Uncertainty quantification in regression neural networks using likelihood-based belief functions - Connaissances, Incertitudes et Données Accéder directement au contenu
Communication Dans Un Congrès Année : 2024

Uncertainty quantification in regression neural networks using likelihood-based belief functions

Résumé

We introduce a new method for quantifying prediction uncertainty in regression neural networks using evidential likelihood-based inference. The method is based on the Gaussian approximation of the likelihood function and the linearization of the network output with respect to the weights. Prediction uncertainty is described by a random fuzzy set inducing a predictive belief function. Preliminary experiments suggest that the approximations are very accurate and that the method allows for conservative uncertainty-aware predictions.
Fichier principal
Vignette du fichier
belief2024_nn_final.pdf (452.65 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04621414 , version 1 (24-06-2024)

Identifiants

  • HAL Id : hal-04621414 , version 1

Citer

Thierry Denoeux. Uncertainty quantification in regression neural networks using likelihood-based belief functions. Eighth International Conference on Belief Functions (BELIEF 2024), Sep 2024, Belfast, United Kingdom. ⟨hal-04621414⟩
0 Consultations
0 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More