AD-601 Deep Learning - May 2024 Question Paper

Under each question, How to answer lists the points an answer should cover, outlined by CampusPrep from the syllabus. Check the details against your textbook.

1. 9 Marks
   aDefine Neuron. Explain McCulloch-Pitts Neural Model in detail.
   How to answer

   • Definition of biological/artificial neuron and its components
   • Structure and working of McCulloch-Pitts model with threshold logic
   • Mathematical formulation and example
   • Capabilities and limitations
2. 5 Marks
   bExplain the difference between Machine Learning and Artificial Intelligence.
   How to answer

   • Definition of Artificial Intelligence and Machine Learning
   • Key differences in scope, goal and techniques
   • Comparison table
   • Examples of each
3. 7 Marks
   aExplain different types of Autoencoders.
   How to answer

   • Definition and structure of autoencoders
   • Types: vanilla, sparse, denoising, contractive, variational
   • Working of each type briefly
   • Applications
4. 7 Marks
   bWrite about Linear and Multilayer Perceptron in detail.
   How to answer

   • Structure of linear perceptron and its decision boundary
   • Architecture of multilayer perceptron with hidden layers
   • Activation functions and backpropagation
   • Comparison and applications
5. 7 Marks
   aBriefly discuss about Convolutional Neural Network with Deep Learning.
   How to answer

   • Definition and motivation of CNN in deep learning
   • Key layers: convolution, pooling, fully connected
   • Working with feature maps and weight sharing
   • Applications
6. 7 Marks
   bWhat is LeNet? Explain the usage of LeNet in object recognition.
   How to answer

   • Definition and history of LeNet architecture
   • Layer-wise architecture of LeNet-5
   • Use in digit and object recognition
   • Significance and limitations
7. 7 Marks
   aWrite the difference between Convolutional Neural Network and Recurrent Neural Network.
   How to answer

   • Basic architecture of CNN for spatial data
   • Basic architecture of RNN for sequential data
   • Differences in connectivity, memory and training
   • Comparison table with applications
8. 7 Marks
   bWhat is Long Short Term Memory in Neural Network? Explain.
   How to answer

   • Need for LSTM to solve vanishing gradient in RNN
   • Structure of LSTM cell: forget, input, output gates
   • Working of gates with equations
   • Applications
9. 7 Marks
   aGive an overview of MCMC Sampling.
   How to answer

   • Concept of Markov Chain Monte Carlo
   • Need for sampling in deep learning
   • Common algorithms: Metropolis-Hastings, Gibbs sampling
   • Applications
10. 7 Marks
    bWhat is the difference between Autoencoders and RBM? Discuss.
    How to answer

    • Principle of Autoencoders: encoder-decoder
    • Principle of RBM: energy-based bipartite model
    • Differences in training, architecture and use
    • Discussion with examples
11. 8 Marks
    aWhat do you understand by deep learning? Explain the different types of Deep Learning.
    How to answer

    • Definition of deep learning and deep networks
    • Types: CNN, RNN, autoencoders, GANs etc.
    • Characteristics of each type
    • Applications
12. 6 Marks
    bWhat is the objective of the empirical risk minimization? Explain the principle of Empirical Risk Minimization.
    How to answer

    • Definition of empirical risk vs true risk
    • Objective of empirical risk minimization
    • Principle and mathematical formulation
    • Overfitting and regularization link
13. 7 Marks
    aWhy do we use convolution network in Deep Learning? Discuss.
    How to answer

    • Need for convolution: parameter sharing, sparse interactions
    • Advantages for image data: translation invariance
    • Role in feature extraction and hierarchy
    • Examples
14. 7 Marks
    bExplain the difference between RNN and Bidirectional RNN.
    How to answer

    • Architecture of unidirectional RNN
    • Architecture of bidirectional RNN with forward and backward layers
    • Differences in context usage and training
    • Applications
15. 14 Marks
    Explain any two of the following:

    1. a)Structured Output in Convolutional Networks
    2. b)AlexNet
    3. c)Deep Boltzmann Machines
    4. d)Self Organizing Maps (SOMs)
    How to answer

    • Select any two topics from the list
    • Definition, architecture and working of each
    • Key features, advantages and applications
    • Diagram where relevant