Local algorithms for interactive clustering

TL;DR

This paper introduces local, interactive clustering algorithms that start from any initial partition, make incremental adjustments, and are both theoretically efficient and effective on real-world datasets.

Contribution

It presents a new class of provably efficient local algorithms for interactive clustering under natural stability assumptions.

Findings

Algorithms produce accurate clusterings.

Algorithms perform well on real-world data.

Efficient in the constrained setting.

Abstract

We study the design of interactive clustering algorithms for data sets satisfying natural stability assumptions. Our algorithms start with any initial clustering and only make local changes in each step; both are desirable features in many applications. We show that in this constrained setting one can still design provably efficient algorithms that produce accurate clusterings. We also show that our algorithms perform well on real-world data.