Roll Number:
Thapar Institute of Engineering and Technology, Patiala
Department of Electrical and Instrumentation Engineering
B. E EIC (VIII' Semester) Course Code : UEI704
ESE Course Name : Soft Computing Techniques
December 08, 2022 Thursday, 04.30 - 07.30 PM
Time: 3 Hours, M. Marks: 40 Name of Faculty: Dr. Saurabh Bhardwaj
Note: Attempt all questions; Assume missing data, if any, suitably
Q.1 (a) Explain Gradient Descent Algorithm (GA) by taking a suitable example (4)
i. Write the GA's weight and bias updating equations.
ii. List each of the tuning parameters.
iii. Explain what will probably happen if a learning rate is used that is too large
and when one is used that is too small.
Q.1 (b) i. List the main components of the Biological Neuron. Compare and contrast (4)
biological neurons and Artificial neurons.
ii. For the network shown below, calculate the net input to the output neuron
1
0.7
0.4
---1, ;-;,,,
-1
0.3 ,-.
x? ‘,..)--r• y
0.2 s(x-3
Q.2 Using the Center of Gravity and Center of Largest Area defuzzification methods, find (8)
the crisp value corresponding to the following fuzzy output sets.
p?,. pi.: . Pc,*
I I I
• -
a 1 2 3 4 5 5 3 1 / 3 4 .5 6 0 1 7 3 4 5
Output Fuzzy set 1 Output Fuzzy set 2 Output Fuzzy set 3
Q.3 (a) Briefly illustrate the fuzzy membership functions with examples and explain their (4)
importance in fuzzy logic. Describe how a neural network is used to obtain fuzzy
membership functions.
Q.3 (b) Write a short note on Fuzzy Relations. Two fuzzy relations are given by (4)
Z1 Z2 Z3
• Y1 Y2
ii = 0.1 0.41
0.11 & :s% = Yi 10.5
x1 [0.3
.r2 I.0.8 0.71 Y2 l0.4
0.2 0.21
Obtain Fuzzy relation l', as a composition between the fuzzy relations r? and S.
Q.4 (a) Find the size of the output volume in a convolutional neural network if the size of the (4)
input image is 28 x 28, and the other parameters are as follows: Filter size is 3 x 3, the
stride is 1, padding is 0, and there are ten filters in all. What should the padding be for
the convoluted matrix to be the same size as the input image?
Q.4 (b) For the following two vectors, A and B: (4)
{vector A: [ — ]} , f V ector B: [i]}
il
i. Plot a decision boundary for a perceptron network to recognize these two vectors.
ii. Find weight and bias that will produce the decision boundary found in part 1
Q.5 (a) Describe the differences between the Mamdani and Takagi-Sugeno fuzzy inference (4)
systems. Explain single-point crossover for permutation encoding using an example.
Q.5 (b) Write short notes on the following main features of Genetic Algorithms (GA). (4)
(I) Encoding (II) Fitness Function (III) Mutation (IV) Sel'ction
Note: Evaluated Answer Sheets will be shown on 190 December 2022 from 12:00 -12:30 PM in F Block.
