Inertia in Optimization: Acceleration and Adaptivity
Hippolyte Labarriere (post-doctorat au Laboratoire d'apprentissage informatique et statistique (LCSL), University of Genova)
| Quand ? |
Le 22/10/2024, de 13:00 à 14:00 |
|---|---|
| Participants |
Hippolyte Labarriere |
| Ajouter un événement au calendrier |
vCal
iCal
|
HIPPOLYTE Labarriere
Title: Inertia in Optimization: Acceleration and Adaptivity
Abstract:
In this talk, we will give an overview of inertial methods in convex optimization. Some useful tools and concepts will be presented before emphasizing the importance of the geometry of the function of interest in choosing the right inertial scheme. This will raise the issue of adaptivity, since such methods require an a priori knowledge that may not be available. In this context, we will provide techniques to overcome this problem such as backtracking and restart.
In this talk, we will give an overview of inertial methods in convex optimization. Some useful tools and concepts will be presented before emphasizing the importance of the geometry of the function of interest in choosing the right inertial scheme. This will raise the issue of adaptivity, since such methods require an a priori knowledge that may not be available. In this context, we will provide techniques to overcome this problem such as backtracking and restart.
Website: https://hippolytelbrrr.github.io/
In Room M7 101, 1st floor, Monod campus, ENSL.