English

Generalizing Graph Neural Networks on Out-Of-Distribution Graphs

Machine Learning 2024-03-12 v4 Artificial Intelligence

Abstract

Graph Neural Networks (GNNs) are proposed without considering the agnostic distribution shifts between training and testing graphs, inducing the degeneration of the generalization ability of GNNs on Out-Of-Distribution (OOD) settings. The fundamental reason for such degeneration is that most GNNs are developed based on the I.I.D hypothesis. In such a setting, GNNs tend to exploit subtle statistical correlations existing in the training set for predictions, even though it is a spurious correlation. However, such spurious correlations may change in testing environments, leading to the failure of GNNs. Therefore, eliminating the impact of spurious correlations is crucial for stable GNNs. To this end, we propose a general causal representation framework, called StableGNN. The main idea is to extract high-level representations from graph data first and resort to the distinguishing ability of causal inference to help the model get rid of spurious correlations. Particularly, we exploit a graph pooling layer to extract subgraph-based representations as high-level representations. Furthermore, we propose a causal variable distinguishing regularizer to correct the biased training distribution. Hence, GNNs would concentrate more on the stable correlations. Extensive experiments on both synthetic and real-world OOD graph datasets well verify the effectiveness, flexibility and interpretability of the proposed framework.

Keywords

Cite

@article{arxiv.2111.10657,
  title  = {Generalizing Graph Neural Networks on Out-Of-Distribution Graphs},
  author = {Shaohua Fan and Xiao Wang and Chuan Shi and Peng Cui and Bai Wang},
  journal= {arXiv preprint arXiv:2111.10657},
  year   = {2024}
}

Comments

IEEE TPAMI 2023