The paper discusses the challenges faced by existing HGNNs in handling complex heterogeneous graphs with numerous relation types. It introduces BG-HGNN as a solution that integrates different relations into a unified feature space, improving efficiency and performance significantly. The empirical studies show that BG-HGNN outperforms existing models in terms of parameter efficiency, training throughput, and accuracy.
他の言語に翻訳
原文コンテンツから
arxiv.org
深掘り質問