Saved in:
Bibliographic Details
Main Authors: Mensch, Zier, Holdijk, Lars, Duffield, Samuel, Aifer, Maxwell, Coles, Patrick J., Welling, Max, Cheng, Miranda C. N.
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.15925
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911453588488192
author Mensch, Zier
Holdijk, Lars
Duffield, Samuel
Aifer, Maxwell
Coles, Patrick J.
Welling, Max
Cheng, Miranda C. N.
author_facet Mensch, Zier
Holdijk, Lars
Duffield, Samuel
Aifer, Maxwell
Coles, Patrick J.
Welling, Max
Cheng, Miranda C. N.
contents Stochastic-gradient MCMC methods enable scalable Bayesian posterior sampling but often suffer from sensitivity to minibatch size and gradient noise. To address this, we propose Stochastic Gradient Lattice Random Walk (SGLRW), an extension of the Lattice Random Walk discretization. Unlike conventional Stochastic Gradient Langevin Dynamics (SGLD), SGLRW introduces stochastic noise only through the off-diagonal elements of the update covariance; this yields greater robustness to minibatch size while retaining asymptotic correctness. Furthermore, as comparison we analyze a natural analogue of SGLD utilizing gradient clipping. Experimental validation on Bayesian regression and classification demonstrates that SGLRW remains stable in regimes where SGLD fails, including in the presence of heavy-tailed gradient noise, and matches or improves predictive performance.
format Preprint
id arxiv_https___arxiv_org_abs_2602_15925
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Robust Stochastic Gradient Posterior Sampling with Lattice Based Discretisation
Mensch, Zier
Holdijk, Lars
Duffield, Samuel
Aifer, Maxwell
Coles, Patrick J.
Welling, Max
Cheng, Miranda C. N.
Machine Learning
Stochastic-gradient MCMC methods enable scalable Bayesian posterior sampling but often suffer from sensitivity to minibatch size and gradient noise. To address this, we propose Stochastic Gradient Lattice Random Walk (SGLRW), an extension of the Lattice Random Walk discretization. Unlike conventional Stochastic Gradient Langevin Dynamics (SGLD), SGLRW introduces stochastic noise only through the off-diagonal elements of the update covariance; this yields greater robustness to minibatch size while retaining asymptotic correctness. Furthermore, as comparison we analyze a natural analogue of SGLD utilizing gradient clipping. Experimental validation on Bayesian regression and classification demonstrates that SGLRW remains stable in regimes where SGLD fails, including in the presence of heavy-tailed gradient noise, and matches or improves predictive performance.
title Robust Stochastic Gradient Posterior Sampling with Lattice Based Discretisation
topic Machine Learning
url https://arxiv.org/abs/2602.15925