Gaussian Approximation and Multiplier Bootstrap for Federated Linear Stochastic Approximation

Ilya Levin ∗, Maksim Shuklin ∗, Eric Moulines, Paul Mangold, Sergey Samsonov · 2026

Read Gaussian Approximation and Multiplier Bootstrap for Federated Linear Stochastic Approximation by Ilya Levin ∗, Maksim Shuklin ∗, Eric Moulines, Paul Mangold, Sergey Samsonov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-19629.txt.

Introduction

Levin, Shuklin, Moulines, Mangold and Samsonov prove Berry-Esseen type bounds for federated linear stochastic approximation. These are the first federated Gaussian approximations to capture trade offs between messages and compute, plus gaps between sites, in plain terms. Results cover constant step sizes and shrinking step sizes with growing local rounds, matching recent rates. Local step size, update count and site gaps enter the rates directly. They also build an online multiplier bootstrap for inference on the last iterate that avoids working out the long run covariance matrix.

Catalog record

CREATORIlya Levin ∗, Maksim Shuklin ∗, Eric Moulines, Paul Mangold, Sergey Samsonov
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE225163
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.19629

Browse the full library · About the Tower of Babel