Training Robust Graph Neural Networks by Modeling Noise Dependencies

TL;DR

This paper introduces DANG, a realistic noise model for graphs, and proposes DA-GNN, a robust GNN that captures noise dependencies, validated by new benchmarks and superior performance over existing methods.

Contribution

It presents a dependency-aware noise model for graphs and a novel robust GNN, DA-GNN, that models causal noise relationships using variational inference.

Findings

DA-GNN outperforms baselines across various noise scenarios.

New benchmarks simulate realistic noise dependencies in graphs.

DA-GNN maintains robustness in both DANG and conventional noise models.

Abstract

In real-world applications, node features in graphs often contain noise from various sources, leading to significant performance degradation in GNNs. Although several methods have been developed to enhance robustness, they rely on the unrealistic assumption that noise in node features is independent of the graph structure and node labels, thereby limiting their applicability. To this end, we introduce a more realistic noise scenario, dependency-aware noise on graphs (DANG), where noise in node features create a chain of noise dependencies that propagates to the graph structure and node labels. We propose a novel robust GNN, DA-GNN, which captures the causal relationships among variables in the data generating process (DGP) of DANG using variational inference. In addition, we present new benchmark datasets that simulate DANG in real-world applications, enabling more practical research on robust GNNs. Extensive experiments demonstrate that DA-GNN consistently outperforms existing baselines across various noise scenarios, including both DANG and conventional noise models commonly considered in this field. Our code is available at https://github.com/yeonjun-in/torch-DA-GNN.