Saved in:
Bibliographic Details
Main Authors: Arous, Gérard Ben, Erdogdu, Murat A., Vural, Nuri Mert, Wu, Denny
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.03688
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909978418216960
author Arous, Gérard Ben
Erdogdu, Murat A.
Vural, Nuri Mert
Wu, Denny
author_facet Arous, Gérard Ben
Erdogdu, Murat A.
Vural, Nuri Mert
Wu, Denny
contents We study the optimization and sample complexity of gradient-based training of a two-layer neural network with quadratic activation function in the high-dimensional regime, where the data is generated as $f_*(\boldsymbol{x}) \propto \sum_{j=1}^{r}λ_j σ\left(\langle \boldsymbol{θ_j}, \boldsymbol{x}\rangle\right), \boldsymbol{x} \sim N(0,\boldsymbol{I}_d)$, $σ$ is the 2nd Hermite polynomial, and $\lbrace\boldsymbolθ_j \rbrace_{j=1}^{r} \subset \mathbb{R}^d$ are orthonormal signal directions. We consider the extensive-width regime $r \asymp d^β$ for $β\in [0, 1)$, and assume a power-law decay on the (non-negative) second-layer coefficients $λ_j\asymp j^{-α}$ for $α\geq 0$. We present a sharp analysis of the SGD dynamics in the feature learning regime, for both the population limit and the finite-sample (online) discretization, and derive scaling laws for the prediction risk that highlight the power-law dependencies on the optimization time, sample size, and model width. Our analysis combines a precise characterization of the associated matrix Riccati differential equation with novel matrix monotonicity arguments to establish convergence guarantees for the infinite-dimensional effective dynamics.
format Preprint
id arxiv_https___arxiv_org_abs_2508_03688
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning quadratic neural networks in high dimensions: SGD dynamics and scaling laws
Arous, Gérard Ben
Erdogdu, Murat A.
Vural, Nuri Mert
Wu, Denny
Machine Learning
We study the optimization and sample complexity of gradient-based training of a two-layer neural network with quadratic activation function in the high-dimensional regime, where the data is generated as $f_*(\boldsymbol{x}) \propto \sum_{j=1}^{r}λ_j σ\left(\langle \boldsymbol{θ_j}, \boldsymbol{x}\rangle\right), \boldsymbol{x} \sim N(0,\boldsymbol{I}_d)$, $σ$ is the 2nd Hermite polynomial, and $\lbrace\boldsymbolθ_j \rbrace_{j=1}^{r} \subset \mathbb{R}^d$ are orthonormal signal directions. We consider the extensive-width regime $r \asymp d^β$ for $β\in [0, 1)$, and assume a power-law decay on the (non-negative) second-layer coefficients $λ_j\asymp j^{-α}$ for $α\geq 0$. We present a sharp analysis of the SGD dynamics in the feature learning regime, for both the population limit and the finite-sample (online) discretization, and derive scaling laws for the prediction risk that highlight the power-law dependencies on the optimization time, sample size, and model width. Our analysis combines a precise characterization of the associated matrix Riccati differential equation with novel matrix monotonicity arguments to establish convergence guarantees for the infinite-dimensional effective dynamics.
title Learning quadratic neural networks in high dimensions: SGD dynamics and scaling laws
topic Machine Learning
url https://arxiv.org/abs/2508.03688