Unsupervised Domain Adaptive Graph Convolutional Networks

This repository contains the author's implementation in PyTorch for the paper "Unsupervised Domain Adaptive Graph Convolutional Networks", WWW-20.

Dependencies

• Python (>=3.6)
• Torch (>=1.2.0)
• numpy (>=1.16.4)
• torch_scatter (>= 1.3.0)
• torch_geometric (>= 1.3.0)

Datasets

The data folder includes different domain data. The preprocessed data can be found in Google Drive.

The orginal datasets can be founded from here.

Implementation

Here we provide the implementation of UDA-GCN, along with two domain datasets. The repository is organised as follows:

• data/ contains the necessary dataset files for DBLP domain and ACM domian(can be found in Google Drive);
• dual_gnn/ contains the implementation of the Global GCN and Local GCN;

Finally, UDAGCN_demo.py puts all of the above together and can be used to execute a full training run on the datasets.

Process

• Place the datasets in data/
• Change the dataset in UDAGCN_demo.py .
• Training/Testing:

python UDAGCN_demo.py

Citation

@inproceedings{wu2020UDAGCN
author={Man Wu and Shirui Pan and Chuan Zhou and Xiaojun Chang and Xingquan Zhu},
title={Unsupervised Domain Adaptive Graph Convolutional Networks},
journal={{WWW} '20: The Web Conference},
year={2020}
}