Spectral Dynamics in Deep Networks: Feature Learning, Outlier Escape, and Learning Rate Transfer

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Lauditi, Clarissa, Pehlevan, Cengiz, Bordelon, Blake
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916034805497856
author Lauditi, Clarissa
Pehlevan, Cengiz
Bordelon, Blake
author_facet Lauditi, Clarissa
Pehlevan, Cengiz
Bordelon, Blake
contents We study the evolution of hidden-weight spectra in wide neural networks trained by (stochastic) gradient descent. We develop a two-level dynamical mean-field theory (DMFT) that jointly tracks bulk and outlier spectral dynamics for spiked ensembles whose spike directions remain statistically dependent on the random bulk. We apply this framework to two settings: (1) infinite-width nonlinear networks in mean-field/$μ$P scaling and (2) deep linear networks in the proportional high-dimensional limit, where width, input dimension, and sample size diverge with fixed ratios. Our theory predicts how outliers evolve with training time, width, output scale, and initialization variance. In deep linear networks, $μ$P yields width-consistent outlier dynamics and hyperparameter transfer, including width-stable growth of the leading NTK mode toward the edge of stability (EoS). In contrast, NTK parameterization exhibits strongly width-dependent outlier dynamics, despite converging to a stable large-width limit. We show that this bulk+outlier picture is descriptive of simple tasks with small output channels, but that tasks involving large numbers of outputs (ImageNet classification or GPT language modeling) are better described by a restructuring of the spectral bulk. We develop a toy model with extensive output channels that recapitulates this phenomenon and show that edge of the spectrum still converges for sufficiently wide networks.
format Preprint
id arxiv_https___arxiv_org_abs_2605_07870
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Spectral Dynamics in Deep Networks: Feature Learning, Outlier Escape, and Learning Rate Transfer
Lauditi, Clarissa
Pehlevan, Cengiz
Bordelon, Blake
Disordered Systems and Neural Networks
Artificial Intelligence
Machine Learning
We study the evolution of hidden-weight spectra in wide neural networks trained by (stochastic) gradient descent. We develop a two-level dynamical mean-field theory (DMFT) that jointly tracks bulk and outlier spectral dynamics for spiked ensembles whose spike directions remain statistically dependent on the random bulk. We apply this framework to two settings: (1) infinite-width nonlinear networks in mean-field/$μ$P scaling and (2) deep linear networks in the proportional high-dimensional limit, where width, input dimension, and sample size diverge with fixed ratios. Our theory predicts how outliers evolve with training time, width, output scale, and initialization variance. In deep linear networks, $μ$P yields width-consistent outlier dynamics and hyperparameter transfer, including width-stable growth of the leading NTK mode toward the edge of stability (EoS). In contrast, NTK parameterization exhibits strongly width-dependent outlier dynamics, despite converging to a stable large-width limit. We show that this bulk+outlier picture is descriptive of simple tasks with small output channels, but that tasks involving large numbers of outputs (ImageNet classification or GPT language modeling) are better described by a restructuring of the spectral bulk. We develop a toy model with extensive output channels that recapitulates this phenomenon and show that edge of the spectrum still converges for sufficiently wide networks.
title Spectral Dynamics in Deep Networks: Feature Learning, Outlier Escape, and Learning Rate Transfer
topic Disordered Systems and Neural Networks
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2605.07870