English

A Spectral Condition for Feature Learning

Machine Learning 2024-05-15 v2

Abstract

The push to train ever larger neural networks has motivated the study of initialization and training at large network width. A key challenge is to scale training so that a network's internal representations evolve nontrivially at all widths, a process known as feature learning. Here, we show that feature learning is achieved by scaling the spectral norm of weight matrices and their updates like fan-out/fan-in\sqrt{\texttt{fan-out}/\texttt{fan-in}}, in contrast to widely used but heuristic scalings based on Frobenius norm and entry size. Our spectral scaling analysis also leads to an elementary derivation of \emph{maximal update parametrization}. All in all, we aim to provide the reader with a solid conceptual understanding of feature learning in neural networks.

Keywords

Cite

@article{arxiv.2310.17813,
  title  = {A Spectral Condition for Feature Learning},
  author = {Greg Yang and James B. Simon and Jeremy Bernstein},
  journal= {arXiv preprint arXiv:2310.17813},
  year   = {2024}
}
R2 v1 2026-06-28T13:03:20.947Z