Weighted-Metapath2Vec

Weighted-Metapath2Vec is a Python package for embedding heterogeneous graphs. It uses a weighted variant of metapath2vec to compute the node embeddings. The embeddings can be used for downstream machine learning.

The package is a work-in-progress. There are bugs, and example notebooks are missing. If you want to use this package, expect to make changes.

Installation

pip install weighted-metapath2vec

Usage

from weighted_metapath2vec import WeightedMetapath2VecModel

G = ...  # Load a networkx graph as G

metapaths = [
    ['Article', 'Author', 'Article'],
    ['Author', 'Article', 'Author']
]

model = WeightedMetapath2VecModel(G,
                                  metapaths,
                                  walk_length=3,
                                  n_walks_per_node=20,
                                  embedding_dim=128)

node_embeddings = model.fit_transform()

...  # downstream task

Contributing

Use GitHub to fork and submit pull requests.

Citation

Please cite this code as follows (BibTeX):

@software{Weighted_Metapath2Vec,
        author = {Ansarinia, Morteza and Cardoso-Leite, Pedro},
        doi = {10.5281/zenodo.7096229},
        month = {6},
        title = {{Weighted Metapath2Vec Graph Embedding}},
        url = {https://github.com/morteza/weighted-metapath2vec},
        version = {v0.1.4},
        year = {2022}
}

Acknowledgements

This project is supported by the Luxembourg National Research Fund (ATTRACT/2016/ID/11242114/DIGILEARN and INTER Mobility/2017-2/ID/11765868/ULALA).

License

MIT License. See the LICENSE file.