Algorithmic Datasets for Graph Learning

该数据集旨在评估循环图神经网络（GNNs）在图结构数据上的外推能力，特别关注路径查找和距离等算法任务。数据集包含用于路径查找和距离等任务的不同实例，并用于验证在不同条件下，具有不同架构的GNNs的性能。这些数据集对规模为10的图进行了评估，并在规模高达原图1000倍的图上进行了测试。其核心任务是图算法的外推能力研究。

This dataset aims to evaluate the extrapolation capability of recurrent Graph Neural Networks (GNNs) on graph-structured data, with a particular focus on algorithmic tasks such as pathfinding and distance calculation. It includes diverse instances for tasks including pathfinding and distance-related problems, and is used to verify the performance of GNNs with different architectures under various conditions. This dataset is evaluated on graphs of size 10, and tested on graphs with scales up to 1000 times that of the original graphs. The core objective of this dataset is to investigate the extrapolation capabilities of graph algorithms.