Pith. sign in

REVIEW 3 major objections 4 minor 43 references

This paper shows that the candidate-generation step of two classic population optimizers—SOMA and Differential Evolution—is, conditional on the sampled state and random masks, an exact linear or affine transformation, and derives closed-for

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 · deepseek-v4-flash

2026-08-03 11:12 UTC pith:LZS27E7H

load-bearing objection Solid, carefully scoped analysis of SOMA/DE proposal geometry with verified closed-form moments; the main caveat is a labeling issue around the 'canonical' SOMA mask assumption. the 3 major comments →

arxiv 2607.29228 v1 pith:LZS27E7H submitted 2026-07-31 cs.NE cs.AI

Linear Proposal Operators and Stochastic Search Geometry in SOMA and Differential Evolution

Vojt\v{e}ch Nov\'ak , Ivan Zelinka This is my paper
classification cs.NE cs.AI MSC 68T2090C59
keywords SOMADifferential Evolutionlinear proposal operatorsstochastic search geometryBernoulli masksfinite-population momentscrossover covarianceBBOB benchmark
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.

Population-based optimizers are usually treated as nonlinear black boxes because selection, replacement, and rank-based decisions are discontinuous. This paper tries to establish that the hidden variation stage is not: for canonical SOMA, each candidate proposal is an exact linear map of an augmented migrant–leader state, and relative to the leader it is simply a coordinate-wise rescaling. Treating the perturbation mask as independent Bernoulli random variables yields closed-form formulas for the proposal mean, covariance, expected step length, expected squared leader distance, active dimensionality, and coordinate coverage; for DE/rand/1/bin the paper derives matching finite-population moments including the extra covariance induced by forced-coordinate crossover. If these results hold, parameter setting (PRT, path parameter, F, CR) can be replaced by direct geometric targeting, and algorithm design can be guided by proposal-level quantities. The paper verifies the identities numerically and shows that repair and selection, while unavoidable, are separate transformations that can substantially distort the raw proposals.

Core claim

The central claim is that the nonlinearity of a full optimization iteration lives in selection, target construction, boundary repair, and adaptation, not in the variation operator. Once the migrant, leader, path parameter, and perturbation mask are fixed, the SOMA proposal z = x + tΠ(L−x) is exactly linear in the augmented state (x,L) through the block matrix M(t,Π), and relative to the leader it reduces to a coordinate-wise rescaling by (1−t) on active coordinates. Under independent Bernoulli masks, the paper derives closed forms: E[z] = x + pt(L−x), covariance p(1−p)t² diag(r²), normalized expected squared leader distance 1−pt(2−t), active dimensionality Dp, and per-coordinate coverage 1−(

What carries the argument

The load-bearing object is the augmented migrant–leader state s = (x,L) in R^{2D} and the block matrix M(t,Π) = [[I−tΠ, tΠ],[0,I]], which turns the proposal into an exact linear map; in leader-relative coordinates it collapses to the diagonal rescaling I−tΠ. For the stochastic analysis, the Bernoulli perturbation mask Π = diag(π_d) with π_d ~ Bernoulli(p) is the random operator that produces all the closed-form moments. For DE, the analogous machinery is the augmented donor/target state and the forced-coordinate binomial crossover mask, whose marginal activation probability becomes q⋆ = q + (1−q)/D and whose covariance has off-diagonal entries −(1−q)²/D².

Load-bearing premise

The closed-form results describe the raw proposal stage under exactly the stated procedural variants—independent regenerated Bernoulli masks and donor sampling without replacement—and they do not by themselves predict the full algorithm because boundary repair and fitness-based selection can substantially reshape the proposal distribution.

What would settle it

Run canonical SOMA with one fixed perturbation mask reused across the entire path; if per-coordinate coverage does not rise with path length J as 1−(1−p)^J but stays close to p, then the independent-regeneration assumption behind Eq. (13) fails. Similarly, sampling DE donors with replacement would change the finite-population covariance factor, providing a direct check of the mutation-moment formula.

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

If this is right

  • SOMA's PRT, path parameter, and path length become designable geometric quantities: expected active dimensionality, covariance, step length, and coordinate coverage are all computable in closed form.
  • The same separation applies to DE/rand/1/bin: finite-population mutation moments and forced-coordinate crossover moments fully describe the raw trial distribution, including the cross-coordinate covariance that the forced coordinate introduces.
  • Because the proposal operator is objective-independent, the derived geometry holds for any fitness function before repair and selection, providing a common basis for comparing masking and covariance across different algorithms.
  • Operator-guided variants—geometry-controlled, rotation-aware, and a composite iSOMA extension—improve canonical SOMA on the noiseless benchmark, with rotation-aware masking strongest at low dimension and high budget.
  • Boundary repair and fitness-based selection can substantially change raw proposal means and covariances, so the exact results describe the variation stage only; predicting full optimizer behavior requires modeling repair and selection as well.

Where Pith is reading between the lines

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

  • If raw proposal moments are exact, then selection and repair are the only remaining sources of nonlinearity; one could try to characterize fitness-based selection as an operator on the proposal covariance, estimated from acceptance rates, to close the gap between raw and selected distributions.
  • The coverage formula suggests a testable design rule: choose path length J and PRT so that 1−(1−p)^J covers a desired fraction of coordinates, turning exploration budgets into explicit geometric targets.
  • Rotation-aware masking in a population-covariance basis is promising in low dimensions, but in higher dimensions the empirical basis may be noisy; shrinkage, low-rank, or rank-one-updated bases are natural extensions to test.
  • The DE/SOMA correspondence via masked affine difference operators hints at a unified analysis of other strategies, such as current-to-best or all-to-random, through the same augmented-state machinery.

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. This paper proposes an operator–selection factorization for SOMA and DE/rand/1/bin, separating raw candidate generation from boundary repair and fitness-based selection. For SOMA, the fixed-mask proposal z = x + tΠ(L−x) is written as an exact linear map on an augmented state (Eq. 4), reducing in leader-relative coordinates to coordinate-wise rescaling by (1−t) (Eq. 7). Under independent Bernoulli perturbation masks, the paper derives closed forms for the proposal mean, covariance, expected squared step length, expected squared leader distance, active dimensionality, and path-level coordinate coverage (Eqs. 9–13). For DE/rand/1/bin, it derives finite-population donor moments with an exact sampling-without-replacement covariance factor, and forced-coordinate binomial crossover moments (Appendices A.3–A.4). The analysis motivates three new SOMA variants: GC-SOMA, RA-SOMA, and an adaptive composite iL-SHOMA-RA, benchmarked on the full noiseless BBOB suite at D∈{5,10,20} and budgets 500D/2500D. The empirical results show improvements over canonical SOMA and mixed but competitive performance against iL-SHADE and SciPy-DE.

Significance. If accepted, the paper provides a clean, reusable decomposition of two widely used optimizers. Its strongest assets are the parameter-free exact identities, the finite-population DE correction factor (1 + 2F²M/(M−1)), and the exact forced-coordinate crossover covariance with negative off-diagonal entries; all are verified by exact enumeration to floating-point precision, and the code is made available. The boundary-repair and selection-distortion studies in Appendix A.5 are honest about the scope of the raw-proposal analysis. The paper also makes a credible design point: proposal-level geometry can be converted into algorithmic controls, even though the composite iL-SHOMA-RA is correctly labeled exploratory. The central derivation is sound; the main risks are terminological overreach and benchmark confounds rather than mathematical error.

major comments (3)
  1. [Section 2.1; Eq. (13); Table C.9]
  2. [Section 5.1; Eq. (21); Appendix A.4]
  3. [Section 6.1; Table 4; Table C.9]
minor comments (4)
  1. [Figure 1]
  2. [Section 5.1, Eq. (21)]
  3. [Section 6.2, Table 5]
  4. [Appendix A.5]

Circularity Check

0 steps flagged

No significant circularity: the proposal moments are parameter-free consequences of the explicitly stated algorithm definitions and are independently verified by enumeration/Monte Carlo.

full rationale

The paper's derivations are self-contained conditional analyses of explicitly stated procedural rules. The SOMA proposal z = x + tΠ(L−x) (Eq. 2) is rewritten exactly as the augmented linear map in Eq. (4) and as leader-relative rescaling in Eq. (7); these are algebraic identities, not fitted results. The closed-form moments (Eqs. 9–13) follow directly from the independent-Bernoulli mask assumption, with no fitted constants, and are checked by exact enumeration and Monte Carlo simulation. The DE results (Appendixes A.3–A.5) are similarly derived from ordered donor sampling without replacement and forced-coordinate binomial crossover, and verified to floating-point precision by enumeration. The GC-SOMA parameterization in Eq. (23) is an inversion of the derived moment relation, not a prediction derived from a fit. The self-citations ([10], [41–43]) are background or future-work references and are not load-bearing for the proposal-moment derivations. The only substantive caveat is the ambiguity of the 'canonical' SOMA mask semantics: the paper explicitly notes competing mask-reuse conventions, and Eq. (13) is conditional on independent regeneration at J path points. Appendix A.2 quantifies the reuse case, and Appendix A.5 explicitly qualifies the raw-proposal scope against boundary repair and selection. These are definitional-consistency and scope limitations, not circular reductions; the claimed derivations do not assume their own conclusions.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

The analytical moment results are parameter-free given the algorithm definitions listed above. The free parameters in this ledger are the hand-set schedules of the newly proposed optimizer variants; they do not enter the theorems but they do enter the benchmark claims. No new physical or algorithmic entities are postulated.

free parameters (3)
  • RA-SOMA mask parameter and path values = PRT = 0.30; t in {0.4, 0.8, 1.2, 1.6, 2.0, 2.4}
    Hand-set in Appendix C to illustrate mask/covariance geometry; not derived from theory and not tuned per BBOB function, but affects the reported RA-SOMA results.
  • GC-SOMA schedule = p = 0.35 + 0.55 s; rho_c = 0.85 - 0.70 s; overshoot branch for s < 0.55
    Budget-dependent schedule chosen by hand in Appendix C; paper labels GC-SOMA a proof of concept, so the performance claim is conditional on this schedule.
  • iL-SHOMA-RA adaptation constants = memory size 6, alpha = 0.7/0.3, Cauchy scale gamma = 0.1, N_min = D, jump = 10
    Inherited from SHADE/iSOMA conventions; because the method introduces several mechanisms at once, the paper does not attribute gains to any single constant.
axioms (5)
  • domain assumption SOMA PRT masks are independent Bernoulli(p) entries, regenerated independently at each path point.
    Used for Eqs. (9)-(13); coverage formula (13) explicitly fails under mask reuse, which the paper notes.
  • domain assumption DE donor indices r1, r2, r3 are sampled uniformly, ordered, without replacement, from the target-excluded population.
    Underlies the finite-population mutant covariance in Appendix A.3; the factor M/(M-1) changes under sampling with replacement.
  • domain assumption Binomial crossover forces exactly one coordinate from the mutant (j_rand).
    Produces the q* and Gamma corrected moments in Appendix A.4; without the forced coordinate the effective marginal is q, not q + (1-q)/D.
  • domain assumption All-zero masks are allowed in the raw proposal distribution.
    The unconditioned Binomial(K) and coverage formulas assume this; K > 0 conditioning is derived separately in Appendix A.2.
  • domain assumption BBOB noiseless instances 1-5 and the fixed seeds are representative for rank comparisons.
    Empirical claims about competitiveness rest on this sampling; the paper reports statistical tests on 120 blocks.

pith-pipeline@v1.3.0-daily-deepseek · 24586 in / 15180 out tokens · 155108 ms · 2026-08-03T11:12:15.208750+00:00 · methodology

0 comments
read the original abstract

Swarm and evolutionary algorithms are usually analyzed as complete procedural systems in which nonlinear selection, replacement, and adaptation obscure simpler structure within candidate generation. This paper introduces an operator--selection factorization that separates objective-independent variation from boundary repair and fitness-dependent selection, and uses it to study the proposal geometry of the Self-Organizing Migrating Algorithm (SOMA) and Differential Evolution (DE). The canonical SOMA proposal is shown to be affine in the search space and exactly linear in an augmented migrant--leader state. In leader-relative coordinates, the resulting operator provides a direct interpretation of interpolation, projection, overshooting, and coordinate masking. Under Bernoulli perturbation masks, we derive closed-form expressions for the proposal mean, covariance, expected squared step length, expected squared distance from the leader, active dimensionality, and coordinate coverage. For canonical DE/rand/1/bin, we derive the finite-population moments of differential mutation and characterize the additional covariance and coordinate dependence induced by forced-coordinate binomial crossover. Exact enumeration and Monte Carlo experiments verify the analytical identities and quantify the effects of mask conditioning, boundary repair, and fitness-based selection. The analysis further motivates geometry-controlled and rotation-aware SOMA variants, together with an adaptive population-reducing extension of iSOMA. Experiments on the complete noiseless BBOB benchmark show that these operator-guided variants substantially improve upon canonical SOMA and are competitive with established DE methods in several dimension--budget regimes. The results demonstrate how proposal-level operator analysis can support both the interpretation and design of population-based optimizers.

Figures

Figures reproduced from arXiv: 2607.29228 by Ivan Zelinka, Vojt\v{e}ch Nov\'ak.

Figure 1
Figure 1. Figure 1: Discrete SOMA proposal paths and selected replacements on the Three-Hump Camel function for three successive [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Variation–selection decomposition of one SOMA migration. The current population first undergoes [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Bernoulli-mask SOMA proposal geometry for fixed migrant [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Influence of the differential weight F and crossover rate CR on the donor-marginalized proposal geometry of DE/rand/1/bin. Both panels use the same population and target vector xi. Gray markers denote the current population, blue points represent possible mutant vectors v = xr1 + F(xr2 − xr3 ), and red points represent the corresponding trial vectors after standard forced-coordinate binomial crossover. The… 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

43 extracted references · 13 canonical work pages

  1. [1]

    Storn, K

    R. Storn, K. Price, Differential evolution—a simple and efficient heuristic for global optimization over continuous spaces, Journal of Global Optimization 11 (4) (1997) 341–359.doi:10.1023/A:1008202821328

    doi:10.1023/a:1008202821328 1997

  2. [2]

    Davendra, I

    D. Davendra, I. Zelinka, Self-Organizing Migrating Algorithm: Methodology and Implementation, Springer International Publish- ing, Cham, 2016.doi:10.1007/978-3-319-28161-2

    work page doi:10.1007/978-3-319-28161-2 2016

  3. [3]

    Skanderova, Self-organizing migrating algorithm: Review, improvements and comparison, Artificial Intelligence Review 56 (1) (2023) 101–172.doi:10.1007/s10462-022-10167-8

    L. Skanderova, Self-organizing migrating algorithm: Review, improvements and comparison, Artificial Intelligence Review 56 (1) (2023) 101–172.doi:10.1007/s10462-022-10167-8

    work page doi:10.1007/s10462-022-10167-8 2023

  4. [4]

    A. E. Eiben, J. E. Smith, Introduction to Evolutionary Computing, 2nd Edition, Springer, Berlin, Heidelberg, 2015.doi:10.1007/ 978-3-662-44874-8

    2015

  5. [5]

    I. Zelinka, A survey on evolutionary algorithms dynamics and its complexity—mutual relations, past, present and future, Swarm and Evolutionary Computation 25 (2015) 2–14.doi:10.1016/j.swevo.2015.06.002

    work page doi:10.1016/j.swevo.2015.06.002 2015

  6. [6]

    Doerr, F

    B. Doerr, F. Neumann, A survey on recent progress in the theory of evolutionary algorithms for discrete optimization, ACM Transactions on Evolutionary Learning and Optimization 1 (4) (2021) 1–43.doi:10.1145/3472304

    work page doi:10.1145/3472304 2021

  7. [7]

    S. Das, P. N. Suganthan, Differential evolution: A survey of the state-of-the-art, IEEE Transactions on Evolutionary Computation 15 (1) (2011) 4–31.doi:10.1109/TEVC.2010.2059031

    arXiv 2011

  8. [8]

    S. Das, S. S. Mullick, P. N. Suganthan, Recent advances in differential evolution—an updated survey, Swarm and Evolutionary Computation 27 (2016) 1–30.doi:10.1016/j.swevo.2016.01.004

    work page doi:10.1016/j.swevo.2016.01.004 2016

  9. [9]

    K. R. Opara, J. Arabas, Differential evolution: A survey of theoretical analyses, Swarm and Evolutionary Computation 44 (2019) 546–558.doi:10.1016/j.swevo.2018.06.010

    work page doi:10.1016/j.swevo.2018.06.010 2019

  10. [10]

    Novák, T

    V. Novák, T. Bezděk, I. Zelinka, S. Das, M. Beseda, A longitudinal analysis of the CEC single-objective competitions (2010–2024) and implications for variational quantum optimization, Swarm and Evolutionary Computation (2026) 102469doi:10.1016/j.swevo. 2026.102469

    arXiv 2010

  11. [11]

    K. R. Opara, J. Arabas, Comparison of mutation strategies in differential evolution—a probabilistic perspective, Swarm and Evolutionary Computation 39 (2018) 53–69.doi:10.1016/j.swevo.2017.12.007

    work page doi:10.1016/j.swevo.2017.12.007 2018

  12. [12]

    K. R. Opara, J. Arabas, The contour fitting property of differential mutation, Swarm and Evolutionary Computation 50 (2019) 100441.doi:10.1016/j.swevo.2018.09.001

    work page doi:10.1016/j.swevo.2018.09.001 2019

  13. [13]

    Zaharie, Influence of crossover on the behavior of differential evolution algorithms, Applied Soft Computing 9 (3) (2009) 1126–1138.doi:10.1016/j.asoc.2009.02.012

    D. Zaharie, Influence of crossover on the behavior of differential evolution algorithms, Applied Soft Computing 9 (3) (2009) 1126–1138.doi:10.1016/j.asoc.2009.02.012

    work page doi:10.1016/j.asoc.2009.02.012 2009

  14. [14]

    Pluháček, A

    M. Pluháček, A. Kazikova, T. Kadavy, A. Viktorin, R. Senkerik, Explaining SOMA: The relation of stochastic perturbation to population diversity and parameter space coverage, in: Proceedings of the Genetic and Evolutionary Computation Conference Companion, Association for Computing Machinery, New York, NY, USA, 2021, pp. 1944–1952.doi:10.1145/3449726.3463211

    arXiv 2021

  15. [15]

    Biedrzycki, J

    R. Biedrzycki, J. Arabas, D. Jagodziński, Bound constraints handling in differential evolution: An experimental study, Swarm and Evolutionary Computation 50 (2019) 100453.doi:10.1016/j.swevo.2018.10.004

    work page doi:10.1016/j.swevo.2018.10.004 2019

  16. [16]

    J. M. Tomczak, E. Węglarz-Tomczak, A. E. Eiben, Differential evolution with reversible linear transformations, in: Proceedings of the 2020 Genetic and Evolutionary Computation Conference Companion, Association for Computing Machinery, New York, NY, USA, 2020, pp. 205–206, extended version available as arXiv:2002.02869.doi:10.1145/3377929.3389972. 33 Table...

    Pith/arXiv arXiv 2020

  17. [17]

    Q. B. Diep, T. C. Truong, S. Das, I. Zelinka, Self-organizing migrating algorithm with narrowing search space strategy for robot path planning, Applied Soft Computing 116 (2022) 108270.doi:10.1016/j.asoc.2021.108270

    arXiv 2022

  18. [18]

    Brest, S

    J. Brest, S. Greiner, B. Bošković, M. Mernik, V. Žumer, Self-adapting control parameters in differential evolution: A comparative study on numerical benchmark problems, IEEE Transactions on Evolutionary Computation 10 (6) (2006) 646–657.doi:10.1109/ TEVC.2006.872133

    arXiv 2006

  19. [19]

    A. K. Qin, V. L. Huang, P. N. Suganthan, Differential evolution algorithm with strategy adaptation for global numerical optimiza- tion, IEEE Transactions on Evolutionary Computation 13 (2) (2009) 398–417.doi:10.1109/TEVC.2008.927706

    arXiv 2009

  20. [20]

    Zhang, A

    J. Zhang, A. C. Sanderson, JADE: Adaptive differential evolution with optional external archive, IEEE Transactions on Evolu- tionary Computation 13 (5) (2009) 945–958.doi:10.1109/TEVC.2009.2014613

    arXiv 2009

  21. [21]

    Tanabe, A

    R. Tanabe, A. S. Fukunaga, Success-history based parameter adaptation for differential evolution, in: 2013 IEEE Congress on Evolutionary Computation, IEEE, 2013, pp. 71–78.doi:10.1109/CEC.2013.6557555

    arXiv 2013

  22. [22]

    Tanabe, A

    R. Tanabe, A. S. Fukunaga, Improving the search performance of SHADE using linear population size reduction, in: 2014 IEEE Congress on Evolutionary Computation, IEEE, 2014, pp. 1658–1665.doi:10.1109/CEC.2014.6900380

    arXiv 2014

  23. [23]

    Brest, M

    J. Brest, M. Sepesy Maučec, B. Bošković, iL-SHADE: Improved L-SHADE algorithm for single-objective real-parameter optimiza- tion, in: 2016 IEEE Congress on Evolutionary Computation, IEEE, 2016, pp. 1188–1195.doi:10.1109/CEC.2016.7743922

    arXiv 2016

  24. [24]

    Guo, C.-C

    S.-M. Guo, C.-C. Yang, Enhancing differential evolution utilizing eigenvector-based crossover operator, IEEE Transactions on Evolutionary Computation 19 (1) (2015) 31–49.doi:10.1109/TEVC.2013.2297160

    arXiv 2015

  25. [25]

    Caraffini, F

    F. Caraffini, F. Neri, A study on rotation invariance in differential evolution, Swarm and Evolutionary Computation 50 (2019) 100436.doi:10.1016/j.swevo.2018.08.013

    work page doi:10.1016/j.swevo.2018.08.013 2019

  26. [26]

    Huxley, Evolution: The Modern Synthesis, George Allen & Unwin, London, 1942

    J. Huxley, Evolution: The Modern Synthesis, George Allen & Unwin, London, 1942

    1942

  27. [27]

    Hansen, A

    N. Hansen, A. Ostermeier, Completely derandomized self-adaptation in evolution strategies, Evolutionary Computation 9 (2) (2001) 159–195.doi:10.1162/106365601750190398

    doi:10.1162/106365601750190398 2001

  28. [28]

    Hansen, R

    N. Hansen, R. Ros, N. Mauny, M. Schoenauer, A. Auger, Impacts of invariance in search: When CMA-ES and PSO face ill- conditioned and non-separable problems, Applied Soft Computing 11 (8) (2011) 5755–5769.doi:10.1016/j.asoc.2011.03.001

    work page doi:10.1016/j.asoc.2011.03.001 2011

  29. [29]

    Virtanen, R

    P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman, P. van Mulbregt, SciPy 1.0 Contributors, SciPy 1.0: Fundamental algorithms for scientific computing in python, Nature Methods 17 (2020) 261–272.doi:10.1038/s41592-019-0686-2

    doi:10.1038/s41592-019-0686-2 2020

  30. [30]

    URLhttps://pypi.org/project/pyade-python/

    xKuZz, PyADE: Python advanced differential evolution algorithms library, PyPI package, version 1.1, accessed 31 July 2026 (2019). URLhttps://pypi.org/project/pyade-python/

    2026

  31. [31]

    J. J. Moré, S. M. Wild, Benchmarking derivative-free optimization algorithms, SIAM Journal on Optimization 20 (1) (2009) 172–191.doi:10.1137/080724083

    doi:10.1137/080724083 2009

  32. [32]

    Beiranvand, W

    V. Beiranvand, W. Hare, Y. Lucet, Best practices for comparing optimization algorithms, Optimization and Engineering 18 (4) (2017) 815–848.doi:10.1007/s11081-017-9366-1

    doi:10.1007/s11081-017-9366-1 2017

  33. [33]

    Hansen, A

    N. Hansen, A. Auger, R. Ros, O. Mersmann, T. Tušar, D. Brockhoff, COCO: A platform for comparing continuous optimizers in a black-box setting, Optimization Methods and Software 36 (1) (2021) 114–144.doi:10.1080/10556788.2020.1808977

    arXiv 2021

  34. [34]

    Finck, N

    S. Finck, N. Hansen, R. Ros, A. Auger, Real-parameter black-box optimization benchmarking 2009: Noiseless functions definitions, Tech. Rep. RR-6829, INRIA, updated version as of February 2019 (2009). URLhttps://inria.hal.science/inria-00362633v2/document

    2009

  35. [35]

    Hansen, A

    N. Hansen, A. Auger, S. Finck, R. Ros, Real-parameter black-box optimization benchmarking 2009: Experimental setup, Tech. Rep. RR-6828, INRIA (2009). URLhttps://inria.hal.science/inria-00362649v2/document

    2009

  36. [36]

    Jamil, X.-S

    M. Jamil, X.-S. Yang, A literature survey of benchmark functions for global optimisation problems, International Journal of Mathematical Modelling and Numerical Optimisation 4 (2) (2013) 150–194.doi:10.1504/IJMMNO.2013.055204

    arXiv 2013

  37. [37]

    Derrac, S

    J. Derrac, S. García, D. Molina, F. Herrera, A practical tutorial on the use of nonparametric statistical tests as a methodology for comparing evolutionary and swarm intelligence algorithms, Swarm and Evolutionary Computation 1 (1) (2011) 3–18.doi: 10.1016/j.swevo.2011.02.002

    doi:10.1016/j.swevo.2011.02.002 2011

  38. [38]

    García, A

    S. García, A. Fernández, J. Luengo, F. Herrera, Advanced nonparametric tests for multiple comparisons in the design of experiments in computational intelligence and data mining: Experimental analysis of power, Information Sciences 180 (10) (2010) 2044–2064. doi:10.1016/j.ins.2009.12.010

    doi:10.1016/j.ins.2009.12.010 2010

  39. [39]

    Hansen, A

    N. Hansen, A. Auger, D. Brockhoff, T. Tušar, Anytime performance assessment in blackbox optimization benchmarking, IEEE Transactions on Evolutionary Computation 26 (6) (2022) 1293–1305.doi:10.1109/TEVC.2022.3210897

    arXiv 2022

  40. [40]

    T. J. Choi, An efficient eigenvector-based crossover for differential evolution: Simplifying with rank-one updates, AIMS Mathe- matics 10 (2) (2025) 3500–3522.doi:10.3934/math.2025162

    work page doi:10.3934/math.2025162 2025

  41. [41]

    Illésová, V

    S. Illésová, V. Novák, T. Bezděk, C. Possel, M. Beseda, Numerical optimization strategies for the variational hamiltonian ansatz in noisy quantum environments (2025).arXiv:2505.22398,doi:10.48550/arXiv.2505.22398

    doi:10.48550/arxiv.2505.22398 2025

  42. [42]

    Novák, S

    V. Novák, S. Illésová, T. Bezděk, I. Zelinka, M. Beseda, Reliable optimization under noise in quantum variational algorithms (2025).arXiv:2511.08289,doi:10.48550/arXiv.2511.08289

    doi:10.48550/arxiv.2511.08289 2025

  43. [43]

    Novák, I

    V. Novák, I. Zelinka, V. Snášel, Optimization strategies for variational quantum algorithms in noisy landscapes, Evolutionary IntelligenceAccepted for publication; preprint available as arXiv:2506.01715 (2026).arXiv:2506.01715. 35

    arXiv 2026