Low Rank Gradients and Where to Find Them

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sonthalia, Rishi, Murray, Michael, Montúfar, Guido
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911188596555776
author Sonthalia, Rishi
Murray, Michael
Montúfar, Guido
author_facet Sonthalia, Rishi
Murray, Michael
Montúfar, Guido
contents This paper investigates low-rank structure in the gradients of the training loss for two-layer neural networks while relaxing the usual isotropy assumptions on the training data and parameters. We consider a spiked data model in which the bulk can be anisotropic and ill-conditioned, we do not require independent data and weight matrices and we also analyze both the mean-field and neural-tangent-kernel scalings. We show that the gradient with respect to the input weights is approximately low rank and is dominated by two rank-one terms: one aligned with the bulk data-residue , and another aligned with the rank one spike in the input data. We characterize how properties of the training data, the scaling regime and the activation function govern the balance between these two components. Additionally, we also demonstrate that standard regularizers, such as weight decay, input noise and Jacobian penalties, also selectively modulate these components. Experiments on synthetic and real data corroborate our theoretical predictions.
format Preprint
id arxiv_https___arxiv_org_abs_2510_01303
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Low Rank Gradients and Where to Find Them
Sonthalia, Rishi
Murray, Michael
Montúfar, Guido
Machine Learning
Artificial Intelligence
This paper investigates low-rank structure in the gradients of the training loss for two-layer neural networks while relaxing the usual isotropy assumptions on the training data and parameters. We consider a spiked data model in which the bulk can be anisotropic and ill-conditioned, we do not require independent data and weight matrices and we also analyze both the mean-field and neural-tangent-kernel scalings. We show that the gradient with respect to the input weights is approximately low rank and is dominated by two rank-one terms: one aligned with the bulk data-residue , and another aligned with the rank one spike in the input data. We characterize how properties of the training data, the scaling regime and the activation function govern the balance between these two components. Additionally, we also demonstrate that standard regularizers, such as weight decay, input noise and Jacobian penalties, also selectively modulate these components. Experiments on synthetic and real data corroborate our theoretical predictions.
title Low Rank Gradients and Where to Find Them
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2510.01303