SavitribaiPhule Pune University

Fourth Year Computer Engineering (2019 Course)

 410251: Deep Learning

Teaching Scheme: Credit Examination Scheme:

TH: 03 Hours/Week 03 In-Sem (Paper): 30 Marks

End-Sem (Paper): 70 Marks

Prerequisite Courses: Machine Learning (410242)

Companion Course: Laboratory Practice V(410254)

Course Objectives:

To understand the basics of neural networks.

1. To understand the basics of neural networks.

2. Comparing different deep learning models.

3. To understand the Recurrent and Recursive nets in Deep Learning

4. To understand the basics of deep reinforcement Learning models.

5. To analyze Types of Networks.

6. To Describe Reinforcement Learning

Course Outcomes:

On completion of the course, student will be able to–

CO1.: Understand the basics of Deep Learning and apply the tools to implement deep learning
applications

CO2.: Evaluate the performance of deep learning models (e.g., with respect to the bias-variance
trade- off, overfitting and underfitting, estimation of test error).

CO3.: To apply the technique of Convolution (CNN) and Recurrent Neural Network (RNN) for
implementing Deep Learning models
Unit I: Foundations of Deep learning (07 Hours)
What is machine learning and deep learning?, Supervised and Unsupervised Learning, bias
variance tradeoff, hyper parameters, under/overfitting regularization, Limitations of machine
learning, History of deep learning, Advantage and challenges of deep learning. Learning
representations from data , Understanding how deep learning works in three figures, Common
Architectural Principles of Deep Network, Architecture Design, Applications of Deep Learning,
Introduction and use of popular industry tools such as Tensor FLow, Keras, PyTorch, Caffe,
Shogun.
Unit II: Deep Neural Networks (DNNs) (07 Hours)
Introduction to Neural Networks :The Biological Neuron, The Perceptron, Multilayer Feed-
Forward Networks , Training Neural Networks :Backpropagation and Forward propagation
Activation Functions :Linear ,Sigmoid, Tannh, Hard Tanh, Softmax, Rectified Linear, Loss
Functions :Loss Function Notation , Loss Functions for Regression , Loss Functions for
Classification, Loss Functions for Reconstruction, Hyperparameters : Learning Rate,
Regularization, Momentum, Sparsity, Deep Feedforward Networks – Example of Ex OR, Hidden
Units, cost functions, error backpropagation, Gradient-Based Learning, Implementing Gradient
Descent, vanishing and Exploding gradient descent, Sentiment Analysis, Deep Learning with
Pytorch, Jupyter, colab.

Unit III: Convolution Neural Network (CNN) (07 Hours)

Introduction, CNN architecture overview, The Basic Structure of a Convolutional Network-

Padding, Strides, Typical Settings, the ReLU layer, Pooling, Fully Connected Layers, The
Interleaving between Layers, Local Response Normalization, Training a Convolutional Network

Unit IV: Recurrent Neural Network (CNN) (07 Hours)

Recurrent and Recursive Nets: Unfolding Computational Graphs, Recurrent Neural Networks,
Bidirectional RNNs, Encoder-Decoder Sequence-to-Sequence Architectures, Deep Recurrent
Networks, Recursive Neural Networks, The Challenge of Long-Term Dependencies, Echo State
Networks, Leaky Units and Other Strategies for Multiple Time Scales, The Long Short-Term
Memory and Other Gated RNNs, Optimization for Long-Term Dependencies, Explicit Memory.
Practical Methodology: Performance Metrics, Default Baseline Models, Determining Whether to
Gather More Data, Selecting Hyper parameters

Unit V: Deep Generative Models (07 Hours)

 Introduction to deep generative model, Boltzmann Machine, Deep Belief Networks, Generative
adversarial network (GAN), discriminator network, generator network, types of GAN, Applications
of GAN networks

Unit VI: Reinforcement Learning (07 Hours)

Introduction of deep reinforcement learning, Markov Decision Process, basic framework of

reinforcement learning, challenges of reinforcement learning, Dynamic programming algorithms for
reinforcement learning,Q Learning and Deep Q-Networks, Deep Q recurrent networks, Simple
reinforcement learning for Tic-Tac-Toe.

Text Books:
1. Good fellow, I., Bengio, Y.,,Courville, A, ―Deep Learning, MIT Press, 2016.
2. Josh Patterson & Adam Gibson, ―Deep Learning
3. Charu Agarwal, ―Neural Networks and deep learning, A textbook
4. Nikhil Buduma, ―Fundamentals of Deep Learning, SPD
5. Francois chollet, ―Deep Learning with Python
References Books:
1. Richard S. Sutton and Andrew G. Barto, ―Reinforcement Learning: An Introduction
2. by SethWeidman, ―Deep Learning from Scratch: Building with Python from First Principles
O‘Reily
3. Francois Duval, ―Deep Learning for Beginners, Practical Guide with Python and Tensorflow
E-Books:
1. http://csis.pace.edu/ctappert/cs855-18fall/DeepLearningPractitionersApproach.pdf .
2. https://www.dkriesel.com/_media/science/neuronalenetze-en-zeta2-1col-dkrieselcom.pdf
 Department of Computer Engineering
TEACHING PLAN
Academic Year: 2023-24(Semester: VIII)
Course Title: Deep Learning Subject Code: Class: B.E. Division: A & B
(DL) 410251
Term: II Date of commencement of classes: Date of conclusion of teaching:
11/12/2023 19/04/2024
Lecture Examination Scheme
Schedule:
3Hrs/ Week Theory:100 M Term Work Practical Oral
In Sem:30 M (1Hr)
End Sem:70 M -
(2Hrs.30 min.) 50 50
Subject Mrs. Sonal Previous 3 Years 2021-22 2022-23 2023-24
Teacher Fatangare & Mrs University Result
Jyoti Ragatwan NA
 UNIT – I: Foundations of Deep learning (07 Hours)
What is machine learning and deep learning?, Supervised and Unsupervised Learning, bias variance tradeoff,
hyper parameters, under/overfitting regularization, Limitations of machine learning, History of deep learning,
Advantage and challenges of deep learning. Learning representations from data , Understanding how deep
learning works in three figures, Common Architectural Principles of Deep Network, Architecture Design,
Applications of Deep Learning, Introduction and use of popular industry tools such as Tensor FLow, Keras,
PyTorch, Caffe, Shogun.
PLAN ACTUAL
Lect. Date Topics Date Topics covered Reasons
No. for Deviation
1 What is machine learning and
deep learning?,Supervised
and Unsupervised Learning,

2 bias variance tradeoff, hyper

parameters,
3 under/overfitting
regularization,
Limitations of machine
learning,
4 History of deep learning,
Advantage and challenges of
deep learning.
5 Learning representations
from data , Understanding
how deep learning works in
three figures,
6 Common Architectural
Principles of Deep Network,
Architecture Design,
Applications of Deep
learning,
7 Introduction and use of
popular industry tools such as
TensorFLow,Keras, PyTorch,
Caffe, Shogun.
Make up Classes

Contents Beyond syllabus

UNIT –II: Deep Neural Networks (DNNs) (07 Hours)

Introduction to Neural Networks :The Biological Neuron, The Perceptron, Multilayer Feed-Forward Networks ,
Training Neural Networks :Backpropagation and Forward propagation Activation Functions :Linear ,Sigmoid,
Tannh, Hard Tanh, Softmax, Rectified Linear, Loss Functions :Loss Function Notation , Loss Functions for
Regression , Loss Functions for Classification, Loss Functions for Reconstruction, Hyperparameters : Learning
Rate, Regularization, Momentum, Sparsity, Deep Feedforward Networks – Example of Ex OR, Hidden Units,
cost functions, error backpropagation, Gradient-Based Learning, Implementing Gradient Descent, vanishing and
Exploding gradient descent, Sentiment Analysis, Deep Learning with Pytorch, Jupyter, colab.
PLAN ACTUAL
Lect. Date Topics Date Topics covered Reasons
No. for Deviation
8 Introduction to Neural
Networks :The Biological
Neuron, The Perceptron,
Multilayer Feed-Forward
Networks
9 Training Neural
Networks :Backpropagation
and Forward propagation
10 Activation
Functions :Linear ,Sigmoid,
Tannh, Hard Tanh, Softmax,
Rectified Linear,
11 Loss Functions :Loss
Function Notation , Loss
Functions for Regression ,
Loss Functions for
Classification, Loss
Functions for Reconstruction,
12 Hyperparameters : Learning
Rate, Regularization,
Momentum, Sparsity, Deep
Feedforward Networks –
Example of Ex OR, Hidden
Units, cost functions, error
backpropagation
13 Gradient-Based Learning,
Implementing Gradient
Descent,
14 vanishing and Exploding
gradient Make up Classes

Contents Beyond syllabus

UNIT –III: Convolution Neural Network (CNN) (07 Hours)

Introduction, CNN architecture overview, The Basic Structure of a Convolutional Network- Padding, Strides,
Typical Settings, the ReLU layer, Pooling, Fully Connected Layers, The Interleaving between Layers, Local
Response Normalization, Training a Convolutional Network

PLAN ACTUAL
Lect. Date Topics Date Topics covered Reasons
No. for Deviation
15 Introduction, CNN
architecture overview,
16 The Basic Structure of a
Convolutional Network-
Padding,
17 Strides, Typical Settings, the
ReLU layer, Pooling,
18 Fully Connected Layers

19 The Interleaving between

Layers,
20 Local Response
Normalization,
21 Training a Convolutional
Network
Make up Classes

Contents Beyond syllabus

UNIT –IV: Recurrent Neural Network(CNN) (07 Hours)

Recurrent and Recursive Nets: Unfolding Computational Graphs, Recurrent Neural Networks, Bidirectional
RNNs, Encoder-Decoder Sequence-to-Sequence Architectures, Deep Recurrent Networks, Recursive Neural
Networks, The Challenge of Long-Term Dependencies, Echo State Networks, Leaky Units and Other Strategies
for Multiple Time Scales, The Long Short-Term Memory and Other Gated RNNs, Optimization for Long-Term
Dependencies, Explicit Memory. Practical Methodology: Performance Metrics, Default Baseline Models,
Determining Whether to Gather More Data, Selecting Hyper parameters
PLAN ACTUAL
Lect. Date Topics Date Topics covered Reasons
No. for Deviation
22 Recurrent and Recursive
Nets: Unfolding
Computational Graphs,
Recurrent Neural Networks
23 Bidirectional RNNs,
Encoder-Decoder Sequence-
to-Sequence Architectures,
 24 Deep Recurrent Networks,
Recursive Neural Networks,
The Challenge of Long-Term
Dependencies,
25 Echo State Networks, Leaky
Units and Other Strategies for
Multiple Time Scales,
26 The Long Short-Term
Memory and Other Gated
RNNs, Optimization for
Long-Term Dependencies
27 Explicit Memory. Practical
Methodology: Performance
Metrics, Default Baseline
Models,
28 Determining Whether to
Gather More Data, Selecting
Hyper parameters
Make up Classes

Contents Beyond syllabus

UNIT –V: Deep Generative Models (07 Hours)

Introduction to deep generative model, Boltzmann Machine, Deep Belief Networks, Generative adversarial
network (GAN), discriminator network, generator network, types of GAN, Applications of GAN networks
PLAN ACTUAL
Lect. Date Topics Date Topics covered Reasons
No. for Deviation
29 Introduction to deep
generative model,
30 Boltzmann Machine

31 Deep Belief Networks

32 Generative adversarial
network (GAN),
33 Generative adversarial
network (GAN),
34 discriminator network,
generator network,
35 types of GAN, Applications
of GAN networks
Make up Classes
 Contents Beyond syllabus

UNIT –VI: Reinforcement Learning (07 Hours)

Introduction of deep reinforcement learning, Markov Decision Process, basic framework of reinforcement
learning, challenges of reinforcement learning, Dynamic programming algorithms for reinforcement learning,Q
Learning and Deep Q-Networks, Deep Q recurrent networks, Simple reinforcement learning for Tic-Tac-Toe

PLAN ACTUAL
Lect. Date Topics Date Topics covered Reasons
No. for Deviation
36 Introduction of deep
reinforcement learning,
37 Markov Decision Process,
basic framework of
reinforcement learning
38 challenges of reinforcement
learning, Dynamic
programming algorithms for
reinforcement learning
39 Q Learning and Deep Q-
Networks, Deep Q recurrent
networks
40 Q Learning and Deep Q-
Networks, Deep Q recurrent
networks
41 Simple reinforcement
learning for Tic-Tac-Toe.
42 Introduction of deep
reinforcement learning,
Make up Classes

Contents Beyond syllabus

SUMMARY

No. of lectures allotted by university 42

Total no. of lectures conducted

Percentage of syllabus covered
Total no. of makeup classes
 UNIT WISE QUESTION BANK
Subject: Deep Learning Course: BE (2019 Pattern)

UNIT 1: Foundations of Deep learning

1. What are Deep stacking networks?

2. What is machine learning and deep learning?

3. Explain Deep Learning Architecture

4. Write advantages and challenges of deep learning.

5. Explain Convolution neural networks

6. Explain Recurrent neural networks

7. How deep learning works in three figures?

8. Short note on applications of Deep learning,

9. Explain Supervised and Unsupervised Learning

10. Short Note on LSTM networks & GRU networks

 11. Explain Restricted Boltzmann Machines

12. Explain Bias-Variance Tradeoff

13. What is Over and underfitting.

14. Write note on:

i. Keras

ii. Pytorch

iii. Caffe

iv. Shogun

v. Tensorflow

UNIT 2: Deep Neural Networks (DNNs)

1. Explain Vanishing Gradient Descent.

2. Write a short note on Sigmoid & Softmax.

3. What is the Biological Neuron & Perceptron?

4. Explain Loss function for Classification .

5. Explain Deep Feed forward Networks .

6. Write a short note on Linear & Tanh.

7. Explain Multilayer Feed-Forward Networks.

8. What is Activation Functions.?Explain any two.

9. Why does the vanishing or exploding gradients problem happen?

10. What are the methods for tuning hyperparameters

11. Explain Gradient Descent.

12. What is the Loss function? Explain Loss function for Regression

13. Draw and Explain the architecture of SOM ( Self Organizing Maps)

14. Draw and Explain the architecture of Single layer Feedforward Networks
 15.What are loss functions?

16. Explain in detail Hyper parameter.

17. Write the example of sentiment analysis in short.

UNIT 3: Convolution Neural Network (CNN)

1. Write short notes on CNN

2. Explain the different layers in CNN.

3. Explain the terms “Valid Padding” and “Same Padding” in CNN.

4. What are the different types of Pooling? Explain their characteristics.

5. Explain padding and strides with examples.

6. What is Stride? What is the effect of high Stride on the feature map?

7. Explain the role of the flattening layer in CNN.

8. What is RELU Layer?

9. Explain Importance of pooling?

10. Explain Local Response Normalization.

11. Explain how to train CNN with examples.

12. Explain Causal Planning.

13. Why do we prefer Convolution Neural networks (CNN) over Artificial Neural networks (ANN)

for image data as input.

 14. Briefly explain the two major steps of CNN i.e, Feature Learning and Classification

UNIT 4: Convolution Neural Network (CNN)

1. What is a recurrent net?

2. Write a short note on a recursive net?

3. Why do we need a Recurrent Neural Network?

4. Explain in short unfolding of computational graphs.

5. What are recurrent neural networks?

6. Write a short note on Encoder-Decoder Sequence to Sequence.

7. Explain Deep Recurrent Networks.

8. Explain the challenge of Long Term Dependencies.

9. Explain Echo State Network.

10. Explain RNN.

11. Explain with example LSTM

12. Discuss the importance of pooling layer in Convolutional Neural Networks (CNN)

13. Describe the shape or dimension of input data for a recurrent neural networks

14. Describe zero padding strategies used in Convolution.

 15. Explain the role of filter, padding, and strides in a convolution layer of a CNN with an example

UNIT 5: Deep Generative Models

1. Write a short note on GAN model

2. What is a Boltzmann machine?

3. What is deep Belief Network?

4. What are GANs?

5. What are applications of GAN?

6. Explain types of GAN.

7. Explain Restricted Boltzmann Machine

8. Explain Advantages & disadvantages of Restricted Boltzmann Machine

9. Explain the types of Deep Generative Model

10. Explain Autoencoders

UNIT 6: Reinforcement Learning
1. What is deep reinforcement learning?

2. What is the Markov Decision Process?

3. What is the basic framework of reinforcement learning?

4. What are the challenges faced in reinforcement learning?

5. Explain the challenges of Reinforcement Learning

6. Explain Q Learning vs Deep Q Learning

7. Define the Q function. How to take action given a Q function

8. Write a short note on Deep Q recurrent Network

9. Explain Model-based vs Model-free learning algorithms

10. Explain Dynamic Programming for of Reinforcement Learning

11. What are the Elements of Reinforcement Learning

12. Explain the Q-Learning Application

 Department of Computer Engineering
A.Y. 2022-23 (Semester-II)
UNIT TEST-I EXAM
Subject: Deep Learning Date: / / 2023 Time: 1 Hr
Maximum Marks: 30

Instructions to Candidates:
1. Attempt Questions Q.1 OR Q.2, Q.3 OR Q.4.
2. Neat diagrams must be drawn wherever necessary.
3. Assume suitable data, if necessary.

Q. No. Questions Marks CO PO

1 PO1
A. What are Deep stacking networks? 5 CO1
PO5

PO1
B. Explain Convolution neural network 5 CO1
PO3

C. Write advantages and challenges of deep learning. 5 CO1 PO3

OR
 2 PO1
A. Short Note on LSTM networks & GRU networks 5 CO1
PO5

PO1
B. Short note on applications of Deep learning, 5 CO1
PO12

C. Explain Bias-Variance Tradeoff 5 CO1 PO4

3 Why does the vanishing or exploding gradients

A. 5 CO2 PO2
problem happen?

Draw and Explain the architecture of SOM ( Self PO3

B. 6 CO2
Organizing Maps) PO4

C. What is the Biological Neuron & Perceptron, 4 CO2 PO2

OR

4 A. Explain Loss function for Classification 5 CO2 PO12

B. Explain Gradient Descent. 5 CO2 PO2

C. Explain Multilayer Feed-Forward Networks. 5 CO2 PO3

Department of Computer Engineering

A.Y. 2023-24 (Semester-II)

UNIT TEST-I EXAM
Subject: Deep Learning Date: / / 2024 Time: 1 Hr
Maximum Marks: 30

Instructions to Candidates:
1. Attempt Questions Q.1 OR Q.2, Q.3 OR Q.4.
2. Neat diagrams must be drawn wherever necessary.
3. Assume suitable data, if necessary.

Q. No. Questions Marks CO PO

 1 A. 5 CO1 PO1
What are Deep stacking networks?

B. 5 CO1 PO2
Explain Convolution neural networks

C. 5 CO1 PO5
Explain Recurrent neural networks

OR

2 A. Short Note on LSTM networks & GRU networks 6 CO1 PO12

B. Explain Restricted Boltzmann Machines 4 CO1 PO1

C. Explain Bias-Variance Tradeoff 5 CO1 PO2

3 A. Explain Multilayer Feed-Forward Networks 5 CO2 PO1

B. 5 CO2 PO2
What is Activation Functions.?Explain any two

C. Why does the vanishing or exploding gradients problem 5 CO2 PO2

happen?

OR

4 A. 5 CO2 PO5
What are the methods for tuning hyperparameters

CO2 PO7
B. Explain Gradient Descent 5

CO2
C. What is the Loss function? Explain Loss function for 5 PO12
Regression

Department of Computer Engineering

A.Y. 2023-24 (Semester-II)
UNIT TEST-II EXAM
Subject: Deep Learning Date: / / 2024 Time: 1 Hr
Maximum Marks: 30

Instructions to Candidates:
1. Attempt Questions Q.1 OR Q.2, Q.3 OR Q.4.
2. Neat diagrams must be drawn wherever necessary.
3. Assume suitable data, if necessary.

Q.
Questions Marks CO PO
No.

1 Explain padding and strides with examples.

A. 5 CO3 PO1

Explain the terms “Valid Padding” and “Same Padding” in

B. 5 CO3 PO2
CNN.

What are the different types of Pooling? Explain their

C. 5 CO3 PO3
characteristics

OR

2 PO3
A. Explain how to train CNN with example 5 CO3
PO5

B. Explain Local Response Normalization 5 CO3 PO4

Explain RELU Layer?

C. 5 CO3 PO1

3 A. Explain with example LSTM 5 CO4 PO2

Describe the shape or dimension of input data for a recurrent

B. 5 CO4 PO2
neural networks

Explain the role of filter, padding, and strides in a

C. 5 CO4 PO3
convolution layer of a CNN with an example

OR

4 A. Explain Deep Recurrent Networks 5 CO4 PO1

B. Explain in short unfolding of computational graphs. 5 CO4 PO3

Write a short note on Encoder-Decoder Sequence to PO2

C. 5 CO4
Sequence. PO3
 Department of Computer Engineering

A.Y. 2022-23 (Semester-II)

UNIT TEST-II EXAM
Subject: Deep Learning Date: / / 2023 Time: 1 Hr
Maximum Marks: 30

Instructions to Candidates:
1. Attempt Questions Q.1 OR Q.2, Q.3 OR Q.4.
2. Neat diagrams must be drawn wherever necessary.
3. Assume suitable data, if necessary.

Q. No. Questions Marks CO PO

1 A. Explain the different layers in CNN 5 CO3 PO1

B. Explain the terms “Valid Padding” and “Same Padding” in 5 CO3 PO2
CNN.

C. What are the different types of Pooling? Explain their 5 CO3 PO3
characteristics
OR
2 A. Explain padding and strides with examples 5 CO3 PO4

B. Explain Local Response Normalization 5 CO3 PO5

C. Explain the role of the flattening layer in CNN. 5 CO3 PO7

3 A. Write a short note on a recursive net? 5 CO4 PO1
B. Explain in short unfolding of computational graphs. 5 CO4 PO2

C. Explain the role of filter, padding, and strides in a 5 CO4 PO3

convolution layer of a CNN with an example
OR
4 A. Describe zero padding strategies used in Convolution 5 CO4 PO4

B. Describe the shape or dimension of input data for a recurrent 5 CO4 PO7
neural networks

C. Write a short note on Encoder-Decoder Sequence to 5 CO4 PO1

Sequence. 2
 Department of Computer Engineering
A.Y. 2023-24 (Semester-II)
PRELIM EXAMINATION
1
Subject: Deep Learning Date: / / 2024 Time: 2( ) Hr
2
Maximum Marks: 70

Instructions to Candidates:
1. Attempt Questions Q.3 OR Q.4, Q.5 OR Q.4, Q.5 OR Q.6 Q. 7 OR Q.8
2. Neat diagrams must be drawn wherever necessary.
3. Assume suitable data, if necessary.

Q. No. Questions Marks CO PO

1
A. Explain importance and different types of pooling. 6 CO3
PO3

Explain the terms “Valid Padding” and “Same Padding” in

B. 6 CO3 PO2
CNN.

C. Explain different Activation Function. 8 CO3 PO2

OR

2 What is Stride? What is the effect of high Stride on the feature

A. 6 CO3 PO4
map

B. Explain Local Response Normalization 6 CO3 PO5

C. Explain how to train CNN with examples. 8 CO3 PO3

3 A. Explain bidirectional RNNs. 6 CO4 PO2

Describe the shape or dimension of input data for a recurrent

B. 5 CO4 PO1
neural networks

C. Explain the challenge of Long Term Dependencies 5 CO4 PO2

OR

4 Describe zero padding strategies used in Convolution

A. 6 CO4 PO1

B. Explain Echo State Network. 5 CO4 PO2

 C. Explain Performance metrics used for classification 5 CO4 PO2

5 A. How does deep Belief Network work? 6 CO5 PO3

B. Explain types of GAN 5 CO5 PO2

Explain Advantages & disadvantages of Restricted Boltzmann

C. 5 CO5 PO2
Machine

OR

6 A. Explain Component Boltzmann Machine. 6 CO5 PO2

PO6
B. What are applications of GAN? 5 CO5
PO7

C. Explain Autoencoders. 5 CO5 PO2

7 A. What are the Elements of Reinforcement Learning 8 CO6 PO3

B. Define the Q function. How to take action given a Q function 6 CO6 P04

C. What is Markov Decision Process 4 CO6 PO4

OR

8 A. Explain Dynamic Programming for of Reinforcement Learning 8 CO6 PO1

PO4,
B. Explain Q Learning vs Deep Q Learning 6 CO6
PO5

C. What is Deep reinforcement learning? 4 CO6 PO3

 Department of Computer Engineering
A.Y. 2023-24 (Semester-II)
PRELIM EXAMINATION
1
Subject: Deep Learning Date: / / 2024 Time: 2( ) Hr
2
Maximum Marks: 70

Instructions to Candidates:
1. Attempt Questions Q.1 OR Q.2, Q.3 OR Q.4 Q.5 OR Q.6 Q. 7 OR Q.8
2. Neat diagrams must be drawn wherever necessary.
3. Assume suitable data, if necessary.

Q. No. Questions Marks CO PO

1 A. Explain the terms “Valid Padding” and “Same Padding” in 6 CO3 PO2
CNN.
B. Explain Importance of pooling? 6 CO3 PO3

C. Explain different layer of CNN 8 CO3 PO1

OR

2 A. Explain Local Response Normalization. 6 CO3 PO5

B. What are the different types of Pooling? Explain their 6 CO3 PO3
characteristics.

C. Explain how to train CNN with example 8 CO3 PO3

3 A. Why do we need a Recurrent Neural Network? 5 CO4 PO4
 B. Explain in short unfolding of computational graphs. 5 CO4 PO3

C. Explain with example LSTM 6 CO4 PO1

OR
4 A. What is a recurrent net? 5 CO4 PO1
B. Write a short note on Encoder-Decoder Sequence to Sequence. 5 CO4 PO2

C. Explain Echo State Network. 6 CO4 PO5

5 A. What do you mean by generative model. 5 CO5 PO5
B. Explain types of GAN. 5 CO5 PO2

C. Write a short note on GAN model 6 CO5 PO1

OR
6 A. What is a Boltzmann machine? 5 CO5 PO2
B. What is deep Belief Network? 5 CO5 PO1

C. What are applications of GAN? 6 CO5 PO6,

PO7
7 A. What is the Markov Decision Process? 6 CO6 PO3
B. What is the basic framework of reinforcement learning? 6 CO6 PO4

C. Write a short note on Deep Q recurrent Network 6 CO6 PO1

OR
8 A. What is deep reinforcement learning? 6 CO6 PO1
B. What are the challenges faced in reinforcement learning? 6 CO6 PO5

C. Explain Q Learning vs. Deep Q Learning 6 CO6 PO4