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.
@article{jovanovic2021groc, title = {Towards Robust Graph Contrastive Learning}, author = {Nikola Jovanović and Zhao Meng and Lukas Faber and Roger Wattenhofer}, year={2021}, journal = {The Workshop on Self-Supervised Learning for the Web (SSL @ WWW)} }