Bonjour à tous,

Vous êtes cordialement invité.e.s à l'évaluation du Predoc III de Juan Ramirez, le 06 septembre, 9h30 am (mode hybride).

Jury

Abstract

Constrained optimization techniques enable us to enforce requirements such as fairness, robustness, or safety on machine learning models. Enforcing constraints is essential for ensuring compliance with government regulations and safety guidelines, which is important given the growing adoption of artificial intelligence tools in safety-critical applications. In this presentation, I will cover fundamental constrained optimization techniques applicable to general nonconvex constrained optimization problems, such as those arising in deep learning applications. Additionally, I will present our previous work aimed at ensuring that training machine learning models with constraints is as easy as training without them.