NoisyGL

NoisyGL是由浙江大学和阿里巴巴集团合作创建的综合性基准数据集，专门用于评估图神经网络在标签噪声环境下的性能。该数据集包含8个来自不同领域的节点分类数据集，如Cora、Citeseer等，涵盖了从生物网络到社交网络的多样化数据。创建过程中，研究团队采用了统一的数据分割和预处理策略，确保了数据集的一致性和公平性。NoisyGL的应用领域主要集中在解决图神经网络在标签噪声影响下的鲁棒性问题，为研究者提供了一个标准化的平台，以深入探讨和改进GNN在噪声环境中的表现。

NoisyGL is a comprehensive benchmark dataset co-developed by Zhejiang University and Alibaba Group, specifically dedicated to evaluating the performance of Graph Neural Networks (GNNs) in label-noisy environments. This dataset includes 8 node classification datasets from diverse domains, such as Cora, Citeseer and others, spanning a wide range of data types from biological networks to social networks. During the dataset construction, the research team adopted a unified data splitting and preprocessing protocol to ensure the consistency and fairness of the benchmark. The primary application focus of NoisyGL is to address the robustness issues of GNNs under the influence of label noise, providing researchers with a standardized platform to deeply explore and improve the performance of GNNs in noisy environments.