Pith. sign in

REVIEW 2 major objections 3 minor 96 references

Exploiting Structure with Anisotropic Consensus-Based Optimization

T0 review · 2 major / 3 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read Anisotropic CBO finds global minima of additively separable high-dimensional objectives with complexity that depends only on the intrinsic dimension, not the ambient one.

desk verdict Solid, correctly scoped theory paper: anisotropic CBO complexity tracks intrinsic d (not ambient D) under additive separability + coordinate alignment; the independence caveat is already the authors' own. read the letter →

arxiv 2607.10205 v1 pith:JKW7KHCA submitted 2026-07-11 math.OC

Sabrina Bonandin , Konstantin Riedl , Sara Veneruso This is my paper
classification math.OC MSC 65K1090C2690C5635Q9035Q84
keywords consensus-basedoptimizationanisotropicdiffusionadditiveseparabilitycurseofdimensionalitymean-fieldlimitglobalnonconvexderivative-freemethods
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

High-dimensional global optimization is usually cursed by dimension: cost grows exponentially with the number of variables. This paper shows that anisotropic consensus-based optimization (CBO)—a swarm of particles that mix a drift toward a consensus point with coordinate-wise random exploration—automatically exploits additive separability. When the objective is a sum of lower-dimensional pieces living on disjoint coordinate blocks, the particle system decouples into independent lower-dimensional problems. Consequently the number of particles and the temperature parameter needed for high-probability global convergence grow only with the largest block size (the intrinsic dimension), not with the full ambient dimension. The same analysis also shows that only the local geometry of each block matters, so a single poorly conditioned coordinate does not force the whole algorithm to pay the worst-case cost. Numerical heat-maps confirm that success rates collapse once the blocks are rotated away from the coordinate axes, underscoring that the noise must be aligned with the separability structure.

What carries the argument

Mean-field decoupling: under additive separability and independent initial projections, the anisotropic Fokker–Planck equation factors into L independent lower-dimensional McKean–Vlasov equations; the quantitative Laplace principle then applies block-wise, producing an α0 that is the maximum of the block-wise thresholds.

What would settle it

Take a fully separable Rastrigin function, apply a dense orthogonal rotation that mixes all coordinates, and check whether the empirical success probability of anisotropic CBO still scales only with the original one-dimensional blocks or collapses to the ambient-dimension scaling.

Watch

Extended reading notes

Core claim

For an objective that is additively separable into L lower-dimensional components of dimensions at most d, anisotropic CBO converges in high probability to the unique global minimizer with particle number and inverse-temperature that scale only with d and with the tractability constants of the individual components, independent of ambient dimension D.

Load-bearing premise

The coordinate-wise noise of the algorithm must line up with the unknown partition of the variables; if the true separable blocks are rotated relative to the axes, the decoupling fails and the dimension-reduction guarantee disappears.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 3 minor

Summary. The paper studies anisotropic consensus-based optimization (CBO) for nonsmooth nonconvex objectives that are additively separable into lower-dimensional components (Assumption 1.1). It proves that the mean-field anisotropic dynamics separate into independent lower-dimensional McKean processes under statistical independence of the projected initial measures (Theorem 3.3), yielding mean-field global convergence of the Lyapunov functional V_\infty with hyperparameters \alpha_0 = max_\ell \alpha_{0,\ell} that depend only on the intrinsic dimension d = max d_\ell and the component-wise inverse-continuity constants of Assumption 2.10 (Theorem 3.5). Combining this with the strong mean-square approximation of [16] produces a high-probability \ell_\infty bound for the practical discrete finite-particle scheme whose particle number N and consensus parameter scale with d (not ambient D) while the time-step contribution depends at most linearly on D (Theorem 3.8, Corollary 3.9). Supporting numerics on (block-rotated) Rastrigin functions illustrate the predicted degradation with increasing d and with increasing non-convexity inside the blocks.

Significance. The result is a genuine advance in the rigorous analysis of multi-agent derivative-free methods: it supplies the first explicit complexity statement showing that anisotropic CBO can break the curse of dimensionality when the objective admits an aligned additive structure. The component-wise quantitative Laplace principle (Proposition 4.7), the refined tractability assumptions, the explicit scaling \alpha_0 ~ O(d^{2} log d), and the product success probability (1-\delta_1)^L are technically clean and useful. The derivation is self-contained once the independence/alignment hypotheses are granted, builds carefully on the existing mean-field and strong-convergence literature, and is corroborated by systematic heat-map experiments. The open problem of learning the noise axes on the fly is clearly motivated by the same analysis.

major comments (2)
  1. [Theorem 3.3, proof of Theorem 3.8 (display (53)–(55))] Theorem 3.3 establishes path-wise factorization only for the mean-field laws under the independence hypothesis on {\rho_0^\ell}. The subsequent high-probability bound of Theorem 3.8 (and the product structure (1-\delta_1)^L) therefore inherits this mean-field factorization via independent copies of the McKean process. While the paper correctly flags that finite-N trajectories do not separate, a quantitative estimate of the rate at which the empirical measure approaches the product structure (or a fully microscopic argument under the same hypotheses) would make the complexity claim for the implementable algorithm more self-contained.
  2. [Corollary 3.9 and the citation to [16, Theorem 2.4]] The linear-in-D factor inside C_NA (arising from the anisotropic diffusion) appears in the admissible step-size of Corollary 3.9. Although this is only polynomial and does not restore an exponential curse, the manuscript never isolates its precise dependence; a short remark quantifying the constant would clarify that the ambient dimension truly enters only mildly.
minor comments (3)
  1. [Theorem 1.2 / Corollary 3.9] The informal statement of Theorem 1.2 and the rigorous Corollary 3.9 use slightly different normalizations of \epsilon; a single consistent expression would improve readability.
  2. [Section 5] Figures 1, 5, 7 and 9–10 are informative but the color-scale legends and exact success criterion (|E(x_\alpha)-E*| < 0.25) are stated only in the text; placing them in the captions would help.
  3. [Throughout] Notation switches between d and bold d for the intrinsic dimension; a uniform choice would avoid momentary confusion.

Circularity Check

1 steps flagged · score 1.0 of 10

No significant circularity: d-only complexity follows from explicit separability + independence assumptions via standard mean-field analysis; self-citations supply reusable tools, not the target claim.

  1. self citation load bearing [Proof of Thm 3.8 (after Eq. 50) and Cor 3.9; also Rem 3.7]
    "the last inequality follows from the strong mean-square convergence estimate in [16, Theorem 2.4], with constants CNA and CMFA depending on the constants appearing in Assumption 2.5... Building on this result, our Theorem 3.8 bounds... [16, Theorem 1.1] and [16, Corollary 1.2]"

    The finite-N/Δt error control that converts mean-field V∞ decay into the high-probability ℓ∞ bound of the implementable scheme is taken from the authors' concurrent preprint [16]. This is a genuine self-citation of a technical lemma, but it is not load-bearing for the novel d-only scaling (which already appears at the mean-field level in Thm 3.5 under Ass. 2.10); the citation merely supplies a black-box rate that any independent strong-convergence result could replace. Hence only a minor, non-central circularity of kind 3.

full rationale

The central claim (Thm 1.2/3.8/Cor 3.9) that anisotropic CBO complexity scales exponentially only in intrinsic d (via α0 = max α0,ℓ) rather than ambient D is obtained by (i) proving mean-field pathwise factorization under additive separability (Ass. 1.1) plus statistical independence of the projected initials hoℓ_0 (Thm 3.3, Eqs. 41-42, Lemma 4.1), (ii) adapting the Lyapunov/V∞ decay + quantitative Laplace argument componentwise (Thm 3.5, Prop 4.7 using Ass. 2.10 on each Eℓ, Prop 4.9), and (iii) lifting to the discrete scheme via the strong mean-square rates of the authors' prior work [16]. Hyperparameters α0,ℓ are explicit functions of the component constants (ηℓ,R0,ℓ, uℓ, hoℓ_0(B)) and ε (Eq. 81/6), never fitted to the success heatmaps of Sec. 5. The independence/alignment hypothesis is stated up front, used as a premise for decoupling, and numerically falsified under rotation (Figs. 1,5,7), so the theorems are correctly scoped rather than circular. Self-citations to [48,49,86,16] supply well-posedness, concentration and discretization estimates that are independent of the new d-scaling; they do not force the result by definition or uniqueness import. No self-definitional loop, fitted-as-prediction, or ansatz-smuggling appears. Score 1 only for the ordinary presence of overlapping-author tool citations that are not load-bearing for the novelty.

Assumptions & free parameters 4 free parameters · 6 assumptions · 1 invented entities

The central complexity claim rests on structural and analytic assumptions standard in CBO analysis plus the paper-specific additive partition and independence of projected initials. Algorithmic hyperparameters are free but their scaling is derived, not fitted to the numerical heatmaps. No new physical entities are postulated; 'intrinsic dimension' is a definition from the partition.

free parameters (4)
  • α (consensus sharpness)
    Must exceed α0 = max_ℓ α0,ℓ; α0,ℓ is derived from ε, d_ℓ, and Assumption 2.10 constants, but the working α is a free algorithmic choice above that threshold.
  • λ, σ (drift and noise intensities)
    User-chosen with constraint 2λ > σ²; enter exponential rates and constants C_NA, C_MFA.
  • N, Δt, H (particles, step, iterations)
    Chosen to meet ε_total, δ1, δ2 via Cor 3.9; free implementation parameters whose sufficient sizes are bounded by the theory.
  • ϑ ∈ (0,1) (rate slack)
    Free analysis parameter controlling the exponential decay window in Thm 3.5/3.8.
assumptions (6)
  • domain assumption Additive separability: E(x)=Σ_ℓ E_ℓ(B_ℓ^⊤ x) with partition of coordinates (Assumption 1.1).
    Load-bearing structural hypothesis; without it the dynamics do not decouple and complexity reverts to ambient D.
  • domain assumption Regularity W1–W3 on E or component-wise W1ℓ–W3ℓ (Assumptions 2.5/2.6).
    Standard for well-posedness of interacting particle and mean-field CBO systems.
  • domain assumption Inverse continuity / far-field tractability on each E_ℓ (Assumption 2.10).
    Needed for quantitative Laplace principle and unique global minimizer per component.
  • ad hoc to paper Projected initial measures ρ^ℓ_0 are statistically independent; x* ∈ supp(ρ0).
    Required for pathwise product structure in Thm 3.3 and product success probability in Thm 3.8; not automatic for arbitrary ρ0.
  • domain assumption Unique global minimizer x* of E exists.
    Stated after (3); uniqueness also follows from the tractability assumptions.
  • standard math Standard SDE/mean-field tools: weak solutions of Fokker–Planck, Grönwall, Burkholder–Davis–Gundy, Wasserstein stability of consensus points.
    Used throughout §§3–4; cited from prior CBO literature.
invented entities (1)
  • Intrinsic dimension d := max_ℓ d_ℓ
    purpose: Captures complexity of additively separable objectives better than the pair (D,L).
    Definitional quantity from the partition; not a new physical object. Independent evidence is definitional only.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exploiting Structure with Anisotropic Consensus-Based Optimization." pith.science (2026). https://pith.science/paper/JKW7KHCA

@misc{pith2026260710205,
  author       = {Pith},
  title        = {Pith review of: Exploiting Structure with Anisotropic Consensus-Based Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JKW7KHCA}},
  note         = {Machine review of arXiv:2607.10205}
}
abstract

Anisotropic consensus-based optimization (CBO), a multi-agent metaheuristic derivative-free optimization method, which reliably finds global minima of nonsmooth and nonconvex objective functions while being amenable to rigorous theoretical analysis, automatically detects and exploits additively separable structures of high-dimensional objective functions. This enables the algorithm to mitigate the curse of dimensionality where the objective function decomposes additively into lower-dimensional components. In this paper, we show this property proving that the computational complexity of anisotropic CBO depends exponentially only on the intrinsic dimension $\mathbf{d}$ of the objective function, rather than the ambient dimension $D\gg\mathbf{d}$. Additionally, we demonstrate that the computational complexity depends only on tractability conditions of the lower-dimensional components rather than on the full energy landscape, allowing for a more refined description of the objective function and algorithmic complexity, as the objective function landscape is captured directly at the level of the individual components. Our results highlight the effectiveness of anisotropic CBO for additively separable objective functions provided sufficient alignment between the structure of the anisotropic noise and the separability structure of the objective. This motivates the design of an enhanced algorithm that learns during optimization how to effectively explore the loss landscape by aligning the noise with the structure of the objective function, which we leave for future research. Numerical experiments validate our theoretical results, accentuating the influence of the intrinsic dimensionality $\mathbf{d}$, the level of separability, and the complexity and non-convexity of the objective within the separable components on performance and computational complexity of the anisotropic CBO algorithm.

Figures

Figures reproduced from arXiv: 2607.10205 by the authors.

Figure 1
Figure 1. The performance of anisotropic CBO progressively deteriorates as the complexity of the objective function (here, the intrinsic dimensionality d) increases (from left to right), see [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. The performance of anisotropic CBO progressively deteriorates as the complexity of the objective function (here, the non-convexity within the separable components) increases (from left to right), see [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Comparison between the standard Rastrigin function (fully separable) and a rotated version of the function by π/6 (non￾separable) with different levels of non-convexity. The rotation mixes the variables, thereby preventing additive separability. a11a12a13 a21a22a23 a31a32a33 b44 c55c56 c65c66 d77d78 d87d88     [PITH_FULL_IMAGE:figures/full_fig_p030_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Conceptual illustration of the block-diagonal structure of the rotation matrix R. Each colored block represents a full rotation matrix Qℓ ∈ O(dℓ). factor Qℓ to define our rotation. By varying the block sizes dℓ and the number of blocks L, we generate a family of rotate…
Figure 5
Figure 5. Figure 5: Success rate of the CBO method with anisotropic noise applied to different versions of the Rastrigin function in dimension D = 4. Each heatmap shows the success rate as a function of α and N. Results are averaged over 100 runs. The progressive deterioration of the succ…
Figure 6
Figure 6. Figure 6: shows that, in the isotropic case, the heatmaps of the standard Rastrigin function, with constant cD = 10, and its rotated version are essentially indistinguishable, confirming our expectation. This suggests that the degree of separability of the objective function doe…
Figure 7
Figure 7. Figure 7: Success rate of the CBO method with anisotropic noise applied to different versions of the Rastrigin function in dimension D = 8. Each heatmap shows the success rate as a function of α and N. Results are averaged over 50 runs. The progressive deterioration of the succe…
Figure 8
Figure 8. Figure 8: Success rate of the CBO method with isotropic noise applied to different versions of the Rastrigin function in dimension D = 8. Each heatmap shows the success rate as a function of α and N. Results are averaged over 50 runs. 5.3 Influence of the complexity of the objec…
Figure 9
Figure 9. Figure 9: Success rate of the CBO method with anisotropic noise applied to different versions of the Rastrigin function in dimension D = 4. Each heatmap shows the success rate as a function of α and N. Results are averaged over 100 runs. The figures highlight a progressive reduc…
Figure 10
Figure 10. Figure 10: Success rate of the CBO method with anisotropic noise applied to different versions of the Rastrigin function in dimension D = 8. Each heatmap shows the success rate as a function of α and N. Results are averaged over 50 runs. The figures highlight a progressive reduc…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

96 extracted references · 10 linked inside Pith

  1. [16]

    Bonandin, K

    S. Bonandin, K. Riedl, and S. Veneruso. Strong global convergence of the consensus-based optimization algorithm.arXiv preprint arXiv:2512.10654, 2025

    arXiv 2025

  2. [1]

    Aarts and J

    E. Aarts and J. Korst.Simulated annealing and Boltzmann machines. A stochastic approach to com- binatorial optimization and neural computing. Wiley-Interscience Series in Discrete Mathematics and Optimization. John Wiley & Sons, Ltd., Chichester, 1989

    1989

  3. [2]

    Aceves-Sanchez, G

    P. Aceves-Sanchez, G. Albi, F. Ferrarese, and M. Herty. Consensus-based optimization with α-stable jump processes.arXiv preprint arXiv:2604.05626, 2026

    arXiv 2026

  4. [3]

    Alcalde, L

    A. Alcalde, L. Bungert, K. Riedl, and T. Roith. Quantifying concentration phenomena of mean-field transformers in the low-temperature regime.arXiv preprint arXiv:2605.10931, 2026

    arXiv 2026

  5. [4]

    Ambrosio, N

    L. Ambrosio, N. Gigli, and G. Savar´ e.Gradient flows in metric spaces and in the space of probability measures. Lectures in Mathematics ETH Z¨ urich. Birkh¨ auser Verlag, Basel, second edition, 2008. 34

    2008

  6. [5]

    Anitescu

    M. Anitescu. Degenerate nonlinear programming with a quadratic growth condition.SIAM Journal on Optimization, 10(4):1116–1135, 2000

    2000

  7. [6]

    B¨ ack, D

    T. B¨ ack, D. B. Fogel, and Z. Michalewicz, editors.Handbook of evolutionary computation. Institute of Physics Publishing, Bristol; Oxford University Press, New York, 1997

    1997

  8. [7]

    Bailo, A

    R. Bailo, A. Barbaro, S. N. Gomes, K. Riedl, T. Roith, C. Totzeck, and U. Vaes. CBX: Python and Julia packages for consensus-based interacting particle methods.Journal of Open Source Software, 9(98):6611, 2024

    2024

Show all 96 references
  1. [8]

    Beddrich, E

    J. Beddrich, E. Chenchene, M. Fornasier, H. Huang, and B. Wohlmuth. Constrained consensus-based optimization and numerical heuristics for the few particle regime.arXiv preprint arXiv:2410.10361, 2024

    2024

  2. [9]

    Bellavia and G

    S. Bellavia and G. Malaspina. A discrete consensus-based global optimization method with noisy objective function.Journal of Optimization Theory and Applications, 206(1):20, 2025

    2025

  3. [10]

    Bellman.Dynamic Programming

    R. Bellman.Dynamic Programming. Press Princeton, 1957

    1957

  4. [11]

    Billingsley.Probability and measure

    P. Billingsley.Probability and measure. John Wiley & Sons, Inc., Hoboken, NJ, 2012

    2012

  5. [12]

    Blum and A

    C. Blum and A. Roli. Metaheuristics in combinatorial optimization: Overview and conceptual comparison. ACM Computing Surveys, 35(3):268–308, 2003

    2003

  6. [13]

    Bolte, T

    J. Bolte, T. P. Nguyen, J. Peypouquet, and B. W. Suter. From error bounds to the complexity of first-order descent methods for convex functions.Mathematical Programming, 165(2, Ser. A):471–507, 2017

    2017

  7. [14]

    Bonandin and M

    S. Bonandin and M. Herty. Consensus-based algorithms for stochastic optimization problems.SIAM Journal on Optimization, 35(4):2572–2598, 2025

    2025

  8. [15]

    Bonandin and M

    S. Bonandin and M. Herty. Kinetic variable-sample methods for stochastic optimization problems. Communications on Pure and Applied Analysis, 31:122–145, 2026

    2026

  9. [17]

    Borghi.Mean-field theory for consensus-based optimization with applications

    G. Borghi.Mean-field theory for consensus-based optimization with applications. PhD thesis, RWTH Aachen University, 2024

    2024

  10. [18]

    Borghi, S

    G. Borghi, S. Grassi, and L. Pareschi. Consensus based optimization with memory effects: random selection and applications.Chaos Solitons Fractals, 174:Paper No. 113859, 17, 2023

    2023

  11. [19]

    Borghi and M

    G. Borghi and M. Herty. Model predictive control strategies using consensus-based optimization.arXiv preprint arXiv:2312.13085, 2023

    2023 arXiv

  12. [20]

    Borghi, M

    G. Borghi, M. Herty, and L. Pareschi. A consensus-based algorithm for multi-objective optimization and its mean-field description. In2022 IEEE 61st Conference on Decision and Control (CDC), pages 4131–4136. IEEE, 2022

    2022

  13. [21]

    Borghi, M

    G. Borghi, M. Herty, and L. Pareschi. An adaptive consensus based method for multi-objective optimiza- tion with uniform Pareto front approximation.Applied Mathematics and Optimization, 88(2):Paper No. 58, 43, 2023

    2023

  14. [22]

    Borghi, M

    G. Borghi, M. Herty, and L. Pareschi. Constrained consensus-based optimization.SIAM Journal on Optimization, 33(1):211–236, 2023

    2023

  15. [23]

    Borghi, M

    G. Borghi, M. Herty, and A. Stavitskiy. Dynamics of measure-valued agents in the space of probabilities. SIAM Journal on Mathematical Analysis, 57(5):5107–5134, 2025

    2025

  16. [24]

    Borghi, H

    G. Borghi, H. Huang, and J. Qiu. A particle consensus approach to solving nonconvex-nonconcave min-max problems.arXiv preprint arXiv:2407.17373, 2024

    2024 arXiv

  17. [25]

    Bottou, F

    L. Bottou, F. E. Curtis, and J. Nocedal. Optimization methods for large-scale machine learning.SIAM Review, 60(2):223–311, 2018. 35

    2018

  18. [26]

    Bungert, F

    L. Bungert, F. Hoffmann, D. Kim, and T. Roith. MirrorCBO: A consensus-based optimization method in the spirit of mirror descent.arXiv preprint arXiv:2501.12189, 2025

    2025 arXiv

  19. [27]

    Bungert, T

    L. Bungert, T. Roith, and P. Wacker. Polarized consensus-based dynamics for optimization and sampling. Mathematical Programming, 2024

    2024

  20. [28]

    Byeon, S.-Y

    J. Byeon, S.-Y. Ha, G. Hwang, D. Ko, and J. Yoon. Consensus, error estimates and applications of first - and second-order consensus-based optimization algorithms.Mathematical Models and Methods in Applied Sciences, 35(2):345–401, 2025

    2025

  21. [29]

    J. A. Carrillo, Y.-P. Choi, C. Totzeck, and O. Tse. An analytical framework for consensus-based global optimization method.Mathematical Models and Methods in Applied Sciences, 28(6):1037–1066, 2018

    2018

  22. [30]

    J. A. Carrillo, N. Garc´ ıa Trillos, S. Li, and Y. Zhu. FedCBO: Reaching group consensus in clustered feder- ated learning through consensus-based optimization.Journal of Machine Learning Research, 25(214):1–51, 2024

    2024

  23. [31]

    J. A. Carrillo, F. Hoffmann, A. M. Stuart, and U. Vaes. Consensus-based sampling.Studies in Applied Mathematics, 148(3):1069–1140, 2022

    2022

  24. [32]

    J. A. Carrillo, S. Jin, L. Li, and Y. Zhu. A consensus-based global optimization method for high dimensional machine learning problems.ESAIM: Control, Optimisation and Calculus of Variations, 27(suppl.):Paper No. S5, 22, 2021

    2021

  25. [33]

    J. A. Carrillo, C. Totzeck, and U. Vaes. Consensus-based optimization and ensemble kalman inversion for global optimization problems with constraints. InModeling and Simulation for Collective Dynamics, pages 195–230. World Scientific, 2023

    2023

  26. [34]

    Chaintron and A

    L.-P. Chaintron and A. Diez. Propagation of chaos: a review of models, methods and applications. I. Models and methods.Kinetic and Related Models, 15(6):895–1015, 2022

    2022

  27. [35]

    Y. Chao, X. Sun, K. Riedl, M. Khadiv, and J. Etesami. Convergence of consensus-based particle methods for nonconvex bi-level optimization. 2026

    2026

  28. [36]

    J. Chen, S. Jin, and L. Lyu. A consensus-based global optimization method with adaptive momentum estimation.Communications in Computational Physics, 31(4):1296–1316, 2022

    2022

  29. [37]

    Chenchene, H

    E. Chenchene, H. Huang, and J. Qiu. A consensus-based algorithm for non-convex multiplayer games. Journal of Optimization Theory and Applications, 206(2):45, 2025

    2025

  30. [38]

    Cipriani, H

    C. Cipriani, H. Huang, and J. Qiu. Zero-inertia limit: from particle swarm optimization to consensus-based optimization.SIAM Journal on Mathematical Analysis, 54(3):3091–3121, 2022

    2022

  31. [39]

    A. R. Conn, K. Scheinberg, and L. N. Vicente.Introduction to Derivative-Free Optimization. Society for Industrial and Applied Mathematics, USA, 2009

    2009

  32. [40]

    De Deyn, M

    W. De Deyn, M. Herty, and G. Samaey. Mean-field limits for two-layer neural networks trained with consensus-based optimization.arXiv preprint arXiv:2511.21466, 2025

    2025 arXiv

  33. [41]

    R. J. de Keijzer, L. Y. Visser, O. Tse, and S. J. Kokkelmans. Consensus-based qubit configuration optimization for variational algorithms on neutral atom quantum systems.npj Quantum Information, 11(1):186, 2025

    2025

  34. [42]

    Dembo and O

    A. Dembo and O. Zeitouni.Large deviations techniques and applications, volume 38 ofApplications of Mathematics (New York). Springer-Verlag, New York, second edition, 1998

    1998

  35. [43]

    L. C. Evans.Measure theory and fine properties of functions. Chapman and Hall/CRC, 1992

    1992

  36. [44]

    Fornasier, H

    M. Fornasier, H. Huang, J. Klemenc, and G. Malaspina. From consensus-based optimization to evolution strategies: Proof of global convergence.arXiv preprint arXiv:2602.11677, 2026

    2026

  37. [45]

    Fornasier, H

    M. Fornasier, H. Huang, L. Pareschi, and P. S¨ unnen. Consensus-based optimization on hypersurfaces: Well-posedness and mean-field limit.Mathematical Models and Methods in Applied Sciences, 30(14):2725– 2751, 2020. 36

    2020

  38. [46]

    Fornasier, H

    M. Fornasier, H. Huang, L. Pareschi, and P. S¨ unnen. Consensus-based optimization on the sphere: convergence to global minimizers and machine learning.Journal of Machine Learning Research, 22:Paper No. 237, 55, 2021

    2021

  39. [47]

    Fornasier, H

    M. Fornasier, H. Huang, L. Pareschi, and P. S¨ unnen. Anisotropic diffusion in consensus-based optimization on the sphere.SIAM Journal on Optimization, 32(3):1984–2012, 2022

    1984

  40. [48]

    Fornasier, T

    M. Fornasier, T. Klock, and K. Riedl. Convergence of anisotropic consensus-based optimization in mean- field law. In J. L. J. Laredo, J. I. Hidalgo, and K. O. Babaagba, editors,Applications of Evolutionary Computation - 25th European Conference, EvoApplications 2022, Held as P...

    2022

  41. [49]

    Fornasier, T

    M. Fornasier, T. Klock, and K. Riedl. Consensus-Based Optimization Methods Converge Globally.SIAM Journal on Optimization, 34(3):2973–3004, 2024

    2024

  42. [50]

    Fornasier, P

    M. Fornasier, P. Richt´ arik, K. Riedl, and L. Sun. Consensus-based optimisation with truncated noise. European Journal of Applied Mathematics, 36(2):292–315, 2025

    2025

  43. [51]

    Fornasier and L

    M. Fornasier and L. Sun. A PDE framework of consensus-based optimization for objectives with multiple global minimizers.Communications in Partial Differential Equations, 50(4):493–541, 2025

    2025

  44. [52]

    Franceschi, L

    J. Franceschi, L. Pareschi, and M. Zanella. Superlinear drift in consensus-based optimization with condensation phenomena.arXiv preprint arXiv:2506.09001, 2025

    2025

  45. [53]

    Gandolfi and M

    M. Gandolfi and M. Ceotto. Molecular dynamics of artificially pair-decoupled systems: An accurate tool for investigating the importance of intramolecular couplings.Journal of Chemical Theory and Computation, 19(18):6093–6108, 2023. PMID: 37698951

    2023

  46. [54]

    Garc´ ıa Trillos, A

    N. Garc´ ıa Trillos, A. Kumar Akash, S. Li, K. Riedl, and Y. Zhu. Defending against diverse attacks in federated learning through consensus-based bi-level optimization.Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 383(2298)...

    2025

  47. [55]

    Gerber, F

    N. Gerber, F. Hoffmann, D. Kim, and U. Vaes. Uniform-in-time propagation of chaos for consensus-based optimization.arXiv preprint arXiv:2505.08669, 2025

    2025

  48. [56]

    N. J. Gerber, F. Hoffmann, and U. Vaes. Mean-field limits for consensus-based optimization and sampling. ESAIM: Control, Optimisation and Calculus of Variations, 31:74, 2025

    2025

  49. [57]

    Graham and D

    C. Graham and D. Talay.Stochastic simulation and Monte Carlo methods: mathematical foundations of stochastic simulation, volume 68. Springer Science & Business Media, 2013

    2013

  50. [58]

    Grassi, H

    S. Grassi, H. Huang, L. Pareschi, and J. Qiu. Mean-field particle swarm optimization. InModeling and Simulation for Collective Dynamics, pages 127–193. World Scientific, 2023

    2023

  51. [59]

    Grassi and L

    S. Grassi and L. Pareschi. From particle swarm optimization to consensus based optimization: stochastic modeling and mean-field limit.Mathematical Models and Methods in Applied Sciences, 31(8):1625–1657, 2021

    2021

  52. [60]

    S.-Y. Ha, S. Jin, and D. Kim. Convergence of a first-order consensus-based global optimization algorithm. Mathematical Models and Methods in Applied Sciences, 30(12):2417–2444, 2020

    2020

  53. [61]

    S.-Y. Ha, S. Jin, and D. Kim. Convergence and error estimates for time-discrete consensus-based optimization algorithms.Numerische Mathematik, 147(2):255–282, 2021

    2021

  54. [62]

    Hastie and R

    T. Hastie and R. Tibshirani. Generalized additive models.Statistical science, 1(3):297–310, 1986

    1986

  55. [63]

    Herty, Y

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

    2025

  56. [64]

    Herty and S

    M. Herty and S. Veneruso. Micro-macro decomposition for constrained particle swarm optimization method.Kinetic and Related Models, 19:95–118, 2025

    2025

  57. [65]

    D. J. Higham. An algorithmic introduction to numerical simulation of stochastic differential equations. SIAM Review, 43(3):525–546, 2001. 37

    2001

  58. [66]

    Huang and H

    H. Huang and H. Kouhkouh. Self-interacting CBO: Existence, uniqueness, and long-time convergence. Applied Mathematics Letters, 161:109372, 2025

    2025

  59. [67]

    Huang and H

    H. Huang and H. Kouhkouh. A derivative-free particle method for optimization in Hilbert spaces.arXiv preprint arXiv:2605.31565, 2026

    2026 arXiv

  60. [68]

    Huang, H

    H. Huang, H. Kouhkouh, and L. Sun. Faithful global convergence for the rescaled consensus-based optimization.arXiv preprint arXiv:2503.08578, 2025

    2025

  61. [69]

    Huang and J

    H. Huang and J. Qiu. On the mean-field limit for the consensus-based optimization.Math. Methods Appl. Sci., 45(12):7814–7831, 2022

    2022

  62. [70]

    Huang, J

    H. Huang, J. Qiu, and K. Riedl. On the global convergence of particle swarm optimization methods. Applied Mathematics and Optimization, 88(2):Paper No. 30, 44, 2023

    2023

  63. [71]

    Huang, J

    H. Huang, J. Qiu, and K. Riedl. Consensus-based optimization for saddle point problems.SIAM Journal on Control and Optimization, 62(2):1093–1121, 2024

    2024

  64. [72]

    Huang, M

    Y. Huang, M. Herty, D. Kalise, and N. Kantas. Fast and robust consensus-based optimization via optimal feedback control.SIAM Journal on Scientific Computing, 48(1):A74–A102, 2026

    2026

  65. [73]

    Kalise, A

    D. Kalise, A. Sharma, and M. V. Tretyakov. Consensus-based optimization via jump-diffusion stochastic differential equations.Mathematical Models and Methods in Applied Sciences, 33(2):289–339, 2023

    2023

  66. [74]

    Karimi, J

    H. Karimi, J. Nutini, and M. Schmidt. Linear convergence of gradient and proximal-gradient methods under the polyak- lojasiewicz condition. In P. Frasconi, N. Landwehr, G. Manco, and J. Vreeken, editors, Machine Learning and Knowledge Discovery in Databases - European Conferen...

    2016

  67. [75]

    Kennedy and R

    J. Kennedy and R. Eberhart. Particle swarm optimization. InProceedings of International Conference on Neural Networks (ICNN’95), Perth, WA, Australia, November 27 - December 1, 1995, pages 1942–1948. IEEE, 1995

    1995

  68. [76]

    Khatab and C

    M. Khatab and C. Totzeck. A consensus-based optimization algorithm using Gaussian processes for global optimization problems in Sobolev spaces.arXiv preprint arXiv:2603.15337, 2026

    2026

  69. [77]

    Klamroth, M

    K. Klamroth, M. Stiglmayr, and C. Totzeck. Consensus-based optimization for multi-objective problems: a multi-swarm approach.Journal of Global Optimization, pages 1–32, 2024

    2024

  70. [78]

    Mao.Stochastic differential equations and applications

    X. Mao.Stochastic differential equations and applications. Elsevier, 2007

    2007

  71. [79]

    H. P. McKean, Jr. Propagation of chaos for a class of non-linear parabolic equations. InStochastic Differential Equations (Lecture Series in Differential Equations, Session 7, Catholic Univ., 1967), volume Session 7 ofLecture Series in Differential Equations, pages 41–57. Air ...

    1967

  72. [80]

    P. D. Miller.Applied asymptotic analysis, volume 75 ofGraduate Studies in Mathematics. American Mathematical Society, Providence, RI, 2006

    2006

  73. [81]

    Necoara, Y

    I. Necoara, Y. Nesterov, and F. Glineur. Linear convergence of first order methods for non-strongly convex optimization.Mathematical Programming, 175(1-2, Ser. A):69–107, 2019

    2019

  74. [82]

    Pinnau, C

    R. Pinnau, C. Totzeck, O. Tse, and S. Martin. A consensus-based model for global optimization and its mean-field limit.Mathematical Models and Methods in Applied Sciences, 27(1):183–204, 2017

    2017

  75. [83]

    Recht, C

    B. Recht, C. R´ e, S. J. Wright, and F. Niu. Hogwild: A lock-free approach to parallelizing stochastic gradient descent. In J. Shawe-Taylor, R. S. Zemel, P. L. Bartlett, F. C. N. Pereira, and K. Q. Weinberger, editors,Advances in Neural Information Processing Systems 24: 25th ...

    2011

  76. [84]

    C. Reeves. Genetic algorithms. InHandbook of metaheuristics, volume 57 ofInternat. Ser. Oper. Res. Management Sci., pages 55–82. Kluwer Acad. Publ., Boston, MA, 2003. 38

    2003

  77. [85]

    K. Riedl. Leveraging memory effects and gradient information in consensus-based optimisation: On global convergence in mean-field law.European Journal of Applied Mathematics, 35(4):483–514, 2024

    2024

  78. [86]

    Riedl.Mathematical Foundations of Interacting Multi-Particle Systems for Optimization

    K. Riedl.Mathematical Foundations of Interacting Multi-Particle Systems for Optimization. PhD thesis, Technische Universit¨ at M¨ unchen, 2024

    2024

  79. [87]

    Riedl, T

    K. Riedl, T. Klock, C. Geldhauser, and M. Fornasier. Gradient is all you need? How consensus- based optimization can be interpreted as a stochastic relaxation of gradient descent.SIAM Journal on Mathematics of Data Science, 2026

    2026

  80. [88]

    Roith, L

    T. Roith, L. Bungert, and P. Wacker. Consensus-based optimization for closed-box adversarial attacks and a connection to evolution strategies.arXiv preprint arXiv:2506.24048, 2025

    2025 arXiv

  81. [89]

    X. Sun, A. Jordana, M. Fornasier, J. Etesami, and M. Khadiv. Consensus-based optimization: Towards global optimality in robotics.arXiv preprint arXiv:2602.06868, 2026

    2026 arXiv

  82. [90]

    Sznitman

    A.-S. Sznitman. Topics in propagation of chaos. In ´Ecole d’ ´Et´ e de Probabilit´ es de Saint-Flour XIX—1989, volume 1464 ofLecture Notes in Math., pages 165–251. Springer, Berlin, 1991

    1989

  83. [91]

    Totzeck and M.-T

    C. Totzeck and M.-T. Wolfram. Consensus -based global optimization with personal best.Mathematical Biosciences and Engineering, 17(5):6026–6044, 2020

    2020

  84. [92]

    N. G. Trillos, S. Li, K. Riedl, and Y. Zhu. CB 2O: Consensus-based bi-level optimization.arXiv preprint arXiv:2411.13394, 2024

    2024 arXiv

  85. [93]

    V. Vapnik. Statistical learning theory. new york: John willey & sons.Inc, 1998

    1998

  86. [94]

    Wahba.Spline Models for Observational Data

    G. Wahba.Spline Models for Observational Data. Society for Industrial and Applied Mathematics, 1990

    1990

  87. [95]

    D. H. Wolpert and W. G. Macready. Coevolutionary free lunches.IEEE Trans. Evol. Comput., 9(6):721– 735, 2005

    2005

  88. [96]

    Zhang, Y

    H. Zhang, Y. Ma, R. Kitichotkul, J. Rapp, and P. T. Boufounos. ProxiCBO: A consensus-based method for composite optimization. InProc. IEEE Int. Conf. Acoust., Speech, and Signal Process. (Barcelona, Spain, 2026), 2026. 39

    2026

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.