Structural Compression (StructComp) is a simple yet effective training framework that significantly improves the scalability of graph contrastive learning by substituting message passing with node compression.
