Sequential stochastic blackbox optimization with zeroth order gradient estimators - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Sequential stochastic blackbox optimization with zeroth order gradient estimators

Résumé

This work considers stochastic optimization problems in which the objective function values can only be computed by a blackbox corrupted by some random noise following an unknown distribution. The proposed method is based on sequential stochastic optimization (SSO): the original problem is decomposed into a sequence of subproblems. Each of these subproblems is solved using a zeroth order version of a sign stochastic gradient descent with momentum algorithm (ZOSignum) and with an increasingly fine precision. This decomposition allows a good exploration of the space while maintaining the efficiency of the algorithm once it gets close to the solution. Under Lipschitz continuity assumption on the blackbox, a convergence rate in expectation is derived for the ZO-signum algorithm. Moreover, if the blackbox is smooth and convex or locally convex around its minima, a convergence rate to an epsilon-optimal point of the problem may be obtained for the SSO algorithm. Numerical experiments are conducted to compare the SSO algorithm with other state-of-the-art algorithms and to demonstrate its competitiveness.
Fichier non déposé

Dates et versions

hal-04292913 , version 1 (17-11-2023)

Identifiants

  • HAL Id : hal-04292913 , version 1

Citer

Jean Bigeon, Charles Audet, Michael Kokkolaras, Romain Couderc. Sequential stochastic blackbox optimization with zeroth order gradient estimators. MOPTA2023, Aug 2023, Bethlehem, PA, USA, United States. ⟨hal-04292913⟩
15 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More