GRAPHTESTER

GRAPHTESTER是一个用于分析图神经网络（GNNs）理论能力的新工具，由苏黎世联邦理工学院的研究团队开发。该数据集包含超过40个不同的图数据集，用于确定不同GNNs的性能上限，支持边缘特征、不同的性能度量、GNN层数和更高阶的Weisfeiler-Lehman测试。此外，GRAPHTESTER还提供了一个合成数据集，用于基准测试节点和边缘特征，如位置编码。数据集的应用领域包括图分类、节点分类和链接预测等，旨在解决图数据分析中的复杂问题。

GRAPHTESTER is a novel tool for analyzing the theoretical capabilities of Graph Neural Networks (GNNs), developed by a research team at ETH Zurich. This dataset includes over 40 distinct graph datasets, which are used to determine the performance upper bounds of different GNNs. It supports edge features, various performance metrics, GNN layer counts, and higher-order Weisfeiler-Lehman tests. Additionally, GRAPHTESTER provides a synthetic dataset for benchmarking node and edge features such as positional encodings. Its application domains include graph classification, node classification, link prediction, and other related tasks, aiming to address complex problems in graph data analysis.