Estimation en ligne de l'inverse du Hessien pour l'optimisation stochastique

Jeudi 14 mars 2024, 10:15 à 11:15

Salle des séminaires (M.0.1)

Wei Lu

LMI (INSA Rouen)

This work addresses second-order stochastic optimization for estimating the minimizer of a convex function written as an expectation. A direct recursive estimation technique for the inverse Hessian matrix using a Robbins-Monro procedure is introduced. This approach enables to drastically reduces computational complexity. Above all, it allows to develop universal stochastic Newton methods and investigate the asymptotic efficiency of the proposed approach.