From Mean-Field Limits to Semiclassical Concentration: Global Convergence of the Canonical Evolutionary Strategy

Luis Mart\'i; Mat\'ias Neto; Nayat Sanchez-Pi; Nicol\'as Garay

arxiv: 2605.30371 · v1 · pith:J3L5C6MLnew · submitted 2026-05-19 · 💻 cs.NE · cs.LG· math.DS

From Mean-Field Limits to Semiclassical Concentration: Global Convergence of the Canonical Evolutionary Strategy

Mat\'ias Neto , Nicol\'as Garay , Luis Mart\'i , Nayat Sanchez-Pi This is my paper

Pith reviewed 2026-06-30 17:57 UTC · model grok-4.3

classification 💻 cs.NE cs.LGmath.DS

keywords canonical evolutionary strategymean-field limitreplicator-mutator equationgeometric selectionsemiclassical limitglobal convergencesurvival of the flattest

0 comments

The pith

The Canonical Evolutionary Strategy converges globally because its mean-field replicator-mutator limit concentrates according to the principal eigenfunction.

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

The paper formulates the Canonical Evolutionary Strategy as a controlled framework and derives a rigorous hierarchy from discrete individual-based dynamics to a deterministic mean-field replicator-mutator equation. Global convergence in this limit is shown to be governed by the principal eigenfunction of the operator, a property the authors call Geometric Selection. This mechanism naturally favors robust, flat optima over narrow local traps and permits intrinsic mass transport even when the global solution lies outside the initial population support. High-dimensional benchmarks confirm advantages over consensus-driven and gradient methods that suffer premature variance collapse.

Core claim

We provide a rigorous hierarchy from a discrete individual-based dynamics to a deterministic mean-field limit, demonstrating that global convergence is governed by the principal eigenfunction of the underlying operator. This property, defined as Geometric Selection, naturally prioritizes robust, flat optima over narrow local traps, offering a mathematical justification for the survival of the flattest phenomenon. Moreover, the replicator-mutator dynamics of CES facilitate intrinsic mass transport, unlike consensus-driven methods.

What carries the argument

Geometric Selection: prioritization of the principal eigenfunction of the operator in the semiclassical limit of the replicator-mutator equation.

Load-bearing premise

The discrete individual-based dynamics of the Canonical Evolutionary Strategy converge to the deterministic mean-field replicator-mutator equation.

What would settle it

A simulation on a landscape with a known principal eigenfunction where the finite-population CES fails to concentrate mass at the predicted flat global optimum while the mean-field equation succeeds.

Figures

Figures reproduced from arXiv: 2605.30371 by Luis Mart\'i, Mat\'ias Neto, Nayat Sanchez-Pi, Nicol\'as Garay.

**Figure 1.** Figure 1: Mean-Field Limit Validation. (a): The population distribution νM(t) follows the PDE dynamics, escaping the local trap. (b): The time-averaged Wasserstein distance between the stochastic CES and the replicator-mutator PDE as a function of population size M. The solid line denotes the mean error over 30 independent seeds, while the shaded area represents the ±1 standard deviation. The log-log scale highlight… view at source ↗

**Figure 2.** Figure 2: Semiclassical Concentration (σ1). Evolution of the stationary density ϕ1 for variable c with fixed a = 40. The mass progressively concentrates at the global maximum, effectively bypassing local trap as h → 0. 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 x 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Prin cip al Eig e nfu n ctio n 1(x) Concentration of 1 for fixed c=0.04 1(x) (c=0.04) Global Maximum Fitness (x) shape 0.0 0.1 0.2… view at source ↗

**Figure 3.** Figure 3: Stationary distributions in complex landscapes ( [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Population mass transport in the Shifted scenario ( [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗

**Figure 5.** Figure 5: Convergence rates for the Shifted d = 30 scenario. CBO reach a premature plateau, while CES maintains a consistent descent. – Failure of Baseline Consensus: The CBO’s failure in the shifted case in d = 2 (CBO 0%vs. CES 43%) suggests that the consensus mechanism is highly sensitive to the initial “gravitational” center of the population. Scalability and Convergence Rates In high-dimensional settings (d = 30… view at source ↗

read the original abstract

We address the issue of global convergence in stochastic continuous optimization. For that purpose, we formulate the Canonical Evolutionary Strategy (CES) as a controlled mathematical framework to analyze global convergence in evolutionary algorithms via the semiclassical limit of a Schr{\"o}dinger-type replicator-mutator equation. We provide a rigorous hierarchy from a discrete individual-based dynamics to a deterministic mean-field limit, demonstrating that global convergence is governed by the principal eigenfunction of the underlying operator. This property, defined as Geometric Selection, naturally prioritizes robust, flat optima over narrow local traps, offering a mathematical justification for the ''survival of the flattest'' phenomenon. Moreover, unlike consensus-driven methods that are prone to premature variance collapse when the global minimizer resides outside the initial support, the replicator-mutator dynamics of CES facilitate intrinsic mass transport. High-dimensional benchmarks (d = 30) confirm this advantage, showing that CES achieves lower residual errors in shifted initialization scenarios where standard consensus-driven and gradient-based methods fail to migrate effectively. By shifting the focus from point-wise consensus to spectral concentration, our framework provides a robust theoretical foundation for global convergence in Evolution Strategies (ES) without the need for additional numerical heuristics.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper sketches a mean-field to semiclassical hierarchy for global convergence in evolutionary strategies via the principal eigenfunction, but leaves the derivations and quantitative evidence unshown.

read the letter

The core claim is that the Canonical Evolutionary Strategy can be cast as a controlled system whose mean-field limit is a replicator-mutator equation, after which a semiclassical limit shows that the principal eigenfunction drives concentration at flat minima. They label this Geometric Selection and argue it explains both global reach and the preference for robust optima over sharp traps. The contrast with consensus methods, which collapse variance and cannot move mass outside the initial support, is the practical payoff they highlight.

What the work does is assemble those three layers—discrete particles, mean-field PDE, semiclassical spectral picture—into one narrative and point to high-dimensional runs where CES keeps moving when other methods stall. That synthesis is not standard in the ES literature they cite.

The soft spots sit right at the center. The abstract asserts a rigorous hierarchy and eigenfunction-governed convergence, yet supplies no operator definitions, no limit theorem statements, and no error analysis. The benchmark claims are stated without numbers, baselines, run counts, or statistical detail, so the reported advantage in shifted-initialization cases cannot be judged. The load-bearing step is the passage from discrete individual-based dynamics to the deterministic mean-field equation; without seeing how regularity conditions or parameter regimes are handled, it is impossible to know whether the subsequent spectral argument rests on solid ground.

This is for readers already working on mean-field models or spectral methods in stochastic optimization. Someone looking for a new angle on why ES can exhibit global behavior might find the framing useful, but the current version does not yet let a reader verify the steps. It deserves a serious referee so the derivations can be examined directly.

Referee Report

2 major / 1 minor

Summary. The paper claims to derive a rigorous hierarchy from discrete individual-based Canonical Evolutionary Strategy (CES) dynamics to a deterministic mean-field replicator-mutator PDE and then to a semiclassical limit of a Schrödinger-type equation; global convergence is asserted to be governed by the principal eigenfunction of the underlying operator (termed Geometric Selection), which explains the survival of the flattest and enables intrinsic mass transport, with supporting high-dimensional (d=30) benchmarks showing lower residual errors than consensus or gradient methods under shifted initializations.

Significance. If the hierarchy and spectral analysis are rigorously established, the work would supply a mathematical foundation linking evolutionary dynamics to semiclassical concentration, offering a parameter-free explanation for preference of flat optima and a justification for global convergence without ad-hoc heuristics; this could influence theoretical analysis of evolution strategies and related population-based optimizers.

major comments (2)

[Abstract / hierarchy derivation] Abstract and the section on the hierarchy: the central claim of a 'rigorous hierarchy' from discrete individual-based dynamics through the mean-field replicator-mutator equation to the semiclassical limit is asserted without any proof sketches, explicit operator definitions, regularity conditions, or error bounds; this is load-bearing for the global-convergence result.
[High-dimensional benchmarks] Benchmark section: the claim that CES achieves lower residual errors in shifted-initialization scenarios (d=30) is presented without quantitative error values, baseline comparisons, statistical measures, or variance reporting, which undermines the empirical support for the advantage over consensus-driven methods.

minor comments (1)

[Introduction / model formulation] Notation for the replicator-mutator operator and the semiclassical scaling parameter should be introduced with explicit definitions early in the text to aid readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments highlighting the need for greater detail on the hierarchy and benchmarks. We address each point below and will revise the manuscript to strengthen these aspects.

read point-by-point responses

Referee: [Abstract / hierarchy derivation] Abstract and the section on the hierarchy: the central claim of a 'rigorous hierarchy' from discrete individual-based dynamics through the mean-field replicator-mutator equation to the semiclassical limit is asserted without any proof sketches, explicit operator definitions, regularity conditions, or error bounds; this is load-bearing for the global-convergence result.

Authors: The full manuscript derives the mean-field replicator-mutator PDE from the individual-based CES via a propagation-of-chaos argument under bounded fitness and mutation kernels, with the operator explicitly given as the sum of a diffusion (mutation) term and a multiplication (selection) term. The semiclassical limit follows from a WKB ansatz yielding the Schrödinger-type equation whose principal eigenfunction governs the concentration. We agree that the abstract is high-level and that explicit sketches, regularity assumptions (e.g., Lipschitz fitness), and quantitative error bounds between levels would strengthen the presentation. In revision we will insert a concise proof outline together with the stated conditions and bounds. revision: partial
Referee: [High-dimensional benchmarks] Benchmark section: the claim that CES achieves lower residual errors in shifted-initialization scenarios (d=30) is presented without quantitative error values, baseline comparisons, statistical measures, or variance reporting, which undermines the empirical support for the advantage over consensus-driven methods.

Authors: We accept that the benchmark section currently reports only qualitative superiority. The revision will add a table with mean residual errors and standard deviations over repeated runs, direct numerical comparisons against the cited consensus and gradient baselines, and explicit variance measures for the d=30 shifted-initialization experiments. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper's central claim is a claimed rigorous derivation chain from discrete individual-based CES dynamics through a mean-field replicator-mutator PDE to a semiclassical limit, with convergence governed by the principal eigenfunction (termed Geometric Selection). The abstract and description present this as an operator-theoretic construction supported by high-dimensional benchmarks, without any quoted equations or steps that reduce by construction to fitted inputs, self-definitions, or load-bearing self-citations. The derivation is self-contained against external mathematical benchmarks and falsifiable claims, yielding no circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the existence of the mean-field limit and the validity of the semiclassical approximation for the replicator-mutator dynamics. No free parameters or new postulated entities are stated in the abstract.

axioms (2)

domain assumption The discrete individual-based dynamics converge to a deterministic mean-field replicator-mutator equation
Invoked when stating the rigorous hierarchy from discrete to mean-field limit.
domain assumption The semiclassical limit of the Schrödinger-type replicator-mutator equation governs global convergence via its principal eigenfunction
Used to define Geometric Selection and mass transport.

pith-pipeline@v0.9.1-grok · 5762 in / 1406 out tokens · 35632 ms · 2026-06-30T17:57:17.011750+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

22 extracted references · 15 canonical work pages · 1 internal anchor

[1]

Journal of Dynamics and Differential Equations 31, 2029–2052 (2019).https: //doi.org/10.1007/s10884-018-9692-9

Alfaro, M., Veruete, M.: Evolutionary branching via replicator-mutator equations. Journal of Dynamics and Differential Equations 31, 2029–2052 (2019).https: //doi.org/10.1007/s10884-018-9692-9

work page doi:10.1007/s10884-018-9692-9 2029
[2]

Series in Mathematical and Computational Biology, John Wiley & Sons, Chichester, UK (2003)

Cantrell, R.S., Cosner, C.: Spatial Ecology via Reaction-Diffusion Equations. Series in Mathematical and Computational Biology, John Wiley & Sons, Chichester, UK (2003)

2003
[3]

Mathematical Models and Meth- ods in Applied Sciences 28(06), 1037–1066 (2018).https://doi.org/10.1142/ S0218202518500276

Carrillo, J.A., Choi, Y.P., Totzeck, C., Tse, O.: An analytical framework for consensus-based global optimization method. Mathematical Models and Meth- ods in Applied Sciences 28(06), 1037–1066 (2018).https://doi.org/10.1142/ S0218202518500276

2018
[4]

Theoretical Population Bi- ology 69(3), 297–321 (2006).https://doi.org/10.1016/j.tpb.2005.10.004

Champagnat, N., Ferrière, R., Méléard, S.: Unifying evolutionary dynamics: From individual stochastic processes to macroscopic models. Theoretical Population Bi- ology 69(3), 297–321 (2006).https://doi.org/10.1016/j.tpb.2005.10.004

work page doi:10.1016/j.tpb.2005.10.004 2006
[5]

Coville, J.: Convergence to equilibrium for positive solutions of some mutation- selection model (2013),https://arxiv.org/abs/1308.6471

work page internal anchor Pith review Pith/arXiv arXiv 2013
[6]

Journal of Mathematical Biology70(4), 829–854 (2015)

Fontbona, J., Méléard, S.: Non local lotka-volterra system with cross-diffusion in an heterogeneous medium. Journal of Mathematical Biology70(4), 829–854 (2015). https://doi.org/10.1007/s00285-014-0781-z

work page doi:10.1007/s00285-014-0781-z 2015
[7]

Fornasier, M., Huang, H., Klemenc, J., Malaspina, G.: From consensus-based opti- mizationtoevolutionstrategies:Proofofglobalconvergence.arXivpreprint(2026), https://arxiv.org/abs/2602.11677

work page arXiv 2026
[8]

The Annals of Applied Prob- ability 14(4) (2004)

Fournier, N., Méléard, S.: A microscopic probabilistic description of a locally reg- ulated population and macroscopic approximations. The Annals of Applied Prob- ability 14(4) (2004). https://doi.org/10.1214/105051604000000882

work page doi:10.1214/105051604000000882 2004
[9]

Springer (2013)

Gendreau, M., Potvin, J.Y.: Handbook of Metaheuristics. Springer (2013)

2013
[10]

Nesterov

Hansen, N., Auger, A., Ros, R., Mersmann, O., Tušar, T., Brockhoff, D.: Coco: a platform for comparing continuous optimizers in a black-box setting. Optimiza- tion Methods and Software 36(1), 114–144 (2020).https://doi.org/10.1080/ 10556788.2020.1808977

work page arXiv 2020
[11]

Evolutionary Computation9(2), 159–195 (2001).https://doi.org/10

Hansen, N., Ostermeier, A.: Completely derandomized self-adaptation in evolution strategies. Evolutionary Computation9(2), 159–195 (2001).https://doi.org/10. 1162/106365601750190398

2001
[12]

Mathematical Models and Methods in Applied Sciences 35(10), 2207–2243 (Jun 2025)

Herty, M., Huang, Y., Kalise, D., Kouhkouh, H.: A multiscale consensus-based algorithm for multilevel optimization. Mathematical Models and Methods in Applied Sciences 35(10), 2207–2243 (Jun 2025). https://doi.org/10.1142/ s021820252550037x

2025
[13]

Neural Computation9(1), 1–42 (01 1997)

Hochreiter, S., Schmidhuber, J.: Flat minima. Neural Computation9(1), 1–42 (01 1997). https://doi.org/10.1162/neco.1997.9.1.1

work page doi:10.1162/neco.1997.9.1.1 1997
[14]

Springer Berlin, Heidel- berg (2016)

Méléard, S.: Modèles aléatoires en Ecologie et Evolution. Springer Berlin, Heidel- berg (2016)

2016
[15]

Interface Focus3(6), 20130054 (12 2013).https://doi.org/ 10.1098/rsfs.2013.0054, https://doi.org/10.1098/rsfs.2013.0054

Morozov, A.: Modelling biological evolution: recent progress, current challenges and future direction. Interface Focus3(6), 20130054 (12 2013).https://doi.org/ 10.1098/rsfs.2013.0054, https://doi.org/10.1098/rsfs.2013.0054

work page doi:10.1098/rsfs.2013.0054 2013
[16]

Pathiraja, S., Wacker, P.: Mathematical description of continuous time and space replicator-mutator equations for quadratic fitness landscapes (2024), https:// arxiv.org/abs/2412.08178

work page arXiv 2024
[17]

Mathematical Models and Methods in Applied Sciences 27(01), 183–204 (2017).https://doi.org/10.1142/s0218202517400061 16 M

Pinnau,R.,Totzeck,C.,Tse,O.,Martin,S.:Aconsensus-basedmodelforglobalop- timization and its mean-field limit. Mathematical Models and Methods in Applied Sciences 27(01), 183–204 (2017).https://doi.org/10.1142/s0218202517400061 16 M. Neto et al

work page doi:10.1142/s0218202517400061 2017
[18]

Roith, T., Bungert, L., Wacker, P.: Consensus-based optimization for closed- box adversarial attacks and a connection to evolution strategies (2025),https: //arxiv.org/abs/2506.24048

work page arXiv 2025
[19]

Japan Journal of Industrial and Applied Mathematics 34, 473–488 (2017).https://doi.org/10.1007/s13160-017-0249-9

Wakano, J.Y., Funaki, T., Yokoyama, S.: Derivation of replicator–mutator equa- tions from a model in population genetics. Japan Journal of Industrial and Applied Mathematics 34, 473–488 (2017).https://doi.org/10.1007/s13160-017-0249-9

work page doi:10.1007/s13160-017-0249-9 2017
[20]

Weiss, Pierre: L’hypothèse du champ moléculaire et la propriété ferromagnétique. J. Phys. Theor.Appl.6(1), 661–690 (1907).https://doi.org/10.1051/jphystap: 019070060066100, https://doi.org/10.1051/jphystap:019070060066100

work page doi:10.1051/jphystap: 1907
[21]

URLhttps://doi.org/10.2514/1.G001921

Williams, G., Aldrich, A., Theodorou, E.A.: Model predictive path integral control: From theory to parallel computation. Journal of Guidance, Control, and Dynamics 40(2), 344–357 (2017).https://doi.org/10.2514/1.G001921

work page doi:10.2514/1.g001921 2017
[22]

Zworski, M.: Semiclassical Analysis, Graduate Studies in Mathematics, vol. 138. American Mathematical Society, Providence, RI (2012)

2012

[1] [1]

Journal of Dynamics and Differential Equations 31, 2029–2052 (2019).https: //doi.org/10.1007/s10884-018-9692-9

Alfaro, M., Veruete, M.: Evolutionary branching via replicator-mutator equations. Journal of Dynamics and Differential Equations 31, 2029–2052 (2019).https: //doi.org/10.1007/s10884-018-9692-9

work page doi:10.1007/s10884-018-9692-9 2029

[2] [2]

Series in Mathematical and Computational Biology, John Wiley & Sons, Chichester, UK (2003)

Cantrell, R.S., Cosner, C.: Spatial Ecology via Reaction-Diffusion Equations. Series in Mathematical and Computational Biology, John Wiley & Sons, Chichester, UK (2003)

2003

[3] [3]

Mathematical Models and Meth- ods in Applied Sciences 28(06), 1037–1066 (2018).https://doi.org/10.1142/ S0218202518500276

Carrillo, J.A., Choi, Y.P., Totzeck, C., Tse, O.: An analytical framework for consensus-based global optimization method. Mathematical Models and Meth- ods in Applied Sciences 28(06), 1037–1066 (2018).https://doi.org/10.1142/ S0218202518500276

2018

[4] [4]

Theoretical Population Bi- ology 69(3), 297–321 (2006).https://doi.org/10.1016/j.tpb.2005.10.004

Champagnat, N., Ferrière, R., Méléard, S.: Unifying evolutionary dynamics: From individual stochastic processes to macroscopic models. Theoretical Population Bi- ology 69(3), 297–321 (2006).https://doi.org/10.1016/j.tpb.2005.10.004

work page doi:10.1016/j.tpb.2005.10.004 2006

[5] [5]

Coville, J.: Convergence to equilibrium for positive solutions of some mutation- selection model (2013),https://arxiv.org/abs/1308.6471

work page internal anchor Pith review Pith/arXiv arXiv 2013

[6] [6]

Journal of Mathematical Biology70(4), 829–854 (2015)

Fontbona, J., Méléard, S.: Non local lotka-volterra system with cross-diffusion in an heterogeneous medium. Journal of Mathematical Biology70(4), 829–854 (2015). https://doi.org/10.1007/s00285-014-0781-z

work page doi:10.1007/s00285-014-0781-z 2015

[7] [7]

Fornasier, M., Huang, H., Klemenc, J., Malaspina, G.: From consensus-based opti- mizationtoevolutionstrategies:Proofofglobalconvergence.arXivpreprint(2026), https://arxiv.org/abs/2602.11677

work page arXiv 2026

[8] [8]

The Annals of Applied Prob- ability 14(4) (2004)

Fournier, N., Méléard, S.: A microscopic probabilistic description of a locally reg- ulated population and macroscopic approximations. The Annals of Applied Prob- ability 14(4) (2004). https://doi.org/10.1214/105051604000000882

work page doi:10.1214/105051604000000882 2004

[9] [9]

Springer (2013)

Gendreau, M., Potvin, J.Y.: Handbook of Metaheuristics. Springer (2013)

2013

[10] [10]

Nesterov

Hansen, N., Auger, A., Ros, R., Mersmann, O., Tušar, T., Brockhoff, D.: Coco: a platform for comparing continuous optimizers in a black-box setting. Optimiza- tion Methods and Software 36(1), 114–144 (2020).https://doi.org/10.1080/ 10556788.2020.1808977

work page arXiv 2020

[11] [11]

Evolutionary Computation9(2), 159–195 (2001).https://doi.org/10

Hansen, N., Ostermeier, A.: Completely derandomized self-adaptation in evolution strategies. Evolutionary Computation9(2), 159–195 (2001).https://doi.org/10. 1162/106365601750190398

2001

[12] [12]

Mathematical Models and Methods in Applied Sciences 35(10), 2207–2243 (Jun 2025)

Herty, M., Huang, Y., Kalise, D., Kouhkouh, H.: A multiscale consensus-based algorithm for multilevel optimization. Mathematical Models and Methods in Applied Sciences 35(10), 2207–2243 (Jun 2025). https://doi.org/10.1142/ s021820252550037x

2025

[13] [13]

Neural Computation9(1), 1–42 (01 1997)

Hochreiter, S., Schmidhuber, J.: Flat minima. Neural Computation9(1), 1–42 (01 1997). https://doi.org/10.1162/neco.1997.9.1.1

work page doi:10.1162/neco.1997.9.1.1 1997

[14] [14]

Springer Berlin, Heidel- berg (2016)

Méléard, S.: Modèles aléatoires en Ecologie et Evolution. Springer Berlin, Heidel- berg (2016)

2016

[15] [15]

Interface Focus3(6), 20130054 (12 2013).https://doi.org/ 10.1098/rsfs.2013.0054, https://doi.org/10.1098/rsfs.2013.0054

Morozov, A.: Modelling biological evolution: recent progress, current challenges and future direction. Interface Focus3(6), 20130054 (12 2013).https://doi.org/ 10.1098/rsfs.2013.0054, https://doi.org/10.1098/rsfs.2013.0054

work page doi:10.1098/rsfs.2013.0054 2013

[16] [16]

Pathiraja, S., Wacker, P.: Mathematical description of continuous time and space replicator-mutator equations for quadratic fitness landscapes (2024), https:// arxiv.org/abs/2412.08178

work page arXiv 2024

[17] [17]

Mathematical Models and Methods in Applied Sciences 27(01), 183–204 (2017).https://doi.org/10.1142/s0218202517400061 16 M

Pinnau,R.,Totzeck,C.,Tse,O.,Martin,S.:Aconsensus-basedmodelforglobalop- timization and its mean-field limit. Mathematical Models and Methods in Applied Sciences 27(01), 183–204 (2017).https://doi.org/10.1142/s0218202517400061 16 M. Neto et al

work page doi:10.1142/s0218202517400061 2017

[18] [18]

Roith, T., Bungert, L., Wacker, P.: Consensus-based optimization for closed- box adversarial attacks and a connection to evolution strategies (2025),https: //arxiv.org/abs/2506.24048

work page arXiv 2025

[19] [19]

Japan Journal of Industrial and Applied Mathematics 34, 473–488 (2017).https://doi.org/10.1007/s13160-017-0249-9

Wakano, J.Y., Funaki, T., Yokoyama, S.: Derivation of replicator–mutator equa- tions from a model in population genetics. Japan Journal of Industrial and Applied Mathematics 34, 473–488 (2017).https://doi.org/10.1007/s13160-017-0249-9

work page doi:10.1007/s13160-017-0249-9 2017

[20] [20]

Weiss, Pierre: L’hypothèse du champ moléculaire et la propriété ferromagnétique. J. Phys. Theor.Appl.6(1), 661–690 (1907).https://doi.org/10.1051/jphystap: 019070060066100, https://doi.org/10.1051/jphystap:019070060066100

work page doi:10.1051/jphystap: 1907

[21] [21]

URLhttps://doi.org/10.2514/1.G001921

Williams, G., Aldrich, A., Theodorou, E.A.: Model predictive path integral control: From theory to parallel computation. Journal of Guidance, Control, and Dynamics 40(2), 344–357 (2017).https://doi.org/10.2514/1.G001921

work page doi:10.2514/1.g001921 2017

[22] [22]

Zworski, M.: Semiclassical Analysis, Graduate Studies in Mathematics, vol. 138. American Mathematical Society, Providence, RI (2012)

2012