Stochastic online optimization. Single-point and multi-point non-linear multi-armed bandits. Convex and strongly-convex case
- Авторлар: Gasnikov A.V.1,2, Krymova E.A.2, Lagunovskaya A.A.3,1, Usmanova I.N.1,2, Fedorenko F.A.1
-
Мекемелер:
- Moscow Institute of Physics and Technology (State University)
- Institute for Information Transmission Problems (Kharkevich Institute)
- Keldysh Institute of Applied Mathematics
- Шығарылым: Том 78, № 2 (2017)
- Беттер: 224-234
- Бөлім: Stochastic Systems, Queueing Systems
- URL: https://journals.rcsi.science/0005-1179/article/view/150534
- DOI: https://doi.org/10.1134/S0005117917020035
- ID: 150534
Дәйексөз келтіру
Аннотация
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.
Негізгі сөздер
Авторлар туралы
A. Gasnikov
Moscow Institute of Physics and Technology (State University); Institute for Information Transmission Problems (Kharkevich Institute)
Хат алмасуға жауапты Автор.
Email: gasnikov@yandex.ru
Ресей, Moscow; Moscow
E. Krymova
Institute for Information Transmission Problems (Kharkevich Institute)
Email: gasnikov@yandex.ru
Ресей, Moscow
A. Lagunovskaya
Keldysh Institute of Applied Mathematics; Moscow Institute of Physics and Technology (State University)
Email: gasnikov@yandex.ru
Ресей, Moscow; Moscow
I. Usmanova
Moscow Institute of Physics and Technology (State University); Institute for Information Transmission Problems (Kharkevich Institute)
Email: gasnikov@yandex.ru
Ресей, Moscow; Moscow
F. Fedorenko
Moscow Institute of Physics and Technology (State University)
Email: gasnikov@yandex.ru
Ресей, Moscow
Қосымша файлдар
