A ZeNN architecture to avoid the Gaussian trap

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Carvalho, Luís, Costa, João L., Mourão, José, Oliveira, Gonçalo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908473728434176
author Carvalho, Luís
Costa, João L.
Mourão, José
Oliveira, Gonçalo
author_facet Carvalho, Luís
Costa, João L.
Mourão, José
Oliveira, Gonçalo
contents We propose a new simple architecture, Zeta Neural Networks (ZeNNs), in order to overcome several shortcomings of standard multi-layer perceptrons (MLPs). Namely, in the large width limit, MLPs are non-parametric, they do not have a well-defined pointwise limit, they lose non-Gaussian attributes and become unable to perform feature learning; moreover, finite width MLPs perform poorly in learning high frequencies. The new ZeNN architecture is inspired by three simple principles from harmonic analysis: i) Enumerate the perceptons and introduce a non-learnable weight to enforce convergence; ii) Introduce a scaling (or frequency) factor; iii) Choose activation functions that lead to near orthogonal systems. We will show that these ideas allow us to fix the referred shortcomings of MLPs. In fact, in the infinite width limit, ZeNNs converge pointwise, they exhibit a rich asymptotic structure beyond Gaussianity, and perform feature learning. Moreover, when appropriate activation functions are chosen, (finite width) ZeNNs excel at learning high-frequency features of functions with low dimensional domains.
format Preprint
id arxiv_https___arxiv_org_abs_2505_20553
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A ZeNN architecture to avoid the Gaussian trap
Carvalho, Luís
Costa, João L.
Mourão, José
Oliveira, Gonçalo
Machine Learning
Probability
68T07, 68T01
I.2.0; G.0
We propose a new simple architecture, Zeta Neural Networks (ZeNNs), in order to overcome several shortcomings of standard multi-layer perceptrons (MLPs). Namely, in the large width limit, MLPs are non-parametric, they do not have a well-defined pointwise limit, they lose non-Gaussian attributes and become unable to perform feature learning; moreover, finite width MLPs perform poorly in learning high frequencies. The new ZeNN architecture is inspired by three simple principles from harmonic analysis: i) Enumerate the perceptons and introduce a non-learnable weight to enforce convergence; ii) Introduce a scaling (or frequency) factor; iii) Choose activation functions that lead to near orthogonal systems. We will show that these ideas allow us to fix the referred shortcomings of MLPs. In fact, in the infinite width limit, ZeNNs converge pointwise, they exhibit a rich asymptotic structure beyond Gaussianity, and perform feature learning. Moreover, when appropriate activation functions are chosen, (finite width) ZeNNs excel at learning high-frequency features of functions with low dimensional domains.
title A ZeNN architecture to avoid the Gaussian trap
topic Machine Learning
Probability
68T07, 68T01
I.2.0; G.0
url https://arxiv.org/abs/2505.20553