ABOUT THE COURSE

This course is an introduction to deep learning, a branch of machine learning concerned with developing and applying modern neural networks. Deep learning algorithms extract layered high-level representations of data to maximize performance on a given task. For example, if asked to recognize faces, a deep neural network may learn to represent image pixels with edges, larger shapes, parts of the face like eyes and ears, and, finally, individual face identities. Deep learning is behind many recent advances in AI, including Siri’s speech recognition, Facebook’s tag suggestions, and self-driving cars.  This course covers the basics of neural networks, different types of neural network architectures, optimization techniques, deep neural networks such as CNN, RNN, LSTM, etc., and sequence modeling, auto-encoders, attention, and transformers. 

COURSE OUTCOMES

CO - 1: Apply knowledge of deep learning principles to create and implement a simple feedforward neural network.
CO - 2: Analyze and evaluate deep learning architectures for specific tasks in computer vision, NLP, and generative modeling.
CO - 3: Design and optimize data preprocessing techniques, loss functions, and regularization methods to train deep learning models effectively.
CO - 4: Synthesize and demonstrate proficiency in advanced activation functions, normalization techniques, transfer learning, and state-of-the-art research in deep learning for complex model development.

COURSE MODULES

MODULE 1: Fundamentals of Deep Learning - Introduction to Deep Learning: Definition and Motivation, Overview and Key Milestones behind Deep Learning, Deep Learning vs. Traditional Machine Learning: Comparison, Advantages and limitations of Deep Learning, Neural Network Basics: Structure and Function of Artificial Neuron, Activation Functions, Building Blocks: Layes, Weights, Bias, Forward and Backward Propagation, Implementation of Feedforward Neural Network.

MODULE 2: Deep Learning Architectures - Multi-Layer Perceptron, Back Propagation Algorithm, Convolutional Neural Networks (CNN): Architecture, Pooling, Usecases: Image Classification, Object Detection, and Image Segmentation, Recurrent Neural Networks (RNNs): Architecture, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Generative Adversarial Networks (GANs): Training to generate images, music, or text, Applications of GAN: art generation and data augmentation, Applications of Deep Learning: Case Studies.

MODULE 3: Deep Learning Operations (DLOps) - Introduction to Deep Learning Operations: Significance and Management, DL Models in Production: Challenges, Complexities, Model Versioning, and Collaboration, Tools for collaborative development and model versioning, Continuous Integration and Continuous Deployment (CI/CD), Scalability and Distributed Training, Monitoring and Logging, Model Governance and Ethics, Automated Machine Learning (AutoML)

MODULE 4: Practical Applications of Deep Learning - Image Classification: Use CNN for Image Classification, Natural Language Processing (NLP): Train RNN for Sentiment Analysis, Computer Vision: Object Detection using Pre-trained Models, Image Generation using GAN

COURSE PREREQUISITES

This is an upper-level undergraduate/graduate course. All students should have knowledge on Calculus, Linear Algebra, Probability & Statistics, and should have the ability to code in Python. 

TEXT BOOKS

1. Deep Learning by Ian Goodfellow, Yoshua Bengio, and Aaron Courville

2. Neural Networks and Deep Learning: A Textbook by Charu Aggarwal

3. Dive into Deep Learning by Aston Zhang, Zack C. Lipton, Mu Li, and Alexander J. Smola

4. Deep Learning with Python by Francois Chollet

LEARNING MATERIALS

• Introduction to Neural Networks

• Back Propagation Neural Network Algorithm

• Deep Neural Networks

• Supplementary Material: Long Short-Term Memory (LSTM)

• Sequence Modeling, Kernel Machines, Statistical Learning Theory

  • • Support Vector Clustering

    • Multiple Kernel Learning

• Autoencoders, RBM, Attention, Transformers, GAN, Deep Kernels 

  • • RBM

    • Autoencoder

LAB EXERCISES

• Exercise 1 - Classification using a simple Artificial Neural Network

• Exercise 2 - Classification using Backpropagation Neural Network

• Exercise 3 - Image Classification using Convolutional Neural Network

• Exercise 4 - Sentiment Classification using RNN/LSTM

• Exercise 5 - Salary Prediction using Support Vector Regression

• Exercise 6 - Dropout Regularization in Deep Learning Models

  • • Dataset

ASSIGNMENTS

• Assignment 1: 

  • • Implement a spam email classifier using simple Perceptron

• Assignment 2: 

  • • Implement a Support Vector Machine Classifier for Sentiment Classification of Twitter Data on COVID-19

VIDEO LECTURES

CONTACT

If you have any questions or need further clarification about the course, please contact me via email or during my office hours. Let's embark on this exciting journey into Deep Learning! Get ready to unlock the potential of AI through the power of deep neural networks.

Looking forward to a rewarding and knowledge-filled semester!

Dr. Anoop V. S.
Course Instructor