Pith. sign in

REVIEW 3 major objections 95 references

Low-fidelity predictions become extra inputs that let a Gaussian process learn a high-fidelity surrogate from scarce data, with better accuracy and lower cost than standard multifidelity methods.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-13 20:27 UTC pith:WSRUITZC

load-bearing objection Practical multifidelity GP recipe that is genuinely useful, with one real fairness soft spot in the baselines that does not sink the contribution. the 3 major comments →

arxiv 2603.22050 v2 pith:WSRUITZC submitted 2026-03-23 stat.ML cs.LG

Multifidelity-Augmented Gaussian Process Inputs for Surrogate Modeling from Scarce Data

Atticus Rex , Elizabeth Qian , David Peterson This is my paper
classification stat.ML cs.LG
keywords multifidelity learningGaussian process regressionsurrogate modelingscarce dataautoregressive estimatorscokriginginput augmentation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

When high-fidelity simulations or experiments are expensive, only a handful of labeled points can be collected, so ordinary surrogates overfit or become useless outside the data. Cheaper low-fidelity models exist, yet classical multifidelity Gaussian-process recipes either invert huge joint covariances or chain models one level at a time under a Markov restriction. This paper shows that simply treating every trained low-fidelity predictor as an extra coordinate of the high-fidelity input space removes both problems: the high-fidelity Gaussian process can combine all of them nonlinearly, while training still proceeds sequentially and low-fidelity models need not themselves be Gaussian processes. On a synthetic nonlinear map, a laminar-flame-speed extrapolation task, and a sparse CFD velocity field, the resulting surrogates cut root-mean-square error relative to Kennedy–O’Hagan, NARGP, and single-fidelity kriging while keeping only an O(N_high^3) kernel factorization. A linear mean that already mixes the low-fidelity predictions further stabilizes extrapolation far from the scarce high-fidelity points.

Core claim

Augmenting the high-fidelity input vector with the predictions of every available low-fidelity surrogate produces a single Gaussian process whose posterior mean and variance are more accurate than both single-fidelity kriging and the leading autoregressive multifidelity estimators, at a training cost that scales only with the scarce high-fidelity sample size.

What carries the argument

The multifidelity feature map φ_l that recursively concatenates the original coordinates with every lower-fidelity predictor already trained; the high-fidelity GP is then fit on φ_1(X_1) and can therefore condition on all low-fidelity information at once.

Load-bearing premise

The reported gains rest on giving the proposed method a linear mean that already mixes low-fidelity predictions while restricting the baseline methods to constant means, and on replacing low-fidelity Gaussian processes by nearest-neighbor models whenever full GPs become too expensive.

What would settle it

Retrain every method on the same three test suites under identical mean functions and identical low-fidelity regressors (full GPs when feasible); if the accuracy and log-marginal-likelihood gaps disappear, the claim of superiority does not hold.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Many-query tasks (optimization, UQ, inverse problems) can be run with far fewer high-fidelity evaluations once low-fidelity features are free to enter the kernel.
  • Low-fidelity surrogates need not be Gaussian processes; cheap alternatives such as nearest neighbors or neural nets keep total offline cost practical even for millions of low-fidelity points.
  • Extrapolation beyond a constrained high-fidelity design space becomes reliable when the mean itself is a linear combination of low-fidelity predictors.
  • Uncertainty estimates of the high-fidelity posterior remain well-calibrated enough to flag regions where additional high-fidelity data would help most.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same feature-augmentation idea can be dropped into sparse or deep Gaussian processes without changing the recursive construction, potentially removing residual oscillations seen in the CFD posterior.
  • When low-fidelity models are themselves hierarchical (e.g., successive mesh refinements), an automatic relevancy determination length-scale on each feature can prune useless fidelities before the high-fidelity kernel is optimized.
  • The method supplies a natural experimental-design loop: allocate the next high-fidelity point where the current multifidelity posterior variance is largest relative to the cost of the remaining low-fidelity models.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 0 minor

Summary. The paper proposes MAGPI, a multifidelity Gaussian-process surrogate that recursively augments the high-fidelity input space with predictions of all available lower-fidelity surrogates (Algorithms 1–2, feature map (11)). Low-fidelity stages may be any point-estimate regressor, while the high-fidelity stage remains a GP whose kernel and (optionally linear) mean act on the augmented features; this is claimed to combine cokriging-style use of all fidelities with the sequential cost of autoregressive methods. Complexity is stated as O(N1^3 + sum of low-fidelity train/predict costs). Three numerical studies (1-D analytic, laminar flame-speed extrapolation, sparse LES/RANS velocity interpolation) report lower RMSE, higher R2 and higher log marginal likelihood than Kennedy–O’Hagan, NARGP and single-fidelity kriging (Tables 2, 4, 6).

Significance. If the accuracy and cost claims hold under fair baselines, MAGPI is a practical and useful addition to the multifidelity GP literature: it removes the Markov restriction of classical autoregressive schemes, permits non-GP low-fidelity models (critical when low-fidelity data are abundant), and supplies a simple, implementable recipe that engineers can drop into existing GP tool-chains. The three application-oriented benchmarks and the explicit complexity comparison are genuine strengths. The theoretical contribution (Proposition 1) is modest—an essentially tautological statement that a larger hypothesis class cannot lower the supremum marginal likelihood—but does not undermine the engineering value of the method.

major comments (3)
  1. §4.2 and Eq. (14): MAGPI is given a linear mean that already mixes the low-fidelity surrogate outputs as features, while KOH, NARGP and single-fidelity kriging receive only constant means “to correct for vertical bias.” Because that linear mean is itself a simple multifidelity model, part of the reported gains in Table 2 (and likewise Tables 4 and 6) may be attributable to the mean rather than to the augmented-input kernel. Equalizing the mean class across all methods (or reporting an ablation with constant mean for MAGPI) is required before the claim of superiority over the state of the art can be accepted.
  2. §4.4: Full-rank GPs at the intermediate LES resolutions are declared prohibitively expensive, so KNN approximations are substituted for the low-fidelity stages of KOH and NARGP as well as MAGPI. The resulting comparison therefore no longer tests the published KOH/NARGP algorithms; it tests three different ways of using the same KNN features. Either a sparse/approximate GP implementation of the baselines or an explicit statement that the comparison is only among KNN-augmented variants is needed for the CFD claim to be load-bearing.
  3. Proposition 1 only guarantees that a larger hypothesis class can achieve a higher marginal likelihood; it does not guarantee better generalization or that the observed numerical superiority survives equalized means and equalized low-fidelity models. The abstract and introduction statements that the method is “more accurate … relative to the state of the art” therefore rest almost entirely on the three tables; those tables must be made robust to the two design choices above.

Circularity Check

0 steps flagged

No load-bearing circularity; methods paper whose accuracy claims rest on external numerical benchmarks, with only a near-tautological existence guarantee (Prop. 1) that does not force the reported gains.

full rationale

The derivation of MAGPI is definitional of a new estimator (augment high-fidelity GP inputs by recursive low-fidelity surrogate predictions, Algorithms 1–2 and eqs. 11–13) rather than a claimed first-principles derivation of a physical law. Hyperparameters are obtained by ordinary Type-II maximum likelihood (eq. 5 / §3.1), not by baking test-set RMSE/R^{2} into the objective. Proposition 1 merely observes that a larger hypothesis class (more features) can attain a marginal likelihood at least as high as a nested smaller class; this is true by construction of the suprema but is not used to claim that the numerical superiority over KOH/NARGP/kriging is forced, nor does it equate any reported prediction to a fitted target. No uniqueness theorem, ansatz, or load-bearing result is imported via self-citation. The three numerical tables compare against external baselines on independent test problems; any fairness concerns about mean-function choice or KNN substitution affect experimental design, not circularity of the method’s own equations. Hence the central claim of improved accuracy/cost does not reduce by construction to its inputs.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 1 invented entities

The method rests on standard GP regression machinery plus the modeling choice that low-fidelity surrogate predictions are useful additional inputs. Free parameters are the usual GP hyperparameters and mean coefficients fit by marginal likelihood. No new physical entities are postulated; the 'multifidelity-augmented features' are a constructed representation, not an ontological claim.

free parameters (4)
  • ARD lengthscales λ and kernel amplitude b (high-fidelity GP)
    Optimized by maximizing log marginal likelihood on high-fidelity training data; control relevance of each original and low-fidelity feature.
  • High-fidelity noise variance σ₁²
    Jointly optimized with kernel/mean hyperparameters via Type-II MLE.
  • Linear mean coefficients α_i, β_ℓ, γ
    Calibrated by maximizing log marginal likelihood; β_ℓ directly weight low-fidelity surrogate predictions in the prior mean (eq. 14).
  • Low-fidelity surrogate hyperparameters (if GPs) or model choices (e.g., KNN k)
    Chosen per fidelity; affect the quality of features passed upward. Not jointly optimized with the high-fidelity GP.
axioms (5)
  • standard math High-fidelity observations are a GP prior plus i.i.d. Gaussian noise; posterior mean/variance follow the standard conditioning formulas.
    §2.2 eqs. (2)–(5); canonical Rasmussen & Williams setup.
  • domain assumption Anisotropic RBF/ARD kernels define a sufficiently rich hypothesis class for the target high-fidelity maps.
    Used throughout experiments; smoothness/infinite differentiability assumed for flame speed and flow fields that contain shocks.
  • domain assumption Low-fidelity surrogate predictions are informative features for the high-fidelity map even when correlations are moderate (~0.4–0.6) and relationships are nonlinear.
    Load-bearing for the method; illustrated in §4.2 analytic example and assumed in chemistry/CFD cases.
  • ad hoc to paper Fidelity ordering need only designate y₁ as highest; other levels may be arbitrarily ordered.
    Stated in §2.1; MAGPI does not require a strict fidelity hierarchy for feature construction, unlike Markovian autoregressive schemes.
  • standard math Enlarging the mean/kernel class by adding features cannot decrease the supremum achievable marginal likelihood (Prop. 1).
    §3.1; true for nested hypothesis classes but does not imply better test error.
invented entities (1)
  • Multifidelity-augmented feature map φ_l (recursive concatenation of inputs with all higher-index surrogate predictions) no independent evidence
    purpose: Defines the input space on which the high-fidelity GP is trained and evaluated so all low-fidelity models can influence predictions non-Markovianly.
    Constructed representation, not a physical object; independent_evidence is false because utility is only shown inside this paper's experiments.

pith-pipeline@v1.1.0-grok45 · 31425 in / 3349 out tokens · 35720 ms · 2026-07-13T20:27:02.070952+00:00 · methodology

0 comments
read the original abstract

Supervised machine learning describes the practice of fitting a parameterized model to labeled input-output data. Supervised machine learning methods have demonstrated promise in learning efficient surrogate models that can (partially) replace expensive high-fidelity models, making many-query analyses, such as optimization, uncertainty quantification, and inference, tractable. However, when training data must be obtained through the evaluation of an expensive model or experiment, the amount of training data that can be obtained is often limited, which can make learned surrogate models unreliable. In many engineering and scientific settings, cheaper low-fidelity models may be available, for example arising from simplified physics modeling or coarse grids. These models may be used to generate additional low-fidelity training data. The goal of multifidelity machine learning is to use both high- and low-fidelity training data to learn a surrogate model which is cheaper to evaluate than the high-fidelity model, but more accurate than any available low-fidelity model. This work proposes a new multifidelity training approach for Gaussian process regression which uses low-fidelity data to define additional features that augment the input space of the learned model. Similarly to cokriging estimators, the proposed approach conditions the high-fidelity surrogate model on the predictions of all available low-fidelity surrogate models, while benefiting from the computational efficiency of autoregressive estimators. Numerical experiments on several test problems demonstrate both increased predictive accuracy and reduced computational cost relative to the state of the art.

Figures

Figures reproduced from arXiv: 2603.22050 by Atticus Rex, David Peterson, Elizabeth Qian.

Figure 1
Figure 1. Figure 1: Results from trained models on the analytical test problem. The target functions are plotted with [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A visualization of the high-fidelity (USC II) training and testing data. The blue dots show the high [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The laminar flame speed experiment captured at four different temperatures. The shaded regions [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The full high-fidelity (125 µm LES) flow field with sparse high-fidelity training data indicated with the “+" symbols. We emphasize that single-fidelity GP regression/kriging, achieving the worst RMSE of all models on the testing data, fully interpolates its training data; with scarce data, low training error provides little guar￾antee of a reliable surrogate model. The proposed method achieves better perf… view at source ↗
Figure 5
Figure 5. Figure 5: Results of the sparse flow-field interpolation experiment. The left half of the plots show ap [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison between model error and predictive uncertainty. The top plot shows the difference [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

95 extracted references · 30 canonical work pages · 2 internal anchors

  1. [1]

    Multi-level CFD-based Airfoil Shape Optimization With Automated Low- fidelity Model Selection,

    S. Koziel and L. Leifsson, “Multi-level CFD-based Airfoil Shape Optimization With Automated Low- fidelity Model Selection,”Procedia Computer Science, 2013 International Conference on Computational Science, vol. 18, pp. 889–898, Jan. 2013,ISSN: 1877-0509.DOI:10 . 1016 / j . procs . 2013 . 05 . 254 Accessed: Aug. 10, 2025. [Online]. Available:https://www.sc...

    2013

  2. [2]

    MULTI-FIDELITY MACHINE LEARNING FOR UNCERTAINTY QUANTIFICATION AND OPTIMIZATION,

    R. Zhang and N. Alemazkoor, “MULTI-FIDELITY MACHINE LEARNING FOR UNCERTAINTY QUANTIFICATION AND OPTIMIZATION,” English,Journal of Machine Learning for Modeling and Computing, vol. 5, no. 4, 2024,ISSN: 2689-3967, 2689-3975.DOI:10.1615/JMachLearnModelComput. 2024055786Accessed: Nov. 18, 2024. [Online]. Available:https : / / www . dl . begellhouse . com / jo...

    doi:10.1615/jmachlearnmodelcomput 2024

  3. [3]

    Data-driven model reduction for the Bayesian solution of inverse problems,

    T. Cui, Y. M. Marzouk, and K. E. Willcox, “Data-driven model reduction for the Bayesian solution of inverse problems,” en,International Journal for Numerical Methods in Engineering, vol. 102, no. 5, pp. 966–990, May 2015,ISSN: 0029-5981, 1097-0207.DOI:10.1002/nme.4748Accessed: Oct. 27, 2025. [Online]. Available:https://onlinelibrary.wiley.com/doi/10.1002/nme.4748

    work page doi:10.1002/nme.4748accessed: 2015

  4. [4]

    Model Reduction of Linear Dynamical Systems via Balancing for Bayesian Inference,

    E. Qian et al., “Model Reduction of Linear Dynamical Systems via Balancing for Bayesian Inference,” en,Journal of Scientific Computing, vol. 91, no. 1, p. 29, Mar. 2022,ISSN: 1573-7691.DOI:10 . 1007 / s10915 - 022 - 01798 - 8Accessed: Oct. 27, 2025. [Online]. Available:https : / / doi . org / 10 . 1007 / s10915-022-01798-8

    2022

  5. [5]

    Recent Advances in Surrogate Modeling Methods for Uncer- tainty Quantification and Propagation,

    C. Wang, X. Qiang, M. Xu, and T. Wu, “Recent Advances in Surrogate Modeling Methods for Uncer- tainty Quantification and Propagation,” en,Symmetry, vol. 14, no. 6, p. 1219, Jun. 2022,ISSN: 2073- 8994.DOI:10.3390/sym14061219Accessed: Oct. 27, 2025. [Online]. Available:https://www.mdpi. com/2073-8994/14/6/1219

    work page doi:10.3390/sym14061219accessed: 2022

  6. [6]

    Bayesian Calibration of Computer Models,

    M. C. Kennedy and A. O’Hagan, “Bayesian Calibration of Computer Models,”Journal of the Royal Statistical Society Series B: Statistical Methodology, vol. 63, no. 3, pp. 425–464, Sep. 2001,ISSN: 1369-7412. DOI:10.1111/1467-9868.00294Accessed: Oct. 27, 2025. [Online]. Available:https://doi.org/10. 1111/1467-9868.00294

    doi:10.1111/1467-9868.00294accessed: 2001

  7. [7]

    Design and Analysis of Computer Experiments,

    J. Sacks, W. J. Welch, T. J. Mitchell, and H. P . Wynn, “Design and Analysis of Computer Experiments,” Statistical Science, vol. 4, no. 4, pp. 409–423, Nov. 1989,ISSN: 0883-4237, 2168-8745.DOI:10.1214/ss/ 1177012413Accessed: Oct. 27, 2025. [Online]. Available:https://projecteuclid.org/journals/ statistical- science/volume- 4/issue- 4/Design- and- Analysis...

    doi:10.1214/ss/ 1989

  8. [8]

    Experimental and Numerical Study of the Laminar Burning Velocity and Pollutant Emissions of the Mixture Gas of Methane and Carbon Dioxide,

    Y. Wang, Y. Wang, X. Zhang, G. Zhou, B. Yan, and R. J. M. Bastiaans, “Experimental and Numerical Study of the Laminar Burning Velocity and Pollutant Emissions of the Mixture Gas of Methane and Carbon Dioxide,”International Journal of Environmental Research and Public Health, vol. 19, no. 4, p. 2078, Feb. 2022,ISSN: 1661-7827.DOI:10.3390/ijerph19042078Acce...

    work page doi:10.3390/ijerph19042078accessed: 2078

  9. [9]

    An overview of statistical learning theory,

    V . N. Vapnik, “An overview of statistical learning theory,” eng,IEEE transactions on neural networks, vol. 10, no. 5, pp. 988–999, 1999,ISSN: 1045-9227.DOI:10.1109/72.788640

    doi:10.1109/72.788640 1999

  10. [10]

    Hastie, R

    T. Hastie, R. Tibshirani, and J. Friedman,The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Second Edition, en. Springer Science & Business Media, Aug. 2009, Google-Books-ID: tVIjmNS3Ob8C,ISBN: 978-0-387-84858-7. 21

    2009

  11. [11]

    C. E. Rasmussen and C. K. I. Williams,Gaussian processes for machine learning(Adaptive computation and machine learning), en, 3. print. Cambridge, Mass.: MIT Press, 2008,ISBN: 978-0-262-18253-9

    2008

  12. [12]

    Survey of Multifidelity Methods in Uncertainty Propagation, Inference, and Optimization,

    B. Peherstorfer, K. Willcox, and M. Gunzburger, “Survey of Multifidelity Methods in Uncertainty Propagation, Inference, and Optimization,”SIAM Review, vol. 60, no. 3, pp. 550–591, Jan. 2018,ISSN: 0036-1445.DOI:10.1137/16M1082469Accessed: Oct. 17, 2025. [Online]. Available:https://epubs. siam.org/doi/10.1137/16M1082469

    work page doi:10.1137/16m1082469accessed: 2018

  13. [13]

    Multifidelity linear regression for scientific machine learning from scarce data,

    E. Qian, D. Kang, V . Sella, and A. Chaudhuri, “Multifidelity linear regression for scientific machine learning from scarce data,” en,Foundations of Data Science, vol. 7, no. 1, pp. 271–297, Mar. 2025.DOI: 10.3934/fods.2024049Accessed: Oct. 15, 2025. [Online]. Available:https://www.aimsciences. org/en/article/doi/10.3934/fods.2024049

    work page doi:10.3934/fods.2024049accessed: 2025

  14. [14]

    Review of multi-fidelity models,

    M. G. Fernández-Godino, “Review of multi-fidelity models,”Advances in Computational Science and Engineering, vol. 1, no. 4, pp. 351–400, 2023, arXiv:1609.07196 [stat],ISSN: 2837-1739.DOI:10.3934/ acse.2023015Accessed: Dec. 31, 2024. [Online]. Available:http://arxiv.org/abs/1609.07196

    Pith/arXiv arXiv 2023

  15. [15]

    Overview of Gaussian process based multi-fidelity tech- niques with variable relationship between fidelities, application to aerospace systems,

    L. Brevault, M. Balesdent, and A. Hebbal, “Overview of Gaussian process based multi-fidelity tech- niques with variable relationship between fidelities, application to aerospace systems,”Aerospace Sci- ence and Technology, vol. 107, p. 106 339, Dec. 2020,ISSN: 1270-9638.DOI:10.1016/j.ast.2020.106339 Accessed: Oct. 15, 2025. [Online]. Available:https://www...

    work page doi:10.1016/j.ast.2020.106339 2020

  16. [16]

    MFNets: MULTI-FIDELITY DATA- DRIVEN NETWORKS FOR BAYESIAN LEARNING AND PREDICTION,

    A. A. Gorodetsky, J. D. Jakeman, G. Geraci, and M. S. Eldred, “MFNets: MULTI-FIDELITY DATA- DRIVEN NETWORKS FOR BAYESIAN LEARNING AND PREDICTION,” en,International Journal for Uncertainty Quantification, vol. 10, no. 6, pp. 595–622, 2020,ISSN: 2152-5080.DOI:10.1615/Int.J. UncertaintyQuantification.2020032978Accessed: Jun. 7, 2025. [Online]. Available:http...

    doi:10.1615/int.j 2020

  17. [17]

    Multifidelity deep operator networks for data-driven and physics-informed problems,

    A. A. Howard, M. Perego, G. E. Karniadakis, and P . Stinis, “Multifidelity deep operator networks for data-driven and physics-informed problems,”Journal of Computational Physics, vol. 493, p. 112 462, Nov. 2023,ISSN: 0021-9991.DOI:10 . 1016 / j . jcp . 2023 . 112462Accessed: Feb. 12, 2026. [Online]. Available:https://www.sciencedirect.com/science/article/...

    2023

  18. [18]

    A multifidelity approach to continual learning for physical systems,

    A. Howard, Y. Fu, and P . Stinis, “A multifidelity approach to continual learning for physical systems,” en,Machine Learning: Science and Technology, vol. 5, no. 2, p. 025 042, May 2024,ISSN: 2632-2153.DOI: 10.1088/2632- 2153/ad45b2Accessed: Feb. 12, 2026. [Online]. Available:https://doi.org/10. 1088/2632-2153/ad45b2

    doi:10.1088/2632- 2024

  19. [19]

    Multifidelity deep neural operators for efficient learning of partial differential equations with application to fast inverse design of nanoscale heat transport,

    L. Lu, R. Pestourie, S. G. Johnson, and G. Romano, “Multifidelity deep neural operators for efficient learning of partial differential equations with application to fast inverse design of nanoscale heat transport,”Physical Review Research, vol. 4, no. 2, p. 023 210, Jun. 2022.DOI:10.1103/PhysRevResearch. 4 . 023210Accessed: Feb. 12, 2026. [Online]. Availa...

    doi:10.1103/physrevresearch 2022

  20. [20]

    Heinlein, A

    A. Heinlein, A. A. Howard, D. Beecroft, and P . Stinis,Multifidelity domain decomposition-based physics- informed neural networks and operators for time-dependent problems, en, arXiv:2401.07888 [math], Jun. 2024.DOI:10.48550/arXiv.2401.07888Accessed: Feb. 12, 2026. [Online]. Available:http://arxiv. org/abs/2401.07888 22

    Pith/arXiv arXiv doi:10.48550/arxiv.2401.07888accessed: 2024

  21. [21]

    Multilevel Monte Carlo methods,

    M. B. Giles, “Multilevel Monte Carlo methods,” en,Acta Numerica, vol. 24, pp. 259–328, May 2015, ISSN: 0962-4929, 1474-0508.DOI:10 . 1017 / S096249291500001XAccessed: Jan. 17, 2026. [Online]. Available:https://www.cambridge.org/core/journals/acta-numerica/article/abs/multilevel- monte-carlo-methods/C5AF9A57ED8FF8FDF08074C1071C5511

    2015

  22. [22]

    A generalized approximate con- trol variate framework for multifidelity uncertainty quantification,

    A. A. Gorodetsky, G. Geraci, M. S. Eldred, and J. D. Jakeman, “A generalized approximate con- trol variate framework for multifidelity uncertainty quantification,”Journal of Computational Physics, vol. 408, p. 109 257, May 2020,ISSN: 0021-9991.DOI:10 . 1016 / j . jcp . 2020 . 109257Accessed: Oct. 17, 2025. [Online]. Available:https : / / www . sciencedire...

    2020

  23. [23]

    Optimal Model Management for Multifidelity Monte Carlo Estimation,

    B. Peherstorfer, K. Willcox, and M. Gunzburger, “Optimal Model Management for Multifidelity Monte Carlo Estimation,” en,SIAM Journal on Scientific Computing, vol. 38, no. 5, A3163–A3194, Jan. 2016, ISSN: 1064-8275, 1095-7197.DOI:10.1137/15M1046472Accessed: Oct. 17, 2025. [Online]. Available: http://epubs.siam.org/doi/10.1137/15M1046472

    work page doi:10.1137/15m1046472accessed: 2016

  24. [24]

    A. A. Gorodetsky, J. D. Jakeman, and M. S. Eldred,Grouped approximate control variate estimators, arXiv:2402.14736 [stat], Feb. 2024.DOI:10.48550/arXiv.2402.14736Accessed: Jan. 12, 2026. [On- line]. Available:http://arxiv.org/abs/2402.14736

    Pith/arXiv arXiv doi:10.48550/arxiv.2402.14736accessed: 2024

  25. [25]

    On Multilevel Best Linear Unbiased Estimators,

    D. Schaden and E. Ullmann, “On Multilevel Best Linear Unbiased Estimators,”SIAM/ASA Journal on Uncertainty Quantification, vol. 8, no. 2, pp. 601–635, Jan. 2020.DOI:10.1137/19M1263534Accessed: Oct. 21, 2025. [Online]. Available:https://epubs.siam.org/doi/10.1137/19M1263534

    work page doi:10.1137/19m1263534accessed: 2020

  26. [26]

    Multifidelity Monte Carlo Es- timation of Variance and Sensitivity Indices,

    E. Qian, B. Peherstorfer, D. O’Malley, V . V . Vesselinov, and K. Willcox, “Multifidelity Monte Carlo Es- timation of Variance and Sensitivity Indices,”SIAM/ASA Journal on Uncertainty Quantification, vol. 6, no. 2, pp. 683–706, Jan. 2018.DOI:10.1137/17M1151006Accessed: Feb. 12, 2026. [Online]. Available: https://epubs.siam.org/doi/abs/10.1137/17M1151006

    work page doi:10.1137/17m1151006accessed: 2018

  27. [27]

    A. A. Popov and A. Sandu,Multifidelity Ensemble Kalman Filtering Using Surrogate Models Defined by Physics-Informed Autoencoders, arXiv:2102.13025 [math], Mar. 2021.DOI:10.48550/arXiv.2102.13025 Accessed: Jan. 26, 2026. [Online]. Available:http://arxiv.org/abs/2102.13025

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2102.13025 2021

  28. [28]

    A transport-based multifidelity preconditioner for Markov chain Monte Carlo

    B. Peherstorfer and Y. Marzouk,A transport-based multifidelity preconditioner for Markov chain Monte Carlo, arXiv:1808.09379 [math], Aug. 2018.DOI:10.48550/arXiv.1808.09379Accessed: Jan. 17, 2026. [Online]. Available:http://arxiv.org/abs/1808.09379

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1808.09379accessed: 2018

  29. [29]

    Bayesian inference of Stochastic reaction networks using Multifidelity Sequential Tempered Markov Chain Monte Carlo,

    T. A. Catanach, H. D. Vo, and B. Munsky, “Bayesian inference of Stochastic reaction networks using Multifidelity Sequential Tempered Markov Chain Monte Carlo,” English,International Journal for Un- certainty Quantification, vol. 10, no. 6, Jun. 2020, Number: SAND2020–7898J; SAND2019–15382J,ISSN: 2152-5080.DOI:10.1615/int.j.uncertaintyquantification.202003...

    work page doi:10.1615/int.j.uncertaintyquantification.2020033241accessed: 2020

  30. [30]

    Markov chain Monte Carlo Using an Approximation,

    J. A. Christen and C. Fox, “Markov chain Monte Carlo Using an Approximation,”Journal of Computa- tional and Graphical Statistics, vol. 14, no. 4, pp. 795–810, Dec. 2005, _eprint: https://doi.org/10.1198/106186005X76983, ISSN: 1061-8600.DOI:10.1198/106186005X76983Accessed: Mar. 5, 2026. [Online]. Available:https: //doi.org/10.1198/106186005X76983

    doi:10.1198/106186005x76983 2005

  31. [31]

    Multi-fidelity Monte Carlo: A pseudo-marginal approach,

    D. Cai and R. P . Adams, “Multi-fidelity Monte Carlo: A pseudo-marginal approach,” en,Advances in Neural Information Processing Systems, vol. 35, pp. 21 654–21 667, Dec. 2022. Accessed: Jan. 17, 2026. [Online]. Available:https://proceedings.neurips.cc/paper_files/paper/2022/hash/8803b9ae0b13011f28e6dd57da2ebbd8- Abstract-Conference.html?utm_source=chatgpt.com 23

    2022

  32. [32]

    Multifidelity multilevel Monte Carlo to accelerate approximate Bayesian parameter inference for partially observed stochastic processes,

    D. J. Warne, T. P . Prescott, R. E. Baker, and M. J. Simpson, “Multifidelity multilevel Monte Carlo to accelerate approximate Bayesian parameter inference for partially observed stochastic processes,” Journal of Computational Physics, vol. 469, p. 111 543, Nov. 2022,ISSN: 0021-9991.DOI:10 . 1016 / j . jcp.2022.111543Accessed: Jan. 17, 2026. [Online]. Avai...

    arXiv 2022

  33. [33]

    A relaxed localized trust-region reduced basis approach for optimization of multiscale problems,

    T. Keil and M. Ohlberger, “A relaxed localized trust-region reduced basis approach for optimization of multiscale problems,” en,ESAIM: Mathematical Modelling and Numerical Analysis, vol. 58, no. 1, pp. 79–105, Jan. 2024,ISSN: 2822-7840, 2804-7214.DOI:10 . 1051 / m2an / 2023089Accessed: Feb. 12,

    2024

  34. [34]

    Available:https://www.esaim-m2an.org/articles/m2an/abs/2024/01/m2an220169/ m2an220169.html

    [Online]. Available:https://www.esaim-m2an.org/articles/m2an/abs/2024/01/m2an220169/ m2an220169.html

    2024

  35. [35]

    A Certified Trust Region Reduced Basis Approach to PDE-Constrained Optimization,

    E. Qian, M. Grepl, K. Veroy, and K. Willcox, “A Certified Trust Region Reduced Basis Approach to PDE-Constrained Optimization,”SIAM Journal on Scientific Computing, vol. 39, no. 5, S434–S460, Jan. 2017,ISSN: 1064-8275.DOI:10.1137/16M1081981Accessed: Feb. 12, 2026. [Online]. Available:https: //epubs.siam.org/doi/abs/10.1137/16M1081981

    work page doi:10.1137/16m1081981accessed: 2017

  36. [36]

    Klein and M

    B. Klein and M. Ohlberger,Multi-fidelity Learning of Reduced Order Models for Parabolic PDE Constrained Optimization, arXiv:2503.21252 [math], Mar. 2025.DOI:10 . 48550 / arXiv . 2503 . 21252Accessed: Feb. 12, 2026. [Online]. Available:http://arxiv.org/abs/2503.21252

    Pith/arXiv arXiv 2025

  37. [37]

    Li and F

    K. Li and F. Li,Multi-Fidelity Methods for Optimization: A Survey, arXiv:2402.09638 [cs], Feb. 2024.DOI: 10.48550/arXiv.2402.09638Accessed: Feb. 12, 2026. [Online]. Available:http://arxiv.org/abs/ 2402.09638

    Pith/arXiv arXiv doi:10.48550/arxiv.2402.09638accessed: 2024

  38. [38]

    A General Framework for Multi-fidelity Bayesian Optimization with Gaussian Processes,

    J. Song, Y. Chen, and Y. Yue, “A General Framework for Multi-fidelity Bayesian Optimization with Gaussian Processes,” en, inProceedings of the Twenty-Second International Conference on Artificial Intel- ligence and Statistics, PMLR, Apr. 2019, pp. 3158–3167. Accessed: Feb. 12, 2026. [Online]. Available: https://proceedings.mlr.press/v89/song19b.html

    2019

  39. [39]

    Multi-fidelity optimization via surrogate modelling,

    A. I. Forrester, A. Sóbester, and A. J. Keane, “Multi-fidelity optimization via surrogate modelling,” Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, vol. 463, no. 2088, pp. 3251–3269, Oct. 2007,ISSN: 1364-5021.DOI:10.1098/rspa.2007.1900Accessed: Feb. 12, 2026. [Online]. Available:https://doi.org/10.1098/rspa.2007.1900

    doi:10.1098/rspa.2007.1900accessed: 2088

  40. [40]

    A multilevel projection-based model order reduction framework for nonlinear dynamic multiscale problems in structural and solid mechanics,

    M. J. Zahr, P . Avery, and C. Farhat, “A multilevel projection-based model order reduction framework for nonlinear dynamic multiscale problems in structural and solid mechanics,” en,International Jour- nal for Numerical Methods in Engineering, vol. 112, no. 8, pp. 855–881, Nov. 2017,ISSN: 0029-5981, 1097- 0207.DOI:10.1002/nme.5535Accessed: Feb. 12, 2026. ...

    work page doi:10.1002/nme.5535accessed: 2017

  41. [41]

    Design optimization using hyper-reduced-order mod- els,

    D. Amsallem, M. Zahr, Y. Choi, and C. Farhat, “Design optimization using hyper-reduced-order mod- els,” en,Structural and Multidisciplinary Optimization, vol. 51, no. 4, pp. 919–940, Apr. 2015,ISSN: 1615- 147X, 1615-1488.DOI:10 . 1007 / s00158 - 014 - 1183 - yAccessed: Feb. 12, 2026. [Online]. Available: http://link.springer.com/10.1007/s00158-014-1183-y

    doi:10.1007/s00158-014-1183-y 2015

  42. [42]

    An optimization-based framework for anisotropic simplex mesh adap- tation,

    M. Yano and D. L. Darmofal, “An optimization-based framework for anisotropic simplex mesh adap- tation,”Journal of Computational Physics, vol. 231, no. 22, pp. 7626–7649, Sep. 2012,ISSN: 0021-9991. DOI:10 . 1016 / j . jcp . 2012 . 06 . 040Accessed: Feb. 12, 2026. [Online]. Available:https : / / www . sciencedirect.com/science/article/pii/S0021999112003749 24

    2012

  43. [43]

    A statistical approach to some basic mine valuation problems on the Witwatersrand,

    D. G. Krige, “A statistical approach to some basic mine valuation problems on the Witwatersrand,” Journal of the Southern African Institute of Mining and Metallurgy, vol. 52, no. 6, pp. 119–139, Dec. 1951. DOI:10.10520/AJA0038223X_4792Accessed: Jun. 23, 2025. [Online]. Available:https://journals. co.za/doi/10.10520/AJA0038223X_4792

    work page doi:10.10520/aja0038223x_4792accessed: 1951

  44. [44]

    Principles of geostatistics,

    G. Matheron, “Principles of geostatistics,”Economic Geology, vol. 58, no. 8, pp. 1246–1266, Dec. 1963, ISSN: 0361-0128.DOI:10.2113/gsecongeo.58.8.1246Accessed: Jun. 23, 2025. [Online]. Available: https://doi.org/10.2113/gsecongeo.58.8.1246

    work page doi:10.2113/gsecongeo.58.8.1246accessed: 1963

  45. [45]

    Predicting the Output from a Complex Computer Code When Fast Approximations Are Available,

    M. C. Kennedy and A. O’Hagan, “Predicting the Output from a Complex Computer Code When Fast Approximations Are Available,”Biometrika, vol. 87, no. 1, pp. 1–13, 2000,ISSN: 0006-3444. Accessed: Nov. 23, 2024. [Online]. Available:https://www.jstor.org/stable/2673557

    arXiv 2000

  46. [46]

    Priors for Infinite Networks,

    R. M. Neal, “Priors for Infinite Networks,” en, inBayesian Learning for Neural Networks, R. M. Neal, Ed., New York, NY: Springer, 1996, pp. 29–53,ISBN: 978-1-4612-0745-0.DOI:10.1007/978-1-4612-0745- 0_2Accessed: Nov. 3, 2025. [Online]. Available:https://doi.org/10.1007/978-1-4612-0745-0_2

    work page doi:10.1007/978-1-4612-0745- 1996

  47. [47]

    Multilayer feedforward networks are universal approx- imators,

    K. Hornik, M. Stinchcombe, and H. White, “Multilayer feedforward networks are universal approx- imators,”Neural Networks, vol. 2, no. 5, pp. 359–366, Jan. 1989,ISSN: 0893-6080.DOI:10.1016/0893- 6080(89)90020- 8Accessed: Nov. 3, 2025. [Online]. Available:https://www.sciencedirect.com/ science/article/pii/0893608089900208

    doi:10.1016/0893- 1989

  48. [48]

    Evaluation of gaussian processes and other methods for non-linear regression,

    C. E. Rasmussen, “Evaluation of gaussian processes and other methods for non-linear regression,” AAINQ28300 ISBN-10: 0612283003, phd, University of Toronto, CAN, 1997

    1997

  49. [49]

    Matrix formulation of co-kriging,

    D. E. Myers, “Matrix formulation of co-kriging,” en,Journal of the International Association for Math- ematical Geology, vol. 14, no. 3, pp. 249–257, Jun. 1982,ISSN: 1573-8868.DOI:10 . 1007 / BF01032887 Accessed: Sep. 10, 2024. [Online]. Available:https://doi.org/10.1007/BF01032887

    work page doi:10.1007/bf01032887 1982

  50. [50]

    Linear coregionalization model: Tools for estimation and choice of cross- variogram matrix,

    M. Goulard and M. Voltz, “Linear coregionalization model: Tools for estimation and choice of cross- variogram matrix,” en,Mathematical Geology, vol. 24, no. 3, pp. 269–286, Apr. 1992,ISSN: 1573-8868. DOI:10.1007/BF00893750Accessed: Sep. 9, 2025. [Online]. Available:https://doi.org/10.1007/ BF00893750

    work page doi:10.1007/bf00893750accessed: 1992

  51. [51]

    Kernels for Vector-Valued Functions: A Review,

    M. A. Álvarez, L. Rosasco, and N. D. Lawrence, “Kernels for Vector-Valued Functions: A Review,” English,Foundations and Trends® in Machine Learning, vol. 4, no. 3, pp. 195–266, Jun. 2012,ISSN: 1935- 8237, 1935-8245.DOI:10.1561/2200000036Accessed: Oct. 15, 2025. [Online]. Available:https:// www.nowpublishers.com/article/Details/MAL-036

    work page doi:10.1561/2200000036accessed: 2012

  52. [52]

    Multi-task Gaussian Process Prediction,

    E. V . Bonilla, K. Chai, and C. Williams, “Multi-task Gaussian Process Prediction,” inAdvances in Neural Information Processing Systems, vol. 20, Curran Associates, Inc., 2007. Accessed: Oct. 15, 2025. [Online]. Available:https://papers.nips.cc/paper_files/paper/2007/hash/66368270ffd51418ec58bd793f2d9b1b- Abstract.html

    arXiv 2007

  53. [53]

    Z. Xu, H. Wang, J. M. Phillips, and S. Zhe,Standard Gaussian Process is All You Need for High-Dimensional Bayesian Optimization, arXiv:2402.02746 [cs], Mar. 2025.DOI:10.48550/arXiv.2402.02746Accessed: Oct. 27, 2025. [Online]. Available:http://arxiv.org/abs/2402.02746

    Pith/arXiv arXiv doi:10.48550/arxiv.2402.02746accessed: 2025

  54. [54]

    Kernel principal component analysis-based Gaussian process regression mod- elling for high-dimensional reliability analysis,

    T. Zhou and Y. Peng, “Kernel principal component analysis-based Gaussian process regression mod- elling for high-dimensional reliability analysis,”Computers & Structures, vol. 241, p. 106 358, Dec. 2020,ISSN: 0045-7949.DOI:10.1016/j.compstruc.2020.106358Accessed: Oct. 27, 2025. [Online]. Available:https://www.sciencedirect.com/science/article/pii/S0045794...

    work page doi:10.1016/j.compstruc.2020.106358accessed: 2020

  55. [55]

    Variable noise and dimensionality reduction for sparse Gaussian pro- cesses,

    E. Snelson and Z. Ghahramani, “Variable noise and dimensionality reduction for sparse Gaussian pro- cesses,” inProceedings of the Twenty-Second Conference on Uncertainty in Artificial Intelligence, ser. UAI’06, Arlington, Virginia, USA: AUAI Press, Jul. 2006, pp. 461–468,ISBN: 978-0-9749039-2-7. Accessed: Oct. 27, 2025

    2006

  56. [56]

    Gaussian processes with built-in dimensionality reduction: Applications to high-dimensional uncertainty propagation,

    R. Tripathy, I. Bilionis, and M. Gonzalez, “Gaussian processes with built-in dimensionality reduction: Applications to high-dimensional uncertainty propagation,”Journal of Computational Physics, vol. 321, pp. 191–223, Sep. 2016,ISSN: 0021-9991.DOI:10.1016/j.jcp.2016.05.039Accessed: Oct. 27, 2025. [Online]. Available:https://www.sciencedirect.com/science/a...

    work page doi:10.1016/j.jcp.2016.05.039accessed: 2016

  57. [57]

    A. G. Wilson, Z. Hu, R. Salakhutdinov, and E. P . Xing,Deep Kernel Learning, arXiv:1511.02222 [cs], Nov. 2015.DOI:10.48550/ arXiv.1511. 02222Accessed: Apr. 16, 2025. [Online]. Available:http: //arxiv.org/abs/1511.02222

    Pith/arXiv arXiv 2015

  58. [58]

    Bayesian Analysis,

    J. O. Berger, “Bayesian Analysis,” en, inStatistical Decision Theory and Bayesian Analysis, J. O. Berger, Ed., New York, NY: Springer, 1985, pp. 118–307,ISBN: 978-1-4757-4286-2.DOI:10.1007/978-1-4757- 4286- 2_4Accessed: Nov. 3, 2025. [Online]. Available:https://doi.org/10.1007/978- 1- 4757- 4286-2_4

    doi:10.1007/978-1-4757- 1985

  59. [59]

    Sparse Gaussian process hyperparam- eters: Optimize or integrate?

    V . Lalchand, W. P . Bruinsma, D. R. Burt, and C. E. Rasmussen, “Sparse Gaussian process hyperparam- eters: Optimize or integrate?” InProceedings of the 36th International Conference on Neural Information Processing Systems, ser. NIPS ’22, Red Hook, NY, USA: Curran Associates Inc., Nov. 2022, pp. 16 612– 16 623,ISBN: 978-1-71387-108-8. Accessed: Oct. 27, 2025

    2022

  60. [60]

    M. L. Stein,Interpolation of Spatial Data(Springer Series in Statistics). New York, NY: Springer, 1999, ISBN: 978-1-4612-7166-6 978-1-4612-1494-6.DOI:10.1007/ 978- 1- 4612- 1494- 6Accessed: Oct. 27,

    1999

  61. [61]

    Available:http://link.springer.com/10.1007/978-1-4612-1494-6

    [Online]. Available:http://link.springer.com/10.1007/978-1-4612-1494-6

    doi:10.1007/978-1-4612-1494-6

  62. [62]

    Practical Bayesian Optimization of Machine Learning Al- gorithms,

    J. Snoek, H. Larochelle, and R. P . Adams, “Practical Bayesian Optimization of Machine Learning Al- gorithms,” inAdvances in Neural Information Processing Systems, vol. 25, Curran Associates, Inc., 2012. Accessed: Nov. 3, 2025. [Online]. Available:https://papers.nips.cc/paper_files/paper/2012/ hash/05311655a15b75fab86956663e1819cd-Abstract.html

    arXiv 2012

  63. [63]

    Spatial Modelling Using a New Class of Nonstationary Covariance Functions,

    C. J. Paciorek and M. J. Schervish, “Spatial Modelling Using a New Class of Nonstationary Covariance Functions,”Environmetrics, vol. 17, no. 5, pp. 483–506, 2006,ISSN: 1180-4009.DOI:10 . 1002 / env . 785Accessed: Oct. 27, 2025. [Online]. Available:https : / / pmc . ncbi . nlm . nih . gov / articles / PMC2157553/

    2006

  64. [64]

    Kındap and S

    Y. Kındap and S. Godsill,Non-Gaussian Process Regression, arXiv:2209.03117 [stat], Sep. 2022.DOI:10. 48550 /arXiv .2209 .03117Accessed: Oct. 27, 2025. [Online]. Available:http: / /arxiv .org /abs / 2209.03117

    Pith/arXiv arXiv 2022

  65. [65]

    Deep Gaussian Processes,

    A. Damianou and N. D. Lawrence, “Deep Gaussian Processes,” en, inProceedings of the Sixteenth International Conference on Artificial Intelligence and Statistics, PMLR, Apr. 2013, pp. 207–215. Accessed: Jun. 20, 2025. [Online]. Available:https://proceedings.mlr.press/v31/damianou13a.html

    2013

  66. [66]

    Sparse Greedy Matrix Approximation for Machine Learning,

    A. J. Smola and B. Schökopf, “Sparse Greedy Matrix Approximation for Machine Learning,” inPro- ceedings of the Seventeenth International Conference on Machine Learning, ser. ICML ’00, San Francisco, CA, USA: Morgan Kaufmann Publishers Inc., Jun. 2000, pp. 911–918,ISBN: 978-1-55860-707-1. Ac- cessed: Mar. 13, 2025. 26

    2000

  67. [67]

    Using the Nyström Method to Speed Up Kernel Machines,

    C. Williams and M. Seeger, “Using the Nyström Method to Speed Up Kernel Machines,” inAdvances in Neural Information Processing Systems, vol. 13, MIT Press, 2000. Accessed: Mar. 13, 2025. [Online]. Available:https://papers.nips.cc/paper_files/paper/2000/hash/19de10adbaa1b2ee13f77f679fa1483a- Abstract.html

    2000

  68. [68]

    Variational Learning of Inducing Variables in Sparse Gaussian Processes,

    M. Titsias, “Variational Learning of Inducing Variables in Sparse Gaussian Processes,” en, inPro- ceedings of the Twelfth International Conference on Artificial Intelligence and Statistics, PMLR, Apr. 2009, pp. 567–574. Accessed: Jun. 23, 2025. [Online]. Available:https : / / proceedings . mlr . press / v5 / titsias09a.html

    2009

  69. [69]

    Efficient SVM Training Using Low-Rank Kernel Representations,

    S. Fine and K. Scheinberg, “Efficient SVM Training Using Low-Rank Kernel Representations,”Jour- nal of Machine Learning Research, vol. 2, no. Dec, pp. 243–264, 2001,ISSN: ISSN 1533-7928. Accessed: Mar. 13, 2025. [Online]. Available:https://www.jmlr.org/papers/v2/fine01a.html

    2001

  70. [70]

    Kernel Interpolation for Scalable Structured Gaussian Processes (KISS- GP),

    A. Wilson and H. Nickisch, “Kernel Interpolation for Scalable Structured Gaussian Processes (KISS- GP),” en, inProceedings of the 32nd International Conference on Machine Learning, PMLR, Jun. 2015, pp. 1775–1784. Accessed: Apr. 30, 2025. [Online]. Available:https://proceedings.mlr.press/v37/ wilson15.html

    2015

  71. [71]

    Random Features for Large-Scale Kernel Machines,

    A. Rahimi and B. Recht, “Random Features for Large-Scale Kernel Machines,” inAdvances in Neural Information Processing Systems, vol. 20, Curran Associates, Inc., 2007. Accessed: Jun. 23, 2025. [Online]. Available:https://papers.nips.cc/paper_files/paper/2007/hash/013a006f03dbc5392effeb8f18fda755- Abstract.html

    2007

  72. [72]

    Scalable Kernel Methods via Doubly Stochastic Gradients,

    B. Dai et al., “Scalable Kernel Methods via Doubly Stochastic Gradients,” inAdvances in Neural In- formation Processing Systems, vol. 27, Curran Associates, Inc., 2014. Accessed: Dec. 20, 2024. [Online]. Available:https://proceedings.neurips.cc/paper_files/paper/2014/hash/95d309f0b035d97f69902e7972c2b2e6- Abstract.html

    2014

  73. [73]

    Multi-fidelity Gaussian process regression for computer experiments,

    L. L. Gratiet, “Multi-fidelity Gaussian process regression for computer experiments,” fr, Ph.D. dis- sertation, Université Paris-Diderot - Paris VII, Oct. 2013. Accessed: Oct. 28, 2024. [Online]. Available: https://theses.hal.science/tel-00866770

    2013

  74. [74]

    Multi-Fidelity for MDO Using Gaussian Pro- cesses,

    N. Garland, R. Le Riche, Y. Richet, and N. Durrande, “Multi-Fidelity for MDO Using Gaussian Pro- cesses,” en, inAerospace System Analysis and Optimization in Uncertainty, L. Brevault, M. Balesdent, and J. Morio, Eds., Cham: Springer International Publishing, 2020, pp. 295–320,ISBN: 978-3-030-39126-3. DOI:10.1007/978-3-030-39126-3_8Accessed: Mar. 5, 2026. ...

    work page doi:10.1007/978-3-030-39126-3_8accessed: 2020

  75. [75]

    Raissi and G

    M. Raissi and G. Karniadakis,Deep Multi-fidelity Gaussian Processes, arXiv:1604.07484 [cs], Apr. 2016. DOI:10.48550/arXiv.1604.07484Accessed: Dec. 31, 2024. [Online]. Available:http://arxiv.org/ abs/1604.07484

    Pith/arXiv arXiv doi:10.48550/arxiv.1604.07484accessed: 2016

  76. [76]

    A combined modeling method for complex multi-fidelity data fusion,

    L. Tang et al., “A combined modeling method for complex multi-fidelity data fusion,” en,Machine Learning: Science and Technology, vol. 5, no. 3, p. 035 071, Sep. 2024,ISSN: 2632-2153.DOI:10 . 1088 / 2632- 2153/ad718fAccessed: Nov. 18, 2024. [Online]. Available:https://dx.doi.org/10.1088/ 2632-2153/ad718f

    2024

  77. [77]

    A generalized hierarchical co-Kriging model for multi- fidelity data fusion,

    Q. Zhou, Y. Wu, Z. Guo, J. Hu, and P . Jin, “A generalized hierarchical co-Kriging model for multi- fidelity data fusion,” en,Structural and Multidisciplinary Optimization, vol. 62, no. 4, pp. 1885–1904, Oct. 2020,ISSN: 1615-1488.DOI:10.1007/s00158- 020- 02583- 7Accessed: Nov. 18, 2024. [Online]. Available:https://doi.org/10.1007/s00158-020-02583-7 27

    doi:10.1007/s00158- 1904

  78. [78]

    A sequential multi-fidelity metamodeling approach for data regression,

    Q. Zhou, Y. Wang, S.-K. Choi, P . Jiang, X. Shao, and J. Hu, “A sequential multi-fidelity metamodeling approach for data regression,”Knowledge-Based Systems, vol. 134, pp. 199–212, Oct. 2017,ISSN: 0950- 7051.DOI:10.1016/j.knosys.2017.07.033Accessed: Nov. 18, 2024. [Online]. Available:https: //www.sciencedirect.com/science/article/pii/S0950705117303556

    work page doi:10.1016/j.knosys.2017.07.033accessed: 2017

  79. [79]

    Gaussian process fusion method for multi-fidelity data with heterogeneity distribution in aerospace vehicle flight dynamics,

    B. Yang, B. Chen, Y. Liu, and J. Chen, “Gaussian process fusion method for multi-fidelity data with heterogeneity distribution in aerospace vehicle flight dynamics,”Engineering Applications of Artificial Intelligence, vol. 138, p. 109 228, Dec. 2024,ISSN: 0952-1976.DOI:10.1016/j.engappai.2024.109228 Accessed: Nov. 18, 2024. [Online]. Available:https://www...

    work page doi:10.1016/j.engappai.2024.109228 2024

  80. [80]

    Nonlinear information fusion algorithms for data-efficient multi-fidelity modelling,

    P . Perdikaris, M. Raissi, A. Damianou, N. D. Lawrence, and G. E. Karniadakis, “Nonlinear information fusion algorithms for data-efficient multi-fidelity modelling,”Proceedings of the Royal Society A: Math- ematical, Physical and Engineering Sciences, vol. 473, no. 2198, p. 20 160 751, Feb. 2017.DOI:10.1098/ rspa.2016.0751Accessed: Nov. 18, 2024. [Online]...

    doi:10.1098/rspa.2016.0751 2017

Showing first 80 references.