Gaussian Relational Graph Transformer
For practitioners of relational graph learning, this method addresses the challenge of capturing long-range dependencies and integrating multiple information types, offering a new state-of-the-art approach.
GelGT introduces a Gaussian relational graph transformer that jointly models structural, semantic, and temporal information, achieving up to 13.8% improvement in predictive performance on relational graph learning tasks.
Relational graph learning models relational databases as graphs and has demonstrated superior performance on a wide range of relational predictive tasks. However, existing methods struggle to capture long-range dependencies due to information decay in their message-passing mechanisms, and recent relational graph transformers remain limited in jointly modeling structural, semantic, and temporal information. In this paper, we propose GelGT, a Gaussian relational graph transformer that explicitly addresses these challenges. GelGT introduces a structure-semantic collaborative sampling strategy to preserve structural connectivity while filtering irrelevant semantic information, and incorporates a Gaussian graph attention mechanism with a learnable Gaussian bias on the sampled subgraphs to dynamically encode temporal dependencies. Extensive experiments on various real-world datasets demonstrate that GelGT achieves state-of-the-art downstream task performance, with up to a 13.8% improvement in predictive performance.