Adversarial Contrastive Graph Augmentation with Counterfactual Regularization

A novel model-based adversarial contrastive graph augmentation (ACGA) method for generating both positive samples with minimal sufficient information and hard negative graph samples automatically.