PhD defense of Vincent LAURENT

Vincent LAURENT

Hybrid Learning Approaches for Industrial Applications

Abstract

This doctoral project is part of a technology transfer initiative between academic and industrial research on asset management topics. In these contexts, data is often missing or of poor quality, and the methods proposed in this thesis aim to overcome the difficulties consistently encountered in applications.

To address these challenges, we tackle two themes: active learning, for which we propose a new approach to a bipartite ranking problems. This method is based on an extension of K-armed bandit problems, for which we propose a theoretical analysis. The second theme deals with model hybridization methods, which we present and illustrate with industrial applications.

Direction

• Mathilde MOUGEOT

Jury

• Mme Madalina OLTEANU, Université Paris Dauphine - PSL, Rapporteur
• M. Jairo CUGLIARI, Université Lumière 2, Rapporteur
• Mme Christine KERIBIN, Université Paris-Saclay, Examinateur
• M. Erwan LE PENNEC, Ecole polytechnique, Examinateur
• Mme Anne SABOURIN, Université Paris Cité, Examinateur