DBATU Final Year Computer syllabus

DBATU · Final Year B.Tech Computer Engineering · Non-NEP

Deep Learning

BTCOE705 · Elective III · Semester VII · 3 credits · Continuous assessment 20 + mid-semester 20 + end-semester 60

One of the Elective III options.

All notes for this subject
  1. Unit 1: Foundations

    6 hours

    History and success stories of deep learning; McCulloch–Pitts neurons and thresholding logic; perceptrons and the perceptron learning algorithm; multilayer perceptrons and their representation power; sigmoid neurons, gradient descent and feed-forward networks.

    Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then

    • History, advantages and challenges of deep learning
    • McCulloch–Pitts neuron and thresholding logic
    • Perceptron and feed-forward networks
  2. Unit 2: Training Feed-forward Networks

    6 hours

    Backpropagation; gradient descent with momentum and Nesterov acceleration, stochastic GD, AdaGrad and RMSProp; principal component analysis and singular value decomposition.

    Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then

    • Forward and backpropagation
    • Optimisers: momentum, Nesterov, AdaGrad, RMSProp and Adam
    • PCA and singular value decomposition
  3. Unit 3: Autoencoders and Regularisation

    6 hours

    Autoencoders and their relation to PCA; denoising, sparse and contractive autoencoders; bias–variance trade-off, L2 regularisation, early stopping, data augmentation, parameter sharing; greedy layer-wise pre-training, better activations and initialisation, batch normalisation.

    Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then

    • Autoencoders
    • Denoising, sparse and contractive autoencoders
    • Regularisation: dropout, batch normalisation and early stopping
    • Weight initialisation and greedy layer-wise pre-training
  4. Unit 4: Convolutional Neural Networks

    6 hours

    CNNs; LeNet, AlexNet, ZF-Net, VGGNet, GoogLeNet and ResNet; learning vector representations of words.

    Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then

    • CNN architecture
    • LeNet and AlexNet
    • CNN architectures: VGG, GoogLeNet, ResNet and DenseNet
    • Learning word representations
  5. Unit 5: Recurrent Networks and Attention

    6 hours

    Recurrent neural networks; backpropagation through time; encoder–decoder models; the attention mechanism and attention over images.

    Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then

    • Recurrent neural networks
    • Backpropagation through time
    • Encoder–decoder sequence-to-sequence models
    • Attention mechanism

Unit contents are summarised from the official DBATU syllabus . AiCanCode.org is not affiliated with Dr. Babasaheb Ambedkar Technological University (DBATU).