JAYA SAKTHI ENGINEERING COLLEGE

[Approved By AICTE, New Delhi & Affiliated to

ANNA UNIVERSITY, Chennai]
Thiruninravur– 602 024, Thiruvallur Dt., Tamil Nadu.

QUESTION BANK (AY: 2024-2025) EVEN SEMESTER

Department : AI&DS Subject Code : AL3451

Year/ Sem : II/ IV Subject : MACHINE LEARNING

Regulation : 2021 Staff In charge : NANDHINI K

UNIT I INTRODUCTION TO MACHINE LEARNING

Review of Linear Algebra for machine learning; Introduction and motivation for machine
learning; Examples of machine learning applications, Vapnik-Chervonenkis (VC) dimension,
Probably Approximately Correct (PAC) learning, Hypothesis spaces, Inductive bias,
Generalization, Bias variance trade-off.

CO1: To understand the basic concepts of machine learning.

PART-A (2 Marks)
S.No Question K CO
Level
1 What is Machine Learning? K1 CO1
2 What are the types of Machine Learning? Explain with examples. K1 CO1

3 What is the difference between classification and regression? K2 CO1

4 Explain the concept of supervised learning K2 CO1

5 Explain the concept of unsupervised learning. K1 CO1

6 What are training data and test data? Why are they important? K1 CO1

7 What is overfitting? How can it be prevented? K1 CO1

8 What is underfitting in Machine Learning? How can it be addressed? K1 CO1

9 What are the basic steps in a typical machine learning project? K2 CO1
10 What is a confusion matrix? K1 CO1

11 What is the bias-variance trade-off? K1 CO1

12 Explain the difference between a model and an algorithm in machine learning K1 CO1

13 What are evaluation metrics in supervised learning? K1 CO1

14 What is cross-validation? How is it useful in machine learning? K1 CO1

15 What is feature selection, and why is it important? K1 CO1

16 Explain the concept of feature scaling. Why is it necessary? K2 CO1

17 What is a decision tree, and how does it work in classification tasks? K1 CO1

18 What are K-nearest neighbors (KNN)? How does KNN classify data points? K1 CO1

19 What is a support vector machine (SVM)? K1 CO1

20 What are the differences between a linear regression model and a logistic regression K1 CO1
model?

PART-B (13 Marks)

S.No Question K CO
Level
1 Explain the working of the k-Nearest Neighbors (KNN) algorithm. K2 CO1

2 What is the difference between classification and clustering algorithms? K2 CO1

3 Explain the concept of linear regression K2 CO1

4 Explain gradient descent and its role in machine learning. K2 CO1

5 What is a cost function in machine learning? K2 CO1

6 Explain the working of logistic regression K2 CO1

7 What is a decision tree? Explain the concept of entropy and information gain in K2 CO1
decision trees.
8 What is random forest, and how does it work? K2 CO1

9 What is the difference between bagging and boosting in ensemble methods? K2 CO1

10 Explain the concept of Support Vector Machines (SVM) and its use in classification. K2 CO1
11 What are kernel functions in SVM, and why are they used? K2 CO1

12 What is Naive Bayes classifier? K2 CO1

13 What is K-means clustering? Explain how the algorithm works. K2 CO1

15 What are principal components in Principal Component Analysis (PCA)? K2 CO1

16 What is the role of regularization in machine learning? Explain L1 and L2 K2 CO1

regularization.
17 What are neural networks, and how do they work? K2 CO1

18 What is deep learning, and how does it differ from traditional machine learning? K2 CO1

20 What is the difference between batch gradient descent and stochastic gradient descent? K2 CO1

21 What is cross-validation, and how is it used to evaluate model performance? K2 CO1

22 What is the ROC curve, and how is it used to evaluate classification models? K2 CO1

PART-C (15 Marks)

S.No Question K CO
Level
1 Explain the concept of model evaluation. What are the different techniques used for K2 CO1
evaluating machine learning models?

2 What is the significance of hyperparameter tuning in machine learning models? K2 CO1

3 What are the different types of activation functions used in neural networks? K2 CO1

4 What is the purpose of dropout in deep learning models? K2 CO1

5 Explain the concept of gradient boosting. How does it improve upon decision trees? K2 CO1

6 . What is the difference between supervised and unsupervised anomaly detection K2 CO1
techniques?
7 What is transfer learning in deep learning? K3 CO1

8 Explain the concept of word embeddings and their use in Natural Language Processing K3 CO1
(NLP).

9 What is the importance of the learning rate in optimization algorithms? K3 CO1

10 How do Convolutional Neural Networks (CNNs) work? Explain their application in K3 CO1
image classification.
 UNIT II ENSEMBLE TECHNIQUES AND UNSUPERVISED LEARNING

Linear Regression Models: Least squares, single & multiple variables, Bayesian linear
regression, gradient descent, Linear Classification Models: Discriminant function – Perceptron
algorithm, Probabilistic discriminative model - Logistic regression, Probabilistic generative
model – Naive Bayes, Maximum margin classifier – Support vector machine, Decision Tree,
Random Forests

CO2: To understand and build supervised learning models.

PART-A (2 Marks)

S.No Question K CO
Level
1 What is the least squares method in linear regression? K1 CO2

2 Define single-variable linear regression. K1 CO2

3 What is the difference between simple and multiple linear regression? K2 CO2

4 How is Bayesian linear regression different from standard linear regression? K2 CO2

5 What is gradient descent in the context of linear regression? K1 CO2

6 Describe the perceptron algorithm. K1 CO2

7 What is a discriminant function in linear classification? K1 CO2

8 What is logistic regression, and how does it differ from linear regression? K1 CO2

9 Define Naive Bayes classifier. K2 CO2

10 What is the maximum margin classifier in SVM? K1 CO2

11 What is the role of a kernel in Support Vector Machines (SVM)? K1 CO2

12 How do Decision Trees make predictions? K1 CO2

13 What is a Random Forest and how does it improve upon Decision Trees? K1 CO2

14 What is overfitting in decision trees? K1 CO2

15 How does the perceptron algorithm handle linearly separable data? K1 CO2

16 What is the loss function in logistic regression? K2 CO2

17 Define the term "hyperplane" in the context of SVM. K1 CO2

18 How does the Naive Bayes classifier handle categorical data? K1 CO2

19 What are the advantages of using Random Forest over a single decision tree? K1 CO2

20 What is the decision boundary in logistic regression? K1 CO2

21 Explain the concept of "boosting" in machine learning. K1 CO2

22 What is meant by the term "margin" in SVM? K1 CO2

23 What are the assumptions made by the Naive Bayes classifier? K1 CO2

24 How does SVM deal with non-linearly separable data? K1 CO2

25 Define "feature importance" in decision trees and Random Forests. K1 CO2

PART-B (13 Marks)

S.No Question K CO
Level
1 Explain the working of Linear Regression and its types (simple and multiple). K2 CO2

2 Derive the normal equation for the least squares method in linear regression. K2 CO2

3 Describe the perceptron algorithm and its convergence criteria. K2 CO2

4 Compare and contrast logistic regression and Naive Bayes classifiers. K2 CO2

5 Explain the concept of the support vector machine (SVM) and how it maximizes the K2 CO2
margin.

6 Describe the working and advantages of a Decision Tree classifier. K2 CO2

7 Discuss the Random Forest algorithm and its applications. K2 CO2

8 What is the role of regularization in linear regression and logistic regression? K2 CO2

9 Derive the gradient descent update rule for linear regression. K2 CO2

10 Explain the concept of overfitting and underfitting in supervised learning. K2 CO2

11 Discuss the implementation of K-fold cross-validation in the evaluation of classifiers. K2 CO2

12 Explain the Naive Bayes classifier with an example. K2 CO2

13 Discuss the logistic regression in detail, including its decision boundary and cost K2 CO2
function.

14 Explain how the perceptron algorithm can be used for binary classification. K2 CO2

15 Describe the difference between discriminative and generative models with examples. K2 CO2

PART-C (15 Marks)

S.No Question K CO
Level
1 Discuss in detail the linear regression model with multiple variables. How does it work K2 CO2
in multivariate scenarios?

2 Explain the concept of Bayesian linear regression and how it can be used for regression K2 CO2
tasks.

3 Elaborate on the working of the perceptron algorithm, including its decision-making K2 CO2
process and convergence.

4 Discuss in detail the probabilistic discriminative model (Logistic Regression) and its K2 CO2
use in binary classification.

5 Explain the maximum margin classifier (SVM) and its relation to the kernel trick. K2 CO2

6 Discuss the concept of decision trees and Random Forests with their advantages and K2 CO2
disadvantages.

7 How does Naive Bayes handle categorical data? Discuss its performance in text K3 CO2
classification.

8 Discuss the importance of regularization techniques such as L1 and L2 in linear K3 CO2

models.

9 Explain the difference between bagging and boosting in ensemble learning. Provide K3 CO2
real-world examples.

10 Compare the use of gradient descent in linear regression versus neural networks. K3 CO2
 UNIT III ENSEMBLE TECHNIQUES AND UNSUPERVISED LEARNING

Combining multiple learners: Model combination schemes, Voting, Ensemble Learning -

bagging, boosting, stacking, Unsupervised learning: K-means, Instance Based Learning: KNN,
Gaussian mixture models and Expectation maximization.

CO3: To understand and build unsupervised learning models.

PART-A (2 Marks)

S.No Question K CO
Level
1 Define ensemble learning. K1 CO3

2 What is bagging in ensemble learning? K1 CO3

3 What is boosting in ensemble learning? K2 CO3

4 What is stacking in ensemble learning? K2 CO3

5 How does voting work in ensemble methods? K1 CO3

6 Define K-means clustering. K1 CO3

7 What is an instance-based learning method? K1 CO3

8 How does K-nearest neighbors (KNN) algorithm work? K1 CO3

9 What is the major drawback of KNN? K2 CO3

10 Explain the concept of Gaussian Mixture Models (GMM). K1 CO3

11 What is expectation maximization? K1 CO3

12 Describe the concept of overfitting in unsupervised learning. K1 CO3

13 What is the difference between supervised and unsupervised learning? K1 CO3

14 Explain the working of hierarchical clustering. K1 CO3

15 How is bagging different from boosting? K1 CO3

16 What is the role of distance metrics in KNN? K2 CO3

17 What are the advantages of K-means over other clustering algorithms? K1 CO3

18 How do decision trees improve in a random forest? K1 CO3

19 What is the use of the softmax function in ensemble methods? K1 CO3

20 Define the term "feature selection" in unsupervised learning. K1 CO3

21 How does K-means handle outliers? K1 CO3

22 What is the purpose of bootstrapping in ensemble learning? K1 CO3

23 Describe a case where boosting would perform better than bagging. K1 CO3

24 What is the primary goal of ensemble learning techniques? K1 CO3

25 Define and explain "weighted voting" in ensemble learning. K1 CO3

PART-B (13 Marks)

S.No Question K CO
Level
1 Explain bagging and boosting in detail and compare their advantages and K2 CO3
disadvantages.

2 Describe how KNN works, including the importance of distance metrics and its impact K2 CO3
on performance.

3 Explain the working of Gaussian Mixture Models (GMM) and its applications. K2 CO3

4 Discuss the K-means algorithm with the steps involved and its limitations. K2 CO3

5 What is the difference between hard and soft clustering in unsupervised learning? K2 CO3

6 Describe the EM (Expectation Maximization) algorithm and its use in Gaussian K2 CO3
Mixture Models.

7 Explain the ensemble method of stacking, with an example. K2 CO3

8 Compare the performance of KNN and K-means in clustering problems. K2 CO3

9 Discuss in detail the voting mechanism in ensemble learning and its variants. K2 CO3

10 How does boosting help reduce bias in machine learning models? K2 CO3

11 Explain the importance of cross-validation in ensemble methods. K2 CO3

12 Describe how the Random Forest algorithm works and the advantages of ensemble K2 CO3
trees.

13 Explain the concept of unsupervised learning and its applications. K2 CO3

14 Discuss the concept of "dimensionality reduction" and its importance in unsupervised K2 CO3
learning.

15 Explain how instance-based learning (like KNN) works, and discuss its computational K2 CO3
complexity.

PART-C (15 Marks)

S.No Question K CO
Level
1 Explain the working of K-means clustering, and discuss the advantages and limitations K2 CO3
of the algorithm.
2 Describe Gaussian Mixture Models (GMM) and their application to clustering K2 CO3
problems.
3 Discuss the bagging and boosting ensemble methods, their working, and applications in K2 CO3
real-world problems.
4 Discuss the importance of hyperparameters in boosting techniques. K2 CO3

5 Explain the concept and working of decision trees and Random Forest in ensemble K2 CO3
learning.
6 Describe the challenges in applying KNN to large datasets and how these can be K2 CO3
mitigated.
7 Compare and contrast clustering algorithms such as K-means, hierarchical clustering, K3 CO3
and DBSCAN.
8 Explain the concept of ensemble learning and its advantages over single learners. CO3

9 Discuss the application of Expectation Maximization in parameter estimation and CO3

clustering.
10 Discuss the concept of dimensionality reduction and its importance in clustering and CO3
ensemble learning.
 UNIT IV NEURAL NETWORKS

Multilayer perceptron, activation functions, network training – gradient descent optimization –

stochastic gradient descent, error backpropagation, from shallow networks to deep networks –
Unit saturation (aka the vanishing gradient problem) – ReLU, hyperparameter tuning, batch
normalization, regularization, dropout.

CO4: To evaluate the algorithms based on corresponding metrics identified

PART-A (2 Marks)
S.No Question K CO
Level
1 What is a Multilayer Perceptron (MLP)? K1 CO4

2 What are activation functions in neural networks? K1 CO4

3 Define the term "backpropagation." K2 CO4

4 What is the role of gradient descent in neural networks? K2 CO4

5 Explain the concept of stochastic gradient descent. K1 CO4

6 What is the vanishing gradient problem? K1 CO4

7 How does ReLU activation function work? K1 CO4

8 What is batch normalization in neural networks? K1 CO4

9 Define the term "dropout" in the context of neural networks. K2 CO4

10 What are the different types of neural network architectures? K1 CO4

11 What is a neural network layer? K1 CO4

12 Define hyperparameters in the context of neural networks. K1 CO4

13 What is an epoch in neural network training? K1 CO4

14 How does the ReLU activation function help with the vanishing gradient problem? K1 CO4

15 What are the types of loss functions used in neural networks? K1 CO4

16 What is a learning rate in neural networks? K2 CO4

17 Explain what is meant by a shallow neural network. K1 CO4

18 What is the purpose of regularization in neural networks? K1 CO4

19 Define the concept of overfitting in neural networks. K1 CO4

20 What is the main goal of training a neural network? K1 CO4

21 What is the difference between supervised and unsupervised neural networks? K1 CO4

22 How is backpropagation implemented in a neural network? K1 CO4

23 What is a hyperparameter tuning in neural networks? K1 CO4

24 Explain the concept of a convolutional neural network (CNN). K1 CO4

25 How does the dropout technique help prevent overfitting in neural networks? K1 CO4

PART-B (13 Marks)

S.No Question K CO
Level
1 Explain the concept and working of a Multilayer Perceptron (MLP). K2 CO4

2 Discuss the process of backpropagation and its role in training neural networks. K2 CO4

3 Explain the role of activation functions in neural networks. Compare and contrast K2 CO4
common activation functions like sigmoid, tanh, and ReLU.

4 Discuss the vanishing gradient problem and how it affects neural network training. K2 CO4

5 Explain stochastic gradient descent and how it differs from batch gradient descent. K2 CO4

6 Discuss the role and implementation of batch normalization in deep networks. K2 CO4

7 Explain the concept of dropout in neural networks and its role in preventing overfitting. K2 CO4

8 Discuss the importance of learning rates in training neural networks and how they K2 CO4
affect convergence.

9 Explain how the gradient descent optimization technique is applied in training neural K2 CO4
networks.

10 Discuss the role of regularization techniques in preventing overfitting in neural K2 CO4

networks.

11 What are the various types of neural network architectures, and how are they suited to K2 CO4
different types of problems?
12 Discuss the significance of hyperparameter tuning in improving the performance of K2 CO4
neural networks.

13 How do deep neural networks differ from shallow networks? What challenges arise K2 CO4
when training deep networks?

14 What is the ReLU activation function, and why is it widely used in deep learning? K2 CO4

15 Explain how different loss functions in neural networks are chosen for different tasks K2 CO4
(classification, regression).

PART-C (15 Marks)

S.No Question K CO
Level
1 Explain in detail the architecture and working of a Multilayer Perceptron (MLP) in K2 CO4
neural networks.

2 Discuss the role of activation functions in neural networks, with a focus on ReLU and K2 CO4
its advantages over other activation functions.

3 Describe the backpropagation algorithm, how it works, and its impact on neural K2 CO4
network training.

4 Explain stochastic gradient descent (SGD) in detail, including its variants like mini- K2 CO4
batch and full-batch gradient descent.

5 Discuss the vanishing gradient problem and its solutions such as ReLU and batch K2 CO4
normalization.

6 Describe the dropout technique and how it is used to prevent overfitting in neural K2 CO4
networks.

7 Explain the process of training deep neural networks and the challenges associated with K3 CO4
deep architectures.

8 Discuss the concept of hyperparameter tuning and how it can improve the performance K2 CO4
of neural networks.

9 Explain batch normalization in the context of deep neural networks and how it helps K2 CO4
with training.

10 Discuss the various challenges and solutions in training deep networks, such as K2 CO4
vanishing gradients and poor convergence.
 UNIT V DESIGN AND ANALYSIS OF MACHINE LEARNING EXPERIMENTS

Guidelines for machine learning experiments, Cross Validation (CV) and resampling – K-fold
CV, bootstrapping, measuring classifier performance, assessing a single classification algorithm
and comparing two classification algorithms – t test, McNemar’s test, K-fold CV paired t test

CO4: To evaluate the algorithms based on corresponding metrics identified

PART-A (2 Marks)

S.No Question K CO
Level
1 What is cross-validation in machine learning experiments? K1 CO4

2 Define K-fold cross-validation. K1 CO4

3 What is bootstrapping in machine learning? K2 CO4

4 Explain the purpose of measuring classifier performance. K2 CO4

5 What is the significance of a t-test in comparing classifiers? K1 CO4

6 Define McNemar’s test and its purpose. K1 CO4

7 What is a paired t-test in K-fold cross-validation? K1 CO4

8 How is classifier accuracy measured? K1 CO4

9 What are the key steps in designing a machine learning experiment? K2 CO4

10 Explain how cross-validation helps in model validation. K1 CO4

11 What is the difference between training and test datasets? K1 CO4

12 What is the importance of randomness in bootstrapping? K1 CO4

13 Define the concept of resampling in machine learning. K1 CO4

14 What is the purpose of a confusion matrix? K1 CO4

15 Explain the use of precision and recall in classifier performance evaluation. K1 CO4

16 What are false positives and false negatives in classification? K2 CO4

17 How does the ROC curve help in classifier performance evaluation? K1 CO4

18 What is a learning curve, and how is it used in machine learning experiments? K1 CO4
19 How do you assess the bias-variance tradeoff in machine learning? K1 CO4

20 What is the significance of the area under the ROC curve (AUC)? K1 CO4

21 Define the term "model overfitting" and how it is detected. K1 CO4

22 How does cross-validation prevent overfitting in machine learning models? K1 CO4

23 What are the advantages of K-fold cross-validation over other validation techniques? K1 CO4

24 What is the importance of feature selection in machine learning experiments? K1 CO4

25 What is the purpose of a random split in training and testing datasets? K1 CO4

PART-B (13 Marks)

S.No Question K CO
Level
1 Discuss the process and significance of cross-validation in machine learning. K2 CO4

2 Explain the concept of K-fold cross-validation and its advantages. K2 CO4

3 Discuss the differences between K-fold cross-validation and bootstrapping. K2 CO4

4 How do you evaluate and compare two machine learning classifiers using statistical K2 CO4
tests?

5 Discuss how McNemar’s test is used to compare classifiers. K2 CO4

6 Explain the procedure and significance of a t-test in classifier comparison. K2 CO4

7 Discuss the importance of classifier performance metrics such as accuracy, precision, K2 CO4
recall, and F1-score.

8 Explain how cross-validation can help in preventing overfitting. K2 CO4

9 Describe the concept of resampling techniques like bootstrapping and how they K2 CO4
improve model evaluation.

10 Discuss the role of a confusion matrix in evaluating classification algorithms. K2 CO4

11 How do you measure and interpret the performance of classifiers using ROC curves? K2 CO4

12 Discuss the importance of regularization techniques in machine learning experiments. K2 CO4

13 What are the challenges in interpreting and comparing model performance in machine K2 CO4
learning?

14 Discuss the purpose and use of learning curves in machine learning experiments. K2 CO4

15 How do you assess a classifier’s performance using precision-recall curves and AUC? K2 CO4

PART-C (15 Marks)

S.No Question K CO
Level
1 Explain the role of cross-validation in the design and analysis of machine learning K2 CO4
experiments.

2 Discuss the process of measuring classifier performance and the importance of K2 CO4
precision and recall.

3 Discuss the use of McNemar’s test and paired t-tests in comparing two classification K2 CO4
algorithms.

4 Explain the significance of K-fold cross-validation in assessing the performance of K2 CO4

machine learning models.

5 How do statistical tests such as the t-test and McNemar's test help in comparing K2 CO4
classifiers?

6 Discuss the bias-variance tradeoff and its impact on model performance. K2 CO4

7 Explain how resampling techniques like bootstrapping help in improving model K3 CO4
performance evaluation.

8 Discuss the importance of model regularization in the context of machine learning K3 CO4
experiments.

9 How do confusion matrices, ROC curves, and AUC help evaluate classifier K3 CO4
performance?

10 Explain the process of designing a machine learning experiment from dataset collection K3 CO4
to result evaluation.
CO1: To understand the basic concepts of machine learning.

CO2: To understand and build supervised learning models.

CO3: To understand and build unsupervised learning models.

CO4: To evaluate the algorithms based on corresponding metrics identified

STAFF INCHARGE AC PRINCIPAL DIRECTOR