Knowledge Integration in XAI with Gödel Integrals - Archive ouverte HAL Access content directly
Conference Papers Year : 2023

Knowledge Integration in XAI with Gödel Integrals

Intégration de connaissances dans le domaine de l'AXI avec les intégrales de Gödel

Abstract

Counterfactual examples constitute a popular form of explanations that are most often generated through the optimisation of a cost function that combines different components of the explanation quality. This paper focuses on the final aggregation of the objective term, that depends on the considered machine learning task, and the subjective term, that depends on the targeted user and more precisely on their knowledge. It discusses the desired properties of this aggregation operator and proposes to use two forms of the Gödel integral operator, highlighting the expressiveness and appropriateness they offer.
Fichier principal
Vignette du fichier
fuzz_IEEE_2023.pdf (577.93 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-04174815 , version 1 (05-09-2023)

Identifiers

  • HAL Id : hal-04174815 , version 1

Cite

Adulam Jeyasothy, Agnès Rico, Marie-Jeanne Lesot, Christophe Marsala, Thibault Laugel. Knowledge Integration in XAI with Gödel Integrals. International Conference on Fuzzy Systems, Aug 2023, Incheon, South Korea. ⟨hal-04174815⟩
33 View
41 Download

Share

Gmail Facebook X LinkedIn More