Variational Graph Auto-Encoders

TL;DR

The paper introduces the variational graph auto-encoder (VGAE), a novel unsupervised learning framework for graph data that leverages latent variables and graph convolutional networks to improve link prediction performance.

Contribution

It presents the first variational auto-encoder framework for graphs that incorporates node features and learns interpretable latent representations.

Findings

Achieves competitive link prediction results on citation networks.

Incorporating node features significantly improves predictive accuracy.

Provides a scalable and flexible approach for unsupervised graph learning.

Abstract

We introduce the variational graph auto-encoder (VGAE), a framework for unsupervised learning on graph-structured data based on the variational auto-encoder (VAE). This model makes use of latent variables and is capable of learning interpretable latent representations for undirected graphs. We demonstrate this model using a graph convolutional network (GCN) encoder and a simple inner product decoder. Our model achieves competitive results on a link prediction task in citation networks. In contrast to most existing models for unsupervised learning on graph-structured data and link prediction, our model can naturally incorporate node features, which significantly improves predictive performance on a number of benchmark datasets.