• Student: Xuan-Truong Hoang
• ID: 1612899
• University: HCM city University of Science

This project is implement the code of Structural Deep Network Embedding (SDNE) and Deep Embedding Method for Dynamic Graphs (DynGEM). Beside, evaluating the performance among three graph embedding algorithms are Node2Vec, SDNE and DynGEM. I focus on the implement of DynGEM.

Go to the folder and use pip to install requirement.txt file.

Run: pip install -r requirement.txt

There are two task to evaluation: stability and link prediction. To run stability:

1. Open statbility_configuration.ini
2. Change dataset_name is your folder dataset in data folder. Dataset should be in a folder and contain a list of graph_0[index].edgelist files. For example: graph_00.edgelist, graph_01.edgelist, graph_23.edgelist,... These files will be read by networkx library of Python.
3. There are 3 algorithms in second group: is_dyge, is_node2vec and is_sdne. Set any algorithm is True to run that algorithm.
4. There are 2 kinds of files are:
   • train_[stability|link_pred].py: for training model to getting weights, models.
   • eval_[stability|link_pred].py: for evaluating model with metric stability or link prediction (ROC AUC Score| Precision@K | mAP)
5. To run a file: python train_link_pred.py

Note:Make sure you have your dataset in data folder with your dataset have a list graph_0*.edgelist files. Remember change dataset_name in configuration file same name your graph dataset.

Run: python train_link_pred.py or python eval_link_pred.py

I had two dynamic graph dataset in examples folder. Run get_soc_wiki_elec_data.py or get_cit_hepth_data.py to download dataset.

You will get a folder saved_data which contains all your trained data:

• Processed data for link prediction;
• DynGEM, SDNE models;
• DynGEM, SDNE, Node2Vec embedding.

If you run evaluation python file, you get the result of metric on your graph dataset with stability and link prediction.

To understand more about algorithms, checking references in the end of page.

@article{goyal2018dyngem, title={Dyngem: Deep embedding method for dynamic graphs}, author={Goyal, Palash and Kamra, Nitin and He, Xinran and Liu, Yan}, journal={arXiv preprint arXiv:1805.11273}, year={2018} }

@inproceedings{wang2016structural, title={Structural deep network embedding}, author={Wang, Daixin and Cui, Peng and Zhu, Wenwu}, booktitle={Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining}, pages={1225--1234}, year={2016} }

@inproceedings{grover2016node2vec, title={node2vec: Scalable feature learning for networks}, author={Grover, Aditya and Leskovec, Jure}, booktitle={Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining}, pages={855--864}, year={2016} }