The Positivity of the Neural Tangent Kernel
SIAM Journal on Mathematics of Data Science · 30.06.2025 · 7 (2) · 495-515
DOI: 10.1137/24m1659534 ↗Kurz gefasst
Fachartikel in der Zeitschrift SIAM Journal on Mathematics of Data Science, online erschienen am 24.04.2025. Verfasst von Luís Carvalho und 3 weiteren Autor:innen. Die Arbeit wurde bisher 6-mal zitiert (Crossref) und nennt 28 Literaturangaben. Gefördert von: Fundação para a Ciência e a Tecnologia.
Abstract (Originalsprache)
The Neural tangent kernel (NTK) has emerged as a fundamental concept in the study of wide neural networks. In particular, it is known that the positivity of the NTK is directly related to the memorization capacity of sufficiently wide networks, i.e., to the possibility of reaching zero loss in training via gradient descent. Here we will improve on previous works and obtain a sharp result concerning the positivity of the NTK of feedforward networks of any depth. More precisely, we will show that, for any nonpolynomial activation function, the NTK is strictly positive definite. Our results are based on a novel characterization of polynomial functions, which is of independent interest.
Förderung
- Fundação para a Ciência e a Tecnologia
Zitieren
APA
Carvalho, L., Costa, J. L., Mourão, J., & Oliveira, G. (2025). The Positivity of the Neural Tangent Kernel. SIAM Journal on Mathematics of Data Science, 7(2), 495-515. https://doi.org/10.1137/24m1659534
Vancouver
Carvalho L, Costa JL, Mourão J, Oliveira G. The Positivity of the Neural Tangent Kernel. SIAM Journal on Mathematics of Data Science. 2025;7(2):495-515. doi:10.1137/24m1659534
BibTeX
@article{Carvalho2025,
author = {Carvalho, Luís and Costa, João L. and Mourão, José and Oliveira, Gonçalo},
title = {The Positivity of the Neural Tangent Kernel},
journal = {SIAM Journal on Mathematics of Data Science},
year = {2025},
volume = {7},
number = {2},
pages = {495--515},
doi = {10.1137/24m1659534}
}
EUVIMED