Link prediction with knowledge graphs has been thoroughly studied in graph machine learning, leading to a rich landscape of graph neural network architectures with successful applications. Nonetheless, it remains challenging to transfer the success of these architectures to link prediction with relational hypergraphs. The presence of relational hyperedges makes link prediction a task between $k$ nodes for varying choices of $k$, which is substantially harder than link prediction with knowledge graphs, where every relation is binary ($k=2$). In this paper, we propose two frameworks for link prediction with relational hypergraphs and conduct a thorough analysis of the expressive power of the resulting model architectures via corresponding relational Weisfeiler-Leman algorithms, and also via some natural logical formalisms. Through extensive empirical analysis, we validate the power of the proposed model architectures on various relational hypergraph benchmarks. The resulting model architectures substantially outperform every baseline for inductive link prediction, and lead to state-of-the-art results for transductive link prediction. Our study therefore unlocks applications of graph neural networks to fully relational structures.

翻译：基于知识图谱的链接预测在图表征学习中已得到深入研究，形成了丰富的图神经网络架构体系并取得广泛应用。然而，将这些架构的成功经验迁移至关系超图的链接预测仍面临挑战。关系超边的存在使得链接预测成为涉及$k$个节点（$k$值可变）的任务，这比每条关系均为二元（$k=2$）的知识图谱链接预测困难得多。本文提出了两种面向关系超图链接预测的框架，通过对应的关系型Weisfeiler-Leman算法及自然逻辑形式化方法，对所构建模型架构的表达能力进行了深入分析。通过大量实验验证，我们证实了所提模型架构在多种关系超图基准测试中的有效性。该模型架构在归纳式链接预测任务上显著超越所有基线方法，并在直推式链接预测任务上取得最优结果。本研究因此解锁了图神经网络在全关系结构中的应用。