GNNs for Node Clustering in Signed and Directed Networks

Yixuan He

Abstract:

With an increasing number of applications where data can be represented as graphs, graph neural networks are a useful tool to apply deep learning to graph data. In particular, node clustering is an important problem in network analysis. Signed and directed networks are important types of networks that are linked to many real-world problems; their asymmetry provides a challenge for many clustering methods.

We propose two graph neural network models for node clustering in signed networks and directed networks, respectively. The methods are end-to-end in combining embedding generation and clustering without an intermediate step. Experimental results on a synthetic signed stochastic block model, a polarized version of it, and real-world data at different scales, demonstrate that our proposed methods can achieve comparable or better results than state-of-the-art node clustering methods, for a wide range of noise and sparsity levels. The introduced models complement existing well-performing methods through the possibility of including exogenous information, in the form of node-level features or labels.