sur ce site

Accueil du site > Résumés des séminaires > Labo > Résumés > Numerical Stochastic Black-box Optimization

Numerical Stochastic Black-box Optimization

This presentation will start by a gentle introduction to the domain of black-box numerical optimization. We will give a few applications where black-box algorithms and particularly stochastic adaptive algorithms are needed. We will explain the main difficulties and challenges that need to be addressed in order to solve black-box optimization problems and how they are currently tackled by state-of-the-art algorithms like CMA-ES.

We will then present some connexions between stochastic adaptive optimization algorithms and MCMC on the one hand and information geometry on the other hand.

We will finish by discussing open problems in terms of algorithm design and theory.

CMAP UMR 7641 École Polytechnique CNRS, Route de Saclay, 91128 Palaiseau Cedex France, Tél: +33 1 69 33 46 23 Fax: +33 1 69 33 46 46