CSE756: Probabilistic Graphical Models

Foundations of Probabilistic Modeling: probability distributions, independence, conditional probability, Bayes theorem; Graph Representations: directed and undirected graphs, factorization, graphical model structure; Bayesian Networks (BNs): structure learning, D-separation, exact and approximate inference, parameter estimation; Markov Networks (MRFs): undirected graphical models, pairwise Markov property, Gibbs distribution; Inference in Graphical Models: variable elimination, belief propagation, junction tree algorithm, Monte Carlo methods (MCMC, Gibbs sampling); Dynamic Graphical Models: Hidden Markov Models (HMMs), Kalman filters, dynamic Bayesian networks; Structured Prediction: CRFs, probabilistic modeling for sequential and spatial data; Deep Learning & Graphical Models: deep generative models, variational inference, hybrid approaches (Neural ODEs, Graph Neural Networks); Applications: medical diagnosis, natural language processing, computer vision, reinforcement learning, recommendation systems; Challenges and Advances: scalability, approximate inference techniques, learning in large-scale models.

Course Objectives

The core objectives of this course are to:
Understand the fundamentals of Probabilistic graphical models: potential frameworks for representing complex domains using probability distributions, with numerous applications in machine learning, computer vision, natural language processing and computational biology
Provide a comprehensive survey of the topic, introducing the key formalisms and main techniques used to construct them, make predictions, and support decision-making under uncertainty
To develop the knowledge and skills necessary to design, implement and apply these models to solve real problems.

List of Books

1. To Be Added

Course Outcome