distance_explainer

XAI method to explain distances in embedded spaces.

Installation

There are 2 ways to install distance_explainer. To install distance_explainer from PyPI (recommended) run:

pip install distance_explainer

To instead install distance_explainer from the GitHub repository, run:

git clone git@github.com:dianna-ai/distance_explainer.git
cd distance_explainer
python3 -m pip install .

How to use

See our documentation and tutorial notebook for how to use this package. In short:

image1 = np.random.random((100, 100, 3))
image2 = np.random.random((100, 100, 3))

image2_embedded = model(image2)
explainer = DistanceExplainer(axis_labels={2: 'channels'})
attribution_map = explainer.explain_image_distance(model, image1, image2_embedded)

If you use, please cite

If you use Distance Explainer for your research, please cite our method paper and the software itself:

• Method paper: "Explainable embeddings with Distance Explainer" — arXiv:2505.15516 (to appear in the XAI26 proceedings)
• Software: distance_explainer — doi:10.5281/zenodo.10018768

@article{meijer2025explainable,
  title={Explainable embeddings with {D}istance {E}xplainer},
  author={Meijer, Christiaan and Bos, E. G. Patrick},
  journal={arXiv preprint arXiv:2505.15516},
  year={2025},
  note={To appear in the XAI26 proceedings}
}

@software{meijer2023distance,
  title={distance_explainer},
  author={Meijer, Christiaan and Bos, Patrick},
  doi={10.5281/zenodo.10018768},
  year={2023}
}

Contributing

If you want to contribute to the development of distance_explainer, have a look at the contribution guidelines.

Credits

This package was created with Cookiecutter and the NLeSC/python-template.