Pith. sign in

REVIEW 4 major objections 5 minor 79 references

Multi-fidelity Bayesian Data-Driven Design of Energy Absorbing Spinodoid Cellular Structures

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Multi-fidelity Bayesian optimization finds spinodoid designs with up to 11% more energy absorption than single-fidelity BO under the same budget.

desk verdict The 11% MFBO advantage likely comes from an uncharged high-fidelity evaluation in the recommender step, not from multi-fidelity transfer learning. read the letter →

arxiv 2507.22079 v1 pith:D7Q43ONV submitted 2025-07-25 cs.LG cond-mat.mtrl-sci

Leo Guo , Hirak Kansara , Siamak F. Khosroshahi , GuoQi Zhang , Wei Tan This is my paper
classification cs.LGcond-mat.mtrl-sci
keywords multi-fidelityBayesianoptimizationspinodoidcellularstructuresenergyabsorptionSobolsensitivityanalysisGaussianprocesssurrogatefiniteelementsimulationdata-drivendesignmetamaterial
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

This paper is a head-to-head test: when the goal is maximizing the energy absorption of spinodoid cellular structures, architected lattice-like materials, does multi-fidelity Bayesian optimization (MFBO) actually beat ordinary single-fidelity Bayesian optimization (BO) under the same computational budget? The authors claim yes, with the best MFBO configuration reaching 11% higher normalized energy absorption than the best BO configuration, and an average gain of 8.6% over the four kernel-acquisition combinations. The design space is first pruned with Sobol' sensitivity analysis, a variance-based method that shows relative density dominates energy absorption, leaving three angular parameters to optimize. If the claim holds, cheap coarse-mesh finite-element simulations can serve as training signal for expensive design optimization, which would matter for any simulation-heavy engineering design problem.

What carries the argument

The load-bearing mechanism is the multi-task Gaussian process (MTGP) surrogate, whose covariance between fidelity levels is written as the product of a single design kernel and a positive semi-definite fidelity correlation matrix, $K = K_{\mathrm{Had}} \odot B$. Paired with variable-fidelity acquisition functions that multiply expected improvement or upper confidence bound by the cost ratio and the inter-fidelity correlation, this lets the optimizer learn from cheap coarse-mesh (resolution 20) simulations while ultimately optimizing the fine-mesh (resolution 30) objective. Sobol' and Saltelli sampling supply both the sensitivity indices that reduce the design space and the initial design of experiments.

What would settle it

Re-run the same budget-matched BO-versus-MFBO comparison after removing the low-fidelity plateau artifact, for example by using a refined coarse mesh or a different meshing routine, and check whether the 8-11% advantage persists; alternatively, repeat each configuration many times with different random seeds and see whether the gap exceeds run-to-run noise.

Watch

Extended reading notes

Core claim

The paper reports that multi-task Gaussian-process Bayesian optimization, a multi-fidelity method that combines a design kernel with a fidelity correlation matrix, outperforms single-fidelity BO when both are given the same total budget of high-fidelity evaluations on a three-parameter spinodoid energy-absorption problem. The best MFBO configuration (logarithmic expected improvement with an RBF kernel) attains a normalized energy absorption of 0.044007, while the best BO configuration (upper confidence bound with an RBF kernel) attains 0.039517, an 11% improvement; the average gain over the four kernel-acquisition combinations is 8.6%. The authors also show that sensitivity analysis over the full four-dimensional design space identifies the density parameter as dominant, which justifies fixing it and optimizing the three angles only. The recommended designs shift from strictly columnar to slightly bending-dominated geometries, which the authors interpret as a sign that the multi-fidelity surrogate explored a different and better region of the angular design space.

Load-bearing premise

The whole advantage rests on the coarse-mesh, low-fidelity simulations being a useful guide to the fine-mesh, high-fidelity behavior, despite their measured correlation of about 0.68 and the numerical artifact that flattens many low-fidelity values.

Editorial extensions

If this is right

  • A designer can spend the same total simulation budget and end up with a spinodoid structure that absorbs more energy, because low-fidelity evaluations are roughly nine times cheaper and still informative.
  • The recommended designs shift from strictly columnar geometries to slightly bending-dominated ones, suggesting the multi-fidelity surrogate explored a different and better region of the angular design space.
  • The sensitivity-analysis-first workflow is reusable: it tells the designer which parameters matter before any optimization begins, reducing the dimensionality of the expensive problem.
  • Because the code and data are open source, the same MFBO pipeline can be applied to other expensive finite-element design problems without re-deriving the machinery.

Reading between the lines

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

  • The reported 8-11% gain is measured against a low-fidelity model with a Pearson correlation of about 0.68 and a numerical plateau artifact; how much of the gain transfers to other simulation pairs is an open question.
  • The comparison runs each configuration once, so without repeated restarts the 11% figure should be read as a point estimate, not as a statistically verified gap.
  • A natural next test is to extend the same MFBO setup to the two-objective problem the authors mention, maximizing energy absorption while minimizing peak force, by scalarizing or using a trust function.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. This paper proposes a data-driven optimization framework for designing spinodoid cellular structures that maximize normalized energy absorption under compressive loading. The framework combines Sobol'/Saltelli sampling with variance-based sensitivity analysis to reduce the 4D design space to 3D (fixing relative density ρ = 0.3), and then compares single-fidelity Bayesian optimization (BO) with multi-fidelity Bayesian optimization (MFBO) using Gaussian-process surrogates at two FE mesh resolutions. The authors report that MFBO outperforms BO by up to 11% in normalized EA across the tested kernel/acquisition configurations, and they release code and data openly. The methodological exposition of sensitivity analysis and Gaussian-process regression is standard, but the empirical comparison supporting the headline claim has several weaknesses that need to be addressed.

Significance. If the headline result holds, the paper would provide a useful application-oriented demonstration that multi-fidelity BO can reduce the computational cost of FE-based metamaterial design, which is a practically relevant contribution. The paper also gives a clear, self-contained treatment of Sobol' sensitivity analysis and multi-task Gaussian processes, and the open-source release is a strength. However, the current evidence is not yet sufficient: the central comparison may not charge all high-fidelity evaluations to the stated budget, the experiments are single runs without error bars, and the reported advantage is not robust across matched settings. These issues are load-bearing for the abstract's claim rather than cosmetic.

major comments (4)
  1. [§2.3, Algorithm 2, and Table 3] The budget accounting for the final recommendation in MFBO is not specified. Algorithm 2's loop only evaluates the fidelity selected by the acquisition function, so a low-fidelity incumbent x*_1 is not necessarily evaluated at high fidelity during the optimization. Yet the recommender described after Algorithm 2 returns fM(x*_1), which requires an additional high-fidelity FE evaluation that may occur after the stated 50-high-fidelity-equivalent budget. The paper should state explicitly whether these final evaluations are charged to the budget, and if so, how; if they are not, the comparison with BO is unfair because MFBO receives free look-ahead by reading high-fidelity values of cheap low-fidelity points. Please report the number of extra high-fidelity runs needed for each MFBO configuration and include them in the cost totals.
  2. [§4.1, §4.2, Tables 2–3, Figures 13–15] Each BO and MFBO configuration appears to have been run once, with no repeated seeds, initializations, or error bars. The statement in Section 4.2 that multi-task Bayesian optimization is 'more reliable' is not supported by a single trajectory per configuration. The authors should either run multiple repetitions (at least for a subset of configurations) with different initial Sobol' subsets or random restarts and report means/variances, or clearly state that no variance information is available and temper the reliability claim accordingly.
  3. [Abstract and §4.2] The headline 'up to 11%' compares the best MFBO configuration (LogEI/RBF, EA 0.044007 in Table 3) with the best BO configuration (UCB/RBF, EA 0.039517 in Table 2), which are not matched in kernel or acquisition function. The matched-pair average of 8.6% is more relevant, and it includes a negative case (UCB/RBF: MFBO 0.038878 vs BO 0.039517), so the claim that MFBO 'is able to outperform BO' across settings should be reformulated to present both the best-case and the matched-pair result, including the variance across configurations.
  4. [§3.3, Figure 12, and §4.2] The effectiveness of MFBO rests on the assumption that the coarse-mesh simulations are a useful proxy for the fine-mesh simulations. The paper itself reports only a Pearson correlation of approximately 0.68 and a numerical artifact that produces identical low-fidelity EA values (around 0.018) for a range of θ designs. Given that the VF-EI acquisition function multiplies the expected improvement by this correlation, the paper should provide evidence that low-fidelity evaluations improve rather than mislead the search. For example, report the fraction of low-fidelity acquisitions selected during each MFBO run and compare against a BO baseline that uses the same total computational budget without low-fidelity data.
minor comments (5)
  1. [Abstract and §1] The abstract contains a grammatical error: 'when faced with with a real-life engineering problem' should read 'when faced with a real-life engineering problem.'
  2. [§2.3, Algorithm 2] The manuscript uses both 'MFBO' and 'MTBO' to refer to the same method; for consistency, one acronym should be used throughout.
  3. [§3.2, Figure 7] The text states that the low-fidelity objective is evaluated on the same 96 design samples plus an additional 288 samples, 'for a total of 386 samples.' However, 96 + 288 = 384; please correct the total or the individual counts.
  4. [§4.2, Eq. (34)–(35)] Please specify how the GP hyperparameters and the multi-task correlation matrix B are optimized in the implementations used for Tables 2 and 3, including any bounds or optimizer settings, so that the open-source results are reproducible.
  5. [§2.1, Eq. (7)–(8)] The notation for the Saltelli estimators is mostly clear, but the definition of the bootstrap confidence intervals used in Figures 9 and 11 is not given; a brief sentence or reference for the bootstrap procedure would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the MFBO-vs-BO comparison is empirical and self-contained; self-citations are not load-bearing.

full rationale

The paper's central claim is an empirical comparison between single-fidelity BO and multi-fidelity BO on a spinodoid energy-absorption problem. The MFBO surrogate equations (multi-task GP, VF-EI, VF-UCB) are taken from the literature, not derived from the paper's own results. The low-fidelity/high-fidelity correlation (Pearson ≈ 0.68) is measured from Saltelli samples, not assumed into the outcome. Sobol' sensitivity indices are computed from simulations, and the 4D-to-3D dimension reduction is justified by those measured indices. The final BO/MFBO optima are obtained by running the respective algorithms and reporting the recommended designs; there is no equation-level identity between an input and an output. The cited self-work [72] is used for context about spinodoid structures and for a related multi-objective study, but it does not carry the MFBO comparison or any uniqueness claim, so it is not load-bearing. A possible concern about whether the MFBO recommender's final high-fidelity evaluations are charged to the stated budget would be a fairness/correctness issue, not a circularity issue, because no fitted parameter or defined quantity is being renamed as a prediction.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claim of MFBO superiority rests primarily on measured quantities (0.68 correlation, 0.11 cost ratio) and hand-set protocol choices (rho=0.3, budgets of 160 and 50 high-fidelity equivalents, sample counts). All mathematical machinery is standard GP/BO/Sobol theory. No new physical entities are introduced.

free parameters (7)
  • GP prior hyperparameters (RBF and Matern: amplitude c, lengthscale lambda, noise s^2) = Not reported in text; estimated by MLE
    Standard GP model parameters fit to the design of experiments via Equation (16); the optimization trajectories depend on these fits.
  • Multi-task GP cross-fidelity correlation matrix B = Not reported; estimated from data
    The multi-task covariance kernel (Equation 30) uses a positive semi-definite B estimated from data; the quality of knowledge transfer depends on it.
  • UCB and VF-UCB acquisition hyperparameters (beta, omega1, omega2) = Not reported in text
    Equations (21) and (35) use beta and weights omega1, omega2; their values are not stated, which affects reproducibility and comparison fairness.
  • Fixed relative density rho = 0.3 = 0.3
    Chosen by hand at the lower bound 'for mass scaling reasons' (Section 3.3), not from the objective; the 3D optimization and the 11% claim apply only to this slice.
  • Computational budgets: initial DoE 160 high-fidelity equivalents, optimization 50 high-fidelity equivalents = 160 and 50
    Hand-set protocol values that define the comparison; the MFBO advantage is measured under this specific budget.
  • Cost ratio CR between low- and high-fidelity simulations = 0.11
    Median coarse-mesh simulation time 251 s versus fine-mesh 2300 s (Figure 7); used in the variable-fidelity acquisition functions (Equations 34 and 35).
  • Saltelli sample counts for sensitivity analysis = 96 high-fidelity and 288 low-fidelity (4D); 980 per fidelity (3D)
    Sample sizes chosen based on stabilization of sensitivity indices; they determine the sensitivity conclusions and the initial DoE.
assumptions (6)
  • domain assumption FE simulations with elastic-plastic material behavior and one-dimensional quasi-static compression accurately represent the energy absorption of the physical spinodoid structure
    No experimental validation is performed; the optimization targets simulated EA only, as acknowledged in Section 5 (future work).
  • domain assumption The low-fidelity coarse-mesh model is a useful proxy for the high-fidelity model despite Pearson correlation 0.68 and a numerical artifact cluster
    MFBO transfer learning relies on this correlation (Section 3.3, Figure 12).
  • standard math Standard Gaussian process and Bayesian optimization mathematics, including posterior formulas, Sobol decomposition, and Saltelli estimators
    Equations (7) to (33) are taken from the GP and BO literature; the paper does not derive them.
  • domain assumption Normalizing energy absorption by a solid-material reference EAs is the appropriate design objective
    Equation (36) defines the objective; the normalization choice affects the optimum and the comparison.
  • domain assumption The design space is sufficiently described by the four parameters rho, theta1, theta2, theta3
    Table 1 defines the search space; graded structures and other spinodoid parameters are excluded.
  • standard math The MLE point estimate is a sufficient approximation to the posterior predictive distribution
    Equation (17) uses theta_MLE instead of marginalizing over theta, a known approximation in GP regression.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-fidelity Bayesian Data-Driven Design of Energy Absorbing Spinodoid Cellular Structures." pith.science (2026). https://pith.science/paper/D7Q43ONV

@misc{pith2026250722079,
  author       = {Pith},
  title        = {Pith review of: Multi-fidelity Bayesian Data-Driven Design of Energy Absorbing Spinodoid Cellular Structures},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D7Q43ONV}},
  note         = {Machine review of arXiv:2507.22079}
}
read the original abstract

Finite element (FE) simulations of structures and materials are getting increasingly more accurate, but also more computationally expensive as a collateral result. This development happens in parallel with a growing demand of data-driven design. To reconcile the two, a robust and data-efficient optimization method called Bayesian optimization (BO) has been previously established as a technique to optimize expensive objective functions. In parallel, the mesh width of an FE model can be exploited to evaluate an objective at a lower or higher fidelity (cost & accuracy) level. The multi-fidelity setting applied to BO, called multi-fidelity BO (MFBO), has also seen previous success. However, BO and MFBO have not seen a direct comparison with when faced with with a real-life engineering problem, such as metamaterial design for deformation and absorption qualities. Moreover, sampling quality and assessing design parameter sensitivity is often an underrepresented part of data-driven design. This paper aims to address these shortcomings by employing Sobol' samples with variance-based sensitivity analysis in order to reduce design problem complexity. Furthermore, this work describes, implements, applies and compares the performance BO with that MFBO when maximizing the energy absorption (EA) problem of spinodoid cellular structures is concerned. The findings show that MFBO is an effective way to maximize the EA of a spinodoid structure and is able to outperform BO by up to 11% across various hyperparameter settings. The results, which are made open-source, serve to support the utility of multi-fidelity techniques across expensive data-driven design problems.

Figures

Figures reproduced from arXiv: 2507.22079 by the authors.

Figure 1
Figure 1. Flowchart diagram of BO. BO can be flexibly adapted into settings with data that adhere to particular input or output structures. Multi-fidelity BO (MFBO), which relies on surro￾gate models that handle the cost-accuracy trade-off between low- and high-fidelity data, allows for accelerated convergence by leveraging inexpensive approximations alongside selective high-fidelity evaluations. This approach enables more co… view at source ↗
Figure 2
Figure 2. A schematic construction of A (i) B given A and B. Finally, let A (i) B := (c (i)⊤ 1 , . . . , c (i)⊤ N ) ⊤ be decomposed into rows. Then, the following approximations hold: Var(E(Y |Xi)) ≈ 1 N X N j=1 f(bj )(f(c (i) j ) − f(aj )), (7) E(Var(Y |X∼i)) ≈ 1 2N X N j=1 (f(aj ) − f(c (i) j ))2 . (8) Moreover, the statistical confidence of these approximations can be numerically tested by means of bootstrapping [35, 37]. … view at source ↗
Figure 3
Figure 3. Saltelli sampling on a unit square with 32, 64 and 128 subsequent samples (different [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Flowchart diagram of the Bayesian data driven framework which includes BO (Figure [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Varying levels of ρ ∈ [0.3, 0.6], increasing from left to right, for fixed values θ1 = 90◦ , θ2 = 0◦ and θ3 = 0◦ . 3.1. Objective As the one-dimensional compressive loading simulation is performed, a force P is applied across a displacement x until a threshold δmax is …
Figure 6
Figure 6. Figure 6: Two mesh resolutions for fixed values ρ = 0.3, θ1 = 15◦ , θ2 = 15◦ and θ3 = 0◦ . The coarse mesh represents the low-fidelity model, while the fine mesh is considered high-fidelity. Intuitively, a low value for θ3 arranges the structure into columns aligned into the loa…
Figure 7
Figure 7. Figure 7: Simulation times for mesh resolution 20 (left) and 30 (right). The median simulation [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]
Figure 8
Figure 8. Figure 8: The collection of Saltelli samples used in Figure [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]
Figure 9
Figure 9. Figure 9: First (S1) and total (ST) order Sobol’ sensitivity indices and corresponding boot [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]
Figure 10
Figure 10. Figure 10: The collection of 980 Saltelli samples (each fidelity) and their normalized EA value [PITH_FULL_IMAGE:figures/full_fig_p022_10.png]
Figure 11
Figure 11. Figure 11: First (S1) and total (ST) order Sobol’ sensitivity indices of the 3D EA objective (fixed [PITH_FULL_IMAGE:figures/full_fig_p023_11.png]
Figure 12
Figure 12. Figure 12: Low- and high-fidelity EA values at the coincident 3D Saltelli design samples used in [PITH_FULL_IMAGE:figures/full_fig_p024_12.png]
Figure 13
Figure 13. Figure 13: A number of standard BO histories (colored dots) and each cumulative maximum [PITH_FULL_IMAGE:figures/full_fig_p025_13.png]
Figure 14
Figure 14. Figure 14: A number of EA optimization histories with multi-task Bayesian optimization. The [PITH_FULL_IMAGE:figures/full_fig_p026_14.png]
Figure 15
Figure 15. Figure 15: The EA-optimal spinodoid structures and θ-values (Tables 2 and 3) found by the best-performing BO (top right) and the best MTBO (top left) schemes, superimposed with the optimization histories. Low-fidelity EA evaluations during the MTBO run are indicated with the sma…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

79 extracted references · 66 canonical work pages

  1. [1]

    Y. Wu, J. Fang, C. Wu, C. Li, G. Sun, Q. Li, Additively manufactured ma- terials and structures: A state-of-the-art review on their mechanical charac- teristics and energy absorption, International Journal of Mechanical Sciences 246 (2023) 108102

    work page 2023

  2. [2]

    Harris, G

    J. Harris, G. McShane, Impact response of metallic stacked origami cellular materials, International Journal of Impact Engineering 147 (2021) 103730

    work page 2021

  3. [3]

    S. C. Han, J. W. Lee, K. Kang, A new type of low density material: Shellular., Advanced Materials (Deerfield Beach, Fla.) 27 (37) (2015) 5506–5511

    work page 2015

  4. [4]

    Guell Izard, J

    A. Guell Izard, J. Bauer, C. Crook, V. Turlo, L. Valdevit, Ultrahigh energy absorption multifunctional spinodal nanoarchitectures, Small 15 (45) (2019) 1903834. 30

    work page 2019

  5. [5]

    Al-Ketan, R

    O. Al-Ketan, R. K. Abu Al-Rub, Multifunctional mechanical metamateri- als based on triply periodic minimal surface lattices, Advanced Engineering Materials 21 (10) (2019) 1900524

    work page 2019

  6. [6]

    J. W. Cahn, On spinodal decomposition, Acta metallurgica 9 (9) (1961) 795– 801

    work page 1961

  7. [7]

    M. P. Bendsoe, O. Sigmund, Topology optimization: theory, methods, and applications, Springer Science & Business Media, 2013

    work page 2013

  8. [8]

    Zhang, M.-T

    Y. Zhang, M.-T. Hsieh, L. Valdevit, Mechanical performance of 3d printed interpenetrating phase composites with spinodal topologies, Composite Struc- tures 263 (2021) 113693

    work page 2021

Show all 79 references
  1. [9]

    Y. Guo, S. Sharma, S. Kumar, Inverse designing surface curvatures by deep learning, Advanced Intelligent Systems (2024) 2300789

    2024

  2. [10]

    X. Yu, J. Zhou, H. Liang, Z. Jiang, L. Wu, Mechanical metamaterials asso- ciated with stiffness, rigidity and compressibility: A brief review, Progress in Materials Science 94 (2018) 114–173

    2018

  3. [11]

    M. A. Bessa, P. Glowacki, M. Houlder, Bayesian machine learning in meta- material design: Fragile becomes supercompressible, Advanced Materials 31 (48) (2019) 1904845. arXiv:https://onlinelibrary.wiley.com/doi/ pdf/10.1002/adma.201904845, doi:10.1002/adma.201904845. URL https:...

    2019 doi

  4. [12]

    D. Shin, A. Cupertino, M. H. de Jong, P. G. Steeneken, M. A. Bessa, R. A. Norte, Spiderweb nanomechanical resonators via bayesian optimiza- tion: inspired by nature and guided by machine learning, arXiv preprint arXiv:2108.04809 (2021)

    2021 arXiv

  5. [13]

    Kuszczak, F

    I. Kuszczak, F. Azam, M. Bessa, P. Tan, F. Bosi, Bayesian optimisation of hexagonal honeycomb metamaterial, Extreme Mechanics Letters 64 (2023) 102078

    2023

  6. [14]

    Raßloff, P

    A. Raßloff, P. Seibert, K. A. Kalina, M. K¨ astner, Inverse design of spinodoid structures using bayesian optimization, Computational Mechanics (2025) 1– 22

    2025

  7. [15]

    Marco, F

    A. Marco, F. Berkenkamp, P. Hennig, A. P. Schoellig, A. Krause, S. Schaal, S. Trimpe, Virtual vs. real: Trading off simulations and physical experiments 31 in reinforcement learning with bayesian optimization, in: 2017 IEEE Inter- national Conference on Robotics and Automation...

    2017

  8. [16]

    G. Pang, P. Perdikaris, W. Cai, G. E. Karniadakis, Discovering variable frac- tional orders of advection–dispersion equations from field data using multi- fidelity bayesian optimization, Journal of Computational Physics 348 (2017) 694–714

    2017

  9. [17]

    Perdikaris, G

    P. Perdikaris, G. E. Karniadakis, Model inversion via multi-fidelity bayesian optimization: a new paradigm for parameter estimation in haemodynamics, and beyond, Journal of The Royal Society Interface 13 (118) (2016) 20151107

    2016

  10. [18]

    Imani, S

    M. Imani, S. F. Ghoreishi, D. Allaire, U. M. Braga-Neto, Mfbo-ssm: Multi- fidelity bayesian optimization for fast inference in state-space models, in: Pro- ceedings of the AAAI Conference on Artificial Intelligence, Vol. 33, 2019, pp. 7858–7865

    2019

  11. [19]

    Q. Wu, H. Wang, W. Hong, Multistage collaborative machine learning and its application to antenna modeling and optimization, IEEE Transactions on Antennas and Propagation 68 (5) (2020) 3397–3409

    2020

  12. [20]

    Charayron, T

    R. Charayron, T. Lefebvre, N. Bartoli, J. Morlier, Towards a multi-fidelity & multi-objective bayesian optimization efficient algorithm, Aerospace Science and Technology 142 (2023) 108673

    2023

  13. [21]

    Ariyarit, M

    A. Ariyarit, M. Sugiura, Y. Tanabe, M. Kanazaki, Hybrid surrogate-model- based multi-fidelity efficient global optimization applied to helicopter blade design, Engineering Optimization 50 (6) (2018) 1016–1040

    2018

  14. [22]

    M. Tyan, N. V. Nguyen, J.-W. Lee, Improving variable-fidelity modelling by exploring global design space and radial basis function networks for aerofoil design, Engineering Optimization 47 (7) (2015) 885–908

    2015

  15. [23]

    J. Kou, W. Zhang, Multi-fidelity modeling framework for nonlinear unsteady aerodynamics of airfoils, Applied Mathematical Modelling 76 (2019) 832–855

    2019

  16. [24]

    X. Peng, J. Kou, W. Zhang, Multi-fidelity nonlinear unsteady aerodynamic modeling and uncertainty estimation based on hierarchical kriging, Applied Mathematical Modelling 122 (2023) 1–21

    2023

  17. [25]

    A. Tran, T. Wildey, S. McCann, smf-bo-2cogp: A sequential multi-fidelity constrained bayesian optimization framework for design applications, Journal of Computing and Information Science in Engineering 20 (3) (2020). 32

    2020

  18. [26]

    M. Aziz, W. Hare, M. Jaberipour, Y. Lucet, Multi-fidelity algorithms for the horizontal alignment problem in road design, Engineering Optimization 52 (11) (2020) 1848–1867

    2020

  19. [27]

    Zhang, W

    S. Zhang, W. Lyu, F. Yang, C. Yan, D. Zhou, X. Zeng, X. Hu, An efficient multi-fidelity bayesian optimization approach for analog circuit synthesis, in: 2019 56th ACM/IEEE Design Automation Conference (DAC), IEEE, 2019, pp. 1–6

    2019

  20. [28]

    S. Li, W. Xing, R. Kirby, S. Zhe, Multi-fidelity bayesian optimization via deep neural networks, Advances in Neural Information Processing Systems 33 (2020)

    2020

  21. [29]

    Cupertino, D

    A. Cupertino, D. Shin, L. Guo, P. G. Steeneken, M. A. Bessa, R. A. Norte, Centimeter-scale nanomechanical resonators with low dissipation, arXiv preprint arXiv:2308.00611 (2023)

    2023 arXiv

  22. [30]

    Lenhart, K

    T. Lenhart, K. Eckhardt, N. Fohrer, H.-G. Frede, Comparison of two different approaches of sensitivity analysis, Physics and Chemistry of the Earth, Parts A/B/C 27 (9-10) (2002) 645–654

    2002

  23. [31]

    Baucells, E

    M. Baucells, E. Borgonovo, Invariant probabilistic sensitivity analysis, Man- agement Science 59 (11) (2013) 2536–2549

    2013

  24. [32]

    Borgonovo, E

    E. Borgonovo, E. Plischke, Sensitivity analysis: A review of recent advances, European Journal of Operational Research 248 (3) (2016) 869–887

    2016

  25. [33]

    I. M. Sobol, Global sensitivity indices for nonlinear mathematical models and their monte carlo estimates, Mathematics and computers in simulation 55 (1-

  26. [34]

    M. M. Opgenoord, D. L. Allaire, K. E. Willcox, Variance-based sensitivity analysis to support simulation-based design under uncertainty, Journal of Me- chanical Design 138 (11) (2016) 111410

    2016

  27. [35]

    Marelli, C

    S. Marelli, C. Lamas, K. Konakli, C. Mylonas, P. Wiederkehr, B. Sudret, Uqlab user manual–sensitivity analysis, Report UQLab-V1 (2019) 2–106

    2019

  28. [36]

    Saltelli, P

    A. Saltelli, P. Annoni, I. Azzini, F. Campolongo, M. Ratto, S. Tarantola, Variance based sensitivity analysis of model output. design and estimator for the total sensitivity index, Computer physics communications 181 (2) (2010) 259–270. 33

    2010

  29. [37]

    Efron, Bootstrap methods: another look at the jackknife, in: Break- throughs in statistics: Methodology and distribution, Springer, 1992, pp

    B. Efron, Bootstrap methods: another look at the jackknife, in: Break- throughs in statistics: Methodology and distribution, Springer, 1992, pp. 569–593

    1992

  30. [38]

    A. E. Hoerl, R. W. Kennard, Ridge regression: Biased estimation for nonorthogonal problems, Technometrics 12 (1) (1970) 55–67

    1970

  31. [39]

    Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society Series B: Statistical Methodology 58 (1) (1996) 267–288

    R. Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society Series B: Statistical Methodology 58 (1) (1996) 267–288

    1996

  32. [40]

    Hutter, H

    F. Hutter, H. H. Hoos, K. Leyton-Brown, Sequential model-based optimiza- tion for general algorithm configuration, in: International conference on learn- ing and intelligent optimization, Springer, 2011, pp. 507–523

    2011

  33. [41]

    Bergstra, R

    J. Bergstra, R. Bardenet, Y. Bengio, B. K´ egl, Algorithms for hyper-parameter optimization, Advances in neural information processing systems 24 (2011)

    2011

  34. [42]

    Snoek, O

    J. Snoek, O. Rippel, K. Swersky, R. Kiros, N. Satish, N. Sundaram, M. Pat- wary, M. Prabhat, R. Adams, Scalable bayesian optimization using deep neu- ral networks, in: International conference on machine learning, 2015, pp. 2171–2180

    2015

  35. [43]

    Shahriari, K

    B. Shahriari, K. Swersky, Z. Wang, R. P. Adams, N. De Freitas, Taking the human out of the loop: A review of bayesian optimization, Proceedings of the IEEE 104 (1) (2015) 148–175

    2015

  36. [44]

    Garnett, Bayesian optimization, Cambridge University Press, 2023

    R. Garnett, Bayesian optimization, Cambridge University Press, 2023

    2023

  37. [45]

    C. M. Bishop, N. M. Nasrabadi, Pattern recognition and machine learning, Vol. 4, Springer, 2006

    2006

  38. [46]

    C. K. Williams, C. E. Rasmussen, Gaussian processes for machine learning, Vol. 2, MIT press Cambridge, MA, 2006

    2006

  39. [47]

    Schulz, M

    E. Schulz, M. Speekenbrink, A. Krause, A tutorial on gaussian process regres- sion: Modelling, exploring, and exploiting functions, Journal of mathematical psychology 85 (2018) 1–16

    2018

  40. [48]

    M. Liu, G. Chowdhary, B. C. Da Silva, S.-Y. Liu, J. P. How, Gaussian pro- cesses for learning and control: A tutorial with examples, IEEE Control Sys- tems Magazine 38 (5) (2018) 53–86

    2018

  41. [49]

    W. Holt, D. Nguyen, Essential aspects to bayesian data imputation, Available at SSRN 4494311 (2023). 34

    2023

  42. [50]

    D. R. Jones, M. Schonlau, W. J. Welch, Efficient global optimization of ex- pensive black-box functions, Journal of Global optimization 13 (4) (1998) 455–492

    1998

  43. [51]

    Srinivas, A

    N. Srinivas, A. Krause, S. M. Kakade, M. Seeger, Gaussian process optimiza- tion in the bandit setting: No regret and experimental design, arXiv preprint arXiv:0912.3995 (2009)

    2009 arXiv

  44. [52]

    Ament, S

    S. Ament, S. Daulton, D. Eriksson, M. Balandat, E. Bakshy, Unexpected improvements to expected improvement for bayesian optimization, arXiv preprint arXiv:2310.20708 (2023)

    2023 arXiv

  45. [53]

    M. G. Fern´ andez-Godino, C. Park, N.-H. Kim, R. T. Haftka, Review of multi- fidelity models, arXiv preprint arXiv:1609.07196 (2016)

    2016 arXiv

  46. [54]

    A. I. Forrester, Black-box calibration for complex-system simulation, Philo- sophical Transactions of The Royal Society A: Mathematical, Physical and Engineering Sciences 368 (1924) (2010) 3567–3579

    1924

  47. [55]

    Balabanov, B

    V. Balabanov, B. Grossman, L. Watson, W. Mason, R. Haftka, Multifi- delity response surface model for hsct wing bending material weight, in: 7th AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and Optimization, 1998, p. 4804

    1998

  48. [56]

    Fidkowski, Quantifying uncertainties in radiation hydrodynamics models, Ann Arbor 1001 (2014) 48109

    K. Fidkowski, Quantifying uncertainties in radiation hydrodynamics models, Ann Arbor 1001 (2014) 48109

    2014

  49. [57]

    A. I. Forrester, A. S´ obester, A. J. Keane, Multi-fidelity optimization via sur- rogate modelling, Proceedings of the royal society a: mathematical, physical and engineering sciences 463 (2088) (2007) 3251–3269

    2007

  50. [58]

    Le Gratiet, C

    L. Le Gratiet, C. Cannamela, B. Iooss, A bayesian approach for global sen- sitivity analysis of (multifidelity) computer codes, SIAM/ASA Journal on Uncertainty Quantification 2 (1) (2014) 336–363

    2014

  51. [59]

    Williams, E

    C. Williams, E. V. Bonilla, K. M. Chai, Multi-task gaussian process predic- tion, Advances in neural information processing systems (2007) 153–160

    2007

  52. [60]

    Kandasamy, G

    K. Kandasamy, G. Dasarathy, J. Schneider, B. P´ oczos, Multi-fidelity bayesian optimisation with continuous approximations, arXiv preprint arXiv:1703.06240 (2017)

    2017 arXiv

  53. [61]

    Zhang, Z.-H

    Y. Zhang, Z.-H. Han, K.-S. Zhang, Variable-fidelity expected improvement method for efficient global optimization of expensive functions, Structural and Multidisciplinary Optimization 58 (4) (2018) 1431–1451. 35

    2018

  54. [62]

    Jiang, J

    P. Jiang, J. Cheng, Q. Zhou, L. Shu, J. Hu, Variable-fidelity lower confidence bounding approach for engineering optimization problems with expensive sim- ulations, AIAA Journal 57 (12) (2019) 5416–5430

    2019

  55. [63]

    B. Do, R. Zhang, Multi-fidelity bayesian optimization in engineering design, arXiv preprint arXiv:2311.13050 (2023)

    2023

  56. [64]

    W. J. Maddox, M. Balandat, A. G. Wilson, E. Bakshy, Bayesian optimization with high-dimensional outputs, arXiv preprint arXiv:2106.12997 (2021)

    2021 arXiv

  57. [65]

    Swersky, J

    K. Swersky, J. Snoek, R. P. Adams, Multi-task bayesian optimization (2013)

    2013

  58. [66]

    Laumanns, J

    M. Laumanns, J. Ocenasek, Bayesian optimization algorithms for multi- objective optimization, in: International Conference on Parallel Problem Solv- ing from Nature, Springer, 2002, pp. 298–307

    2002

  59. [67]

    Schonlau, Computer experiments and global optimization (1997)

    M. Schonlau, Computer experiments and global optimization (1997)

    1997

  60. [68]

    Gonz´ alez, Z

    J. Gonz´ alez, Z. Dai, P. Hennig, N. Lawrence, Batch bayesian optimization via local penalization, in: Artificial intelligence and statistics, PMLR, 2016, pp. 648–657

    2016

  61. [69]

    Huang, T

    D. Huang, T. T. Allen, W. I. Notz, R. A. Miller, Sequential kriging opti- mization using multiple-fidelity evaluations, Structural and Multidisciplinary Optimization 32 (5) (2006) 369–382

    2006

  62. [70]

    Y. Liu, S. Chen, F. Wang, F. Xiong, Sequential optimization using multi- level cokriging and extended expected improvement criterion, Structural and Multidisciplinary Optimization 58 (3) (2018) 1155–1173

    2018

  63. [71]

    X. Ruan, P. Jiang, Q. Zhou, J. Hu, L. Shu, Variable-fidelity probability of improvement method for efficient global optimization of expensive black-box problems, Structural and Multidisciplinary Optimization 62 (6) (2020) 3021– 3052

    2020

  64. [72]

    Kansara, S

    H. Kansara, S. F. Khosroshahi, L. Guo, M. A. Bessa, W. Tan, Multi-objective bayesian optimisation of spinodoid cellular structures for crush energy absorp- tion, arXiv preprint arXiv:2411.14508 (2024)

    2024 arXiv

  65. [73]

    J. W. Cahn, Phase separation by spinodal decomposition in isotropic systems, The Journal of chemical physics 42 (1) (1965) 93–99

    1965

  66. [74]

    Zheng, S

    L. Zheng, S. Kumar, D. M. Kochmann, Data-driven topology optimization of spinodoid metamaterials with seamlessly tunable anisotropy, Computer Methods in Applied Mechanics and Engineering 383 (2021) 113894. 36

    2021

  67. [75]

    Y. Liu, H. Wang, L. Yan, J. Huang, Y. Liang, Mechanical properties of homo- geneous and functionally graded spinodal structures, International Journal of Mechanical Sciences 269 (2024) 109043

    2024

  68. [76]

    R. T. Marler, J. S. Arora, The weighted sum method for multi-objective optimization: new insights, Structural and multidisciplinary optimization 41 (2010) 853–862

    2010

  69. [77]

    J. Knowles, Parego: A hybrid algorithm with on-line landscape approximation for expensive multiobjective optimization problems, IEEE transactions on evolutionary computation 10 (1) (2006) 50–66

    2006

  70. [78]

    Ellison, F

    M. Ellison, F. DiazDelaO, N. Ince, M. Willetts, Robust optimisation of com- putationally expensive models using adaptive multi-fidelity emulation, Ap- plied Mathematical Modelling 100 (2021) 92–106

    2021

  71. [79]

    Irshad, S

    F. Irshad, S. Karsch, A. D¨ opp, Leveraging trust for joint multi-objective and multi-fidelity optimization, Machine Learning: Science and Technology 5 (1) (2024) 015056. 37

    2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.