Identifiability of Deep Polynomial Neural Networks

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Usevich, Konstantin, Borsoi, Ricardo, Dérand, Clara, Clausel, Marianne
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917235761610752
author Usevich, Konstantin
Borsoi, Ricardo
Dérand, Clara
Clausel, Marianne
author_facet Usevich, Konstantin
Borsoi, Ricardo
Dérand, Clara
Clausel, Marianne
contents Polynomial Neural Networks (PNNs) possess a rich algebraic and geometric structure. However, their identifiability -- a key property for ensuring interpretability -- remains poorly understood. In this work, we present a comprehensive analysis of the identifiability of deep PNNs, including architectures with and without bias terms. Our results reveal an intricate interplay between activation degrees and layer widths in achieving identifiability. As special cases, we show that architectures with non-increasing layer widths are generically identifiable under mild conditions, while encoder-decoder networks are identifiable when the decoder widths do not grow too rapidly compared to the activation degrees. Our proofs are constructive and center on a connection between deep PNNs and low-rank tensor decompositions, and Kruskal-type uniqueness theorems. We also settle an open conjecture on the dimension of PNN's neurovarieties, and provide new bounds on the activation degrees required for it to reach the expected dimension.
format Preprint
id arxiv_https___arxiv_org_abs_2506_17093
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Identifiability of Deep Polynomial Neural Networks
Usevich, Konstantin
Borsoi, Ricardo
Dérand, Clara
Clausel, Marianne
Machine Learning
Artificial Intelligence
Algebraic Geometry
68T07, 62R01, 15A69, 14M99
Polynomial Neural Networks (PNNs) possess a rich algebraic and geometric structure. However, their identifiability -- a key property for ensuring interpretability -- remains poorly understood. In this work, we present a comprehensive analysis of the identifiability of deep PNNs, including architectures with and without bias terms. Our results reveal an intricate interplay between activation degrees and layer widths in achieving identifiability. As special cases, we show that architectures with non-increasing layer widths are generically identifiable under mild conditions, while encoder-decoder networks are identifiable when the decoder widths do not grow too rapidly compared to the activation degrees. Our proofs are constructive and center on a connection between deep PNNs and low-rank tensor decompositions, and Kruskal-type uniqueness theorems. We also settle an open conjecture on the dimension of PNN's neurovarieties, and provide new bounds on the activation degrees required for it to reach the expected dimension.
title Identifiability of Deep Polynomial Neural Networks
topic Machine Learning
Artificial Intelligence
Algebraic Geometry
68T07, 62R01, 15A69, 14M99
url https://arxiv.org/abs/2506.17093