csdnlzh / RandNE

Implementation of a scalable embedding method suggested in ICDM2018

Info:

Implementation of the scalable graph embedding method suggested in [1]

NOTE THAT THIS IS NOT THE IMPLEMENTATION OF THE AUTHORS OF [1] AND IT IS NOT GUARANTEED TO WORK. The implementation of the authors of [1] is/will be here: https://github.com/ZW-ZHANG/RandNE.

to compile:

gcc RandNE.c -o RandNE -O9 -lm

to execute:

./RandNE net.txt emb.txt d q a0 a1 ... aq

• net.txt should contain on each line: "u v\n" that is the input undirected graph.
• emb.txt contains the resulting embedding (d floats on each line)
• d is the dimension of the embedding
• q is the order of the embedding
• ak's are the coefficient of A^k matrix such as defined in [1]

running time on a commodity machine:

On http://snap.stanford.edu/data/com-Orkut.html

• d=32 and q=3: 5 minutes
• d=32 and q=5: 10 minutes
• d=128 and q=3: 20 minutes

Reference:

[1] : "Billion-scale Network Embedding with Iterative Random Projection", Ziwei Zhang, Peng Cui, Haoyang Li, Xiao Wang and Wenwu Zhu, ICDM2018. https://papers-gamma.link/paper/110

First contributor:

Maximilien Danisch
Septembre 2018
http://bit.ly/danisch
[e-mail]