**** Lien zoom: https://univ-artois-fr.zoom.us/j/93828206912 ****
Le prochain séminaire GDR RADIA aura lieu Jeudi 28 Novembre à 11h
Il s’agira d’une intervention de Leila Amgoud (IRIT – CNRS)
Titre: Axiomatic Foundations of Explainability
Résumé: Explaining black-box classification models is a hot topic in AI, with the overall goal of improving confidence in the decisions made by such models. Various explanation functions have been proposed. However, their formal properties and connections have not been sufficiently investigated. In this talk, I will present some contributions to the axiomatic foundations of explainability. I will discuss key axioms that explainers would satisfy, their links and compatibility, and shed light on the axioms that distinguish the two main types of reasons: sufficient and necessary. I also present different characterisation results in constrained and unconstrained settings, and when explanations are generated from the whole feature space or from a sample (or dataset). The final result is a catalogue of different forms of explanation, with different complexities and formal guarantees.
Le séminaire suivant aura lieu le 19 Décembre, Patrice Perny (LIP6) parlera de « Modélisation des préférences et apprentissage de modèles de décision en présence d’interactions entre critères».
Tous les détails sur le séminaire du GDR et les vidéos des précédents séminaires ici:
https://gdr-radia.cnrs.fr/seminaire/
———————————————————————
Desinscription: envoyez un message a: bull-ia-unsubscribe@gdria.fr