Graph Prompt Learning

Official implementation of Graph structure prompt learning: A novel methodology to improve performance of graph neural networks published by Applied Intelligence

Requirements

• torch==1.10.0
• torch_geometric==2.0.0

Datasets

• The node classification datasets include Cora,Citeseer,Pubmed,Computers,CS,Photo,Physics and DBLP.
• The graph classification datasets include DD，PROTEINS,NCI1 and NCI109. Specificially,the download url of datasets can refer to https://pytorch-geometric.readthedocs.io/en/latest/modules/datasets.html

Code

• For node classification, the code is in Node_Classification_GPL, and the code for GCN and GAT on the Cora dataset according to the original division method is in Py_GCN_GPL and Py_GAT_GPL.
• For graph classification, the code is in Graph_Classification_GPL. DiffPool_GPL, SAGPool_GPL and Mewispool_GPL are implemented separately.

Cite

@article{gpl2025,
  title={Graph structure prompt learning: A novel methodology to improve performance of graph neural networks},
  author={Huang, Zhenhua and Li, Kunhao and Wang, Shaojie and Jia, Zhaohong and Zhu, Wentao and Mehrotra, Sharad},
  journal={Applied Intelligence},
  year={2025},
  publisher={Springer}
}