Teaching | Somdyuti Paul

Spring 2024

Course Instructors: Somdyuti Paul, Mahesh Mohan M R, Jiaul Hoque Paik
Syllabus:

Introduction and historical trends in deep learning
Perceptron and deep feedforward neural networks, backpropagation and optimization using gradient descent based methods
Regularization techniques for deep networks
Introduction to convolutional neural networks, standard CNN architectures and their use for classification and regression
Visualization of CNN filters and gradients, and adversarial attacks
Autoencoders and their applications
Sequence modeling using recurrent neural networks, types of recurrent units and their roles in resolving long-term dependency issues
Introduction to transformers, language modeling using BERT
Neural architecture search
Deep generative models, Boltzmann machines and deep belief networks
Adversarial learning and generative adversarial networks
Multi-task learning and federated learning.

Course Instructors: Mahesh Mohan M R, Somdyuti Paul, Sudeshna Sarkar
Syllabus

Introduction to machine learning, basics of supervised, unsupervised and reinforcement learning
k nearest neighbours
Linear models for classification and regression
Bayesian learning and naive Bayes model
Bias-variance tradeoff
Decision Trees
MLE and MAP estimation
Support vector machines
Feedforward neural networks
Convolutional and Recurrent Neural networks
Ensemble methods
Clustering
Dimensionality reduction using PCA and LDA
Hidden Markov Models
Overview of large language models