PreprintCC BY 4.0arXiv
Polynomial Residual Networks are Generically Identifiable
Abstract
We prove identifiability results for constant-width residual networks with polynomial activations and bias terms. For activations of degree at least five with generic coefficients, the fiber over a generic parameter is a singleton. For activations of degree at least three with no linear term after a suitable translation of the input, the generic fiber is given by translating biases and then scaling weights and translated biases by real roots of unity. These results specialize to networks without bias terms. With the same proof technique, we further derive necessary conditions for critical points of the ResNet parametrization, that are central in singular learning theory.
Related
- Approximation of solutions of parameter-dependent problems by residual neural networks
- LAIR-Net: Leaky Alignment-Impulse Residual Networks for Tabular Regression
- Width and Depth Limits Commute During Training in Linear Residual Networks
- Adaptive Randomized Neural Networks for PDEs: Structure-Aware Feature Redistribution and Residual-Driven Block Enrichment
- Algebraic Expressivity Certificates for Shallow Polynomial Neural Networks
BibTeX
@misc{dahlgren2026polynomial,
title = {{Polynomial Residual Networks are Generically Identifiable}},
author = {Isabel Dahlgren and Giovanni Luca Marchetti and Kathlén Kohn},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.33738},
eprint = {2610.33738},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.33738},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Polynomial Residual Networks are Generically Identifiable
AU - Isabel Dahlgren
AU - Giovanni Luca Marchetti
AU - Kathlén Kohn
PY - 2026
DA - 2026/10/24
PB - arXiv
DO - 10.48550/arxiv.2610.33738
UR - https://arxiv.org/abs/2610.33738
AB - We prove identifiability results for constant-width residual networks with polynomial activations and bias terms. For activations of degree at least five with generic coefficients, the fiber over a generic parameter is a singleton. For activations of degree at least three with no linear term after a suitable translation of the input, the generic fiber is given by translating biases and then scaling weights and translated biases by real roots of unity. These results specialize to networks without bias terms. With the same proof technique, we further derive necessary conditions for critical points of the ResNet parametrization, that are central in singular learning theory.
ER -