gHAWK: Local and Global Structure Encoding for Scalable Training of Graph Neural Networks on Knowledge Graphs

2025-12-09

AI Summary

Existing message-passing Graph Neural Networks (GNNs) face scalability bottlenecks when trained on large-scale knowledge graphs (KGs): mini-batch training limits node perception to local neighborhood subsets, while iterative message passing suffers from low efficiency. This paper proposes gHAWK, a scalable GNN framework that precomputes and fuses both local and global structural priors to alleviate reliance on message passing. Specifically, gHAWK encodes local neighborhood topology via Bloom filters and captures global relational structure using TransE embeddings, jointly injecting these priors into node representations. This design enables efficient mini-batch training and inference, and is architecture-agnostic—compatible with diverse GNN backbones. Evaluated on multiple large-scale OGB-KG benchmarks, gHAWK achieves state-of-the-art performance on three graph learning leaderboards, attaining higher accuracy and significantly reduced training time for both node and link prediction tasks.

Technology Category

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

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

Knowledge Graphs (KGs) are a rich source of structured, heterogeneous data, powering a wide range of applications. A common approach to leverage this data is to train a graph neural network (GNN) on the KG. However, existing message-passing GNNs struggle to scale to large KGs because they rely on the iterative message passing process to learn the graph structure, which is inefficient, especially under mini-batch training, where a node sees only a partial view of its neighborhood. In this paper, we address this problem and present gHAWK, a novel and scalable GNN training framework for large KGs. The key idea is to precompute structural features for each node that capture its local and global structure before GNN training even begins. Specifically, gHAWK introduces a preprocessing step that computes: (a)~Bloom filters to compactly encode local neighborhood structure, and (b)~TransE embeddings to represent each node's global position in the graph. These features are then fused with any domain-specific features (e.g., text embeddings), producing a node feature vector that can be incorporated into any GNN technique. By augmenting message-passing training with structural priors, gHAWK significantly reduces memory usage, accelerates convergence, and improves model accuracy. Extensive experiments on large datasets from the Open Graph Benchmark (OGB) demonstrate that gHAWK achieves state-of-the-art accuracy and lower training time on both node property prediction and link prediction tasks, topping the OGB leaderboard for three graphs.

Problem

• Scalable GNN training on large knowledge graphs
• Precomputing local and global structural node features
• Reducing memory usage and accelerating convergence

Innovation

• Precomputes local Bloom filters for neighborhood encoding
• Uses TransE embeddings to capture global graph structure
• Fuses structural features with domain-specific node attributes