Hybrid Graph Benchmark (HGB)
收藏资源简介:
Hybrid Graph Benchmark (HGB) 是一个包含23个真实世界混合图数据集的综合性基准,涵盖生物学、社交媒体和电子商务等多个领域。这些数据集不仅包含简单的节点对关系,还包含更复杂的节点交互,如超图和层次图结构。HGB的创建旨在为评估和训练图神经网络(GNNs)提供一个全面的框架,特别是在处理复杂图结构时。通过这些数据集,研究者可以探索和验证不同GNN模型在处理高阶图数据时的性能和局限性,从而推动图表示学习领域的进一步发展。
Hybrid Graph Benchmark (HGB) is a comprehensive benchmark comprising 23 real-world hybrid graph datasets spanning multiple domains including biology, social media, and e-commerce. These datasets not only include simple pairwise node relationships but also more complex node interactions such as hypergraph and hierarchical graph structures. The HGB was created to provide a comprehensive framework for evaluating and training graph neural networks (GNNs), particularly when handling complex graph structures. With these datasets, researchers can explore and validate the performance and limitations of diverse GNN models when processing high-order graph data, thus promoting further advancements in the field of graph representation learning.

- 1Hybrid Graph: A Unified Graph Representation with Datasets and Benchmarks for Complex Graphs帝国理工学院生物工程系 · 2024年



