Course

Module 1:

Fundamentals of Neural Networks – Types of Machine Learning, Types of Artificial Neural Networks – The McCulloch−Pitts Network, Perceptron, The Sigmoid Neuron, Linear Separable Problem, Multilayer Perceptron (MLP), Optimization Techniques, Exploding Gradient Problem, Weight Initialization, Deep Learning. Convolutional Neural Network (CNN) – Components of CNN Architecture, Rectified Linear Unit (ReLU) Layer, Exponential Linear Unit (ELU, or SELU), Unique Properties of CNN, Architectures of CNN, Applications of CNN.

Module 2:

Recurrent Neural Network (RNN) – Simple Recurrent Neural Network, LSTM Implementation, Gated Recurrent Unit (GRU), Deep Recurrent Neural Network. Autoencoder – Features of Autoencoder, Types of Autoencoder

Module 3:

Restricted Boltzmann Machine – Boltzmann Machine, RBM Architecture, Types of RBM.

Open-Source Frameworks of Deep learning – Python, TensorFlow, Keras, PyTorch.

Applications with Deep Learning – Image Classification using CNN, Visual Speech Recognition using 3D – CNN, Stock Market Prediction using RNN, Next Word Prediction using RNN-LSTM.

Suggested Lab Sessions:

·         Implement and analyse perceptron and multilayer perceptron (MLP) models for linear and non-linear classification tasks. Python with TensorFlow or Keras.

·         Build a CNN model to classify handwritten digits using the MNIST dataset. Include: Convolution, pooling, ReLU/ELU layers. Evaluation of accuracy and loss curves

·         Build a simple RNN to predict the next value in a sine wave sequence. Implement an LSTM model for next word prediction using a sample text corpus.

·        Develop an autoencoder to perform dimensionality reduction and reconstruction.

Course Outcomes:

CO1: Understand the fundamentals of neural networks.

CO2: Design feed-forward networks using different techniques

CO3: Model Convolution Neural Networks and articulate RNN and Autoencoder.

CO4: Understand the architecture of Boltzmann Machines and Frame works for deep learning

CO5: Apply Neural Networks to practical problems.

Textbooks / References:

1.      S Lovelyn Rose, L Ashok Kumar, D Karthika Renuka, “Deep Learning using Python”, 1st Edn.  2019, John Wiley & Sons.

2.      Goodfellow I, Bengio Y, Courville A, & Bengio Y, “Deep learning”, Cambridge: MIT Press, 1st Edition, 2016.

3.      Michael Nielsen, “Neural Networks and Deep Learning”, Online book, 2016.

4.      Umberto Michelucci, “Applied Deep Learning. A Case-based Approach to Understanding Deep Neural Networks” A Press, 2018.

5.      Francois Chollet “Deep Learning with Python”, Manning Publications, 2017.