arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2303.14859cs.LGcs.AI
收藏

基于增广的图OOD泛化中的标签偏移

Mind the Label Shift of Augmentation-based Graph OOD Generalization

Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);University of Chinese Academy of Sciences(中国科学院大学);ShanghaiTech University(上海科技大学)

Junchi Yu, Jian Liang, Ran He

摘要:

分布外泛化是图神经网络的一个重要问题。最近的工作采用不同的图形版本来生成增强环境,并学习一个不变的GNN进行泛化。然而,由于图的结构化编辑不可避免地要改变图的标号,所以在扩充过程中经常会发生标签偏移。这带来了增强环境之间的不一致的预测关系,这是有害的泛化。为了解决这个问题,我们提出了\textbf{LiSA},它产生标签不变的扩增,以促进图OOD推广。LiSA利用训练图的\textbf{L} abel- \textbf{i} nvariant \textbf{S}子图来构造\textbf{A}增强环境,而不是诉诸图版本。具体来说,LiSA首先设计了变分子图生成器,以有效地提取局部预测模式并构造多个标签不变子图。然后,收集由不同生成器产生的子图,以构建不同的增强环境。为了促进增强环境之间的多样性,LiSA进一步引入了一种易于处理的基于能量的正则化,以扩大环境分布之间的成对距离。通过这种方式,LiSA生成具有一致预测关系的各种增强环境,并有助于学习不变的GNN。在节点级和图级OOD基准测试上的大量实验表明,LiSA在不同的GNN主干上实现了令人印象深刻的泛化性能。代码可在\url{https://github.com/Samyu0304/LiSA}上获得。

英文摘要:

Out-of-distribution (OOD) generalization is an important issue for Graph Neural Networks (GNNs). Recent works employ different graph editions to generate augmented environments and learn an invariant GNN for generalization. However, the label shift usually occurs in augmentation since graph structural edition inevitably alters the graph label. This brings inconsistent predictive relationships among augmented environments, which is harmful to generalization. To address this issue, we propose \textbf{LiSA}, which generates label-invariant augmentations to facilitate graph OOD generalization. Instead of resorting to graph editions, LiSA exploits \textbf{L}abel-\textbf{i}nvariant \textbf{S}ubgraphs of the training graphs to construct \textbf{A}ugmented environments. Specifically, LiSA first designs the variational subgraph generators to extract locally predictive patterns and construct multiple label-invariant subgraphs efficiently. Then, the subgraphs produced by different generators are collected to build different augmented environments. To promote diversity among augmented environments, LiSA further introduces a tractable energy-based regularization to enlarge pair-wise distances between the distributions of environments. In this manner, LiSA generates diverse augmented environments with a consistent predictive relationship and facilitates learning an invariant GNN. Extensive experiments on node-level and graph-level OOD benchmarks show that LiSA achieves impressive generalization performance with different GNN backbones. Code is available on \url{https://github.com/Samyu0304/LiSA}.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑