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 2306.15902cs.LGcs.AIcs.CV
收藏

面向分布外泛化的个体与结构图信息瓶颈

Individual and Structural Graph Information Bottlenecks for Out-of-Distribution Generalization

Peking University(北京大学);Ant Group(蚂蚁集团);Peking University (Qingdao)(北京大学(青岛));Mila - Québec AI Institute(米拉-魁北克人工智能研究所);HEC Montréal(蒙特利尔高等商学院)

Ling Yang, Jiayi Zheng, Heyuan Wang, Zhongyi Liu, Zhilin Huang, Shenda Hong, Wentao Zhang, Bin Cui

AI总结:

针对分布外图泛化中虚假特征及实例级学习局限,提出统一框架IS-GIB,结合I-GIB丢弃无关信息与S-GIB从高阶视角学习类关系不变特征,在节点与图级任务上展现卓越泛化能力。

摘要:

分布外(OOD)图泛化对许多现实应用至关重要。现有方法忽略了丢弃与标签无关的虚假或噪声特征。此外,它们主要进行实例级类不变图学习,未能利用图实例间的结构类关系。在此,我们致力于在统一框架中解决这些问题,称为个体与结构图信息瓶颈(IS-GIB)。为消除分布偏移导致的虚假特征,我们提出个体图信息瓶颈(I-GIB),通过最小化输入图与其嵌入间的互信息来丢弃无关信息。为利用结构域内与跨域相关性,我们提出结构图信息瓶颈(S-GIB)。针对多域图批次,S-GIB首先计算成对的输入-输入、嵌入-嵌入及标签-标签相关性。随后,它最小化输入图与嵌入对间的互信息,同时最大化嵌入与标签对间的互信息。S-GIB的核心洞察在于,通过在多重分布偏移下维持类关系,从高阶视角同时丢弃虚假特征并学习不变特征。我们将I-GIB与S-GIB统一构成互补框架IS-GIB。在节点级和图级任务上的大量实验一致表明了IS-GIB卓越的泛化能力。代码见 https://github.com/YangLing0818/GraphOOD。

英文摘要:

Out-of-distribution (OOD) graph generalization are critical for many real-world applications. Existing methods neglect to discard spurious or noisy features of inputs, which are irrelevant to the label. Besides, they mainly conduct instance-level class-invariant graph learning and fail to utilize the structural class relationships between graph instances. In this work, we endeavor to address these issues in a unified framework, dubbed Individual and Structural Graph Information Bottlenecks (IS-GIB). To remove class spurious feature caused by distribution shifts, we propose Individual Graph Information Bottleneck (I-GIB) which discards irrelevant information by minimizing the mutual information between the input graph and its embeddings. To leverage the structural intra- and inter-domain correlations, we propose Structural Graph Information Bottleneck (S-GIB). Specifically for a batch of graphs with multiple domains, S-GIB first computes the pair-wise input-input, embedding-embedding, and label-label correlations. Then it minimizes the mutual information between input graph and embedding pairs while maximizing the mutual information between embedding and label pairs. The critical insight of S-GIB is to simultaneously discard spurious features and learn invariant features from a high-order perspective by maintaining class relationships under multiple distributional shifts. Notably, we unify the proposed I-GIB and S-GIB to form our complementary framework IS-GIB. Extensive experiments conducted on both node- and graph-level tasks consistently demonstrate the superior generalization ability of IS-GIB. The code is available at https://github.com/YangLing0818/GraphOOD.

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。

↑