UEI704
Uploaded by
vivesod153UEI704
Uploaded by
vivesod153Roll Number:
Thapar Institute of Engineering and Technology, Patiala
Department of Electrical and Instrumentation Engineering
B.E EIC (HP Semester) Course Code : UEI704
Auxiliary Examination Course Name : Soft Computing Techniques
August 28, 2023 Thursday, 05.30 = 08.30 PM
Time: 3 Hours, M. Marks: 100 Name of Faculty: Dr. Saurabh Bhardwaj
Note: Attempt all questions; Assume missing data, if any, suitably
Q.1 (a) What is your interpretation of the term Soft-Computing. How does Soft Computing (10)
differ from Hard Computing? Elaborate on the primary components of a biological
neuron.
Q.1 (b) In a multilayer perceptron with an input layer of ten neurons, a hidden layer of six (5)
neurons, and an output layer of three neurons, all using the ReLU activation
function, how many total parameters (weights and biases) are there in the
network?
Q.2 (a) Discuss the difference between crisp 'output and fuzzy output in the context of (10)
defuzzification. Briefly explain the concept of centroid defuzzification. How is the
centroid of an output fuzzy set computed?
Q.2 (b) Given two fuzzy sets A and B with membership functions A = [0.2, 0.5, 0.7] and (10)
that of B = [0.4, 0.6, 0.9], compute the crisp output using the Mean of Maxima
method.
Q.3 (a) Write down the unified learning rule of perceptron. Consider a perceptron with (10)
inputs [0.5, -0.3], weights [0.2, -0.4], and a bias of 0.1. The target output is 1, but
the perceptron's output is -0.2. Using the unified learning rule, calculate the
updated weights and bias.
Q.3 (b) Explain the role of the rectified linear unit (ReLU) activation function in neural (10)
networks. What advantages does it offer over traditional activation functions like
the sigmoid? For a neuron with an input of 2.0, a weight of 1.5, a bias of -0.5, and
using the sigmoid activation function calculate the output of the neuron.
Q.4 (a) Describe the process of selection in a genetic algorithm. Explain the following (15)
selection techniques: roulette wheel selection, tournament selection, and rank-
based selection? Compare and contrast these techniques, highlighting their
advantages and potential drawbacks. '
Q.4 (b) Describe the Hebbian learning rule. What is the underlying principle of Hebbian (10)
learning, and how does it contribute to strengthening connections between
neurons?
Q.5 (a) You have a perceptron with two input features: xi=0.6 and x2=0.8, and weights w i (10)
=0.4 and w2=0.7. The threshold is 0.5. Apply the perceptron learning rule to update
the weights for a target output of 1. Show the step-by-step calculations.
Q.5 (b) Solve the following questions in relation to the fuzzy logic (10)
a. Given a triangular membership function with points A(2,0), B(4,1), and C(6,0),
calculate,the membership value for x = 5.
b. Given two fuzzy sets for "Temperature" and "Humidity" with membership
values T=0.6, H=0.4, and a rule ''If Temperature is Cold and Humidity is Low, then
Output is Yes," calculate the fuzzy output using the min. Gperator.
