B.Tech (Semester V) | Theory Examination 2023–24 | AKTU
Time: 3 Hours | Maximum Marks: 100
This exam paper assesses a student’s understanding of Machine Learning (ML) principles, algorithms, and applications through conceptual, analytical, and coding-based questions.
It covers essential ML concepts such as supervised and unsupervised learning, SVM, decision trees, ANNs, CNNs, reinforcement learning, and genetic algorithms.
The paper is divided into three sections (A, B, C) that evaluate both theoretical knowledge and practical reasoning through detailed explanations and real-world applications.
Section A — Short Answer Questions (10 × 2 = 20 Marks)
This section contains 10 short conceptual questions, each testing fundamental understanding of ML algorithms, key definitions, and learning techniques.
Students are required to attempt all.
Sample Questions:
Discuss the important objectives of Machine Learning.
What is the difference between overfitting and underfitting in decision tree learning?
Explain the role of support vectors in an SVM model.
What is the gradient descent delta rule?
Define ANN and CNN.
Differentiate between lazy learning and eager learning.
Concepts Covered:
Overfitting vs underfitting, gradient descent, support vectors, decision tree suitability, ANN & CNN basics, and genetic algorithm terms like chromosome and gene.
Section B — Descriptive / Medium-Length Questions (3 × 10 = 30 Marks)
Students must attempt any three questions. This section evaluates the ability to explain and compare ML algorithms, demonstrate understanding through examples, and interpret different ML model types.
Sample Questions:
Compare Supervised and Unsupervised Learning techniques with suitable examples.
Explain Maximum Likelihood and Least Squared Error Hypothesis with examples.
Compare and contrast Information Gain, Gain Ratio, and Gini Index used in decision tree learning.
Explain different layers of a Convolutional Neural Network (CNN) with suitable examples.
Discuss applications of Reinforcement Learning and specify real-world problems where it is best used.
Concepts Covered:
Supervised vs unsupervised learning, probabilistic modeling, decision tree measures, CNN architecture, and reinforcement learning applications.
Section C — Long / Analytical Questions (5 × 10 = 50 Marks)
Students must attempt any one question from each part (Q3–Q7).
This section focuses on comprehensive understanding, algorithm design, and practical application through long analytical and theoretical responses.
Sample Questions:
Q3.
a. Compare Regression, Classification, and Clustering with real-life applications.
b. Explain the Concept Learning task with an example.
Q4.
a. Explain hyperplane (decision boundary) in SVM and categorize various popular kernel functions.
b. Differentiate between Naïve Bayes Classifier and Bayesian Belief Networks with real-world examples.
Q5.
a. Discuss the Decision Tree algorithm and explain its working in detail.
b. Demonstrate the K-Nearest Neighbors (KNN) algorithm for classification using an example.
Q6.
a. Illustrate the Backpropagation Algorithm and derive training rules for weights in hidden and output layers.
b. Write short notes on the Probably Approximately Correct (PAC) learning model.
Q7.
a. Explain Q-Learning, its key terms, features, and real-world applications.
b. Define the Genetic Algorithm and explain its working with a flowchart.
Concepts Covered:
Regression, classification, clustering, SVM, Naïve Bayes, KNN, Decision Trees, Backpropagation, Reinforcement Learning, Genetic Algorithms, and PAC model theory.
Learning Outcomes:
After completing this paper, students will be able to:
Understand and differentiate between core ML models and algorithms.
Apply statistical and mathematical reasoning in algorithm design.
Implement and analyze neural networks, decision trees, and SVMs.
Grasp key ML paradigms — supervised, unsupervised, and reinforcement learning.
Interpret ML’s real-world applications in automation, data prediction, and AI systems.
