(SEM VI) THEORY EXAMINATION 2023-24 MACHINE LEARNING TECHNIQUES

KAI601 – MACHINE LEARNING TECHNIQUES (B.Tech Sem VI)

The answers are written in simple, humanized language, not in short bullet points, and prepared strictly as per the uploaded question paper (both pages).
Reference: Uploaded Question Paper 

KAI601-MACHINE-LEARNING-TECHNIQ…

SECTION A

Attempt all questions in brief (2 × 10 = 20 marks)

(a) Name two types of learning commonly used in machine learning.

Two commonly used types of learning in machine learning are supervised learning and unsupervised learning. Supervised learning works with labeled data, while unsupervised learning finds hidden patterns in unlabeled data.

(b) Give an example of a supervised learning problem.

An example of a supervised learning problem is email spam detection, where emails are labeled as “spam” or “not spam” and the model learns to classify new emails based on past examples.

(c) What is logistic regression, and how does it differ from linear regression?

Logistic regression is a classification algorithm used to predict categorical outcomes such as yes/no or true/false. Unlike linear regression, which predicts continuous values, logistic regression uses a sigmoid function to output probabilities between 0 and 1.

(d) What are the three types of support vector kernels commonly used in SVMs?

The three commonly used kernels are linear kernel, polynomial kernel, and radial basis function (RBF) kernel. These kernels help transform data into higher-dimensional space for better class separation.

(e) Define inductive bias in the context of decision tree learning.

Inductive bias refers to the set of assumptions a learning algorithm makes to generalize beyond the training data. In decision tree learning, the inductive bias favors smaller trees and attributes with higher information gain.

(f) Describe the process of locally weighted regression in instance-based learning.

Locally Weighted Regression predicts output by fitting a model around a query point using nearby training instances. Closer data points are given higher weights, allowing flexible and accurate local predictions.

(g) Define perceptron and their role in artificial neural networks.

A perceptron is the simplest form of a neural network unit that takes inputs, applies weights, sums them, and passes the result through an activation function. It forms the basic building block of artificial neural networks.

(h) What are the key characteristics of the Self-Organizing Map (SOM) algorithm?

SOM is an unsupervised learning algorithm that maps high-dimensional data into a lower-dimensional grid while preserving topological relationships. It is widely used for clustering and visualization.

(i) Define Reinforcement Learning (RL) and explain its key components.

Reinforcement Learning is a learning technique where an agent learns by interacting with an environment and receiving rewards or penalties. Its key components include agent, environment, state, action, and reward.

(j) Discuss the components of a genetic algorithm.

A genetic algorithm consists of population initialization, fitness evaluation, selection, crossover, mutation, and termination. These components mimic natural evolution to find optimal solutions.

SECTION B

Attempt any three (10 × 3 = 30 marks)

(a) Supervised vs unsupervised vs reinforcement learning

Supervised learning uses labeled data to predict outcomes and is commonly used in classification and regression problems. Unsupervised learning works on unlabeled data and identifies hidden patterns or clusters. Reinforcement learning focuses on decision-making by learning through rewards and penalties over time. Each approach suits different real-world problems.

(b) Mathematical formulation of linear regression

Linear regression models the relationship between input and output using a linear equation. The hypothesis function predicts output as a weighted sum of input features. The cost function measures prediction error, usually using mean squared error. Gradient descent is used to minimize the cost and optimize parameters.

(c) ID3 algorithm for decision tree construction

The ID3 algorithm constructs decision trees using information gain as the selection criterion. It calculates entropy for each attribute and selects the attribute that provides maximum information gain. This process continues recursively until all samples are classified or stopping conditions are met.

(d) Backpropagation algorithm and its importance

Backpropagation is a learning algorithm used to train neural networks by minimizing error. It propagates error backward from output to input layers and updates weights using gradient descent. This enables deep networks to learn complex patterns.

(e) Q-learning algorithm in reinforcement learning

Q-learning is a model-free reinforcement learning algorithm that learns optimal actions using a Q-table. It updates Q-values based on reward and future expectations, enabling agents to learn optimal policies over time.

SECTION C

Attempt any one (10 marks)

(a) Bayesian networks and probabilistic relationships

Bayesian networks represent variables and their conditional dependencies using directed acyclic graphs. They are constructed using probability distributions and updated using Bayes’ theorem. Bayesian networks are used in medical diagnosis, risk assessment, and decision support systems.

(b) Model evaluation metrics

Model evaluation measures how well a machine learning model performs. Accuracy measures overall correctness, precision measures correctness of positive predictions, recall measures ability to identify positives, and F1-score balances precision and recall. Each metric is useful depending on the problem context.