Pith. sign in

REVIEW 4 major objections 6 minor 5 cited by

CB$^2$O: Consensus-Based Bi-Level Optimization

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A particle consensus method provably solves nonconvex bi-level optimization problems, converging exponentially to the unique good global minimizer.

desk verdict Genuinely new CBO variant with a serious mean-field convergence proof; the gap to the finite-particle algorithm and the strength of Assumption A5 are the real caveats. read the letter →

arxiv 2411.13394 v2 pith:3PRAAASY submitted 2024-11-20 math.OC math.AP

Nicolás García Trillos , Sixu Li , Konstantin Riedl , Yuhua Zhu This is my paper · ORCID
classification math.OCmath.AP MSC 65K1090C2690C5635Q9035Q84
keywords bi-leveloptimizationconsensus-basedmean-fieldlimitFokker-Planckequationsderivative-freenonconvexquantileselectionLaplaceprinciple
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper introduces a derivative-free, particle-based method for simple bi-level optimization: minimize an upper-level objective $G$ over the set of global minimizers of a lower-level objective $L$, with both functions possibly nonconvex. The algorithm steers particles toward a consensus point computed in two steps: keep only the fraction $\beta$ of particles with the lowest values of $L$, then average those survivors with Gibbs weights $\exp(-\alpha G)$. The paper proves that in the mean-field limit the particle law satisfies a nonlinear nonlocal Fokker-Planck equation, that regular solutions exist, and that for sufficiently small $\beta$ and large $\alpha$ the law converges exponentially fast in Wasserstein-2 distance to a Dirac mass at the unique good minimizer $\theta^*_{\mathrm{good}}$. Numerical experiments on constrained optimization and sparse representation learning demonstrate the method in practice.

What carries the argument

The load-bearing object is the consensus point $m^{G,L}_{\alpha,\beta}(\rho) = \int \theta \, \exp(-\alpha G(\theta)) / \|\exp(-\alpha G)\|_{L^1(I^L_\beta[\rho])} \, dI^L_\beta[\rho](\theta)$, where $I^L_\beta[\rho]$ is $\rho$ restricted to the quantile set $Q^L_\beta[\rho] = \{\theta \in B_R(0) : L(\theta) \le \tfrac{2}{\beta}\int_{\beta/2}^{\beta} q^L_a[\rho]\, da + \delta_q\}$; the implemented version uses the simpler sublevel set of the $\beta$-quantile of $L$. The paper's convergence proof combines a quantitative quantiled Laplace principle (Q2LP, Proposition 4.2) with the time evolution of $V(\rho_t) = \tfrac12 W_2^2(\rho_t, \delta_{\theta^*_{\mathrm{good}}})$, controlling $\|m^{G,L}_{\alpha,\beta}(\rho_t) - \theta^*_{\mathrm{good}}\|_2$ as a sum of two terms: one made small by $\beta$ (how well the quantile set approximates $\Theta$) and one made small by $\alpha$ (how well the soft-min approximates the minimizer of $G$ on that set). Existence of solutions is carried by a stability estimate for the consensus point under a combination of Wasserstein and $L^2$ perturbations, replacing the standard Wasserstein Lipschitz estimate that fails for the quantile-selected consensus point.

What would settle it

Run the mean-field Fokker-Planck dynamics on a two-dimensional problem where $L$ has two global minimizers at points $a$ and $b$, $G(a) < G(b)$, but on some radius-$r$ neighborhood of $\{a,b\}$ the minimizer of $G$ lies at a point in $B_r(b)$ rather than near $a$; with $\beta$ and $\alpha$ chosen as in Theorem 2.7, observe that $W_2^2(\rho_t, \delta_a)$ stops decreasing and does not reach the claimed tolerance $\varepsilon$, contradicting the exponential convergence statement.

Watch

Extended reading notes

Core claim

The central claim is Theorem 2.7: under assumptions A1-A5 on $L$ and $G$, for any target accuracy $\varepsilon$, if the quantile parameter $\beta$ is small enough, the inverse-temperature parameter $\alpha$ is large enough, and the drift-to-noise ratio satisfies $2\lambda > d\sigma^2$, then every sufficiently regular weak solution $\rho$ of the mean-field Fokker-Planck equation reaches $W_2^2(\rho_T, \delta_{\theta^*_{\mathrm{good}}}) = \varepsilon$ at some time $T$ in an explicitly bounded interval, and on the way to that time the squared Wasserstein distance decays at least exponentially with rate $(1-\vartheta)(2\lambda - d\sigma^2)$. The mechanism is the consensus point $m^{G,L}_{\alpha,\beta}(\rho)$: a $\beta$-quantile sublevel set of $L$ approximates the minimizer set $\Theta$, and a soft-min weighted average with weights $\exp(-\alpha G)$ picks out the point in that set that is best for $G$. The paper also proves (Theorem 2.4) existence of weak solutions with regularity, using a new stability estimate for the consensus point under combined Wasserstein and $L^2$ perturbations, since the quantile truncation destroys the pure Wasserstein stability that standard consensus-based optimization enjoys.

Load-bearing premise

The upper-level objective must be stably minimized on every sufficiently small neighborhood of the entire set of lower-level minimizers: the minimizer of $G$ on that neighborhood must stay within the neighborhood's radius of the target point, and away from that ball the value of $G$ must be larger by a fixed gap; if this structure fails, the quantile-selected consensus point can track a wrong $G$-minimizer near the lower-level solution set.

Editorial extensions

If this is right

  • For nonconvex simple bi-level problems, a derivative-free particle method converges in mean-field law to the unique good minimizer with an explicit exponential rate once $\beta$ is small and $\alpha$ is large.
  • Constrained global optimization can be handled without projection or penalty tuning when the constraint is encoded as a lower-level objective; experiments on circular and star constraints support this.
  • The method is invariant under increasing transformations of $L$, matching the invariance of the feasible set $\Theta$.
  • The existence theorem supplies regular weak solutions to the associated nonlocal Fokker-Planck equation, extending consensus-based optimization well-posedness to quantile-selected consensus points.
  • High-dimensional sparse representation learning experiments show the method is tractable in practice and that choosing $\beta$ so that $\lceil \beta N \rceil = 2$ gives a simple and effective tuning rule.

Reading between the lines

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

  • The two-scale selection mechanism, where $\beta$ identifies an approximate minimizer set and $\alpha$ chooses among its elements, is a generic recipe for constrained consensus methods and could extend to constraints encoded by stochastic or adversarially chosen lower-level objectives.
  • The combined Wasserstein plus $L^2$ stability estimate for the consensus point is a transferable analytical tool for other selection-based interacting particle dynamics, such as quantile-truncated sampling or min-max consensus methods.
  • The experiments suggest a testable scaling prediction: the optimal quantile parameter for finite particles should roughly satisfy $\beta N$ constant, aligning with the practical rule $\lceil \beta N \rceil = 2$.
  • A practical diagnostic for the hard structural condition on $G$ would be to shrink $\beta$ slightly and check whether the consensus point jumps to a different element of $\Theta$; a jump would indicate that the stability condition underlying the convergence theorem fails.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes CB2O, a consensus-based particle metaheuristic for simple bi-level optimization problems of the form (1.1), where the upper-level objective G is minimized over the minimizer set Θ of a possibly nonconvex lower-level objective L. The algorithm computes a consensus point by selecting a lower-level quantile of particles and applying a Gibbs/Laplace weight in the upper-level objective. The main theoretical results are (i) an existence theorem for weak solutions of the mean-field Fokker-Planck equation (Theorem 2.4), obtained via a new stability estimate for the consensus point under a combination of Wasserstein and L2 perturbations, and (ii) a mean-field convergence theorem (Theorem 2.7) showing that, under Assumptions A1–A5 and suitable hyperparameter choices, the Wasserstein-2 distance to the target θ*_good decays at least exponentially until it reaches a prescribed accuracy ε. Numerical experiments on constrained optimization, sparse representation learning, and clustered federated learning are reported, with code provided. The paper is transparent about several open problems, notably the mean-field approximation of the finite-particle system and non-uniqueness of mean-field solutions.

Significance. If the main results hold, the paper makes a valuable contribution to derivative-free global optimization for bilevel problems: it proposes a principled quantile-based selection mechanism, proves existence and convergence for the associated mean-field dynamics, and provides explicit, a priori hyperparameter conditions. The stability estimate of Proposition 3.2 and the quantitative quantiled Laplace principle of Proposition 4.2 are technical novelties that may be useful beyond this paper. The numerical section is extensive and the authors provide open-source code, which supports reproducibility. However, the central convergence theorem is a statement about the mean-field PDE, not about the finite-particle algorithm implemented in practice, and its key assumption A5 is strong and not verified on the numerical examples. These points substantially temper the practical claims.

major comments (4)
  1. [Assumption A5; Eqs. (2.15a)–(2.15b); Proposition 4.2; Theorem 2.7] Assumption A5 is load-bearing for Theorem 2.7, since Proposition 4.2 uses it to guarantee that the minimizer ˜θgood of G on Q_L^β[ϱ] ⊂ N_{r_G}(Θ) lies within r_G of θ*_good, and the error bound (4.18) collapses without it. Yet A5 is not implied by A1, which only says θ*_good uniquely minimizes G on Θ. For example, if Θ consists of two separated global minimizers θ_bad and θ*_good with G(θ_bad) < G(θ*_good), then for every sufficiently small r_G the minimizer of G on N_{r_G}(Θ) is near θ_bad, at distance roughly 2r_G or larger from θ*_good, so (2.15b) fails while A1 holds. The manuscript does not verify A5, or the related A3, for any of the problems in Section 5; in particular the sparse representation experiment in Section 5.3 involves a neural-network landscape where such inverse-continuity conditions are highly nontrivial to check. The theorem's applicability to the reported experiments is therefore unestablished. I ask the authors to either prove A5 in the numerical settings, provide a tractable sufficient condition, or explicitly state the verified domain of Theorem 2.7.
  2. [Remark 1.1; Theorem 2.7; Section 6; Section 5.1] The central convergence guarantee (Theorem 2.7) applies to any sufficiently regular weak solution ρ of the mean-field Fokker-Planck equation (1.13), not to the finite-particle Algorithm 1. The paper explicitly leaves the mean-field approximation limit (1.14)/(1.15) as an open problem, and Section 6 lists it among future work. Additionally, the algorithm implemented in Section 5 uses the unregularized consensus point (1.6), whereas the analyzed dynamics use the regularized quantile set (2.3) with truncation radius R, slack δ_q, and averaged quantile. Thus there are two gaps between the proved statement and the implemented method: finite N and regularized-versus-implemented consensus. Since the abstract and Section 1.1 present CB2O as a method that 'converges' to θ*_good, the manuscript should either prove or at least clearly delineate a finite-particle or propagation-of-chaos result, or restrict the claims to the mean-field regularized dynamics and state this limitation in the abstract.
  3. [Section 5.2, Section 5.3; Assumptions A2–A5] The numerical experiments compare CB2O with several baselines and report good precision, but they do not check the theoretical assumptions A2–A5 for the chosen L and G. For the star constraint (5.7b), the lower-level objective has a nonconvex one-dimensional minimizer set and it is not demonstrated that A3 (the inverse-continuity/error-bound condition (2.13a)) holds. For the sparse representation task, both L and G are computed from stochastic mini-batches and random network initializations, and it is not clear that A5 holds for the actual objective landscape. Since Theorem 2.7 is the theoretical justification for the algorithm, the experimental section should include at least a discussion of why the assumptions are plausibly satisfied, or present numerical evidence for the key inequalities (2.13a), (2.15a)–(2.15b) on the relevant regions.
  4. [Theorem 2.4; Section 3.3] The existence theorem does not prove uniqueness of weak solutions of the mean-field dynamics, and the proof relies on a compactness argument that yields only a subsequential limit. While the paper acknowledges this (Section 6) and Theorem 2.7 is stated for 'any weak solution', the lack of uniqueness means the convergence statement does not identify a canonical limit dynamics and the continuity requirement on t ↦ m_{α,β}^{G,L}(ρ_t) is an extra regularity condition. This is not fatal, but it should be highlighted more prominently in the statement of Theorem 2.7 and in the discussion of the algorithm's theoretical guarantees.
minor comments (6)
  1. [Section 1 (around Eq. (1.8))] The word 'obtaiened' in the sentence following Eq. (1.8) is a typo.
  2. [Section 4.3 (Eq. (4.43))] The definition of er_ε uses 'max_{s∈[0,r_G,ε]}' with a condition in the argument; this should be written as a supremum over admissible s and seems to be intended as the largest s such that sup_{θ∈B_s(θ*_good)}(G(θ)-G(θ*_good)) ≤ u_ε.
  3. [Section 4.3 and Remark 4.3] The notation eG_r and eGr is introduced in the proof of Proposition 4.2 without a clear definition; please define it explicitly before first use.
  4. [Section 5.2, Tables 1–4] The precision numbers are reported to two decimals in some rows (e.g., 4×10⁻³) and to three decimals in others; a uniform significant-digit convention would improve readability.
  5. [Section 2.2, Assumption 2.3] In W3 the variable q is used both as a threshold and later as the quantile function q_a^L; this notational overlap could confuse readers.
  6. [Section 6 and Remark 1.1] The open-problem paragraph lists the mean-field approximation as future work, but the abstract's final sentence presents the convergence result without this caveat; please align the abstract with the actual scope of the theorem.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the convergence theorem is proved with explicit error bounds and a priori hyperparameter choices; self-citation to prior CBO work supplies only non-load-bearing lemmas.

full rationale

The paper's central claim (Theorem 2.7) is that every sufficiently regular weak solution of the mean-field Fokker-Planck equation (1.13) converges exponentially fast to a Dirac measure at theta*_good under Assumptions A1-A5 with suitable a priori hyperparameter choices. Walking the derivation chain: Lemma 4.1 derives the differential inequality for V(rho_t) = (1/2) W2^2(rho_t, delta_{theta*_good}) by direct computation from the weak formulation (2.5) with the test function (1/2)||theta - theta*_good||^2; Proposition 4.2 (the quantitative quantiled Laplace principle) is proved from Assumptions A2-A5 with explicit error terms, the first controlled by the inverse-continuity conditions A3/A5 and the second by the quantitative Laplace principle combined with a lower bound on the mass near theta*_good; Proposition 4.4 imports [51, Proposition 4.6], a published result whose stated assumptions concern standard CBO and neither contain nor presuppose the present target statement; and the proof of Theorem 2.7 assembles these ingredients with explicit formulas for delta_q, beta_0, and alpha_0 that depend only on the initial distribution and problem constants. The error decomposition (2.22) is a triangle-inequality bound whose two terms are respectively controlled by A5 (via the quantile beta) and the Laplace principle (via alpha); it is not an identity forced by the definition of the consensus point, since the claim that the quantile-restricted, G-weighted average mG,L_alpha,beta(rho_t) approximates theta*_good is a theorem (Proposition 4.2) under structural conditions, not a definitional consequence. No fitted parameter is relabeled as a prediction, and the hyperparameters are not tuned to the convergence statement. The admitted gap (Remark 1.1) that no mean-field approximation result (1.14)-(1.15) is proven for CB2O is a limitation of the theory relative to the implemented finite-particle algorithm, but it is not circularity. The self-citations to [51] (whose author K. Riedl overlaps with the present paper) provide the analytical framework and two technical propositions; these are peer-reviewed statements with assumptions that do not include the target result, and the paper contributes a new stability estimate (Proposition 3.2) and a new selection mechanism rather than assuming the conclusion.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central theoretical results rest on strong domain assumptions (A1-A5, W1-W6) about the objective functions, plus several external theorems from the CBO literature. The hyperparameters α, β, δq, R are not fitted to data but are subject to existential conditions in the main theorems. No new physical or ontological entities are introduced.

free parameters (5)
  • alpha (upper-level weighting exponent) = not fitted; chosen by hand, e.g., 10, 30, 50 in experiments; required large in Theorem 2.7
    Controls the Laplace-type approximation of the upper-level minimizer within the quantile set. The theorem requires alpha > alpha0, an existential bound depending on problem parameters.
  • beta (quantile fraction) = not fitted; e.g., 0.01-0.05 in experiments; required small in Theorem 2.7
    Controls the particle selection: only the best beta fraction w.r.t. L is used for the consensus point. The theorem requires beta < beta0, again existential.
  • delta_q (slack parameter in theoretical quantile set) = not fitted; required sufficiently small per Remark 2.8
    Introduced as a regularization in the theoretical definition of Q_L^beta. Needed for the convergence proof but not used in the implemented algorithm.
  • R (radius of truncation in theoretical quantile set) = not fitted; assumed large enough to contain θ*_good plus margin
    Part of the regularized quantile set (2.3); ensures consensus point remains bounded. Not present in the implemented algorithm.
  • lambda, sigma (drift and diffusion coefficients) = set to 1 and 1 in experiments; condition 2*lambda > d*sigma^2 in Theorem 2.7
    Standard CBO hyperparameters controlling attraction strength and exploration noise. Required to satisfy a spectral condition for exponential convergence.
assumptions (5)
  • domain assumption Assumptions A1-A5 on L and G, including uniqueness of θ*_good, local Hölder continuity, and inverse continuity properties (2.13) and (2.15)
    These are the central structural conditions for Theorem 2.7. They are stated as Assumption 2.6 and are not implied by the problem formulation alone. Assumption A5 in particular is strong and hard to verify.
  • domain assumption Assumptions W1-W6 on L and G, including Lipschitz continuity of L, a volume regularity condition W3 on level sets, and growth bounds on G
    Used for the existence theorem (Theorem 2.4). W3 is an unusual condition ruling out full-dimensional level sets and is needed for the stability estimate in Lemma 3.4.
  • standard math Quantitative Laplace principle from [51, Proposition 4.5]
    Used as a black-box tool in the proof of the quantitative quantiled Laplace principle (Proposition 4.2). It is an external published result.
  • standard math Regularity and positivity results for linear Fokker-Planck equations from [54, Theorem 2.5]
    Used in the Picard iteration proof of Theorem 2.4. This is a recent preprint, so the dependency is somewhat less established than a published theorem.
  • standard math Aubin-Lions compactness lemma and standard functional analysis embedding theorems
    Used to extract convergent subsequences in the Picard iteration. These are classical tools.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CB$^2$O: Consensus-Based Bi-Level Optimization." pith.science (2026). https://pith.science/paper/3PRAAASY

@misc{pith2026241113394,
  author       = {Pith},
  title        = {Pith review of: CB$^2$O: Consensus-Based Bi-Level Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3PRAAASY}},
  note         = {Machine review of arXiv:2411.13394}
}
abstract

Bi-level optimization problems, where one wishes to find the global minimizer of an upper-level objective function over the globally optimal solution set of a lower-level objective, arise in a variety of scenarios throughout science and engineering, machine learning, and artificial intelligence. In this paper, we propose and investigate, analytically and experimentally, consensus-based bi-level optimization (CB$^2$O), a multi-particle metaheuristic derivative-free optimization method designed to solve bi-level optimization problems when both objectives may be nonconvex. Our method leverages within the computation of the consensus point a carefully designed particle selection principle implemented through a suitable choice of a quantile on the level of the lower-level objective, together with a Laplace principle-type approximation w.r.t. the upper-level objective function, to ensure that the bi-level optimization problem is solved in an intrinsic manner. We give an existence proof of solutions to a corresponding mean-field dynamics, for which we first establish the stability of our consensus point w.r.t. a combination of Wasserstein and $L^2$ perturbations, and consecutively resort to PDE considerations extending the classical Picard iteration to construct a solution. For such solution, we provide a global convergence analysis in mean-field law showing that the solution of the associated nonlinear nonlocal Fokker-Planck equation converges exponentially fast to the unique solution of the bi-level optimization problem provided suitable choices of the hyperparameters. The practicability and efficiency of our CB$^2$O algorithm is demonstrated through extensive numerical experiments in the settings of constrained global optimization, sparse representation learning, and robust (clustered) federated learning.

Figures

Figures reproduced from arXiv: 2411.13394 by the authors.

Figure 1
Figure 1. An illustration of the CB2O algorithm (1.2) and its working principles for solving bi-level optimization problems of the form (1.1). We depict a typical setting of (1.1) in two dimensions with the lower-level objective function L being a Himmelblau function (plotted as contours in the xy-plane) and with the upper-level objective function G being a parabola (plotted as a surface)1 . The set of global minimizers of L,… view at source ↗
Figure 2
Figure 2. Illustration of Assumptions A3 and A5 on the lower-level objective function L and the upper level objective function G. Upper-level objective function G: Assumption A4 imposes that the objective function G is hG-H¨older continuous with constant HG locally around θ ∗ good. In addition, Assumption A5 requires the upper-level objective function G to admit an inverse continuity condition [48,50,51,83], which has to hold… view at source ↗
Figure 3
Figure 3. The influence of the hyperparameter β. By decreasing the value of β from left to right, we shrink the quantile set QL β [ϱ] (illustrated as yellow shadows on the xy plane and projected onto the upper-level objective), thereby ensuring an increasingly finer approximation of the set Θ of global minimizers of L. In order to technically approximate the set Θ through QL β [ϱ] sufficiently well, we need to choose in Theor… view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: The influence of the hyperparameter α. By increasing the value of α from left to right, we improve on the approximation of ˜θgood, i.e., how well m G,L α,β (ϱ), depicted as the orange dot in all three plots, eventually approximates ˜θgood. In order to technically appro…
Figure 5
Figure 5. Figure 5: An illustration that the consensus point [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: 2D Ackley function with two different constraint sets. [PITH_FULL_IMAGE:figures/full_fig_p040_6.png]
Figure 7
Figure 7. Figure 7: Comparison of different algorithms for the two-dimensional Ackley function with a circular [PITH_FULL_IMAGE:figures/full_fig_p043_7.png]
Figure 8
Figure 8. Figure 8: Comparison of different algorithms for the two-dimensional Ackley function with a star con [PITH_FULL_IMAGE:figures/full_fig_p044_8.png]
Figure 9
Figure 9. Figure 9: The influence of different hyperparameters on the performance of the CB [PITH_FULL_IMAGE:figures/full_fig_p045_9.png]
Figure 10
Figure 10. Figure 10: The influence of the hyperparameter β on the precision of the CB2O algorithm for different numbers of particles N averaged over 100 simulations. Influence of the Hyperparameters on the Running Time. Figures 11a, 11b and 11c reveal that the running time of the CB2O alg…
Figure 11
Figure 11. Figure 11: The influence of different hyperparameters on the running time of the CB [PITH_FULL_IMAGE:figures/full_fig_p046_11.png]
Figure 12
Figure 12. Figure 12: Comparison between standard CBO and CB2O for different constant values of the quantile hyperparameter β. All algorithms employ N = 100 particles. The sub-figures from left to right are comparing the training loss L, the norm of the feature representation G, and the te…
Figure 13
Figure 13. Figure 13: Comparison of CB2O for different number N of particles and quantile hyperparameter β. In the top row, we compare CB2O using N = 200 particles and different constant values of the parameter β with CB2O using N = 100 particles and β = 0.02. In the bottom row, we compare…
Figure 14
Figure 14. Figure 14: Comparison between CB2O with β = 0.02 and Penalized CBO with different Lagrange multipliers χ. The sub-figures from left to right are comparing the training loss L, the norm of the feature representation G, and the test accuracy against training epochs, respectively. …
Figure 15
Figure 15. Figure 15: Comparison between standard CBO and CB2O for decaying β with different decay factors. The sub-figures from left to right are comparing the training loss L, the norm of the feature representation G, and the test accuracy against training epochs, respectively. 6 Conclus…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Regularity and positivity of solutions of the Consensus-Based Optimization equation: unconditional global convergence

    math.AP 2025-02 conditional novelty 7.0 of 10

    For d>1, smooth solutions of the CBO Fokker-Planck equation are positive away from the consensus point, so the usual initial-support condition for global convergence can be dropped.

  2. MirrorCBO: A consensus-based optimization method in the spirit of mirror descent

    math.OC 2025-01 conditional novelty 7.0 of 10

    MirrorCBO generalizes consensus-based optimization to non-Euclidean geometries via mirror descent, with mean-field convergence guarantees and numerical wins on sparse and constrained optimization.

  3. Exploiting Structure with Anisotropic Consensus-Based Optimization

    math.OC 2026-07 accept novelty 6.0 of 10

    Anisotropic CBO's computational complexity depends exponentially only on the intrinsic dimension of an additively separable objective, not the ambient dimension, under aligned anisotropic noise.

  4. Defending Against Diverse Attacks in Federated Learning Through Consensus-Based Bi-Level Optimization

    cs.LG 2024-12 conditional novelty 6.0 of 10

    Consensus-based bi-level optimization converges in mean-field to the target model despite malicious agents, and a practical variant, FedCB2O, partially mitigates label-flipping attacks in clustered federated learning.

  5. Consensus-based optimization for closed-box adversarial attacks and a connection to evolution strategies

    math.OC 2025-06 conditional novelty 5.0 of 10

    Consensus-based optimization matches or beats natural evolution strategies as a closed-box adversarial attack method in easier attack scenarios, and consensus hopping is shown to be a gradient-descent-like limit of CBO.

Reference graph

Works this paper leans on

104 extracted references · 57 canonical work pages · cited by 5 Pith papers

  1. [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

  2. [2]

    D. J. Albers, P.-A. Blancquart, M. E. Levine, E. E. Seylabi, and A. Stuart. Ensemble kalman methods with constraints. Inverse Problems, 35(9):095007, 2019

    2019

  3. [3]

    Albi, Y.-P

    G. Albi, Y.-P. Choi, M. Fornasier, and D. Kalise. Mean field control hierarchy. Applied Mathematics & Optimization , 76(1):93–135, 2017

    2017

  4. [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

    2008

  5. [5]

    Amini and F

    M. Amini and F. Yousefian. An iterative regularized incremental projected subgradient method for a class of bilevel optimization problems. In 2019 American Control Conference (ACC) , pages 4069–4074. IEEE, 2019

    2019

  6. [6]

    Anandalingam

    G. Anandalingam. Artificial intelligence based approaches for solving hierarchical optimization problems. Impacts of recent computer advances on operations research , pages 289–301, 1989

    1989

  7. [7]

    J. S. Angelo and H. J. Barbosa. A study on the use of heuristics to solve a bilevel programming problem. International Transactions in Operational Research, 22(5):861–882, 2015

    2015

  8. [8]

    Anitescu

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

    2000

Show all 104 references
  1. [9]

    L. Arnold. Stochastic differential equations: Theory and applications . Wiley-Interscience [John Wiley & Sons], New York-London-Sydney, 1974. Translated from the German

    1974

  2. [10]

    J.-P. Aubin. Un th´ eoreme de compacit´ e.CR Acad. Sci. Paris , 256(24):5042–5044, 1963. 51

    1963

  3. [11]

    Bae, S.-Y

    H.-O. Bae, S.-Y. Ha, M. Kang, H. Lim, C. Min, and J. Yoo. A constrained consensus based optimization algorithm and its application to finance. Appl. Math. Comput. , 416:Paper No. 126726, 10, 2022

    2022

  4. [12]

    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

  5. [13]

    Beck and S

    A. Beck and S. Sabach. A first order method for finding minimal norm-like solutions of convex optimization problems. Mathematical Programming, 147(1):25–46, 2014

    2014

  6. [14]

    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

  7. [15]

    Bellavia and G

    S. Bellavia and G. Malaspina. A discrete consensus-based global optimization method with noisy objective function. arXiv preprint arXiv:2408.10078 , 2024

    2024 arXiv

  8. [16]

    Blum and A

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

    2003

  9. [17]

    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. Math. Program., 165(2, Ser. A):471–507, 2017

    2017

  10. [18]

    Bonandin and M

    S. Bonandin and M. Herty. Consensus-based algorithms for stochastic optimization problems. arXiv preprint arXiv:2404.10372, 2024

    2024

  11. [19]

    G. Borghi. Repulsion dynamics for uniform pareto front approximation in multi-objective opti- mization problems. PAMM, 23(1):e202200285, 2023

    2023

  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 optimization with uniform Pareto front approximation. Appl. Math. Optim. , 88(2):Paper No. 58, 43, 2023

    2023

  14. [22]

    Borghi, M

    G. Borghi, M. Herty, and L. Pareschi. Constrained consensus-based optimization. SIAM J. Optim. , 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 proba- bilities. arXiv preprint arXiv:2407.06389 , 2024

    2024 arXiv

  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]

    Bungert, T

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

    2024

  18. [26]

    J. V. Burke. An exact penalization viewpoint of constrained optimization. SIAM Journal on control and optimization , 29(4):968–998, 1991

    1991

  19. [27]

    Byeon, S.-Y

    J. Byeon, S.-Y. Ha, and J.-H. Won. Discrete consensus-based optimization. arXiv preprint arXiv:2403.03430, 2024

    2024 arXiv

  20. [28]

    A. Cabot. Proximal point algorithm controlled by a slowly vanishing term: applications to hierar- chical minimization. SIAM Journal on Optimization , 15(2):555–572, 2005

    2005

  21. [29]

    J. Cao, R. Jiang, N. Abolfazli, E. Yazdandoost Hamedani, and A. Mokhtari. Projection-free meth- ods for stochastic simple bilevel optimization with convex lower-level problem. Advances in Neural Information Processing Systems, 36, 2024

    2024

  22. [30]

    J. Cao, R. Jiang, E. Y. Hamedani, and A. Mokhtari. An accelerated gradient method for simple bilevel optimization with convex lower-level problem. arXiv preprint arXiv:2402.08097 , 2024. 52

    2024 arXiv

  23. [31]

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

    2018

  24. [32]

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

    2024

  25. [33]

    J. A. Carrillo, F. Hoffmann, A. M. Stuart, and U. Vaes. Consensus-based sampling. Stud. Appl. Math., 148(3):1069–1140, 2022

    2022

  26. [34]

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

    2021

  27. [35]

    J. A. Carrillo, S. Jin, H. Zhang, and Y. Zhu. An interacting particle consensus method for con- strained global optimization. arXiv preprint arXiv:2405.00891 , 2024

    2024

  28. [36]

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

    2023

  29. [37]

    Chaintron and A

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

    2022

  30. [38]

    Chaintron and A

    L.-P. Chaintron and A. Diez. Propagation of chaos: a review of models, methods and applications. II. Applications. Kinet. Relat. Models , 15(6):1017–1173, 2022

    2022

  31. [39]

    Chaudhry, M

    A. Chaudhry, M. Ranzato, M. Rohrbach, and M. Elhoseiny. Efficient lifelong learning with a-GEM. In International Conference on Learning Representations , 2019

    2019

  32. [40]

    P. Chen, X. Shi, R. Jiang, and J. Wang. Penalty-based methods for simple bilevel optimization under H¨ olderian error bounds.arXiv preprint arXiv:2402.02155 , 2024

    2024 arXiv

  33. [41]

    Cipriani, H

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

    2022

  34. [42]

    Dembo and O

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

    1998

  35. [43]

    Dempe, N

    S. Dempe, N. Dinh, and J. Dutta. Optimality conditions for a simple convex bilevel programming problem. Variational Analysis and Generalized Differentiation in Optimization and Control: In Honor of Boris S. Mordukhovich , pages 149–161, 2010

    2010

  36. [44]

    D. L. Donoho. Compressed sensing. IEEE Trans. Inf. Theory , 52(4):1289–1306, 2006

    2006

  37. [45]

    Dutta and T

    J. Dutta and T. Pandit. Algorithms for simple bilevel programming. Bilevel Optimization: Advances and Next Challenges , pages 253–291, 2020

    2020

  38. [46]

    Elsken, J

    T. Elsken, J. H. Metzen, and F. Hutter. Neural architecture search: a survey. J. Mach. Learn. Res., 20:Paper No. 55, 21, 2019

    2019

  39. [47]

    Fornasier, H

    M. Fornasier, H. Huang, L. Pareschi, and P. S¨ unnen. Consensus-based optimization on hypersur- faces: Well-posedness and mean-field limit. Math. Models Methods Appl. Sci. , 30(14):2725–2751, 2020

    2020

  40. [48]

    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. J. Mach. Learn. Res. , 22:Paper No. 237, 55, 2021

    2021

  41. [49]

    Fornasier, H

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

    1984

  42. [50]

    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

  43. [51]

    Fornasier, T

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

    2024

  44. [52]

    From integro-differential models to data-oriented ap- proaches for emergent phenomena

    M. Fornasier, P. Richt´ arik, K. Riedl, and L. Sun. Consensus-based optimization with truncated noise. Eur. J. Appl. Math. (special issue “From integro-differential models to data-oriented ap- proaches for emergent phenomena”) , 2024

    2024

  45. [53]

    Fornasier and L

    M. Fornasier and L. Sun. A pde framework of consensus-based optimization for objectives with multiple global minimizers. arXiv preprint arXiv:2403.06662 , 2024

    2024 arXiv

  46. [54]

    Fornasier and L

    M. Fornasier and L. Sun. Regularity and positivity of solutions of the consensus-based optimization equation: unconditional global convergence. arXiv preprint arXiv:2502.01434 , 2025

    2025 arXiv

  47. [55]

    Franceschi, P

    L. Franceschi, P. Frasconi, S. Salzo, R. Grazzi, and M. Pontil. Bilevel programming for hyperparam- eter optimization and meta-learning. In J. G. Dy and A. Krause, editors, Proceedings of the 35th International Conference on Machine Learning, ICML 2018, Stockholmsm¨ assan, Sto...

    2018

  48. [56]

    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 , 2025

    2025

  49. [57]

    N. J. Gerber, F. Hoffmann, and U. Vaes. Mean-field limits for consensus-based optimization and sampling. arXiv preprint arXiv:2312.07373 , 2023

    2023

  50. [58]

    Giang-Tran, N

    K.-H. Giang-Tran, N. Ho-Nguyen, and D. Lee. Projection-free methods for solving convex bilevel optimization problems. arXiv preprint arXiv:2311.09738 , 2023

    2023 arXiv

  51. [59]

    C. Gong, X. Liu, and Q. Liu. Automatic and harmless regularization with constrained and lexico- graphic optimization: A dynamic barrier approach. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P. Liang, and J. W. Vaughan, editors, Advances in Neural Information Processing Systems ...

    2021

  52. [60]

    Grassi, H

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

    2023

  53. [61]

    E. S. Helou and L. E. Sim˜ oes. ε-subgradient algorithms for bilevel convex optimization. Inverse problems, 33(5):055020, 2017

    2017

  54. [62]

    Herty, Y

    M. Herty, Y. Huang, D. Kalise, and H. Kouhkouh. A multiscale consensus-based algorithm for multi-level optimization. arXiv preprint arXiv:2407.09257 , 2024

    2024 arXiv

  55. [63]

    Herty and S

    M. Herty and S. Veneruso. Micro-macro decomposition for constrained particle swarm optimization method. in preparation, 2024

    2024

  56. [64]

    Huang and H

    H. Huang and H. Kouhkouh. Uniform-in-time mean-field limit estimate for the consensus-based optimization. arXiv preprint arXiv:2411.03986 , 2024

    2024 arXiv

  57. [65]

    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

  58. [66]

    Huang, J

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

    2023

  59. [67]

    Huang, J

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

    2024

  60. [68]

    Jiang, W.-X

    J.-J. Jiang, W.-X. Wei, W.-L. Shao, Y.-F. Liang, and Y.-Y. Qu. Research on large-scale bi-level particle swarm optimization algorithm. Ieee Access, 9:56364–56375, 2021

    2021

  61. [69]

    Jiang, N

    R. Jiang, N. Abolfazli, A. Mokhtari, and E. Y. Hamedani. A conditional gradient-based method for simple bilevel optimization with convex lower-level problem. In International Conference on Artificial Intelligence and Statistics , pages 10305–10323. PMLR, 2023. 54

    2023

  62. [70]

    H. D. Kaushik and F. Yousefian. A method with convergence rates for optimization problems with variational inequality constraints. SIAM Journal on Optimization , 31(3):2171–2198, 2021

    2021

  63. [71]

    Kennedy and R

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

    1995

  64. [72]

    Ko, S.-Y

    D. Ko, S.-Y. Ha, S. Jin, and D. Kim. Convergence analysis of the discrete consensus-based opti- mization algorithm with random batch interactions and heterogeneous noises. Mathematical Models and Methods in Applied Sciences , 32(06):1071–1107, 2022

    2022

  65. [73]

    LeCun, L

    Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner. Gradient-based learning applied to document recognition. Proc. IEEE, 86(11):2278–2324, 1998

    1998

  66. [74]

    X. Li, P. Tian, and X. Min. A hierarchical particle swarm optimization for solving bilevel program- ming problems. In International Conference on Artificial Intelligence and Soft Computing , pages 1169–1178. Springer, 2006

    2006

  67. [75]

    J. Lions. Quelques m´ ethodes de r´ esolution des probl` emes aux limites non lin´ eaires. Collection ´ etudes math´ ematiques. Dunod, 1969

    1969

  68. [76]

    R. Liu, J. Gao, J. Zhang, D. Meng, and Z. Lin. Investigating bi-level optimization for learning and vision from a unified perspective: A survey and beyond. IEEE Trans. Pattern Anal. Mach. Intell. , 44(12):10045–10067, 2022

    2022

  69. [77]

    Malitsky

    Y. Malitsky. The primal-dual hybrid gradient method reduces to a primal method for linearly constrained optimization problems. arXiv preprint arXiv:1706.02602 , 2017

    2017 arXiv

  70. [78]

    Mathieu, L

    R. Mathieu, L. Pittard, and G. Anandalingam. Genetic algorithm based approach to bi-level linear programming. RAIRO-Operations Research, 28(1):1–21, 1994

    1994

  71. [79]

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

    2006

  72. [80]

    Necoara, Y

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

    2019

  73. [81]

    Oduguwa and R

    V. Oduguwa and R. Roy. Bi-level optimisation using genetic algorithm. In Proceedings 2002 IEEE International Conference on Artificial Intelligence Systems (ICAIS 2002) , pages 322–327. IEEE, 2002

    2002

  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. Math. Models Methods Appl. Sci. , 27(1):183–204, 2017

    2017

  75. [83]

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

    2024

  76. [84]

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

    2024

  77. [85]

    Riedl, T

    K. Riedl, T. Klock, C. Geldhauser, and M. Fornasier. Gradient is All You Need? arXiv preprint arXiv:2306.09778, 2023

    2023

  78. [86]

    Riedl, T

    K. Riedl, T. Klock, C. Geldhauser, and M. Fornasier. How Consensus-Based Optimization can be Interpreted as a Stochastic Relaxation of Gradient Descent. ICML Workshop Differentiable Almost Everything: Differentiable Relaxations, Algorithms, Operators, and Simulators , 2024

    2024

  79. [87]

    W. Rudin. Functional Analysis. International series in pure and applied mathematics. McGraw-Hill, 1991

    1991

  80. [88]

    Sabach and S

    S. Sabach and S. Shtern. A first order method for solving convex bilevel optimization problems. SIAM Journal on Optimization , 27(2):640–660, 2017

    2017

  81. [89]

    Samadi, D

    S. Samadi, D. Burbano, and F. Yousefian. Achieving optimal complexity guarantees for a class of bilevel convex optimization problems. In 2024 American Control Conference (ACC) , pages 2206–

    2024

  82. [90]

    Shehu, P

    Y. Shehu, P. T. Vuong, and A. Zemkoho. An inertial extrapolation method for convex simple bilevel optimization. Optimization Methods and Software , 36(1):1–19, 2021

    2021

  83. [91]

    L. Shen, N. Ho-Nguyen, and F. Kılın¸ c-Karzan. An online convex optimization-based framework for convex bilevel optimization. Mathematical Programming, 198(2):1519–1582, 2023

    2023

  84. [92]

    Sinha, P

    A. Sinha, P. Malo, and K. Deb. A review on bilevel optimization: From classical to evolutionary approaches and applications. IEEE transactions on evolutionary computation , 22(2):276–295, 2017

    2017

  85. [93]

    M. Solodov. An explicit descent method for bilevel convex optimization.Journal of Convex Analysis, 14(2):227, 2007

    2007

  86. [94]

    Tikhonov and V

    A. Tikhonov and V. Arsenin. Solutions of Ill-posed Problems . Halsted Press book. Winston, 1977

    1977

  87. [95]

    C. Villani. Optimal transport: Old and new , volume 338 of Grundlehren der mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences] . Springer-Verlag, Berlin, 2009

    2009

  88. [96]

    J. Wang, X. Shi, and R. Jiang. Near-optimal convex simple bilevel optimization with a bisection method. In International Conference on Artificial Intelligence and Statistics , pages 2008–2016. PMLR, 2024

    2008

  89. [97]

    Wogrin, S

    S. Wogrin, S. Pineda, and D. A. Tejada-Arango. Applications of bilevel optimization in energy and electricity markets. Bilevel Optimization: Advances and Next Challenges , pages 139–168, 2020

    2020

  90. [98]

    Y. Yin. Genetic-algorithms-based approach for bilevel programming models. Journal of transporta- tion engineering, 126(2):115–120, 2000

    2000

  91. [99]

    M. B. Zafar, I. Valera, M. G. Rogriguez, and K. P. Gummadi. Fairness constraints: Mechanisms for fair classification. In Artificial intelligence and statistics , pages 962–970. PMLR, 2017

    2017

  92. [100]

    X. Zhu, Q. Yu, and X. Wang. A hybrid differential evolution algorithm for solving nonlinear bilevel programming with linear constraints. In 2006 5th IEEE International Conference on Cognitive Informatics, volume 1, pages 126–131. IEEE, 2006. 56 A Technical Lemmas A.1 Technical...

    2006

  93. [101]

    limt→s ∥ρt∥2 L2(Rd) = ∥ρs∥2 L2(Rd) for any s ∈ [0, T]

  94. [102]

    lim t→s Z Rd φ(θ)ρt(θ)dθ = Z Rd φ(θ)ρs(θ)dθ, (A.15) for every φ ∈ L2(Rd)

    For any s ∈ [0, T], ρt weakly converges to ρs in L2(Rd) as t → s, i.e. lim t→s Z Rd φ(θ)ρt(θ)dθ = Z Rd φ(θ)ρs(θ)dθ, (A.15) for every φ ∈ L2(Rd). Then ρt strongly converges to ρs as t → s in L2(Rd). In other words, ρ ∈ C [0, T], L2(Rd) . Proof. For any s ∈ [0, T], we have lim t...

  95. [103]

    ϱn strongly converges to ϱ in L2 loc(Rd),

  96. [104]

    Then {ϱn}n∈N converges to ϱ in the Wasserstein- 2 distance, i.e., lim n→∞ W2(ϱn, ϱ) = 0

    supn∈N R ∥θ∥4 2ϱn(θ)dθ ≤ C for some constant C >0. Then {ϱn}n∈N converges to ϱ in the Wasserstein- 2 distance, i.e., lim n→∞ W2(ϱn, ϱ) = 0. Proof. Since weak convergence of probability measures and convergence in second moments implies con- vergence in the Wasserstein-2 distan...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.