R.V.S.
TECHNICAL CAMPUS
DEPARTMENT OF
COMPUTER SCIENCE AND ENGINEERING
ACADEMIC YEAR 2024-2025
QUESTION BANK
SUBJECT : AD3501 – DEEP LEARNING
SEM/YEAR : V/III
UNIT I DEEP NETWORKS 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.
QUESTIONS
PART - A
1 Differentiate scalar and vector.
2 What is stochastic gradient descent
3 Differentiate supervised and unsupervised learning.
4 What is Discrete data? Give example.
5 What is Continuous data? Give example.
6 What is Categorical data? Give example.
7 What are the types of Categorical data?
8 What is Polytomous variable?
9 What are the Measurement Scales available?
10 What is qualitative data?
11 What is Nominal?
12 What is Ordinal?
13 What is Interval?
14 What is Ratio?
15 What are the software tools available for EDA?
16 What is Histogram?
17 What is Data transformation?
18 What are the transformation techniques?
19 How to handle the missing data?
20 Illustrate identity matrix
PART – B
1. Discuss the Bias - Variance trade off.
2. Dicuss overfitting and underfitting with an Example.
3. Explain the operations of deep feed forward network with a diagram.
4. Compare EDA with Classical and Bayesian analysis with diagram.
5. Explain Software Tools available for EDA in Detail.
6. Explain Visual Aids for EDA.
7. Explain Merging database – style dataframes.
8. Explain different Data Transformation Techniques in detail.
9. Explain Line chart and Bar chart in detail with graph.
10. Explain Polar chart and Histogram in detail.
UNIT I I 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.
QUESTIONS
PART – A
1. What is convolutional networks?
2. List three stages of a convolutional network.
3. Illustrate reverse correlation.
4. Discussabout parameter sharing in neural network.
5. Howpooling handles inputs of varying size?
6. List three important ideas that help to improve a machine learning
system.
7. Explain complex layer terminology.
8. Differentiate complex layer terminology and simple layer terminology in
Convolutional network.
9. Give example for convolution.
10. Examine equivariance to translation.
11. List out various formats of data that can be used with convolutional
Networks.
12. Explain feature map.
13. Give three properties of V1 that a convolutional network layer is
14. designed
to capture.
15.Simulatethe idea behind reverse correlation.
15. Howto reduce the cost of convolutional network training?
16. Defineprimary visual cortex.
17. Whatisunshared convolution?
18. Createa chart that demonstrates convolution with a stride.
20.Show three basic strategies for obtaining convolution kernels without
Supervised training.
PART – B
1. Write an example function for Convolution operation and explain in detail.
2. Sparse interactions with suitable diagram
3. Parameter sharing with suitable diagram
4. Describe Pooling with suitable example.
5. Evaluate variants of the basic convolution function.
6. Create a graphical demonstration for parameter sharing and explain it in Detail.
7. Discuss local connections, convolution and full connections with diagram?.
8. Illustrate unshared convolution with suitable examples.
9. Write short notes Max Pooling.
10. Develop table with examples of different formats of data that can be used
with convolutional networks.
UNIT III CONVOLUTIONAL 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.
QUESTIONS
PART - A
1. What is Recurrent Neural Networks?
2. Illustrate echo state networks.
3. What are leaky units?
4. Illustrate block diagram of LSTM recurrent network “cell”.
5. Illustrate echo state networks.
6. Develop block diagram for LSTM.
7. What is Bidirectional Recurrent Neural Networks?
8. What is decoder?
9. Predict the concept of gated RNNs.
10. Illustrate important design patterns for recurrent neural networks.
11. Point out the advantage of introducing depth in Deep recurrent Networks.
12. Classify the different strategies for Multiple Time Scales.
13. Compare echo state network and liquid state machines.
14. Describe Recursive Neural Networks.
15. Give the blocks of decomposition of computation of most Recurrent Neural
Network.
16. Assess explicit memory.
17. Develop a schematic diagram of a network with an explicit memory.
18. Summarize about echo state networks.
19. What is Encoder?
20. Give the advantage of recursive nets over recurrent nets.
PART – B
1. Compute the gradient in a Recurrent Neural Network.
2. Explain Optimization for Long-Term Dependencies.
3. Point out various features of Echo state networks.
4. Explain Leaky Units and Other Strategies for Multiple Time Scales.
5. Illustrate Encoder-Decoder sequence-to-sequence Architecture.
6. Describe Deep Recurrent Networks in detail.
7. Discuss Recurrent Neural Networks in detail.
8. Explain Bidirectional RNNs.
9. Explain variousGated RNNs.
10. Explain how to compute the gradient in a Recurrent Neural Network.
UNIT IV MODELEVALUATION
Performancemetrics--BaselineModels--Hyperparameters:ManualHyperparameter --Automatic
Hyperparameter -- Grid search -- Random search -- Debugging strategies.
QUESTIONS
PART – A
1. PERFORMANCE METRICS
2. In machine learning, how each task or problem is divided into
3. Performance Metrics for Classification
4. Accuracy
5. When to Use Accuracy?
6. When to Use Accuracy?
7. Define Confusion Matrix
8. Define Precision
9. Define Recall or Sensitivity
10. When to use Precision and Recall?
11. Define F-Scores
12. Define AUC-ROC
13. Define Performance Metrics for Regression
14. Define Mean Absolute Error (MAE)
15. Define Mean Squared Error
16. Define R Squared Score
17. Define adjusted R Squared
18. What is a Baseline Model?
19. Give the types of baseline models
20. Define Hyperparameters
UNIT V AUTOENCODERSANDGENERATIVEMODELS
Autoencoders: Undercomplete autoencoders -- Regularized autoencoders -- Stochastic encoders
and decoders -- Learning withautoencoders; DeepGenerativeModels: Variational autoencoders –
Generative adversarial networks.
QUESTIONS
PART – A
1. What is Probabilistic PCA and Factor Analysis?
2. Define Linear Factor Model.
3. Give the various generalizations of ICA.
4. What is Independent Component Analysis?
5. Give major advantage of slow feature analysis.
6. Name the various tasks than can be done by probabilistic models.
7. What is Denoising Autoencoders?
8. Predict the primary disadvantage of the non-parametric encoder.
9. Point out the trade-off faced in representation learning problems.
10. Distinguish between one-shot learning and zero-shot learning.
11. Develop distribution equation for energy based model.
12. Which are undirected models?
13. Summarize Distributed representations.
14. Compare directed models and undirected models.
15. How many task does the learner must perform in transfer learning?
16. List the two different ideas combined by Unsupervised pre-training.
17. Evaluate Undirected models.
18. Compare distributed representation and a symbolic one.
19. Slow Feature Analysis is an efficient application of slowness principle?
20. Point out the reason for why Greedy layer-wise pre-training called Greedy
PART – B
1. Assess Independent Component Analysis.
2. Develop various graphs to describe Model Structure.
3. Explain Autoencoders.
4. Explain Monte Carlo methods.
5. Develop a short notes on Separation and D-Separation.
6. Explain Markov random fields.
7. Write about representation learning.
8. Discuss about Slow Feature Analysis.
9. Describe Distributed Representation.
10. Discuss in detail about transfer learning and Domain Adaptation.
