M.Sc.

(DS) Sem III

Deep Learning
Question Bank

Unit I
1. Describe Machine Learning and its evolution and Discuss how does machine learn. Give
examples. (7/8/10). Bloom’s level – 2.
2. Describe any 5 real-time applications of Deep learning. (5). Bloom’s level – 2
3. Discuss the following with suitable examples: (10). Bloom’s level – 2.
i. Scalars
ii. Vectors
iii. Matrices
iv. Tensors
v. Hyperplanes
4. What is probability? Discuss conditional probability considering a suitable example. (8/10)
Bloom’s level – 2.
5. Describe probability distributions. Explain Bootstrapping and cross-validation resampling
methods. Give examples. (10) Bloom’s level – 2.
6. Discuss the different resampling methods available in probability distribution. (10) Bloom’s level
– 2.
7. Define Regression. Explain Linear Regression giving a suitable example. (10) Bloom’s level – 2.
8. Explain Classification considering a suitable example. (10) Bloom’s level – 2.
9. Differentiate Regression and Classification problems of Machine Learning. (5) Bloom’s level –
4.
10. Write a short note on Clustering. (5) Bloom’s level – 2.
11. Describe Errors in Machine Learning. (5) Bloom’s level – 2.
12. What is Bias? Explain the concept of Bias with respect to machine learning models. (7/8)
Bloom’s level – 2.
13. What is Variance? Explain the concept of Variance with respect to machine learning models.
(7/8) Bloom’s level – 2.
14. Write a short note on Bias Variance Trade-off. (5) Bloom’s level – 2.
15. What is Underfitting? Explain the reasons that causes underfitting and the techniques to reduce it.
(10) Bloom’s level – 2.
16. What is Overfitting? Explain the reasons that causes Overfitting and the techniques to reduce it.
(10) Bloom’s level – 2.
17. Distinguish between Model Parameters and Hyperparameters. Give examples for both. (8)
Bloom’s level – 2.
18. Explain Model Parameters and Hyperparameters. Give examples. (8) Bloom’s level – 2.
19. Define Optimization. Explain the working of Gradient Descent algorithm. (10) Bloom’s level –
2.
20. Explain the working of Stochastic Gradient Descent algorithm. (10) Bloom’s level – 2.
21. Explain Training Data Set, Validation Data Set and Test Data Set. Give examples. (10) Bloom’s
level – 2.
22. Differentiate between Training Data Set, Validation Data Set and Test Data Set. (10) Bloom’s
level – 2.

SHREETHA BHAT, ASSISTANT PROFESSOR, SCHOOL OF CSA, 1

REVA UNIVERSITY, BENGALURU.
 M.Sc. (DS) Sem III

23. What is an Estimator? Explain the Use and types of Estimators in Machine Learning. (10)
Bloom’s level – 2.
24. Discuss the challenges faced by Deep Learning in current scenario. (10) Bloom’s level – 2.

Unit II
1. Define Neural Network. Explain its architecture with a neat diagram. Describe the linear equation
Ax=b in context of Neural Network. (8) (Bloom’s Level -2)
2. Explain the structure of a Biological neuron cell with a neat diagram. (8) (Bloom’s Level -2)
3. Define a Perceptron. Discuss single layer perceptron with a neat diagram. (8) (Bloom’s Level -2)
4. Write the Single layer Perceptron learning algorithm. (5) (Bloom’s Level -3)
5. List down the limitations of the perceptron. (3) (Bloom’s Level -1)
6. Distinguish between a perceptron and a multilayer feed forward network. (5) (Bloom’s Level -2)
7. Discuss the structure of an artificial neuron for a multilayer perceptron with a neat diagram. (8)
(Bloom’s Level -2)
8. Describe the following terms with reference to multilayer perceptron: (9) (Bloom’s Level -2)
i. Connection Weights
ii. Biases
iii. Activation Function
9. Explain with a neat diagram the Feed Forward Neural Network Architecture. (10) (Bloom’s Level
-2).
10. Outline the steps to be followed in training neural networks. (5) (Bloom’s Level -1).
11. Write the pseudocode for the general neural network training. (5) (Bloom’s Level -3).
12. What do you mean by Back Propagation learning? Write the pseudocode for Back Propagation
algorithm for updating weights. (10) (Bloom’s Level -1,3).
13. Write the meaning for the following neural network notations:
i, ni, j, ai, Ai, g, Erri, Wi, Wj,I, input_sumi (10) (Bloom’s Level -1,3).
14. Explain the steps followed to update weights using Back Propagation algorithm. (12) (Bloom’s
Level -1,3).
15. Describe the Activation Function. Explain its types. (10) (Bloom’s Level -2).
16. Write short notes on the following Activation Functions: (10) (Bloom’s Level -3).
i. Linear
ii. Sigmoid
iii. Tanh
iv. ReLu
v. Softmax
17. Loss function is a method of evaluating how well your machine learning algorithm models your
featured data set. Discuss Logistic Loss and Hinge Loss used for Classification problems. (10)
(Bloom’s Level -2).
18. List and describe the loss functions used for regression problems. (10) (Bloom’s Level -2).
19. Write a short note on the following regression loss functions: (10) (Bloom’s Level -2).
i. Mean squared error loss
ii. Mean absolute error loss.
iii. Mean squared log error loss.
iv. Mean absolute percentage error loss
20. Discuss the role of fine-tuning learning rate to optimize the performance of an neural network
model. (8) (Bloom’s Level -2)

SHREETHA BHAT, ASSISTANT PROFESSOR, SCHOOL OF CSA, 2

REVA UNIVERSITY, BENGALURU.
 M.Sc. (DS) Sem III

21. Describe the role of Regularization used in controlling overfitting in machine learning. (5)
(Bloom’s Level -2).
22. Write a short note on Momentum and Sparsity hyperparameters in machine learning. (5) (Bloom’s
Level -2).
23. Define Deep Learning. Outline the advantages of Deep Neural Networks. (7) (Bloom’s Level -2).
24. What do do mean by Regularization in Deep Learning? Explain the following regularization
techniques:
a. Dropout
b. DropConnect
c. L1 penalty
d. L2 penalty

25. What are Restricted Boltzmann Machines (RBM)? Explain the process of Reconstruction
using RBM.
26. Write a short note on Autoencoder

SHREETHA BHAT, ASSISTANT PROFESSOR, SCHOOL OF CSA, 3

REVA UNIVERSITY, BENGALURU.
 M.Sc. (DS) Sem III

Unit III
1. Explain the architecture of an Autoencoder. Illustrate it with a neat diagram. (10) (Bloom’s Level -
4).
2. Describe an Autoencoder. Explain its variants. (5) (Bloom’s Level - 2).
3. What is a Boltzman Machine? Explain it with a supporting diagram. (7) (Bloom’s Level - 3).
4. Describe the Restricted Boltzman Machine. Explain its working with a neat diagram. Give an
example of its applications. (10) (Bloom’s Level - 3).
5. What is Deep Belief Network (DBN)? Explain its architecture with a neat diagram. (10) (Bloom’s
Level - 3).
6. Describe the steps in extracting features using Restricted Boltzman Machine (RBM). (7) (Bloom’s
Level - 2).
7. What is Generative Adversarial Network (GAN)? Illustrate its working with a suitable diagram
and example. (10) (Bloom’s Level - 4).
8. Outline the popular use cases of Generative Adversarial Networks (GAN). (5) (Bloom’s Level - 4).
9. Describe Convolutional Neural Networks in brief. Outline its applications in real life. (5/7)
(Bloom’s Level - 2).
10. Explain the architecture of the Convolutional Neural Network (CNN) with a neat diagram. (10)
(Bloom’s Level - 3).
11. Write a brief note on the Input layer of CNN architecture. (5) (Bloom’s Level - 2).
12. Define Convolution. Discuss the role of the Convolution layer in feature extraction. (7/8) (Bloom’s
Level - 2).
13. Discuss the four major components of the Convolutional Layer. (Filters, Activation maps,
Parameter sharing Layer-specific hyperparameters) (10) (Bloom’s Level - 2).
14. What are Filters? Explain its function in Convolutional Layer. (5) (Bloom’s Level - 2).
15. Activation is a numerical result if a neuron decided to let information pass through. Discuss the
working of Activation Maps in Convolutional Layer. (7/8) (Bloom’s Level - 2).
16. Discuss the following hyperparameters w.r.t convolutional layer. (10) (Bloom’s Level - 2).
i. Filter Size
ii. Output Depth
iii. Stride
iv. Zero Padding.
17. Pooling layers reduce the data representation progressively over the network and help control
overfitting. Explain its working with a suitable example. (7/8) (Bloom’s Level - 2).
18. Write a short note on Fully Connected Layer. (5) (Bloom’s Level - 2).

SHREETHA BHAT, ASSISTANT PROFESSOR, SCHOOL OF CSA, 4

REVA UNIVERSITY, BENGALURU.
 M.Sc. (DS) Sem III

Unit IV

1. Compare and contrast Feed Forward Neural Networks with Recurrent Neural Networks. (8)
(Bloom’s Level - 4).
2. Justify the usage of Recurrent Neural Network (RNN) for handling sequential Data. (7) (Bloom’s
Level - 6).
3. Outline the issues in Feed Forward Neural Network. Suggest a suitable solution for the problems
listed. (5) (Bloom’s Level - 4).
4. Discuss the applications of RNN in real life scenario. (5) (Bloom’s Level - 2).
5. Explain the working of Recurrent Neural Network. Illustrate it with a neat diagram. (10) (Bloom’s
Level - 2).
6. Discuss the types of Recurrent Neural Networks. Give example for each. (8). (Bloom’s Level - 2).
7. Explain the two standard issues in RNN (Vanishing Gradient Problem, Exploding Gradient
Problem.) (10) (Bloom’s Level - 2).
8. Outline the issues in RNN. (5) (Bloom’s Level - 2)
9. What is Long-Short Term memory (LSTM). With a illustrative diagram explain its architecture.
(10) (Bloom’s Level - 2)
10. Write a brief note on Recursive neural Network. 5((Bloom’s Level - 2)
11. Discuss any two applications of Recursive Neural Network. 10 (Bloom’s Level – 2)
12. Explain the architecture of a Transformer with a neat diagram. 10 (Bloom’s Level – 2)
13. Explain the different components in transformer architecture. 10 (Bloom’s Level – 2)
14. Explain the Attention mechanism. How is it used in transformers?
15. Explain different attention mechanisms and which attention is generally used in transformer
architecture.
16.

SHREETHA BHAT, ASSISTANT PROFESSOR, SCHOOL OF CSA, 5

REVA UNIVERSITY, BENGALURU.