Learning to Make Predictions on Graphs with Autoencoders

TL;DR

This paper introduces a novel autoencoder architecture for graph representation learning that simultaneously performs link prediction and node classification, showing significant improvements on multiple benchmark datasets.

Contribution

The paper presents a new autoencoder model that jointly learns local graph structure and node features in a single end-to-end training process for multiple graph tasks.

Findings

Significant performance improvements over existing methods on nine benchmark datasets.

Efficient end-to-end training for multi-task graph learning.

Effective joint representation of graph structure and node features.

Abstract

We examine two fundamental tasks associated with graph representation learning: link prediction and semi-supervised node classification. We present a novel autoencoder architecture capable of learning a joint representation of both local graph structure and available node features for the multi-task learning of link prediction and node classification. Our autoencoder architecture is efficiently trained end-to-end in a single learning stage to simultaneously perform link prediction and node classification, whereas previous related methods require multiple training steps that are difficult to optimize. We provide a comprehensive empirical evaluation of our models on nine benchmark graph-structured datasets and demonstrate significant improvement over related methods for graph representation learning. Reference code and data are available at https://github.com/vuptran/graph-representation-learning