On the computation of counterfactual explanations - A survey

This repository contains the implementation of the methods proposed in the paper On the computation of counterfactual explanations - A survey by André Artelt and Barbara Hammer.

The methods are implemented in separatinghyperplane.py, glm.py, gnb.py and qda.py - the implementations depend on counterfactual.py, convexprogramming.py and qcqp.py. A minimalistic usage example is given in each file.

The default solver for solving LPs, convex QCQPs and SDPs is SCS. If you want to use a different solver, you have to overwrite the _solve method.

Requirements

• Python3.6
• Packages as listed in requirements.txt

License

MIT license - See LICENSE.md

How to cite

You can cite the version on arXiv.