arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.11083cs.LG
收藏

赋能图不变性学习与深度虚假信息最大化

Empowering Graph Invariance Learning with Deep Spurious Infomax

Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学);The Chinese University of Hong Kong(香港中文大学);Carnegie Mellon University(卡内基梅隆大学)

Tianjun Yao, Yongqiang Chen, Zhenhao Chen, Kai Hu, Zhiqiang Shen, Kun Zhang

AI总结:

针对现有图不变性学习对虚假相关强度假设敏感的问题,提出基于信息最大化原理的EQuAD框架,通过分离虚假特征学习不变特征,在合成和真实数据集上性能稳定提升,最高达31.76%。

摘要:

近年来,利用图上的不变性原理来泛化分布外(OOD)数据的图神经网络开发引起了广泛兴趣。由于对OOD数据的了解有限,现有方法常常对底层虚假特征与目标标签之间的相关性强度做出假设。然而,这种先验信息往往不可用,并且在现实场景中会任意变化,这可能导致现有图不变性学习方法严重失败。为弥补这一差距,我们引入了一种新颖的图不变性学习范式,该范式诱导出稳健且通用的归纳偏置。该范式建立在信息最大化原理鼓励学习虚假特征而不管虚假相关性强度的观察之上。我们进一步提出了EQuAD框架,该框架实现了这一学习范式,并采用量身定制的学习目标,通过将不变特征与通过信息最大化学习到的虚假特征分离来可证明地引出不变特征。值得注意的是,EQuAD在合成数据集和具有挑战性的现实世界数据集上,在不同程度的偏差下表现出稳定且增强的性能,最高可达31.76%。我们的代码可在\url{https://github.com/tianyao-aka/EQuAD}获取。

英文摘要:

Recently, there has been a surge of interest in developing graph neural networks that utilize the invariance principle on graphs to generalize the out-of-distribution (OOD) data. Due to the limited knowledge about OOD data, existing approaches often pose assumptions about the correlation strengths of the underlying spurious features and the target labels. However, this prior is often unavailable and will change arbitrarily in the real-world scenarios, which may lead to severe failures of the existing graph invariance learning methods. To bridge this gap, we introduce a novel graph invariance learning paradigm, which induces a robust and general inductive bias. The paradigm is built upon the observation that the infomax principle encourages learning spurious features regardless of spurious correlation strengths. We further propose the EQuAD framework that realizes this learning paradigm and employs tailored learning objectives that provably elicit invariant features by disentangling them from the spurious features learned through infomax. Notably, EQuAD shows stable and enhanced performance across different degrees of bias in synthetic datasets and challenging real-world datasets up to $31.76\%$. Our code is available at \url{https://github.com/tianyao-aka/EQuAD}.

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。

↑