A. V. Gasnikov, E. A. Krymova, A. A. Lagunovskaya, I. N. Usmanova, F. A. Fedorenko, “Stochastic online optimization. Single-point and multi-point non-linear multi-armed bandits. Convex and strongly-convex case”, Avtomat. i Telemekh., 2017, Issue 2,Pages <nobr>36

This article is cited in 35 papers

Stochastic Systems, Queuing Systems

Stochastic online optimization. Single-point and multi-point non-linear multi-armed bandits. Convex and strongly-convex case

A. V. Gasnikov^ab, E. A. Krymova^b, A. A. Lagunovskaya^ca, I. N. Usmanova^ab, F. A. Fedorenko^a

^a Moscow Institute of Physics and Technology (State University), Moscow, Russia
^b Institute for Information Transmission Problems (Kharkevich Institute), Russian Academy of Sciences, Moscow, Russia
^c Keldysh Institute of Applied Mathematics, Russian Academy of Sciences, Moscow, Russia

Abstract: In this paper the gradient-free modification of the mirror descent method for convex stochastic online optimization problems is proposed. The crucial assumption in the problem setting is that function realizations are observed with minor noises. The aim of this paper is to derive the convergence rate of the proposed methods and to determine a noise level which does not significantly affect the convergence rate.

Keywords: online optimization, gradient-free, inexact oracle, stochastic optimization.

Presented by the member of Editorial Board: P. S. Shcherbakov

Received: 16.10.2014