README.md

Model-probing mislabeled examples detection in machine learning datasets

EN. Detect mislabeled examples in machine learning dataset, using the 4 components framework described in the paper Mislabeled examples detection viewed as probing machine learning models: concepts, survey and extensive benchmark, which allows the implementation of a variety of model-probing detection methods.

FR. Détection d'exemples mal-étiquetés dans des jeux de données d'apprentissage automatique, en utilisant les 4 composants décrits dans l'article Mislabeled examples detection viewed as probing machine learning models: concepts, survey and extensive benchmark, qui permet d'implémenter une multitude de méthodes de détection par sondage de modèle.

Paper

If you use this library in a research project, please consider citing the corresponding paper with the following bibtex entry:

@article{george2024mislabeled,
  title={Mislabeled examples detection viewed as probing machine learning models: concepts, survey and extensive benchmark},
  author={Thomas George and Pierre Nodet and Alexis Bondu and Vincent Lemaire},
  journal={Transactions on Machine Learning Research},
  issn={2835-8856},
  year={2024},
  url={https://openreview.net/forum?id=3YlOr7BHkx},
  note={}
}

Development

Install hatch.

To format and lint:

hatch fmt

To run tests:

hatch test