IONE-Aligning-Users-across-Social-Networks-Using-Network-Embedding

Aligning Users across Social Networks Using Network Embedding(IJCAI)，paper author uses Java(https://github.com/ColaLL/IONE)

For wider application, we have updated the python version，use tensorflow1.14.0

Source Code and anonymous twitter_foursquare data for IJCAI 2016 paper "Aligning Users Across Social Networks Using Network Embedding"

Welcom to contact me (Zihan Yan [email protected]) when you have any problems about the code(Python tensorflow1.14.0)

You just need run 'IONE_tf_train.py'

In order to be fair (actually lazy), we convert the output embedding into an embedding file in ione java version format. You can test the accuracy in java files.run 'emd_to_ione_emd.py'

Before, some students talked privately about the code I wanted tensorflow. Because I'm a little busy, the code didn't come up. So apologize!

The code style may not be very good, Sry!

requirements

numpy==1.14
networkx==2.0
scipy==0.19.1
tensorflow>=1.12.1
gensim==3.0.1
scikit-learn==0.19.0

test result：

1- The tensorflow version of ione because using GPU runs faster than the Java version.
2- The tensorflow version of ione is almost as accurate as the Java version. Sometimes it is better. It may be because of the use of the deep learning optimizer.

The code refers to the open source tool 'OpenNE' of Tsinghua University. OpenNE: An open source toolkit for Network Embedding https://github.com/thunlp/OpenNE

如果感兴趣或者帮助到你 欢迎star ~ paper和code有任何问题 也欢迎讨论.