A Lipschitz Spaces View of Infinitely Wide Shallow Neural Networks
SIAM Journal on Mathematical Analysis · 30.06.2026 · 58 (3) · 2786-2828
DOI: 10.1137/24m1718615 ↗Kurz gefasst
Fachartikel in der Zeitschrift SIAM Journal on Mathematical Analysis, online erschienen am 01.06.2026. Verfasst von Francesca Bartolucci und 5 weiteren Autor:innen. Die Arbeit nennt 73 Literaturangaben. Gefördert von: Royal Society, Engineering and Physical Sciences Research Council, Air Force Office of Scientific Research.
Abstract (Originalsprache)
We revisit the mean field parametrization of shallow neural networks, using signed measures on unbounded parameter spaces and duality pairings that take into account the regularity and growth of activation functions. This setting directly leads to the use of unbalanced Kantorovich–Rubinstein norms defined by duality with Lipschitz functions, and of spaces of measures dual to those of continuous functions with controlled growth. These allow us to make transparent the need for total variation and moment bounds or penalization to obtain the existence of minimizers of variational formulations, under which we prove a compactness result in a strong Kantorovich–Rubinstein norm, and in the absence of which we show several examples demonstrating undesirable behavior. Further, the Kantorovich–Rubinstein setting enables us to combine the advantages of a completely linear parametrization and ensuing reproducing kernel Banach space framework with optimal transport insights. We showcase this synergy with representer theorems and uniform large data limits for empirical risk minimization, and in proposed formulations for distillation and fusion applications.
Förderung
- Royal Society
- Engineering and Physical Sciences Research Council
- Air Force Office of Scientific Research
- Università degli Studi di Genova
- Gruppo Nazionale per l'Analisi Matematica, la Probabilità e le loro Applicazioni
- Università degli Studi di Roma Tor Vergata
- National Physical Laboratory
- PRIN
Zitieren
APA
Bartolucci, F., Carioni, M., Iglesias, J. A., Korolev, Y., Naldi, E., & Vigogna, S. (2026). A Lipschitz Spaces View of Infinitely Wide Shallow Neural Networks. SIAM Journal on Mathematical Analysis, 58(3), 2786-2828. https://doi.org/10.1137/24m1718615
Vancouver
Bartolucci F, Carioni M, Iglesias JA, Korolev Y, Naldi E, Vigogna S. A Lipschitz Spaces View of Infinitely Wide Shallow Neural Networks. SIAM Journal on Mathematical Analysis. 2026;58(3):2786-2828. doi:10.1137/24m1718615
BibTeX
@article{Bartolucci2026,
author = {Bartolucci, Francesca and Carioni, Marcello and Iglesias, José A. and Korolev, Yury and Naldi, Emanuele and Vigogna, Stefano},
title = {A Lipschitz Spaces View of Infinitely Wide Shallow Neural Networks},
journal = {SIAM Journal on Mathematical Analysis},
year = {2026},
volume = {58},
number = {3},
pages = {2786--2828},
doi = {10.1137/24m1718615}
}
EUVIMED