Saved in:
Bibliographic Details
Main Authors: Choi, Hee-Sun, Han, Beom-Seok
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.06374
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910022912442368
author Choi, Hee-Sun
Han, Beom-Seok
author_facet Choi, Hee-Sun
Han, Beom-Seok
contents We introduce a multiplicative neural network architecture in which multiplicative interactions constitute the fundamental representation, rather than appearing as auxiliary components within an additive model. We establish a universal approximation theorem for this architecture and analyze its approximation properties in terms of locality and regularity in Bessel potential spaces. To complement the theoretical results, we conduct numerical experiments on representative targets exhibiting sharp transition layers or pointwise loss of higher-order regularity. The experiments focus on the spatial structure of approximation errors and on regularity-sensitive quantities, in particular, the convergence of Zygmund-type seminorms. The results show that the proposed multiplicative architecture yields residual error structures that are more tightly aligned with regions of reduced regularity and exhibit more stable convergence in regularity-sensitive metrics. These results demonstrate that adopting a multiplicative representation format has concrete implications for the localization and regularity behavior of neural network approximations, providing a direct connection between architectural design and analytical properties of the approximating functions.
format Preprint
id arxiv_https___arxiv_org_abs_2602_06374
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Multiplicative Neural Network Architecture: Locality and Regularity of Approximation
Choi, Hee-Sun
Han, Beom-Seok
Functional Analysis
Machine Learning
46E35, 41A30
We introduce a multiplicative neural network architecture in which multiplicative interactions constitute the fundamental representation, rather than appearing as auxiliary components within an additive model. We establish a universal approximation theorem for this architecture and analyze its approximation properties in terms of locality and regularity in Bessel potential spaces. To complement the theoretical results, we conduct numerical experiments on representative targets exhibiting sharp transition layers or pointwise loss of higher-order regularity. The experiments focus on the spatial structure of approximation errors and on regularity-sensitive quantities, in particular, the convergence of Zygmund-type seminorms. The results show that the proposed multiplicative architecture yields residual error structures that are more tightly aligned with regions of reduced regularity and exhibit more stable convergence in regularity-sensitive metrics. These results demonstrate that adopting a multiplicative representation format has concrete implications for the localization and regularity behavior of neural network approximations, providing a direct connection between architectural design and analytical properties of the approximating functions.
title A Multiplicative Neural Network Architecture: Locality and Regularity of Approximation
topic Functional Analysis
Machine Learning
46E35, 41A30
url https://arxiv.org/abs/2602.06374