This is a pytorch implementation of the VANE framework from paper "A View-Adversarial Framework for Multi-View Network Embedding".
We test our codes with python 3.6, pytorch 1.6, and scikit-learn 0.23.2.
We provide preprocessed Twitter-Rugby data for reproduction. Links for original data: Aminer (https://www.aminer.cn/citation), Twitter-Rugby (http://mlg.ucd.ie/aggregation).
The files in the 'walks' folder are generated based on the codes from the node2vec repo (https://github.com/aditya-grover/node2vec) and the first number of each line indicates the index (0-based index) of view this path comes from. The walks file used for link prediction is generated from a graph deleted 10% of its links.
Ground truth for node classification is in the form of "index of node : index of class". Ground truth for link prediction is in the form of "index of node 1 \t index of node 2 \t linked (1) / non-linked (0)".
If you use the corresponding codes, please cite the following paper:
@inproceedings{DBLP:conf/cikm/FuXLTH20,
author = {Dongqi Fu and
Zhe Xu and
Bo Li and
Hanghang Tong and
Jingrui He},
editor = {Mathieu d'Aquin and
Stefan Dietze and
Claudia Hauff and
Edward Curry and
Philippe Cudr{\'{e}}{-}Mauroux},
title = {A View-Adversarial Framework for Multi-View Network Embedding},
booktitle = {{CIKM} '20: The 29th {ACM} International Conference on Information
and Knowledge Management, Virtual Event, Ireland, October 19-23, 2020},
pages = {2025--2028},
publisher = {{ACM}},
year = {2020},
url = {https://doi.org/10.1145/3340531.3412127},
doi = {10.1145/3340531.3412127},
timestamp = {Mon, 19 Oct 2020 18:49:47 +0200},
biburl = {https://dblp.org/rec/conf/cikm/FuXLTH20.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}