Syllabus: CS 8803 Sequence Prediction

Course Description

This course explores the theoretical foundations of sequential prediction and decision making. We cover a range of topics from classical information theory and algorithmic complexity (Kolmogorov complexity, Solomonoff Induction) to modern generative models (Transformers, Discrete Diffusion Models). The course emphasizes mathematical rigor and the connections between different frameworks.

The course is designed for mathematically-trained graduate students who are interested in engaging in modern research in these areas.

Instructor Information

• Instructor: Jacob Abernethy
• Office Hours: Mondays 3:30 PM (2nd floor of CODA). Note: Please give at least one hour prior warning if you plan to attend.

Material organization

The course will be broken down into three segments

• Part I: Lecture content, one problem set, midterm exam
• Part II: Lecture content, 2nd problem set, 2nd midterm exam
• Part III: Final project, presentations and report due.

The final three lecture meetings of the course will be reserved for project presentations.

Grading

• Problem Sets: 20%
• Exams: 40%
• Final Project: 40%

Policies

(Standard GT policies apply)