OSSOBUCCO
ANR JCJC project ANR-25-CE23-5526 (2025-2029)
Description
OSSOBUCCO
(On Strict Saddle Optimization: Benchmarking, Unified Classification and COmplexity) aims at developing benchmarks for tractable nonconvex optimization formulations. Unlike existing, generic nonlinear optimization test sets, we plan on focusing on classes of nonconvex problems with so-called benign landscape. This benchmark will enable a proper classification of such nonconvex instances as well as development of efficient algorithms in a complexity sense. A starting point will be the strict saddle property that was established for a number of nonconvex data science tasks.
Resources and publications
C. Royer gave a seminar talk on strict saddle optimization in April, 2026.
Open positions
Postdoctoral position A postdoctoral position is available in the context of OSSOBUCCO.
The successful applicant will have a PhD in applied mathematics, computer science, or a related area, with strong experience with continuous optimization methods. The applicant will be expected to conduct research in the area of nonconvex optimization for data science. Topics of interest include high-order optimization methods and landscape analysis.
Interested candidates should contact Clément Royer for further discussion on this opening.
La version française de cette page est disponible ici.