The Co-Evolution Model for Social Network Evolving and Opinion Migration

Implementation of the paper RaRE: Social Rank Regulated Large-scale Network Embedding

Usage:

javac *.java
java Main <options>

or

java Main -help

to show the help command below.

options:

• -k: dimension of embedding; 2 by default
• -data: the location of data file
• -weighted: whether the network is weighted (0: unweighted; other: weighted); 0 by default
• -out: the directory of output files; "." (current directory) by default
• -lr: initial stepsize; 0.05 by default
• -lambda_r: coefficient of dr (first component), must be positive; 1.0 by default
• -lambda_z: coefficient of dz (second component), must be positive; 1.0 by default
• -lambda_0: bias term (third component), should be negative for sparse networks; -1.0 by default
• -alpha_r: parameter for power law prior (on r), recommended value: 1.0~2.0; 1.5 by default
• -reg_z: l2 regularization coefficient (on z); 1e-6 by default
• -tp: fraction of nodes for training, must be between 0 and 1; 0.9 by default
• -tolerance: stopping criterion on log-likelihood improvement (negative number for no tolerance); 1e-6 by default
• -nsw: negative sample weight, must be positive; 5 by default
• -iter: value of maximum edges to be sampled (in thousands); 10000 thousand by default
• -verbose: whether see verbose output or not (0: show limited outputs; other: show all outputs); 1 by default
• -h or -help: show this help command

The -data argument is necessary.

Citation:

@inproceedings{gu2018rare,
  title={RaRE: Social Rank Regulated Large-scale Network Embedding},
  author={Gu, Yupeng and Sun, Yizhou and Li, Yanen and Yang, Yang},
  booktitle={Proceedings of the 2018 World Wide Web Conference on World Wide Web},
  pages={359--368},
  year={2018},
  organization={International World Wide Web Conferences Steering Committee}
}

Contact:

Yupeng Gu