Linearization Turns Neural Operators into Function-Valued Gaussian Processes
Emilia Magnani, Marvin Pförtner, Tobias Weber, Philipp Hennig
Abstract
Neural operators generalize neural networks to learn mappings between function spaces from data. They are commonly used to learn solution operators of parametric partial differential equations (PDEs) or propagators of time-dependent PDEs. However, to make them useful in highstakes simulation scenarios, their inherent predictive error must be quantified reliably. We introduce LUNO, a novel framework for approximate Bayesian uncertainty quantification in trained neural operators. Our approach leverages model linearization to push (Gaussian) weight-space uncertainty forward to the neural operator's predictions. We show that this can be interpreted as a probabilistic version of the concept of currying from functional programming, yielding a function-valued (Gaussian) random process belief. Our framework provides a practical yet theoretically sound way to apply existing Bayesian deep learning methods such as the linearized Laplace approximation to neural operators. Just as the underlying neural operator, our approach is resolution-agnostic by design. The method adds minimal prediction overhead, can be applied posthoc without retraining the network, and scales to large models and datasets. We evaluate these aspects in a case study on Fourier neural operators.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
2 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Multiwavelet-based Operator Learning for Differential Equations cites Multipole Graph Neural Operator for Parametric Partial Differential Equations
- Multiwavelet-based Operator Learning for Differential Equations cites Fourier Neural Operator for Parametric Partial Differential Equations
- Generative Neural Operators through Diffusion Last Layer cites Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks
- Generative Neural Operators through Diffusion Last Layer cites Fourier Neural Operator for Parametric Partial Differential Equations
- Laplace Redux - Effortless Bayesian Deep Learning cites Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks
- Fourier Neural Operator for Parametric Partial Differential Equations cites Multipole Graph Neural Operator for Parametric Partial Differential Equations
Cited by top-tier papers2
- Generative Neural Operators through Diffusion Last LayerSungwon Park, Anthony Zhou, Hongjoong Kim, Amir Barati FarimaniICML 2026 · 1 citation
- Infinite Neural Operators: Gaussian processes on functionsDaniel Augusto de Souza, Yuchen Zhu, Jake Cunningham, Yuri F. Saporito et al.NeurIPS 2025 · 1 citation
Builds on11
- Fourier Neural Operator for Parametric Partial Differential EquationsZongyi Li, Nikola Borislavov Kovachki, Kamyar Azizzadenesheli, Burigede Liu et al.ICLR 2021 · 3,911 citations
- Multipole Graph Neural Operator for Parametric Partial Differential EquationsZongyi Li, Nikola B. Kovachki, Kamyar Azizzadenesheli, Burigede Liu et al.NeurIPS 2020 · 569 citations
- Laplace Redux - Effortless Bayesian Deep LearningErik A. Daxberger, Agustinus Kristiadi, Alexander Immer, Runa Eschenhagen et al.NeurIPS 2021 · 508 citations
- Multiwavelet-based Operator Learning for Differential EquationsGaurav Gupta, Xiongye Xiao, Paul BogdanNeurIPS 2021 · 355 citations
- Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU NetworksAgustinus Kristiadi, Matthias Hein, Philipp HennigICML 2020 · 344 citations
Related papers
- Generic bounds on the approximation error for physics-informed (and) operator learningTim De Ryck, Siddhartha MishraNeurIPS 2022 · 93 citations
- Activation-level uncertainty in deep neural networksPablo Morales-Alvarez, Daniel Hernández-Lobato, Rafael Molina, José Miguel Hernández-LobatoICLR 2021 · 16 citations
- Derivative-enhanced Deep Operator NetworkYuan Qiu, Nolan Bridges, Peng ChenNeurIPS 2024 · 25 citations
- Light-Weight Diffusion Multiplier and Uncertainty Quantification for Fourier Neural OperatorsAlbert Matveev, Sanmitra Ghosh, Aamal Hussain, James-Michael Leahy et al.NeurIPS 2025 · 8 citations
- Thompson Sampling in Function Spaces via Neural OperatorsRafael Oliveira, Xuesong Wang, Kian Ming A. Chai, Edwin V. BonillaNeurIPS 2025 · 2 citations
