Approximation of solutions of parameter-dependent problems by residual neural networks

DOI10.1515/jncds-2025-0019MaRDI QIDQ7284052arXivFDO

Authors Ana Carpio

Publication date 15 July 2026

Published in arXiv

Tldr We develop a convergent scheme to train neural networks involving analytic activation functions based on gradient flows using Lojasiewicz theory; this approach allows for simple implementation and accurate approximation of solutions of parametric problems, including inverse problems with severe ill-posedness.

Full work available at URL https://doi.org/10.1515/jncds-2025-0019

arXiv classification

cs.LGmath.APmath.NA

Cites work

• A Comparison of Three Methods for Selecting Values of Input Variables in the Analysis of Output from a Computer Code
• Approximation by neural networks with a bounded number of nodes at each level
• Approximation by superpositions of a sigmoidal function
• Bayesian approach to inverse scattering with topological priors
• Convergence of the Iterates of Descent Methods for Analytic Cost Functions
• Deep learning via dynamical systems: an approximation perspective
• Deep Network Approximation for Smooth Functions
• Linear Convergence of Gradient and Proximal-Gradient Methods Under the Polyak-\L{}ojasiewicz Condition
• Lower bounds for approximation by MLP neural networks
• Multilayer feedforward networks are universal approximators
• Numerical Optimization
• Optimal approximation rate of ReLU networks in terms of width and depth
• Texts in Computational Science and Engineering
• The Łojasiewicz exponent of an analytic function at an isolated zero
• Universal Approximation Power of Deep Residual Neural Networks Through the Lens of Control
• Universal Approximation Theorems for Differentiable Geometric Deep Learning
• Universal Approximation with Deep Narrow Networks

This page was built for publication: Approximation of solutions of parameter-dependent problems by residual neural networks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284052)