RBF-SCORE+

Description

This algorithm finds the optimal shaping parameter and condition number of various graph RBF kernels for community detection algorithm SCORE+ and its variants. We also optimize the SCORE+ algorithm with higher order proximity of the affinity matrix.

How to run?

1. Go to the basic_function.py under the folder rbf-score.
2. Scroll down and find the main() function.
3. When running on real-world data sets, for instance, running the football.mat data under /data/datasets with a MQ RBF, just set fn= 'football' and RBF='MQ'.

 '''
Run real-world data set
fn: data name
RBF: RBFs, {'MQ', 'iMQ', 'gaussian'}
'''
data, y, k = import_real_data(fn='football')
run_networks(data, y, k, RBF='MQ')

4. Then run the code, done!

Outputs

Your will get outputs like this if you run football data.

0.303 289.3264 0.957 0.62

The outputs tell us that the optimal shaping parameter is 0.303 and the corresponding condition number will be 289.3264, with NMI = 0.957 and Q = 0.62.

If you use this code, please cite:

@ARTICLE{10373106,
  author={Zhu, Yanhui and Hu, Fang and Kuo, Lei Hsin and Liu, Jia},
  journal={IEEE Transactions on Big Data}, 
  title={SCOREH+: A High-Order Node Proximity Spectral Clustering on Ratios-of-Eigenvectors Algorithm for Community Detection}, 
  year={2023},
  volume={},
  number={},
  pages={1-12},
  doi={10.1109/TBDATA.2023.3346715}}