arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2306.00899cs.LGcs.IRcs.SI
收藏

图神经网络链接预测中的陷阱:理解目标链接纳入的影响与更优实践

Pitfalls in Link Prediction with Graph Neural Networks: Understanding the Impact of Target-link Inclusion & Better Practices

University of Michigan, Ann Arbor(密歇根大学安娜堡分校);University of Maryland, College Park(马里兰大学帕克分校);AWS AI Research and Education(AWS人工智能研究与教育)

Jing Zhu, Yuhang Zhou, Vassilis N. Ioannidis, Shengyi Qian, Wei Ai, Xiang Song, Danai Koutra

AI总结:

本文揭示GNN链接预测中纳入目标边导致的过拟合、分布偏移与测试泄漏问题,并提出SpotTarget框架,通过训练和测试时排除相应目标边显著提升稀疏与稠密图中低度节点的预测性能。

摘要:

尽管图神经网络(GNN)在多种高影响力应用中取得了显著成功,但我们证明,在链接预测中,在训练和/或测试时将待预测边纳入图中的常见做法,会对低度节点的性能产生过大影响。我们从理论和实证两方面研究这些做法如何影响不同度数下的节点级性能。具体而言,我们探讨了由此产生的三个问题:(I1)过拟合;(I2)分布偏移;以及(I3)隐式测试泄漏。前两个问题导致对测试数据的泛化能力较差,而后者则导致模型性能被高估,并直接影响 GNN 的部署。为系统解决这些问题,我们提出了一个有效且高效的 GNN 训练框架 SpotTarget,该框架利用了我们对低度节点的洞察:(1)在训练时,如果一条待预测(训练)边至少与一个低度节点相连,则将其排除;(2)在测试时,排除所有待预测的测试边(从而模拟使用 GNN 的真实场景,即测试数据不包含在图中)。SpotTarget 帮助研究人员和从业者遵循从图数据中学习的最佳实践,而这些实践即使是最广泛使用的框架也经常忽视。我们在多个真实世界数据集上的实验表明,SpotTarget 使 GNN 在稀疏图中的准确率最高提升至 15 倍,并显著提升了其在稠密图中针对低度节点的性能。

英文摘要:

While Graph Neural Networks (GNNs) are remarkably successful in a variety of high-impact applications, we demonstrate that, in link prediction, the common practices of including the edges being predicted in the graph at training and/or test have outsized impact on the performance of low-degree nodes. We theoretically and empirically investigate how these practices impact node-level performance across different degrees. Specifically, we explore three issues that arise: (I1) overfitting; (I2) distribution shift; and (I3) implicit test leakage. The former two issues lead to poor generalizability to the test data, while the latter leads to overestimation of the model's performance and directly impacts the deployment of GNNs. To address these issues in a systematic way, we introduce an effective and efficient GNN training framework, SpotTarget, which leverages our insight on low-degree nodes: (1) at training time, it excludes a (training) edge to be predicted if it is incident to at least one low-degree node; and (2) at test time, it excludes all test edges to be predicted (thus, mimicking real scenarios of using GNNs, where the test data is not included in the graph). SpotTarget helps researchers and practitioners adhere to best practices for learning from graph data, which are frequently overlooked even by the most widely-used frameworks. Our experiments on various real-world datasets show that SpotTarget makes GNNs up to 15x more accurate in sparse graphs, and significantly improves their performance for low-degree nodes in dense graphs.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑