discern-xai

DisCERN: Discovering Counterfactual Explanations using Relevance Features from Neighbourhoods


Keywords
machine-learning, explanation, interpretability, counterfactual
License
MIT
Install
pip install discern-xai==0.0.25

Documentation

DisCERN-XAI

DisCERN: Discovering Counterfactual Explanations using Relevance Features from Neighbourhoods

Installing DisCERN

DisCERN supports Python 3+. The stable version of DisCERN is available on PyPI:

pip install discern-xai

To install the dev version of DisCERN and its dependencies, clone this repo and run pip install from the top-most folder of the repo:

pip install -e .

DisCERN requires the following packages:
numpy
pandas
lime
shap
scikit-learn

Compatible Libraries

Attribution Explainer scikit-learn TensorFlow/Keras PyTorch
LIME N/A
SHAP ✓ shap.TreeExplainer ✓ shap.DeepExplainer N/A
Integrated Gradients N/A

Getting Started with DisCERN

Binary Classification example on the Adult Income dataset using RandomForest and Keras Deep Neural Net classifiers are here

Multi-class Classification example on the Cancer risk dataset using RandomForest and Keras Deep Neural Net classifiers are here

Citing

Please cite it follows:

  1. Wiratunga, N., Wijekoon, A., Nkisi-Orji, I., Martin, K., Palihawadana, C., & Corsar, D. (2021, November). Discern: discovering counterfactual explanations using relevance features from neighbourhoods. In 2021 IEEE 33rd International Conference on Tools with Artificial Intelligence (ICTAI) (pp. 1466-1473). IEEE.

  2. Wijekoon, A., Wiratunga, N., Nkisi-Orji, I., Palihawadana, C., Corsar, D., & Martin, K. (2022, August). How Close Is Too Close? The Role of Feature Attributions in Discovering Counterfactual Explanations. In Case-Based Reasoning Research and Development: 30th International Conference, ICCBR 2022, Nancy, France, September 12–15, 2022, Proceedings (pp. 33-47). Cham: Springer International Publishing.

Bibtex:

@misc{wiratunga2021discerndiscovering,
  title={DisCERN:Discovering Counterfactual Explanations using Relevance Features from Neighbourhoods}, 
  author={Nirmalie Wiratunga and Anjana Wijekoon and Ikechukwu Nkisi-Orji and Kyle Martin and Chamath Palihawadana and David Corsar},
  year={2021},
  eprint={2109.05800},
  archivePrefix={arXiv},
  primaryClass={cs.LG}

}

@inproceedings{wijekoon2022close,
    title={How Close Is Too Close? The Role of Feature Attributions in Discovering Counterfactual Explanations},
    author={Wijekoon, Anjana and Wiratunga, Nirmalie and Nkisi-Orji, Ikechukwu and Palihawadana, Chamath and Corsar, David and Martin, Kyle},
    booktitle={Case-Based Reasoning Research and Development: 30th International Conference, ICCBR 2022, Nancy, France, September 12--15, 2022, Proceedings},
    pages={33--47},
    year={2022},
    organization={Springer}
}




drawing

drawing




This research is funded by the iSee project which received funding from EPSRC under the grant number EP/V061755/1. iSee is part of the CHIST-ERA pathfinder programme for European coordinated research on future and emerging information and communication technologies.