Adjusted Shuffling SARAH: Advancing Complexity Analysis via Dynamic Gradient Weighting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Nguyen, Duc Toan, Tran, Trang H., Nguyen, Lam M.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917536965066752
author Nguyen, Duc Toan
Tran, Trang H.
Nguyen, Lam M.
author_facet Nguyen, Duc Toan
Tran, Trang H.
Nguyen, Lam M.
contents In this paper, we propose Adjusted Shuffling SARAH, a novel algorithm that integrates shuffling strategies into the recursive SARAH framework using a dynamic weighting mechanism to enhance exploration. We analyze the algorithm under two operating modes. First, we show that the Exact Mode matches the best-known theoretical guarantees for shuffling variance-reduced methods in both strongly convex and non-convex settings. Second, to address large-scale regimes, we introduce an Inexact Mode that utilizes mini-batch estimators. A key contribution of our work is proving that this Inexact Mode achieves a total complexity independent of the dataset size, making it significantly more scalable than existing shuffling methods when the sample size is large.
format Preprint
id arxiv_https___arxiv_org_abs_2506_12444
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Adjusted Shuffling SARAH: Advancing Complexity Analysis via Dynamic Gradient Weighting
Nguyen, Duc Toan
Tran, Trang H.
Nguyen, Lam M.
Optimization and Control
Machine Learning
In this paper, we propose Adjusted Shuffling SARAH, a novel algorithm that integrates shuffling strategies into the recursive SARAH framework using a dynamic weighting mechanism to enhance exploration. We analyze the algorithm under two operating modes. First, we show that the Exact Mode matches the best-known theoretical guarantees for shuffling variance-reduced methods in both strongly convex and non-convex settings. Second, to address large-scale regimes, we introduce an Inexact Mode that utilizes mini-batch estimators. A key contribution of our work is proving that this Inexact Mode achieves a total complexity independent of the dataset size, making it significantly more scalable than existing shuffling methods when the sample size is large.
title Adjusted Shuffling SARAH: Advancing Complexity Analysis via Dynamic Gradient Weighting
topic Optimization and Control
Machine Learning
url https://arxiv.org/abs/2506.12444