Pith. sign in

REVIEW 4 major objections 4 minor 68 references

Kernel Methods for Learning Operators with Multiple Inputs and Outputs

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Multi-task operator error is set by the hardest task alone.

desk verdict Solid kernel-based multi-operator framework with an honest but overstated abstract: the reconstruction rates are task-wise, but the learning term depends on the encoded dimension and an unquantified fill distance, so the dimension-free claim covers only part of the bound. read the letter →

arxiv 2608.11831 v1 pith:YHRKOBKM submitted 2026-08-12 cs.LG math.STstat.MLstat.TH

Adrien Weihs , Chunyang Liao , Jingmin Sun , Hayden Schaeffer This is my paper · ORCID
classification cs.LGmath.STstat.MLstat.TH MSC 46E2265D1541A05
keywords multi-inputmulti-outputoperatorlearningkernelmethodsencoder-decoderframeworkoperator-valuedkernelsapproximationtheorymultipleMOparametricpartialdifferentialequations
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that learning a map between products of function spaces—multiple inputs, multiple outputs—does not have to pay a curse of dimensionality in the number of tasks. It proposes an encoder–decoder architecture in which inputs and outputs are first measured, a kernel surrogate is learned between the encoded spaces, and predictions are reconstructed by minimum-norm recovery. The main theorem (Theorem 3.15) decomposes the total error into reconstruction errors from encoding, a kernel learning error, and a data-consistency residual, and shows the reconstruction rates depend only on the most demanding individual component, not on the combined number of inputs and outputs. If correct, this makes kernel methods a mathematically grounded and computationally cheap alternative to neural operators in moderate-data scientific computing, supported by closed-form training and inference.

What carries the argument

The load-bearing mechanism is the encoder–decoder error decomposition (Theorem 3.7), paired with minimum-norm recovery as the canonical decoder. The encoder maps each function to point evaluations; the decoder (Theorem 2.1) reconstructs the minimum-norm element consistent with those measurements. A kernel surrogate is learned between the encoded measurement spaces, and Theorem 3.15 combines Sobolev sampling inequalities for the reconstruction terms with a Sobolev-embedding estimate for the learning term, isolating the contribution of each architectural choice. This separation is what allows the reconstruction rates to depend only on the worst componentwise fill distance while the learning rate depends on the encoded input dimension.

What would settle it

Train any kernel model from the paper's setting with parameters drawn from a distribution that maps into a low-dimensional or clustered subset of the latent space, leaving a fixed region in $\Upsilon$ unsampled as $N$ grows; if test inputs whose encodings fall in that region show persistent error even at large $N$, the $h_{\mathrm{tr}}$-controlled learning term has not vanished, contradicting the implied dependence. A second check is to add one very low-regularity input component to an otherwise smooth multi-task family and verify whether the measured rate matches that component's reconstruction exponent exactly.

Watch

Extended reading notes

Core claim

The central claim is Theorem 3.15: for operators $G:X_1\times\cdots\times X_m\to Y_1\times\cdots\times Y_n$, with pointwise encoders and minimum-norm decoders in Sobolev-embedded Hilbert spaces, the approximation error at $x$ is bounded by $\omega\left(C R \max_j h_{X,j}^{\alpha_{X,j}}\right) + C M_R \max_k h_{Y,k}^{\alpha_{Y,k}} + \|D_Y\| C \left(h_{\mathrm{tr}}^{\tau-d_X/2}+\sqrt{\lambda}\right) \|G_{\mathrm{enc}}\|_{H_\Gamma}$ plus a data-consistency term. The first two terms are input and output reconstruction errors; each maximum runs over the component spaces, so convergence in the reconstruction part is controlled by the hardest single approximation task rather than by the total number of tasks. The third term is the kernel learning error, which depends on the dimension $d_X$ of the encoded input space through the fill distance $h_{\mathrm{tr}}$ of encoded training points in the latent domain $\Upsilon$. The paper also claims (Theorem 3.3 and Corollary 3.4) that any operator-valued kernel on the latent space induces a kernel on the original spaces, making the encoded-space kernel problem equivalent to a kernel method acting directly on the original function spaces. On five parametric PDE families, the resulting KernelMO variants attain the lowest errors in most comparisons and reduce training time by two or more orders of magnitude relative to neural operator baselines.

Load-bearing premise

The quantitative guarantee depends on the encoded training points having a small fill distance in the latent space, and the paper does not show how many training operators are needed to ensure that; if the encoded inputs cluster instead of spreading out, the learning-error term may not decay at all.

Editorial extensions

If this is right

  • Adding input or output components to a problem does not compound the reconstruction convergence rate; only the least regular (or least sampled) component matters.
  • Learning in the latent encoded space is equivalent to kernel learning on the original function spaces, so existing RKHS theory transfers to operator learning.
  • KernelMO-OV and KernelMO-PS provide complementary trade-offs: operator-valued training scales with the number of operators, product-space training scales with parameter–input pairs, and both retain the same approximation guarantees.
  • Closed-form kernel training and inference can reduce training time by orders of magnitude compared with neural operator architectures while matching or beating their predictive error on the tested parametric PDEs.
  • The mesh-transferability of the encoder–decoder setup means a trained surrogate can be evaluated on discretizations different from the training grid, without retraining.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The theorem's prediction that the hardest task governs reconstruction could be tested directly: construct multi-task families where one component has much lower Sobolev regularity than the rest and check whether the total error tracks that component's rate while easy components are added.
  • The unquantified fill distance $h_{\mathrm{tr}}$ suggests a design rule the paper does not state: deliberately choose training inputs so their encoded representations are quasi-uniform in the latent space; such experimental design could convert the bound into an explicit $N$-dependence and a sensor-versus-operator sample trade-off.
  • The framework points toward kernel-based 'moderate-data foundation models' for PDE families, but exploiting that possibility requires the missing quantitative link between the number of training operators $N$ and $h_{\mathrm{tr}}$; this is the main gap to close before kernel methods can claim general scalability.
  • One can also test whether the operator-valued formulation remains preferable when each operator is queried repeatedly, since its kernel matrix scales with the number of operators rather than with the number of input–output pairs; this is an engineering consequence the paper describes but does not optimize over.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper introduces a kernel-based encoder-decoder framework for learning operators between products of function spaces. It proposes a four-stage decomposition (observation, representation, learning, reconstruction), proves an abstract error decomposition in Theorem 3.7, and derives quantitative Sobolev-space bounds in Theorem 3.15 that separate reconstruction errors, a kernel learning error, and a data-consistency term. The framework is specialized to multiple operator learning, yielding two kernel methods, KernelMO-OV and KernelMO-PS, and is evaluated on five parametric PDE benchmarks where the methods are reported to match or outperform neural operator baselines with substantially lower training cost.

Significance. If Theorem 3.15 were a complete quantitative guarantee, the paper would make a useful contribution: it connects encoder-decoder operator learning to standard RKHS interpolation and sampling inequalities, gives explicit closed-form estimators, and provides reproducible code and a broad experimental comparison. The decomposition in Theorem 3.7 is conceptually clean and the Sobolev reconstruction estimates are a natural extension of existing scalar and vector-valued kernel results. However, the advertised headline claim that convergence is governed by the hardest constituent task rather than the overall dimension is only partially supported: the two reconstruction terms in Theorem 3.15 do have this form, but the learning term depends explicitly on the total encoded input dimension d_X and on an unquantified fill distance h_tr. In addition, the PCA-based variants that produce several of the strongest experimental results are not covered by the assumptions of Theorem 3.15. The central framework appears internally consistent, but the paper's abstract and contribution statements overstate what the theorems actually deliver.

major comments (4)
  1. [Abstract; Theorem 3.15; Remark 3.17] The abstract claims that the convergence rate is governed by the most challenging constituent approximation problem rather than the overall problem dimension. In Theorem 3.15, however, the learning-error term is ||D_Y|| C (h_tr^{tau-d_X/2} + sqrt(lambda)) ||G_enc||_{H_Gamma}, with d_X = sum_j m_{X,j}, the total number of pointwise input measurements. This term depends explicitly on d_X, and the bound is meaningful only when tau > d_X/2. Thus the dimension-free statement holds at best for the two reconstruction terms; Remark 3.17 correctly concedes that the learning rate may deteriorate with the number of tasks. The abstract and Section 1.1 should be revised to state that the reconstruction part is governed by the hardest task, while the learning part depends on the encoded dimension.
  2. [Assumption L.3; Remark 3.16; Theorem 3.15] Theorem 3.15 is presented as a convergence guarantee, but it is not a finite-sample rate in the number N of training operators. The bound depends on h_tr, the fill distance of the encoded training points U_i = E_X x_i in the encoded domain Upsilon, and Assumption L.3 imposes no relationship between h_tr and N. Remark 3.16 explicitly states that this fill distance is not quantified as a function of N. Consequently, the theorem does not establish decay of the learning error as N increases under the paper's sampling scheme. The paper should either provide a quantitative bound on h_tr (e.g., under a quasi-uniform or space-filling design assumption on the encoded points) or present Theorem 3.15 as a conditional bound depending on h_tr rather than as a convergence rate.
  3. [Theorem 3.7(2); Theorem 3.15] The final bound in Theorem 3.15 contains the data-consistency term ||D_Y|| ||R_{U,lambda} eta(E_X x)||_2, which is not quantified. Since the training outputs are S_i = E_Y G(x_i), the residuals eta_i = S_i - G_enc(U_i) are generally nonzero unless D_X E_X x_i = x_i, a condition that is not imposed. The paper notes in Remark 3.8 that the term vanishes under exact input reconstruction, but for the general setting of Theorem 3.15 no bound on this term is supplied. The reconstruction guarantee is therefore incomplete: it is an inequality with an uncontrolled remainder. The authors should bound this term using Assumption O.2 and Lemma 3.11, or explicitly state it as an additional assumption.
  4. [Section 4; Remark 4.3; Assumption M.2] The strongest experimental results include the PCA-based variants KernelMO-OV (PCA) and KernelMO-PS (PCA). In these methods the encoders are PCA projections of the sampled observations and the decoders are inverse PCA maps. Assumption M.2, which underpins Theorem 3.15, defines encoders as pointwise evaluation maps with measurement spaces Ran(E_j) in R^{m_j}. PCA encoders are not point-evaluation maps, so the quantitative guarantees of Theorem 3.15 do not apply to these configurations. The paper should either extend the theory to general linear encoders satisfying a suitable assumption, or clearly state that the PCA experiments are heuristic extensions outside the proven regime.
minor comments (4)
  1. [Assumption S.1(b)] There is a duplicated definite article: 'equipped with the the product norm' should read 'equipped with the product norm'.
  2. [Remark 3.2] Remark 3.2 refers to 'Assumption S.1' where the assumption being discussed is the membership G_enc in H_Gamma from Assumption L.3; the reference should be corrected.
  3. [Section 4.1.4] In the operator-valued learning discussion for the parametric diffusion-reaction equation, the text reads 'KernelMO-OV / M and KernelMO-OV / M×M', but M×M is a product-space kernel notation; this appears to be a typo for KernelMO-PS / M×M.
  4. [Table 2] The table header 'N test' is inconsistent with the rest of the text; consider using 'Number of test samples' for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the main error bound is assembled from external sampling inequalities, minimum-norm recovery, and explicit regularity assumptions, with no fitted prediction or load-bearing self-citation.

full rationale

The paper's central quantitative result, Theorem 3.15, is derived by chaining independent ingredients: the minimum-norm recovery formula (Theorem 2.1), componentwise Sobolev sampling inequalities (Lemma 3.11, citing [3,4]), vector-valued RKHS sampling estimates (Lemma 3.13, citing [19]), and a standard kernel interpolation/residual estimate (cited [42]). The regulatory assumptions, including G_enc in H_Gamma and H_Gamma embedded in H^tau, are stated as explicit hypotheses rather than as consequences of the desired error bound; Remark 3.2 explains how they could be verified from regularity of G. No experimental constant enters the bounds, and no parameter is fitted to the empirical results and then reported as a prediction. The self-citations [51-53] are contextual or comparative: Remark 3.17 cites [52] only as an analogy for the reconstruction-term scaling, not as the source of the kernel-learning estimate. The paper explicitly acknowledges that the learning-error term depends on the encoded dimension d_X (Remark 3.17) and that the fill distance h_tr is not quantified as a function of N (Remark 3.16). Those admissions weaken the advertised dimension-free finite-sample claim, but they are correctness/completeness limitations, not circular reasoning. The PCA-based experimental configurations fall outside the point-evaluation assumptions of Theorem 3.15, but that is a gap in coverage, not a case of the result being defined in terms of its own inputs. Overall, no specific reduction of the claimed result to its own assumptions or to a self-citation chain is present.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The theory is built from standard functional analysis, RKHS theory, and external sampling inequalities; no fitted constants enter the convergence bounds. The main burden is carried by stated regularity and sampling assumptions (G_enc in the RKHS, controlled fill distances, Sobolev embeddings), not by invented entities or fitted parameters.

free parameters (4)
  • RBF/Matérn kernel length scale γ = grid-searched over {0.01,0.1,1,10,100} per benchmark (Tables 17-26)
    Empirical performance and OOD behavior depend on this choice; theory is independent of the fitted value.
  • Matérn smoothness ν = grid-searched over {1/2,3/2,5/2,7/2}
    Chosen per benchmark on validation data; affects smoothness of the kernel surrogate.
  • PCA component count = 10 components per encoded function family
    PCA variants in the experiments fix the latent dimension by hand; this is not derived from the theory.
  • Ridge/interpolation regularization λ = 0 in all reported experiments (minimum-norm interpolation)
    The theory allows λ>0 but the empirical KernelMO models use λ=0, which requires invertibility of the kernel matrix.
assumptions (6)
  • standard math Sobolev embedding theorem and sampling inequalities from [3, Theorem 4.1] and [19, Theorem 17]
    Used in Lemma 3.11 and Lemma 3.13 to convert pointwise samples into Sobolev-norm reconstruction and learning error bounds.
  • domain assumption Assumption O.2: G maps B_R(H_X) into H_Y with sup_{x in B_R(H_X)} ||G(x)||_{H_Y} ≤ M_R and a modulus of continuity ω
    Needed to control the input reconstruction term in Theorem 3.15; it is a regularity assumption on the target operator, not a structural part of the framework.
  • domain assumption Assumption L.3: G_enc belongs to H_Γ and H_Γ embeds continuously into H^τ(Υ;R^{d_Y}) for τ>d_X/2
    This directly gives the learning-rate exponent in Lemma 3.13 and Theorem 3.15; if the encoded operator is not in the RKHS, the bound is not available.
  • domain assumption Assumption M.2: encoders are point-evaluation maps and decoders are minimum-norm recovery maps on Sobolev spaces
    This is the concrete setting for Lemma 3.11 and Theorem 3.15; PCA-based encoders used in experiments do not satisfy the point-evaluation premise.
  • domain assumption Assumption S.1: product Sobolev/Hilbert spaces with the stated embedding and regularity conditions, plus finite sampling grids with fill distance h_j
    Defines the function spaces and sampling regimes under which the reconstruction rates are proven.
  • domain assumption Invertibility of the kernel Gram matrix Γ(U,U) when λ=0
    Corollary 3.4 and the experiments rely on existence of the minimum-norm interpolant; this is assumed and not established for the specific parameter-function kernels used in practice.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Kernel Methods for Learning Operators with Multiple Inputs and Outputs." pith.science (2026). https://pith.science/paper/YHRKOBKM

@misc{pith2026260811831,
  author       = {Pith},
  title        = {Pith review of: Kernel Methods for Learning Operators with Multiple Inputs and Outputs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YHRKOBKM}},
  note         = {Machine review of arXiv:2608.11831}
}
read the original abstract

Learning mappings between infinite-dimensional objects is a central challenge in scientific machine learning. We introduce a general kernel-based encoder-decoder framework for operator learning that separates observation, representation, learning, and reconstruction. We develop this framework for multi-input, multi-output operator learning, where operators map between products of potentially distinct function spaces. Our approximation theory shows that, although the number of inputs and outputs can increase, the convergence rate is governed by the most challenging constituent approximation problem rather than the overall problem dimension. The framework leads to practical kernel methods with closed-form training and inference, combining mathematical tractability with computational efficiency. We further specialize the approach to multiple operator learning by introducing KernelMO, a family of kernel methods with complementary operator-valued and product-space formulations. Across five families of parametric partial differential equations, the proposed methods achieve competitive or state-of-the-art predictive accuracy while reducing training and inference costs relative to neural operator architectures and deep learning based models, offering an efficient and lightweight alternative.

Figures

Figures reproduced from arXiv: 2608.11831 by the authors.

Figure 1
Figure 1. Conceptual illustration of the trade-offs between theoretical guarantees, empirical performance, and scalability [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Schematic description of the proposed learning framework. (a) Encoder–decoder formulation for learning a map [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Measurement transferability of the encoder-decoder learning framework. The maps [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (18 more)
Figure 4
Figure 4. Figure 4: Encoder–decoder formulations for classical operator learning and multiple operator learning. (a) Classical [PITH_FULL_IMAGE:figures/full_fig_p018_4.png]
Figure 5
Figure 5. Figure 5: Complete product-space formulation of multiple operator learning. Instead of learning the operator-valued map [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]
Figure 6
Figure 6. Figure 6: Summary of normalized performance in the operator-valued-learning experiments. For each method [PITH_FULL_IMAGE:figures/full_fig_p024_6.png]
Figure 7
Figure 7. Figure 7: Summary of normalized performance in the product-space-learning experiments. For each method [PITH_FULL_IMAGE:figures/full_fig_p025_7.png]
Figure 8
Figure 8. Figure 8: Qualitative summary across PDEs for operator-valued learning. Columns correspond to conservation law, [PITH_FULL_IMAGE:figures/full_fig_p025_8.png]
Figure 9
Figure 9. Figure 9: Qualitative summary across PDEs for product-space learning. Columns correspond to conservation law, [PITH_FULL_IMAGE:figures/full_fig_p026_9.png]
Figure 10
Figure 10. Figure 10: Operator-valued learning: qualitative prediction and uncertainty comparison for the conservation law on the [PITH_FULL_IMAGE:figures/full_fig_p028_10.png]
Figure 11
Figure 11. Figure 11: Product-space learning: qualitative prediction and uncertainty comparison for the conservation law on the [PITH_FULL_IMAGE:figures/full_fig_p029_11.png]
Figure 12
Figure 12. Figure 12: Operator-valued learning: qualitative prediction and uncertainty comparison for the diffusion-reaction [PITH_FULL_IMAGE:figures/full_fig_p030_12.png]
Figure 13
Figure 13. Figure 13: Product-space learning: qualitative prediction and uncertainty comparison for the diffusion-reaction-advection [PITH_FULL_IMAGE:figures/full_fig_p031_13.png]
Figure 14
Figure 14. Figure 14: Operator-valued learning: qualitative prediction and uncertainty comparison for the Nonlinear Klein-Gordon [PITH_FULL_IMAGE:figures/full_fig_p032_14.png]
Figure 15
Figure 15. Figure 15: Product-space learning: qualitative prediction and uncertainty comparison for the nonlinear Klein-Gordon [PITH_FULL_IMAGE:figures/full_fig_p033_15.png]
Figure 16
Figure 16. Figure 16: Operator-valued learning: qualitative prediction and uncertainty comparison for the parametric diffusion [PITH_FULL_IMAGE:figures/full_fig_p034_16.png]
Figure 17
Figure 17. Figure 17: Product-space learning: qualitative prediction and uncertainty comparison for the parametric diffusion [PITH_FULL_IMAGE:figures/full_fig_p035_17.png]
Figure 18
Figure 18. Figure 18: Operator-valued learning: qualitative prediction and uncertainty comparison for the parametric wave equation [PITH_FULL_IMAGE:figures/full_fig_p037_18.png]
Figure 19
Figure 19. Figure 19: Product-space learning: qualitative prediction and uncertainty comparison for the parametric wave equation on [PITH_FULL_IMAGE:figures/full_fig_p038_19.png]
Figure 20
Figure 20. Figure 20: Operator-valued learning: trade-off between predictive accuracy and computational cost on the conservation [PITH_FULL_IMAGE:figures/full_fig_p039_20.png]
Figure 21
Figure 21. Figure 21: Product-space learning: trade-off between predictive accuracy and computational cost on the conservation-law [PITH_FULL_IMAGE:figures/full_fig_p040_21.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 39 canonical work pages

  1. [1]

    Adams and J.J.F

    R.A. Adams and J.J.F. Fournier.Sobolev Spaces. Pure and Applied Mathematics. Academic Press, 2003

    work page 2003

  2. [2]

    Álvarez, Lorenzo Rosasco, and Neil D

    Mauricio A. Álvarez, Lorenzo Rosasco, and Neil D. Lawrence. Kernels for vector-valued functions: A review.Foundations and Trends in Machine Learning, 4(3):195–266, March 2012

    work page 2012

  3. [3]

    An extension of a bound for functions in sobolev spaces, with applications to (m, s)-spline interpolation and smoothing.Numerische Mathematik, 107(2):181–211, 2007

    Rémi Arcangéli, María Cruz López de Silanes, and Juan José Torrens. An extension of a bound for functions in sobolev spaces, with applications to (m, s)-spline interpolation and smoothing.Numerische Mathematik, 107(2):181–211, 2007

    work page 2007

  4. [4]

    Extension of sampling inequalities to sobolev semi-norms of fractional order and derivative data.Numerische Mathematik, 121(3):587–608, 2012

    Rémi Arcangéli, María Cruz López de Silanes, and Juan José Torrens. Extension of sampling inequalities to sobolev semi-norms of fractional order and derivative data.Numerische Mathematik, 121(3):587–608, 2012

    work page 2012

  5. [5]

    Theory of reproducing kernels.Transactions of the American Mathematical Society, 68:337–404, 1950

    Nachman Aronszajn. Theory of reproducing kernels.Transactions of the American Mathematical Society, 68:337–404, 1950

    work page 1950

  6. [6]

    Vladyslav Babenko, Vira Babenko, and Oleg Kovalenko. Korneichuk-stechkin lemma, ostrowski and landau inequalities, and optimal recovery problems for l-space valued functions.Numerical Functional Analysis and Optimization, 44(12):1309–1341, 2023

    work page 2023

  7. [7]

    Sorokin, Xianjin Yang, Théo Bourdais, Edoardo Calvello, Matthieu Darcy, Alexander Hsu, Bamdad Hosseini, and Houman Owhadi

    Aras Bacho, Aleksei G. Sorokin, Xianjin Yang, Théo Bourdais, Edoardo Calvello, Matthieu Darcy, Alexander Hsu, Bamdad Hosseini, and Houman Owhadi. Operator learning at machine precision, 2025

    2025

  8. [8]

    Pau Batlle, Yifan Chen, Bamdad Hosseini, Houman Owhadi, and Andrew M. Stuart. Error analysis of kernel/GP methods for nonlinear and parametric pdes.Journal of Computational Physics, 520:113488, 2025. 41

    work page 2025

Show all 68 references
  1. [9]

    Kernel methods are competitive for operator learning.Journal of Computational Physics, 496:112549, 2024

    Pau Batlle, Matthieu Darcy, Bamdad Hosseini, and Houman Owhadi. Kernel methods are competitive for operator learning.Journal of Computational Physics, 496:112549, 2024

    2024

  2. [10]

    Brezis.Functional Analysis, Sobolev Spaces and Partial Differential Equations

    H. Brezis.Functional Analysis, Sobolev Spaces and Partial Differential Equations. Universitext. Springer New York, 2010

    2010

  3. [11]

    Vicon: Vision in- context operator networks for multi-physics fluid dynamics prediction.arXiv preprint arXiv:2411.16063, 2024

    Yadi Cao, Yuxuan Liu, Liu Yang, Rose Yu, Hayden Schaeffer, and Stanley Osher. Vicon: Vision in- context operator networks for multi-physics fluid dynamics prediction.arXiv preprint arXiv:2411.16063, 2024

    2024

  4. [12]

    Carmeli, E

    C. Carmeli, E. De Vito, A. Toido, and V . Umanità. Vector valued reproducing kernel hilbert spaces and universality.Analysis and Applications, 08(01):19–61, 2010

    2010

  5. [13]

    Yifan Chen, Bamdad Hosseini, Houman Owhadi, and Andrew M. Stuart. Solving and learning nonlinear pdes with gaussian processes.Journal of Computational Physics, 447:110668, 2021

    2021

  6. [14]

    Conway.A Course in Functional Analysis, volume 96 ofGraduate Texts in Mathematics

    John B. Conway.A Course in Functional Analysis, volume 96 ofGraduate Texts in Mathematics. Springer, New York, NY , 2 edition, 2007

    2007

  7. [15]

    Springer, New York, 3 edition, 2002

    Brian Davies.Integral Transforms and Their Applications. Springer, New York, 3 edition, 2002

    2002

  8. [16]

    Duffy.Green’s Functions with Applications

    Dean G. Duffy.Green’s Functions with Applications. CRC Press, Boca Raton, FL, 2 edition, 2015

    2015

  9. [17]

    Evans.Partial Differential Equations, volume 19 ofGraduate Studies in Mathematics

    Lawrence C. Evans.Partial Differential Equations, volume 19 ofGraduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2 edition, 2010

    2010

  10. [18]

    Cambridge University Press, 2022

    Simon Foucart.Mathematical Pictures at a Data Science Exhibition. Cambridge University Press, 2022

    2022

  11. [19]

    Vector-valued gaussian processes for ap- proximating divergence- or rotation-free vector fields.Journal of Machine Learning Research, 27(74):1– 36, 2026

    Quoc Thong Le Gia, Ian Hugh Sloan, and Holger Wendland. Vector-valued gaussian processes for ap- proximating divergence- or rotation-free vector fields.Journal of Machine Learning Research, 27(74):1– 36, 2026

    2026

  12. [20]

    Kernel methods for bayesian elliptic inverse problems on manifolds.SIAM/ASA Journal on Uncertainty Quantification, 8(4):1414–1445, 2020

    John Harlim, Daniel Sanz-Alonso, and Ruiyi Yang. Kernel methods for bayesian elliptic inverse problems on manifolds.SIAM/ASA Journal on Uncertainty Quantification, 8(4):1414–1445, 2020

    2020

  13. [21]

    Poseidon: Efficient foundation models for PDEs

    Maximilian Herde, Bogdan Raonic, Tobias Rohner, Roger Käppeli, Roberto Molinaro, Emmanuel de Bezenac, and Siddhartha Mishra. Poseidon: Efficient foundation models for PDEs. InThe Thirty- eighth Annual Conference on Neural Information Processing Systems, 2024

    2024

  14. [22]

    Sparse learning of dynamical systems in RKHS: An operator-theoretic approach

    Boya Hou, Sina Sanjari, Nathan Dahlin, Subhonmesh Bose, and Umesh Vaidya. Sparse learning of dynamical systems in RKHS: An operator-theoretic approach. InProceedings of the 40th International Conference on Machine Learning, volume 202 ofProceedings of Machine Learning Research...

    2023

  15. [23]

    Data-efficient kernel methods for learning differential equations and their solution operators: Algorithms and error analysis, 2025

    Yasamin Jalalian, Juan Felipe Osorio Ramirez, Alexander Hsu, Bamdad Hosseini, and Houman Owhadi. Data-efficient kernel methods for learning differential equations and their solution operators: Algorithms and error analysis, 2025

    2025

  16. [24]

    Minimax optimal kernel operator learning via multilevel training

    Jikai Jin, Yiping Lu, Jose Blanchet, and Lexing Ying. Minimax optimal kernel operator learning via multilevel training. InThe Eleventh International Conference on Learning Representations, 2023

    2023

  17. [25]

    Mionet: Learning multiple-input operators via tensor product

    Pengzhan Jin, Shuai Meng, and Lu Lu. Mionet: Learning multiple-input operators via tensor product. SIAM Journal on Scientific Computing, 44(6):A3490–A3514, 2022

    2022

  18. [26]

    Time-series forecasting and refine- ment within a multimodal pde foundation model.Journal of Machine Learning for Modeling and Com- puting, 6(2):77–89, 2025

    Derek Jollie, Jingmin Sun, Zecheng Zhang, and Hayden Schaeffer. Time-series forecasting and refine- ment within a multimodal pde foundation model.Journal of Machine Learning for Modeling and Com- puting, 6(2):77–89, 2025

    2025

  19. [27]

    Operator-valued kernels for learning from functional response data.Journal of Machine Learning Research, 17(20):1–54, 2016

    Hachem Kadri, Emmanuel Duflos, Philippe Preux, Stéphane Canu, Alain Rakotomamonjy, and Julien Audiffren. Operator-valued kernels for learning from functional response data.Journal of Machine Learning Research, 17(20):1–54, 2016. 42

    2016

  20. [28]

    Kernel-based operator learning: Error analysis, budget allocation, and a physics- informed extension, 2026

    Rüdiger Kempf. Kernel-based operator learning: Error analysis, budget allocation, and a physics- informed extension, 2026

    2026

  21. [29]

    Operator learning with pca-net: upper and lower complexity bounds.J

    Samuel Lanthaler. Operator learning with pca-net: upper and lower complexity bounds.J. Mach. Learn. Res., 24(1), January 2023

    2023

  22. [30]

    Fourier neural operator for parametric partial differential equations

    Zongyi Li, Nikola Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Fourier neural operator for parametric partial differential equations. In International Conference on Learning Representations (ICLR), 2021. preprint ar...

    2021 arXiv

  23. [31]

    Cauchy random features for operator learning in sobolev space.arXiv: 2503.00300, 2025

    Chunyang Liao, Deanna Needell, and Hayden Schaeffer. Cauchy random features for operator learning in sobolev space.arXiv: 2503.00300, 2025

    2025 arXiv

  24. [32]

    Prose-fd: A multimodal pde foundation model for learning multiple operators for forecasting fluid dynamics.arXiv preprint arXiv:2409.09811, 2024

    Yuxuan Liu, Jingmin Sun, Xinjie He, Griffin Pinney, Zecheng Zhang, and Hayden Schaeffer. Prose-fd: A multimodal pde foundation model for learning multiple operators for forecasting fluid dynamics.arXiv preprint arXiv:2409.09811, 2024

    2024 arXiv

  25. [33]

    Bcat: A block causal transformer for pde foundation models for fluid dynamics.arXiv preprint arXiv:2501.18972, 2025

    Yuxuan Liu, Jingmin Sun, and Hayden Schaeffer. Bcat: A block causal transformer for pde foundation models for fluid dynamics.arXiv preprint arXiv:2501.18972, 2025

    2025 arXiv

  26. [34]

    Prose: Predicting multiple operators and symbolic expressions using multimodal transformers.Neural Networks, 180:106707, 2024

    Yuxuan Liu, Zecheng Zhang, and Hayden Schaeffer. Prose: Predicting multiple operators and symbolic expressions using multimodal transformers.Neural Networks, 180:106707, 2024

    2024

  27. [35]

    Learning nonlinear operators via deeponet based on the universal approximation theorem of operators.Nature Machine Intelligence, 3(3):218–229, 2021

    Lu Lu, Pengzhan Jin, Guofei Pang, Zhongqiang Zhang, and George Em Karniadakis. Learning nonlinear operators via deeponet based on the universal approximation theorem of operators.Nature Machine Intelligence, 3(3):218–229, 2021

    2021

  28. [36]

    Optimal recovery of functions and their derivatives from Fourier coefficients prescribed with an error.Sbornik

    G G Magaril-Il’yaev and K Yu Osipenko. Optimal recovery of functions and their derivatives from Fourier coefficients prescribed with an error.Sbornik. Mathematics, 193(3), January 2025

    2025

  29. [37]

    Multiple physics pretraining for physical surrogate models.arXiv preprint arXiv:2310.02994, 2023

    Michael McCabe, Bruno Régaldo-Saint Blancard, Liam Holden Parker, Ruben Ohana, Miles Cranmer, Alberto Bietti, Michael Eickenberg, Siavash Golkar, Géraud Krawezik, Francois Lanusse, et al. Multiple physics pretraining for physical surrogate models.arXiv preprint arXiv:2310.02994, 2023

    arXiv

  30. [38]

    Op- erator learning with gaussian processes.Computer Methods in Applied Mechanics and Engineering, 434:117581, 2025

    Carlos Mora, Amin Yousefpour, Shirin Hosseinmardi, Houman Owhadi, and Ramin Bostanabad. Op- erator learning with gaussian processes.Computer Methods in Applied Mechanics and Engineering, 434:117581, 2025

    2025

  31. [39]

    A multimodal pde founda- tion model for prediction and scientific text descriptions.arXiv preprint arXiv:2502.06026, 2025

    Elisa Negrini, Yuxuan Liu, Liu Yang, Stanley J Osher, and Hayden Schaeffer. A multimodal pde founda- tion model for prediction and scientific text descriptions.arXiv preprint arXiv:2502.06026, 2025

    2025 arXiv

  32. [40]

    Nelsen and Andrew M

    Nicholas H. Nelsen and Andrew M. Stuart. Operator learning using random features: A tool for scientific computing.SIAM Review, 66(3):535–571, 2024

    2024

  33. [41]

    On optimal recovery methods in hardy-sobolev spaces.Sbornik: Mathematics, 192(2):225, feb 2001

    K Yu Osipenko. On optimal recovery methods in hardy-sobolev spaces.Sbornik: Mathematics, 192(2):225, feb 2001

    2001

  34. [42]

    Do ideas have shape? idea registration as the continuous limit of artificial neural networks.Physica D: Nonlinear Phenomena, 444:133592, 2023

    Houman Owhadi. Do ideas have shape? idea registration as the continuous limit of artificial neural networks.Physica D: Nonlinear Phenomena, 444:133592, 2023

    2023

  35. [43]

    Cam- bridge Monographs on Applied and Computational Mathematics

    Houman Owhadi and Clint Scovel.Operator-Adapted Wavelets, Fast Solvers, and Numerical Homoge- nization: From a Game Theoretic Approach to Numerical Approximation and Algorithm Design. Cam- bridge Monographs on Applied and Computational Mathematics. Cambridge University Press, 2019

    2019

  36. [44]

    Robey and J

    R. Robey and J. K. Lundquist. Behavior and mechanisms of doppler wind lidar error in varying stability regimes.Atmospheric Measurement Techniques, 15(15):4585–4622, 2022

    2022

  37. [45]

    Andrew M. Stuart. Inverse problems: A bayesian perspective.Acta Numerica, 19:451–559, 2010. 43

    2010

  38. [46]

    Towards a foundation model for partial differential equations: Multioperator learning and extrapolation.Physical Review E, 111(3):035304, 2025

    Jingmin Sun, Yuxuan Liu, Zecheng Zhang, and Hayden Schaeffer. Towards a foundation model for partial differential equations: Multioperator learning and extrapolation.Physical Review E, 111(3):035304, 2025

    2025

  39. [47]

    Lemon: Learning to learn multi-operator networks

    Jingmin Sun, Zecheng Zhang, and Hayden Schaeffer. Lemon: Learning to learn multi-operator networks. arXiv preprint arXiv:2408.16168, 2024

    2024 arXiv

  40. [48]

    Pdebench: an extensive benchmark for scientific machine learning

    Makoto Takamoto, Timothy Praditia, Raphael Leiteritz, Dan MacKinlay, Francesco Alesiani, Dirk Pflüger, and Mathias Niepert. Pdebench: an extensive benchmark for scientific machine learning. In Proceedings of the 36th International Conference on Neural Information Processing Sy...

    2022

  41. [49]

    Non-local observations and information transfer in data assimilation.Frontiers in Applied Mathematics and Statistics, V olume 5 - 2019, 2019

    Peter Jan van Leeuwen. Non-local observations and information transfer in data assimilation.Frontiers in Applied Mathematics and Statistics, V olume 5 - 2019, 2019

    2019

  42. [50]

    Opinf-llm: Parametric pde solving with llms via operator inference, 2026

    Zhuoyuan Wang, Hanjiang Hu, Xiyu Deng, Saviz Mowlavi, and Yorie Nakahira. Opinf-llm: Parametric pde solving with llms via operator inference, 2026

    2026

  43. [51]

    Generalization bounds and statistical guarantees for multi-task and multiple operator learning with mno networks, 2026

    Adrien Weihs and Hayden Schaeffer. Generalization bounds and statistical guarantees for multi-task and multiple operator learning with mno networks, 2026

    2026

  44. [52]

    Multiple neural operators achieve near-optimal rates for multi-task learning, 2026

    Adrien Weihs and Hayden Schaeffer. Multiple neural operators achieve near-optimal rates for multi-task learning, 2026

    2026

  45. [53]

    A deep learning framework for multi-operator learning: Architectures and approximation theory, 2025

    Adrien Weihs, Jingmin Sun, Zecheng Zhang, and Hayden Schaeffer. A deep learning framework for multi-operator learning: Architectures and approximation theory, 2025. arXiv:2510.25379

    2025

  46. [54]

    Cambridge Monographs on Applied and Computa- tional Mathematics

    Holger Wendland.Scattered Data Approximation. Cambridge Monographs on Applied and Computa- tional Mathematics. Cambridge University Press, 2004

    2004

  47. [55]

    In-context operator learning with data prompts for differential equation problems.Proceedings of the National Academy of Sciences, 120(39):e2310142120, 2023

    Liu Yang, Siting Liu, Tingwei Meng, and Stanley J Osher. In-context operator learning with data prompts for differential equation problems.Proceedings of the National Academy of Sciences, 120(39):e2310142120, 2023

    2023

  48. [56]

    Prompting in-context operator learning with sensor data, equations, and natural language.arXiv preprint arXiv:2308.05061, 2023

    Liu Yang, Tingwei Meng, Siting Liu, and Stanley J Osher. Prompting in-context operator learning with sensor data, equations, and natural language.arXiv preprint arXiv:2308.05061, 2023

    2023 arXiv

  49. [57]

    Generalization guarantees for multi-input neural operator learning in sobolev spaces, 2026

    Yahong Yang, Zecheng Zhang, Wei Zhu, Wenjing Liao, and Hao Liu. Generalization guarantees for multi-input neural operator learning in sobolev spaces, 2026

    2026

  50. [58]

    Pdeformer-2: A versatile foundation model for two- dimensional partial differential equations.arXiv preprint arXiv:2507.15409, 2025

    Zhanhong Ye, Zining Liu, Bingyang Wu, Hongjie Jiang, Leheng Chen, Minyan Zhang, Xiang Huang, Qinghe Meng Zou, Hongsheng Liu, and Bin Dong. Pdeformer-2: A versatile foundation model for two- dimensional partial differential equations.arXiv preprint arXiv:2507.15409, 2025

    2025 arXiv

  51. [59]

    Regularized random fourier features and finite element reconstruction for operator learning in sobolev space.Journal of Machine Learning for Modeling and Computing, 7(3):1–47, 2026

    Xinyue Yu and Hayden Schaeffer. Regularized random fourier features and finite element reconstruction for operator learning in sobolev space.Journal of Machine Learning for Modeling and Computing, 7(3):1–47, 2026

    2026

  52. [60]

    Probabilistic operator learn- ing: generative modeling and uncertainty quantification for foundation models of differential equations

    Benjamin J Zhang, Siting Liu, Stanley J Osher, and Markos A Katsoulakis. Probabilistic operator learn- ing: generative modeling and uncertainty quantification for foundation models of differential equations. arXiv preprint arXiv:2509.05186, 2025

    2025 arXiv

  53. [61]

    Modno: Multi-operator learning with distributed neural operators.Computer Methods in Applied Mechanics and Engineering, 431:117229, 2024

    Zecheng Zhang. Modno: Multi-operator learning with distributed neural operators.Computer Methods in Applied Mechanics and Engineering, 431:117229, 2024

    2024

  54. [62]

    A discretization-invariant extension and analysis of some deep operator networks.Journal of Computational and Applied Mathematics, 456:116226, 2025

    Zecheng Zhang, Wing Tat Leung, and Hayden Schaeffer. A discretization-invariant extension and analysis of some deep operator networks.Journal of Computational and Applied Mathematics, 456:116226, 2025. 44

    2025

  55. [63]

    D2no: Efficient handling of heterogeneous input function spaces with distributed deep neural operators.Computer Methods in Applied Mechanics and Engineering, 428:117084, 2024

    Zecheng Zhang, Christian Moya, Lu Lu, Guang Lin, and Hayden Schaeffer. D2no: Efficient handling of heterogeneous input function spaces with distributed deep neural operators.Computer Methods in Applied Mechanics and Engineering, 428:117084, 2024

    2024

  56. [64]

    Deeponet as a multi-operator extrapolation model: Distributed pretraining with physics-informed fine-tuning.Journal of Computa- tional Physics, page 114537, 2025

    Zecheng Zhang, Christian Moya, Lu Lu, Guang Lin, and Hayden Schaeffer. Deeponet as a multi-operator extrapolation model: Distributed pretraining with physics-informed fine-tuning.Journal of Computa- tional Physics, page 114537, 2025

    2025

  57. [65]

    Belnet: basis enhanced learning, a mesh-free neural operator.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 479(2276):20230043, 2023

    Zecheng Zhang, Leung Wing Tat, and Hayden Schaeffer. Belnet: basis enhanced learning, a mesh-free neural operator.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 479(2276):20230043, 2023

    2023

  58. [66]

    Pi-mfm: Physics-informed mul- timodal foundation model for solving partial differential equations.arXiv preprint arXiv:2512.23056, 2025

    Min Zhu, Jingmin Sun, Zecheng Zhang, Hayden Schaeffer, and Lu Lu. Pi-mfm: Physics-informed mul- timodal foundation model for solving partial differential equations.arXiv preprint arXiv:2512.23056, 2025. 45 A Proofs A.1 Proofs of the Background Section Proof of Theorem 2.1.We f...

    2025

  59. [67]

    Ifx 0∈A S, then (7)A S =x 0 + kerL

    SinceLis surjective, the affine constraint setA S :={x∈X:Lx=S}is nonempty. Ifx 0∈A S, then (7)A S =x 0 + kerL. Ifh∈kerL, thenL(x 0 +h) =S, and conversely, ifx∈A S, thenL(x−x 0) = 0, sox−x 0∈kerL. Letx∈A S and decompose it as x=x ⊥ +x ker, x ⊥∈(kerL) ⊥, x ker∈kerL. Then,Lx⊥ =Lx...

  60. [68]

    Define the functional Jγ :X→R, J γ(x) :=∥x∥ 2 X +γ−1∥Lx−S∥ 2 Z

    We start by proving (8) xγ(S) = (L∗L+γI X)−1L∗S. Define the functional Jγ :X→R, J γ(x) :=∥x∥ 2 X +γ−1∥Lx−S∥ 2 Z. The functionalJ γ is strictly convex, sincex7→∥x∥ 2 X is strictly convex. HenceJ γ has at most one minimizer. We compute the first-order optimality condition. Leth∈...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.