This is the pytorch implementation of Augmenting Knowledge Transfer across Graphs (ICDM 2022).

We introduce a novel notion named trinity signal that can naturally formulate various graph signals at different granularity (e.g., node attributes, edges, and subgraphs). With that, we further propose a domain unification module together with a trinity-signal mixup scheme to jointly minimize the domain discrepancy and augment the knowledge transfer across graphs. Comprehensive empirical results show that TransNet outperforms all existing approaches on seven benchmark datasets by a significant margin.

Here, we only provide datasets M2 and A1. Please download other datasets from the original papers listed in our paper.

TransNet is firstly pre-trained on the source dataset for 2000 epochs; then it is fine-tuned on the target dataset for 800 epochs using limited labeled data in each class. We use Adam optimizer with learning rate 3e-3. α in the beta-distribution of trinity-signal mixup is set to 1.0 and the output dimension of MLP in domain unification module is set to 100 by default. Precision is used as the evaluation metric.

 python ./src/transnet.py --datasets='M2+A1' --finetune_epoch=800 --mu=1e-2 --seed=100 --gnn='gcn' --few_shot=5  --epoch=2000  --batch_size=-1   --finetune_lr=0.01  --pre_finetune=200 --ratio=0.7 --disc='3' --_lambda=0.02  --_lambda=0.05 --_alpha=0.01 --_alpha=0.01

Please cite the following paper if you found this library useful in your research:

Yuzhen Mao, Jianhui Sun, Dawei Zhou
IEEE International Conference on Data Mining (ICDM), 2022

@inproceedings{mao2022augmenting,
  title={Augmenting Knowledge Transfer across Graphs},
  author={Mao, Yuzhen and Sun, Jianhui and Zhou, Dawei},
  booktitle={2022 IEEE International Conference on Data Mining (ICDM)},
  pages={1101--1106},
  year={2022},
  organization={IEEE}
}