On the SAGA Algorithm with Decreasing Step
SIAM Journal on Control and Optimization · 31.12.2026 · 64 (6) · 4228-4253
DOI: 10.1137/25m179748x ↗Kurz gefasst
Fachartikel in der Zeitschrift SIAM Journal on Control and Optimization, online erschienen am 16.11.2026. Verfasst von Bernard Bercu und 2 weiteren Autor:innen. Die Arbeit nennt 44 Literaturangaben. Gefördert von: Agence Nationale de la Recherche.
Abstract (Originalsprache)
Stochastic optimization naturally appears in many application areas, including machine learning. Our goal is to go further in the analysis of the stochastic average gradient accelerated (SAGA) algorithm. To achieve this, we introduce a new [Formula: see text]-SAGA algorithm which interpolates between the stochastic gradient descent ([Formula: see text]) and the SAGA algorithm ([Formula: see text]). First, we investigate the almost sure convergence of this new algorithm with decreasing step which allows us to avoid the restrictive strong convexity and Lipschitz gradient hypotheses associated with the objective function. Second, we establish a central limit theorem for the [Formula: see text]-SAGA algorithm. Finally, we provide the nonasymptotic [Formula: see text] rates of convergence.
Förderung
- Agence Nationale de la Recherche
Zitieren
APA
Bercu, B., Fredes, L., & Gbaguidi, E. (2026). On the SAGA Algorithm with Decreasing Step. SIAM Journal on Control and Optimization, 64(6), 4228-4253. https://doi.org/10.1137/25m179748x
Vancouver
Bercu B, Fredes L, Gbaguidi E. On the SAGA Algorithm with Decreasing Step. SIAM Journal on Control and Optimization. 2026;64(6):4228-4253. doi:10.1137/25m179748x
BibTeX
@article{Bercu2026,
author = {Bercu, Bernard and Fredes, Luis and Gbaguidi, Eméric},
title = {On the SAGA Algorithm with Decreasing Step},
journal = {SIAM Journal on Control and Optimization},
year = {2026},
volume = {64},
number = {6},
pages = {4228--4253},
doi = {10.1137/25m179748x}
}
EUVIMED