Please cite the following work if you find the code useful.
@inproceedings{yang2018meta,
Author = {Yang, Carl and Feng, Yichen and Li, Pan and Shi, Yu and Han, Jiawei},
Booktitle = {ICDM},
Title = {Meta-graph based hin spectral embedding: methods, analyses and insights},
Year = {2018}
}
Contact: Carl Yang (yangji9181@gmail.com)
For example, we use DBLP dataset. The data is processed using https://github.com/macks22/dblp . Please put the processed data under the data folder.
Filter DBLP based on Vocab-Label author list
python dblp_data.py
Match the metapath/metagraph instances from HIN. One need to specify the dataset flag. You need to install wine
before running the commend!!!
python3 submodule.py dblp
Calculated eigenvalues and plot the final eigenvalue plots using python. The 2nd argument is the dataset name, and the 3rd and 4th arguments are
python3 eigen_cal_all.py dblp
Compress the original embedding using the autoencoder. The parameters could be changed inside the autoencoder_dblp.py script.
python3.5 -u autoencoder_dblp.py --dataset dblp
- dblp.q : This is the selected metagraphs for SubMatch Program, please reference SubMatch about how to write this kind of *.q file.
- metapath_dblp.txt : This is the text version of the dblp.q. Their order should be the same.