Empowering trustworthy AI with reliable predictive uncertainty quantification

Webinaire – Vendredi 12 janvier de 15h à 16h

The growing adoption of AI models in critical applications (such as transportation and healthcare) amplifies the risk of their inherent uncertainty, potentially resulting in erroneous decisions and catastrophic failures. Therefore, uncertainty quantification (UQ) plays a central role in guarantying AI model reliability and trustworthiness in these domains.

To address this critical need, we introduce our new open-source Python library Puncc. Puncc provides a comprehensive framework for conformal prediction, a powerful UQ technique that guarantees reliable evaluation of predictive uncertainty under mild assumptions. Leveraging Puncc’s user-friendly API, users can effortlessly obtain uncertainty estimates without the need to retrain their models, enabling informed decision-making and risk assessment for diverse machine learning tasks including regression, classification, object detection, and anomaly detection.

The presentation will be given by Mouchine MENDIL Ingénieur IA – Data Scientist IRT Saint Exupéry

Veuillez laisser ce champ vide

Ne manquez rien !

Inscrivez-vous pour recevoir l'actualité d'ANITI chaque mois.

Nom *

Prénom *

Adresse e-mail *

Nous n’envoyons pas de messages indésirables !

Vérifiez votre boite de réception ou votre répertoire d’indésirables pour confirmer votre abonnement.