Course Description

The goal of the course is to introduce students to reason about machine learning in a formal way. This involves developing algorithms that come with guarantees on various performance parameters such as solution quality, running time, sample complexity etc. Traditional models from theoretical computer science are not always suitable for this purpose as most machine learning problems are highly intractable in the worst case. Yet, real world data often exhibits special structure that makes these problems efficiently solvable in practice. The course will be aimed at getting students familiar with this style of algorithmic analysis which formulates and utilizes special structure in data to design efficient algorithms with provable guarantees.


Lectures


Project Presentation Schedule

Related Papers