
times in order to grasp all the information you can.

Example
Our array A = {3, 4, 1, 5, 2} and I have asked you the location of the
number ‘1’ in A.

We start from index A[1] and check the number there. It’s 3, so we
go on checking. A[2] is 4, A[3] is… what we’re looking for. :D
Return that index i.e 3.

What if the number we’re looking for is ‘7’? Well then we check
A[1], A[2] .. A[5]. No dice. We return 0.

With that, here’s the code for Linear Search.

Alright, brace yourselves.

Now, let’s say we have an array consisting of a 100 elements, what
is the worst-case time that Linear Search will take in order to
determine if an element exists in the array or not? Take a minute
to think about this and come up with an answer before reading
further.

Hint: Think of how many times the for loop will run before
returning a value.

The answer is 100. The worst-case is when the element we are

looking for is not present in the array, thus the for loop will run a
100 times before saying, “This ain’t here, bruh.”.

Now, let the number of elements in the array be a variable i.e N.

Now, in the worst-case, the for loop will run N times before
returning an answer.

Introducing Big-O notation - O ( )

Big-O notation is a way to express the worst-case complexity of an
algorithm. You’ll be seeing a lot more of it from now on, and the
idea will seem complex at first, but it’ll get easier in a bit.

With reference to the case in Linear Search — where the number of

elements is N — we will express our complexity as O ( N ) . Basically,
we’re saying that, in worst-case scenario, our algorithm will take N
steps.

In simpler terms, the block of code inside the Linear Search for
loop runs N times, with each execution being called one operation
i.e O (1) . So, we say that the complexity of our code is O (1× N ) or O ( N ) .

Fear not, you’ll get familiar with the idea as you solve more
problems.

With that, let’s move on to…

Binary Search

Binary Search works on the principle that, in a sorted array (an

array in ascending/descending order), we need not check every
single number in the entire array to check if a number is present
like we did in Linear Search.

Let’s say we’re looking for a number V in an array A consisting of N

elements. Let’s define 3 variables as follows -

1. Left = 1
2. Right = N
3. Mid = (Left + Right) / 2

We check the element present at A[Mid].

If A[Mid] > V, we can say that all elements from Mid to N (i.e
A[Mid], A[Mid+1] … A[N]) are greater than V, hence we don’t need
to look for V between Mid and N.

If A[Mid] < V, we can say that all elements from 1 to Mid (i.e
A[1], A[2] .. A[N]) are lower than V, hence, we don’t need to look
for V between 1 and N.

If A[Mid] = V, we have found our element and return the index.

We repeat this process over and over until -

1. A[Mid] = V, which is when we return the index OR
2. Left > Right, which means that we could not find the number V
in our array and return 0.
Example
Let our array A = {10, 20, 30, 40, 50, 60, 70}, thus number of elements N = 7.
Let’s say we’re looking for the number V = 50.

By our definitions above, Left = 1, Right = 7. This means that Mid = 4.

At A[Mid], we find 40. Since 40 < 50, we can safely say that A[1] upto A[Mid]
are also lower than 50. Why? Because our array is sorted. (Don’t worry, we’ll
discuss sorting soon)

We update Left to be equal to Mid+1 i.e equal to 5 and repeat the process,
making Mid = (5+7)/2 = 6.

At A[Mid], we find 60. Since 60 > 50, we can safely say that A[Mid] upto
A[Right] are also greater than 50.

This time, we update Right to be equal to Mid-1 i.e equal to 5 and repeat the
process, making Mid = (5+5)/2 = 5.

Check A[5], we see A[5] = 50, which is the number we were looking for. We
return the index 5.

What if we were looking for a number not present in the array i.e 55? Well,
the process would go just about the same as above, except on the last step, we
would return 0 since A[5] =/= 50.

Because I’m a super-nice person, I’ll give you the code for Binary Search, but
make sure you implement it on your own before reading further. *virtual
pinky promise?*
But Aulene-senpai, I already know Linear Search, why do I need to
know Binary Search?


Patience, youngling.

Binary Search v. Linear Search

Remember when we spoke about analysing the worst-case

complexity of an algorithm (Big-O Notation)?

Well, it turns out Binary Search is a LOT faster than Linear Search.
This is because we’re eliminating almost half of our possible indices
in one operation, rather than looking through each and every
element like we did in Linear Search.

Instead of taking N steps like we did in Linear Search, we’re taking

LogN steps. Hence, the complexity of Binary Search is O ( LogN ) , and
you can see the difference in this beautiful graph below.
Revisiting Complexity Analysis

Consider the block of code below and try to estimate the

complexity our Algorithm.

Simply put, since Binary Search takes O ( LogN ) time to execute

and our for loop runs in O(N ) , we can say that the time complexity
of our algorithm is O(N • LogN ) .

Similarly, the following code has a complexity of O(N 2 • LogN ) .

Now, before we move on to the next section, I want all of you to

solve the following problems in the order below. It’s fine if you can’t
solve a problem, don’t look for solutions/editorials immediately.
Try to think until you feel exhausted of ideas. After that, get a hot
dog and read the editorial, then close the editorial and implement
what you learnt until you receive the Accepted verdict.

1. BSEARCH1
2. FORESTGA
3. AGGRCOW
An Introduction to Recursion

“Alright, so recursion is everything, and recursion is recursion. Everything

in programming is recursion. I am recursion. YOU are recursion. 42 is
recursion and recursion is 42.” - Yours truly

Quotes aside, understanding recursion is key if you want to be a good

competitive programmer.

Well sensei, what is recursion?

Recursion is to use small sub-problems in order to solve larger problems.

You’ve all played with Legos, right? These little fun-filled delightful little
creatures until you step on them and see your life flashing before your eyes?

Anyway, so Legos start small; you put one block on top of the other until you
build the Walt Disney castle :3.
Recursion is similar to Legos. You take a very small problem i.e a
‘block’ and use it to solve/find the answer to larger problems (i.e A
Disneyland made of Lego blocks. I’m sorry I had a sheltered childhood.)

Alright, enough jokes. Recursion is serious business.

Understanding recursion would make this entire process a lot
simpler, so be prepared to read this section over and over again.

A Simple Series

Let’s take a look at Fibonacci numbers. If you don’t know what

they are, Fibonacci numbers are defined as follows -

F ⎡⎣ i ⎤⎦ = F ⎡⎣ i −1⎤⎦ + F ⎡⎣ i − 2 ⎤⎦
F ⎡⎣ 0 ⎤⎦ = 0
F ⎡⎣1⎤⎦ =1

.. where F[i] denotes the ith Fibonacci number. The Fibonacci

series looks like this -

0,1,1,2, 3,5,8,13,21, 34,55,89...

Now, let’s look at our first equation. We know the values of the first
and second Fibonacci numbers, and we can use them to compute
more. We can define a function as follows in order to generate
Fibonacci numbers, where N signifies the Nth element in F[].
Let’s take a look at the code.
If N = 0, return 0.
If N = 1, return 1.

However, since we don’t know what the Fibonacci number at the

Nth position could be, we tell the computer to call the Fibonacci
function again to calculate the Fibonacci number at the (N-2)th
and the (N-1)th positions, until we finally reach either N = 0 or N =
2.

Here’s a very beautiful and intricately crafted tree in order for you
to understand this -

Now, developing recursive thinking is challenging and it won’t

make sense to you in a day. Don’t fret. Try to define the problem in
simpler terms and work from there. Recursion is super important,
and you’ll come to appreciate it’s elegance and splendour and all
qualities you dream to have in your significant other. ;)

A nice visual way to see the above is linked here.

We’ll be using recursive functions when we take a peek at Merge

Sort in the next section and recursion is the underlying principle
of Dynamic Programming.
Sorting
As defined earlier, sorting is simply putting an array of numbers in
ascending order.

Note: I won’t be going in-depth into sorting. This is simply because,

most of the time, you’ll be using the C++ sort() STL function to
accomplish the task.

I’ll be discussing Merge Sort here, because it has a very elegant

recursive definition and serves as a nice introduction to the Divide
and Conquer paradigm. As the note stated, we won’t be spending
too much time here because I have to get to the g o o d s t u f f.
Merge Sort works in 3 steps -
1. Divide an N-element sequence to be sorted into two
subsequences of n/2 elements.
2. Conquer - sort the two sequences recursively using Merge Sort.
3. Merge the two subsequences to produce the sorted answer.

Anyway, this diagram should explain everything. It’s straight out

of CLRS as well. Again, this section is pretty useless, but really it’s
only to demonstrate how Merge Sort works.
You can find the code for Merge Sort here and find a visual way to
see it in action here. You’ll NEVER implement it programming
contests, but implement it once and just use the C++ Sort function
afterwards.

Alright, time for some problems! Feel stuck? Use Google :)

Do ALL of these. They will strengthen your Searching and Sorting
concepts tremendously.

1. TSORT
2. INCPRO4
3. BOOKSHELVES - ZCO (Sorting)
4. VARIATION - ZCO (Sorting)
5. WORMHOLES (Sorting + Binary Search) - ZCO
6. PYRAMID - (Sorting + Binary Search)

Prefix Sums
Alright, so Prefix Sums are super important. Well, Aulene-kun,
what are Prefix Sums?

Prefix Sums can be defined as, for the ith element in the in an
array, Prefix[i] is the sum of all elements upto and including the
nth element.

Formally, we can create a prefix array with the following

definitions -

Prefix[1] = A[1]
Prefix[i] = Prefix[i-1] + A[i]

Example
Let’s say we have an array A = [1, 3, 5, 6, 10].
Going by our definitions, Prefix = [1, 4, 9, 15, 25].

So why are they important? Well, computing sums from an

element i to another element j will result in O ( N ) worst-case
complexity. Let’s assume that we have to compute sums K times
throughout the course of a problem, our complexity is suddenly
O(N i K ).
Now that’s bad. You know why it’s bad? Because we can compute
prefix sums once in O ( N ) and find all other required sums in
constant time or O (1) .

OMG AULENE-SAN, PLEASE TEACH ME YOUR WAYS, WHAT

SORCERY, WHAT GRACE, SO MUCH BEAUT, MUCH APPRECI8.

Alright, so think back, how would you use prefix sums to compute
sums in constant time?

Let’s look at the example above. To reiterate..

Example that I mentioned above but I’m mentioning again because

all of you are filthy casuals (haha. jokes. so funny)

Let’s say we have an array A = [1, 3, 5, 6, 10].

Going by our definitions, Prefix = [1, 4, 9, 15, 25].

Let’s say I have to compute the sum from i = 2 to i = 4. This would

be equal to.. 3 + 5 + 6 = 14.

Now, what if I subtract Prefix[1] from Prefix[4]?

This would be equal to 15 - 1 = 14, which is our answer. One less for-
loop = very elegant code.

1. CSUMQ
2. JURYMARKS
Anyway, you need to know prefix sums for what’s about to follow.
It’s time for my favourite section, the most important one for the
INOI and relatively tough so be prepared for l o t s of problems.

If you felt that the guide was like a low-effort YouTube video up
until now, it’s because everything upto this point doesn’t matter
nearly as much as what the following does.


Really, if you want to clear the INOI, these next 2 sections are
everything. By everything, I mean..

E V E R Y T H I N G.
Scene 6
Enter: Dynamic Programming

You remember how Recursion works, right? You use small

problems and use them to find solutions to bigger problems.

Here’s the problem with Recursion. Let’s look at the recursive calls
for the Fibonacci problem again, shown below.

Try to analyse the complexity of the above. Hint: try to think in

terms of how many recursive calls are generated at each step!

If you guessed that the complexity would be exponential,

congratulations, you’re eligible for one of my freshly-baked virtual
chocolate chip cookies. To put it formally, if the time to compute
the Nth fibonacci number is T ( N ) , then T ( N ) = T ( N − 1) + T ( N − 2 ) .

We can see that, unless we hit F(0) or F(1), we don’t return any
value. For example, we have to compute F(3) over and over again.
Don’t you think it would be faster if we somehow ‘remembered’ the
value of F(3)?

Let’s take it one step further: what if, every single time we
computed a new value for F[], (be it F[3], F[4], F[whatever]), we
stored that value somewhere and returned it as needed?
Here’s how our fibonacci calls look with the optimisation:

With the above optimisation, we have reduced our complexity to

O ( N ) , big difference!

Say hello to Dynamic Programming. As I like to call it, Dynamic

Programming is ‘smart’ recursion. Rather than just pursuing
recursive calls, we try to store every single unique value we
compute and then just return that value.

Now, the best way to learn DP is by doing DP problems. This applies

to DP more than any other problem type, because you see, you’re a
DP problem. You need to solve smaller, easier DP problems in order
to be able to solve harder DP problems. Fun, is it not?

So here’s what you do. You solve each of these problems in the
order that I give you, and you solve them all. Fair warning, you’re
going to get a lot of problems here on out. Chit-chat’s over.
The first thing I want you to do before crossing names off a list like
Oliver Queen (Arrow reference, anyone?) is to head over to the
following links and read about all sorts of different Dynamic
Programming problem types. They’re explained beautifully, and do
a better job of explaining than my little hands ever could. DON’T
read the code presented at ANY of those links.

Here’s exactly what I want you to do.

1. Open the problem first.

2. Try to come up with a base case and how you’ll build a solution
on top of it.
3. Implement your approach. If you can’t think of one, proceed to
step 4.
4. Read the solution presented at GeeksForGeeks. Again, DON’T
look at their code. Analyse complexity. Implement their solution, in
your own code.
5. This is the most important one: close the solution, and re-
implement.

1. Longest Increasing Subsequence - Problem

2. Longest Common Subsequence - Problem
3. Edit Distance - Problem
4. Grid-Type Min-Cost Path
5. Coin Change
6. 0-1 Knapsack Problem
7. Longest Palindromic Subsequence

Trust me, you do this and the problems that follow, you’ll make
your DP lives easier by miles.
Give a lot of time to all of the following problems. There’s a lot to
learn here.

1. CHOOSING CHOCOLATES (ZCO’09, test cases are unavailable

though. Still worth trying)
2. SUPW (ZCO’14)
3. IPL (ZCO’13)
4. TRIANGLE (IOI’94)
5. ROUNDTABLE (ZCO’12)
6. BREAK UP (ZCO’15)
7. BYTESM2
8. TILINGS
9. COINS
10. CALVINSGAME (INOI’13)
11. LONGPALIN
12. EDIST
13. BUFFALOES
14. FLOWERS
15. KTREE
16. TETRAHEDRON
17. BRACKETS (INOI’16)
18. POSTOFFICE (IOI’00)
19. TRAINING (INOI’17)
20.OFFICEKEYS
21. STARSKY

With that, I’ll conclude the Dynamic Programming section of this

guide. This is probably the topic that will require you to challenge
yourself the most, so do as many DP problems as you can outside of
those mentioned on this guide.

Also, STOP.
If you haven’t done so already, I recommend you check out my
document on STL. Graphs will require heavy STL implementations,
thus it’s a good time to know basic data structures and STL now.
Playing with Graphs by Amit M. Agrawal Austin
LegendFlame (I dare you to Google ‘Austin LegendFlame’)

Let’s answer the first question: what are graphs? Well, graphs are
basically a set of objects, some of which are linked to each other.

Example
Let’s say we have five cities, Paris (City 1), New Delhi (City 2),
New York (City 3), London (City 4) and Shanghai (City 5). I have
to go to New York (City 3), and I’m currently in New Delhi (City
2). Here’s the flight plan for today:-

• New Delhi to Paris (2 -> 1)

• New Delhi to Shanghai (2 -> 5)
• Paris to New York (1 -> 3)
• Paris to Shanghai (1 -> 5)
• Shanghai to London (5 -> 4)
• London to New York (4 -> 3)

We could represent these connections pictorially i.e as a graph.

There’s a lot of terminology here, so let’s break it down one by one.

• Node: The cities on our graph i.e New York, New Delhi, etc are
called nodes. They are the objects on our graph.
• Edge: See the connections between cities? They’re called edges.
They represent a link between two objects, in this case, a flight
going from one city to another. You’ll encounter two types of
edges, Directed and Undirected.
• Directed Edge: This implies that there is a one-way connection
between two nodes. For example, in our graph, there is an edge
from Paris to Shanghai. You cannot use the same edge to travel
from Shanghai to Paris.
• Undirected Edge: This kind of edge symbolises a two-way
connection, meaning that you can use an edge from one node to
another in any direction.

Here’s how our graph would look like if all the edges were
undirected edges.

The above graphs are unweighted graphs. This means that
traversing any edge has no cost associated with it. However,
weighted graphs have a cost associated with traversing each edge.
In our example example, the edge between two cities could have a
weight equal to the distance between them. This would make our
graph look something like this -
Now, how do we represent graphs? Glad you asked. Turns out,
there are two traditional ways to represent graphs: the Adjacency
Matrix and the Adjacency List.

The Adjacency Matrix Representation

This representation is a lot easier to implement. We represent our

graph as a 2-dimensional grid, with the rows and the columns
equal to the number of nodes in our graph.

If the graph is unweighted, we can simply use graph[x][y] = true to

signify that there is an edge between nodes x and y and graph[x]
[y] = false to signify that there is no edge between x and y.

If the graph is weighted, we initialize all graph[x][y] to be equal to

a very high value (say, INT_MAX). Then, we change the value of
graph[x][y] to the weight of the edge, given that the edge exists.

Here’s how our example would look like in an adjacency matrix

representation -

Weighted, Undirected Adjacency Matrix

1 2 3 4 5

1 ∞ 6597 5837 ∞ 9257

2 6597 ∞ ∞ ∞ 4245

3 5837 ∞ ∞ 5585 ∞

4 ∞ ∞ 5585 ∞ 9199

5 9257 4245 ∞ 9199 ∞

Unweighted, Undirected Adjacency Matrix

1 2 3 4 5

1 0 1 1 0 1

2 1 0 0 0 1

3 1 0 0 1 0

4 0 0 1 0 1

5 1 1 0 1 0
The Adjacency List Representation

Aulene-senpai, I already know the Adjacency Matrix

representation, why do I need to know the Adjacency List
representation?

I’ll answer that after class.

In the Adjacency List representation, rather than creating a 2-

dimensional grid to store each edge of a graph, we use a vector for
each node to store it’s neighbours and the weight of the edge
between them. If the graph is unweighted, we just store the
neighbour.

Thus, for our weighted, undirected graph, the Adjacency List

representation will look like this -

Weighted, Undirected Adjacency List (Neighbour, Weight of Edge to Neighbour)

1 2 3 4 5
2, 6597 1, 6597 1, 5837 3, 5585 1, 9257
3, 5837 5, 4245 4, 5585 4, 9199 2, 4245
5, 9257 4, 9199
Why do we need to know the Adjacency List representation
AND the Adjacency Matrix representation?

Here’s the thing: the Adjacency Matrix representation requires

( )
O N 2 space, so if N >= 105, where N is the number of nodes. So
when you do try to use an Adjacency Matrix, your computer is
going to curse you like a sailor. On the other hand, the Adjacency
List representation has an O ( N + E ) space complexity, where E is
the number of edges.

Alright senpai, so if the Adjacency List is s o g o o d, why should I

ever use an Adjacency Matrix?

Well, you see, if you ever had to check if an edge exists between two
nodes, it’s complexity would be O ( N ) in an Adjacency List, while it
would be O (1) in an Adjacency Matrix.

Graph Traversal
Now that you know how to represent a graph, we can move on to
how you should explore a graph. What’s ‘Traversal’? In simple
terms, think of nodes as ‘cities’ and edges as ‘roads’ that are
connecting those cities. We use these roads to go from one city to
another. In formal terms, this is exactly what traversal is.

What better way than to start with a problem?

Problem Statement (GREATESC on IARCS):

Given an undirected, unweighted graph comprising of N nodes (1
<= N <= 3500), M edges and two nodes, u and v, calculate the
minimum distance to reach v from u. Since the graph is
unweighted, assume that all edges have weight = 1.

Example Case:
u = 1, v = 4. The answer for this case will be 3, with the shortest
path going from 1 -> 3 -> 5 -> 4.
Spoilers on how to solve this problem follow.
Let’s see.. in order to find the shortest path from A to B, we need to
explore A’s neighbours, correct?

So clearly, if we have to find the minimum distance between node 1

and node 4, we need to traverse every single edge going out from 1,
then we need to traverse every edge going out from 1’s neighbours,
then we need to traverse every edge going out from 1’s neighbour’s
neighbours and so on until we reach node 4.

How would we ever do that?

Depth-First Search (DFS)

How 2 DFS like a Pr0 (IN THE HOOD) -

1. We create a boolean array visited[], where visited[i] = 1 implies
that we have already visited a node and explored it’s
neighbours, while visited[i] = 0 implies that we haven’t
explored a node.
2. We’ll also create an array distance[], where distance[i]
represents the minimum distance of node i from node v.
Initially, we set all distance[i] to INF (a very high value, say
1000000).
3. We create an Adjacency Matrix/List and use it to represent our
graph.
4. We start our DFS from node 1.
5. We follow the process mentioned above (explore 1’s first
neighbour in-depth until we return to 1, and then explore the
rest of 1’s neighbours.)
6. You can find a visual representation of DFS here.

At the end of this ‘process’, here’s how our distance[] array will
look like.

1 2 3 4 5
3 3 2 0 1

Also, because I’m really nice, here’s the code for the Depth-First
Search on an Adjacency Matrix.
 Complexity Analysis of Depth-First Search
Adjacency List Adjacency Matrix

O (V + E ) O(V 2 )

Breadth-First Search (BFS)

BFS is another graph traversal algorithm, like BFS. Fair warning:
the following two paragraphs are plagiarised. A ‘bit’, anyway.

Using DFS, we start the traversal from the root node and explore
the search as far as possible from it (aka ‘depth-wise’). We keep on
exploring a node’s neighbours until we hit a leaf node or a node
which isn’t connected to any non-visited nodes. Exploration of a
node is suspended as soon as another unexplored node is found.

However, in BFS, we go in a level-by-level manner. What this means

is that rather than going deep into the graph, we completely
explore the neighbour’s of a single node before moving onto
exploring the others.

It’s probably better with a visual representation. Since this (as of

now, anyway), is a PDF, I can’t embed a video file (more
importantly, I can’t embed in GIF memes… this is a nightmare), so I
want you to head over to this VisuAlgo link and try DFS and BFS
on the same graph and hit play.
That’s about it for BFS. You won’t use it much (more like, ever…
but it’s good to know anyway 👍 ).

Problems -
1. TASKFORCE
2. GREATESC
3. PPRIME (Hint: Sieve of Eratosthenes)
4. CRITINTS
5. FENCE (INOI’17)
6. PT07Y
7. PT07Z
8. PARTY
9. NEWREFORM
10. SPECIES

Shortest Paths
Simply put, in a graph, what is the shortest ‘distance’ between two
nodes U and V, where the distance between two nodes is defined as
the sum of edge weights in the path from U to V.

Dijkstra’s Algorithm (Single-Source Shortest Paths)

Let’s consider our favourite Airport World graph from earlier -

Here’s our problem statement: what is the minimum distance that

I need to travel to get from Paris to any other city? Note that the
graph is weighted.

Sample Case:
U = 1, V = 4.
Answer is 11422 km. (Path is 1 -> 3 -> 4)

To formalise the problem a bit more, the problem is asking for

single-source shortest paths: basically, given any source node U,
you can be asked to find the shortest path to any other node V.
The answer to this problem can be found using Dijkstra’s Shortest-
Path Algorithm. It works as follows -

1. We select a starting node. In this case, let our starting node be

U = 1.
2. We go through all edges outgoing from that node and reduce
distances to connected nodes using them.
3. After step 2, we select the node who has the minimum distance
from node U, and repeat step 2.
4. Now, we still again an unexplored node with the minimum
distance, repeat step 2 and 3 until we have explored all nodes.

( )
The complexity of the above algorithm would be O N 2 . However,
we can reduce this by using a max-heap, which would reduce the
complexity to O ( NLogN ) . The following is how I implement
Dijkstra’s.
Problems -
1. EZDIJKST (Dijkstra’s Sample Problem)
2. SEQUENCELAND
3. WEALTHDISPARITY
4. SHPATH
5. WORDHOP
6. CLIQUED
 Epilogue
So… that’s about it from what you’ll learn from this. Hope you had
fun.

This guide was my effort to give back to the community I am very

glad to have been a part of over the past year. I started this project
exactly on 17th of May, 2017 and finished it on the 2nd of August,
2017 after lots of breaks in writing.

Where do you go from here?

Firstly, give yourself a pat on the back — you’ve earned it.

You’ve done more than 50 problems just from this guide, each one
harder than the next. Even if you’ve not, you still have time to do
them now.

There’s a lot you can learn now - start with Segment Trees. Or try
your hand at a couple IOI problems. Whatever floats your boat :P,
but do have fun. I also hope that you, with whatever little skill you
have, give back to the programming community.

If you have solved all the problems mentioned throughout this

manual and have learnt the lost art of Googling, you’ll probably be
able to handle even harder algorithmic problems and techniques.

I’ll leave that to you.

I really do hope this guide helped you in some shape or form.

If you have any critique for this guide or any questions, please
email me at [email protected], or contact me over Facebook.

See you starside,

Aulene / Aul.
*dramatic noise as I press this virtual log out button*