Towards Robust Graph Contrastive Learning
This work addresses robustness in graph contrastive learning, which is an incremental step for applications in graph-based machine learning.
The paper tackles the problem of adversarially robust self-supervised learning on graphs by introducing a new method in the contrastive learning framework that uses adversarial transformations and edge modifications to increase robustness, with preliminary experiments showing promising results.
We study the problem of adversarially robust self-supervised learning on graphs. In the contrastive learning framework, we introduce a new method that increases the adversarial robustness of the learned representations through i) adversarial transformations and ii) transformations that not only remove but also insert edges. We evaluate the learned representations in a preliminary set of experiments, obtaining promising results. We believe this work takes an important step towards incorporating robustness as a viable auxiliary task in graph contrastive learning.