graph contrastive learning (GCL) has emerged as a state-of-the-art strategy
for learning representations of diverse graphs including social and biomedical
networks. GCL widely uses stochastic graph topology augmentation, such as
uniform node dropping, to generate augmented graphs. Howe