Total No. of Questions : 4] SEAT No.
8
23
P8507 [Total No. of Pages : 1
ic-
Oct-22/BE/Insem - 105
tat
6s
B.E. (Information Technology)
8:1
DEEP LEARNING
02 91
(2019 Pattern) (Semester - VII) (414443)
3:3
0
21
Time : 1 Hour]
6/1 13 [Max. Marks : 30
0
Instructions to the candidates:
0/2
.23 GP
1) Answer Q1 or Q2, Q3 or Q4.
2) Neat diagrams must be drawn wherever necessary.
E
80
3) Figures to the right side indicate full marks.
8
C
23
4) Assume suitable data, if necessary.
ic-
16
tat
8.2
Q1) a) Draw and explain the architecture of Multilayered Feedforward Neural
6s
.24
network. [5]
8:1
91
b) What is the need of Regularization? Explain Dropout Regularization.[5]
49
3:3
c) Explain the concept of gradient based Learning. [5]
30
21
01
02
OR
0/2
GP
6/1
Q2) a) What is the problem of vanishing Gradient? Describe various solutions
CE
80
8
to this problem. [7]
23
.23
ic-
b) Explain the working of an Artificial neuron. Also explain the activation
16
functions ReLU and LReLU. [8]
tat
8.2
6s
.24
8:1
Q3) a) Illustrate Convolution operation in CNN with an example. [5]
91
b) Explain the use of padding and strides in pooling layers. [5]
49
3:3
30
c) What is the advantage of weight sharing in CNN. [5]
21
01
02
OR
0/2
GP
6/1
Q4) a) What are pooling layers in CNN? Illustrate Max pooling with an example.[5]
CE
80
b) Discuss applications of CNN. [5]
.23
c) Write short note on AlexNet. [5]
16
8.2
.24
49
