📝 Publications

Improving Augmentation Consistency for Graph Contrastive Learning
Weixin Bu*, Xiaofeng Cao*, Yizhen Zheng, Shirui Pan
[Paper] | [Code]

• A novel augmentation consistency perspective in GCL
• Integrate semantic and structural properties to better capture node consistency
• An effective consistency improvement loss to maintain augmentation consistency among positive node pairs

Nonparametric Teaching for Graph Property Learners
Chen Zhang*, Weixin Bu*, Zeyi Ren, Zhengwu Liu, Yik-Chung Wu, Ngai Wong
[Project] | [Code]

• A novel paradigm that interprets graph property learning within the theoretical context of nonparametric teaching (NT)
• Reveal the consistency between the evolution of GCN driven by parameter updates and that under functional gradient descent in NT
• Demonstrate the effectiveness of GraNT through extensive experiments (graph/node-level regression / classification) in graph property learning