Enhancing Logical Expressiveness in Graph Neural Networks via Path-Neighbor Aggregation

2025-11-11

AI Summary

Graph Neural Networks (GNNs) exhibit limited logical expressivity for knowledge graph (KG) reasoning, particularly in capturing multi-hop logical rules. Method: This paper proposes Path-Neighbor Enhanced GNN (PN-GNN), the first framework to systematically analyze and enhance GNNs’ logical expressivity in KGs. PN-GNN explicitly models path structures and neighborhood semantics by aggregating node representations along inference paths as well as their multi-hop neighbors. Contribution/Results: We theoretically prove that PN-GNN is strictly more expressive than canonical GNNs (C-GNNs) and that its (k+1)-hop expressivity strictly surpasses k-hop expressivity, significantly improving multi-hop logical reasoning capability. Experiments on six synthetic and two real-world KG benchmarks demonstrate that PN-GNN achieves superior logical expressivity and generalization, yielding state-of-the-art performance on link prediction and other KG reasoning tasks.

Technology Category

Knowledge Representation and Reasoning: Computational Complexity of ReasoningData Mining & Knowledge Management: Graph Mining, Social Network Analysis & CommunityMachine Learning: Graph-based Machine Learning

Application Category

Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphsSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Graph neural networks (GNNs) can effectively model structural information of graphs, making them widely used in knowledge graph (KG) reasoning. However, existing studies on the expressive power of GNNs mainly focuses on simple single-relation graphs, and there is still insufficient discussion on the power of GNN to express logical rules in KGs. How to enhance the logical expressive power of GNNs is still a key issue. Motivated by this, we propose Path-Neighbor enhanced GNN (PN-GNN), a method to enhance the logical expressive power of GNN by aggregating node-neighbor embeddings on the reasoning path. First, we analyze the logical expressive power of existing GNN-based methods and point out the shortcomings of the expressive power of these methods. Then, we theoretically investigate the logical expressive power of PN-GNN, showing that it not only has strictly stronger expressive power than C-GNN but also that its $(k+1)$-hop logical expressiveness is strictly superior to that of $k$-hop. Finally, we evaluate the logical expressive power of PN-GNN on six synthetic datasets and two real-world datasets. Both theoretical analysis and extensive experiments confirm that PN-GNN enhances the expressive power of logical rules without compromising generalization, as evidenced by its competitive performance in KG reasoning tasks.

Problem

• Enhancing logical expressiveness of Graph Neural Networks
• Addressing limitations in knowledge graph reasoning capabilities
• Improving path-neighbor aggregation for complex logical rules

Innovation

• Enhances GNNs via path-neighbor aggregation
• Theoretically proves superior logical expressiveness over C-GNN
• Maintains generalization while improving KG reasoning performance