图神经网络(GNNs)凭借其通过聚合局部结构信息学习节点表示的能力,在建模图结构数据方面日益流行。但学界普遍认为,测试图结构可能与训练图结构存在差异,进而引发结构偏移。本文通过实验发现,当发生结构偏移时,GNNs的性能会显著下降,这表明所学习的模型可能对特定结构模式存在偏倚。为应对这一挑战,我们提出了簇信息迁移(Cluster Information Transfer,CIT)机制(代码可在https://github.com/BUPT-GAMMA/CITGNN获取),该机制能够为GNNs学习不变表示,从而提升其对存在结构偏移的各类未知测试图的泛化能力。CIT机制通过将不同的簇信息与节点相结合,同时保留节点的簇无关信息来实现这一目标。通过生成跨不同簇的节点,该机制显著提升了节点的多样性,帮助GNNs学习到不变表示。我们对CIT机制进行了理论分析,结果表明,在结构偏移过程中,簇变化带来的影响可在迁移后得到缓解。此外,所提出的机制是一种即插即用组件,可便捷地用于改进现有GNNs。我们在三种典型的结构偏移场景下对所提方法进行了全面评估,证明了其在提升GNNs性能方面的有效性。
英文摘要:
Graph neural networks (GNNs) have become increasingly popular in modeling graph-structured data due to their ability to learn node representations by aggregating local structure information. However, it is widely acknowledged that the test graph structure may differ from the training graph structure, resulting in a structure shift. In this paper, we experimentally find that the performance of GNNs drops significantly when the structure shift happens, suggesting that the learned models may be biased towards specific structure patterns. To address this challenge, we propose the Cluster Information Transfer (CIT) mechanism (Code available at https://github.com/BUPT-GAMMA/CITGNN), which can learn invariant representations for GNNs, thereby improving their generalization ability to various and unknown test graphs with structure shift. The CIT mechanism achieves this by combining different cluster information with the nodes while preserving their cluster-independent information. By generating nodes across different clusters, the mechanism significantly enhances the diversity of the nodes and helps GNNs learn the invariant representations. We provide a theoretical analysis of the CIT mechanism, showing that the impact of changing clusters during structure shift can be mitigated after transfer. Additionally, the proposed mechanism is a plug-in that can be easily used to improve existing GNNs. We comprehensively evaluate our proposed method on three typical structure shift scenarios, demonstrating its effectiveness in enhancing GNNs' performance.
Learning to Learn Domain-invariant Parameters for Domain Generalization
Institute of Computing Technology, Chinese Academy of Sciences(中国科学院计算技术研究所);University of Chinese Academy of Sciences(中国科学院大学);Southeast University(东南大学);Lenovo Research(联想研究院);Lenovo Ltd(联想有限公司)
Out-of-distribution Generalization for Total Variation based Invariant Risk Minimization
College of Cyber Security, Jinan University(暨南大学网络安全学院);Pratt School of Engineering, Duke University(杜克大学普拉特工程学院);Sino-French Engineer School, Beihang University(北京航空航天大学中法工程师学院)
Counterfactual Supervision-based Information Bottleneck for Out-of-Distribution Generalization
South China University of Technology(华南理工大学)
学习不变(因果)特征的分布外(OOD)推广最近引起了广泛的关注,其中不变风险最小化(VRM)是一个值得注意的解决方案。尽管线性回归理论上有希望,但在线性分类问题中使用递归的挑战仍然存在。通过将信息瓶颈(IB)原理引入到机器学习中,IB-SVM方法已经证明了它解决这些挑战的能力。本文从两个方面进一步改进了IB-SVM。首先,我们证明了在IB-EQUIPMENT中使用的不变特征的支持重叠的关键假设对于保证OOD泛化是强有力的,并且在没有这个假设的情况下仍然有可能实现最优解。其次,我们说明了两种故障模式,IB-ESTA(和ESTA)可能会失败的学习不变的功能,并解决这样的故障,我们提出了一个\textit{Counterfactual Supervision-based Information Bottleneck (CSIB)}学习算法,可证明恢复不变的功能。通过要求反事实推理,CSIB即使在从单个环境访问数据时也能工作。在多个数据集上的实验验证了我们的理论结果。
Improved Generalization Bounds of Group Invariant / Equivariant Deep Networks via Quotient Feature Spaces
RIKEN Center for Advanced Intelligence Project(理化学研究所先进智能项目中心);Keio University(庆应义塾大学);The University of Tokyo(东京大学)
许多不变(或等变)神经网络已经成功地处理不变的数据,如点云和图形。然而,神经网络的推广理论还没有得到很好的发展,因为该理论的几个基本因素,如网络大小和边缘分布,没有深入联系的不变性和等变性。在这项研究中,我们开发了一个新的泛化误差界不变和等变深度神经网络。为了描述不变性和等变性对推广的影响,我们提出了一个\textit{quotient feature space}的概念,它度量了群作用对性质的影响。我们的主要结果证明了商特征空间的体积可以描述泛化误差。此外,该界表明,不变性和等变性显着改善的领先项的界。我们将我们的结果应用于特定的不变和等变网络,例如DeepSets(Zaheer et al.(2017)),并表明它们的泛化界被$\sqrt{n!}$大大提高,其中$n!$是排列的数量。我们还讨论了不变DNN的表达能力,并表明它们可以实现最佳近似率。我们的实验结果支持了我们的理论主张。
Invariant Information Bottleneck for Domain Generalization
Microsoft Research Asia(微软亚洲研究院);Hong Kong University of Science and Technology(香港科技大学);Washington University in St. Louis(圣路易斯华盛顿大学);University of California, Berkeley(加州大学伯克利分校);University of Illinois at Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校)
Domain-invariant Feature Exploration for Domain Generalization
Institute of Computing Technology, Chinese Academy of Sciences(中国科学院计算技术研究所);Microsoft Research Asia(微软亚洲研究院);City University of Hong Kong(香港城市大学);Peng Cheng Laboratory(鹏城实验室)
Winning Prize Comes from Losing Tickets: Improve Invariant Learning by Exploring Variant Parameters for Out-of-Distribution Generalization
Sydney AI Centre, The University of Sydney(悉尼大学悉尼人工智能中心);JD Explore Academy(京东探索研究院);University of Science and Technology of China(中国科学技术大学);Nanjing University of Science and Technology(南京理工大学);Hong Kong Baptist University(香港浸会大学)
Raising the Bar in Graph OOD Generalization: Invariant Learning Beyond Explicit Environment Modeling
The School of Artificial Intelligence, Jilin University(吉林大学人工智能学院);The School of Information and Communication Technology, Griffith University(格里菲斯大学信息与通信技术学院)
Adversarial Label Invariant Graph Data Augmentations for Out-of-Distribution Generalization
Department of Computer Science, Purdue University(普渡大学计算机科学系);Department of Computer Science and Engineering, The Ohio State University(俄亥俄州立大学计算机科学与工程系);Department of Industrial and Systems Engineering, The Ohio State University(俄亥俄州立大学工业与系统工程系)
Out-of-distribution Generalization with Causal Invariant Transformations
University of Chinese Academy of Sciences(中国科学院大学);Academy of Mathematics and Systems Science, Chinese Academy of Sciences(中国科学院数学与系统科学研究院);Huawei Noah’s Ark Lab(华为诺亚方舟实验室)
Domain Generalization via Conditional Invariant Representation
University of Science and Technology of China(中国科学技术大学);The University of Sydney(悉尼大学);Carnegie Mellon University(卡内基梅隆大学);University of Pittsburgh(匹兹堡大学)
Class-Invariant Test-Time Augmentation for Domain Generalization
School of Computing and Artificial Intelligence(计算机与人工智能学院);ANGEL Lab(ANGEL实验室);Southwest Jiaotong University(西南交通大学);Nanyang Technological University(南洋理工大学)
Invariant Graph Learning Meets Information Bottleneck for Out-of-Distribution Generalization
School of Cyber Science and Technology, University of Science and Technology of China(中国科学技术大学网络空间安全学院);Tencent(腾讯);School of Artificial Intelligence and Data Science, University of Science and Technology of China(中国科学技术大学人工智能与数据科学学院)