Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur

Department of Computer Science & Engineering

Assignment Sheet-II
Session 2023_24 (Even Semester)

Subject: Machine Learning (-02)

Max Marks- 40

(Set-I)

Q.No Question BL CO MM
1 3 3 10
What is the curse of Dimensionality in machine learning? How PCA is
used for dimensionality reduction?
2 What is the difference between Forward propagation and Backward Propagation in Neural 6 4 10
Networks?
3 Explain any four activation functions used in a neural network. 3 5 10

4 3 3 10
Compute the principal component of following data-
CLASS 1: X=2,3,4 Y=1,5,3
CLASS 2: X = 5 , 6 , 7. Y=6,7,8
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Assignment Sheet-II
Session 2023_24 (Even Semester)

Subject: Machine Learning (6CAI4-02)

Max Marks- 40

(Set-II)

Q.No Question BL CO MM
1 4 3 10
Compare Feature Extraction and Feature Selection techniques. Explain how
dimensionality can be reduced using a subset selection procedure
2 What is HMM? Explain HMM with suitable diagram 3 4 10

3 Consider the following Neural Network diagram and find the updated 4 5 10
parameters of the network using oneiteration. Sigmoid activation function is
used at hidden and output layer. Consider only the parameters given in the
network. (use η = 0.1 and target output = 0.7).

4 find the optimal solution using the Bellman equation. Also update the table and find value 4 4 10
at each state of the given environment. Here Flag represent the Reward(+1) and Fire(-1).
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Assignment Sheet-II
Session 2023_24 (Even Semester)

Subject: Machine Learning (6CAI4-02)

Max Marks- 40

(Set-III)

Q.No Question BL CO MM
1 (a)What is training, testing and cross validation of machine learning models 3 3 10
(b) What is overfitting and underfitting? What are the catalysts of overfitting?
c) Elaborate Bias Variance dilemma.
2 4 4 10
What s Temporal Learning Reinforcement? When using Temporal Difference
learning, why is it better to learn action values (Qvalues) rather than state
values (V -values)?
3 (a) Explain the differences between Collaborative filtering and Content based filtering 4 5 10
(b)What is TF and IDF? How document is represented with TF-IDF explain with an
example.
4 4 5 10
Refer the neural network given in the figure. Determine the updated
weight w5 using backpropagation algorithm. Consider the learning rate
α = 0.5. (Refer given figure ).
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Assignment Sheet-II
Session 2023_24 (Even Semester)

Subject: Machine Learning (6CAI4-02)

Max Marks- 40

(Set-IV)

Q.No Question BL CO MM
1 Compare Feature Extraction and Feature Selection techniques. Explain how 3 3 10
dimensionality can be reduced using a subset selection procedure
2 How exactly a Markov Decision Process work in reinforcement learning? 3 4 10

3 How deep learning overcomes the challenges in conventional machine learning techniques? 3 5 10
Draw and explain the architecture of different Neural networks.

4 Apply Backpropagation algorithm for following scenario: 3 3 10

 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Assignment Sheet-I
Session 2023_24 (Even Semester)

Subject: Machine Learning (6CAI4-02)

Max Marks- 40

(Set-V)

Q.No Question BL CO MM
1 Explain the following terms: 3 3 10
(i) Precision
(ii) Recall
(iii) F-score
(iv) Sensitivity
(v) Specificity
2 3 4 10

3 b) What is the significance of bias in perceptron networks? Obtain the output of neuron Y 4 5 10
for the given network if binary sigmoid activation function is used. (Refer figure 1).

4 Q1) List various environments for an agent. Compare between following environments. [7] 4 4 10
a) Fully observable versus partially observable b) Deterministic versus stochastic
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Assignment Sheet-II
Session 2023_24 (Even Semester)

Subject: Machine Learning (6CAI4-02)

Max Marks- 40

(Set-VI)

Q.No Question BL CO MM
1 What is Feature Engineering? How do you apply Feature Transformation and feature 3 3 10
selection in the machine Learning
2 Suppose we have 5 rooms in a building connected by doors as shown in the fig below. Note 4 4 10
that door 1 and 4 leads into the building from room 5(outside).Goal state is Room 5(outside).
Find the optimal path for an agent using Q learning Algorithm. the reward from door 1to 5
and 4 to 5 is hundred(100), expect that all reward are equivalent to zero(0)

3 Explain the architecture of feed forward neural network. Give its limitations. [8] 3 5 10
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering
4 Fill in the blanks/Answer the following. 4 5 10
(i) Equation for Sigmoidal basis function is ________
(ii) Following figure implement which logic

(iii) Following figure implement which logic?

(iv) Logical Expression for z1 is __________

 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Assignment Sheet-II
Session 2023_24 (Even Semester)

Subject: MACHINE LEARNING (6CS4-02)

Max Marks- 40

(Set-VII)
Q. No. Question BL CO
1. Illustrate Principal component analysis. Demonstrate how PCA is applied in 3 3
ML.
2. Apply bellman equation and calculate value functions for all states of the 3 4
environment given below.

Rewards:
1: (for goal state),
0: (for non-goal state)
-1: (for DANGER state)

3. Apply Backpropagation algorithm for following scenario: 3 5

4. Illustrate Markov Decision Process (MDP). Demonstrate how MDP is 3 4

applied in Reinforcement Learning.
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Assignment Sheet-II
Session 2023_24 (Even Semester)

Subject: MACHINE LEARNING (6CS4-02)

Max Marks- 40

(Set-VIII)
Q. No. Question BL CO
1. Find the Singular Value Decomposition (SVD) of A, UΣV T , 3 3
where A = ([3, 2, 2], [2, 3, −2], [5, 2, 9])
2. Apply Backpropagation algorithm for following scenario: 3 5

3. Differentiate between Policy evaluation and Value evaluation in 3 4

Reinforcement Learning. Illustrate how these are applied.
4. How Machine Learning algorithms are evaluated ? Demonstrate the evaluation 3 3
measures required for regression problems.
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Assignment Sheet-II
Session 2023_24
(Even Semester)

Subject: MACHINE LEARNING (6CS4-02)

Max Marks- 40

(Set-IX)
Q. No. Question BL CO
1. How Machine Learning algorithms are evaluated? Demonstrate the evaluation 3 3
measures required for classification problems.
2. Apply Backpropagation algorithm for following scenario: 3 5

3. Find the Singular Value Decomposition (SVD) of X, UΣV T , 3 4

where X = ([6, -8, 2, 2], [-1, 5, 9, 0], [4, 5, -2, 9])
4. Illustrate Q-Learning in Reinforcement Learning? Demonstrate how Q- 3 4
function is evaluated?
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering
Assignment
Sheet-II
Session 2023_24
(Even Semester)

Subject: MACHINE LEARNING (6CS4-02)

Max Marks- 40

(Set-X)
Q. No. Question BL CO
1. In a binary classification task for detecting spam emails, a model 3 3
correctly predicted 450 instances as spam (True Positive), correctly
predicted 850 instances as non-spam (True Negative), mispredicted 30
instances as spam (False Positive), and mispredicted 20 instances as
non-spam (False Negative). Evaluate the model's performance and
discuss its effectiveness in distinguishing between spam and non-spam
emails.
2. Consider a 3x3 grid maze where the agent starts at position (0, 0) and 3 4
the goal is at position (2, 2). The agent can move in four directions: up,
down, left, and right. The numbers in the maze represent the rewards:
100 for reaching the goal, -1 for each step taken, and -100 for hitting an
obstacle. The agent aims to learn the optimal policy to navigate from the
start to the goal while maximizing its total reward.

Determine the optimal action based on its current state using Q-

Learning.
3. Consider a simple neural network with one input layer, hidden layer, and 3 5
output layer. The input layer has two neurons, the hidden layer has two
neurons, and the output layer has one neuron. We'll denote the weights
connecting the neurons as wij where i represents the neuron in the
previous layer and j represents the neuron in the current layer. The biases
for each neuron are denoted as bj.

Consider the followings:

Illustrate the iterative process of backpropagation.

4. Imagine you have a dataset containing measurements of 3 different 3 3
Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering
houses, with features including: Square footage (sq ft), Number of
bedrooms (bedrooms), Average income in the neighborhood
($10,000s).

House Square Footage (sq ft) Bedrooms Average Income ($10,000s)

1 2000 3 5
2 1500 2 7
3 2500 4 8

Apply PCA to reduce the dimensionality of the data while retaining

most of the variance.
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering
Subject: MACHINE LEARNING (6CS4-02)
Max Marks- 40

(Set-XI)
Q. No. Question BL CO
1. An image recognition model is deployed for identifying cats and dogs 3 3
in images. The model accurately classified 900 images of cats (True
Positive), correctly classified 800 images without cats (True Negative),
misclassified 30 images of dogs as cats (False Positive), and failed to
identify 40 images of cats (False Negative). Evaluate the model's
performance and discuss its reliability in animal recognition tasks.
2. Consider a 3x3 grid maze where the agent starts at position (0, 0) and 3 4
the goal is at position (2, 2). The agent can move in four directions: up,
down, left, and right. The numbers in the maze represent the rewards:
100 for reaching the goal, -1 for each step taken, and -100 for hitting an
obstacle. The agent aims to learn the optimal policy to navigate from the
start to the goal while maximizing its total reward.

Determine the optimal action based on its current state using Q-

Learning.
3. Consider a simple neural network with one input layer, hidden layer, and 3 5
output layer. The input layer has two neurons, the hidden layer has two
neurons, and the output layer has one neuron. We'll denote the weights
connecting the neurons as wij where i represents the neuron in the
previous layer and j represents the neuron in the current layer. The biases
for each neuron are denoted as bj.

Consider the followings:

Illustrate the iterative process of backpropagation.

4. Imagine you have a dataset containing measurements of 5 students, 3 3
with features including: Hours spent studying (hours), Scores on Math
exam (out of 100), Scores on English exam (out of 100)

Student Hours Studied Math Score English Score

1 5 80 75
2 8 90 85
Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering
3 3 70 65
4 6 85 80
5 9 95 90

Apply PCA to reduce the dimensionality of the data while retaining

most of the variance, aiming to identify underlying factors that might
influence both Math and English scores.
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering

Subject: MACHINE LEARNING (6CS4-02)

Max Marks- 40

(Set-XII)
Q. No. Question BL CO
1. A medical diagnostic model is trained to classify patients as having a certain 3 3
disease or not. The model correctly identified 180 patients with the disease (True
Positive), correctly identified 750 patients without the disease (True Negative),
misdiagnosed 20 healthy patients as having the disease (False Positive), and
missed 50 patients with the disease (False Negative). Assess the model's
performance and its suitability for medical diagnosis.
2. Consider a 3x3 grid maze where the agent starts at position (0, 0) and the goal is 3 4
at position (2, 2). The agent can move in four directions: up, down, left, and right.
The numbers in the maze represent the rewards: 100 for reaching the goal, -1 for
each step taken, and -100 for hitting an obstacle. The agent aims to learn the
optimal policy to navigate from the start to the goal while maximizing its total
reward.

Determine the optimal action based on its current state using Q-Learning.
3. Consider a simple neural network with one input layer, hidden layer, and output 3 5
layer. The input layer has two neurons, the hidden layer has two neurons, and the
output layer has one neuron. We'll denote the weights connecting the neurons as
wij where i represents the neuron in the previous layer and j represents the neuron
in the current layer. The biases for each neuron are denoted as bj.

Consider the followings:

Illustrate the iterative process of backpropagation.

4. Imagine you have a dataset containing measurements of 4 different fruits, with 3 3

features including: Weight (grams), Sugar content (grams), Vitamin C content
(mg), Here's the data:

Fruit Weight Sugar Content Vitamin C

Apple 180 10 8
 Swami Keshvanand Institute of Technology, Management &Gramothan, Jaipur
Department of Computer Science & Engineering
Banana 130 22 85
Orange 150 18 70
Mango 300 40 60

Apply PCA to reduce the dimensionality of the data while retaining most of the
variance, aiming to identify potential relationships between size, sweetness, and
vitamin content.