Neural collapse in the orthoplex regime

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Alcala, James, Andreeva, Rayna, Kobzar, Vladimir A., Mixon, Dustin G., Na, Sanghoon, Sule, Shashank, Xie, Yangxinyu
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908903434878976
author Alcala, James
Andreeva, Rayna
Kobzar, Vladimir A.
Mixon, Dustin G.
Na, Sanghoon
Sule, Shashank
Xie, Yangxinyu
author_facet Alcala, James
Andreeva, Rayna
Kobzar, Vladimir A.
Mixon, Dustin G.
Na, Sanghoon
Sule, Shashank
Xie, Yangxinyu
contents When training a neural network for classification, the feature vectors of the training set are known to collapse to the vertices of a regular simplex, provided the dimension $d$ of the feature space and the number $n$ of classes satisfies $n\leq d+1$. This phenomenon is known as neural collapse. For other applications like language models, one instead takes $n\gg d$. Here, the neural collapse phenomenon still occurs, but with different emergent geometric figures. We characterize these geometric figures in the orthoplex regime where $d+2\leq n\leq 2d$. The techniques in our analysis primarily involve Radon's theorem and convexity.
format Preprint
id arxiv_https___arxiv_org_abs_2603_20587
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Neural collapse in the orthoplex regime
Alcala, James
Andreeva, Rayna
Kobzar, Vladimir A.
Mixon, Dustin G.
Na, Sanghoon
Sule, Shashank
Xie, Yangxinyu
Machine Learning
Information Theory
Metric Geometry
When training a neural network for classification, the feature vectors of the training set are known to collapse to the vertices of a regular simplex, provided the dimension $d$ of the feature space and the number $n$ of classes satisfies $n\leq d+1$. This phenomenon is known as neural collapse. For other applications like language models, one instead takes $n\gg d$. Here, the neural collapse phenomenon still occurs, but with different emergent geometric figures. We characterize these geometric figures in the orthoplex regime where $d+2\leq n\leq 2d$. The techniques in our analysis primarily involve Radon's theorem and convexity.
title Neural collapse in the orthoplex regime
topic Machine Learning
Information Theory
Metric Geometry
url https://arxiv.org/abs/2603.20587