This talk will explore the importance of uncertainty quantification in AI, highlighting its role in assessing the reliability and robustness of machine learning models.

We will discuss different types of uncertainty, such as aleatoric and epistemic, along with key methods used to quantify them with a focus on conformal prediction, a versatile and model-agnostic approach for generating reliable prediction intervals with guaranteed coverage.

To illustrate its practical utility, the talk will conclude with an example application, demonstrating how conformal prediction can enhance decision-making in real-world scenarios

Lieu
LOCEAN, 4^e étage, salle 417