The paradox of over-parameterization in learning algorithm
Annals of Mathematics and Computer Science · 24.05.2026 · 34
DOI: 10.56947/amcs.v34.824 ↗Kurz gefasst
Fachartikel in der Zeitschrift Annals of Mathematics and Computer Science, online erschienen am 24.05.2026. Verfasst von Fabrizio Morlando. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Traditional learning theory fails to explain why over-parameterized networks generalize. We propose a physical synthesis treating learning as a dissipative process. By integrating Differential Geometry, Dynamical Systems, and Algorithmic Stability, we show generalization depends on dynamical stability rather than architecture. Utilizing the Polyak-Lojasiewicz condition and Restricted Strong Convexity, we establish that gradient flow acts as a contractive operator, preventing trajectory divergence. Finally, McDiarmid’s inequalities convert this contraction into rigorous probabilistic bounds. Our framework proves that over-parameterization facilitates geometric properties that inherently regularize the learning process, ensuring robust generalization despite high model capacity.
Zitieren
APA
Morlando, F. (2026). The paradox of over-parameterization in learning algorithm. Annals of Mathematics and Computer Science, 34. https://doi.org/10.56947/amcs.v34.824
Vancouver
Morlando F. The paradox of over-parameterization in learning algorithm. Annals of Mathematics and Computer Science. 2026;34. doi:10.56947/amcs.v34.824
BibTeX
@article{Morlando2026,
author = {Morlando, Fabrizio},
title = {The paradox of over-parameterization in learning algorithm},
journal = {Annals of Mathematics and Computer Science},
year = {2026},
volume = {34},
doi = {10.56947/amcs.v34.824}
}
EUVIMED