10-726 Machine Learning Theory: The Art of Problem Formulation and Theorem Proving

12 units, Carnegie Mellon University

This course is an introduction to theoretical machine learning for students who either want to pursue research in machine learning theory, or want to be able to include rigorous theoretical contributions in methods and empirical projects. Rather than cover an exhaustive list of topics in machine learning theory, our aim is to walk students through a few select modules which will emphasize both (a) how to write down a crisp and interesting theoretical question, and (b) how to resolve the question via a rigorous proof. Topics will include: PAC Learning, Online Learning, Active Learning, The SQ Model and Computational/Statistical Tradeoffs, Finite Sample Analysis, basics of deep learning theory, and learning of more complex objects (e.g., algorithms)

Prerequisites: none listed

Offered: Fall 2026, Fall 2025

Courseweave needs JavaScript to show sections, workloads and planning tools.