This is the reconstructed code for Unsupervised Adversarial Graph Alignment with Graph Embedding.
This new code optimized and simplified the original code.
Part of iUAGA's code will be updated later.
- Providing more optional parameters and you can adjust some common parameters easily in the cmd now;
- Providing execute.sh so you can run a series of commands automatically, which will reduce the number of operations;
- Providing a variety of similarity calculation methods;
- Providing complete data, which had processed;
- The calculation process of cosine similarity is optimized, which makes the code run more efficiently;
- Added a lot of Chinese remarks.
- initial_data_processing.py : Transforming the raw data into a suitable format for code to run.
- deepwalk : Embedding the nodes of graph in an unsupervised fashion.
- main.py : The main function to run the code.
You can also run the execute.sh in cmd to launch the program, including the three steps above:
sh execute.sh
For more details, please refer to the execute.sh.
We used three public datasets in the paper:
You can also get the dataset in -graph data.
For more details, please refer to here.
We utilized DeepWalk to learn the source and target graph embeddings in this work, so the format of input data followed the output of deepwalk. You can learn more about the detail from here.