The Frame Kernel Method for Multiscale Operator Learning

DOI10.48550/arXiv.2608.25084MaRDI QIDQ7284982arXivFDO

Authors M. Keith Ballard, Robert M. Kirby, Varun Shankar, Branden Frieden, Ryan Whitehead

Publication date 25 August 2026

Published in arXiv

Tldr We present a natively multiscale operator learning method for the surrogate modeling of multiscale partial differential equations (PDEs) using a novel multiscale kernel frame function approximation technique. The method generalizes to multiscale outputs and is demonstrated on challenging problems with improved accuracy over neural operators, while also admitting an a posteriori multiscale decomposition upon generalization.

Full work available at URL https://doi.org/10.48550/arXiv.2608.25084

arXiv classification

cs.LGmath.NA

Cites work

• A Class of Nonharmonic Fourier Series
• A comprehensive and fair comparison of two neural operators (with practical extensions) based on FAIR data
• A multiscale finite element method for elliptic problems in composite materials and porous media
• A survey of projection-based model reduction methods for parametric dynamical systems
• An introduction to frames and Riesz bases
• An Optimal Weighted Least-Squares Method for Operator Learning
• Basis-to-basis operator learning using function encoders
• Bayesian calibration of computer models. (With discussion)
• Clustering to minimize the maximum intercluster distance
• Design and analysis of computer experiments
• Doubling the rate -- improved error bounds for orthogonal projection with application to numerical analysis
• Error estimates for interpolation by compactly supported radial basis functions of minimal degree
• Fourier Neural Operator for Parametric Partial Differential Equations
• Generalized multiscale finite element methods (GMsFEM)
• Improved error bounds for scattered data interpolation by radial basis functions
• Kernel methods are competitive for operator learning
• Kernel techniques: From machine learning to meshless methods
• Kernel-based approximation methods using MATLAB
• Locally Subspace-Informed Neural Operators for Efficient Multiscale PDE Solving
• Machine learning for fluid mechanics
• Matrix Computations
• MscaleFNO: multi-scale Fourier neural operator learning for oscillatory functions and wave scattering problems
• Multidimensional binary search trees used for associative searching
• Multiscale analysis in Sobolev spaces on bounded domains
• Multiscale approximation and reproducing kernel Hilbert space methods
• Multiscale kernels
• Multiscale scattered data analysis in samplet coordinates
• Multiwavelet-based Operator Learning for Differential Equations
• Observations on the behavior of radial basis function approximations near boundaries
• Operator-valued kernels for learning from functional response data
• Piecewise polynomial, positive definite and compactly supported radial functions of minimal degree
• Principles of multiscale modeling
• Scattered Data Approximation
• Sobolev bounds on functions with scattered zeros, with applications to radial basis function surface fitting
• Sobolev Spaces
• Spectral Methods in MATLAB
• Survey of multifidelity methods in uncertainty propagation, inference, and optimization
• Ten Lectures on Wavelets
• The heterogeneous multiscale methods
• The Mathematical Theory of Finite Element Methods
• Tight frame expansions of multiscale reproducing kernels in Sobolev spaces
• Wavelet neural operator for solving parametric partial differential equations in computational mechanics problems
• Zooming from global to local: a multiscale RBF approach

This page was built for publication: The Frame Kernel Method for Multiscale Operator Learning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284982)