Grad-Align

Source code for the papers

• Jin-Duk et al. "Grad-Align: Gradual Network Alignment via Graph Neural Networks" AAAI-22 (Student Poster Program)
• Jin-Duk et al. "On the Power of Gradual Network Alignment Using Dual-Perception Similarities". IEEE TPAMI

• Also check our paper (https://github.com/jindeok/GradAlign_plus): Jin-Duk et al. "GradAlign+: Empowering gradual network alignment using attribute augmentation." CIKM 2022.

Pytorch_geometric (https://pytorch-geometric.readthedocs.io/en/latest/) package is used for the implementation of graph neural networks (GNNs).

Dependancy

• Pytorch > 1.8
• torch_geometric: https://pytorch-geometric.readthedocs.io/en/latest/notes/installation.html
• numpy
• pands
• sklearn
• tqdm
• networkx
• matplotlib

Running

run main.py script file

or in your prompt, type

python main.py --graphname 'fb-tt' --k_hop 2 --mode 'not_perturbed'

• --graphname can be either one dataset of the three ('fb-tt' for Facebook vs. Twitter dataset, 'douban' for Douban online vs. offline dataset, 'econ' for Econ perturbed pair.)
• Other description for each arguments are typed in '--help' arguments in main.py argparse.
• for the implemenation of the graphname 'econ', --mode should be changed to 'perturbed' instead of 'not_perturbed'

etc.

If you need any further information, please e-mail me : [email protected]