Graph Condensation for Graph Neural Networks

TL;DR

This paper introduces a graph condensation method that creates small, synthetic graphs capturing essential information of large graphs, enabling efficient GNN training with minimal performance loss.

Contribution

It proposes a novel framework for graph condensation that imitates GNN training trajectories via gradient matching, significantly reducing graph size while maintaining accuracy.

Findings

Achieves 95.3% test accuracy on Reddit with over 99.9% size reduction

Maintains 99.8% and 99.0% accuracy on Flickr and Citeseer respectively

Effective across multiple GNN architectures and datasets

Abstract

Given the prevalence of large-scale graphs in real-world applications, the storage and time for training neural models have raised increasing concerns. To alleviate the concerns, we propose and study the problem of graph condensation for graph neural networks (GNNs). Specifically, we aim to condense the large, original graph into a small, synthetic and highly-informative graph, such that GNNs trained on the small graph and large graph have comparable performance. We approach the condensation problem by imitating the GNN training trajectory on the original graph through the optimization of a gradient matching loss and design a strategy to condense node futures and structural information simultaneously. Extensive experiments have demonstrated the effectiveness of the proposed framework in condensing different graph datasets into informative smaller graphs. In particular, we are able to approximate the original test accuracy by 95.3% on Reddit, 99.8% on Flickr and 99.0% on Citeseer, while reducing their graph size by more than 99.9%, and the condensed graphs can be used to train various GNN architectures.Code is released at https://github.com/ChandlerBang/GCond.