Superconvergence in Neural Network Operators: A Fractional Calculus Perspective

Fuente: Zenodo
Saved in:
Bibliographic Details
Main Authors: Santos, Rômulo Damasclin Chaves dos, Sales, Jorge
Format: Recurso digital
Language:English
Published: Zenodo 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866902093350043648
author Santos, Rômulo Damasclin Chaves dos
Sales, Jorge
author_facet Santos, Rômulo Damasclin Chaves dos
Sales, Jorge
contents <p>This paper presents a comprehensive extension of the Vonoroviskaya-Damasclin theorem, integrating fractional calculus into neural network operators to refine convergence analysis and improve approximation accuracy. By employing advanced fractional calculus techniques, we establish precise error bounds and significantly enhanced convergence rates, contributing to the theoretical development of neural network approximation in complex mathematical contexts. Our approach provides a rigorous framework for solving differential equations, particularly in fluid dynamics, where non-local and memory effects are essential. The incorporation of fractional differentiation into neural network operators demonstrates superior performance compared to traditional approximation methods, ensuring greater accuracy and adaptability. Supported by rigorous mathematical proofs, our findings introduce an adaptive framework that dynamically adjusts to local function smoothness, optimizing convergence rates and stability. Beyond its theoretical impact, this work lays the groundwork for future studies in stochastic models, high-dimensional approximations, and advanced numerical methods. The proposed methodology establishes a solid foundation for further advancements in neural network-based fractional approximations, highlighting the critical role of fractional calculus in refining modern approximation theories and computational techniques.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_15079589
institution Zenodo
language eng
publishDate 2025
publisher Zenodo
record_format zenodo
spellingShingle Superconvergence in Neural Network Operators: A Fractional Calculus Perspective
Santos, Rômulo Damasclin Chaves dos
Sales, Jorge
Fractional Calculus
Neural Network Operators
Fractional Superconvergence
Voronoviskaya Expansions
Adaptive Approximation
<p>This paper presents a comprehensive extension of the Vonoroviskaya-Damasclin theorem, integrating fractional calculus into neural network operators to refine convergence analysis and improve approximation accuracy. By employing advanced fractional calculus techniques, we establish precise error bounds and significantly enhanced convergence rates, contributing to the theoretical development of neural network approximation in complex mathematical contexts. Our approach provides a rigorous framework for solving differential equations, particularly in fluid dynamics, where non-local and memory effects are essential. The incorporation of fractional differentiation into neural network operators demonstrates superior performance compared to traditional approximation methods, ensuring greater accuracy and adaptability. Supported by rigorous mathematical proofs, our findings introduce an adaptive framework that dynamically adjusts to local function smoothness, optimizing convergence rates and stability. Beyond its theoretical impact, this work lays the groundwork for future studies in stochastic models, high-dimensional approximations, and advanced numerical methods. The proposed methodology establishes a solid foundation for further advancements in neural network-based fractional approximations, highlighting the critical role of fractional calculus in refining modern approximation theories and computational techniques.</p>
title Superconvergence in Neural Network Operators: A Fractional Calculus Perspective
topic Fractional Calculus
Neural Network Operators
Fractional Superconvergence
Voronoviskaya Expansions
Adaptive Approximation
url https://doi.org/10.5281/zenodo.15079589