This repository provides a reference implementation of the paper "SPINE: Structural Identity Preserved Inductive Network Embedding". SPINE is an inductive embedding method which can simultaneously preserve the local proximity and the structural identity of nodes. Details can be found here.
- tensorflow > 1.2.1
- networkx == 1.11
- gensim > 2.3.0
- fastdtw
To run SPINE on Cora dataset, execute as:
python main.py --input data/cora_id_edge.txt --train_prefix cora --preprocess True
To evaluate the performance by Pearson and Spearman correlation instead of classification accuracy, set --CORR True.
To run SPINE and SPINE-p, deactive and active --PRETRAIN respectively.
For more options, please check main.py.
We refer to GraphSAGE and GCN while constructing code framework and preprocessing datasets. Many thanks to the authors for making their code available.
Please cite our paper if you find SPINE useful in your research.
@inproceedings{guo2019spine,
title={SPINE: Structural Identity Preserved Inductive Network Embedding},
author={Guo, Junliang and Xu, Linli and Liu, Jingchang},
booktitle={Twenty-Eighth International Joint Conference on Artificial Intelligence},
year={2019}
}
This is only a reference implementation of SPINE, feel free to ask any question by opening an issue or email me at leoguojl@gmail.com.