Course

Neural Networks basics – Linear Separable Problems and Perceptron – Multi layer neural networks and Back Propagation, Practical aspects of Deep Learning: Train/ Dev / Test sets, Bias/variance, Vanishing/exploding gradients, Gradient checking, Hyper Parameter Tuning

Convolutional Neural Networks – Basics and Evolution of Popular CNN architectures – Transfer Learning–Applications: Object Detection and Localization, Face Recognition, Neural Style Trans- fer Recurrent Neural Networks – GRU – LSTM – NLP – Word Embeddings – Transfer Learning – Attention Models – Applications: Sentinel Classification, Speech Recognition, Action Recognition

Restricted Boltzmann Machine – Deep Belief Network – Auto Encoders – Applications: Semi- Supervised classification, Noise Reduction, Non-linear Dimensionality Reduction. Goal Oriented Decision Making – Policy and Target Networks – Deep Quality Network for Reinforcement Learning

Introduction to GAN – Encoder/Decoder, Generator/Discriminator architectures. Challenges in NN training – Data Augmentation – Hyper parameter Settings – Transfer Learning– Developing and Deploying ML Models (e.g., Matlab/Tensor Flow/PyTorch)

Preamble

With the advent of high-end computing facilities and with the availability of huge amount of data, deep learning became the de facto standard machine learning strategy to learn complicated patterns and is offering the state of the art results in diverse fields including but not limited to automatic language translation, speech processing, medical diagnoses and in almost all fields of computer vision. This course provides core understanding in different deep learning architectures, design principles, learning strategies and encourages the usage of many deep learning tools in designing and deploying solutions.

Course Objectives

• To introduce to students, different deep neural network architectures, training strategies/ algorithms, possible challenges, tools and techniques available in designing and deploying solutions to different practical/Engineering problems.

Course Outcomes

Prerequisites

• Computational Linear Algebra

CO-PO Mapping

Evaluation Pattern – 70:30

• Midterm Exam – 20%
• Lab Assignments & Case Study – 50%
• End Semester Exam – 30%

Text Book / References

1. Ian Goodfellow, YoshuaBengio and Aeron Courville,” Deep Learning”, MIT Press, First Edi- tion, 2016.
2. Adam Gibson and Josh Patterson,” Deep Learning, A practitioner’s approach”, O’Reilly, First Edition, 2017.
3. Francois Chollet,” Deep Learning with Python”, Manning Publications Co, First Edition, 2018.
4. Research Papers on Relevant Topics and Internet Resources