Local SC-FC coupling pattern

Subgraph extraction is based on deep coding tree and gSpan algorithm.

Data availability statement

The data is not publicly available due to permission reasons. The data are available from the corresponding author on reasonable request.

Counterfactual Explanation

The counterfactual interpretation based on local SC-FC coupling is based on the DiCE framework. We only give the relevant test code here, and the detailed code will be updated after the paper is accepted. For more details, please visit: https://github.com/interpretml/DiCE.

How to install

This project was implemented on Python 3.7 and MATLAB R2018b.

Install this project using pip:

pip install gspan-mining
pip install dice-ml

How to run

python -m subgraph_extraction -s [minsup] -l [vertex_lower_limit] -u [vertex_upper_limit] ./graphdata/test_graph.txt
python -m subgraph_extraction -s 2 -l 3 -u 4 ./graphdata/test_graph.txt

Reference

• [1]Paper

Tsuda, Koji. "Entire regularization paths for graph data." Proceedings of the 24th international conference on Machine learning. 2007.

• [2]Paper

Yan, Xifeng, and Jiawei Han. "gspan: Graph-based substructure pattern mining." 2002 IEEE International Conference on Data Mining, 2002. Proceedings.. IEEE, 2002.

• [3]Paper

Mothilal, Ramaravind K., Amit Sharma, and Chenhao Tan. "Explaining machine learning classifiers through diverse counterfactual explanations." Proceedings of the 2020 conference on fairness, accountability, and transparency. 2020.

Additional information

Currently, only the prototype code of this study has been uploaded. The corresponding complete code and some details of the experiment will be supplemented after the paper is accepted.