10/3/24, 11:30 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 3 - Week 0: Prerequisite

Answer Submitted.
X

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Week 0: Assignment 0

Assignment not submitted
About NPTEL ()
1) For what value of x will the matrix given below become singular? 1 point

⎡ 8 x 0⎤
⎢ 4 2⎥
How does an NPTEL
0
⎣ 12 0⎦
online course work? ()

Week 0: Prerequisite ()
6
2
Practice: Week 0:
Assignment 0 3
4

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=16&assessment=133 1/6
10/3/24, 11:30 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 3 - Week 0: Prerequisite

(assessment? 5
name=133)
Yes, the answer is correct.
Week 1: Introduction () Score: 1
Accepted Answers:
4
Week 2: Bayesian
Decision Theory () 2) If A is a real square matrix then AAT is 1 point

Week 3- Parametric Unsymmetric

Approaches and Non- Always symmetric
Parametric Density
Skew-Symmetric
Estimation ()
Sometimes symmetric
Week 4: Perceptron Yes, the answer is correct.
Criteria and Score: 1
Discriminative Models Accepted Answers:
() Always symmetric

Week 5: Logistic 3) Let A be 3X3 matrix with rank 2. Then AX=0 has 1 point
Regression, Decision
Trees and Hidden
Only trivial solution X=0
Markov Model () One independent solution
Two independent solution
Week 6 - Ensemble Three independent solution
Methods ()
Yes, the answer is correct.
Score: 1
Week 7:
Accepted Answers:
Dimensionality
One independent solution
Problem ()

⎡1 1 1⎤
4) 1 point

The eigen value of ⎢ 1 1 ⎥ are

Week 8: Mixture Model
1
⎣1 1⎦
and Clustering ()

1
Week 9: Clustering ()

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=16&assessment=133 2/6
10/3/24, 11:30 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 3 - Week 0: Prerequisite

Week 10: Artificial 0,0,0

Neural Network () 0,0,1
0,0,3
Week 11: Introduction
1,1,1
to Deep Neural
Networks () Yes, the answer is correct.
Score: 1
Accepted Answers:
Problem Solving
Session - July 2024 ()
0,0,3
5) How many solution does the following system of linear equation have 1 point
Download () −x + 5y = −1
x−y = 2
x + 3y = 3
Live session ()
Infinitely many
Two distinct solution
Unique
None
Yes, the answer is correct.
Score: 1
Accepted Answers:
Unique

6) Four fair coins are tossed simultaneously. The probability that at least one head and one tails turn up is 1 point

1
16

1
8

7
8

15
16

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=16&assessment=133 3/6
10/3/24, 11:30 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 3 - Week 0: Prerequisite

Yes, the answer is correct.

Score: 1
Accepted Answers:
7
8

7) A bag I contains 4 white balls and 6 black balls. Another bag II contains 4 white balls and 3 black balls. One ball is selected 1 point
at random from one of the bags and it is found to be black. Find the probability that it was drawn from bag I?

2
3

4
9

3
5

7
12
Yes, the answer is correct.
Score: 1
Accepted Answers:
7
12

8) The standard deviation of a uniformly distributed random variable between 0 and 1 is 1 point

1
√12

1
√3

5
√12

7
√12

Yes, the answer is correct.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=16&assessment=133 4/6
10/3/24, 11:30 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 3 - Week 0: Prerequisite

Score: 1
Accepted Answers:
1
√12

9) If a random variable X satisfies the Poisson’s distribution with a mean value of 2, then the probability that X>2 is 1 point

2e−2

1 − 2e−2

3e−2

1 − 3e−2
Yes, the answer is correct.
Score: 1
Accepted Answers:
1 − 3e−2
′
10) Compute the derivative f (x) of the logistic sigmoid 1 point

1
1+e− x

1
1
e− x

0
Yes, the answer is correct.
Score: 1
Accepted Answers:
1
1+e− x

Check Answers and Submit

Your score is: 10/10

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=16&assessment=133 5/6
10/3/24, 11:30 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 3 - Week 0: Prerequisite

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=16&assessment=133 6/6
 10/3/24, 11:32 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 4 - Week 1: Introduction

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Assignment 1

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-08-07, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-08-05, 14:42 IST
1) In a binary classification problem, the confusion matrix is a __________ matrix. 1 point
Week 0: Prerequisite ()
1x1
Week 1: Introduction () 2x2
3x3
Lec 1: Introduction to
Machine Learning (unit? 1x2

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=18&assessment=134 1/5
10/3/24, 11:32 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 4 - Week 1: Introduction

unit=18&lesson=21) Yes, the answer is correct.

Score: 1
Lec 2: Performance Accepted Answers:
Measures of Classification 2x2
(unit?unit=18&lesson=22)
2) Precision is defined as 1 point
Lec 3: Bias-Variance
Tradeoff (unit? TP / (TP + TN)
unit=18&lesson=23)
TP / (TP + FN)
Lec 4: Regression (unit? TP / (TP + FP)
unit=18&lesson=24)
TN / (TN + FP)
Feedback form (unit?
Yes, the answer is correct.
unit=18&lesson=25) Score: 1
Lecture Notes (unit? Accepted Answers:
unit=18&lesson=88) TP / (TP + FP)

Quiz: Assignment 1 3) In a binary classification problem, a classifier correctly predicts 90 instances as positive, incorrectly predicts 15 1 point
(assessment?
instances as positive when they are negative, correctly predicts 90 instances as negative, and incorrectly predicts 10
name=134)
instances as negative when they are positive. What is the accuracy of the classifier?
Assignment Solution
(unit?unit=18&lesson=69) 80
85
Week 2: Bayesian
87.8
Decision Theory ()
95
Week 3- Parametric Yes, the answer is correct.
Approaches and Non- Score: 1
Parametric Density Accepted Answers:
Estimation () 87.8

Week 4: Perceptron 4) For the above question find the F1 score? 1 point
Criteria and
Discriminative Models 78.2%
() 85%
87.8%

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=18&assessment=134 2/5
10/3/24, 11:32 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 4 - Week 1: Introduction

Week 5: Logistic 90.2%

Regression, Decision Yes, the answer is correct.
Trees and Hidden Score: 1
Markov Model () Accepted Answers:
87.8%
Week 6 - Ensemble 5) Consider a dataset with actual values (Y) and predicted values (Y_pred) given below: 1 point
Methods ()
Y = [5, 8, 12, 10, 15],
Y_pred = [4, 7, 10, 11, 13].
Week 7:
What is the bias of the model?
Dimensionality
Problem () 0
-1
Week 8: Mixture Model
and Clustering () 2.2
None of the above
Week 9: Clustering () No, the answer is incorrect.
Score: 0
Week 10: Artificial Accepted Answers:
Neural Network () -1

Week 11: Introduction 6) What is the variance of the model for the data given in the above question? 1 point
to Deep Neural
Networks () 0
-1
Problem Solving 2.2
Session - July 2024 ()
None of the above

Download () No, the answer is incorrect.

Score: 0
Accepted Answers:
Live session ()
2.2

7) Given X = {-2, -1, 0, 1, 2, 3, 4, 5, 6, 7} and the corresponding Y = {-0.5267, 1.3517, 3.8308, 5.5853, 7.5497, 9.9172, 1 point
11.2858, 13.7572, 15.7537, 17.3804}. Find the parameters of the linear regression model.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=18&assessment=134 3/5
10/3/24, 11:32 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 4 - Week 1: Introduction

2.0065, 4.0312
2.0065, 3.5722
1.9214, 3.5722
None of the above

No, the answer is incorrect.

Score: 0
Accepted Answers:
2.0065, 3.5722
8) Find the MSE for the above question. 1 point

0.05783
0.04247
0.04876
None of the above

No, the answer is incorrect.

Score: 0
Accepted Answers:
0.04247

9) A model with high variance and low bias means 1 point

It can be too simple to understand the patterns of the data used in the training.
An excellent performance in the training data, but has a significant decrease in performance when
evaluating the test data.
The model fits the test data better.
The model becomes less sensitive to the training data.
Yes, the answer is correct.
Score: 1
Accepted Answers:
An excellent performance in the training data, but has a significant decrease in performance when
evaluating the test data.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=18&assessment=134 4/5
10/3/24, 11:32 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 4 - Week 1: Introduction

10) Which of the following techniques is used to prevent overfitting in machine learning? 1 point

To create complex machine learning models.

Train the model for more epochs.
Using a regularization to the model.
To increase the variance of the model.
Yes, the answer is correct.
Score: 1
Accepted Answers:
Using a regularization to the model.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=18&assessment=134 5/5
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment 1: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
3. Given, TP = 90, TN = 90, FP = 15, FN = 10

gy
𝑇𝑃 + 𝑇𝑁 90 + 90
Accuracy = 𝑇𝑃 + 𝑇𝑁 + 𝐹𝑃 + 𝐹𝑁
= 90 + 90 + 15 + 10
= 87.8%

lo
4. Precision ( P ) =
𝑇𝑃
𝑇𝑃 + 𝐹𝑃
=
90
90 + 15
no
= 0.8571
ch
Te

𝑇𝑃 90
Recall ( R ) = = = 0.9
of

𝑇𝑃 + 𝐹𝑃 90 + 10
te
itu

𝑃𝑥𝑅 2 𝑥 0.8571 𝑥 0.9

F1 = 2 x 𝑃+𝑅
= 0.8571 + 0.9
= 0.878 = 87.8%
st

5. 𝑌𝑝𝑟𝑒𝑑 − 𝑌 = [ -1, -1, -2, 1, -2]

In
an

2
(𝑌𝑝𝑟𝑒𝑑 − 𝑌 ) = [1, 1, 4, 1, 4]
di
In

Average of (𝑌𝑝𝑟𝑒𝑑 − 𝑌 ) = -1 -> This is the bias

2
6. Average of (𝑌𝑝𝑟𝑒𝑑 − 𝑌 ) = 2.2 -> This is the variance

7. From the linear regression model we have two equations to obtain the optimum value of a & b,
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment 4: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
1 1
𝑛
∑ 𝑦 = 𝑎( 𝑛 ∑ 𝑘) + 𝑏 ⇒ 8. 58851 = 2. 5𝑎 + 𝑏
𝑖 𝑖

Gu
gy
1 2
∑ 𝑥𝑖𝑦𝑖 = 𝑛𝑏( 𝑛 ∑ 𝑥) + 𝑎 ∑ 𝑥𝑖 ⇒ 380. 2522 = 145𝑎 + 25𝑏

lo
𝑖 𝑖 𝑖

Solving these two equations we get:

no
ch
a = 2.0065, b = 3.5722
Te
of

8. Using the equation 𝑦 = 2.0065x + 3.5722, we get the predicted values of y i.e., 𝑦
te

Thus, 𝑦 = [-0.4408, 1.5657, 3.5722, 5.5786, 7.5852, 9.5917, 11.5982, 13.6047, 15.6112,
itu

17.6177]
st
In

1 2
Now MSE is given by, E = 10
∑ (𝑦[𝑖] − 𝑦[𝑖]) = 0.04247
𝑖=1
an
di
In
In
di
an
In
st
itu
te
of
Te
ch
no
lo
gy
Gu
wa
ha
ti
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 10/3/24, 11:33 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 5 - Week 2: Bayesian Decision Theory

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Assignment 2

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-08-07, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-08-05, 15:08 IST
1) Consider a binary classification problem with two classes, A and B with prior probability P (A) = 0.6 and P (B) = 0.4 .Let X be 1 point
Week 0: Prerequisite () a single binary feature that can take values 0 or 1 . Given: P (X = 1|A) = 0.8 and P (X = 0|B) = 0.7. Determine which class the classifier
will classify when X = 1.
Week 1: Introduction ()
Class A
Week 2: Bayesian Class B

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=26&assessment=135 1/5
10/3/24, 11:33 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 5 - Week 2: Bayesian Decision Theory

Decision Theory () Equiprobable for Class A and Class B

Lec 5: Bayesian Decision Not enough information

Theory (Part I) (unit? Yes, the answer is correct.
unit=26&lesson=27) Score: 1
Accepted Answers:
Lec 6: Bayesian Decision
Class A
Theory (Part II) (unit?
unit=26&lesson=28) 2) Consider the following Bayesian network, where F = having the flu and C = coughing: 1 point
Lec 7: Bayes Decision
Theory - Binary Features
(Part III) (unit?
Determine the probability P(F|C) for the following Bayesian network so that it specifies the same joint probabilities as the given network.
unit=26&lesson=31)

Lec 8: Bayesian Decision

Theory (Part IV) (unit?
unit=26&lesson=32)
0.23
Lec 9: Bayesian Decision
Theory (Part V) (unit?
0.03
unit=26&lesson=33) 0.35

Lec 10: Bayesian Belief None of the above.

Networks (unit? Yes, the answer is correct.
unit=26&lesson=34) Score: 1
Accepted Answers:
Feedback form (unit?
0.23
unit=26&lesson=35)

Assignment Solution 3) For the above question, Are C and F independent in the given Bayesian network? 1 point
(unit?unit=26&lesson=67)
Yes.
Lecture Notes (unit?
unit=26&lesson=89) No.
Can't say.
Quiz: Assignment 2
(assessment? Insufficient information.
name=135)
Yes, the answer is correct.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=26&assessment=135 2/5
10/3/24, 11:33 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 5 - Week 2: Bayesian Decision Theory

Score: 1
Week 3- Parametric Accepted Answers:
Approaches and Non- No.
Parametric Density
Estimation () 4) Bayes' decision theory assumes that: 1 point

The feature vectors are dependent on each other.

Week 4: Perceptron
Criteria and The feature vectors are normally distributed.
Discriminative Models The feature vectors are identically distributed.
()
The feature vectors are uniformly distributed.

Week 5: Logistic No, the answer is incorrect.

Score: 0
Regression, Decision
Accepted Answers:
Trees and Hidden
The feature vectors are identically distributed.
Markov Model ()

5) Assume that the word ‘offer’ occurs in 80% of the spam messages in my account. Also, let’s assume ‘offer’ occurs in 10% of my 1 point
Week 6 - Ensemble
desired e-mails. If 30% of the received e-mails are considered as a scam, and I will receive a new message which contains ‘offer’, what is the
Methods ()
probability that it is spam?

Week 7: 0.778
Dimensionality
0.774
Problem ()
0.668
Week 8: Mixture Model 0.664
and Clustering ()
Yes, the answer is correct.
Score: 1
Week 9: Clustering () Accepted Answers:
0.774
Week 10: Artificial
Neural Network () 1.
6) The optimal decision in Bayes Decision Theory is the one that 1 point

Week 11: Introduction Minimizes the error rate.

to Deep Neural Maximizes the error rate.
Networks ()
Minimizes the loss function.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=26&assessment=135 3/5
10/3/24, 11:33 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 5 - Week 2: Bayesian Decision Theory

Problem Solving Maximizes the loss function.

Session - July 2024 () Yes, the answer is correct.
Score: 1
Download () Accepted Answers:
Minimizes the loss function.
Live session () 7) The risk function in Bayesian decision theory combines: 1 point

The prior probabilities and the likelihood function.

The decision boundaries and the feature vectors.
The training set and the test set.
The loss function and the decision rule

Yes, the answer is correct.

Score: 1
Accepted Answers:
The loss function and the decision rule

8) The loss function used in risk-based Bayesian decision theory: 1 point

Quantifies the cost of different types of errors.

Is equal to the likelihood function.
Ignores the prior probabilities of the classes.
Is not used in the decision-making process.

Yes, the answer is correct.

Score: 1
Accepted Answers:
Quantifies the cost of different types of errors.

9) The risk-based Bayesian decision rule accounts for the consequences of different decisions by considering the: 1 point

Number of features in the dataset

The complexity of the classifier
Uncertainty in the data and the associated losses

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=26&assessment=135 4/5
10/3/24, 11:33 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 5 - Week 2: Bayesian Decision Theory

Mean and standard deviation of the feature vectors

Yes, the answer is correct.

Score: 1
Accepted Answers:
Uncertainty in the data and the associated losses

10) The generalized form of a Bayesian network that represents and solves decision problems under uncertain knowledge is known 1 point
as an?

Directed Acyclic Graph

Table of conditional probabilities
Influence diagram
None of the above

Yes, the answer is correct.

Score: 1
Accepted Answers:
Influence diagram

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=26&assessment=135 5/5
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
<Assignment NO 2>: Detailed Solution
Indian Institute of Technology Guwahati
 10/3/24, 11:34 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 6 - Week 3- Parametric Approaches and Non- Parametric Density Estimation

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Week 03 : Assignment 03

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-08-14, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-08-08, 13:33 IST
1) The bandwidth parameter in the Parzen Window method determines: 1 point
Week 0: Prerequisite ()
The number of neighbors to consider for classification
Week 1: Introduction () The size of the neighborhood around a test instance
The dimensionality of the feature space.
Week 2: Bayesian
The complexity of the classifier

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=38&assessment=141 1/6
10/3/24, 11:34 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 6 - Week 3- Parametric Approaches and Non- Parametric Density Estimation

Decision Theory () Yes, the answer is correct.

Score: 1
Accepted Answers:
Week 3- Parametric
The size of the neighborhood around a test instance
Approaches and Non-
Parametric Density 2) If the number of data samples becomes very large. 1 point
Estimation ()
Bayesian Estimation is worse than MLE
Lec 11: Parameter
Maximum Likelihood estimates are slightly bad
Estimation and Maximum
Likelihood Estimation Bayesian Estimation performs same as MLE
(unit?unit=38&lesson=37) None
Lec 12: Parameter Yes, the answer is correct.
Estimation and Bayesian Score: 1
Estimation (unit? Accepted Answers:
unit=38&lesson=39) Bayesian Estimation performs same as MLE
Lec 13: Concept of non-
3) What happens when k = 1 in k -Nearest Neighbor algorithm: 1 point
parametric techniques
(unit?unit=38&lesson=40)
Underfitting
Lec 14: Density Overfitting
Estimation by Parzen
High testing accuracy
Window (unit?
unit=38&lesson=41) All the above

Lec 15: Parzen Window No, the answer is incorrect.

Score: 0
and K nearest neighbor
algorithm (unit? Accepted Answers:
unit=38&lesson=43)
Overfitting

Lecture Notes (unit? 4) There are 18 points in an axis plane namely - 1 point
unit=38&lesson=90)
[(0.8, 0.8)t , (1, 1)t , (1.2, 0.8)t , (0.8, 1.2)t , (1.2, 1.2)t ],
Quiz: Week 03 : belong to class 1;
Assignment 03 [(4, 3)t , (3.8, 2.8)t , (4.2, 2.8)t , (3.8, 3.2)t (4.2, 3.2)t , (4.4, 2.8)t , (4.4, 4.4)t ] ,
(assessment? belong to class 2;
name=141) [(3.2, 0.4)t , (3.2, 0.7)t , (3.8, 0.5)t , (3.5, 1)t , (4, 1)t , (4, 0.7)t ],
belong to class 3.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=38&assessment=141 2/6
10/3/24, 11:34 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 6 - Week 3- Parametric Approaches and Non- Parametric Density Estimation

Feedback form 3 (unit? A new point

unit=38&lesson=44) P = (4.2, 1.8)t introduces into the map. The point P belongs to which class? Use k -nearest neighbor technique with k = 5 to calculate
the result.
Assignment Solution
(unit?unit=38&lesson=66) Class 1

Week 4: Perceptron Class 2

Criteria and Class 3
Discriminative Models None of the above
()
Yes, the answer is correct.
Score: 1
Week 5: Logistic
Accepted Answers:
Regression, Decision
Class 2
Trees and Hidden
Markov Model () 5) Suppose we have two training data points located at 0.5 and 0.7, and we use 0.3 as its rectangle window width. Using the Parzen 1 point
window technique, what would be the probability density if we assume the query point is 0.5?
Week 6 - Ensemble
Methods () 0.5
0.75
Week 7: 2.22
Dimensionality
1.67
Problem ()
Yes, the answer is correct.
Score: 1
Week 8: Mixture Model
Accepted Answers:
and Clustering ()
1.67

Week 9: Clustering ()
6) Suppose that X is a discrete random variable with the following probability 1 point
mass function: where is a parameter.
Week 10: Artificial
(0 ≤ θ ≤ 1)
Neural Network ()

Week 11: Introduction

to Deep Neural The following 10 independent observations were taken from such a distribution:
Networks () (3, 0, 2, 1, 3, 2, 1, 0, 2, 1) . What is the maximum likelihood estimate of θ ?

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=38&assessment=141 3/6
10/3/24, 11:34 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 6 - Week 3- Parametric Approaches and Non- Parametric Density Estimation

Problem Solving 2
Session - July 2024 () 1
0.5
Download ()
0

Live session () Yes, the answer is correct.

Score: 1
Accepted Answers:
0.5

7) Which of the following statements are true about k - nearest neighbor (KNN)- 1 point

Odd value of “K” preferred over even values.

Does more computation on test time rather than train time.
Work well with high dimension.
The optimum value of K for KNN is highly independent on the data.
Yes, the answer is correct.
Score: 1
Accepted Answers:
Odd value of “K” preferred over even values.
Does more computation on test time rather than train time.

8) The disadvantage of using k -NN as classifier: 1 point

Fails while handling large dataset

Fails while handling small dataset
Sensitive to outliers
Training is required
Yes, the answer is correct.
Score: 1
Accepted Answers:
Sensitive to outliers

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=38&assessment=141 4/6
10/3/24, 11:34 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 6 - Week 3- Parametric Approaches and Non- Parametric Density Estimation

9) Consider single observation X that depends on a random parameter .Suppose θ has a prior distribution 1 point

fθ (θ) = λe−λθ for θ ≥ 0, λ > 0

f x (x) = θe−θx |x| > 0
θ

Find the MAP estimation of θ

1
λ+X
1
λ−X
λ
X
X
Yes, the answer is correct.
Score: 1
Accepted Answers:
1
λ+ X

10) The MLE for the data samples X = {x 1 , x 2 , … , x i, … , x k } with the Bernoulli distribution is 1 point

n ⋅ xk
xk
n

Mean of x i
None

Yes, the answer is correct.

Score: 1
Accepted Answers:
Mean of x i

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=38&assessment=141 5/6
10/3/24, 11:34 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 6 - Week 3- Parametric Approaches and Non- Parametric Density Estimation

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=38&assessment=141 6/6
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati

Q.1. Solution - The size of the neighborhood around a test instance

Q.2. Solution - Bayesian Estimation performs same as MLE

Q.3. Solution - Overfitting

Q.4. Solution - Given distances from point 𝑃 to the points 𝑋1 to 𝑋18 are as follows:
𝑑(𝑋1 , 𝑃) = 3.54, 𝑑(𝑋2 , 𝑃) = 3.29, 𝑑(𝑋3 , 𝑃) = 3.16, 𝑑(𝑋4 , 𝑃) = 3.45, 𝑑(𝑋5 , 𝑃) = 3.06,
𝑑(𝑋6 , 𝑃) = 1.22, 𝑑(𝑋7 , 𝑃) = 1.08, 𝑑(𝑋8 , 𝑃) = 1.00, 𝑑(𝑋9 , 𝑃) = 1.46, 𝑑(𝑋10 , 𝑃) = 1.40,
𝑑(𝑋11 , 𝑃) = 1.01, 𝑑(𝑋12, 𝑃) = 2.60, 𝑑(𝑋13 , 𝑃) = 1.72, 𝑑(𝑋14 , 𝑃) = 1.49, 𝑑(𝑋15 , 𝑃) = 1.36,
𝑑(𝑋16 , 𝑃) = 1.06, 𝑑(𝑋17, 𝑃) = 0.824, 𝑑(𝑋18 , 𝑃) = 1.12.
For 𝑘 = 5, the nearest neighbors to point 𝑃 are:
{(𝑋17 , 𝑐𝑙𝑎𝑠𝑠 − 3), (𝑋8 , 𝑐𝑙𝑎𝑠𝑠 − 2), (𝑋11 , 𝑐𝑙𝑎𝑠𝑠 − 2), (𝑋16 , 𝑐𝑙𝑎𝑠𝑠 − 3), (𝑋7 , 𝑐𝑙𝑎𝑠𝑠 − 2)}
Since the majority of the nearest neighbors belong to class 2, the point 𝑃 is classified as belonging
to class 2.

Q.5. Solution - Given:

• Training data points: 0.5 and 0.7

• Rectangle window width (ℎ): 0.3

• Query point: 0.5

The Parzen window counts the number of training points that fall within the window centered
around the query point. The window is defined by the interval:
ℎ ℎ
[0.5 − , 0.5 + ] = [0.5 − 0.15,0.5 + 0.15] = [0.35,0.65]
2 2
Step 1: Determine the points within the window:
The window is [0.35,0.65]. The training point 0.5 lies within this window, but 0.7 does not.
Step 2: Count the points within the window:
There is 1 point within the window (0.5).
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati

Step 3: Calculate the probability density:

The probability density function 𝑓(𝑥) is given by:
Number of points within the window
𝑓 (𝑥 ) =
Total number of points × Window width
Substituting the values:
1 1
𝑓(𝑥) = = = 1.67
2 × 0.3 0.6
Thus, the probability density at the query point 0.5 is 1.67.

Q.6. Solution - Given that the samples are (3,0,2,1,3,2,1,0,2,1), the likelihood function is:
𝐿(𝜃) = 𝑃(𝑋 = 3)𝑃(𝑋 = 0)𝑃(𝑋 = 2)𝑃(𝑋 = 1)𝑃(𝑋 = 3)𝑃(𝑋 = 2)𝑃(𝑋 = 1)𝑃(𝑋 = 0)𝑃(𝑋
= 2)𝑃(𝑋 = 1)
𝑛 3
2𝜃 2 𝜃 3 2(1 − 𝜃) 1−𝜃 2
𝐿(𝜃) = ∏ 𝑃 (𝑋𝑖 ∣ 𝜃) = ( ) ( ) ( ) ( )
3 3 3 3
𝑖=1

The log-likelihood function is:

𝑛

𝑙(𝜃) = ln𝐿(𝜃) = ∑ ln 𝑃(𝑋𝑖 ∣ 𝜃)

𝑖=1

2 1 2 1
𝑙(𝜃) = 2 (ln + ln𝜃) + 3 (ln + ln𝜃) + 3 (ln + ln(1 − 𝜃)) + 2 (ln + ln(1 − 𝜃))
3 3 3 3

= 𝐶 + 5ln𝜃 + 5ln(1 − 𝜃)
2 1 2 1
where 𝐶 = 2ln 3 + 3ln 3 + 3ln 3 + 2ln 3 is a constant.

Taking the derivative of 𝑙(𝜃) with respect to 𝜃:

𝑑𝑙(𝜃) 5 5
= − =0
𝑑𝜃 𝜃 1−𝜃
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati

Thus, the maximum likelihood estimate 𝜃̂MLE is:

𝜃̂MLE = 0.5

Q.7. Solution - Odd value of “K” preferred over even values. and
Does more computation on test time rather than train time.

Q.8. Solution - Sensitive to outliers

Q.9. Solution - The MAP estimation is given by

𝜃MAP = argmax[𝑓𝑋|𝜃 (𝑥) ⋅ 𝑓𝜃 (𝜃)].

𝜃

Substituting the given functions:

𝜃MAP = argmax[𝜃𝑒 −𝜃𝑥 ⋅ 𝜆𝑒 −𝜆𝜃 ]

𝜃

Simplifying, we get:

𝜃MAP = argmax[𝜆𝜃𝑒 −𝜃(𝑥+𝜆) ].

𝜃

Taking the natural logarithm of the function:

ln(𝜃MAP ) = ln(𝜆𝜃) − 𝜃(𝑥 + 𝜆)
ln(𝜃MAP ) = ln(𝜆) + ln(𝜃) − 𝜃(𝑥 + 𝜆).
Differentiate with respect to 𝜃:
𝑑 1
ln(𝜃MAP ) = − (𝑥 + 𝜆) = 0.
𝑑𝜃 𝜃
Solve for 𝜃:
1
=𝑥+𝜆
𝜃
1
𝜃MAP = .
𝑥+𝜆
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati

Thus, the MAP estimate for 𝜃 is

1
𝜃MAP = .
𝜆+𝑋
Q.10. Solution - The likelihood function for 𝑛 independent Bernoulli trials can be expressed as:
𝑛

𝐿(𝑝) = ∏ 𝑝 𝑥𝑖 (1 − 𝑝)1−𝑥𝑖
𝑖=1

Taking the logarithm of the likelihood function:

𝑛

ln𝐿(𝑝) = ∑(𝑥𝑖 ln𝑝 + (1 − 𝑥𝑖 )ln(1 − 𝑝))

𝑖=1
𝑛 𝑛

ln𝐿(𝑝) = ∑ 𝑥𝑖 ln𝑝 + ∑(1 − 𝑥𝑖 ) ln(1 − 𝑝)

𝑖=1 𝑖=1

Simplifying: Let 𝑘 = ∑𝑛𝑖=1 𝑥𝑖 be the number of successes (1’s) in the samples:

ln𝐿(𝑝) = 𝑘ln𝑝 + (𝑛 − 𝑘)ln(1 − 𝑝)
Taking the Derivative: Setting the derivative of the log-likelihood function to zero to find the MLE:
𝑑 𝑘 𝑛−𝑘
ln𝐿(𝑝) = − =0
𝑑𝑝 𝑝 1−𝑝
Solving for 𝑝:
𝑘 𝑛−𝑘
=
𝑝 1−𝑝
𝑘(1 − 𝑝) = (𝑛 − 𝑘)𝑝
𝑘 − 𝑘𝑝 = 𝑛𝑝 − 𝑘𝑝
𝑘 = 𝑛𝑝
𝑘
𝑝=
𝑛
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati

Thus, the MLE for 𝑝 is the sample mean:

𝑛
1
𝑝̂ = ∑ 𝑥𝑖
𝑛
𝑖=1

From the provided options, the correct answer is: Mean of 𝑥𝑖 (i.e., 𝑝̂ = Mean of𝑥𝑖 )
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 03: Detailed Solution
Indian Institute of Technology Guwahati
 10/3/24, 11:35 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 7 - Week 4: Perceptron Criteria and Discriminative Models

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Week 4 : Assignment 4

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-08-21, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-08-21, 15:49 IST
1) How many decision boundaries in one vs all classification? 1 point
Week 0: Prerequisite ()
c
Week 1: Introduction ()
c(c−1)
2
Week 2: Bayesian

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=45&assessment=142 1/5
10/3/24, 11:35 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 7 - Week 4: Perceptron Criteria and Discriminative Models

Decision Theory () c(c+1)

2
None of the above
Week 3- Parametric
Approaches and Non- Yes, the answer is correct.
Parametric Density Score: 1
Estimation () Accepted Answers:
c
Week 4: Perceptron 2) To avoid the problem of ambiguous region of linear discriminant function for c categories, we can 1 point
Criteria and
Discriminative Models Define c linear function gi (x), one for each class for i = 1, 2, . . . , c
()
Assign x to wj if gi (x) < gj (x) for all i ≠ j
Lec 16: Linear
Discriminant Functions Take a linear machine classifier
and Perceptron Criteria All the above
(Part I) (unit?
Yes, the answer is correct.
unit=45&lesson=46)
Score: 1
Lec 17: Linear Accepted Answers:
Discriminant Functions All the above
and Perceptron Criteria
(Part II) (unit? 3) Which of the following statements is true about the learning rate in Gradient Descent? 1 point
unit=45&lesson=47)
A very high learning rate may lead to oscillation
Lec 18: Linear
Discriminant Functions A lower learning rate may lead to faster convergence
and Perceptron Criteria The learning rate doesn't determine the size of the steps taken towards the minimum
(Part III) (unit?
The learning rate has no effect on the convergence of Gradient Descent
unit=45&lesson=48)
Yes, the answer is correct.
Lec 19: Support Vector Score: 1
Machine (Part I) (unit? Accepted Answers:
unit=45&lesson=49) A very high learning rate may lead to oscillation
Lec 20: Support Vector
Machine (Part II) (unit? 4) In the Perceptron algorithm for a binary classifier, what happens to the weights when a positive misclassified point is 1 point
unit=45&lesson=50) encountered?

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=45&assessment=142 2/5
10/3/24, 11:35 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 7 - Week 4: Perceptron Criteria and Discriminative Models

Lecture Notes (unit? It remains the same

unit=45&lesson=91)
It is increased
Quiz: Week 4 : It is decreased
Assignment 4
It is multiplied by a constant
(assessment?
name=142) Yes, the answer is correct.
Score: 1
Feedback form 4 (unit? Accepted Answers:
unit=45&lesson=52) It is increased
Assignment Solution 5) Let wij represents weight between node i at layer k and node j at layer (k-1) of a given multilayer perceptron. The weight 1 point
(unit?unit=45&lesson=93)
updation using gradient descent method is given by: (α and E represent learning rate and Error in the output respectively)

Week 5: Logistic
Regression, Decision ∂E
Wij (t + 1) = Wij (t) + α ∂W ,0 ≤ α ≤ 1
ij
Trees and Hidden
Markov Model () ∂E
Wij (t + 1) = Wij (t) − α ∂W ,0 ≤ α ≤ 1
ij

Week 6 - Ensemble
∂E
Methods () Wij (t + 1) = α ∂W ,0 ≤ α ≤ 1
ij

∂E
Week 7: Wij (t + 1) = −α ∂W ,0 ≤ α ≤ 1
ij
Dimensionality
Problem () Yes, the answer is correct.
Score: 1
Accepted Answers:
Week 8: Mixture Model ∂E
and Clustering ()
Wij (t + 1) = Wij (t) − α ∂W ,0 ≤ α ≤ 1
ij

Week 9: Clustering () 6) A 4-input neuron has weights 3, 4, 5 and 6. The transfer function is linear with the constant of proportionality being equal to 3. 1 point
The inputs are 6, 12, 10 and 20 respectively. What will be the output?
Week 10: Artificial
Neural Network () 238
76
Week 11: Introduction 708
to Deep Neural
123

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=45&assessment=142 3/5
10/3/24, 11:35 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 7 - Week 4: Perceptron Criteria and Discriminative Models

Networks () Yes, the answer is correct.

Score: 1
Accepted Answers:
Problem Solving
708
Session - July 2024 ()
7) Which of these is true about discriminant classifiers? 1 point
Download ()
Assume conditional independence of features
Live session () Robust to outliers
Can perform classification if some missing data points are present
All the above
No, the answer is incorrect.
Score: 0
Accepted Answers:
Robust to outliers

8) A set of training samples are given below- 1 point

Using Support vector machine algorithm, the Marginal line for the classification can be calculated as-

−5.32x1 − 7.193x2 + 9.09 = 0

−6.67x1 + 8.134x2 − 9.09 = 0

−7.21x1 − 9.173x2 + 9.09 = 0

8.21x1 + 7.12x2 − 9.09 = 0

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=45&assessment=142 4/5
10/3/24, 11:35 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 7 - Week 4: Perceptron Criteria and Discriminative Models

No, the answer is incorrect.

Score: 0
Accepted Answers:
−7.21x1 − 9.173x2 + 9.09 = 0
9) In refer to Q.8, A new test sample (0.5,0.5) is found. The class of the given sample is- 1 point

Positive
Negative
Both class
Can't say

Yes, the answer is correct.

Score: 1
Accepted Answers:
Positive

10) What is the main objective of a Support Vector Machine (SVM)? 1 point

To maximize the number of support vectors

To minimize the margin between classes
To maximize the training accuracy
To find a hyperplane that separates classes with the maximum margin

Yes, the answer is correct.

Score: 1
Accepted Answers:
To find a hyperplane that separates classes with the maximum margin

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=45&assessment=142 5/5
 NPTEL Online Certification Course
Machine Learning and Deep Learning” Fundamentals and Applications
Assignment 4: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
6. The output is found by multiplying the weights with their respective inputs, summing the results

Gu
and multiplying with the transfer function. Therefore,
Output = 3 x (3 x 6 + 4 x 12 + 5 x 10 + 6 x 20)

gy
= 3 x (18 + 48 + 50 + 120)

lo
= 708
no
8.
ch
Te
of
te
itu
st
In
an
di
In
9. For a new sample (0.5, 0.5), the class is determined by substituting it in the marginal line
calculated in the above solutions.

Thus, -7.20 (0.5) - 9.173 (0.5) + 9.085 = 0.895, which is a positive number.

Therefore, it belongs to the positive class.

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 10/3/24, 11:36 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 8 - Week 5: Logistic Regression, Decision Trees and Hidden Markov Model

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Week 05 : Assignment 05

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-08-28, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-08-27, 14:11 IST
1) Which of the following statements is true about logistic regression? 1 point
Week 0: Prerequisite ()
It is robust to extreme outliers in the data
Week 1: Introduction () It is not suitable for binary classification problems
It does not make any assumptions about the data distribution
Week 2: Bayesian
It requires a large amount of training data to perform well

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=53&assessment=146 1/6
10/3/24, 11:36 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 8 - Week 5: Logistic Regression, Decision Trees and Hidden Markov Model

Decision Theory () Yes, the answer is correct.

Score: 1
Accepted Answers:
Week 3- Parametric
It does not make any assumptions about the data distribution
Approaches and Non-
Parametric Density 2) Logistic regression is a machine learning algorithm that is used to predict the probability of a _? 1 point
Estimation ()
Categorical independent variable
Week 4: Perceptron Categorical dependent variable
Criteria and
Numerical independent variable
Discriminative Models
() Numerical dependent variable

Yes, the answer is correct.

Week 5: Logistic Score: 1
Regression, Decision Accepted Answers:
Trees and Hidden Categorical dependent variable
Markov Model ()
3) What’s the the hypothesis of logistic regression? 1 point
Lec 21: Logistic
Regression (unit? To limit the cost function between 0 and 1
unit=53&lesson=54) To limit the cost function between -1 and 1
Lec 22: Decision Tree To limit the cost function between -infinity and +infinity
(unit?unit=53&lesson=55)
To limit the cost function between 0 and +infinity
Lec 23: Hidden Markov Yes, the answer is correct.
Model (HMM) (unit? Score: 1
unit=53&lesson=56) Accepted Answers:
To limit the cost function between 0 and 1
Lecture Notes (unit?
unit=53&lesson=92)

Quiz: Week 05 :
Assignment 05
(assessment?
name=146)

Feedback form 5 (unit?

unit=53&lesson=58)

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=53&assessment=146 2/6
10/3/24, 11:36 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 8 - Week 5: Logistic Regression, Decision Trees and Hidden Markov Model

Assignment Solution 4) The dataset of pass or fail in an exam for five students are given in the table. 1 point
(unit?unit=53&lesson=96)

Week 6 - Ensemble
Methods ()

Week 7:
Dimensionality
Problem ()

Week 8: Mixture Model

and Clustering ()

Week 9: Clustering () Calculate the probability of pass for student who studies 34 hours.

Week 10: Artificial Assume the model suggested by the optimizer for odds of passing the course is:
Neural Network ()
log(odds) = - 64 + 2 x hours

Week 11: Introduction

to Deep Neural
0.932
Networks ()
0.952
Problem Solving 0.982
Session - July 2024 () 0.992

No, the answer is incorrect.

Download ()
Score: 0
Accepted Answers:
Live session () 0.982

5) Which of the following is used to measure the quality of a split in a decision tree? 1 point

Information gain
Gini index

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=53&assessment=146 3/6
10/3/24, 11:36 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 8 - Week 5: Logistic Regression, Decision Trees and Hidden Markov Model

Entropy
All of the above
Yes, the answer is correct.
Score: 1
Accepted Answers:
All of the above
6) There are six instances for attributes a1 and a2, with positive and negative classification results given below: 1 point

What is the information gain of a1 relative to these training instances?

0.0523
0.0817
0.025
0

No, the answer is incorrect.

Score: 0
Accepted Answers:
0.0817

7) How is pruning used to prevent overfitting in decision trees? 1 point

By limiting the depth of the tree

By removing branches with low information gain
By setting a minimum number of instances per leaf
By limiting the maximum number of leaves

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=53&assessment=146 4/6
10/3/24, 11:36 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 8 - Week 5: Logistic Regression, Decision Trees and Hidden Markov Model

Yes, the answer is correct.

Score: 1
Accepted Answers:
By removing branches with low information gain
8) Where does the additional variables are added in HMM? 1 point

Temporal model
Reality model
Probabilistic model
All the above

No, the answer is incorrect.

Score: 0
Accepted Answers:
Temporal model

9) Which allows for a simple and matrix implementation of all the basic algorithm? 1 point

HMM
Restricted structure of HMM
Temporary model
Reality model

Yes, the answer is correct.

Score: 1
Accepted Answers:
Restricted structure of HMM

10) Once a day (e.g. at noon), the weather is observed as one of 1 point
State 1: Rainy (R); State 2: Cloudy (C); State 3: Sunny (S)

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=53&assessment=146 5/6
10/3/24, 11:36 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 8 - Week 5: Logistic Regression, Decision Trees and Hidden Markov Model

The state transition probabilities are:

⎡ 0.4 0.3 0.3 ⎤
A = ⎢ 0.2 0.6 0.2 ⎥
⎣ 0.1 0.1 0.8 ⎦

Given that the weather on day 1 is sunny, what is the probability that the weather for the next 7 days will be "sunny-sunny-rainy-rainy-rainy-
cloudy-sunny"?

1.536 ∗ 10−4

1.536 ∗ 10−3

1.356 ∗ 10−4

1.356 ∗ 10−3
No, the answer is incorrect.
Score: 0
Accepted Answers:
1.536 ∗ 10−4

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=53&assessment=146 6/6
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
<Assignment NO 5>: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati
 10/3/24, 11:37 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 9 - Week 6 - Ensemble Methods

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Week 6 : Assignment 6

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-09-04, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-09-02, 10:27 IST
1) In AdaBoost, the final prediction is determined by 1 point
Week 0: Prerequisite ()
A simple average of weak learners' predictions
Week 1: Introduction () The prediction of the last weak
A weighted sum of weak learners' predictions
Week 2: Bayesian
A random selection of weak learners' predictions

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=59&assessment=149 1/5
10/3/24, 11:37 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 9 - Week 6 - Ensemble Methods

Decision Theory () Yes, the answer is correct.

Score: 1
Accepted Answers:
Week 3- Parametric
Approaches and Non-
A weighted sum of weak learners' predictions
Parametric Density 2) Boosting algorithms assign: 1 point
Estimation ()
Equal weights to all training instances
Week 4: Perceptron Weights based on the difficulty of each instance
Criteria and
Weights inversely proportional to the class distribution
Discriminative Models
() Random weights to each base model

Yes, the answer is correct.

Week 5: Logistic Score: 1
Regression, Decision Accepted Answers:
Trees and Hidden Weights based on the difficulty of each instance
Markov Model ()
3) Consider a binary classification problem where a single weak classifier has an accuracy of 60%. By combining multiple classifiers 1 point
Week 6 - Ensemble through majority voting, the overall accuracy improves. If 5 independent classifiers are combined, what is the probability that the ensemble's
Methods () accuracy exceeds 60%?

Lec 24: Ensemble 50%

Classifiers (Part I) (unit? 60.25%
unit=59&lesson=61)
90.63%
Lec 25: Ensemble 68.26%
Classifiers (Part II) (unit?
unit=59&lesson=62) Yes, the answer is correct.
Score: 1
Feedback form 6 (unit? Accepted Answers:
unit=59&lesson=70) 68.26%

Lecture Notes (unit?

unit=59&lesson=114) 4) _______ is useful when high variance models are available. 1 point

Quiz: Week 6 : AdaBoost

Assignment 6
Stacking
Boosting

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=59&assessment=149 2/5
10/3/24, 11:37 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 9 - Week 6 - Ensemble Methods

(assessment? Bagging
name=149)
Yes, the answer is correct.
Assignment Solution Score: 1
(unit?unit=59&lesson=97) Accepted Answers:
Bagging
Week 7:
Dimensionality 5) Bagging techniques combine the predictions of base models by: 1 point
Problem ()
Taking the average of their predictions

Week 8: Mixture Model Taking the maximum or majority vote of their predictions
and Clustering () Weighting the predictions based on their performance
Ignoring the predictions of the base models
Week 9: Clustering ()
Yes, the answer is correct.
Score: 1
Week 10: Artificial
Accepted Answers:
Neural Network () Taking the maximum or majority vote of their predictions

Week 11: Introduction 6) In the third iteration of AdaBoost, the weight assigned to a misclassified data point is 0.4. If the initial weight 1 point
to Deep Neural
for all data points is 1, what was the misclassification rate of this data point at the end of the second iteration?
Networks ()
70%
Problem Solving
90%
Session - July 2024 ()
81%

Download () 86%

No, the answer is incorrect.

Live session () Score: 0
Accepted Answers:
86%

7) Which of the following does bagging help to reduce in machine learning models? (Select all that apply) 1 point

Model variance
Bias

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=59&assessment=149 3/5
10/3/24, 11:37 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 9 - Week 6 - Ensemble Methods

Overfitting
Model complexity

Yes, the answer is correct.

Score: 1
Accepted Answers:
Model variance
Overfitting

8) Random Forest can handle 1 point

Large datasets
Linear and nonlinear data
Classification and regression tasks
All of the above

Yes, the answer is correct.

Score: 1
Accepted Answers:
All of the above

9) The feature importance in Random Forest is measured by: 1 point

The number of instances that use a particular feature for splitting

The average depth of a feature in the decision trees
The coefficient value assigned to each feature
The number of decision trees that use a particular feature

Yes, the answer is correct.

Score: 1
Accepted Answers:
The number of instances that use a particular feature for splitting

10) In stacking, what is a meta-learner? 1 point

A model that performs data preprocessing

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=59&assessment=149 4/5
10/3/24, 11:37 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 9 - Week 6 - Ensemble Methods

A base model trained on the original dataset

A model trained on the predictions of base models
A model used to aggregate the predictions of base models
Yes, the answer is correct.
Score: 1
Accepted Answers:
A model trained on the predictions of base models

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=59&assessment=149 5/5
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati

Q.1. Solution - A weighted sum of weak learners' predictions.

Q.2. Solution - Weights based on the difficulty of each instance.

Q.3. Solution - The probability of a classifier being correct is p = 0.6 and incorrect is 1-p = 0.4.
The ensemble will be correct if at least 3 out of 5 classifiers are correct.
We can calculate the probability for 3, 4, and 5 classifiers being correct using the binomial
distribution formula:

P ( X= k )= n pk (1− p )
() n− k

For, k= 3,4 and 5

P ( X= 3 )= 5 ( 0.6 ) ( 0.4 ) = 10× 0.216× 0.16= 0.3456

() 3 2

P ( X= 4 )= (5)( 0.6) ( 0.4) = 5× 0.1296× 0.4= 0.2592

4 1

P ( X= 5 )= (5)( 0.6 ) ( 0.4 ) = 1× 0.07776× 1= 0.07776

5 0

The total probability that the ensemble is correct is:

𝑃(𝑋 ≥ 3) = 𝑃(𝑋 = 3) + 𝑃(𝑋 = 4) + 𝑃(𝑋 = 5)
𝑃(𝑋 ≥ 3) = 0.3456 + 0.2592 + 0.07776 = 0.68256
Thus, the probability that the ensemble’s accuracy exceeds 60% is 68.26%.
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati

Q.4. Solution – Bagging.

Q.5. Solution - Taking the maximum or majority vote of their predictions.

Q.6. Solution - In AdaBoost, the weight of a misclassified data point is updated as:
(𝑡+1) (𝑡)
𝑤𝑖 = 𝑤𝑖 × exp(𝛼𝑡 )
Given that the weight for a misclassified data point in the third iteration is 0.4, we have:
0.4 = 1 × exp(𝛼2 )
Thus,
𝛼2 = ln(0.4) ≈ −0.916
Using:
1 1 − 𝑒2
𝛼2 = ln ( )
2 𝑒2

we can solve for e2 :

1. Set up the equation:
1 − 𝑒2
−1.832 = ln ( )
𝑒2
2. Exponentiate:
1 − 𝑒2
0.160 =
𝑒2
3. Rearranging gives:
1
𝑒2 = ≈ 0.862
1.160
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati

Q.7. Solution - Model variance and Overfitting.

Q.8. Solution - All of the above.

Q.9. Solution - The number of instances that use a particular feature for splitting.

Q.10. Solution - A model trained on the predictions of base models.

 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 06: Detailed Solution
Indian Institute of Technology Guwahati
 10/3/24, 11:38 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 10 - Week 7: Dimensionality Problem

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Week 07 : Assignment 07

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-09-11, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-09-09, 14:54 IST
1) Determine the principle component of the 2D data- (−3, −3), (−1, −1), (1, 1), (3, 3) 1 point
Week 0: Prerequisite ()
0
Week 1: Introduction () 40
20
Week 2: Bayesian

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=71&assessment=150 1/5
10/3/24, 11:38 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 10 - Week 7: Dimensionality Problem

Decision Theory () 1
√2

Week 3- Parametric Yes, the answer is correct.

Score: 1
Approaches and Non-
Accepted Answers:
Parametric Density 1
Estimation () √2

2) ⎡4 2⎤
⎡1 2⎤
1 point
⎢5 0⎥
Week 4: Perceptron
⎢2 3⎥ ⎢ ⎥
⎢ ⎥ ⎢5 2⎥
Criteria and
For class independent LDA for two classes w1 = ⎢
⎢3 3 ⎥ and w2 = ⎢
⎥
⎢3
⎥ to cluster into two groups. The new
2⎥
Discriminative Models
⎢ ⎥ ⎢ ⎥
() ⎢4 5⎥ ⎢ ⎥
⎣5 ⎦ ⎢5 3⎥
Week 5: Logistic
5 ⎣6 3⎦
Regression, Decision transformation point will be
Trees and Hidden
Markov Model () (−2.453, 2.671)
(−1.457, 1.598)
Week 6 - Ensemble (−1.326, 1.473)
Methods ()
(−2.142, 2.351)
Week 7: No, the answer is incorrect.
Dimensionality Score: 0
Problem () Accepted Answers:
(−1.457, 1.598)
Lec 26: Dimensionality
Problem and Principal 3) After applying PCA, one of the components is (0.5, 0.5, 0.5, 0.5). The other components are? 1 point
Component Analysis 1. (0.5,0.5,0,0)
(unit?unit=71&lesson=72) 2. (0,0,-0.5,-0.5)
3. (0.5,0.5,-0.5,-0.5)
Lec 27: Principal
4. (-0.5,-0.5,0.5,0.5)
Component Analysis
(unit?unit=71&lesson=73)
1, 2
Lec 28: Linear
Discriminant Analysis 1, 3
2, 4

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=71&assessment=150 2/5
10/3/24, 11:38 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 10 - Week 7: Dimensionality Problem

(LDA) (Part I) (unit? 3, 4

unit=71&lesson=74)
Yes, the answer is correct.
Lec 29: Linear Score: 1
Discriminant Analysis Accepted Answers:
(LDA) (Part II) (unit? 3, 4
unit=71&lesson=75)
4) You have a dataset with two classes. Class 1 has 50 data points with a mean of [3, 5] and a 1 point
Lecture Notes (unit? 9 3 12 4
unit=71&lesson=115) covariance matrix of [ ]. Class 2 has 70 data points with a mean of [6, 8] and a covariance matrix of [ ]. What
3 6 4 8
Quiz: Week 07 : is the between-class scatter matrix?
Assignment 07
(assessment? [[18]]
name=150)
1 1
Feedback form 7 (unit? [ ]
1 1
unit=71&lesson=77)

9 9
Assignment Solution
[ ]
(unit? 9 9
unit=71&lesson=160)
18 18
Week 8: Mixture Model [ ]
18 18
and Clustering ()
No, the answer is incorrect.
Score: 0
Week 9: Clustering ()
Accepted Answers:
9 9
Week 10: Artificial [ ]
9 9
Neural Network ()

5) __________seeks directions that are efficient for discrimination. 1 point

Week 11: Introduction
to Deep Neural PCA
Networks ()
SVD
Discriminant analysis
Problem Solving
Session - July 2024 () L2 regularization

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=71&assessment=150 3/5
10/3/24, 11:38 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 10 - Week 7: Dimensionality Problem

Yes, the answer is correct.

Download () Score: 1
Accepted Answers:
Live session () Discriminant analysis

6) If all eigen values are same then 1 point

PCA won't be able to select principal components.

PCA will select multiple principal components
PCA acts as discriminant operation
None of the above

No, the answer is incorrect.

Score: 0
Accepted Answers:
PCA won't be able to select principal components.

7) In PCA, if you choose to retain 'k' principal components, what is the result in terms of dimensionality? 1 point

Dimensionality increases
Dimensionality decreases
Dimensionality remains unchanged
Dimensionality is multiplied by 'k'
Yes, the answer is correct.
Score: 1
Accepted Answers:
Dimensionality decreases

8) PCA is sensitive to the scale of the data. To ensure accurate results, you should- 1 point

Only apply PCA to binary categorical data

Apply PCA without any preprocessing
Normalize the data by dividing by the mean
Standardize the data before applying PCA

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=71&assessment=150 4/5
10/3/24, 11:38 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 10 - Week 7: Dimensionality Problem

Yes, the answer is correct.

Score: 1
Accepted Answers:
Standardize the data before applying PCA

9) The goal of LDA is to:- 1 point

Maximize within-class variance and minimize between-class variance

Minimize within-class variance and maximize between-class variance
Minimize both within-class and between-class variance
Maximize both within-class and between-class variance

Yes, the answer is correct.

Score: 1
Accepted Answers:
Minimize within-class variance and maximize between-class variance

10) The number of linear discriminants that can be obtained in LDA is equal to: 1 point

The number of classes

The number of features
The number of data points
The minimum of the number of classes and features
No, the answer is incorrect.
Score: 0
Accepted Answers:
The number of classes

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=71&assessment=150 5/5
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment 7: Detailed Solution

ti
Indian Institute of Technology Guwahati

ha
wa
Gu
1. Here, 𝑥 = 0 and 𝑦 = 0

gy
𝑉𝑎𝑟(𝑋) 𝐶𝑜𝑣(𝑋,𝑌) 1 20 20
Therefore, C = ⎡⎢ ⎤⎥ =
3
⎡⎢ ⎤⎥
⎣ 𝐶𝑜𝑣(𝑌,𝑋) 𝑉𝑎𝑟(𝑌) ⎦ ⎣ 20 20 ⎦

lo
Correspnding eigen values: |C - λ𝐼| = 0 ⇒ λ1 =
no 20
3
and λ2 = 0.
ch
Te

As λ1 > λ2, corresponding to λ1, principal component is the eigen vector 𝑣1

Therefore, C𝑣1 = λ1𝑣1

of
te

1
⎡ ⎤
itu

2
By solving the above equation, 𝑣1 = ⎢ ⎥
⎢ 1
⎥
⎣ 2 ⎦
st
In

2. For ω1,
3
μ1 = ⎡⎢ ⎤⎥
an

⎣ 3.6 ⎦
di

𝑇 2.5 2
𝑆1 = ∑ (x - μ1)(𝑥 − μ1) = ⎡⎢ ⎤⎥
⎣ 2 1.8 ⎦
In

𝑥 ∈ ω1

For ω2,
4.66
μ2 = ⎡⎢ ⎤⎥
⎣ 2 ⎦
𝑇 1.06 0.2
𝑆2 = ∑ (x - μ2)(𝑥 − μ2) = ⎡⎢ ⎤⎥
𝑥 ∈ ω2 ⎣ 0.2 1.2 ⎦
 3.56 2.2
Now, 𝑆𝑊 = 𝑆1 + 𝑆2 = ⎡⎢ ⎤⎥
⎣ 2.2 3 ⎦

𝑇 2.75 −2.65
𝑆𝐵 = (μ1 − μ2 )(μ1 − μ2) = ⎡⎢ ⎤⎥
⎣ −2.65 2.56 ⎦

ti
ha
* −1 −1.457
𝑤 = 𝑆𝑊 (μ1 − μ2 ) = ⎡⎢ ⎤⎥

wa
⎣ 1.598 ⎦

Gu
3. One of the component is given as (0.5, 0.5, 0.5, 0.5), the other components needs to be
orthonormal, which means their dot product should be 0 and their length should be 1.

gy
Now, Option 1 - (0.5, 0.5, 0, 0)

lo
Orthogonality - (0.5)(0.5) + (0.5)(0.5) + 0 + 0 = 0.5 ≠ 0
Therefore, not orthogonal
no
ch
Te

Option 2 - (0, 0, -0.5, -0.5)

of

Orthogonality - 0 + 0 + (-0.5)(0.5) + (-0.5)(0.5) = -0.5 ≠ 0

Therefore, not orthogonal
te
itu

Option 3 - (0.5, 0.5, -0.5, -0.5)

st

Orthogonality - (0.5)(0.5) + (0.5)(0.5) + (-0.5)(0.5) + (-0.5)(0.5) = 0

In

2 2 2 2
Length = (0. 5) + (0. 5) + (− 0. 5) + (− 0. 5) = 1
an
di

Option 4 - (-0.5, -0.5, 0.5, 0.5)

In

Orthogonality - (-0.5)(0.5) + (-0.5)(0.5) + (0.5)(0.5) + (0.5)(0.5) = 0

2 2 2 2
Length = (− 0. 5) + (− 0. 5) + (0. 5) + (0. 5) = 1

Therefore, option 3 and option 4 are the other components as they are orthonormal.
4. Here,
3 6
μ1 = ⎡⎢ ⎤⎥ and μ = ⎡⎢ ⎤⎥
⎣ 5 ⎦ 2 ⎣ 8 ⎦
Now,
−3
μ1- μ2 = ⎡⎢ ⎤⎥

ti
⎣ −3 ⎦

ha
wa
𝑇 9 9
𝑆𝐵 = (μ1 − μ2 )(μ1 − μ2) = ⎡⎢ ⎤⎥
⎣ 9 9 ⎦

Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 10/3/24, 11:39 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 11 - Week 8: Mixture Model and Clustering

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Week 8 : Assignment 8

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-09-18, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-09-16, 14:06 IST
1) The EM algorithm is based on the principle of 1 point
Week 0: Prerequisite ()
Maximum likelihood estimation
Week 1: Introduction () Bayesian theory
Feature selection
Week 2: Bayesian
Regularization

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=78&assessment=151 1/6
10/3/24, 11:39 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 11 - Week 8: Mixture Model and Clustering

Decision Theory () Yes, the answer is correct.

Score: 1
Accepted Answers:
Week 3- Parametric
Approaches and Non-
Maximum likelihood estimation
Parametric Density 2) Each component in a GMM is characterized by 1 point
Estimation ()
Mean, median, and mode
Week 4: Perceptron Mean and covariance matrix
Criteria and
Variance, skewness, and kurtosis
Discriminative Models
() Standard deviation and correlation coefficient
Yes, the answer is correct.
Week 5: Logistic Score: 1
Regression, Decision Accepted Answers:
Trees and Hidden Mean and covariance matrix
Markov Model ()
3) Which of the following statements is true about GMMs? 1 point
Week 6 - Ensemble
Methods ()
GMMs are only used for regression tasks
GMMs can only model data with a single cluster
Week 7: GMMs are sensitive to the initial placement of the Gaussian components
Dimensionality GMMs are a type of neural network
Problem ()
Yes, the answer is correct.
Score: 1
Week 8: Mixture Model
Accepted Answers:
and Clustering ()
GMMs are sensitive to the initial placement of the Gaussian components
Lec 30: Gaussian Mixture
Model and EM Algorithm 4) What does each component in a GMM represent? 1 point
(unit?unit=78&lesson=79)
A data point in the dataset
Lec 31: K-means
A principal component
clustering (unit?
unit=78&lesson=80) A cluster center
A Gaussian distribution with its mean and covariance

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=78&assessment=151 2/6
10/3/24, 11:39 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 11 - Week 8: Mixture Model and Clustering

Yes, the answer is correct.

Lecture Notes (unit?
Score: 1
unit=78&lesson=117)
Accepted Answers:
Quiz: Week 8 : A Gaussian distribution with its mean and covariance
Assignment 8
(assessment?
5) Let there be 10 data points: {−1, 27, 31, 2, 59, 3, 61, 34, 0, 12}. Use K-means clustering for 3 iterations to 0 points
name=151) cluster them into 3 clusters. The initial labels are {1, 1, 1, 2, 2, 2, 3, 3, 3, 1}. List out the labels after 3 iterations.

Feedback form 8 (unit?

unit=78&lesson=82) {1, 2, 2, 1, 3, 1, 3, 2, 1, 1}

Week 9: Clustering () {1, 1, 2, 1, 3, 1, 3, 2, 1, 2}

{1, 2, 3, 1, 3, 1, 3, 2, 3, 1}
Week 10: Artificial
Neural Network ()
{1, 1, 2, 1, 3, 1, 3, 3, 1, 1}
Week 11: Introduction No, the answer is incorrect.
Score: 0
to Deep Neural
Accepted Answers:
Networks ()
{1, 2, 2, 1, 3, 1, 3, 2, 1, 1}

Problem Solving
6) For the above question, find the centroids after 3 iterations. 0 points
Session - July 2024 ()

Download () {5, 16.5, 56}

{3, 21.33, 42.4}

Live session ()
3.2, 30.67, 60

3.2, 29, 51.33

Yes, the answer is correct.
Score: 0
Accepted Answers:
3.2, 30.67, 60

7) How does K-Means initialize cluster centroids? 1 point

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=78&assessment=151 3/6
10/3/24, 11:39 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 11 - Week 8: Mixture Model and Clustering

At the maximum data point values

Based on the mean of the data points
At the origin (0,0)
Randomly
Yes, the answer is correct.
Score: 1
Accepted Answers:
Randomly
8) What is the best way to determine the optimal number of clusters (K) in K-Means? 1 point

Trial and error

Using the Elbow Method
Setting K equal to the number of data points
Using the Silhouette score
Yes, the answer is correct.
Score: 1
Accepted Answers:
Using the Elbow Method

9)Suppose we have a dataset with two observed features (X1 and X2) and missing values for some 1 point
data points.
The dataset is shown below:

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=78&assessment=151 4/6
10/3/24, 11:39 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 11 - Week 8: Mixture Model and Clustering

We want to estimate the missing values in X1 and X2 using a GMM with two components through the EM
algorithm.
Assume equal mixing coefficients, i.e.,
π1 = π2 = 0.5 and posterior probabilities as Gaussian probability density function.

After 1st iteration estimate the mean of X1?

3.07
3.87
3.57
3.67

No, the answer is incorrect.

Score: 0
Accepted Answers:
3.07

10) For the data in question 9, after 1st iteration estimate the variance of X2? 1 point

0.87
0.174
0.32

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=78&assessment=151 5/6
10/3/24, 11:39 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 11 - Week 8: Mixture Model and Clustering

No, the answer is incorrect.

Score: 0
Accepted Answers:
1

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=78&assessment=151 6/6
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 08: Detailed Solution
Indian Institute of Technology Guwahati

Q.1. Solution – Maximum Likelihood Estimation

Q.2. Solution – Mean and covariance matrix

Q.3. Solution – GMMs are sensitive to the initial placement of the Gaussian components

Q.4. Solution – A Gaussian distribution with its mean and covariance

Q.5. Solution –
Problem: Given 10 data points:
{−1,27,31, 2,59, 3,61,34, 0,12}
We are tasked with applying K-means clustering for 3 iterations to cluster them into 3
clusters. The initial cluster labels are:
{1,1,1,2,2,2,3,3,3,1}
We will proceed step by step, calculating the centroids and reassigning labels.
Step 1: Calculate Initial Centroids
Based on the initial labels, we calculate the centroids of the clusters as follows:
- Cluster 1 (𝐶1 ) contains points: {−1,27,31,12}
−1 + 27 + 31 + 12 69
Centroid of 𝐶1 = = = 17.25
4 4
- Cluster 2 (𝐶2 ) contains points: {2,59, 3}
2 + 59 + 3 64
Centroid of 𝐶2 = = ≈ 21.33
3 3
- Cluster 3 (𝐶3 ) contains points: {61,34, 0}
61 + 34 + 0 95
Centroid of 𝐶3 = = ≈ 31.67
3 3
Step 2: Reassign Labels Based on Closest Centroid
Now, we compute the distance of each point from the centroids and reassign it to the closest
cluster:

Points Distance Distance Distance Closest

from C1 from C2 from C3 center
(17.25) (21.33) (31.67)

-1 18.25 22.33 32.67 C1

27 9.75 5.67 4.63 C3

31 13.75 9.67 0.67 C3

2 15.25 19.33 29.67 C1

59 41.75 37.67 27.33 C3

3 14.25 18.33 28.67 C1

61 43.75 39.67 29.33 C3

34 16.75 12.67 2.33 C3

0 17.25 21.33 31.67 C1

12 5.25 9.33 19.67 C1

New Labels After Iteration 1:

{1,3,3,1,3,1,3,3,1,1}
Step 3: Recalculate Centroids
Now, based on the new labels, we recalculate the centroids.
Cluster 1 (𝐶1 ) contains points: {−1,2,3,0,12}
 −1 + 2 + 3 + 0 + 12 16
Centroid of𝐶1 = = = 3.2
5 5
Cluster 2 (𝐶2 ) has no points, so the center is the previous center
Centroid of C2 = 21.33

Cluster 3 (𝐶3 ) contains points: {27,31,59,61,34}

27 + 31 + 59 + 61 + 34 212
Centroid of𝐶3 = = = 42.4
5 5

Points Distance from Distance Distance from Closest Center

center C1 (3.2) from center center C3 (42.4)
C2 (21.33)

-1 4.2 22.33 43.4 C1

27 23.8 5.67 15.4 C2

31 27.8 9.67 11.4 C2

2 1.2 19.33 40.4 C1

59 55.8 37.67 16.6 C3

3 0.2 18.33 39.4 C1

61 57.8 39.67 18.6 C3

34 30.8 12.67 8.4 C3

0 3.2 21.33 42.4 C1

12 8.8 9.33 30.4 C1

New Labels After Iteration 2:

 {1,2,2,1,3,1,3,3,1,1}

Step 4: Recalculate Centroids

Now, based on the new labels, we recalculate the centroids.
Cluster 1 (𝐶1 ) contains points: {−1,2,3,0,12}
−1 + 2 + 3 + 0 + 12 16
Centroid of𝐶1 = = = 3.2
5 5
Cluster 2 (𝐶2 ) contains points: {27, 31}
Centroid of C2 = (27+31) / 2 = 29

Cluster 3 (𝐶3 ) contains points: {59,61,34}

59 + 61 + 34 154
Centroid of𝐶3 = = = 51.33
3 3

Points Distance from Distance Distance from Closest Center

center C1 (3.2) from center center C3
C2 (29) (51.33)

-1 4.2 30 52.33 C1

27 23.8 2 24.33 C2

31 27.8 2 20.33 C2

2 1.2 27 49.33 C1

59 55.8 30 7.67 C3

3 0.2 26 48.33 C1

61 57.8 32 9.67 C3
 34 30.8 5 17.33 C2

0 3.2 29 51.33 C1

12 8.8 17 39.33 C1

So final labels are

{1,2,2,1,3,1,3,2,1,1}

Q.6. Solution – Centroid C1 = (-1+2+3+0+12)/ 5 = 3.2

Centroid C2 = (27+31+34)/3 = 30.66
Centroid C3 = (59+61) / 2 = 60

Q.7. Solution – Randomly

Q.8. Solution – Using the Elbow Method

Q.9. Solution – 3.07

Q.10. Solution - 1
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 10/3/24, 11:40 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 12 - Week 9: Clustering

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Week 09 : Assignment 09

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-09-25, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-09-24, 11:12 IST
1) Which of the following statements is true about HAC with complete linkage? 1 point
Week 0: Prerequisite ()
It tends to produce elongated clusters.
Week 1: Introduction () It produces spherical clusters.
It tends to merge clusters with the minimum distance between centroids.
Week 2: Bayesian
It considers the maximum distance between points in different clusters.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=83&assessment=152 1/8
10/3/24, 11:40 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 12 - Week 9: Clustering

Decision Theory () Yes, the answer is correct.

Score: 1
Accepted Answers:
Week 3- Parametric
It considers the maximum distance between points in different clusters.
Approaches and Non-
Parametric Density 2) Lets consider we have cluster points P1(1,3) , P2(2,2) , P3(5,8) , P4(8,5) , P5(3,9) , P6(10,7) , 1 point
Estimation ()
P7(3,3) , P8(9,4) , P9(3,7). First, we take our K value as 3 and we assume that our Initial cluster centers
Week 4: Perceptron are P7(3,3), P9(3,7), P8(9,4) as C1, C2, C3. Then find out the new centroids after 1 iterations for the
Criteria and above data points.
Discriminative Models
()
(2, 2.7), (3, 6), (9, 5.3)

Week 5: Logistic (2, 2.7), (3.7, 8), (9, 5.3)

Regression, Decision
Trees and Hidden (2, 6), (3.7, 8), (4, 9.86)
Markov Model ()
(2, 9), (3, 2), (9, 7)
Week 6 - Ensemble Yes, the answer is correct.
Methods () Score: 1
Accepted Answers:
Week 7: (2, 2.7), (3.7, 8), (9, 5.3)
Dimensionality
Problem () 3) What is the primary idea behind Mean-Shift clustering? 1 point

Iteratively shifting data points towards the mean of the entire dataset.
Week 8: Mixture Model
and Clustering () Assigning data points to clusters based on predefined centroids.
Minimizing the sum of squared distances within each cluster.
Week 9: Clustering () Shifting data points towards the mode (highest density area) of the data distribution.

Lec 32: Fuzzy K-means Yes, the answer is correct.

clustering (unit? Score: 1
unit=83&lesson=84) Accepted Answers:
Shifting data points towards the mode (highest density area) of the data distribution.
Lec 33: Hierarchical
Agglomerative Clustering
4) What is the main objective of Hierarchical Agglomerative Clustering (HAC)? 1 point

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=83&assessment=152 2/8
10/3/24, 11:40 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 12 - Week 9: Clustering

and Mean-shift Clustering Minimize within-cluster variance

(unit?unit=83&lesson=85)
Maximize between-cluster variance
Lecture Notes (unit?
Minimize the number of clusters
unit=83&lesson=118)
Maximize the number of data points in each cluster
Quiz: Week 09 :
No, the answer is incorrect.
Assignment 09
Score: 0
(assessment?
Accepted Answers:
name=152)
Minimize within-cluster variance
Feedback form 9 (unit?
5) When m=1 in Fuzzy K-means, the algorithm behaves similarly to: 1 point
unit=83&lesson=87)

Assignment Solution Hierarchical Clustering.

(unit? Mean-Shift Clustering.
unit=83&lesson=162)
Standard K-means.
Week 10: Artificial DBSCAN.
Neural Network ()
Yes, the answer is correct.
Score: 1
Week 11: Introduction Accepted Answers:
to Deep Neural Standard K-means.
Networks ()
6) In Mean Shift clustering, what is the "kernel" used for? 1 point
Problem Solving
Session - July 2024 () It defines the number of clusters.
It specifies the dimensionality of the data.
Download () It represents the shape of the density function.
It determines the number of iterations.
Live session ()
Yes, the answer is correct.
Score: 1
Accepted Answers:
It represents the shape of the density function.

7) What is the key difference between Fuzzy K-Means and the traditional K-Means clustering algorithm? 1 point

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=83&assessment=152 3/8
10/3/24, 11:40 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 12 - Week 9: Clustering

Fuzzy K-Means allows each data point to belong to multiple clusters with varying degrees of membership.
Fuzzy K-Means does not require the number of clusters to be specified.
Fuzzy K-Means assigns each data point to exactly one cluster.
Fuzzy K-Means is a supervised learning algorithm, while K-Means is unsupervised.

Yes, the answer is correct.

Score: 1
Accepted Answers:
Fuzzy K-Means allows each data point to belong to multiple clusters with varying degrees of membership.

8) If we utilize the Hierarchical Agglomerative Clustering method on given below data with the aim of 1 point
identifying clusters based on the smallest values obtained from Euclidean distance calculations, the dendrogram

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=83&assessment=152 4/8
10/3/24, 11:40 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 12 - Week 9: Clustering

will be -

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=83&assessment=152 5/8
10/3/24, 11:40 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 12 - Week 9: Clustering

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=83&assessment=152 6/8
10/3/24, 11:40 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 12 - Week 9: Clustering

Yes, the answer is correct.

Score: 1
Accepted Answers:

9) How does hierarchical agglomerative clustering work? 1 point

It randomly assigns data points to different clusters, and then iteratively adjusts the assignments until a stopping criterion is met.
It applies a clustering algorithm to different subsets of the data, and then combines the results using an ensemble method.
It starts with each data point as its own cluster, and then iteratively merges the closest pairs of clusters until a stopping criterion is met.
It uses a decision tree to recursively split the data into smaller clusters until a stopping criterion is met.

Yes, the answer is correct.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=83&assessment=152 7/8
10/3/24, 11:40 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 12 - Week 9: Clustering

Score: 1
Accepted Answers:
It starts with each data point as its own cluster, and then iteratively merges the closest pairs of clusters until a stopping criterion is met.

10) Which parameter in Fuzzy K-Means controls the fuzziness of the clustering? 1 point

Number of clusters (K).

Membership exponent (m).
Number of iterations.
Initial cluster centroids.
Yes, the answer is correct.
Score: 1
Accepted Answers:
Membership exponent (m).

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=83&assessment=152 8/8
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati

Q.1. Solution – It considers the maximum distance between points in different clusters.

Q.2. Solution – Find the distance between data points and Centroids. which data points have a
minimum distance that points moved to the nearest cluster centroid.

Cluster 1 => P1(1,3) , P2(2,2) , P7(3,3)

Cluster 2 => P3(5,8) , P5(3,9) , P9(3,7)
Cluster 3 => P4(8,5) , P6(10,7) , P8(9,4)
New center of Cluster 1 => (1+2+3)/3 , (3+2+3)/3 => 2,2.7
New center of Cluster 2 => (5+3+3)/3 , (8+9+7)/3 => 3.7,8
New center of Cluster 3 => (8+10+9)/3 , (5+7+4)/3 => 9,5.3
So the new cluster centers are (2,2.7), (3.7,8), (9,5.3)
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati

Q.3. Solution – Shifting data points towards the mode (highest density area) of the data
distribution.

Q.4. Solution – Maximize between-cluster variance

Q.5. Solution – Standard K-means.

Q.6. Solution – It represents the shape of the density function.

Q.7. Solution – Fuzzy K-Means allows each data point to belong to multiple clusters with varying
degrees of membership.

Q.8. Solution – Look at the distance matrix to find the smallest non-zero value, as this indicates
the closest pair of points that should be merged first.

From the provided matrix:

• The smallest distance is 0.1100 between p3p3p3 and p6p6p6.

First Merge

Merge the clusters containing p3 and p6:

• New Cluster 7: {p3,p6}

Now, update the distance matrix to reflect this new cluster. The distance between this new cluster
and any other point will be the minimum distance from any member of the new cluster to that
point (assuming we are using single linkage).
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati

Calculate the new distances:

d(C7,p1) = min(d(p3,p1),d(p6,p1))=min(0.2218,0.2347)=0.

d(C7,p2) = min(d(p3,p2),d(p6,p2))=min(0.1483,0.2540)=0.1483

d(C7,p4)= min(d(p3,p4),d(p6,p4))=min(0.1513,0.2216)=0.1513

d(C7,p5) = min(d(p3,p5),d(p6,p5))=min(0.2843,0.3921)=0.2843

The updated matrix will have a new row and column for C7C7C7 and will no longer include rows
and columns for p3 and p6.

Now, look for the smallest distance in the updated distance matrix:

The smallest distance is 0.1483 between C7 (the cluster containing p3 and p6) and p2.

Repeat the process of updating the distance matrix and finding the smallest distance to merge
clusters until all points are merged into a single cluster.
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati

Q.9. Solution – It starts with each data point as its own cluster, and then iteratively merges the closest
pairs of clusters until a stopping criterion is met.

Q.10. Solution – Membership exponent (m).

 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 09: Detailed Solution
Indian Institute of Technology Guwahati
 10/3/24, 11:41 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 13 - Week 10: Artificial Neural Network

gireesh218@gmail.com 

(https://swayam.gov.in)
(https://swayam.gov.in/nc_details/NPTEL)

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning - Fundamentals and Applications (course)

Announcements (announcements) About the Course (preview) Q&A (forum) Progress (student/home) Mentor (student/mentor)

Review Assignment (assignment_review) Course Recommendations (/course_recommendations)

Course outline Assignment 10

The due date for submitting this assignment has passed.
About NPTEL () Due on 2024-10-02, 23:59 IST.

How does an NPTEL

online course work? ()
Assignment submitted on 2024-09-30, 07:21 IST
Week 0: Prerequisite ()

Week 1: Introduction ()

Week 2: Bayesian

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=98&assessment=157 1/7
10/3/24, 11:41 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 13 - Week 10: Artificial Neural Network

Decision Theory () 1) 1 point

Week 3- Parametric
Approaches and Non-
Parametric Density
Suppose you have input as x,y and z to the network given below-
Estimation ()

Week 4: Perceptron
Criteria and Calculate the gradients of the output with respect to x, y and z.
Discriminative Models
()
( −3
4
, −1
2
, 1)
Week 5: Logistic
Regression, Decision ( −3
4
, 12 , 1)
Trees and Hidden
−1
Markov Model () ( 34 , 2
, 1)

Week 6 - Ensemble −1
( 34 , 2
, −1)
Methods ()
Yes, the answer is correct.
Score: 1
Week 7:
Accepted Answers:
Dimensionality
Problem () ( −3
4
, 12 , 1)

Week 8: Mixture Model 2) Which of the following is not a type of neural network architecture? 1 point
and Clustering ()
Convolutional Neural Network
Week 9: Clustering () Long Short-Term Memory
Self-Organization Map
Week 10: Artificial Decision Tree Network
Neural Network ()
Yes, the answer is correct.
Score: 1
Lec 34: Artificial Neural
Networks for Pattern Accepted Answers:
Decision Tree Network

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=98&assessment=157 2/7
10/3/24, 11:41 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 13 - Week 10: Artificial Neural Network

Classification (PART 1)
(unit?unit=98&lesson=99)
3) What is overfitting in the context of neural networks? 1 point

Lec 35: Artificial Neural When the variance of the model is high
Networks for Pattern When the model has too few layers
Classification (PART 2)
(unit?
When the activation function is not properly defined
unit=98&lesson=100) When the learning rate is too high

Lec 36: Artificial Neural Yes, the answer is correct.

Networks for Pattern
Score: 1
Classification (PART 3) Accepted Answers:
(unit? When the variance of the model is high
unit=98&lesson=101)
4) What is the purpose of backpropagation in training a neural network? 1 point
Lecture Notes (unit?
unit=98&lesson=119)
Forward pass calculation
Quiz: Assignment 10 Calculating the cost function
(assessment?
Updating weights and biases based on prediction errors
name=157)
Regularization of the network
Feedback form 10 (unit?
unit=98&lesson=103)
Yes, the answer is correct.
Score: 1
Accepted Answers:
Week 11: Introduction
Updating weights and biases based on prediction errors
to Deep Neural
Networks ()
5) You are using the sigmoid activation function in a neural network. If the network's weighted input is 2.5, what is the output of the 1 point
Problem Solving sigmoid activation function?
Session - July 2024 ()
0.75

Download () 0.07
0.92
Live session () None

Yes, the answer is correct.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=98&assessment=157 3/7
10/3/24, 11:41 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 13 - Week 10: Artificial Neural Network

Score: 1
Accepted Answers:
0.92

6) In a feedforward neural network with one hidden layer, the input layer has 10 neurons, the hidden layer has 5 neurons, and the output 1 point
layer has 3 neurons. How many total weights (including biases) are there in the network?

70
73
50
65

Yes, the answer is correct.

Score: 1
Accepted Answers:
73

7) Consider a standard three-layer backpropagation net with d input units, h hidden units, c output units, and bias, find 1 point
the expression for output of a node k is

∑hj=1 f(wki f(∑di=1 wji xi + wj0 ) + wko )

∑hj=1 f(wkj f(∑di=1 wji xi + wj0 ) + wko )

∑hj=1 f(wkj f(∑di=1 wij xi + wj0 ) + wko )

∑hj=1 f(wki f(∑di=1 wji xi + wj0 ))

Yes, the answer is correct.
Score: 1
Accepted Answers:
∑hj=1 f(wkj f(∑di=1 wji xi + wj0 ) + wko )

8) You are training a neural network using a fixed learning rate of 0.01. After several epochs, you notice that the loss is 1 point
decreasing too slowly, indicating that the learning rate is too small. If you want to change the learning rate to speed up convergence, what

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=98&assessment=157 4/7
10/3/24, 11:41 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 13 - Week 10: Artificial Neural Network

value should you try next?

0.01
0.1
0.001
None

Yes, the answer is correct.

Score: 1
Accepted Answers:
0.1

9) For a Kohonen Self-organizing map (SOM) to cluster four vectors. The online vectors to be clustered are 1 point
(1, 1, 0, 0); (0, 0, 0, 1); (1, 0, 0, 0); (0, 0, 1, 1) sequentially. The maximum number of clusters to be formed is m=2. Suppose learning rate
⎡ 0.3 0.7 ⎤
(Geometric decrease) is α(0) = 0.6 and α(t + 1) = 0.6α(t). Initial weight matrix is ⎢
⎢ 0.4 0.3 ⎥. The first iteration for KSOM to cluster
⎥
⎢ 0.6 0.6 ⎥
⎣ 0.8 0.3 ⎦
the vector (1, 0, 0, 0) will be

⎡ 0.83 0.88 ⎤
⎢ 0.42 0.72 ⎥
⎢ ⎥
⎢ 0.21 0.24 ⎥
⎣ 0.12 0.12 ⎦

⎡ 0.12 0.952 ⎤
⎢ 0.16 0.29 ⎥
⎢ ⎥
⎢ 0.24 0.096 ⎥
⎣ 0.92 0.048 ⎦

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=98&assessment=157 5/7
10/3/24, 11:41 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 13 - Week 10: Artificial Neural Network

⎡ 0.12 0.88 ⎤
⎢ 0.16 0.72 ⎥
⎢ ⎥
⎢ 0.24 0.24 ⎥
⎣ 0.92 0.12 ⎦

⎡ 0.63 0.35 ⎤
⎢ 0.18 0.21 ⎥
⎢ ⎥
⎢ 0.25 0.15 ⎥
⎣ 0.61 0.6 ⎦
No, the answer is incorrect.
Score: 0
Accepted Answers:
⎡ 0.12 0.952 ⎤
⎢ 0.16 0.29 ⎥
⎢ ⎥
⎢ 0.24 0.096 ⎥
⎣ 0.92 0.048 ⎦

10) Suppose Mean Squared Error (MSE) is used as the loss function for training your neural network model. Your model predicts a 1 point
value of 8 for a data point with an actual target value of 10. What is the MSE loss for this data point?

4
2
1
0

Yes, the answer is correct.

Score: 1
Accepted Answers:
4

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=98&assessment=157 6/7
10/3/24, 11:41 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 13 - Week 10: Artificial Neural Network

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=98&assessment=157 7/7
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications

ti
Assignment 10: Detailed Solution

ha
Indian Institute of Technology Guwahati

wa
Gu
5. Given, weighted input (x) = 2.5

gy
Therefore, considering sigmoid function,

lo
1 1
Output (y) = −𝑥 = −2.5 = 0. 92
1+𝑒 1+𝑒
no
6. Given: Input Layer = 10 Neurons; Hidden Layer = 5 Neurons; Output Layer = 3 Neurons
ch
Weights between Input Layer and Hidden Layer:
Number of weights=Number of input neurons×Number of hidden neurons=10×5=50
Te

Number of biases (hidden layer)=5

of

Therefore, total weights between input and hidden layer = 50 + 5 = 55

Weights between Hidden Layer and Output Layer:
te

Number of weights=Number of hidden neurons×Number of output

itu

neurons=5×3=15
st

Number of biases (hidden layer)=3

In

Therefore, total weights between hidden and out layer = 15 + 3 = 18

Therefore,
an

Total number of weights = 55 + 18 = 73

di

9. Given: 𝑤11 = 0. 3, 𝑤12 = 0. 4, 𝑤13 = 0. 6, 𝑤14 = 0. 8

In

𝑤21 = 0. 7, 𝑤22 = 0. 3, 𝑤23 = 0. 6, 𝑤24 = 0. 3

For first vector, X = [1,1,0,0]

𝑛
2
Euclidean distance, 𝐷(𝑗) = ∑ (𝑤𝑖𝑗 − 𝑥𝑖)
𝑖=1

Now,
 2 2 2 2
𝐷(1) = (1 − 0. 3) + (1 − 0. 4) + (0 − 0. 6) + (0 − 0. 8) = 1.85
2 2 2 2
𝐷(2) = (1 − 0. 7) + (1 − 0. 3) + (0 − 0. 6) + (0 − 0. 3) = 1.03

Here, 𝐷(1) > 𝐷(2), therefore we will update 𝑤𝑖2 as follows

𝑤𝑖2 (new) = 𝑤𝑖2 (old) + 0.6(𝑥𝑖 - 𝑤𝑖2 (old))

ti
Therefore,

ha
𝑤12 (new) = 0.7 + 0.6(1 - 0.7) = 0.88

wa
𝑤22 (new) = 0.3 + 0.6(1 - 0.3) = 0.72

Gu
𝑤32 (new) = 0.6 + 0.6(0 - 0.6) = 0.24

𝑤42 (new) = 0.3 + 0.6(0 - 0.3) = 0.12

gy
lo
For second vector, X = [0,0,0,1]
no 𝑛
Euclidean distance, 𝐷(𝑗) = ∑ (𝑤𝑖𝑗 − 𝑥𝑖)
2
ch
𝑖=1

Now,
Te

2 2 2 2
𝐷(1) = (0 − 0. 3) + (0 − 0. 4) + (0 − 0. 6) + (1 − 0. 8) = 0.65
of

2 2 2 2
𝐷(2) = (0 − 0. 88) + (0 − 0. 72) + (0 − 0. 24) + (1 − 0. 12) = 2.12
te

Here, 𝐷(1) < 𝐷(2), therefore we will update 𝑤𝑖1 as follows

itu

𝑤𝑖1 (new) = 𝑤𝑖1 (old) + 0.6(𝑥𝑖 - 𝑤𝑖1 (old))

st

Therefore,
In

𝑤11 (new) = 0.3 + 0.6(0 - 0.3) = 0.12

an

𝑤21 (new) = 0.4 + 0.6(0 - 0.4) = 0.16

di

𝑤31 (new) = 0.6 + 0.6(0 - 0.6) = 0.24

In

𝑤41 (new) = 0.8 + 0.6(1 - 0.8) = 0.92

For third vector, X = [1,0,0,0]

𝑛
2
Euclidean distance, 𝐷(𝑗) = ∑ (𝑤𝑖𝑗 − 𝑥𝑖)
𝑖=1
 Now,
2 2 2 2
𝐷(1) = (1 − 0. 12) + (0 − 0. 16) + (0 − 0. 24) + (0 − 0. 92) = 1.7
2 2 2 2
𝐷(2) = (1 − 0. 88) + (0 − 0. 72) + (0 − 0. 24) + (0 − 0. 12) = 0.6

Here, 𝐷(1) > 𝐷(2), therefore we will update 𝑤𝑖2 as follows

ti
𝑤𝑖2 (new) = 𝑤𝑖2 (old) + 0.6(𝑥𝑖 - 𝑤𝑖2 (old))

ha
Therefore,

wa
𝑤12 (new) = 0.88 + 0.6(1 - 0.88) = 0.952

𝑤22 (new) = 0.72 + 0.6(0 - 0.72) = 0.29

Gu
𝑤32 (new) = 0.24 + 0.6(0 - 0.24) = 0.096

gy
𝑤42 (new) = 0.12 + 0.6(0 - 0.12) = 0.048

lo
10. We know, MSE =
1
𝑛
𝑛
∑ (𝑦𝑖 − 𝑦𝑖)
2
no
ch
𝑖=1

Here, y = 10 and 𝑦 = 8
Te

1 2 2
Therefore, MSE = (10 − 8) = 2 = 4
of

1
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 10/16/24, 9:54 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 14 - Week 11: Introduction to Deep Neural Net…

(https://swayam.gov.in) (https://swayam.gov.in/nc_details/NPTEL)

gireesh218@gmail.com 

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning -

Fundamentals and Applications (course)

Course Week 11 : Assignment 11

outline The due date for submitting this assignment has passed.
Due on 2024-10-09, 23:59 IST.
About
NPTEL ()
Assignment submitted on 2024-10-05, 15:21 IST
How does an 1) What is the receptive field? 1 point
NPTEL
online It is the name given to the connectivity of neurons only to a local region of
course the output volume.
work? () It is the name given to the relationship of neurons.
It is a hyperparameter that helps neurons when they have difficulty in
Week 0:
finding connections.
Prerequisite
() It is the name given to the connectivity of neurons only to a local region of
the input volume.
Week 1: Yes, the answer is correct.
Introduction Score: 1
() Accepted Answers:
It is the name given to the connectivity of neurons only to a local region of the
Week 2: input volume.
Bayesian
Decision 2) Why is it preferred to keep Kernel Size small? 1 point
Theory ()
The number of features decreases quadratically with kernel size.
Week 3- The number of parameters grows quadratically with kernel size.
Parametric The number of parameters decreases quadratically with kernel size.
Approaches
The number of features grows quadratically with kernel size.
and Non-
Parametric Yes, the answer is correct.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=105&assessment=161 1/4
10/16/24, 9:54 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 14 - Week 11: Introduction to Deep Neural Net…

Density Score: 1
Estimation () Accepted Answers:
The number of parameters grows quadratically with kernel size.
Week 4:
3) You have an input volume that is 63 × 63 × 16 , and convolve it with 1 point
Perceptron
32 filters that are each 7 × 7, using a stride of 2 and no padding. What is the
Criteria and
output volume?
Discriminativ
e Models ()
16 × 16 × 32
Week 5:
16 × 16 × 16
Logistic
Regression, 29 × 29 × 32
Decision
Trees and 29 × 29 × 16
Hidden Yes, the answer is correct.
Score: 1
Markov
Accepted Answers:
Model ()
29 × 29 × 32
Week 6 - 4)Which activation function is less prone to the vanishing gradient 1 point
Ensemble
problem?
Methods ()
Sigmoid
Week 7: Tanh
Dimensionali
ReLU
ty Problem ()
Leaky ReLU
Week 8: Yes, the answer is correct.
Mixture Score: 1
Model and Accepted Answers:
Clustering () Leaky ReLU

Week 9: 5)What is the typical configuration of VGG networks regarding the 1 point
Clustering () convolutional kernel size and stride?

Week 10: 5x5 kernels with stride 2

Artificial 3x3 kernels with stride 1
Neural 7x7 kernels with stride 3
Network ()
1x1 kernels with stride 1
Week 11: Yes, the answer is correct.
Score: 1
Introduction
to Deep Accepted Answers:
Neural 3x3 kernels with stride 1
Networks ()
6) What is the primary function of the Inception module in 1 point
Lec 37: GoogleNet?
Introduction to
Deep Learning It introduces skip connections between layers.
and It combines multiple neural networks into a single architecture.
Convolutional
Neural
It incorporates parallel convolutional operations of different kernel
Network sizes.
(CNN) (unit? It reduces the spatial resolution of feature maps.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=105&assessment=161 2/4
10/16/24, 9:54 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 14 - Week 11: Introduction to Deep Neural Net…

unit=105&less Yes, the answer is correct.

on=107) Score: 1
Accepted Answers:
Lec 38: It incorporates parallel convolutional operations of different kernel sizes.
Vanishing and
Exploding 7) To mitigate the exploding gradient problem, which technique can be 1 point
Gradients in applied to gradient updates during training?
Deep Neural
Networks Standardization
(unit? Gradient Clipping
unit=105&less
on=108)
Decreasing the learning rate
None
Lec 39: CNN
Architectures - Yes, the answer is correct.
LeNet-5 and
Score: 1
AlexNet (unit? Accepted Answers:
unit=105&less Gradient Clipping
on=109)
8)In a CNN architecture, if a max-pooling layer with a 2x2 window 1 point
Lec 40: CNN
Architectures –
and a stride of 2 is applied to the output of a previous convolutional layer,
VGG 16, how does the spatial dimensions of the feature maps change?
GoogLeNet
and ResNet
The spatial dimensions are reduced by half (e.g., from 64x64 to 32x32).
(unit? The spatial dimensions remain the same.
unit=105&less The spatial dimensions are doubled (e.g., from 64x64 to 128x128).
on=110)
The spatial dimensions are quadrupled (e.g., from 64x64 to 256x256).
Lecture Notes
Yes, the answer is correct.
(unit? Score: 1
unit=105&less
Accepted Answers:
on=120)
The spatial dimensions are reduced by half (e.g., from 64x64 to 32x32).
Quiz: Week
11 : 9)How many convolutional layers are there in the AlexNet 1 point
Assignment architecture?
11
(assessment? 2
name=161)
5
Feedback form 8
11 (unit?
11
unit=105&less
on=111) Yes, the answer is correct.
Score: 1
Week 12: Accepted Answers:
Recent 5
Trends in
Deep 10) Which network has the maximum number of training parameters? 1 point
Learning ()
GoogleNet
Problem VGG16
Solving AlexNet
Session -
July 2024 ()
LeNet
Yes, the answer is correct.

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=105&assessment=161 3/4
10/16/24, 9:54 AM Machine Learning and Deep Learning - Fundamentals and Applications - - Unit 14 - Week 11: Introduction to Deep Neural Net…

Score: 1
Download () Accepted Answers:
VGG16
Live session
()

https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=105&assessment=161 4/4
 NPTEL Online Certification Course

ti
Machine Learning and Deep Learning - Fundamentals and Applications

ha
Assignment 11: Detailed Solution
Indian Institute of Technology Guwahati

wa
Gu
gy
3. For a given input size n, kernel of size f, padding of size p and stride as s,

lo
𝑛 − 𝑓 + 2𝑝 𝑛 − 𝑓 + 2𝑝
The output 𝑜 = ( 𝑠
+ 1) x ( 𝑠
+ 1) x number of filters
no
Here, n = 63, f = 7, s = 2, p = 0, number of filters = 32
ch
63 − 7 63 − 7
Therefore, o = ( 2
+ 1) x ( 2
+ 1) x 32 = 29 x 29 x32
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
 NPTEL Online Certification Course
<COURSE NAME>
<Assignment Number>: Detailed Solution
Indian Institute of Technology Guwahati

ti
ha
wa
Gu
gy
lo
no
ch
Te
of
te
itu
st
In
an
di
In
Machine Learning and Deep Learning - Fundamentals and Applications ... https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=112&assessme...

(https://swayam.gov.in) (https://swayam.gov.in/nc_details/NPTEL)

gireesh218@gmail.com 

NPTEL (https://swayam.gov.in/explorer?ncCode=NPTEL) » Machine Learning and Deep Learning -

Fundamentals and Applications (course)

Course Week 12 : Assignment 12

outline The due date for submitting this assignment has passed.
Due on 2024-10-16, 23:59 IST.
About NPTEL
()
Assignment submitted on 2024-10-05, 15:26 IST
How does an 1) How does PCA reduce dimensionality? 1 point
NPTEL online
course work? By removing features that contribute the least variance
() By applying random noise to the data
By performing non-linear transformations
Week 0: By reconstructing input using encoded latent space
Prerequisite
() Yes, the answer is correct.
Score: 1
Accepted Answers:
Week 1:
By removing features that contribute the least variance
Introduction
() 2) What is the main objective of the Discriminator in a GAN? 1 point

To generate new data

Week 2:
Bayesian To encode the input into a latent space
Decision To classify real from generated data
Theory () To reconstruct the input data

Yes, the answer is correct.

Week 3- Score: 1
Parametric Accepted Answers:
Approaches To classify real from generated data
and Non-

1 of 4 17-10-2024, 10:08 am
Machine Learning and Deep Learning - Fundamentals and Applications ... https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=112&assessme...

Parametric 3) In RNNs, what is the primary reason for vanishing gradients? 1 point
Density
Estimation () Use of ReLU activation functions
Use of shallow network architectures
Week 4: Repeated multiplication of small values over long sequences
Perceptron Limited training data
Criteria and
Yes, the answer is correct.
Discriminativ
Score: 1
e Models ()
Accepted Answers:
Repeated multiplication of small values over long sequences
Week 5:
Logistic 4) In an autoencoder, what is the bottleneck layer responsible for? 1 point
Regression,
Decision
Reducing overfitting
Trees and Increasing model complexity
Hidden Controlling the learning rate
Markov Model Extracting and encoding useful features
()
Yes, the answer is correct.
Score: 1
Week 6 -
Accepted Answers:
Ensemble
Extracting and encoding useful features
Methods ()
5) What are skip connections in U-Net used for? 1 point
Week 7:
Creating shortcuts for gradient flow and preserving fine-grained details
Dimensionalit
y Problem () Skipping over layers with no useful information
Reducing computational complexity
Week 8: Avoiding overfitting
Mixture
Yes, the answer is correct.
Model and Score: 1
Clustering () Accepted Answers:
Creating shortcuts for gradient flow and preserving fine-grained details
Week 9:
Clustering () 6) Recurrent Neural Networks (RNNs) are primarily used for: 1 point

Image segmentation
Week 10:
Time-series and sequence data processing
Artificial
Dimensionality reduction
Neural
Network () Data generation

Yes, the answer is correct.

Week 11: Score: 1
Introduction Accepted Answers:
to Deep Time-series and sequence data processing
Neural
Networks ()
7) What is the primary limitation of GANs in real-world applications? 1 point

They can only generate grid-like data, such as images.

Week 12:
They can only generate low-resolution images
Recent
They are not effective for image classification tasks
Trends in

2 of 4 17-10-2024, 10:08 am
Machine Learning and Deep Learning - Fundamentals and Applications ... https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=112&assessme...

Deep They are computationally expensive to train

Learning ()
Yes, the answer is correct.
Score: 1
Lec
41:Generative Accepted Answers:
Adversarial They are computationally expensive to train
Networks
8) Which part of the U-Net architecture is responsible for capturing context 1 point
(GAN) -
Fundamentals
informa�on?
and Contrac�ng path
Applications
(unit?
Bo�leneck
unit=112&lesso Expansive path
n=121) Skip connec�ons
Lec 42: U-Net: Yes, the answer is correct.
Convolutional Score: 1
Networks for Accepted Answers:
Image Contrac�ng path
Segmentation
9) Which of the following is a key challenge in training GANs? 1 point
(unit?
unit=112&lesso
Vanishing gradients
n=122)
Mode collapse
Lec 43: Overfitting
Introduction to
Unsupervised learning
Autoencoder
and Recurrent Yes, the answer is correct.
Neural Score: 1
Networks Accepted Answers:
(RNN) (unit? Mode collapse
unit=112&lesso
10) Advantagesof using autoencoder to reduce the dataset’s 1 point
n=123)
dimensionality are-
Lec 44:
Programming Many instances
Concepts (unit? Many features
unit=112&lesso
They can handle large datasets
n=124)
All of these
Lec 45:
Yes, the answer is correct.
Programming
Score: 1
Concepts (unit?
Accepted Answers:
unit=112&lesso
n=125)
All of these

Lec 46:
Problem
Solving
Session (unit?
unit=112&lesso
n=126)

Lec 47:
Problem

3 of 4 17-10-2024, 10:08 am
Machine Learning and Deep Learning - Fundamentals and Applications ... https://onlinecourses.nptel.ac.in/noc24_ee146/unit?unit=112&assessme...

Solving
Session (unit?
unit=112&lesso
n=127)

Lec 48:
Problem
Solving
Session (unit?
unit=112&lesso
n=128)

Lecture Notes
(unit?
unit=112&lesso
n=130)

Quiz: Week 12
: Assignment
12
(assessment?
name=164)

Feedback form
12 (unit?
unit=112&lesso
n=129)

Problem
Solving
Session -
July 2024 ()

Download ()

Live session
()

4 of 4 17-10-2024, 10:08 am
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati

Q.1. Solution – By removing features that contribute the least variance

Q.2. Solution – To classify real from generated data

Q.3. Solution – Repeated multiplication of small values over long sequences

Q.4. Solution – Extracting and encoding useful features

Q.5. Solution – Creating shortcuts for gradient flow and preserving fine-grained details

Q.6. Solution – Time-series and sequence data processing

Q.7. Solution – They are computationally expensive to train

Q.8. Solution – Contracting path

Q.9. Solution – Mode collapse

Q.10. Solution - All of these

 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati
 NPTEL Online Certification Course
Machine Learning and Deep Learning - Fundamentals and Applications
Assignment Number – 12: Detailed Solution
Indian Institute of Technology Guwahati