You are on page 1of 2

Enrollment No:_____________

GSFC UNIVERSITY
SCHOOL OF TECHNOLOGY
B.Tech. Computer Science & Engineering Regular End Sem Examination (Even Session 2021-22)
Semester: VI Date of Exam: 10/06/2022
Course Code: BTCS604 Time: 10:00 am – 12:00 pm
Course Name: Deep Learning Total Marks: 40
Instructions
1. The paper consists total 06 (six) questions.
2. Attempt any 04 (four) questions, each question carries equal marks.
3. Figures to the right indicate full marks.
4. Make suitable assumptions wherever necessary.

1. Answer the following [10]


a. What is the importance of weights and biases in a neural network? What are the things to keep in
mind while initializing weights and biases?
b. What do you mean by activation function in neural network? Explain Sigmoid, Tanh and ReLU
activation functions.
2. Explain the difference between [10]
a. Sigmoid and SoftMax activation functions.
b. Batch and Stochastic Gradient Descent
c. Local and Global Minima
d. Vanishing and exploding gradients
3. Explain in brief [10]
a. Explain RNN with suitable diagram. Why RNN is best suited for sequential data?
b. What are the various types of RNN? Explain with example: One to One, One to Many, Many to
One and Many to Many RNN.
c. What is difference between CNN and RNN?
4. Answer the following [10]
a. What is the difference between LSTM and GRU?
b. What is the main cause of overfitting and underfitting in a neural network?
c. What is dropout and early stopping? How does it prevent overfitting in a neural network?
5. Compute the following [10]
a. An input image has been converted into a matrix of size 15 X 15 with a filter of size 5 X 5 with
stride of 2. Determine the size of the convolution matrix
b. Given the input matrix as (A)
2 1 0 2 3
9 5 4 2 0
2 3 4 5 6
1 2 3 1 0
0 4 4 2 8
And the filter is (F)
-1 0 1
-1 0 1
-1 0 1

i. Find the convolution Matrix when filter F is applied on Matrix A


ii. Given the input matrix A, write the final matrix when filter size = 2, stride = 1, Pooling =
Max Pooling

Page 1 of 2
6. Let us Consider a CNN having 3 different layers in its architecture as [10]
Layer 1: Filter size – 3 X 3, No. of filters – 10, stride – 1, Padding -0
Layer 2 : Filter size – 5 X 5, No. of filters – 20, Stride – 2, Padding -0
Layer 3 Filter size – 5 X 5, No of filters – 40, stride – 2, Padding -0

If we give input a 3D image to the network of dimension 39X 39 , then determine the dimension of vector
passing through a fully connected layer in the architecture.
Mention the dimensions of all the hidden layers as well as draw suitable diagram to justify your answer.

*********

Page 2 of 2

You might also like