Open Graph Benchmark (OGB)

Open Graph Benchmark (OGB) 是一个多样化的挑战性实际基准数据集，旨在促进图机器学习（ML）研究的规模化、鲁棒性和可重复性。OGB 数据集规模大，涵盖多个重要的图 ML 任务，并覆盖从社交和信息网络到生物网络、分子图、源代码 AST 和知识图谱等多个领域。

Open Graph Benchmark (OGB) is a diverse and challenging real-world benchmark dataset designed to advance the scalability, robustness, and reproducibility of graph machine learning (ML) research. The OGB dataset has a large scale, encompasses a range of critical graph ML tasks, and spans multiple domains ranging from social and information networks, biological networks, molecular graphs, abstract syntax trees (ASTs) of source code, to knowledge graphs.

搜集汇总

背景与挑战

背景概述

Open Graph Benchmark (OGB) 是一个专为图机器学习设计的基准数据集集合，提供现实、大规模且多样化的数据集，覆盖节点、链接和图属性预测等任务。它集成了自动化的数据加载器和统一的评估工具，方便研究人员进行模型比较和开发。