SIR ISSAC NEWTON COLLEGE

OF
ENGINEERING AND TECHNOLOGY

DEPARTMENT OF
ARTIFICIAL INTELLIGENCE
AND DATA SCIENCE

SUBJECT CODE: AD3501

SUBJECT NAME: DEEP LEARNING

QUESTION BANK
BrainKart lOMoARcPSD|453 335 8

DEPARTMENT OF ARTIFICIAL INTELLIGENCE AND DATA SCIENCE

SUBJECT :AD3501 DEEP LEARNING

SEM/YEAR : V/III
UNIT - I: INTRODUCTION TO MACHINE LEARNING BASICS
Linear Algebra: Scalars -- Vectors -- Matrices and tensors; Probability Distributions -- Gradient-based
Optimization – Machine Learning Basics: Capacity -- Overfitting and underfitting --Hyperparameters
and validation sets -- Estimators -- Bias and variance -- Stochastic gradient descent -- Challenges
Motivating deep learning; Deep Networks: Deep feedforward networks; Regularization -- Optimization.
PART – A

Q.N Question BT Competence

o Level
1 What is Deep Learning? BTL-1 Remember
2 Differentiate scalar and vector. BTL-2 Understand
3 Compare Deep Learning and rule based system. BTL-4 Analyze
4 Give example of a feed forward neural network. BTL-2 Understand
5 List out the special kind of matrices. BTL-1 Remember
6 What is representation learning? BTL-1 Remember
7 Compare Deep Learning with Machine learning. BTL-5 Evaluate
8 State the Bayes rule. BTL-1 Remember
9 Point out different set of layers in Feed forward networks. BTL-4 Analyze
10 What is Bias? BTL-1 Remember
11 Give the drawback of rectified linear units. BTL-2 Understand
12 Give Venn diagram for Deep Learning. BTL-2 Understand
13 Specify the formula for conditional probability. BTL-6 Create
14 Illustrate support vector machine. BTL-3 Apply
15 Show the formula for Variance. BTL-3 Apply
16 What is Deep Feed Forward networks? BTL-1 Remember
17 List reasonably common hidden unit types. BTL-1 Remember
Point out the default activation function for modern neural networks. BTL-4 Analyze
18
19 Probability theory is a fundamental tool of many disciplines of BTL-5 Evaluate
science and engineering. Justify.

2
BrainKart lOMoARcPSD|453 335 8

20 Develop a formula for unbiased sample covariance matrix associated with BTL-6 Create
an m x n dimensional matrix X. Assume E[x] =0.
21 Analyze about Tensor. BTL-4 Analyze
22 Illustrate identity matrix? BTL-3 Apply
23 Discuss about matrix inverse. BTL-2 Understand
24 Assess linear dependence of vectors. BTL-5 Evaluate

PART – B
i. What is Deep Learning? (3)
1 ii. Describe how different parts of an Artificial Intelligence system BTL-1 Remember
relate to each other within different AI disciplines in detail with
diagram.(10)
2 Describe how deep learning is a kind of representation learning with the BTL-1 Remember
Venn diagram. (13)
3 List and explain the historical trends in Deep Learning. BTL-1 Remember

i..Discuss about scalars.(7)

4 BTL-2 Understand
ii.Give detail description of vectors. (6)
i.Give the Difference between deep learning and machine
5 BTL-2 Understand
learning.(7)
ii.Give the various concepts of probability. (6)
6 i.Demonstrate linear dependence and independence of vectors.(7) BTL-3 Analyze
ii.Explain span of vectors. (6)
7 Analyze and write short notes on the following. BTL-4 Analyze
i. Vectors. (6)
ii. Matrices.(7)
i.Give Softmax units for Multinoulli Output Distributions. (7) BTL-2 Understand
8 ii.Discuss about Hidden Units. (6)
9 i.Describe Back Propagation algorithm. (7) BTL-1 Remember
ii.Explain regularization for deep learning. (6)
Extrapolate conditional probability and Develop a summary of various
10 BTL-6 Create
common probability distribution. (13)
11 Describe Stochastic Gradient Descent. (13) BTL-1 Remember
12 i. Illustrate the importance of principal components analysis. (6) BTL-3 Analyze
ii. Explain support vector machines in detail. (7)
13 Explain supervised learning algorithm. (13) BTL-4 Apply
14 Discuss unsupervised learning algorithm. (13) BTL-2 Understand
15 Explain output units of feed forward networks. (13) BTL-3 Apply
16 Explain Probability Mass function and Probability Density function. (13) BTL-3 Analyze
17 Explain cost function ingredient based learning. (6) BTL-3 Apply

3
BrainKart lOMoARcPSD|453 335 8

Explain learning conditional distributions with maximum likelihood. (7)

PART – C
1 Develop short notes on following with respect to deep learning BTL-6 Create
withexamples.
i) Scalar and Vectors. (6)
ii) Matrices. (7)
Assess the following with respect to deep learning examples.
2 i) Random Variables. (6) BTL-5 Evaluate
ii) Probability. (7)
3 Develop a Deep Feed forward network and explain. (15) BTL-6 Create

4 Assess the routines to implement forward propagation computation.

(15) BTL-5 Evaluate
5 Assess the difference between linear models and neural networks. (15) BTL-5 Evaluate
UNIT - II: CONVOLUTIONAL NETWORKS
UNIT II CONVOLUTIONAL NEURAL NETWORKS Convolution Operation -- Sparse Interactions -- Parameter
Sharing – Equivariance -- Pooling -- Convolution Variants: Strided -- Tiled -- Transposed and dilated convolutions;
CNN Learning: Nonlinearity Functions -- Loss Functions -- Regularization -- Optimizers --Gradient Computation.

Q No Questions BT Competence
Level
PART – A
1 An essential feature of any convolutional network implementation is the
BTL-5 Evaluate
ability to implicitly zero-pad the input V. Justify
2 The output layer of convolutional network is usually relatively inexpensive
BTL-5 Evaluate
to learning layer. Justify.
3 What is convolutional networks? BTL-1 Remember
4 Create a chart that demonstrates convolution with a stride. BTL-6 Create
5 How pooling handles inputs of varying size? BTL-4 Analyze
6 What is meant by convolution? BTL-1 Understand
7 List three important ideas that help to improve a machine learning system. BTL-1 Remember
8 What is unshared convolution? BTL-2 Understand
9 Define primary visual cortex. BTL-1 Remember
10 How to reduce the cost of convolutional network training? BTL-2 Understand
11 Simulate the idea behind reverse correlation. BTL-6 Create
12 Discuss about parameter sharing in neural network. BTL-2 Understand
13 Give three properties of V1 that a convolutional network layer is designed
BTL-2 Understand
to capture.
14 Explain feature map. BTL-4 Analyze
Explain how a convolutional layer have a property called equivariance to
15 BTL-3 Apply
translation?

4
BrainKart lOMoARcPSD|453 335 8

16 List three stages of a convolutional network. BTL-1 Remember

17 List out various formats of data that can be used with convolutional BTL-1 Remember
networks.
18 Illustrate pooling stage in convolutional network. BTL-3 Apply
Differentiate complex layer terminology and simple layer terminology in
19 BTL-4 Analyze
convolutional network.
20 Show three basic strategies for obtaining convolution kernels without BTL-3 Apply
supervised training.
21 Give example for convolution. BTL-2 Understand
22 Illustrate reverse correlation. BTL-3 Apply
23 Explain complex layer terminology. BTL-4 Analyze
24 Examine equivariance to translation. BTL-5 Evaluate
PART – B
1 Write an example function for Convolution operation and explain in detail. BTL-1 Remember
(13)
2 Explain the following with suitable diagram. BTL-4 Analyze
i. Sparse interactions. (6)
ii. Parameter sharing. (7)
3 Describe Pooling with suitable example. (13) BTL-1 Remember
4 Write an expression for Unshared convolution with explanation and BTL-1 Remember
explain Tiled convolution.(13)
5 Discuss in detail the variants of the Basic Convolution Function. (13) BTL-2 Understand
6 Construct an architecture that show complex layer terminology and BTL-5 Evaluate
Simple layer terminology in convolutional neural network.
7 Discuss local connections, convolution and full connections with diagram? BTL-2 Understand
(13)
8 Develop a table with examples of different formats of data that can be BTL-6 Create
usedwith convolutional networks. (13)
9 Describe in detail about the following. BTL-1 Remember
i. Parameter Sharing. (7)
ii. Equivariant representation. (6)
10 Differentiate locally connected layers, tiled convolution and BTL-4 Analyze
standardconvolution with suitable examples and diagram. (13)
11 i. Write short notes Max Pooling. (6) BTL-2 Remember
ii. Explain Pooling with down sampling. (7)
12 Explain random or Unsupervised Features.(13) BTL-4 Analyze
13 Illustrate unshared convolution with suitable examples. (13) BTL-3 Apply
14 i. Show three properties of V1 that a convolutional network layer BTL-3 Apply
is designed to capture. (6)
ii. Prove the working learned invariances with necessary
exampleand diagram. (7)
15 Discuss parameter sharing. (13) BTL-2 Understand

5
BrainKart lOMoARcPSD|453 335 8

16 Illustrate Equivariant representation. (13) BTL-3 Analyze

17 Evaluate the working learned invariances with necessary example and BTL-5 Evaluate
diagram. (13)
PART – C
1 Construct a graphical demonstration for sparse connectivity and explain it BTL-5 Evaluate
in detail. (15)
2 Create a graphical demonstration for parameter sharing and explain it in BTL-6 Create
detail. (15)
3 Evaluate variants of the basic convolution function. (15) BTL-5 Evaluate
4 Construct a convolutional network to demonstrate the effect of zero BTL-6 Create
padding on network size.
Explain Neuro scientific basis for Convolutional Networks. (15)
5 Create a table with examples of different formats of data that can be used BTL-6 Create
with convolutional networks. (15)
UNIT III - RECURRENT NEURAL NETWORKS

. Unfolding Graphs -- RNN Design Patterns: Acceptor -- Encoder --Transducer; Gradient Computation -- Sequence
Modeling Conditioned on Contexts -- Bidirectional RNN -- Sequence to Sequence RNN – Deep Recurrent Networks --
Recursive Neural Networks -- Long Term Dependencies; Leaky Units: Skip connections and dropouts; Gated
Architecture: LSTM.

PART – A
Q.No Questions BT Competence
Level
1 What is Recurrent Neural Networks? BTL-1 Remember
2 What is Encoder? BTL-1 Remember
3 Give the blocks of decomposition of computation of most Recurrent Neural BTL-2 Understand
Networks.
4 What is Bidirectional Recurrent Neural Networks? BTL-1 Remember
5 Give the advantage of recursive nets over recurrent nets. BTL-2 Understand
6 What is decoder? BTL-1 Remember
7 Describe Recursive Neural Networks. BTL-1 Remember
8 Predict the concept of gated RNNs. BTL-2 Understand

9 Compare echo state network and liquid state machines. BTL-4 Analyze
10 Distinguish content based addressing and location based addressing in BTL-2 Understand
memory networks.
11 Classify the different strategies for Multiple Time Scales. BTL-3 Apply

6
BrainKart lOMoARcPSD|453 335 8

12 Develop a block diagram for LSTM. BTL-6 Create

13 Illustrate important design patterns for recurrent neural networks. BTL-3 Apply
14 Summarize about echo state networks. BTL-5 Evaluate
15 Point out the advantage of introducing depth in Deep recurrent Networks. BTL-4 Analyze
16 Compare gradient descent with and without gradient clipping using BTL-4 Analyze
diagram.
17 Justify the major advantages of unfolding process in computational BTL-5 Evaluate
graphs.
18 Illustrate block diagram of LSTM recurrent network “cell”. BTL-3 Apply
19 What are leaky units? BTL-1 Remember
20 Develop a schematic diagram of a network with an explicit memory. BTL-6 Create
21 Give a block diagram for Long Short Term Memory. BTL-2 Understand
22 Illustrate echo state networks. BTL-3 Apply
23 Explain liquid state machines. BTL-4 Analyze
24 Assess explicit memory. BTL-5 Evaluate
PART – B
1 i. Describe Unfolding Computational Graphs. (6) BTL-1 Remember
ii. Explain Bidirectional RNNs. (7)
2 Describe the following. BTL-1 Remember
i. Teacher Forcing in Recurrent Neural Networks. (7)
ii. Networks with Output Recurrence. (6)
3 i. Echo State Networks. (7) BTL-1 Remember
ii. Explain challenge of Long-Term Dependencies.(6)
4 Discuss Recurrent Neural Networks in detail.(13) BTL-2 Understand
5 Describe Deep Recurrent Networks in detail.(13) BTL-2 Understand
6 Illustrate Encoder-Decoder sequence-to-sequence Architecture. (13) BTL-3 Apply
7 Explain Leaky Units and Other Strategies for Multiple Time Scales. (13) BTL-4 Analyze
8 Point out various features of Echo state networks. (13) BTL-4 Analyze
9 Explain Optimization for Long-Term Dependencies. (13) BTL-5 Evaluate
10 Compute the gradient in a Recurrent Neural Network. (13) BTL-6 Create

11 i. Illustrate Clipping Gradients. (7) BTL-3 Apply

ii. Illustrate Regularizing to Encourage Information Flow.( 6)

7
BrainKart lOMoARcPSD|453 335 8

12 Describe the following. BTL-1 Remember

i. Long Short-Term Memory. (7)
ii. Other Gated RNNs. (6)
13 Explain in detail about the following. BTL-4 Analyze
i. Adding Skip Connections through Time. (6)
ii. Leaky Units and a Spectrum of Different Time Scales .(7)
14 Describe Explicit memory. (13) BTL-2 Understand
15 Discuss Echo State Networks. (13) BTL-2 Understand
16 Illustrate Bidirectional RNNs.(13) BTL-3 Analyze
17 Explain challenge of Long-Term Dependencies. (13) BTL-5 Evaluate
PART – C
1 Develop an example for Unfolding Computational Graphs and describe the BTL-6 Create
major advantages of unfolding process. (15)
2 Explain how to compute the gradient in a Recurrent Neural Network.(15) BTL-5 Evaluate
3 Explain a modeling sequences Conditioned on Context with RNNs. (15) BTL-5 Evaluate
4 Prepare an example of Encoder- Decoder or sequence-to-sequence RNN BTL-6 Create
architecture.(15)
5 Explain various Gated RNNs. (15) BTL-5 Evaluate