Pith. sign in

REVIEW 3 major objections 4 minor 85 references

Pareto-optimizing surrogate models for error and complexity yields an ensemble that disagrees like an uncertainty estimate and beats state-of-the-art SAEAs.

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 ↗

EPOS constructs an ensemble of Pareto-optimal RBFN surrogates via NSGA-II trading accuracy against complexity, and uses LCB with ensemble disagreement to outperform state-of-the-art SAEAs on 1000-FE budgets.

T0 review reviewed 2026-08-04 challenge →

load-bearing objection A well-ablated, honest SAEA paper with a genuinely new idea—building ensembles from Pareto-optimal RBFN structures—though the headline comparison is somewhat inflated by same-suite hyperparameter tuning. the 3 major comments →

arxiv 2608.01777 v1 pith:DID4F7FW submitted 2026-08-03 cs.NE

An Evolutionary Algorithm Assisted by an Ensemble of Pareto-Optimal Surrogate Models

Kei Nishihara , Yaochu Jin , Masaya Nakata This is my paper
classification cs.NE
keywords surrogate-assisted evolutionary algorithmensemble surrogate modelPareto-optimal surrogatesradial basis function networkdifferential evolutionlower confidence boundexpensive optimizationmodel complexity
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.

The reading

The paper proposes EPOS, an evolutionary algorithm for expensive optimization problems in which the surrogate model is not a single fitted curve but an ensemble built automatically. Each generation, EPOS trains many radial basis function networks that differ in the number of hidden nodes and in the smoothness of their fitted landscape, and it uses NSGA-II to keep only those that form the trade-off front between prediction error on held-out data and model complexity. The ensemble is then the average of these Pareto-optimal surrogates, and the disagreement among them is used as a stand-in for uncertainty in a lower-confidence-bound rule that selects which candidate solution to evaluate with the expensive objective. On the CEC 2020 suite with 30, 50, and 100 dimensions and a budget of 1,000 evaluations, the paper reports that EPOS statistically beats ten existing surrogate-assisted algorithms in 251 of 300 pairwise comparisons, with 25 ties, and obtains the best average rank; on the CEC 2011 real-world suite it again ranks first on average. The claim matters because it suggests that a simple network model, shaped by a diversity-producing trade-off, can replace hand-picked surrogate types and probability-based uncertainty models when evaluations are scarce.

Core claim

EPOS is an adaptive ensemble SAEA. In the surrogate adaptation phase, NSGA-II solves a bi-objective problem: minimize the test RMSE of an RBFN and minimize the number of hidden-layer nodes n, with the spread parameter σ as a second design variable. The Pareto set obtained after ω_max generations is used directly as the ensemble. For prescreening, the mean prediction and the unbiased sample standard deviation of the ensemble are computed; the candidate with the smallest LCB = mean − α·std is sent to the real fitness function. The reported experimental results show that this combination yields statistically better solutions than the comparison algorithms on most CEC 2020 instances (aggregate +

What carries the argument

The central object is the Pareto-optimal surrogate ensemble: a set of RBFNs found by NSGA-II that minimize the two objectives (RMSE, n). Because n and σ control the roughness of the fitted landscape, Pareto-optimal members automatically span a range of smoothness while staying accurate. The second load-bearing piece is the LCB infill criterion LCB(x)=mean(x)−α·std(x), where the standard deviation of the ensemble members' outputs (Eq. 8) emulates uncertainty without Kriging's probabilistic machinery. The combination is what makes the algorithm adaptive (a new ensemble is built every generation from the current archive) and ensemble-based (all Pareto members vote).

Load-bearing premise

That the disagreement among the Pareto-optimal surrogates is a valid stand-in for prediction uncertainty; if the spread of model outputs does not track the actual error, the LCB infill criterion reduces to a plain average and the algorithm's edge over simpler ensemble variants is unexplained.

What would settle it

Take the CEC 2020 functions at D=30, fit the EPOS ensemble, and on the held-out test set compute the Pearson correlation between the per-point ensemble standard deviation (Eq. 8) and the absolute error of the ensemble mean. If the correlation is close to zero across functions, the disagreement is not acting as an uncertainty measure, and the reported advantage over the Mean variant would need a different explanation.

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

If this is right

  • Adaptive SAEAs can stop relying on a small set of hand-picked model types or RBF kernels: a single flexible model class, tuned by a trade-off objective, can supply the diversity the ensemble needs.
  • A disagreement-based LCB offers a practical uncertainty estimate in high dimensions, where full probabilistic Kriging variance is costly and fragile; this should benefit any acquisition-based search, not only DE.
  • The advantage over the compared algorithms grows with dimension (D=30 to D=100), so the method is most promising for medium- to high-dimensional expensive problems.
  • Ablation results support using the whole Pareto front rather than just the most accurate or the most extreme models: ensembles of size two underperform, and using all Pareto members gives the most stable best-rank results.
  • The same surrogate-ensemble idea transfers to Bayesian optimization with Kriging/DACE models, where the Pareto version with emulated LCB outperformed plain LCB-based BO in the paper's tests.

Where Pith is reading between the lines

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

  • I would expect the disagreement signal to be most informative when the training data are sparse and the landscape multimodal; on smooth, low-dimensional problems the LCB term should add little beyond the mean, so the algorithm's advantage over a Mean variant may shrink or vanish.
  • A straightforward test of the paper's uncertainty story is to calibrate the ensemble standard deviation against empirical residuals on held-out test data; if the relationship is weak, the gains likely come from averaging rather than from the LCB exploration term.
  • The complexity objective could be replaced by a more direct smoothness measure (e.g., the Lipschitz constant or curvature of the fitted surface), which may make the trade-off interpretable and transferable to other surrogate model classes such as Gaussian processes.
  • Because the paper reports source code availability, a third party can rerun the CEC 2020/2011 comparisons and test whether the advantage persists with a different optimizer (e.g., CMA-ES) in place of DE.
Share X Bluesky LinkedIn Reddit HN

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 / 4 minor

Summary. The paper proposes EPOS, a surrogate-assisted evolutionary algorithm that constructs an ensemble of RBFN surrogates by solving a bi-objective problem: minimizing RMSE on a held-out test set and minimizing model complexity (number of hidden nodes), with the RBF spread parameter as a second decision variable. The Pareto-optimal surrogate set produced by NSGA-II is averaged to form the ensemble prediction, and the sample standard deviation of member predictions is used as an uncertainty measure in an LCB infill criterion for prescreening DE offspring. Experiments on CEC2020 (D=30,50,100; 1000 FEs; 31 trials) and CEC2011 real-world problems report that EPOS statistically outperforms ten state-of-the-art SAEAs, with aggregate +/−/∼ = 24/251/25 on CEC2020 and best average ranks of 2.300 and 2.955 on CEC2020 and CEC2011, respectively. Extensive ablations and sensitivity analyses are included in the main text and supplementary material.

Significance. If the empirical claims hold, the core idea—automatically constructing surrogate ensembles via Pareto optimization of error and complexity—is a novel and potentially useful contribution to SAEAs. The study is unusually thorough in its experimental design: 31 independent trials, Wilcoxon signed-rank tests, standard deviations, ablations of each algorithmic component (Acc-only, Smp-only, NoMOEA, NoSigma, Mean, EI), sensitivity analyses of key hyperparameters, and a real-world benchmark suite. The source code is publicly available. The CEC2011 results provide some out-of-sample support, and the BO extension shows the ensemble principle transfers to Kriging. However, two issues are load-bearing for the central claim: (a) the default hyperparameters appear to be evaluated and possibly selected on the same CEC2020 suite used for the headline comparison, and the sensitivity analyses show strong performance variation; (b) the emulated uncertainty used in the LCB criterion is explicitly acknowledged in Section VII to be not formally grounded in probability theory and to fail in certain scenarios. While both are acknowledged, they are not resolved, and they directly affect the strength of th

major comments (3)
  1. [Section V-A-2; Supplementary Sections S-VIII and S-IX; Tables S-IV and S-V] The paper does not report how EPOS's default hyperparameters (N_M=10, omega_max=10, F=0.5, CR=0.9, alpha=2, etc.) were selected. The sensitivity analyses in S-VIII and S-IX, conducted on the same CEC2020 suite, show substantial performance variation: average rank ranges from 2.467 to 7.533 across F/CR settings, and smaller N_M/omega_max values degrade results. If the defaults were chosen after inspecting CEC2020 results, the headline comparison is in-sample for EPOS while the baselines use hyperparameters from their original papers. This threatens the general claim of statistical superiority. The authors should either specify the tuning protocol, use a separate tuning set or nested validation, or demonstrate that EPOS maintains superiority across a broad range of reasonable hyperparameter settings.
  2. [Equations (8)-(9); Section VII; Table V c) in Section VI-C] The uncertainty term in Eq. (8) is the sample standard deviation of the Pareto-surrogate outputs, used in the LCB infill criterion of Eq. (9). The paper explicitly states in Section VII that this emulated uncertainty 'is not formally grounded in probability theory, unlike Kriging' and 'may not work when there are many extremely inaccurate models or when all models make the same prediction.' The 'Mean' ablation (Table V c) shows that EPOS with only the average surrogate is competitive or better at early FE budgets, so the demonstrated benefit of the uncertainty term is concentrated at later FEs and is not guaranteed. To support the claim that the LCB mechanism is a substantive contribution, the authors should provide empirical calibration evidence (e.g., coverage of prediction intervals, correlation between ensemble disagreement and actual error) or a formal justification. As it stands, t
  3. [Table S-VIII (F6 row) and Table VIII] For CEC2011 problem F6, the paper states 'GL-SADE failed in running' and Table S-VIII lists N/A for GL-SADE. However, the corresponding +/−/∼ counts in Table VIII for GL-SADE sum to 22 (6/12/4), which is the full problem count. It is unclear how the N/A value was treated in the Wilcoxon signed-rank test and in the average rank calculation. This is a data-handling detail that must be specified, as it affects the statistical comparison against GL-SADE.
minor comments (4)
  1. [Table I] The last row lists '110 Pareto-optimal surrogates obtained by NSGA-II.' In fact, 110 is the total number of surrogate models constructed per generation (N_M + N_M*omega_max), but only the nondominated solutions of the final NSGA-II population (at most N_M=10) form the Pareto set used for the ensemble. Please rephrase to avoid overstating the number of Pareto-optimal models.
  2. [Section VI-C, Eq. (10)] In the EI formula, the second term uses the standard normal density function, but the text and equation use the same symbol Phi for both CDF and PDF. Replace the second Phi with phi (or clarify the notation).
  3. [Section I] 'bless of uncertainty' appears to be a typo for 'blessing of uncertainty'.
  4. [Equation (5)] The lower bound of the number of hidden nodes, x_M,1 = |D_train|/2, is stated without justification. A brief explanation of why half the training-data size is a principled minimum would improve accessibility.

Circularity Check

0 steps flagged

No significant circularity: EPOS's central claim is an empirical algorithm comparison, and its uncertainty/infill equations are defined statistics, not fitted inputs or self-citation-derived results.

full rationale

The paper's central claim is an empirical comparison of a new algorithm (EPOS) against state-of-the-art SAEAs on CEC2020 and CEC2011. The core mechanism—minimizing RBFN approximation error (RMSE) and model complexity (n) via NSGA-II—is a design choice expressed in Eq. (3), not a quantity fitted to the headline result. The uncertainty measure in Eq. (8) is defined as the sample standard deviation of the Pareto-optimal surrogate outputs, and Eq. (9) defines LCB as mean minus alpha times that standard deviation; these are explicit definitions, and the paper openly states in Section VII that this emulated uncertainty 'is not formally grounded in probability theory, unlike Kriging.' Thus there is no hidden derivation of the conclusion from the input. The comparisons against ten external SAEAs use hyperparameters from the original papers, and EPOS's settings are stated in Table S-II; while the sensitivity analyses in Sections S-VIII and S-IX show performance variation across hyperparameter choices on CEC2020, no evidence in the text indicates that the defaults were chosen by tuning on the same benchmark used for the headline comparison, and such a protocol concern would be a benchmarking issue rather than a circular derivation. The self-citations, including SADE-ATDSC [7], are used as baselines and related work, not as load-bearing support for the main claim, and no 'uniqueness theorem' is imported from the authors' prior work. The paper also supports its component choices with controlled ablations (Acc-only, Smp-only, NoMOEA, NoSigma, Mean, EI), which further reduces any concern that the central result is forced by construction.

Axiom & Free-Parameter Ledger

7 free parameters · 5 axioms · 0 invented entities

The method introduces no new physical entities. Its empirical support rests on several unverified modeling assumptions, the strongest being the engineered mapping from (n, sigma) to landscape smoothness and the validity of ensemble disagreement as uncertainty. These are acknowledged as ungrounded by the authors in the conclusion.

free parameters (7)
  • LCB coefficient alpha = 2
    Chooses exploration strength in Eq. (9); fixed in all runs; sensitivity not reported.
  • NSGA-II population size N_M = 10
    Together with omega_max sets total surrogate models (110); chosen by hand; sensitivity in S-VIII shows 10/10 among best.
  • NSGA-II generations omega_max = 10
    Chosen by hand; sensitivity in S-VIII.
  • test data rate delta = 0.2
    Ratio of archive data held out for RMSE calculation in surrogate selection.
  • training data size N_data = 5D
    Number of top solutions used to build surrogates; scales with dimension.
  • DE scaling factor F = 0.5
    Sensitivity in S-IX shows impact on performance.
  • DE crossover rate CR = 0.9
    Sensitivity in S-IX; CR=1.0 substantially worse.
axioms (5)
  • domain assumption RBFN output smoothness is effectively controlled by hidden node count n and spread sigma
    Section III-C and IV-A; the entire EPOS design relies on (n, sigma) as a handle on smoothness, but no direct smoothness measure is reported.
  • domain assumption The ensemble standard deviation (Eq. 8) is a usable proxy for model uncertainty in LCB
    Section IV.B.3 and VII; authors explicitly note it is not formally grounded in probability theory, unlike Kriging.
  • domain assumption RMSE on a holdout split of the top-N_data archived solutions is a reliable surrogate-selection criterion
    Section IV.B.1, Eq. (3); the test data is drawn from the elite region, whose distribution shifts during search.
  • domain assumption NSGA-II with 10 population and 10 generations returns a representative Pareto set of surrogates
    Section IV.B.1; the small budget is justified only by sensitivity analysis, not by convergence evidence.
  • domain assumption k-means cluster centers provide adequate RBF centers for any surrogate structure
    Section IV.A; centers set to k-means of training data without validation.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of An Evolutionary Algorithm Assisted by an Ensemble of Pareto-Optimal Surrogate Models." pith.science (2026). https://pith.science/paper/DID4F7FW

@misc{pith2026260801777,
  author       = {Pith},
  title        = {Pith review of: An Evolutionary Algorithm Assisted by an Ensemble of Pareto-Optimal Surrogate Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DID4F7FW}},
  note         = {Machine review of arXiv:2608.01777}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

An ensemble of surrogate models helps improve the prediction quality and robustness of surrogate models, and in turn, the search performance of surrogate-assisted evolutionary algorithms (SAEAs). Although different degrees of smoothness of the approximated fitness landscapes need to be carefully designed for an effective ensemble, little attention has been paid to the explicit tuning of the degree of smoothness derived by surrogate models. This study proposes an adaptive ensemble SAEA, which automatically constructs plausible ensemble models by optimizing their parameter settings. Unlike existing adaptive/ensemble SAEAs, which consider prediction accuracy alone, the proposed algorithm optimizes the structure of radial basis function networks (RBFNs) by solving bi-objective minimization problems of approximation error and model complexity, resulting in robust ensemble models of accurate surrogate models with different degrees of smoothness of the approximated fitness landscapes. As a result, the over/under-fittings are reduced. Additionally, an infill criterion is designed so that surrogate models with different degrees of smoothness can contribute to the solution prescreening. The experimental results demonstrated the statistical superiority of our algorithm over state-of-the-art SAEAs on a single-objective benchmark and real-world problem sets under an expensive optimization scenario. The source code of the proposed algorithm is available at https://github.com/haranychan/EPOS

Figures

Figures reproduced from arXiv: 2608.01777 by Kei Nishihara, Masaya Nakata, Yaochu Jin.

Figure 1
Figure 1. Figure 1: A demonstration of the effectiveness of the ensemble surrogate model [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A diagram of the proposed EPOS, where the solid arrows stand for [PITH_FULL_IMAGE:figures/full_fig_p007_2.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

85 extracted references · 80 canonical work pages

  1. [1]

    Managing approximate models in evolutionary aerodynamic design optimization,

    Y . Jin, M. Olhofer, and B. Sendhoff, “Managing approximate models in evolutionary aerodynamic design optimization,” inIEEE Congr. Evol. Comput. (CEC), vol. 1, May 2001, pp. 592–599

    2001

  2. [2]

    Evolution by adapting surrogates,

    M. N. Le, Y . S. Ong, S. Menzel, Y . Jin, and B. Sendhoff, “Evolution by adapting surrogates,”Evol. Comput., vol. 21, no. 2, pp. 313–340, 2013

    2013

  3. [3]

    Surrogate-assisted teaching- learning-based optimization for high-dimensional and computationally expensive problems,

    H. Dong, P. Wang, X. Yu, and B. Song, “Surrogate-assisted teaching- learning-based optimization for high-dimensional and computationally expensive problems,”Appl. Soft Comput., vol. 99, p. 106934, Feb. 2021

    2021

  4. [4]

    EMONAS-Net: Efficient multiob- jective neural architecture search using surrogate-assisted evolutionary algorithm for 3D medical image segmentation,

    M. B. Calisto and S. K. Lai-Yuen, “EMONAS-Net: Efficient multiob- jective neural architecture search using surrogate-assisted evolutionary algorithm for 3D medical image segmentation,”Artif. Intell. Med., vol. 119, p. 102154, Sep. 2021

    2021

  5. [5]

    NSGA-Net: neural architecture search using multi- objective genetic algorithm,

    Z. Lu, I. Whalen, V . Boddeti, Y . Dhebar, K. Deb, E. Goodman, and W. Banzhaf, “NSGA-Net: neural architecture search using multi- objective genetic algorithm,” inAnnu. Conf. Genet. Evol. Comput. (GECCO), Jul. 2019, pp. 419–427

    2019

  6. [6]

    Surrogate-assisted evolutionary optimisa- tion: a novel blueprint and a state of the art survey,

    M. I. E. Khaldi and A. Draa, “Surrogate-assisted evolutionary optimisa- tion: a novel blueprint and a state of the art survey,”Evol. Intell., vol. 17, no. 4, pp. 2213–2243, Aug. 2024. NISHIHARAet al.: EA ASSISTED BY AN EPOS MODELS 13

    work page 2024

  7. [7]

    Surrogate-assisted Differential Evolution with Adaptation of Training Data Selection Criterion,

    K. Nishihara and M. Nakata, “Surrogate-assisted Differential Evolution with Adaptation of Training Data Selection Criterion,” inIEEE Symp. Ser. Comput. Intell. (SSCI), Dec. 2022, pp. 1675–1682

    work page 2022

  8. [8]

    Incorporating gradient information into dimension perturbation mutation for high- dimensional expensive optimization,

    Z. Yang, S. Chu, J. Liu, H. Qiu, M. Xiao, and L. Gao, “Incorporating gradient information into dimension perturbation mutation for high- dimensional expensive optimization,”Swarm Evol. Comput., vol. 84, p. 101446, Feb. 2024

    work page 2024

  9. [9]

    A Fast Kriging-Assisted Evolutionary Algorithm Based on Incremental Learning,

    D. Zhan and H. Xing, “A Fast Kriging-Assisted Evolutionary Algorithm Based on Incremental Learning,”IEEE Trans. Evol. Comput., vol. 25, no. 5, pp. 941–955, Oct. 2021

    work page 2021

  10. [10]

    DACE: A MAT- LAB Kriging Toolbox,

    S. N. Lophaven, H. B. Nielsen, and J. Søndergaard, “DACE: A MAT- LAB Kriging Toolbox,” Informatics and Mathematical Modelling, DTU, Tech. Rep. IMM-REP-2002-12, Aug. 2002

    work page 2002

  11. [11]

    A supervised surrogate- assisted evolutionary algorithm for complex optimization problems,

    X. Zhao, X. Jia, T. Zhang, T. Liu, and Y . Cao, “A supervised surrogate- assisted evolutionary algorithm for complex optimization problems,” IEEE Trans. Instrum. Meas., vol. 72, 2023

    work page 2023

  12. [12]

    Are you concerned about limited function evaluations: Data-augmented Pareto set learning for expensive multi- objective optimization,

    Y . Lu, B. Li, and A. Zhou, “Are you concerned about limited function evaluations: Data-augmented Pareto set learning for expensive multi- objective optimization,”Proc. Conf. AAAI Artif. Intell., vol. 38, no. 13, pp. 14 202–14 210, Mar. 2024

    work page 2024

  13. [13]

    A two-phase kriging- assisted evolutionary algorithm for expensive constrained multiobjective optimization problems,

    Z. Zhang, Y . Wang, J. Liu, G. Sun, and K. Tang, “A two-phase kriging- assisted evolutionary algorithm for expensive constrained multiobjective optimization problems,”IEEE Trans. Syst. Man Cybern. A Syst., vol. 54, no. 8, pp. 4579–4591, Aug. 2024

    work page 2024

  14. [14]

    A distribution information- based kriging-assisted evolutionary algorithm for expensive many- objective optimization problems,

    Z. Zhang, Y . Wang, G. Sun, and T. Pang, “A distribution information- based kriging-assisted evolutionary algorithm for expensive many- objective optimization problems,”IEEE Trans. Evol. Comput., vol. 29, no. 6, pp. 2656–2670, Dec. 2025

    work page 2025

  15. [15]

    Adaptive basis function selection enhanced multisurrogate-assisted evolutionary algorithm for production optimization,

    J. Wang, K. Zhang, L. Zhang, J. Wang, W. Peng, X. Yan, H. Wang, H. Zhang, Y . Yang, H. Sun, P. Liu, H. Chen, and X. Fang, “Adaptive basis function selection enhanced multisurrogate-assisted evolutionary algorithm for production optimization,”SPE J., vol. 28, no. 06, pp. 3192–3209, Dec. 2023

    work page 2023

  16. [16]

    Surrogate-assisted evolutionary computation: Recent advances and future challenges,

    Y . Jin, “Surrogate-assisted evolutionary computation: Recent advances and future challenges,”Swarm Evol. Comput., vol. 1, no. 2, pp. 61–70, Jun. 2011

    work page 2011

  17. [17]

    A study on metamodeling techniques, ensembles, and multi-surrogates in evolutionary computa- tion,

    D. Lim, Y .-S. Ong, Y . Jin, and B. Sendhoff, “A study on metamodeling techniques, ensembles, and multi-surrogates in evolutionary computa- tion,” inAnnu. Conf. Genet. Evol. Comput. (GECCO), Jul. 2007, pp. 1288–1295

    work page 2007

  18. [18]

    Ensemble of surrogates,

    T. Goel, R. T. Haftka, W. Shyy, and N. V . Queipo, “Ensemble of surrogates,”Struct. Multidiscip. Optim., vol. 33, no. 3, pp. 199–216, Mar. 2007

    work page 2007

  19. [19]

    Generalizing Surrogate- Assisted Evolutionary Computation,

    D. Lim, Y . Jin, Y .-S. Ong, and B. Sendhoff, “Generalizing Surrogate- Assisted Evolutionary Computation,”IEEE Trans. Evol. Comput., vol. 14, no. 3, pp. 329–355, Jun. 2010

    work page 2010

  20. [20]

    A twofold infill criterion-driven heterogeneous ensemble surrogate-assisted evolutionary algorithm for computationally expensive problems,

    M. Yu, J. Liang, Z. Wu, and Z. Yang, “A twofold infill criterion-driven heterogeneous ensemble surrogate-assisted evolutionary algorithm for computationally expensive problems,”Knowl. Based. Syst., vol. 236, p. 107747, Jan. 2022

    work page 2022

  21. [21]

    Universal approximation using radial-basis- function networks,

    J. Park and I. W. Sandberg, “Universal approximation using radial-basis- function networks,”Neural Comput., vol. 3, no. 2, pp. 246–257, 1991

    work page 1991

  22. [22]

    An XGBoost- assisted evolutionary algorithm for expensive multiobjective optimiza- tion problems,

    F. Mao, M. Chen, K. Zhong, J. Zeng, and Z. Liang, “An XGBoost- assisted evolutionary algorithm for expensive multiobjective optimiza- tion problems,”Inf. Sci., vol. 666, p. 120449, May 2024

    work page 2024

  23. [23]

    XGBoost: A Scalable Tree Boosting System,

    T. Chen and C. Guestrin, “XGBoost: A Scalable Tree Boosting System,” inInt. Conf. Knowl. Discov. Data Min. (KDD), Aug. 2016, pp. 785–794

    work page 2016

  24. [24]

    Heterogeneous ensemble-based infill criterion for evolutionary multiobjective optimization of expensive problems,

    D. Guo, Y . Jin, J. Ding, and T. Chai, “Heterogeneous ensemble-based infill criterion for evolutionary multiobjective optimization of expensive problems,”IEEE Trans. Cybern., vol. 49, no. 3, pp. 1012–1025, Mar. 2019

    work page 2019

  25. [25]

    R. H. Myers and D. C. Montgomery,Response surface methodology– process and product optimization using designed experiments. John Wiley & Sons, 1995

    work page 1995

  26. [26]

    Ensemble of surrogates assisted particle swarm optimization of medium scale expensive problems,

    F. Li, X. Cai, and L. Gao, “Ensemble of surrogates assisted particle swarm optimization of medium scale expensive problems,”Appl. Soft Comput., vol. 74, pp. 291–305, Jan. 2019

    work page 2019

  27. [27]

    Reducing Fitness Evaluations Using Clustering Techniques and Neural Network Ensembles,

    Y . Jin and B. Sendhoff, “Reducing Fitness Evaluations Using Clustering Techniques and Neural Network Ensembles,” inAnnu. Conf. Genet. Evol. Comput. (GECCO), 2004, pp. 688–699

    work page 2004

  28. [28]

    A theory of adaptive pattern classifiers,

    S. Amari, “A theory of adaptive pattern classifiers,”IEEE Trans. Electron. Comput., vol. EC-16, no. 3, pp. 299–307, Jun. 1967

    work page 1967

  29. [29]

    Surrogate models in evolutionary single-objective optimization: A new taxonomy and experimental study,

    H. Tong, C. Huang, L. L. Minku, and X. Yao, “Surrogate models in evolutionary single-objective optimization: A new taxonomy and experimental study,”Inf. Sci., vol. 562, pp. 414–437, Jul. 2021

    work page 2021

  30. [30]

    A dynamic surrogate-assisted evolutionary algorithm framework for expensive structural optimization,

    M. Yu, X. Li, and J. Liang, “A dynamic surrogate-assisted evolutionary algorithm framework for expensive structural optimization,”Struct. Multidiscip. Optim., vol. 61, no. 2, pp. 711–729, Feb. 2020

    work page 2020

  31. [31]

    Evolutionary Computation for Expensive Optimization: A Survey,

    J.-Y . Li, Z.-H. Zhan, and J. Zhang, “Evolutionary Computation for Expensive Optimization: A Survey,”Mach, Intell. Res., vol. 19, no. 1, pp. 3–23, Feb. 2022

    work page 2022

  32. [32]

    Classification of adaptive memetic algorithms: a comparative study,

    Y .-S. Ong, M.-H. Lim, N. Zhu, and K.-W. Wong, “Classification of adaptive memetic algorithms: a comparative study,”IEEE Trans. Syst. Man Cybern. B Cybern., vol. 36, no. 1, pp. 141–152, Feb. 2006

    work page 2006

  33. [33]

    Surrogate-assisted evo- lutionary algorithm with model and infill criterion auto-configuration,

    L. Xie, G. Li, Z. Wang, L. Cui, and M. Gong, “Surrogate-assisted evo- lutionary algorithm with model and infill criterion auto-configuration,” IEEE Trans. Evol. Comput., vol. 28, no. 4, pp. 1114–1126, Aug. 2024

    work page 2024

  34. [34]

    A decision-theoretic method for surrogate model selection,

    R. V . Field, “A decision-theoretic method for surrogate model selection,” J. Sound Vib., vol. 311, no. 3, pp. 1371–1390, Apr. 2008

    work page 2008

  35. [35]

    Size-reducing RBF networks,

    J. Murata, S. Itoh, and K. Hirasawa, “Size-reducing RBF networks,” in IEEE Int. Jt. Conf. Neural Netw. (IJCNN), vol. 2, 1999, pp. 1308–1312

    work page 1999

  36. [36]

    Pareto-based multiobjective machine learning: An overview and case studies,

    Y . Jin and B. Sendhoff, “Pareto-based multiobjective machine learning: An overview and case studies,”IEEE Trans. Syst. Man Cybern. C Appl. Rev., vol. 38, no. 3, pp. 397–415, May 2008

    work page 2008

  37. [37]

    A fast and elitist multiobjective genetic algorithm: NSGA-II,

    K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, “A fast and elitist multiobjective genetic algorithm: NSGA-II,”IEEE Trans. Evol. Comput., vol. 6, no. 2, pp. 182–197, Apr. 2002

    work page 2002

  38. [38]

    Differential Evolution – A Simple and Efficient Heuristic for Global Optimization over Continuous Spaces,

    R. Storn and K. Price, “Differential Evolution – A Simple and Efficient Heuristic for Global Optimization over Continuous Spaces,”J. Global Optimiz., vol. 11, pp. 341–359, 1997

    work page 1997

  39. [39]

    A pairwise ranking estimation model for surrogate-assisted evolutionary algorithms,

    T. Harada, “A pairwise ranking estimation model for surrogate-assisted evolutionary algorithms,”Complex Intell. Syst., vol. 9, no. 6, pp. 6875– 6890, Dec. 2023

    work page 2023

  40. [40]

    Problem Definitions and Evaluation Criteria for the CEC 2020 Special Session and Competition on Sin- gle Objective Bound Constrained Numerical Optimization,

    C. T. Yue, K. V . Price, P. N. Suganthan, J. J. Liang, M. Z. Ali, B. Y . Qu, N. H. Awad, and P. P. Biswas, “Problem Definitions and Evaluation Criteria for the CEC 2020 Special Session and Competition on Sin- gle Objective Bound Constrained Numerical Optimization,” Zhengzhou Univ., China, and Nanyang Tech. Univ., Singapore, Tech. Rep., 2019

    work page 2020

  41. [41]

    Evolutionary model type selection for global surrogate modeling,

    D. Gorissen, T. Dhaene, and F. Turck, “Evolutionary model type selection for global surrogate modeling,”J. Mach. Learn. Res., vol. 10, pp. 2039–2078, Dec. 2009

    work page 2039

  42. [42]

    Radial Basis Function Assisted Optimization Method with Batch Infill Sampling Criterion for Expensive Optimization,

    G. Li, Q. Zhang, J. Sun, and Z. Han, “Radial Basis Function Assisted Optimization Method with Batch Infill Sampling Criterion for Expensive Optimization,” inIEEE Congr. Evol. Comput. (CEC), Jun. 2019, pp. 1664–1671

    work page 2019

  43. [43]

    An aRBF surrogate-assisted neighborhood field optimizer for expensive problems,

    M. Yu, J. Liang, K. Zhao, and Z. Wu, “An aRBF surrogate-assisted neighborhood field optimizer for expensive problems,”Swarm Evol. Comput., vol. 68, p. 100972, Aug. 2021

    work page 2021

  44. [44]

    An efficient surrogate-assisted hybrid optimization algorithm for expensive optimization problems,

    J.-S. Pan, N. Liu, S.-C. Chu, and T. Lai, “An efficient surrogate-assisted hybrid optimization algorithm for expensive optimization problems,”Inf. Sci., vol. 561, pp. 304–325, Jun. 2021

    work page 2021

  45. [45]

    A data-driven evolutionary algorithm with multi-evolutionary sampling strategy for expensive optimization,

    F. Yu, W. Gong, and H. Zhen, “A data-driven evolutionary algorithm with multi-evolutionary sampling strategy for expensive optimization,” Knowl. Based. Syst., vol. 242, p. 108436, Apr. 2022

    work page 2022

  46. [46]

    Evolutionary Sampling Agent for Expensive Problems,

    H. Zhen, W. Gong, and L. Wang, “Evolutionary Sampling Agent for Expensive Problems,”IEEE Trans. Evol. Comput., vol. 27, no. 3, pp. 716–727, Jun. 2023

    work page 2023

  47. [47]

    Surrogate-assisted evolutionary framework with an ensemble of teaching-learning and differential evolution for expensive optimization,

    X. Lin and Z. Meng, “Surrogate-assisted evolutionary framework with an ensemble of teaching-learning and differential evolution for expensive optimization,”Inf. Sci., vol. 680, p. 121137, Oct. 2024

    work page 2024

  48. [48]

    Angle information enhanced kriging-assisted adaptive evolutionary algorithm for compu- tationally highly expensive high-dimensional problems,

    H. Lu, Z. Yang, J. Huang, L. Chen, and X. Cai, “Angle information enhanced kriging-assisted adaptive evolutionary algorithm for compu- tationally highly expensive high-dimensional problems,”Swarm Evol. Comput., vol. 96, p. 101992, Jul. 2025

    work page 2025

  49. [49]

    Automated surrogate- assisted particle swarm optimizer with an adaptive parental guidance strategy for expensive engineering optimization problems,

    R. Dai, J. Jie, Z. Wang, H. Zheng, and W. Wang, “Automated surrogate- assisted particle swarm optimizer with an adaptive parental guidance strategy for expensive engineering optimization problems,”J. Comput. Des. Eng., vol. 12, no. 3, pp. 145–183, Feb. 2025

    work page 2025

  50. [50]

    A Hierarchical Surrogate-Assisted Differential Evolution With Core Space Localization,

    L. Yu, Z. Meng, and H. Zhu, “A Hierarchical Surrogate-Assisted Differential Evolution With Core Space Localization,”IEEE Trans. Cybern., vol. 55, no. 2, pp. 939–952, Feb. 2025

    work page 2025

  51. [51]

    Improving the search performance of SHADE using linear population size reduction,

    R. Tanabe and A. S. Fukunaga, “Improving the search performance of SHADE using linear population size reduction,” inIEEE Congr. Evol. Comput. (CEC), Jul. 2014, pp. 1658–1665

    work page 2014

  52. [52]

    Co- evolution of large language models and configuration strategies to enhance surrogate-assisted evolutionary algorithm,

    L. Xie, Y . Zhang, Z. Tang, E. Chung, G. Li, and Z. Wang, “Co- evolution of large language models and configuration strategies to enhance surrogate-assisted evolutionary algorithm,” inInt. Conf. Knowl. Discov. Data Min. (KDD), Aug. 2025, pp. 3321–3332

    work page 2025

  53. [53]

    (θ l, θu)-Parametric Multi-Task Optimization: Joint Search in Solution and Infinite Task Spaces,

    T. Wei, J. Liu, A. Gupta, P. S. Tan, and Y .-S. Ong, “(θ l, θu)-Parametric Multi-Task Optimization: Joint Search in Solution and Infinite Task Spaces,”IEEE Trans. Evol. Comput., vol. 30, no. 3, pp. 1270–1283, 2025. 14 IEEE TRANSACTIONS ON CYBERNETICS (A PREPRINT OF THE “ACCEPTED ARTICLE” VERSION)

    work page 2025

  54. [54]

    Local ensemble surrogate assisted crowding differential evolution,

    C. Jin, A. K. Qin, and K. Tang, “Local ensemble surrogate assisted crowding differential evolution,” inIEEE Congr. Evol. Comput. (CEC), May 2015, pp. 433–440

    work page 2015

  55. [55]

    Model-based methods for contin- uous and discrete global optimization,

    T. Bartz-Beielstein and M. Zaefferer, “Model-based methods for contin- uous and discrete global optimization,”Appl. Soft Comput., vol. 55, pp. 154–167, Jun. 2017

    work page 2017

  56. [56]

    Committee-Based Active Learning for Surrogate-Assisted Particle Swarm Optimization of Expensive Prob- lems,

    H. Wang, Y . Jin, and J. Doherty, “Committee-Based Active Learning for Surrogate-Assisted Particle Swarm Optimization of Expensive Prob- lems,”IEEE Trans. Cybern., vol. 47, no. 9, pp. 2664–2677, Sep. 2017

    work page 2017

  57. [57]

    A novel surrogate- assisted evolutionary algorithm with an uncertainty grouping based infill criterion,

    Q. Liu, X. Wu, Q. Lin, J. Ji, and K.-C. Wong, “A novel surrogate- assisted evolutionary algorithm with an uncertainty grouping based infill criterion,”Swarm Evol. Comput., vol. 60, p. 100787, Feb. 2021

    work page 2021

  58. [58]

    A surrogate-assisted evolutionary algorithm with Gaussian process regression and diversity search for large-scale expensive optimization,

    X. Ma, Y . Li, Y . Zhuang, Y . Li, J. Fan, L. Wang, Y . Qi, and J. Xiong, “A surrogate-assisted evolutionary algorithm with Gaussian process regression and diversity search for large-scale expensive optimization,” Appl. Soft Comput., vol. 182, no. 113440, p. 113440, Oct. 2025

    work page 2025

  59. [59]

    A surrogate-assisted differential evolution algorithm with a dual-space-driven selection strategy for expensive optimization problems,

    H. Liu, Z. Ren, C. He, and W. Du, “A surrogate-assisted differential evolution algorithm with a dual-space-driven selection strategy for expensive optimization problems,”Complex Intell. Syst., vol. 11, no. 5, p. 230, May 2025

    work page 2025

  60. [60]

    A classifier-ensemble- based surrogate-assisted evolutionary algorithm for distributed data- driven optimization,

    X.-Q. Guo, F.-F. Wei, J. Zhang, and W.-N. Chen, “A classifier-ensemble- based surrogate-assisted evolutionary algorithm for distributed data- driven optimization,”IEEE Trans. Evol. Comput., vol. 29, no. 3, pp. 711–725, 2025

    work page 2025

  61. [61]

    Distributed data- driven evolutionary optimization with surrogate ensemble strategies,

    F.-F. Wei, Q.-Y . Zeng, J. Zhang, and W.-N. Chen, “Distributed data- driven evolutionary optimization with surrogate ensemble strategies,” IEEE Trans. Emerg. Top. Comput. Intell., Early Access, 2026

    work page 2026

  62. [62]

    Decision space partition based surrogate- assisted evolutionary algorithm for expensive optimization,

    Y . Liu, J. Liu, and S. Tan, “Decision space partition based surrogate- assisted evolutionary algorithm for expensive optimization,”Expert Syst. Appl., vol. 214, p. 119075, Mar. 2023

    work page 2023

  63. [63]

    Probability selection-based surrogate-assisted evolutionary algorithm for expensive optimization,

    S. Wang and J.-Y . Li, “Probability selection-based surrogate-assisted evolutionary algorithm for expensive optimization,”Appl. Sci. (Basel), vol. 15, no. 21, p. 11404, Oct. 2025

    work page 2025

  64. [64]

    An improved teaching-learning-based opti- mization algorithm for solving unconstrained optimization problems,

    R. V . Rao and V . Patel, “An improved teaching-learning-based opti- mization algorithm for solving unconstrained optimization problems,” Sci. Iran., vol. 20, no. 3, pp. 710–720, Jun. 2013

    work page 2013

  65. [65]

    Particle swarm optimization,

    J. Kennedy and R. Eberhart, “Particle swarm optimization,” inInt. Conf. Neural Netw., vol. 4, Nov. 1995, pp. 1942–1948

    work page 1995

  66. [66]

    AIEA: An asynchronous influence-based evolutionary algorithm for expensive many-objective optimization,

    F.-F. Wei, W.-N. Chen, and J. Zhang, “AIEA: An asynchronous influence-based evolutionary algorithm for expensive many-objective optimization,”IEEE Trans. Cybern., vol. 55, no. 2, pp. 786–799, Feb. 2025

    work page 2025

  67. [67]

    Comparative studies of metamod- elling techniques under multiple modelling criteria,

    R. Jin, W. Chen, and T. W. Simpson, “Comparative studies of metamod- elling techniques under multiple modelling criteria,”Struct. Multidiscip. Optim., vol. 23, no. 1, pp. 1–13, Dec. 2001

    work page 2001

  68. [68]

    J. H. Holland,Adaptation in Natural and Artificial Systems: An Intro- ductory Analysis with Applications to Biology, Control, and Artificial Intelligence. MIT Press, Apr. 1992

    work page 1992

  69. [69]

    A dual ensemble- surrogate assisted evolutionary algorithm with state information driven infill criterion for expensive many-objective optimization,

    W. Zhang, J. Liu, J. Tian, Y . Liu, and H. Wang, “A dual ensemble- surrogate assisted evolutionary algorithm with state information driven infill criterion for expensive many-objective optimization,”Swarm Evol. Comput., vol. 97, p. 102019, Aug. 2025

    work page 2025

  70. [70]

    Surrogate-assisted multi-objective evo- lutionary optimization with Pareto front model-based local search method,

    F. Li, L. Gao, and W. Shen, “Surrogate-assisted multi-objective evo- lutionary optimization with Pareto front model-based local search method,”IEEE Trans. Cybern., vol. 54, no. 1, pp. 173–186, Jan. 2024

    work page 2024

  71. [71]

    Evolutionary Optimization of Computationally Expensive Problems via Surrogate Modeling,

    Y . S. Ong, P. B. Nair, and A. J. Keane, “Evolutionary Optimization of Computationally Expensive Problems via Surrogate Modeling,”AIAA J., vol. 41, no. 4, pp. 687–696, Apr. 2003

    work page 2003

  72. [72]

    ExTrEMO: Transfer Evolu- tionary Multiobjective Optimization with Proof of Faster Convergence,

    J. Liu, A. Gupta, C. Ooi, and Y .-S. Ong, “ExTrEMO: Transfer Evolu- tionary Multiobjective Optimization with Proof of Faster Convergence,” IEEE Trans. Evol. Comput., vol. 29, no. 1, pp. 102–116, Feb. 2025

    work page 2025

  73. [73]

    Multi- layer perceptron grouping and sparse Gaussian process-based surrogate- assisted evolutionary algorithm for expensive multiobjective optimiza- tion,

    J.-S. Pan, Y . Peng, J. Li, J. Zhao, L. Kong, and S.-C. Chu, “Multi- layer perceptron grouping and sparse Gaussian process-based surrogate- assisted evolutionary algorithm for expensive multiobjective optimiza- tion,”IEEE Trans. Cybern., vol. 56, no. 4, pp. 2148–2161, Nov. 2025

    work page 2025

  74. [74]

    Simple and scalable predictive uncertainty estimation using deep ensembles,

    B. Lakshminarayanan, A. Pritzel, and C. Blundell, “Simple and scalable predictive uncertainty estimation using deep ensembles,” inAnnu. Conf. Neural Inf. Process. Syst. (NeurIPS), vol. 30, Dec. 2016, pp. 6402–6413

    work page 2016

  75. [75]

    Sustainable Performance Improvement of Surrogate-assisted Evolutionary Algorithms using Tabu Search,

    K. Nishihara, M. Nakata, and S. Watanabe, “Sustainable Performance Improvement of Surrogate-assisted Evolutionary Algorithms using Tabu Search,” inInt. Jt. Conf. Comput. Intell. (IJCCI), 2026, pp. 386–404

    work page 2026

  76. [76]

    Utilizing the Expected Gradient in Surrogate-assisted Evolutionary Algorithms,

    K. Nishihara and M. Nakata, “Utilizing the Expected Gradient in Surrogate-assisted Evolutionary Algorithms,” inAnnu. Conf. Genet. Evol. Comput. (GECCO) Companion, Jul. 2023, pp. 447–450

    work page 2023

  77. [77]

    Emulation-based adaptive differential evolution: fast and auto- tunable approach for moderately expensive optimization problems,

    ——, “Emulation-based adaptive differential evolution: fast and auto- tunable approach for moderately expensive optimization problems,” Complex Intell. Syst., vol. 10, no. 3, pp. 3633–3656, Jun. 2024

    work page 2024

  78. [78]

    PlatEMO: A MATLAB plat- form for evolutionary multi-objective optimization [educational forum],

    Y . Tian, R. Cheng, X. Zhang, and Y . Jin, “PlatEMO: A MATLAB plat- form for evolutionary multi-objective optimization [educational forum],” IEEE Comput. Intell. Mag., vol. 12, no. 4, pp. 73–87, Nov. 2017

    work page 2017

  79. [79]

    A Surrogate-Assisted Differential Evolution Algorithm for High-Dimensional Expensive Optimization Problems,

    W. Wang, H.-L. Liu, and K. C. Tan, “A Surrogate-Assisted Differential Evolution Algorithm for High-Dimensional Expensive Optimization Problems,”IEEE Trans. Cybern., vol. 53, no. 4, pp. 2685–2697, Apr. 2023

    work page 2023

  80. [80]

    Evolutionary algorithm with individual-distribution search strategy and regression-classification surrogates for expensive optimization,

    G. Li, L. Xie, Z. Wang, H. Wang, and M. Gong, “Evolutionary algorithm with individual-distribution search strategy and regression-classification surrogates for expensive optimization,”Inf. Sci., vol. 634, pp. 423–442, Jul. 2023

    work page 2023

Showing first 80 references.

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.