FEM and Multi-Layered FEM: Feature Explanation Methods with Statistical Filtering of Important Features - Archive ouverte HAL
Chapitre D'ouvrage Année : 2024

FEM and Multi-Layered FEM: Feature Explanation Methods with Statistical Filtering of Important Features

Résumé

Deep learning (DL) approaches have become essential in data analysis and classification but they appear as black boxes, the results being given without any explanation. Then, the need for explanations of Deep Neural Network (DNN) decisions has led to an active research in eXplainable Artificial Intelligence (XAI). Here we developed Multi-Layered FEM that is an improvement of the Feature Explanation method that was based on the activation values of the target layer of the network.
Fichier non déposé

Dates et versions

hal-04698686 , version 1 (16-09-2024)

Identifiants

Citer

Alexey Zhukov, Jenny Benois-Pineau, Romain Giot, Romain Bourqui, Luca Bourroux. FEM and Multi-Layered FEM: Feature Explanation Methods with Statistical Filtering of Important Features. Emerging Topics in Pattern Recognition and Artificial Intelligence, 09, WORLD SCIENTIFIC, pp.295-327, 2024, Series on Language Processing, Pattern Recognition, and Intelligent Systems, ⟨10.1142/9789811289125_0012⟩. ⟨hal-04698686⟩

Collections

CNRS UNIV-BORDEAUX
18 Consultations
0 Téléchargements

Altmetric

Partager

More