Non-Homophilous-Benchmarks

Non-Homophilous-Benchmarks数据集由康奈尔大学创建，旨在为非同质性图上的机器学习任务提供基准。该数据集包含7个子数据集，涵盖多种应用场景，如社交网络、生物信息和欺诈检测。每个数据集都具有不同的节点和边特征，以及复杂的标签-拓扑关系。这些数据集的创建旨在解决现有数据集在非同质性图学习中的局限性，如规模小和应用范围狭窄。通过这些数据集，研究者可以更好地理解和开发适用于非同质性图的机器学习模型。

The Non-Homophilous-Benchmarks dataset was developed by Cornell University to provide benchmarks for machine learning tasks on non-homophilous graphs. This dataset includes seven sub-datasets covering diverse application scenarios such as social networks, bioinformatics, and fraud detection. Each sub-dataset features distinct node and edge attributes, as well as complex label-topology relationships. These datasets are designed to address the limitations of existing datasets in non-homophilous graph learning, such as small scale and narrow application scope. These datasets enable researchers to better understand and develop machine learning models tailored for non-homophilous graphs.