You are on page 1of 16

RIPHAH INTERNATIONAL

UNIVERSITY ISLAMABAD
Faculty of Computing
Final Term Online Examination Fall 2020 Semester

Course Name: Introduction to Machine Learning Course Code:


Program: BS SE
Semester: 7th Section:
Date of
Teacher’s Name Mr. Shahbaz Hassan 24th January
Examination
Exam Duration: 06 Hours Total Marks 30
Vetted By:

INSTRUCTIONS
 It’s an open book test, students can look at relevant resources andnotes
 Attempt all questions; marks of each question are dulymentioned.
 Write your name and SAP ID on all answer sheets and embed your signature on all answersheets.
 Write page number on allsheets.
 Make Pdf file of answer sheets and rename the file with your name & SAPID.
 Upload Pdf file on Moellim within the giventime
 Please be analytical and specific in answeringquestions.
 Image must bereadable
 The students should retain the answer sheet until the announcement ofresults.

Question 1: [10 Marks] [Python Implementation]


This question is related to clustering problem. Download the CVL single digit database. The database comprises
7000 Training Images, 7000 validation Images and more than 21000 Test images. Refer to the readme file with
the database for more information. You are required to implement the K-means clustering in python where
youwill show the cluster centers. For this, you may consider (Training, Test, Validation, or any sub set of the
dataset). Sample output with 7000 training images is illustrated in Figure 1.
Make a copy
Of pictures:
Question 2: [10 Marks][Handwritten]
Back propagation is the core of training neural networks. The objective of back propagation is to optimize the
weights so that the network can learn how to correctly map arbitrary inputs to outputs. You need to apply
back propagation to the given neural network (by hand) and show the weight updates resulting from the
3rditeration.

Network Input:

 To provide an input to the network you will consider youSAP-ID. First, you will split your SAP- ID into two
digits and then you will divide the numbers with 100. For example if your SAP-ID is 1234 then at your first
step you will split into 12 & 34 and then you will divide with 100 i.e. 12/ 100= 0.12 and 34/100=0.34.

 To obtain the weight values of Wac, Wbc, Wad, Wbd you should consider the first digit of your SAP-ID/100
And for weight values of Wce and Wde you should consider last digit of your SAP-ID/100

 Value of learning rate = 1 and Target value is0.7


Question 3: [10 Marks] [Descriptive]

Answer the following questions.

 Write the name of the application that you should want to design after studying this course.

Answer: I want to make an application which identifies the number plate of any vehicle.

 What type of features you would like to include in your application?

Answer:

1) Identify the numbers on the number plate through OCR (Optical Character Recognition)
of any vehicle.

 Identify the model of any vehicle.

 What type of dataset you will use or how you will collect the dataset for your application?

Answer: Collecting images data of cars from internet. 

 What do you think how much instance of dataset should be enough for your application.

Answer: 2000 images for every different vehicle then we can easily identify the number plate
and model of vehicle.

 Which machine learning approaches you will adopt (classification or clustering).


Answer: Classification

 Justify your selected approach with suitable comments.

Answer:

1) I will use YOLO (You Only Look Once) algorithm for this application.

2) It detects faster as compared to others.

3) Its accuracy is also good.

4) This algorithm is suitable for live object detection.

You might also like