A Potts model approach to unsupervised graph clustering with Graph Neural Networks

TL;DR

This paper introduces a novel GNN framework based on the Potts model to improve unsupervised graph clustering, effectively addressing the resolution limit problem inherent in modularity-based methods.

Contribution

The paper proposes a new GNN approach inspired by the Potts model, overcoming the resolution limit in modularity-based graph clustering.

Findings

Achieves state-of-the-art clustering results on real-world datasets.

Overcomes the resolution limit problem in modularity-based clustering.

Demonstrates improved detection of smaller clusters.

Abstract

Numerous approaches have been explored for graph clustering, including those which optimize a global criteria such as modularity. More recently, Graph Neural Networks (GNNs), which have produced state-of-the-art results in graph analysis tasks such as node classification and link prediction, have been applied for unsupervised graph clustering using these modularity-based metrics. Modularity, though robust for many practical applications, suffers from the resolution limit problem, in which optimization may fail to identify clusters smaller than a scale that is dependent on properties of the network. In this paper, we propose a new GNN framework which draws from the Potts model in physics to overcome this limitation. Experiments on a variety of real world datasets show that this model achieves state-of-the-art clustering results.