Tracy–Widom fluctuations for the row-centered conjugate kernel of a random network with polynomial activation

Let $Y=\sigma(WX/\sqrt d)$, where W is an m × d matrix with independent standard Gaussian entries, the N columns of X are independent and uniform on the sphere of radius $\sqrt d$, and σ is a fixed polynomial with $\mathbb{E}\sigma(Z)=\mathbb{E}[Z\sigma(Z)]=0$ for a standard Gaussian Z. Let $m,d,N\to\infty$ proportionally with $m\ne N$ asymptotically, and let K_P = YPY^T/N, where P removes the mean of each row of Y over the N data points. We prove that the largest eigenvalue of K_P has Tracy–Widom fluctuations of order N^-2/3 around $v(1+\sqrt{m/N})^2$, where $v=\mathbb{E}\sigma(Z)^2$. The same holds for the largest eigenvalue of YY^T/N when a_2^2N/d stays below $v\sqrt{N/m}$, where a_2 is the second Hermite coefficient of σ. The proof applies, conditionally on the data, the edge universality theorem of Chen and Lyu for sample covariance matrices whose sample vectors satisfy a bound on quadratic forms. For polynomial features this bound follows from the Gaussian Poincaré inequality and hypercontractivity, with constants independent of the dimension.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-15
DOI
https://doi.org/10.5281/zenodo.24246776
Primary Topic
Random Matrices and Applications
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Tracy–Widom fluctuations for the row-centered conjugate kernel of a random network with polynomial activation

Yitzchak Shmalo
Zenodo (CERN European Organization for Nuclear Research)
Random Matrices and Applications
preprint

Tracy–Widom fluctuations for the row-centered conjugate kernel of a random network with polynomial activation

Yitzchak Shmalo
preprint en

Abstract

Let $Y=\sigma(WX/\sqrt d)$, where W is an m × d matrix with independent standard Gaussian entries, the N columns of X are independent and uniform on the sphere of radius $\sqrt d$, and σ is a fixed polynomial with $\mathbb{E}\sigma(Z)=\mathbb{E}[Z\sigma(Z)]=0$ for a standard Gaussian Z. Let $m,d,N\to\infty$ proportionally with $m\ne N$ asymptotically, and let K_P = YPY^T/N, where P removes the mean of each row of Y over the N data points. We prove that the largest eigenvalue of K_P has Tracy–Widom fluctuations of order N^-2/3 around $v(1+\sqrt{m/N})^2$, where $v=\mathbb{E}\sigma(Z)^2$. The same holds for the largest eigenvalue of YY^T/N when a_2^2N/d stays below $v\sqrt{N/m}$, where a_2 is the second Hermite coefficient of σ. The proof applies, conditionally on the data, the edge universality theorem of Chen and Lyu for sample covariance matrices whose sample vectors satisfy a bound on quadratic forms. For polynomial features this bound follows from the Gaussian Poincaré inequality and hypercontractivity, with constants independent of the dimension.

Zenodo (CERN European Organization for Nuclear Research)
Hebrew University of Jerusalem (IL)
Random Matrices and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon