Course Syllabus

CS274B: Learning in Graphical Models

University of California, Irvine, Spring 2021

Prof. Erik Sudderth

Probabilistic graphical models provide a flexible framework for describing large, complex, heterogeneous collections of random variables. This course surveys state-of-the-art methods for statistical learning and inference in graphical models, as motivated by applications in image and video analysis, text and language processing, sensor networks, autonomous robotics, biological structure prediction, social networks, and more.

We will study efficient inference algorithms based on optimization-based variational methods, and simulation-based Monte Carlo methods. Several approaches to learning from data will be covered, including conditional models for discriminative learning, and Bayesian methods for controlling model complexity. Motivating applications will be explored via homework assignments and a final project. 

Course Materials

• Piazza Links to an external site. will be used for all course announcements, discussions, and questions.  All enrolled students should sign up, watch for important announcements, and post questions (anonymously if you prefer) about course content.
• Lecture Calendar, including readings, slides, and how to access lecture recordings
• Homework Assignments are distributed via Canvas, but submitted and graded via gradescope Links to an external site.
• Homework Policies and Resources, including the Collaboration and Academic Honesty Policy
• Final Project deadlines and requirements
• Resources and References for help with homeworks and projects

Course Information

• Lectures:  Tuesdays and Thursdays from 12:30-1:50pm, distributed via videos and webinars.
• Instructor:  Prof. Erik Sudderth.  Office hours Wednesdays from 2:00-3:00pm, via Zoom: https://uci.zoom.us/j/91450586290 Links to an external site..
• Teaching Assistant:  Dahai Hao.  Office hours Tuesdays and Thursdays from 4:00-5:00pm, via Zoom:  https://uci.zoom.us/j/96112554639 Links to an external site.
• Grading:  50% homework assignments, 50% final projects.

Course Prerequisites

A previous course in probabilistic machine learning (UCI CS274A).  Comfort with multivariable calculus, linear algebra, probability theory, and mathematical writing.  Python programming, and some experience with optimization algorithms, required for homeworks and projects.