Pith. sign in

REVIEW 1 major objections 7 minor 46 references

Convergence Analysis of the Restarted Moving-Anchored Extra-Gradient Method in the Absence of Local Lipschitz Continuity

T0 review · 1 major / 7 minor · reviewed 2026-07-09 · glm-5.2

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

Pith's one-line read Extra-gradient method converges without Lipschitz continuity

desk verdict Removing Lipschitz continuity from accelerated extra-gradient methods is the real contribution here. The proof structure is sound. read the letter →

arxiv 2607.07585 v1 pith:TLCJCJLW submitted 2026-07-08 math.OC

Defeng Sun , Liping Zhang , Wei Zhao This is my paper · ORCID
classification math.OC MSC 47H0549J4065K1590C25
keywords monotoneinclusionextra-gradientmethodmovinganchorrestartstrategynon-Lipschitzconvergenceiterationcomplexitycompositenaturalresidualmaximaloperator
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 the Moving-Anchored Extra-Gradient (MAEG) method for solving monotone inclusion problems of the form 0 ∈ (F+B)(x), where F is continuous and monotone and B is maximal monotone. The central structural innovation is that the anchor point's distance to the solution set is monotonically non-increasing across iterations, a property absent in prior moving-anchor methods. Under Lipschitz continuity of F, MAEG achieves O(1/k) non-asymptotic and o(1/k) asymptotic convergence rates. The paper then exploits the non-increasing anchor distance to design a restart strategy (MAEG-R) that provably converges to a solution when F is merely continuous and monotone, requiring neither global nor local Lipschitz continuity. When F does happen to be Lipschitz, MAEG-R restarts only finitely many times and preserves the full complexity guarantees of the base method. The key mechanism enabling convergence in the non-Lipschitz regime is a compactness argument (via the Tube Lemma applied to joint continuity of the resolvent) showing that the line-search condition is satisfiable at the first iteration of each restart epoch, combined with a contradiction argument showing that infinitely many restarts would force divergence of a bounded quantity.

What carries the argument

The composite natural residual R_nat^c(x, λ) = (1/λ)(x - J_{λB}(x - λF(x))) serves as the optimality measure. The auxiliary function L_k = ⟨u_k - y_k, d_k⟩ - (1/2)(1-2ρ)Λ_k ||d_k||^2 governs the recursion. The restart anchor update u_{r+1,0} = Π_{cl(dom B)}(u_{r,k-1} - κ⟨u_{r,k-1}-y_{r,k-1}, d_{r,k-1}⟩/||d_{r,k-1}||^2 · d_{r,k-1}) preserves the non-increasing distance property for κ ∈ [0,2].

What would settle it

Construct a continuous monotone operator F and maximal monotone B with nonempty solution set for which the restart strategy triggers infinitely many times and the restart points remain bounded away from the solution set, contradicting the contradiction argument in Theorem 3 Case (ii). This would require violating the compactness-based line-search guarantee of Proposition 5.

Watch

Extended reading notes

Core claim

The non-increasing distance from the anchor point to the solution set is the load-bearing property. It simultaneously (a) yields the O(1/k) and o(1/k) rates under Lipschitz continuity, (b) guarantees that restarts do not move the anchor farther from the solution set, and (c) enables a contradiction proof that the restart points must converge to the solution set even when F lacks any Lipschitz property. The restart strategy works because at each restart's first iteration, the line-search condition is guaranteed to be satisfiable by a compactness-plus-continuity argument (Proposition 5), and if restart points stayed bounded away from the solution set forever, the squared first-iteration norms,

Load-bearing premise

The convergence proof for the case of infinitely many restarts relies on a compactness argument showing that the line-search condition is always satisfiable at the first iteration of each restart epoch, using the Tube Lemma applied to the joint continuity of the resolvent mapping on a compact set. If this compactness argument fails or the joint continuity property does not hold in a given setting, the infinite-restart convergence proof collapses. Additionally, the number of重启

Editorial extensions

If this is right

  • Operators arising in p-Laplacian obstacle problems (p > 2, locally but not globally Lipschitz) and ℓ_p^2-regularized problems (1 < p < 2, not even locally Lipschitz) become directly solvable by an accelerated extra-gradient method with convergence guarantees, whereas prior accelerated methods are inapplicable.
  • The restart framework is agnostic to the specific restart trigger: any additional restart criterion can be added provided the anchor is updated per the prescribed rule and the restart occurs after an accepted MAEG step.
  • The counterexample in the appendix shows that a prior line-search-based extra-gradient convergence proof (Khobotov 1987) fails without local Lipschitz continuity, clarifying that the present restart strategy is not merely a convenience but a necessary mechanism for the non-Lipschitz regime.
  • The O(1/k) rate for the natural residual matches the best known order for this problem class under Lipschitz continuity, and the paper achieves it without requiring boundedness of the feasible set.

Reading between the lines

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

  • If the compactness argument in Proposition 5 could be extended to infinite-dimensional settings (e.g., Hilbert spaces), the restart strategy might yield convergence guarantees for non-Lipschitz monotone inclusions in PDE-constrained optimization and infinite-dimensional variational inequalities, though the Tube Lemma step is inherently finite-dimensional.
  • The number of restarts in the non-Lipschitz case is not theoretically bounded, which means practical performance depends on how quickly the restart points approach the solution set. A worst-case bound on restart count would close the gap between the convergence guarantee and a non-asymptotic complexity result for the non-Lipschitz regime.
  • The moving-anchor mechanism may extend to other algorithm families (e.g., primal-dual methods, ADMM variants) where a distance-to-solution-set monotonicity property can be established, potentially removing Lipschitz requirements there as well.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 7 minor

Summary. This paper introduces the Moving-Anchored Extra-Gradient (MAEG) method for monotone inclusion problems $0 in (F+B)(x)$, where $F$ is continuous and monotone and $B$ is maximal monotone. The method features a moving anchor whose distance to the solution set is monotonically non-increasing (Theorem 1(ii)). Under Lipschitz continuity of $F$, MAEG achieves $O(1/k)$ iteration complexity (Theorem 1(iii)) and, with a positive anchor-update parameter $rho$, an $o(1/k)$ asymptotic rate (Remark 3). The main contribution is a restart strategy (Algorithm 2, MAEG-R) that ensures convergence under Assumption 1 (nonempty solution set) alone, without any Lipschitz continuity of $F$. When $F$ is Lipschitz, MAEG-R restarts only finitely many times and preserves the complexity guarantees of MAEG (Remark 4). The key technical tool is Proposition 5, which uses the Tube Lemma and the joint continuity of the resolvent (Lemma 1(iii)) to guarantee well-definedness of the first iteration of each epoch on compact sets excluding the solution set.

Significance. The paper makes a genuine contribution to the theory of accelerated extra-gradient methods. Eliminating the Lipschitz continuity requirement for an anchored/accelerated EG method is novel; existing methods in this class (EAG, FEG, SEG) all require global Lipschitz continuity. The convergence proof for the infinite-restart case (Theorem 3, Case ii) is carefully constructed: the contradiction argument via the finite sum bound (39), the lower bound on $|d_{r,1}|$ from (40)-(41) and Lemma 2, and the cross-epoch non-increasing anchor distance all hold up under scrutiny. The Lyapunov function $L_k$ (14) and its recursive inequality (Proposition 2) are well-constructed and yield the $O(1/k)$ bound directly without fitted constants. The counterexample in the appendix (showing that Khobotov's proof requires local Lipschitz continuity) is a useful clarification. The numerical experiments are comprehensive, covering globally Lipschitz, locally Lipschitz, and genuinely non-Lipschitz regimes, with a GPU-optimized implementation.

major comments (1)
  1. Section 4, proof of Theorem 3, Case (ii): The application of Proposition 4 to conclude $|y_{r,k} - u_{r,k}| to 0$ uniformly in $k$ is correct, but the argument could be stated more explicitly. Proposition 4 gives $|y_{r,k} - u_{r,k}|^2 le C |u_{r,0} - x^*|^2$ where $C$ depends on $M, sigma, rho$ but not on $r$ or $k$. Since $|u_{r,0} - x^*| to 0$ as $r to infty$, this gives uniform convergence. The paper states this conclusion in one sentence; expanding it to make the role of the uniform bound explicit would improve clarity, though the logic is sound.
minor comments (7)
  1. Notation clash: Algorithm 2 uses $alpha_1$ and $bar{alpha}_1$ as stepsize bounds, while Section 5 uses $alpha_1, alpha_2, alpha_3$ for adaptive restart criteria parameters with different meanings. Consider using distinct symbols (e.g., $eta_1, eta_2, eta_3$) for the restart criteria in Section 5.
  2. Remark 3 establishes the $o(1/k)$ asymptotic rate, but this result is only a remark rather than a formal theorem. Given that it is highlighted in the abstract, elevating it to a corollary or theorem would be appropriate.
  3. Section 5: The adaptive restart criteria (i)-(iii) are not covered by the theoretical analysis in Section 4. Remark 5 states that additional restart criteria can be incorporated, but a brief comment in Section 5 explicitly connecting these practical criteria to Remark 5 would help the reader.
  4. Table 1: The forward evaluation count for MFBS on the Ran family (e.g., $3.0 times 10^6$ at $n=2^{10}$) is orders of magnitude larger than for MAEG variants. A brief explanation of why MFBS requires so many more forward evaluations (e.g., line search behavior) would be helpful.
  5. Section 3, between Remark 1 and Assumption 1: The role of the set $X$ in Algorithm 1 is mentioned briefly in Remark 1, but the theoretical analysis does not discuss how the choice of $X$ affects convergence. The numerical experiments (Table 1) show that $X$ can significantly impact performance. A remark on this would strengthen the paper.
  6. Reference [32] (Sun, 1993) is cited as showing convergence of EG with line search under only continuity and monotonicity. The relationship between this earlier result and the present contribution could be clarified, particularly regarding what is new beyond [32] (the accelerated/anchored framework and complexity guarantees).
  7. Minor typo: In the abstract, 'ano(1/k)' should be 'an $o(1/k)$'.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and the positive assessment. The referee's single major comment is well-taken and concerns a clarity issue in the proof of Theorem 3, Case (ii), regarding the uniform convergence of ||y_{r,k} - u_{r,k}|| → 0. We agree that the argument can and should be stated more explicitly.

read point-by-point responses

  1. Referee: Section 4, proof of Theorem 3, Case (ii): The application of Proposition 4 to conclude |y_{r,k} - u_{r,k}| → 0 uniformly in k is correct, but the argument could be stated more explicitly. Proposition 4 gives |y_{r,k} - u_{r,k}|^2 ≤ C |u_{r,0} - x*|^2 where C depends on M, σ, ρ but not on r or k. Since |u_{r,0} - x*| → 0 as r → ∞, this gives uniform convergence. The paper states this conclusion in one sentence; expanding it to make the role of the uniform bound explicit would improve clarity, though the logic is sound.

    Authors: We agree with the referee that the argument is currently stated too tersely. In the revised manuscript, we will expand the relevant passage in the proof of Theorem 3, Case (ii), to make the role of the uniform bound explicit. Specifically, we will add the following clarification: Proposition 4 establishes that ||y_{r,k} - u_{r,k}||^2 ≤ C ||u_{r,0} - x*||^2 for all k ≥ 1, where C = (1-ρ)/ρ + σ^2 M / ((1-σ^2)(1-2ρ)^2) depends only on M, σ, and ρ, and is independent of both r and k. Since we have already shown that ||u_{r,0} - x*|| → 0 as r → ∞ (because the distance sequence is monotonically non-increasing and the subsequence converges to x* ∈ Sol(F,B)), it follows that sup_{k≥1} ||y_{r,k} - u_{r,k}||^2 ≤ C ||u_{r,0} - x*||^2 → 0 as r → ∞. This establishes the uniform (in k) convergence of ||y_{r,k} - u_{r,k}|| to 0, and consequently y_{r,k} → x* as well. We believe this expanded explanation will make the proof more transparent. revision: yes

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity found; derivation is self-contained

full rationale

The paper's theoretical core is parameter-free and self-contained. The O(1/k) complexity bound (Theorem 1(iii)) follows from the Lyapunov function L_k (Eq. 14) and the recursive inequality in Proposition 2, which is derived from monotonicity of (F+B), nonexpansiveness of the resolvent, and the stepsize condition (11) — all standard, none fitted to data. The convergence proof (Theorem 2) uses Propositions 3–4, which follow from the same Lyapunov analysis. The restart convergence proof (Theorem 3, Case ii) uses a genuine contradiction argument: the finite bound (39) on Σ λ²_{r,1}||d_{r,1}||² is derived from Theorem 1(ii); the lower bound comes from Proposition 5, which itself relies on the Tube Lemma (cited from Munkres [24], an external textbook), Lemma 1(iii) (cited from [4, Proposition 3.4] by non-overlapping authors), and Lemma 2 (a generalization of [12, 7]). No step reduces to its own inputs by construction. Self-citations ([8, 9, 32, 36, 41, 46]) appear only in the introduction for background context (LP/QP applications, prior EG work) and are not load-bearing for any proof. The stepsize condition (11) is a standard backtracking condition, not a fitted parameter renamed as a prediction. The proof of Theorem 2 is described as 'inspired by [44, Theorem 5]' but is carried out in full detail within the paper. No uniqueness theorem is invoked to forbid alternatives. The only reason for a score of 1 rather than 0 is the presence of multiple self-citations in the introduction, but none are load-bearing for the mathematical results.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

No new physical entities, particles, forces, or dimensions are introduced. The algorithmic constructs (moving anchor u_k, descent direction d_k, auxiliary function L_k) are standard mathematical objects in optimization. All parameters are algorithmic hyperparameters with stated valid ranges, not fitted constants. The theory is parameter-free in the sense that the convergence guarantees hold for any valid parameter choices.

free parameters (7)
  • σ = 0.99 (experiments)
    Line-search acceptance parameter in (11); any value in (0,1) is valid. Not fitted to data.
  • ρ = 0 or 0.2 (experiments)
    Anchor-update parameter in (13); any value in [0, 1/2) is valid. Controls whether anchor moves.
  • β = 0.7 (experiments)
    Backtracking contraction factor in Algorithm 2; standard choice.
  • κ = 2 (experiments)
    Restart projection parameter in (35); any value in [0,2] is valid.
  • M = not specified numerically
    Stepsize upper-bound ratio in Proposition 4; theoretical parameter.
  • ε_r = →0
    Restart threshold sequence in Algorithm 2; must converge to 0.
  • α₁, α₂, α₃ = 0.1, 0.6, 0.2
    Adaptive restart heuristic parameters in Section 5; empirical choices.
assumptions (4)
  • domain assumption Sol(F, B) is nonempty.
    Assumption 1. Standard for convergence analysis of monotone inclusion.
  • domain assumption F is continuous and monotone on cl(dom B); B is maximal monotone; cl(dom B) ⊆ dom F.
    Problem setting in Section 1. Ensures T = F + B is maximal monotone by [23, Proposition A.1].
  • domain assumption For each k, there exists λ_k > 0 satisfying (11).
    Assumption 2. Shown to hold under Lipschitz or local Lipschitz continuity (Remark 2); in the non-Lipschitz case, replaced by the restart mechanism.
  • domain assumption Stepsizes satisfy λ_k ∈ (λ̲, M λ₁) for some λ̲ > 0, M > 1.
    Used in Proposition 4 and Theorem 2(ii) for the ρ > 0 case. Ensures bounded stepsizes for convergence of {y_k} and {u_k}.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Convergence Analysis of the Restarted Moving-Anchored Extra-Gradient Method in the Absence of Local Lipschitz Continuity." pith.science (2026). https://pith.science/paper/TLCJCJLW

@misc{pith2026260707585,
  author       = {Pith},
  title        = {Pith review of: Convergence Analysis of the Restarted Moving-Anchored Extra-Gradient Method in the Absence of Local Lipschitz Continuity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TLCJCJLW}},
  note         = {Machine review of arXiv:2607.07585}
}
abstract

In this paper, we introduce the moving-anchored extra-gradient (MAEG) method for solving monotone inclusion problems involving the sum of a continuous monotone operator and a maximal monotone operator. Notably, the distance from the anchor point to the solution set is designed to be monotonically non-increasing. Under Lipschitz continuity of the forward operator, MAEG attains an $\mathcal{O}(1/k)$ non-asymptotic iteration complexity, and when a positive anchor-update parameter is used, it further achieves an $o(1/k)$ asymptotic rate. Furthermore, leveraging the specific behavior of the anchor point, we propose a tailored restart strategy. We demonstrate that this strategy ensures convergence even in the absence of local Lipschitz continuity, while preserving the original iteration complexity guarantees whenever the Lipschitz condition holds.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 46 canonical work pages

  1. [1]

    Moving Anchor Extragradient Methods For Smooth Structured Minimax Problems

    James K Alcala, Yat Tin Chow, and Mahesh Sunkula. Moving anchor extragradient methods for smooth structured minimax problems.arXiv preprint arXiv:2308.12359, 2023

    work page Pith review arXiv 2023

  2. [2]

    Practical large-scale linear programming using primal-dual hybrid gradient

    David Applegate, Mateo D´ ıaz, Oliver Hinder, Haihao Lu, Miles Lubin, Brendan O’Donoghue, and Warren Schudy. Practical large-scale linear programming using primal-dual hybrid gradient. Advances in Neural Information Processing Systems, 34:20243–20257, 2021

    work page 2021

  3. [3]

    Faster first-order primal-dual meth- ods for linear programming using restarts and sharpness.Mathematical Programming, 201(1):133– 184, 2023

    David Applegate, Oliver Hinder, Haihao Lu, and Miles Lubin. Faster first-order primal-dual meth- ods for linear programming using restarts and sharpness.Mathematical Programming, 201(1):133– 184, 2023

    work page 2023

  4. [4]

    Relocated Fixed-Point Iterations with Applications to Variable Stepsize Resolvent Splitting

    Felipe Atenas, Heinz H Bauschke, Minh N Dao, and Matthew K Tam. Relocated fixed-point iter- ations with applications to variable stepsize resolvent splitting.arXiv preprint arXiv:2507.07428, 2025

    work page Pith review arXiv 2025

  5. [5]

    Springer, 2017

    Heinz H Bauschke and Patrick L Combettes.Convex Analysis and Monotone Operator Theory in Hilbert Spaces. Springer, 2017

    work page 2017

  6. [6]

    Accelerated algorithms for constrained nonconvex-nonconcave min-max optimization and comonotone inclusion

    Yang Cai, Argyris Oikonomou, and Weiqiang Zheng. Accelerated algorithms for constrained nonconvex-nonconcave min-max optimization and comonotone inclusion. InProceedings of the 41st International Conference on Machine Learning, volume 235 ofProceedings of Machine Learn- ing Research, pages 5312–5347. PMLR, 21–27 Jul 2024

    work page 2024

  7. [7]

    Projected gradient methods for linearly constrained problems

    Paul H Calamai and Jorge J Mor´ e. Projected gradient methods for linearly constrained problems. Mathematical Programming, 39(1):93–116, 1987

    work page 1987

  8. [8]

    HPR-LP: An implementation of an HPR method for solving linear programming.Mathematical Programming Computation, pages 1–28, 2025

    Kaihuang Chen, Defeng Sun, Yancheng Yuan, Guojun Zhang, and Xinyuan Zhao. HPR-LP: An implementation of an HPR method for solving linear programming.Mathematical Programming Computation, pages 1–28, 2025. 20

    work page 2025

Show all 46 references
  1. [9]

    HPR-QP: A dual halpern peaceman-rachford method for solving large-scale convex composite quadratic programming.arXiv preprint arXiv:2507.02470, 2025

    Kaihuang Chen, Defeng Sun, Yancheng Yuan, Guojun Zhang, and Xinyuan Zhao. HPR-QP: A dual halpern peaceman-rachford method for solving large-scale convex composite quadratic programming.arXiv preprint arXiv:2507.02470, 2025

    2025 arXiv

  2. [10]

    Fast projection onto the simplex and theℓ 1 ball.Mathematical Programming, 158(1):575–585, 2016

    Laurent Condat. Fast projection onto the simplex and theℓ 1 ball.Mathematical Programming, 158(1):575–585, 2016

    2016

  3. [11]

    Efficient projections onto theℓ 1-ball for learning in high dimensions

    John Duchi, Shai Shalev-Shwartz, Yoram Singer, and Tushar Chandra. Efficient projections onto theℓ 1-ball for learning in high dimensions. InProceedings of the 25th International Conference on Machine Learning, pages 272–279, 2008

    2008

  4. [12]

    Two-metric projection methods for constrained optimiza- tion.SIAM Journal on Control and Optimization, 22(6):936–964, 1984

    Eli M Gafni and Dimitri P Bertsekas. Two-metric projection methods for constrained optimiza- tion.SIAM Journal on Control and Optimization, 22(6):936–964, 1984

    1984

  5. [13]

    Fixed points of nonexpanding maps.Bulletin of the American Mathematical Society, 73:957–961, 1967

    Benjamin Halpern. Fixed points of nonexpanding maps.Bulletin of the American Mathematical Society, 73:957–961, 1967

    1967

  6. [14]

    Evgenii Nikolaevich Khobotov. Modification of the extra-gradient method for solving variational inequalities and certain optimization problems.USSR Computational Mathematics and Mathe- matical Physics, 27(5):120–127, 1987

    1987

  7. [15]

    Accelerated proximal point method for maximally monotone operators.Mathe- matical Programming, 190(1):57–87, 2021

    Donghwan Kim. Accelerated proximal point method for maximally monotone operators.Mathe- matical Programming, 190(1):57–87, 2021

    2021

  8. [16]

    The extragradient method for finding saddle points and other problems

    Galina M Korpelevich. The extragradient method for finding saddle points and other problems. Matecon, 12:747–756, 1976

    1976

  9. [17]

    Fast extra gradient methods for smooth structured nonconvex- nonconcave minimax problems.Advances in Neural Information Processing Systems, 34:22588– 22600, 2021

    Sucheol Lee and Donghwan Kim. Fast extra gradient methods for smooth structured nonconvex- nonconcave minimax problems.Advances in Neural Information Processing Systems, 34:22588– 22600, 2021

    2021

  10. [18]

    On the convergence rate of the Halpern-iteration.Optimization Letters, 15(2):405– 418, 2021

    Felix Lieder. On the convergence rate of the Halpern-iteration.Optimization Letters, 15(2):405– 418, 2021

    2021

  11. [19]

    Splitting algorithms for the sum of two nonlinear operators.SIAM Journal on Numerical Analysis, 16(6):964–979, 1979

    Pierre-Louis Lions and Bertrand Mercier. Splitting algorithms for the sum of two nonlinear operators.SIAM Journal on Numerical Analysis, 16(6):964–979, 1979

    1979

  12. [20]

    Projected reflected gradient methods for monotone variational inequalities.SIAM Journal on Optimization, 25(1):502–520, 2015

    Yura Malitsky. Projected reflected gradient methods for monotone variational inequalities.SIAM Journal on Optimization, 25(1):502–520, 2015

    2015

  13. [21]

    Golden ratio algorithms for variational inequalities.Mathematical Programming, 184(1):383–410, 2020

    Yura Malitsky. Golden ratio algorithms for variational inequalities.Mathematical Programming, 184(1):383–410, 2020

    2020

  14. [22]

    On the complexity of the hybrid proximal extragra- dient method for the iterates and the ergodic mean.SIAM Journal on Optimization, 20(6):2755– 2787, 2010

    Renato DC Monteiro and Benar Fux Svaiter. On the complexity of the hybrid proximal extragra- dient method for the iterates and the ergodic mean.SIAM Journal on Optimization, 20(6):2755– 2787, 2010

    2010

  15. [23]

    Renato DC Monteiro and Benar Fux Svaiter. Complexity of variants of tseng’s modified fb splitting and korpelevich’s methods for hemivariational inequalities with applications to saddle-point and convex optimization problems.SIAM Journal on Optimization, 21(4):1688–1720, 2011

    2011

  16. [24]

    Munkres.Topology

    James R. Munkres.Topology. Prentice Hall, Upper Saddle River, NJ, 2nd edition, 2000

    2000

  17. [25]

    John F. Nash. Non-cooperative games.Annals of Mathematics, 54(2):286–295, 1951

    1951

  18. [26]

    Prox-method with rate of convergence o (1/t) for variational inequalities with lipschitz continuous monotone operators and smooth convex-concave saddle point problems

    Arkadi Nemirovski. Prox-method with rate of convergence o (1/t) for variational inequalities with lipschitz continuous monotone operators and smooth convex-concave saddle point problems. SIAM Journal on Optimization, 15(1):229–251, 2004

    2004

  19. [27]

    High-order reduced-gradient methods for composite variational inequalities.arXiv preprint arXiv:2311.15154, 2023

    Yurii Nesterov. High-order reduced-gradient methods for composite variational inequalities.arXiv preprint arXiv:2311.15154, 2023

    2023 arXiv

  20. [28]

    Lower complexity bounds of first-order methods for convex- concave bilinear saddle-point problems.Mathematical Programming, 185(1):1–35, 2021

    Yuyuan Ouyang and Yangyang Xu. Lower complexity bounds of first-order methods for convex- concave bilinear saddle-point problems.Mathematical Programming, 185(1):1–35, 2021. 21

    2021

  21. [29]

    A modification of the arrow-hurwitz method of search for saddle points

    Leonid Denisovich Popov. A modification of the arrow-hurwitz method of search for saddle points. Mat. Zametki, 28(5):777–784, 1980

    1980

  22. [30]

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

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

    2017

  23. [31]

    A hybrid approximate extragradient–proximal point algorithm using the enlargement of a maximal monotone operator.Set-Valued Analysis, 7(4):323– 345, 1999

    Mikhail V Solodov and Benar F Svaiter. A hybrid approximate extragradient–proximal point algorithm using the enlargement of a maximal monotone operator.Set-Valued Analysis, 7(4):323– 345, 1999

    1999

  24. [32]

    Projected extragradient method for finding saddle points of general convex program- ming.Journal of Qufu Normal University, 19(4):10–17, 1993

    Defeng Sun. Projected extragradient method for finding saddle points of general convex program- ming.Journal of Qufu Normal University, 19(4):10–17, 1993

    1993

  25. [33]

    A projection and contraction method for the nonlinear complementarity problem and its extensions.Mathematica Numerica Sinica, 16(2):183–194, 1994

    Defeng Sun. A projection and contraction method for the nonlinear complementarity problem and its extensions.Mathematica Numerica Sinica, 16(2):183–194, 1994

    1994

  26. [34]

    A new step-size skill for solving a class of nonlinear projection equations.Journal of Computational Mathematics, pages 357–368, 1995

    Defeng Sun. A new step-size skill for solving a class of nonlinear projection equations.Journal of Computational Mathematics, pages 357–368, 1995

    1995

  27. [35]

    A class of iterative methods for solving nonlinear projection equations.Journal of Optimization Theory and Applications, 91:123–140, 1996

    Defeng Sun. A class of iterative methods for solving nonlinear projection equations.Journal of Optimization Theory and Applications, 91:123–140, 1996

    1996

  28. [36]

    Accelerating preconditioned admm via degenerate proximal point mappings.SIAM Journal on Optimization, 35(2):1165– 1193, 2025

    Defeng Sun, Yancheng Yuan, Guojun Zhang, and Xinyuan Zhao. Accelerating preconditioned admm via degenerate proximal point mappings.SIAM Journal on Optimization, 35(2):1165– 1193, 2025

    2025

  29. [37]

    Sublinear convergence rates of extragradient-type methods: A survey on classical and recent developments.arXiv preprint arXiv:2303.17192, 2023

    Quoc Tran-Dinh. Sublinear convergence rates of extragradient-type methods: A survey on classical and recent developments.arXiv preprint arXiv:2303.17192, 2023

    2023 arXiv

  30. [38]

    A modified forward-backward splitting method for maximal monotone mappings

    Paul Tseng. A modified forward-backward splitting method for maximal monotone mappings. SIAM Journal on Control and Optimization, 38(2):431–446, 2000

    2000

  31. [39]

    Approximation of fixed points of nonexpansive mappings.Archiv der mathe- matik, 58:486–491, 1992

    Rainer Wittmann. Approximation of fixed points of nonexpansive mappings.Archiv der mathe- matik, 58:486–491, 1992

    1992

  32. [40]

    Iterative algorithms for nonlinear operators.Journal of the London Mathematical Society, 66(1):240–256, 2002

    Hong-Kun Xu. Iterative algorithms for nonlinear operators.Journal of the London Mathematical Society, 66(1):240–256, 2002

    2002

  33. [41]

    Bo Yang, Xinyuan Zhao, Xudong Li, and Defeng Sun. An accelerated proximal alternating direction method of multipliers for optimal decentralized control of uncertain systems.Journal of Optimization Theory and Applications, 204(1):9, 2025

    2025

  34. [42]

    Accelerated algorithms for smooth convex-concave minimax problems withO(1/k 2) rate on squared gradient norm

    TaeHo Yoon and Ernest K Ryu. Accelerated algorithms for smooth convex-concave minimax problems withO(1/k 2) rate on squared gradient norm. InInternational Conference on Machine Learning, pages 12098–12109. PMLR, 2021

    2021

  35. [43]

    Accelerated minimax algorithms flock together.SIAM Journal on Optimization, 35(1):180–209, 2025

    TaeHo Yoon and Ernest K Ryu. Accelerated minimax algorithms flock together.SIAM Journal on Optimization, 35(1):180–209, 2025

    2025

  36. [44]

    Symplectic extra-gradient type method for solving general non- monotone inclusion problem.arXiv preprint arXiv:2406.10793, 2024

    Ya-xiang Yuan and Yi Zhang. Symplectic extra-gradient type method for solving general non- monotone inclusion problem.arXiv preprint arXiv:2406.10793, 2024

    2024 arXiv

  37. [45]

    Symplectic discretization approach for developing new proximal point algorithm.Computational Optimization and Applications, pages 1–37, 2025

    Ya-xiang Yuan and Yi Zhang. Symplectic discretization approach for developing new proximal point algorithm.Computational Optimization and Applications, pages 1–37, 2025

    2025

  38. [46]

    An efficient HPR algorithm for the Wasserstein barycenter problem withO(Dim(P)/ε) computational complexity.arXiv preprint arXiv:2211.14881, 2022

    Guojun Zhang, Yancheng Yuan, and Defeng Sun. An efficient HPR algorithm for the Wasserstein barycenter problem withO(Dim(P)/ε) computational complexity.arXiv preprint arXiv:2211.14881, 2022. 22 A A Counterexample to the Claim in [14] Consider the following variational inequali...

    2022 arXiv

Pith tools

Reviewed July 9, 2026 · model on record in the stance chip above.