EUVIMEDEUVIMED entdecken
Crossref

VARIATIONAL METHODS FOR NEURAL NETWORK TRAINING: APPLICATIONS OF STURM-LIOUVILLE ENERGY ESTIMATES

Massimiliano Ferrara

Far East Journal of Mathematical Sciences (FJMS) · 2025

DOI: 10.17654/0972087125015 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Far East Journal of Mathematical Sciences (FJMS), 2025. Verfasst von Massimiliano Ferrara.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper establishes a novel connection between local minimization principles for Sturm-Liouville equations and optimization techniques used in training neural networks. By interpreting the training of neural networks as a variational problem, we demonstrate how recent results on energy estimates for mixed boundary value problems in Sturm-Liouville theory can be adapted to analyze and improve neural network convergence. We present two main theorems: the first establishes conditions for guaranteed convergence to non-zero local minima in neural network training, and the second demonstrates the existence of multiple critical points with energy estimates. Our theoretical results are supported by experimental validation on benchmark datasets, showing improved performance in avoiding trivial solutions during training. This work bridges the gap between classical differential equation theory and modern machine learning optimization.

Zitieren

APA

Ferrara, M. (2025). VARIATIONAL METHODS FOR NEURAL NETWORK TRAINING: APPLICATIONS OF STURM-LIOUVILLE ENERGY ESTIMATES. Far East Journal of Mathematical Sciences (FJMS). https://doi.org/10.17654/0972087125015

Vancouver

Ferrara M. VARIATIONAL METHODS FOR NEURAL NETWORK TRAINING: APPLICATIONS OF STURM-LIOUVILLE ENERGY ESTIMATES. Far East Journal of Mathematical Sciences (FJMS). 2025. doi:10.17654/0972087125015

BibTeX
@article{Ferrara2025,
  author = {Ferrara, Massimiliano},
  title = {VARIATIONAL METHODS FOR NEURAL NETWORK TRAINING: APPLICATIONS OF STURM-LIOUVILLE ENERGY ESTIMATES},
  journal = {Far East Journal of Mathematical Sciences (FJMS)},
  year = {2025},
  doi = {10.17654/0972087125015}
}
Kostenlos starten →