Date : Nov. 20, 2024, 2 p.m. - Type : Thesis - Babacar SOW - Amphi A 104 - Espace Fauriel
Optimizing and metamodeling functions defined over clouds of points |
In this thesis, we address functions defined on sets of vectors.
Examples of such functions include electricity production simulators based on wind turbines and criteria for design of experiments.
We focus on the optimization of these functions when they are black-box type and their metamodeling when they are assumed to be costly.
For non-costly functions, the optimization methods used are evolutionary algorithms based on optimal transport. Metamodeling employs Gaussian processes with tailored kernels.
This work contributes to the field of optimization in both computer science and applied mathematics.
Jury:
Nathalie BARTOLI, Professor, ISAE-SUPAERO, Reviewer
Sébastien LE DIGABEL, Professor, Polytechnique Montréal, Reviewer
Sébastien DA VEIGA, Professor, ENSAI, Examiner
Viet Hung NGUYEN, Professor, UCA, Examiner
Rodolphe LE RICHE, Research Director, CNRS\LIMOS, Thesis Director
Julien PELAMATTI, R&D Engineer, EDF R&D, Co-supervisor
Merlin KELLER, R&D Engineer, EDF R&D, Co-supervisor
Sanaa ZANNANE, R&D Engineer, EDF R&D, Co-supervisor.