Skip to content
AI News HubLIVE
Original source2 min read

Classification with Large-Step Gradient Descent: Unifying Self-Stabilization and Convergence Rates

Summary

This paper extends a unified geometric perspective to exponential-tail losses in separable linear classification, identifies a curvature-defined reference subspace replacing the minimizer manifold, derives explicit iterate bounds revealing a three-phase damping structure, and uncovers a transient implicit bias induced by large step sizes.

SourcearXiv Machine LearningAuthor: Salma Tarmoun, Lachlan Ewen MacDonald, Konstantinos Emmanouilidis, Ziqing Xu, Ren\'e Vidal
Classification with Large-Step Gradient Descent: Unifying Self-Stabilization and Convergence Rates
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 12 Nov 2026]

Title:Classification with Large-Step Gradient Descent: Unifying Self-Stabilization and Convergence Rates

View a PDF of the paper titled Classification with Large-Step Gradient Descent: Unifying Self-Stabilization and Convergence Rates, by Salma Tarmoun and 4 other authors

View PDF HTML (experimental)

Abstract:Neural networks are often trained with gradient descent (GD) using step sizes that exceed classical stability thresholds, a regime where standard convergence theory no longer applies. Despite exhibiting non-monotonic loss and oscillatory dynamics, this regime frequently leads to faster convergence and better generalization. Existing theoretical perspectives offer complementary but incomplete insights: one identifies a self-stabilization mechanism that drives GD toward flat regions but does not yield convergence rates; another derives rigorous rates for specific losses but does not explain the underlying stabilization mechanism; and a third unifies these views for overparameterized least squares, but relies on the existence of a minimizer manifold, which is absent in classification settings. In this work, we extend this unified geometric perspective to exponential-tail losses, including logistic and cross-entropy losses in separable linear classification, where the loss infimum is not attained at any finite parameter. We identify a curvature-defined reference subspace that replaces the role of the minimizer manifold, yielding a coupled dynamical system in which orthogonal oscillations are damped by parallel progress along directions of decreasing sharpness. Under regularity conditions, which we verify for logistic and multiclass cross-entropy losses on symmetric data models, we derive explicit bounds for the iterates, not just the loss, covering both the unstable and stable regimes, and revealing a three-phase damping structure in the latter. Our analysis further uncovers a transient implicit bias induced by large step sizes: although GD converges asymptotically to the max-margin direction, large steps can leave persistent orthogonal residuals and, in certain geometries, amplify them, affecting finite-time alignment with the max-margin direction.

Subjects:

Machine Learning (cs.LG); Optimization and Control (math.OC)

Cite as: arXiv:2611.17459 [cs.LG]

(or arXiv:2611.17459v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2611.17459

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Salma Tarmoun [view email] [v1] Thu, 12 Nov 2026 19:36:52 UTC (1,709 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Classification with Large-Step Gradient Descent: Unifying Self-Stabilization and Convergence Rates, by Salma Tarmoun and 4 other authors

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.LG

new | recent | 2026-11

Change to browse by:

cs math math.OC

References & Citations

SciX

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

IArxiv recommender toggle

IArxiv Recommender (What is IArxiv?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

ResearchersAdvanced

Key points

  • Extends unified geometric analysis of large-step GD to exponential-tail losses (logistic, multiclass cross-entropy) in separable linear classification
  • Identifies a curvature-defined reference subspace replacing the minimizer manifold, yielding a coupled dynamical system where orthogonal oscillations are damped by parallel progress along decreasing sharpness
  • Derives explicit bounds for the iterates under regularity conditions, covering unstable and stable regimes, and reveals a three-phase damping structure
  • Uncovers a transient implicit bias: large steps can leave persistent orthogonal residuals and affect finite-time alignment with the max-margin direction

Highlights and analysis are generated automatically and may contain errors. Check the original source.