Principled approaches for extending neural architectures to function spaces for operator learning
- 1NVIDIA Corporation, Santa Clara, CA, USA
- 2Department of Computing and Mathematical Sciences, California Institute of Technology, Pasadena, CA, USA
Abstract
Deep learning has achieved remarkable success in computer vision and natural language processing, where tasks are commonly formulated as mappings between finite-dimensional representations. Many scientific problems, however, including those governed by partial differential equations, are naturally posed on infinite-dimensional function spaces. This mismatch has limited conventional neural networks from achieving comparable success in scientific applications. Here we identify and distil key principles for constructing practical neural architectures for mappings between function spaces. Neural operators provide a principled extension of neural networks to such settings, offering a path towards bringing deep learning’s transformative impact to science. Because deep learning’s success has relied heavily on architectural refinements, extending these advances to neural operators allows operator learning to benefit from refined designs. Guided by the principles we outline, we propose a recipe for converting popular neural architectures into neural operators with minimal modifications. We also discuss practical steps for making these models effective. This perspective offers a systematic bridge between finite-dimensional network design and operator learning for scientific applications.
Related
- Physics-informed random neural architectures on compact boundaryless multiply connected manifolds for random-operator models in machine learning and uncertainty quantification
- Quantum Variational Approaches to Plasma Equilibrium: Cost Function Design for the Grad-Shafranov Equation
- General and concise operator approach to the dyadic Green's function of layered media
- Toward Principled Generative Data Assimilation of Turbulent Flows from Sparse Observations
- Extending the Similarity Spectrum to the SWIR for quality control of above water reflectance measurements
BibTeX
@article{berner2026principled,
title = {{Principled approaches for extending neural architectures to function spaces for operator learning}},
author = {Julius Berner and Miguel Liu-Schiaffini and Jean Kossaifi and Valentin Duruisseaux and Boris Bonev and Kamyar Azizzadenesheli and Animashree Anandkumar},
journal = {Nature Machine Intelligence},
year = {2026},
month = jul,
doi = {10.1038/s42256-026-01267-z},
url = {https://doi.org/10.1038/s42256-026-01267-z},
note = {CC BY},
}
RIS
TY - JOUR
TI - Principled approaches for extending neural architectures to function spaces for operator learning
AU - Julius Berner
AU - Miguel Liu-Schiaffini
AU - Jean Kossaifi
AU - Valentin Duruisseaux
AU - Boris Bonev
AU - Kamyar Azizzadenesheli
AU - Animashree Anandkumar
PY - 2026
DA - 2026/07/03
JO - Nature Machine Intelligence
DO - 10.1038/s42256-026-01267-z
UR - https://doi.org/10.1038/s42256-026-01267-z
AB - Deep learning has achieved remarkable success in computer vision and natural language processing, where tasks are commonly formulated as mappings between finite-dimensional representations. Many scientific problems, however, including those governed by partial differential equations, are naturally posed on infinite-dimensional function spaces. This mismatch has limited conventional neural networks from achieving comparable success in scientific applications. Here we identify and distil key principles for constructing practical neural architectures for mappings between function spaces. Neural operators provide a principled extension of neural networks to such settings, offering a path towards bringing deep learning’s transformative impact to science. Because deep learning’s success has relied heavily on architectural refinements, extending these advances to neural operators allows operator learning to benefit from refined designs. Guided by the principles we outline, we propose a recipe for converting popular neural architectures into neural operators with minimal modifications. We also discuss practical steps for making these models effective. This perspective offers a systematic bridge between finite-dimensional network design and operator learning for scientific applications.
SN - 2522-5839
ER -