The source code of "Dual-level Mixup for Graph Few-shot Learning with Fewer Tasks"
Here we provide three datasets for reproducibility, including Amazon-Clothing, CoraFull, and DBLP, with two datasets in the zip file and the other available for automatic download.
You can run the following the command.
cd SMILE
unzip few_shot_data
For in-domain setting
python train.py --dataset Amazon_clothing --way 5 --shot 5 --num_tasks 5
You can change the --dataset to corafull or dblp to train other datasets.
For cross-domain setting
python train_cross_domain.py --dataset Amazon_clothing --dataset_cr corafull
You can change the --dataset to corafull and the --dataset_cr to Amazon_clothing to switch the cross-domain setting.
If you find our work can help your research, please cite our work!
@inproceedings{liu2025dual,
title={Dual-level Mixup for Graph Few-shot Learning with Fewer Tasks},
author={Liu, Yonghao and Li, Mengyu and Giunchiglia, Fausto and Huang, Lan and Li, Ximing and Feng, Xiaoyue and Guan, Renchu},
booktitle={The Web Conference},
year={2025}
}
If you have any question, feel free to contact via email.