OGB-LSC: A Large-Scale Challenge for Machine Learning on Graphs
Weihua Hu1, Matthias Fey2, Hongyu Ren1, Maho Nakata3, Yuxiao Dong4, Jure Leskovec1 1Department of Computer Science, Stanford University 2Department of Computer Science, TU Dortmund University 3RIKEN, 4Facebook AI
Abstract
Enabling effective and efficient machine learning (ML) over large-scale graph data (e.g., graphs with billions of edges) can have a great impact on both industrial and scientific applications. However, existing efforts to advance large-scale graph ML have been largely limited by the lack of a suitable public benchmark. Here we present OGB Large-Scale Challenge (OGB-LSC), a collection of three real-world datasets for facilitating the advancements in large-scale graph ML. The OGB-LSC datasets are orders of magnitude larger than existing ones, covering three core graph learning tasks—link prediction, graph regression, and node classification. Furthermore, we provide dedicated baseline experiments, scaling up expressive graph ML models to the massive datasets. We show that expressive models significantly outperform simple scalable baselines, indicating an opportunity for dedicated efforts to further improve graph ML at scale. Moreover, OGB-LSC datasets were deployed at ACM KDD Cup 2021 and attracted more than 500 team registrations globally, during which significant performance improvements were made by a variety of innovative techniques. We summarize the common techniques used by the
中文速览
大规模图数据上的机器学习一直面临"好模型跑不动、能跑的模型效果差"的两难困境，根本原因在于缺乏真正大规模的公开基准数据集。为此，研究者推出了OGB大规模挑战赛（OGB-LSC），提供三个来自真实世界的超大图数据集——涵盖学术论文引用网络上的节点分类、知识图谱补全的链接预测、以及分子性质预测的图回归，规模比此前同类数据集大10到100倍。大量基线实验表明，表达能力强的图神经网络（Graph Neural Network, GNN）虽然扩展到大规模数据有难度，但其性能显著优于为了追求可扩展性而简化的模型，说明在大图上探索强表达力模型仍有巨大潜力。这些数据集作为2021年ACM KDD Cup的竞赛题目，吸引了全球500余支队伍参赛，各团队通过标签传播、自监督预训练、模型集成等创新手段大幅刷新了性能上限，为大规模图机器学习的最佳实践提供了重要参考。
原文 arXiv:2103.09430;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2103.09430v3