graph contrastive learning (GCL), as a self-supervised learning method, can
solve the problem of annotated data scarcity. It mines explicit features in
unannotated graphs to generate favorable graph representations for downstream
tasks. Most existing GCL methods focus on the design of