This repository contains code and example notebooks related to the following paper :
- How to make the most of local explanations: effective clustering based on influences, E. Escriva et al. (accepted at the ADBIS 2023 Conference)
- Data exploration based on local attribution explanation: a medical use case, E. Escriva et al. (accepted at the EXEC-MAN 2023 Workshop, host at the ADBIS 2023 Conference)
These papers present the first global method to explore influences as a new data space.
Available Code Files:
- Clustering Methods : Methods adapted to clusters influence explanations.
- Selection Methods : Methods developped or adapted to select representative instances.
- Metrics : Clustering metrics used during experiments.
- MMD-Critic : Implementation of MMD-Critic, based on the one from authors. (consider leaving a ⭐ on their repository to support authors releasing their code.)
Examples: