B.Tech (Semester V) | Theory Examination 2023–24 | AKTU / UPTU
Subject Code: KCS056 | Maximum Marks: 100 | Time: 3 Hours
Description:
This question paper assesses a student’s understanding of Soft Computing concepts — including Neural Networks, Fuzzy Logic, and Genetic Algorithms — as applied in intelligent systems and data-driven computation.
The exam is structured into three sections (A, B, C), moving from fundamental definitions to analytical and applied design problems. It aims to test a student’s grasp of theoretical concepts, logical reasoning, and problem-solving in soft computing frameworks.
SECTION A — Short Answer Questions (10 × 2 = 20 Marks)
Students must attempt all ten short questions, each carrying two marks.
This section focuses on basic definitions, models, and conceptual clarity related to neurons, perceptrons, fuzzy sets, and genetic algorithms.
Sample Questions from Section A (Page 1):
Explain a neuron with its structure.
Define an artificial neural network (ANN).
Draw the multilayer perceptron model.
Discuss the different features of a single-layer perceptron.
Explain the roles of crisp sets.
Write one key difference between crisp and fuzzy sets.
List the basic fuzzy set operations.
Explain fuzzy relations with an example.
Define a genetic algorithm (GA).
List different types of encoding in genetic algorithms.
Concepts covered: Neuron models, ANN basics, perceptrons, fuzzy logic foundation, crisp vs fuzzy systems, and genetic algorithm encoding mechanisms.
SECTION B — Descriptive / Medium-Length Questions (3 × 10 = 30 Marks)
Students are required to attempt any three out of the five given questions.
This section tests the ability to describe and analyze core Soft Computing techniques and their mathematical principles.
Sample Questions from Section B (Page 1):
Explain activation functions and their role in the neuron model.
How is a linearly separable task defined in two-dimensional space? Discuss the XOR problem.
Explain all fuzzy set properties with examples.
Verify De Morgan’s Law using a truth table for three states.
Explain different selection methods in a genetic algorithm for choosing the next generation population.
Concepts covered: Activation functions, XOR problem, fuzzy logic laws, truth tables, and population selection methods in GA (like roulette wheel, tournament, and rank selection).
SECTION C — Analytical / Long Answer Questions (5 × 10 = 50 Marks)
This section evaluates applied understanding — students must attempt one part from each question (Q3–Q7).
It covers neural network models, fuzzy logic reasoning, and genetic computation mechanisms in detail.
Sample Questions from Section C (Pages 1–2):
Q3.
a. Draw a single-layer feed-forward network and explain its working.
b. Explain the working of a recurrent network and compare it with a multilayer neural network.
Q4.
a. Explain the McCulloch-Pitts model and list its disadvantages.
b. Draw a network for solving the XOR problem using perceptrons.
Q5.
a. Define the membership function and explain its importance in fuzzy logic.
b. Discuss two important fuzzy inference procedures (such as Mamdani and Sugeno models).
Q6.
a. Explain the attributes of predicate logic used in soft computing.
b. Define fuzziness of fuzzy sets and explain what a fuzzy function is.
Q7.
a. Discuss genetic operators and their role in GA.
b. Explain why mutation is important in GA and describe different types of mutation.
Concepts covered: Feedforward and recurrent networks, McCulloch-Pitts model, XOR neural implementation, fuzzy inference systems, predicate logic, genetic operators, and mutation types (bit flip, swap, inversion, etc.).
Learning Outcomes:
After completing this paper, students should be able to:
Understand the architecture and learning process of neural networks.
Apply fuzzy logic concepts to uncertain or imprecise problems.
Analyze and implement genetic algorithms for optimization problems.
Integrate soft computing techniques to develop hybrid intelligent systems.
