Nonsmooth optimization

The activities around nonsmooth optimization are driven by application problems (in operation research, numerical mechanics, decision making, robotics, identification...), which require the development of algorithms, the latter motivating in turn fundamental research. These three aspects are interspersed in our research work.

Main research themes

• Connexion between continuous and combinatorial optimization - via duality in convex optimization.
• Optimization methods over the cone of positive semidefinite matrices - applications in combinatorial optimization, in control theory.
• Algorithms for large-scale machine learning.
• Second-order approximation of structured nonsmooth functions.

International and industrial relations

• Industrial firms: EDF, meteorological and geophysical institutes.
• Other INRIA projects as Idopt, Mystis in particular.
• Academic laboratories: Operations Research Department of Cornell University, Math departments of Guelph University and Barcelona Autonoma Univeristy, LAAS (CNRS Toulouse), Leibniz Institute (Grenoble University)...