Department of Computational and Data Sciences

DS226: Introduction to Computing for Artificial Intelligence and Machine Learning

October 13, 2022
Assignment 2
Instructions:
• Submit a typed report in PDF format on Moodle. Use LATEXpreferably. Handwritten
reports will not be accepted.
• Solutions should be in the same order as the questions.
• Discussion is encouraged, but answers should be your own. Do not copy answers.
Plagiarism will be penalised severely.
• For late submissions, the following late submission policy:

Delay % of Credit
0-72 hours 50
72 hours - 1 week 25
Beyond 1 week No credit

Deadline: 27th October, 11:59 AM

Total Credits: 100

Question 1. (30)

Take any dataset from https://data.gov.in/ and produce visualisations using matplotlib, pandas,seaborn
and numpy.
Consider yourself as an independent evaluator presenting the findings based on the data to Government
authorities who are going to make decisions based on the data(visualisations) that you present. The
objective here is to create a Data Story and to make sure each plot conveys some inference (can also be
some hand-wavy inference) related to the data. Feel free to use more plots than the ones mentioned below
if you feel that it is meaningful and helps in explaining your data more concretely.
1) Submit one jupyter notebook with all code plots and any comments/notes you have to make (com-
ments can be made via markdown in jupyter)
2) You are required to do at least the following
• Scatter plot using matplotlib
• Line plot
• Bar plot or dot plot
• Box plot or violin plot
• Pair plot using seaborn (make sure you justify the plots that you choose to show in the pairplot)
3) For each plot write what inferences you can draw from the plot.
4) Each plot should be legible and must be properly labelled.
5) Finally, a single jupyter notebook is required for this question. Once you save your jupyter notebook
make sure all the outputs are seen when it is reopened (without the need for compiling it again).

1
 2

Question 2. (30)
Consider a vector v ∈ Rn ,
 
v1
 v2 
 
v=
 .. 

 . 
vn
Pn
stored as a 1-dimensional array. Recall that the sum of all elements of this vector, i=1 vi , can be computed
with a parallel implementation using OpenMP
#pragma omp parallel for default(shared) reduction(+: sum_arr)
for (i = 0; i <= ARRAY_SIZE; i++)
sum_arr += arr[i]; // Sum up the array
Now, consider the dot product between two vectors, v, w ∈ Rn ,

v · w = vT w
 
w1
 w2 
i
h 
= v1 v2 . . . vn  .. 

 . 


wn
n
X
= vi wi
i=1

The dot product is used widely to calculate vector projections, vector decompositions and to determine
orthogonality. In the context of neural networks, the dot product can be used to calculated the weighted
sum of inputs to a neuron.
I. Write a routine to calculate the dot product of two 1-dimensional arrays with double val-
ues. The routine should be an efficient, parallel implementation in C using OpenMP.

Another commonly used linear algebra routine is DAXPY. For two vectors v, w ∈ Rn and the scalar
α ∈ R, DAXPY performs the following operation

v ← v + αw

Assume that v and w are both stored as 1-dimensional arrays of double values and α is a double constant.
II. Write a parallel implementation of DAXPY using OpenMP in C. The routine should take
two 1-dimensional arrays with double values and a double scalar and perform the DAXPY
operation mentioned above.

Next, we consider the L2 norm of the difference between two vectors v, w ∈ Rn ,

v
u n
uX
∥v − w∥L2 = t (v
i − wi )2
i=1

This can be used to calculate the error between a predicted and observed value, and finds extensive usage
in ML algorithms. Assume v and w to be two 1-dimensional arrays with double values.
III. Write a routine to calculate the L2 norm error between two 1-dimensional arrays with
double values. Use the parallel dot product and DAXPY functions you have implemented
earlier to build this routine.
Lastly, recall that a matrix-vector product appears in the solution of the normal equation for a linear
regression problem. For a given matrix A ∈ Rm×n and a vector x ∈ Rn , the matrix-vector product is given
 3

by
   
a11 a12 ··· a1n x1
 a21 a22

··· a2n   x2 
  
Ax = 
 .. .. × . 
..   
 . . ··· .   .. 
am1 am2 ··· amn xn
 
a11 x1 + a12 x2 + · · · + a1n xn

 a21 x1 + a22 x2 + · · · + a2n xn 

=  .. 
.
 
 
am1 x1 + am2 x2 + · · · + amn xn
Assume A to be stored as a 1-dimensional array of double values with length m × n in the row-major form
and x to be stored as a 1-dimensional array of length n
IV. Write a routine to compute the matrix-vector product Ax. Use the parallel dot product
function you have implemented earlier to build this routine.

Question 3. (40)

This is a programming question to recreate a very simplified version of the classic Snake game. In this
simplified version, you have a snake with a goal to move around and eat pieces of food in the game area.
Once the game starts, the snake moves continuously in the user input direction. Each time snake eats
food, it grows i.e. its length increases. You lose the game if the snake runs into itself.

Setup:
The essential objects for building the game are,
1) Screen or the game window
2) Snake Head
3) Snake Body
4) Food pieces
You have been provided template for the program written in Python. The template takes care of the game
window and other GUI implementations using turtle. You have to write the main game logic components
like movement of snake’s head, random food placement in game area and consumption condition for food,
increase of snake’s length after successful consumption of food piece and the check for the game end
condition using OOP concepts in Python. Rest instructions are given in the template. Each function
is accompanied by a set of comments that will tell you what the function implements, what it accepts as
parameters and what it returns. You are given the function definitions in the template.
Objectives:
1) Your code must run after you submit it.
2) Food should appear at random places inside the game window.
3) Snake length should increase by one unit after consuming the food piece.
4) Game should end if the snake runs into itself.