ExplorerArtificial IntelligenceAI
Research PaperResearchia:202602.14024

On the implicit regularization of Langevin dynamics with projected noise

Govind Menon

Abstract

We study Langevin dynamics with noise projected onto the directions orthogonal to an isometric group action. This mathematical model is introduced to shed new light on the effects of symmetry on stochastic gradient descent for over-parametrized models. Our main result identifies a novel form of implicit regularization: when the initial and target density are both invariant under the group action, Langevin dynamics with projected noise is equivalent in law to Langevin dynamics with isotropic diff...

Submitted: February 14, 2026Subjects: AI; Artificial Intelligence

Description / Details

We study Langevin dynamics with noise projected onto the directions orthogonal to an isometric group action. This mathematical model is introduced to shed new light on the effects of symmetry on stochastic gradient descent for over-parametrized models. Our main result identifies a novel form of implicit regularization: when the initial and target density are both invariant under the group action, Langevin dynamics with projected noise is equivalent in law to Langevin dynamics with isotropic diffusion but with an additional drift term proportional to the negative log volume of the group orbit. We prove this result by constructing a coupling of the two processes via a third process on the group itself, and identify the additional drift as the mean curvature of the orbits.


Source: arXiv:2602.12257v1 - http://arxiv.org/abs/2602.12257v1 PDF: https://arxiv.org/pdf/2602.12257v1 Original Link: http://arxiv.org/abs/2602.12257v1

Please sign in to join the discussion.

No comments yet. Be the first to share your thoughts!

Access Paper
View Source PDF
Submission Info
Date:
Feb 14, 2026
Topic:
Artificial Intelligence
Area:
AI
Comments:
0
Bookmark
On the implicit regularization of Langevin dynamics with projected noise | Researchia