NAME OF PROGRAMME: B.

TECH(CSE) - AIML
SEMESTER / YEAR: V / IIIRD
SUBJECT NAME (SUBJECT CODE): DEEP LEARNING AND NEURAL NETWORK (NTID
357)
Some Important Guidelines for the Question Bank Setter:
1. Question bank must cover: Course (subject) Learning Outcomes (CLOs) and bloom's taxonomy
(L1 : Remember, L2 : Understand, L3 : Apply, L4 : Analyse) details in this regard are attached with
the mail.
2. The question bank should be prepared in the given format would also be attached.
3. Each question and parts of the questions should be written in clear language. Also break the questions
relatively in shorter sentences if they contain brief information.
4. Repetition of a question is not allowed.
5. The file should be sent in MS-Word format.
6. The font size of the content should be Arial (font size 12) for English & Kruti Dev 010 (font size 14) for Hindi.
7. Wherever the question papers have been prepared in both Hindi and English languages, the Hindi
version of the question should be written immediately after English version of each question.
8. In case of MBA Course, Section-C must contain Case studies (one case study per unit or numerical
type questions as per the format).
9. If the syllabus contains more than 5 units or less than 5 units then update the format accordingly.
10. A question bank moderation committee will be formed by the Dean of the concerned college under
the supervision of the concerned department HOD. Committee will check & ensure that the question
bank is prepared according to the guidelines. After that they will make a folder according to the
Program, Branch & semester/year wise, and would ensure that all question banks are available
according to the evaluation scheme. Hods will submit all folders to the CoE Office.
SECTION-A (Very Short Answer Type Questions)
UNIT-I
S.No. Question CO Bloom's
Taxonomy
a) In case of multiclass classification which activation function should L1
be used in the output layer in the neural network? Write down it’s
formula?
b) Differentiate between Feedforward and Back propagation? L1
c) What are activation functions and its use? L2
d) Explain Binary Step Activation Function? L2
e) Name one Linear and Non-Linear Activation Function? L3
f) Differentiate between Shallow Neural Network and Deep Neural L3
Network?
g) Why do we introduce hidden layer in the neural network? L4
h) Differentiate between Perceptron and Multi-Layer Perceptron? L4
i) What is the main limitation of a perceptron? L5
j) How does a perceptron learn to make accurate predictions? L5
UNIT-II

S.No. Question CO Bloom's

Taxonomy
a) Give one difference between lazy execution and eager execution? L1
b) In computational graph nodes represents the …………… and L1
edges represents the flow of ……….. form one to another.
c) In which year Tensorflow 2.0 was announced? L2
d) Do we need Placeholder in Tensorflow 2.0? L2
e) Are 3d tensors and 3d vectors are same? L3
f) What is Computational Graph? L3
g) Compare Machine Learning and Deep Learning? L4
h) In TensorFlow 1.0, which command can be used to run a L4
TensorFlow graph?
i) Draw a simple neural network which consist of an input layer(4 L5
nodes) ,one hidden layer(3 nodes) and one output layer(1 node).
Calculate total number of trainable parameter (weight and bias) in
that neural network.
j) Give any two real world example of 2d tensor. L5

UNIT-III
S.No. Question CO Bloom's
Taxonomy
a) What is Global minimum? Are Global minimum and local minimum L1
are same?
b) What is Gradient Descent? L1
c) Show the local minima, local maxima, Global minima and Global L2
maxima using a Graph?
d) Define Learning Rate? L2
e) Is time series data a 3D tensor? If yes than, give any two real world L3
example of time series data.
f) Write down the purpose of introducing the hidden layers in the L3
neural network.
g) How many layers does shallow feed forward neural network? Draw L4
a deep feedforward neural network.
h) What is Global Maxima? L4
i) What is learning rate in neural network? L5
j) What is Saddle point? L5

UNIT-IV
S.No. Question CO Bloom's
Taxonomy
a) Which of the following is true about momentum optimization for L1
finding minima of a function?
a. It uses an exponentially weighted moving average of past
gradients to update the parameters.
b. It always finds the global minimum of the function.
c. It requires the function to be convex.
d. It is slower than gradient descent for large datasets.
b) What is Keras? L1
c) Write down the name of any two Optimizers? L2
d) Write down name of any two organization which are currently using L2
Keras?
e) In how many ways Keras can build the model? L3
f) Why is there a need for keras? L3
g) In which layer of Neural network Sigmoid activation function should L4
be used?
h) Write down the update formula for weights and bias used in L4
Gradient Descent in Backward propagation
i) Write down the code to disable the eager execution in Tensorflow. L5
j) Which of the following is true about gradient descent optimization L5
for finding minima of a function?
a. It always finds the global minimum of the function.
b. It may get stuck at local minima.
 c. It only works for convex functions.
d. It does not require the function to be differentiable.

SECTION-B (Short Answer Type Questions)

UNIT-I
S.No. Question CO Bloom's
Taxonomy
a) What Is the Difference Between Epoch, Batch, and Iteration in L1
Deep Learning?
b) Compare Biological Neuron with Artificial Neuron? L1
c) Explain ReLU activation function with its advantages and L2
disadvantages?
d) Give the activation function required ? L2
I. Regression
II. Binary Classification
III. Multiclass Classification
IV. Convolutional Neural Network (CNN)
Recurrent Neural Network
e) What are the advantages of TensorFlow Session? L3
f) Explain the structure of ANN? L3
g) Compare Biological Neuron with Artificial Neuron? L4
h) What is the use of Bias and weights in Neural Network? L4
i) Write down some drawbacks of single layer perceptron L5
j) Do you think Deep learning is better than Machine learning? If so, L5
explain

UNIT-II
S.No. Question CO Bloom's
Taxonomy
a) What are constants, Variable and placeholder in Tensorflow L1
version1.0? Write down a program to add two variable using
Tensorflow version1.0.
b) What is optimization in neural network? Explain any 3 difference L1
between batch gradient descent and stochastic gradient descent.
c) What is Backward Propagation? How does it help to optimize the L2
value of weight and bias in neural network? Explain with an
example
d) What is Momentum Gradient descent? Explain the difference L2
between the Adagrad and Adadelta..
e) Explain computational graph and its benefits with an example. How L3
does tensorflow version 2.0 enhance deployment capabilities.
f) Is feature scaling required in decision tree or random forest L3
algorithm?
g) Which version of Tensorflow does not support Lazy execution? L4
h) What is activation function? Explain any three type of activation L4
function.
i) What are Global maxima and Global minima? How it is different, L5
form the local maxima and minima?
j) What is Keras? Explain its characteristics. L5
 UNIT-III
S.No. Question CO Bloom's
Taxonomy
a) What is Saddle point? How it is different from the local minima or L1
maxima?
b) What is learning rate in Neural Network? L1
c) Give any two real world example of 2D tensor? L2
d) Write down the code to disable the eager execution in Tensorflow. L2
e) If you have a dataset in which dependent variable has more than 2 L3
class so in that case can we use Linear perceptron to classify the
data? Justify your answer.
f) Explain information theory. L3
g) ] Write down difference between sequential and functional model. L4
h) Explain gradient descent. How to train a neural network using L4
forward propagation and backward propagation?
i) Explain the architecture of simple neural network. Draw a neat L5
diagram.
j) Explain the difference between the Adagrad and Adadelta.. L5

UNIT-IV
S.No. Question COURSE
OUTCOMES
a) Explain information theory. If the probability of an event is 0.6 than CO1
calculate the amount of information of that event.
b) Which is the simplest model of Keras? CO1
c) How do you avoid overfitting in deep learning? CO2
d) Explain information theory. If the probability of an event is 0.6 than CO2
calculate the amount of information of that event.
e) Write down some drawbacks of single layer perceptron. CO3
f) What are constants, Variable and placeholder in Tensorflow CO3
version1.0 ? Explain in detail with example
g) How do you train a Keras model, and what are the steps involved in CO4
training a neural network using Keras?
h) What is Momentum Gradient descent? What are some algorithms CO4
have been developed using the idea of using different learning rates
for different weights
i) What is sequential model in keras? CO5
j) Explain Keras and its features? CO5

SECTION-C [Descriptive Answer Type Questions / Case Study (for MBA COURSES only)]
UNIT-I
S.No. Question CO Bloom's
Taxonomy
a) Give the comparison between Deep Learning and Machine L1
Learning? Give any two applications of Deep Learning?
b) Explain types of data layers while working on data pipeline in L1
tensorflow?
c) Draw a simple neural network which consist of an input layer(4 L2
nodes) ,two hidden layer(3 nodes) and one output layer(2 node).
 Calculate total number of trainable parameter (weight and bias) in
that neural network.
d) Draw and explain shallow and deep artificial neural network? L2
e) Explain different type of activation function used in neural network. L3
f) What is Backward Propagation? How do you update the value of L3
Weights and bias In neural network?
g) Explain the concept of tensors in Machine learning and deep L4
learning along with the code.
h) How do you build Neural network using Keras? Explain it with an L4
example along with necessary code
i) Explain Tensorflow on the basis of computational graph? L5
j) What is Activation Function? Explain the types of activation L5
functions used in the Neural network

UNIT-II
S.No. Question CO Bloom's
Taxonomy
a) Explain Graph and Session and its relation to Neural Networks? L1
b) What are Tensors? Explain its different dimensions with examples? L1
c) What is Activation Function? Explain the types of activation L2
functions used in the Neural network.
d) Explain the concept of optimization in Neural network. Explain any L2
3 difference between batch gradient descent and stochastic
gradient descent.
e) What is Shallow feed forward network and deep feed forward L3
neural network.
f) Explain gradient descent. How to train a neural network using L3
forward propagation and backward propagation?
g) Write down the code to create a sequential model using Keras. L4
What is epochs in Neural network training.
h) Explain Tensorflow on the basis of computational graph? L4
i) Explain types of data layers while working on data pipeline in L5
tensorflow?
j) What are constants, Variable and placeholder in Tensorflow L5
version1.0 ? Explain in detail with example?

UNIT-III
S.N Question C Bloom'
o. O s
Taxono
my
a) What is Backward Propagation? Write down all the steps involved in L1
updating the value of Weights and bias using Gradient Descent along
with the equations in Backward propagation
b) Write short note on Gradient Descent ? L1
c) How to declare a variable and placeholder in tensorflow version1.0? L2
Write down the necessary code.
d) Explain the concept of saddle point and how to overcome it? L2
e) Write down the necessary steps for the transformation along with code L3
(Encoding, scaling, and filling null values) applied for the given Toy
Dataset. (No need to write data and the Output).
 Product Id Product Price State Purchase

001 Car 400 UP Yes

002 Bike 150 UP No

003 Bicycle 100 UK Yes

004 Car UK Yes

005 Bike 160 Delhi Yes

006 Bike 170 UP Yes

007 Car 500 UP No

008 Bicycle 150 UK No

009 Bicycle 160 Delhi No

010 Car 600 UK Yes

011 Car 650 UP No

012 Bicycle 200 Delhi No

013 Bike UK Yes

014 Bike 180 UK Yes

015 Car 700 UP Yes

f) Explain the concept of feature engineering. Why do we use drop out L3

layers in neural network?
g) Explain gradient descent. How to train a neural network using forward L4
propagation and backward propagation?
h) Explain Information Theory and its relevance in Artificial Intelligence? L4
i) Explain Loss functions in Deep Learning? L5
j) Explain in detail about the Gradient Descent. How the batch Gradient L5
descent is different form the Stochastic and mini batch Gradient
descent?

UNIT-IV
S.No. Question CO Bloom's
Taxonomy
a) What is Momentum Gradient descent? What are some algorithms L1
have been developed using the idea of using different learning
rates for different weights
b) Explain the complete layered structure of Keras API? L1
c) Explain how back propogation works in detail? L2
d) Explain Gradient Descent and its challenges in detail? L2
e) What is Gradient Descent and explain its three approaches? L3
f) What is Momentum Gradient descent? What are some key L3
differences between the Adadelta, Adagrad and Adam optimizer?