Many mathematical models are non-deterministic: it means that for several runs of the model with the same input values do not lead to the same output. In this framework, a natural indicator about the performance of the model for certain input values is the expectation of its associated ouputs. From an optimization point of view, it can be interesting to find the input values associated with the highest ouput mean. This presentation is about a state of the art to address optimization problems in this framework. We focus particularly on discrete optimization, i.e. when the set of input values is finite.
Concentration inequalities and bandit algorithms or how to find the perfect darts' partner.
- Se connecter pour poster des commentaires
Nom de l'orateur
Arthur Macherey
Etablissement de l'orateur
LMJL
Date et heure de l'exposé
Lieu de l'exposé
Salle des séminaires