Towards generalizable Graph Contrastive Learning: An information theory perspective

Yige Yuan, Bingbing Xu, Huawei Shen, Qi Cao, Keting Cen, Wen Zheng, Xueqi Cheng

Abstract: Highlights•We introduce the GCL-GE metric for the generalization of GCL.•We prove a mutual information upper bound for GCL-GE.•We propose InfoAdv framework based on this bound to enhance GCL generalization.•We experimentally demonstrate the superior performance of the InfoAdv framework.