Hierarchical Maximum-Margin Clustering

TL;DR

This paper introduces a hierarchical maximum-margin clustering approach that recursively segments data in a top-down fashion, leveraging feature regularizers to produce meaningful hierarchies, outperforming flat clustering methods.

Contribution

The paper proposes a novel hierarchical clustering algorithm with a greedy splitting criterion and regularizers for feature sharing, advancing beyond existing flat maximum-margin clustering techniques.

Findings

Outperforms flat and hierarchical baselines on four datasets

Produces semantically meaningful and clean cluster hierarchies

Effective in capturing data semantics through feature regularization

Abstract

We present a hierarchical maximum-margin clustering method for unsupervised data analysis. Our method extends beyond flat maximum-margin clustering, and performs clustering recursively in a top-down manner. We propose an effective greedy splitting criteria for selecting which cluster to split next, and employ regularizers that enforce feature sharing/competition for capturing data semantics. Experimental results obtained on four standard datasets show that our method outperforms flat and hierarchical clustering baselines, while forming clean and semantically meaningful cluster hierarchies.