Virtual Topology Augmentation/虚拟拓扑增强
:根据当前预测,为每个类别创建一个 prototype 虚拟节点(virtual super node)来表示该类的 general pattern。根据前两步的结果,在高风险节点和其对应的实际标签之间构建虚拟边。注意 BAT 具有线性复杂度,因此计算十分高效,我们在每一步训练中都会重新计算 BAT 来实现动态数据增强。
算法流程图,从左至右
六、Experiment/实验
6.1 Performance/性能
我们在论文中报告了 BAT 在 5(数据集)x 5(GNN架构)x 7(类别不平衡图学习技术)x 3(分类性能/偏差指标)x 4(类别不平衡种类)= 2100 种实验组合下的结果,以展示 BAT 的泛用性和有效性。
[1] Krawczyk, Bartosz. "Learning from imbalanced data: open challenges and future directions." Progress in Artificial Intelligence 5.4 (2016): 221-232.
[2] Chen, Deli, et al. "Topology-imbalance learning for semi-supervised node classification." Advances in Neural Information Processing Systems 34 (2021): 29885-29897.
[3] Zhao, Tianxiang, Xiang Zhang, and Suhang Wang. "Graphsmote: Imbalanced node classification on graphs with graph neural networks." Proceedings of the 14th ACM international conference on web search and data mining. 2021.
[4] Park, Joonhyung, Jaeyun Song, and Eunho Yang. "Graphens: Neighbor-aware ego network synthesis for class-imbalanced node classification." International conference on learning representations. 2021.
[5] Liu, Zemin, et al. "A survey of imbalanced learning on graphs: Problems, techniques, and future directions." arXiv preprint arXiv:2308.13821 (2023).
作者:
刘芷宁
来源:公众号【PaperWeekly】
llustration From IconScout By IconScout Store
-The End-