CLL308 / GraphDA

Unsupervised Domain Adaptation on Graphs/DSR

Graph Domain Adaptation: A Generative View

The official implementation of Graph Domain Adaptation: A Generative View. The model is a combination of Graph Neural Networks and some well-known DA frameworks, e.g, replacing the ResNet feature extractor with GCNs.

Requirements

• numpy==1.15.1
• scikit_learn==0.24.1
• torch==1.8.0

Run the demo

python run_dann.py

Data

The egonet dataset with 4 domains: OAG, Digg, Twitter and Weibo

• raw: https://github.com/xptree/DeepInf
• processed: https://pan.baidu.com/s/1BLxGa7785xuE17j_uVd2bA psw:9ins

Compared Models:

• DANN: https://www.jmlr.org/papers/volume17/15-239/15-239.pdf
• MDD: https://arxiv.org/abs/1904.05801
• DIVA: https://arxiv.org/abs/1905.10427
• DSR: https://arxiv.org/abs/2012.11807