Subgroup Generalization and Fairness of Graph Neural Networks
Subgroup Generalization and Fairness of Graph Neural Networks
复制标题
DOI：
--
复制
发表时间：
2021-06
影响因子：
2.7
通讯作者：
Jiaqi Ma;Junwei Deng;Qiaozhu Mei
中科院分区：
文献类型：
--
作者：
Jiaqi Ma;Junwei Deng;Qiaozhu Mei
尽管图神经网络（GNN）的应用取得了巨大的成功，但对其泛化能力的理论理解，特别是对于数据不独立且相同分布（IID）的节点级任务，一直很稀疏。泛化性能的理论研究有利于理解GNN模型的基本问题（如公平性）和设计更好的学习方法。在本文中，我们提出了一种新的PAC贝叶斯分析GNNs下的非IID半监督学习设置。此外，我们分析了未标记节点的不同子群上的泛化性能，这使我们能够从理论角度进一步研究GNN的准确性-（不）公平性。在合理的假设下，我们证明了测试子组和训练集之间的距离可能是影响GNN在该子组上性能的关键因素，这需要特别注意公平学习的训练节点选择。在多个GNN模型和数据集上的实验支持了我们的理论结果。
Despite enormous successful applications of graph neural networks (GNNs), theoretical understanding of their generalization ability, especially for node-level tasks where data are not independent and identically-distributed (IID), has been sparse. The theoretical investigation of the generalization performance is beneficial for understanding fundamental issues (such as fairness) of GNN models and designing better learning methods. In this paper, we present a novel PAC-Bayesian analysis for GNNs under a non-IID semi-supervised learning setup. Moreover, we analyze the generalization performances on different subgroups of unlabeled nodes, which allows us to further study an accuracy-(dis)parity-style (un)fairness of GNNs from a theoretical perspective. Under reasonable assumptions, we demonstrate that the distance between a test subgroup and the training set can be a key factor affecting the GNN performance on that subgroup, which calls special attention to the training node selection for fair learning. Experiments across multiple GNN models and datasets support our theoretical results.