CS 6180 - Clustering

Clustering is the most basic unsupervised learning task: Organize a collection of objects into groups so that objects in the same group are similar. But almost every word in the sentence above has different specific meanings, leading to different ways to thinking about clusterings. What is similar? What is a group? What are the objects being grouped? And what does it even mean to organize them? In this course, we will develop a conceptual understanding of the problems of clustering such as: What's the best clustering formulation for a given problem? What kinds of clusters might I expect (or will I end up with) if I use a particular formulation? What kinds of formal guarantees can we provide on the behavior of a particular algorithm for a clustering formulation? How do we extract meaning from the results? And given that clustering is usually unsupervised, how do we know that we have a good answer?