Authors: Ryan Hsiang, Robert Joseph George, Anima Anandkumar
Organizations: Department of Electrical Engineering, National Taiwan University · Department of Computing and Mathematical Sciences, California Institute of Technology
Certifying the robustness and physical consistency of scientific machine-learning models ensures that their predictions remain reliable beyond the test set. For neural operators, which learn PDE solution operators by mapping infinite-dimensional function spaces, certification means establishing rigorous, domain-wide guarantees for their output functions. However, standard neural network (NN) verification methods are fundamentally designed for vector-to-vector mappings, and they can only certify neural operators at fixed grid resolutions and provide no guarantees over the entire domain or across another resolution. We close this gap by introducing a function-space certification framework that verifies neural operators across the entire continuous domain. We formulate interval bound propagation (IBP) and CROWN in function space, propagating upper and lower bound functions through each layer using rigorous numerical integration. In addition, we extend bound propagation to higher-order derivatives to certify PDE residuals and solution error bounds. Experiments across several neural operator architectures and scientific problems, including the 572.5-million-parameter spherical Fourier neural operator for global weather forecasting, show that our method produces tight, domain-wide guarantees efficiently, providing a foundation for trustworthy operator learning.
Figures & tables
Figure 1: Overview of function-space certification for neural operators. Function-space IBP and CROWN propagate the input through the neural operator. Each spectral convolution is an integral, which is enclosed with a rigorous error bound. The resulting upper and lower bound functions, and their derivatives, hold at every point of the domain and certify robustness, PDE residuals, boundary/initial conditions, and the solution error through a stability estimate.
Figure 2: Grid versus function-space certification counterexample. Given two input functions that match at grid points, running CROWN on grid points (grid CROWN) can overlook perturbations between points, while function-space CROWN soundly bounds the outputs across the entire domain.
Figure 3: Grid certificates cover only grid points, function-space certificates cover every position. Top: Width of the output bounds at each position x for input perturbations of size ε=10−3 , from function-space IBP (purple) and function-space CROWN (green). Gray shows MC sample widths. Smaller bound widths are tighter. Burgers is shown at the final time t=1 . Bottom: Maximum grid CROWN width across different resolutions N , divided by the function-space width (dashed line). Overall, function-space bounds are comparable to or tighter than grid bounds.
PDE
Scope
Method
Bound width
Empirical width
Gap
Time (s)
Memory (GiB)
1D
Poisson
Grid
IBP
5.330×10−3
9.771×10−5
54.54
0.08
0.008
CROWN
3.981×10−4
9.771×10−5
4.07
0.36
0.010
Function-space
IBP
5.364×10−3
9.580×10−5
55.98
0.55
3.844
CROWN
4.689×10−4
9.580×10−5
4.89
1833.87
18.699
Helmholtz
Grid
IBP
1.593×10−1
1.000×10−1
1.59
0.03
0.008
CROWN
1.273×10−1
1.000×10−1
1.27
0.15
0.009
Table 1: Robustness certification for grid and function-space IBP/CROWN on an NVIDIA B300 GPU, using Gaussian quadrature with Q=1,048,576 . Function-space IBP uses linearized perturbation bounds at 2048×2048 for 2D Poisson and 4096×4096 for space–time Burgers.
(a) PDE residual bounds
PDE
Quantity
Upper bound
Bound width
Time (s)
Memory (GiB)
1D
Poisson
∥−∂xxuθ−a∥∞
1.105×10−1
1.061×10−5
29.11
2.013
∥−∂xxuθ−a∥2
5.515×10−2
1.451×10−5
Helmholtz
∥−∂xxuθ−κ2uθ−a∥∞
2.833×10−1
5.875×10−4
37.67
2.013
∥−∂xxuθ−κ2uθ−a∥2
1.177×10−1
6.864×10−4
2D
Poisson
∥−∂xxuθ−∂yyuθ−a∥∞
2.935
1.124×10−3
533.43
3.223
Table 2: Residual and solution error certification. (a) Certified residual bounds and their widths. (b) Certified solution error bounds from PDE stability estimates. In (b), bound width measures the uncertainty in the stability expression, the true solution error may be smaller than the lower bound.
Figure 4: Residual bounds across 25 autoregressive rollout steps of the Kuramoto–Sivashinsky equation. We evaluate a neural operator trained on the KS equation with step size τ=0.1 for 25 time steps. Shaded bands indicate certified intervals for spatial L∞ (left) and L2 (right) residuals. The predicted field develops sharp spatial features around steps 9–11, causing the residual to jump from ∼101 to ∼105 . This massive residual indicates that the model has failed to maintain physical consistency under long-horizon rollout, a failure mode our certified bounds expose.
Figure 5: Temperature heat map of California Circles show the SFNO’s temperature predictions at ERA5’s grid resolution of 0.25∘ . Our framework extends beyond the grid, providing maximum temperature bounds over the entire region of California.
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
Benchmark
Method
Max width ↓
Runtime
Device
1D (1 layer)
Rectangle
6.27×10−6
4.92 s
GPU
Gaussian
8.56×10−76
9.58 s
GPU
Arb
2.46×10−71
9.16 s
CPU
1D (2 layers)
Rectangle
3.29×10−2
49.23 s
GPU
Gaussian
1.72×10−187
53.34 s
GPU
Arb
2.39×10−14
53.90 s
CPU
Appendix
Table 3: Comparison of certified integration backends across 64 query points ( ε=0 ).
Figure 6: Illustration of bilinear and linear extension.
PDE
Bilinear
Linearized
Certified width
Gap
Time (s)
Memory (GiB)
2D Poisson
Off
Off
1.275×10−1
2068.70
1.31
10.54
On
Off
4.009×10−2
650.51
0.77
15.60
Off
On
8.725×10−2
1415.56
2.97
13.44
On
On
4.645×10−4
7.54
2.93
18.51
Burgers
Off
Off
5.008
987.23
9.81
78.76
On
Off
4.291
846.03
9.39
116.26
Appendix
Table 4: Ablation of bilinear cell bounds and linearized perturbation propagation.
Neural network methods for PDEs require reliable error control in function space norms. However, trained neural networks can typically only be probed at a finite number of point values. Without strong assumptions, point evaluations alone do not provide enough information to derive tight deterministic and guaranteed bounds on function space norms. In this work, we move beyond a purely black-box setting and exploit the neural network structure directly. We present a framework for the certified and accurate computation of integral quantities of neural networks, including Lebesgue and Sobolev norms, by combining interval arithmetic enclosures on axis-aligned boxes with adaptive marking/refinement and quadrature-based aggregation. On each box, we compute guaranteed lower and upper bounds for function values and derivatives, and propagate these local certificates to global lower and upper bounds for the target integrals. Our analysis provides a general convergence theorem for such certified adaptive quadrature procedures and instantiates it for function values, Jacobians, and Hessians, yielding certified computation of Lp, W1,p, and W2,p norms. We further show how these ingredients lead to practical certified bounds for PINN interior residuals. Numerical experiments illustrate the accuracy and practical behavior of the proposed methods.
Johannes Gründler, Moritz Maibaum, Philipp Petersen
Faculty of Mathematics, University of Vienna, Kolingasse 14–16, 1090 Vienna, Austria
Neural operators provide fast surrogate models for approximating operators between function spaces, but their predictions often lack uncertainty quantification. We develop a split conformal framework to guarantee that a calibrated pointwise band around the neural operator output contains the true solution on at least a 1−γ fraction of the evaluation domain, with probability at least 1−α over test and calibration inputs, where α,γ∈(0,1). Our method reduces a normalized residual field to its spatial (1−γ)-quantile and computes a scaling factor using a held-out calibration dataset. We prove marginal coverage guarantees for measurable residual fields defined on arbitrary probability spaces, covering both continuum domains and fixed discretizations. Under mild assumptions on the data distribution, we show that the coverage conditional on the calibration set follows a Beta distribution, which we verify with numerical experiments on Darcy flow and Navier--Stokes equations, where our calibration yields bands consistently tighter than existing corrections while retaining the target coverage.
Tom Stent, Nicolas Boullé
Department of Mathematics Imperial College London London, SW7 2AZ, UK
We develop a rigorous framework for extending neural operators to handle out-of-distribution input functions. We leverage kernel approximation techniques and provide theory for characterizing the input-output function spaces in terms of Reproducing Kernel Hilbert Spaces (RKHSs). We provide theorems on the requirements for reliable extensions and their predicted approximation accuracy. We also establish formal relationships between specific kernel choices and their corresponding Sobolev Native Spaces. This connection further allows the extended neural operators to reliably capture not only function values but also their derivatives. Our methods are empirically validated through the solution of elliptic partial differential equations (PDEs) involving operators on manifolds having point-cloud representations and handling geometric contributions. We report results on key factors impacting the accuracy and computational performance of the extension approaches.
Blaine Quackenbush, Paul J. Atzberger
Department of Mathematics, University of California Santa Barbara (UCSB). · Department of Mathematics, Department of Mechanical Engineering, University of California Santa Barbara (UCSB)