Clustering on the Edge: Learning Structure in Graphs

TL;DR

This paper introduces a graph clustering method that leverages edge features to automatically determine the number of clusters and handle large-scale data, with applications in image segmentation, community discovery, and entity resolution.

Contribution

It extends the planted partition model to incorporate edge features, enabling simultaneous structure learning and cluster number estimation.

Findings

Achieves partitions close to the true clustering log-likelihood

Effectively handles large numbers of clusters

Utilizes labeled edges for improved clustering accuracy

Abstract

With the recent popularity of graphical clustering methods, there has been an increased focus on the information between samples. We show how learning cluster structure using edge features naturally and simultaneously determines the most likely number of clusters and addresses data scale issues. These results are particularly useful in instances where (a) there are a large number of clusters and (b) we have some labeled edges. Applications in this domain include image segmentation, community discovery and entity resolution. Our model is an extension of the planted partition model and our solution uses results of correlation clustering, which achieves a partition O(log(n))-close to the log-likelihood of the true clustering.