How to get more compact representation space by SUGRL?

[JiaLonghao1997]

SUGRL has shown impressive performance and speed. However, the paper Eliciting Structural and Semantic Global Knowledge in Unsupervised Graph Contrastive Learning found that the representation space produced by SUGRL is not compact enough. We hope to use SUGLR for representation learning of weighted biological networks and then perform node clustering. How can we improve it?

[YujieMo]

Thanks for your interests on our work! You can consider to add some cluster-level contrastive learing to enhance the compactness of the learned representations.