CS 281A / STAT 241A: Statistical Learning Theory

Fall 2026

Course Description

This graduate-level course studies the foundations and frontiers of machine learning theory. Topics include empirical processes, concentration of measure, margin-based algorithms, and recent developments in deep learning and optimization theory. The course is designed primarily for PhD students working in machine learning theory and assumes strong mathematical maturity and prior research exposure in the area.

Logistics

Instructor: Jason D. Lee
GSI: Hengyu Fu
Time: Tuesday/Thursday, 3:30–5:00 p.m. Lectures begin at 3:40 p.m. and meet without a break.
Location: Social Sciences Building 20

Enrollment

Students who would like to take the course but are currently unable to enroll should complete the enrollment permission request form.

Prerequisites

Students should be comfortable with graduate-level probability and statistics, algorithms, optimization, and control. Relevant Berkeley courses include STAT 210A/B, CS 270, CS 272, EE 227, and EE 221A.

Coursework

The course grade is based on lecture scribing (10%), a proof-based midterm exam (40%), and a research project with proposals and peer review (50%). The final project should present novel research related to the course topics; literature surveys are outside the project scope.