Skip to content
EndsemNov 20242019 patternBE IT

Deep Learning (DL) SPPU Endsem Question Paper Nov 2024

All 8 questions from this SPPU endsem paper, as printed. Read them here or open the PDF.

Open the PDFMost repeated DL questions
Subject code
414443
Pattern
2019 pattern
Exam
Endsem
Session
Nov 2024
Max. marks
70
Time
2½ Hours
Semester
VII
Paper code
[6354]-574

Printed for: B.E. (Information Technology)

Instructions to the candidates

  1. Answer Q1 or Q2, Q3 or Q4, Q5 or Q6, Q7 or Q8.
  2. Neat diagrams must be drawn wherever necessary.
  3. Figures to the right indicate full marks.
  4. Assume suitable data, if necessary.

Questions

  1. Q1

    1. a)

      What is Recurrent Neural Network (RNN)? Draw and Explain Simple Recurrent Neural Network. Explain four types of RNN with proper diagram.

      Also asked:May 2024 ·Nov 2023 ·Nov 2022

      [8]
    2. b)

      Explain in detail the following advantages of RNN over other types of Neural Network. i) Ability To Handle Variable-Length Sequences ii) Memory of Past Inputs

      [9]

  2. Q2

    1. a)

      Discuss how the RNNs suffer from the problem of vanishing gradients. And how is it overcome with Long Short-Term Memory Network (LSTM)?

      [9]
    2. b)

      What is meant by Recursive Neural Network? How is it different from Recurrent Neural Network?

      [8]

  3. Q3

    1. a)

      What is Autoencoder? Elaborate its two networks Encoder and Decoder.

      [5]
    2. b)

      How do autoencoders differ from other unsupervised learning techniques like PCA or clustering algorithms?

      [5]
    3. c)

      Explain the following listed hyperparameters and methods which are used to optimize the training of autoencoders : i) Activation functions ii) Loss functions iii) Learning rate iv) Batch size v) Early stopping vi) Initialization

      [8]

  4. Q4

    1. a)

      Explain the working of denoising autoencoder. In what way, it differ from regular autoencoders?

      [5]
    2. b)

      What is a contractive autoencoder? How does it differs from regular autoencoders? List the most appropriate applications of contractive autoencoders.

      [8]
    3. c)

      Differentiate between Undercomplete Autoencoder and Sparse Autoencoder.

      Also asked:May 2024

      [5]

  5. Q5

    1. a)

      What is a DenseNet? Why do we need Densenet? List it’s advantages.

      [5]
    2. b)

      What is transfer learning in CNN?

      [7]
    3. c)

      Categorize Domain Adaptation methods - Homogeneous DA, Heterogeneous DA.

      [5]

  6. Q6

    1. a)

      What are the representation learning challenges?

      [5]
    2. b)

      “Is Densenet good for image classification?” Extend your idea about it.

      [5]
    3. c)

      What is the general framework of Transfer Learning?

      [7]

  7. Q7

    1. a)

      Explain content based, collaborative and hybrid recommender system with pros and cons.

      Also asked:May 2024 ·Nov 2022

      [9]
    2. b)

      State different applications of Deep Learning? With suitable diagram explain use of CNN for image classification.

      [9]

  8. Q8

    1. a)

      State applications of NLP. Why RNN is preferred as compare to CNN for Natural Language Processing? State different NLP tasks where RNN is used.

      [9]
    2. b)

      State and Explain different centrality measures used in social network analysis. 

      [9]