Towards Reliable Link Prediction with Robust Graph Information Bottleneck

Link prediction on graphs has achieved great success with the rise of deep graph learning. However, the potential robustness under the edge noise is less investigated. We reveal that the inherent edge noise that naturally perturbs both input topology and target label leads to severe performance degradation and representation collapse. Here, we propose an information-theory guided principle, Robust Graph Information Bottleneck (RGIB), to extract reliable supervision signals and avoid representation collapse. Different from the general information bottleneck, RGIB decouples and balances the mutual dependence among graph topology, edge label, and representation, building a new learning objective for robust representation. We also provide two implementations, RGIB-SSL and RGIB-REP, that benefit from different methodologies, i.e., self-supervised learning and data reparametrization, for indirect and direct data denoising, respectively. Extensive experiments on six benchmarks with various scenarios verify the effectiveness of the proposed RGIB.