Overview

Graduate-level course in theory of ML and optimization in/for ML. Tentative focuses include:

Out-of-sample, aka generalization performance of optimization algorithms in ML

• How to define out-of-sample performance?
• Algorithmic tools for designing optimization methods that “generalize”?
• Fundamental limits and pitfalls?
• Questions at the interface of optimization with statistics

Sequential decision making

• Online learning, adversarial environments
• Partial feedback, multi-armed bandits, zero-order optimization
• Reinforcement learning, policy optimization algorithms

Prerequisites

In general: graduate-level mathematical / algorithmic maturity will be assumed

• Material is mostly theoretical/algorithmic/mathematical in nature
• Mathematical maturity: calculus, linear algebra, probability, basic algorithms & complexity
• Be familiar with basics of ML, basics of convex optimization (at the level of “Intro to Machine Learning” or equivalent)

Requirements

• (20%) Attendance and participation