Home › AKTU PYQ › Deep Learning › PYQ Questions

Deep Learning (BAI701) AKTU previous year questions 2023–2026

Every Deep Learning question from 5 AKTU papers, tagged by unit, topic and marks. A few recent questions from each unit are listed below; open the page to filter by unit, topic or mark type.

Unit 1: Introduction to Neural Networks – AKTU PYQs

  • Define model in machine learning. (2 marks, 2026, Introduction to Machine Learning)
  • What is the difference between linear regression and logistic regression? (2 marks, 2026, Linear Models (SVM, Perceptron, Logistic Regression))
  • Derive the mathematical expression for the output of a single hidden-layer neural network. Explain the role of activation functions. (7 marks, 2026, Neural Networks & Shallow Networks)
  • Explain the backpropagation algorithm in detail. Derive the weight update rule for a multilayer perceptron. (7 marks, 2026, Training a Network (Loss, Backprop, SGD))
  • Discuss the significance of activation functions in the Universal Approximation Theorem. (7 marks, 2026, Universal Function Approximation)

Unit 2: Deep Networks – AKTU PYQs

  • Discuss convolutional neural network (CNN)? (2 marks, 2026, Convolutional Networks)
  • Name any two activation functions used in deep networks. (2 marks, 2026, Convolutional Networks)
  • Name two applications of Word2Vec. (2 marks, 2026, Generative Adversarial Networks (GAN))
  • Describe how probabilistic models help reduce overfitting. Explain with examples. (7 marks, 2026, Regularization & Batch Normalization)
  • Compare batch normalization with layer normalization and instance normalization. (7 marks, 2026, Regularization & Batch Normalization)

Unit 3: Dimensionality Reduction – AKTU PYQs

  • Describe the role of regularization during optimization. (2 marks, 2026, Training a ConvNet (Weight Init, Hyperparameters))
  • Compare PCA and LDA. Explain how LDA maximizes class separability using scatter matrices. (7 marks, 2026, Linear Dimensionality Reduction (PCA, LDA))
  • Explain convolution operation mathematically, including stride, padding, and feature maps. (7 marks, 2026, ConvNet Architectures (AlexNet, VGG, ResNet))
  • Compare AlexNet, VGG, Inception, and ResNet in terms of depth, performance, and computation. (7 marks, 2026, ConvNet Architectures (AlexNet, VGG, ResNet))
  • Explain the impact of weight initialization on the performance of deep networks. (2 marks, 2025, Training a ConvNet (Weight Init, Hyperparameters))

Unit 4: Optimization & Generalization – AKTU PYQs

  • Discuss the generalization in machine learning. (2 marks, 2026, Generalization in Neural Networks)
  • Explain the optimization problem in deep learning. Why are deep learning loss surfaces highly non-convex? Discuss with suitable diagrams. (7 marks, 2026, Optimization in Deep Learning)
  • Compare RNN, LSTM, and GRU networks. Highlight their advantages and use cases. (7 marks, 2026, Recurrent Networks & LSTM)
  • What is deep reinforcement learning? Explain its architecture involving policy networks, value networks, and environment interaction. (7 marks, 2026, Deep Reinforcement Learning)
  • Define non-convex optimization. (2 marks, 2025, Optimization in Deep Learning)

Unit 5: Case Studies & Applications – AKTU PYQs

  • Describe the ImageNet dataset in detail. Explain its scale, structure, challenges, and importance in the evolution of deep learning. (7 marks, 2026, Image Classification & Detection)
  • Explain how WaveNet differs from RNNs for sequential audio generation. (7 marks, 2026, Audio WaveNet)
  • Explain the Word2Vec model. Discuss CBOW and Skip-Gram architectures with diagrams. (7 marks, 2026, NLP & Word2Vec)
  • Describe the architecture and training of WaveNet for audio generation. (2 marks, 2025, Audio WaveNet)
  • How deep learning models have revolutionized computer vision tasks? (2 marks, 2025, Image Classification & Detection)