Work

Last-Iterate Convergence of Clipped SGD with Linear Decay

Year: 2026

Type: preprint

Abstract: In this paper, we study the last-iterate convergence of clipped stochastic gradient descent under heavy-tailed noise. We assume that the stochastic gradients are unbiased and that the noise has a bounded central $p$-th moment, where $1

Source: arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.18555

Authors: Jiang, Wei, Bai, Haomin, Li, Zechao

Institutions: Nanjing University of Science and Technology, Nanjing University

Language: English

Topic: Stochastic Gradient Optimization Techniques

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7221794827

Cites:

Cited by: 1

Related to: 10