(SEM V) THEORY EXAMINATION 2024-25 MACHINE LEARNING TECHNIQUES

Subject Code: BCS055
Maximum Marks: 70
Time: 3 Hours
Paper ID: 310309

Question Paper Overview

SECTION A (2 × 7 = 14 Marks)

(Conceptual short-answer questions — basic ML theory and applications)

a. What constitutes a well-defined learning problem in Machine Learning? Give an example.
b. How does Machine Learning differ from Data Science, and how do they complement each other?
c. Explain the role of the Sigmoid Function in Logistic Regression.
d. What is the significance of the decision surface in Support Vector Machines (SVM)?
e. What is overfitting in decision tree learning, and how can it be avoided?
f. Explain the role of the activation function in a Perceptron.
g. Describe the role of the reward function in Reinforcement Learning.

SECTION B (Attempt any three × 7 = 21 Marks)

a. Describe the steps involved in designing a Machine Learning system. Illustrate each step with a real-world example.
b. Differentiate between Simple Linear Regression and Multiple Linear Regression. How does including multiple features affect model complexity?
c. Illustrate the steps of the ID3 algorithm, explaining attribute selection and tree construction.
d. Discuss the Self-Organizing Map (SOM) algorithm and how it performs clustering and dimensionality reduction.
e. Explain the role of the Q-learning function in Reinforcement Learning. How does it help agents learn optimal actions and policies?

SECTION C (Attempt one part from each question × 7 = 35 Marks)

Q3

(a) Discuss the history of Machine Learning with major milestones and advancements.
OR
(b) Compare Supervised, Unsupervised, and Reinforcement Learning with appropriate examples.

Q4

(a) Apply the Naïve Bayes Classifier to the given dataset (Play Tennis example).
Dataset:

Predict: {Outlook = Sunny, Temperature = Cool, Humidity = High, Wind = Strong}

OR
(b) Fit a Linear Regression model for the dataset

(x,y):(1,1.5),(2,3.0),(3,4.5),(4,6.0)(x, y): (1, 1.5), (2, 3.0), (3, 4.5), (4, 6.0)(x,y):(1,1.5),(2,3.0),(3,4.5),(4,6.0)

and predict y when x = 5.

Q5

(a) Describe Locally Weighted Regression (LWR) and how it differs from traditional regression models. Discuss its applications.
OR
(b) Given the dataset:

Compute entropy before split and after splitting on A. Determine which attribute (A or B) gives highest information gain.

Q6

(a) Derive the mathematical steps of Backpropagation for training a neural network.
OR
(b) Explain the architecture and working of Convolutional Neural Networks (CNNs).

Q7

(a) Explain the concept of Reinforcement Learning (RL) and how it differs from Supervised and Unsupervised Learning.
OR
(b) Describe the components of a Genetic Algorithm (GA) — chromosomes, genes, fitness function, and population — and explain the GA cycle.

Key Topics for Revision

1. Machine Learning Overview

Definition: Field of AI that enables systems to learn patterns from data.

Core Components: Data, Model, Learning Algorithm, Evaluation Metric.

Example: Email spam detection, recommender systems.

2. Types of Machine Learning

3. Logistic Regression & Sigmoid Function

Sigmoid Function:

• f(x)=11+e−xf(x) = \frac{1}{1 + e^{-x}}f(x)=1+e−x1​

Converts linear output into probabilities (0–1 range).

Use: Classification problems such as spam vs. non-spam.

4. SVM Decision Surface

Separates classes with a maximum-margin hyperplane.

Non-linear data handled via kernel trick (e.g., RBF kernel).

5. Overfitting in Decision Trees

Definition: Tree fits noise, not trend.

Prevention:                       Pruning (post/pre).

Limiting depth.                   Minimum samples per leaf.

6. Self-Organizing Map (SOM)

Unsupervised neural network for clustering and visualization.

Maps high-dimensional data to 2D grid preserving topological properties.

Useful in pattern recognition and dimensionality reduction.

7. Q-Learning

Model-free Reinforcement Learning algorithm.

Q(s, a) = Expected cumulative reward for taking action a in state s.

Update rule:

• Q(s,a)=Q(s,a)+α[r+γmax⁡a′Q(s′,a′)−Q(s,a)]Q(s,a) = Q(s,a) + \alpha [r + \gamma \max_{a'} Q(s',a') - Q(s,a)]Q(s,a)=Q(s,a)+α[r+γa′max​Q(s′,a′)−Q(s,a)]

8. Naïve Bayes Classifier

Based on Bayes’ theorem and feature independence assumption.

• P(C∣X)=P(X∣C)P(C)P(X)P(C|X) = \frac{P(X|C)P(C)}{P(X)}P(C∣X)=P(X)P(X∣C)P(C)​

Commonly used in text classification and sentiment analysis.

9. Locally Weighted Regression (LWR)

A non-parametric regression technique.

Assigns weights to nearby data points (closer points have higher influence).

Used in robotics and real-time predictive modeling.

10. Backpropagation

Gradient descent-based algorithm for training neural networks.

Steps:      Forward propagation (compute output).

Compute loss function (e.g., MSE).

Backpropagate errors to update weights.

11. Convolutional Neural Networks (CNNs)

Composed of convolutional, pooling, and fully connected layers.

Extracts spatial hierarchies of features.

Used in image classification, object detection, and medical imaging.

12. Genetic Algorithms (GAs)

Evolution-inspired optimization approach:

Initialize random population.

Evaluate fitness.

Apply selection, crossover, mutation.

Repeat until convergence.

Used in feature selection, scheduling, and parameter tuning.

Exam Tips

Draw diagrams for Neural Network, SVM margin, Decision Tree, CNN layers.

Remember entropy/information gain formulas for ID3.

Practice Naïve Bayes and Linear Regression numericals.

Revise short definitions: overfitting, reward function, activation function, Q-value, feature scaling.

Learn backpropagation steps and Reinforcement Learning cycle diagram.