Skip to main content
Aller à la page d’accueil de la Commission européenne (s’ouvre dans une nouvelle fenêtre)
français fr
CORDIS - Résultats de la recherche de l’UE
CORDIS

Unlocking Reliable Small-data AI through Bayesian Deep Learning

Objectif

While AI for science has enabled remarkable breakthroughs in recent years, many scientists have been left underwhelmed when trying to apply AI methods in their domains. This is mainly because most scientific datasets are too small for the data-hungry modern foundation model approaches. Bayesian deep learning (BDL), combining rigorous Bayesian principles with the flexibility of neural networks, has positioned itself as a powerful remedy for this lack of data, but its adoption is hampered by some tedious design choices, most importantly, the selection of model, prior, and inference method. This proposal aims to bring the revolution of automated machine learning, which has radically boosted the adoption of conventional AI methods, into the realm of BDL. The goal is to develop an automated end-to-end pipeline that domain practitioners can autonomously use to select the best model architectures and data invariances, incorporate their own prior knowledge in an efficient and natural way, and choose the best inference method tailored to their computational resources. While many research challenges will have to be overcome to reach this ambitious goal, recent advances in BDL research, including some of my own, put me in the ideal position now to tackle these questions. Throughout this process, the research will stay grounded in real-world scientific requirements via close collaboration with domain experts in healthcare, materials science, and astrophysics. A successful completion of this project will democratize access to BDL solutions by cutting out the machine learning middlemen, thus unlocking small-data AI applications across scientific domains.

Champ scientifique (EuroSciVoc)

Ce projet n'a pas encore été classé par EuroSciVoc.
Proposez les domaines scientifiques qui vous semblent les plus pertinents et aidez-nous à améliorer notre service de classification.

Mots‑clés

Programme(s)

Thème(s)

Régime de financement

HORIZON-ERC - HORIZON ERC Grants

Voir tous les projets financés dans le cadre de ce programme de financement

Appel à propositions

(s’ouvre dans une nouvelle fenêtre) ERC-2026-STG

Voir tous les projets financés au titre de cet appel

Institution d’accueil

TECHNISCHE UNIVERSITAT NURNBERG
Contribution nette de l'UE
€ 1 499 115,00
Adresse
Dr.-Luise-Herzberg-Straße 4
90461 NUREMBERG
Allemagne
Région
Bayern Mittelfranken Nürnberg, Kreisfreie Stadt
Type d’activité
Higher or Secondary Education Establishments
Liens
Coût total
€ 1 499 115,00

Bénéficiaires (1)