Convergence of Nonmonotone Proximal Gradient Methods under the Kurdyka-Lojasiewicz Property without a Global Lipschitz Assumption

Christian Kanzow; Leo Lehmann

Convergence of Nonmonotone Proximal Gradient Methods under the Kurdyka-Lojasiewicz Property without a Global Lipschitz Assumption

T0 review · 0 major / 5 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 nonmonotone proximal gradient method converges with KL rates without global Lipschitz assumptions.

desk verdict Solid theory paper: proves KL-based whole-sequence convergence and rates for mean-type nonmonotone proximal gradient without global Lipschitz or bounded iterates; main caveat is the conditional subsequence assumption and an imported lemma. read the letter →

arxiv 2411.12376 v1 pith:TCNVTCIU submitted 2024-11-19 math.OC

Christian Kanzow , Leo Lehmann This is my paper · ORCID

classification math.OC MSC 49J5290C30

keywords compositeoptimizationnonsmoothproximalgradientmethodKurdyka–Łojasiewiczpropertynonmonotonelinesearchglobalconvergencelinearrateof

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 proves that the mean-type nonmonotone proximal gradient method, applied to a composite objective $\psi=f+\varphi$, converges to a single limit point whenever one accumulation point satisfies the Kurdyka--Lojasiewicz property, and it does so without requiring the gradient of $f$ to be globally Lipschitz or the iterates to be bounded in advance. Previously, whole-sequence convergence and rate guarantees of this kind were available only for monotone stepsize rules or under stronger assumptions. The proof shows that the nonmonotone reference values $R_k$ decrease monotonically and control the inter-iterate gaps, and that the KL property at an accumulation point forces the whole sequence to be a Cauchy sequence. With a power desingularization $\chi(t)=ct^\kappa$, the method then converges Q-linearly for the reference values and R-linearly for the iterates when $\kappa\in[1/2,1)$, and polynomially when $\kappa\in(0,1/2)$. If correct, nonmonotone proximal gradient methods inherit the same global convergence guarantees as their monotone counterparts, which matters because nonmonotone stepsizes often take larger steps.

What carries the argument

The mean-type nonmonotone rule maintains reference values $R_{k+1}=(1-p_{k+1})R_k+p_{k+1}\psi(x_{k+1})$ with $p_{k+1}\ge p_{\min}>0$, so that $\{R_k\}$ is monotonically decreasing even though $\{\psi(x_k)\}$ is not. The gap $\Xi_{k-1}=\sqrt{R_{k-1}-R_k}$ bounds the step length $\|x_k-x_{k-1}\|$ up to a constant, and a fixed memory window of length $m$ chosen from $p_{\min}$ keeps the KL-based recursion finite. The KL inequality, applied to the reference values rather than directly to the function values, yields a telescoping bound on sums of $\Xi_j$; that bound makes the iterates a Cauchy sequence without any global Lipschitz constant. A separate local argument keeps the stepsizes $\gamma_k$ bounded away from zero near the accumulation point using only local Lipschitz continuity of $f'$.

What would settle it

Run Algorithm 3.2 with $f(x)=e^x+x^4/4$, $\varphi(x)=|x|^{1/2}$, and parameters chosen inside the allowed ranges, watching whether the iterates with one finite accumulation point ever split toward a second accumulation point. Observing two distinct accumulation points would falsify Theorem 4.5, while observing convergence in all such runs would support it.

Watch

Extended reading notes

Core claim

The central claim is Theorem 4.5 and Theorem 4.6: under Assumption 3.1 (the objective is bounded below, $\varphi$ is bounded below by an affine function, and $f'$ is locally Lipschitz), if a subsequence of iterates generated by Algorithm 3.2 converges to $x^*$ and $\psi$ satisfies the Kurdyka--Lojasiewicz property at $x^*$, then the whole sequence $\{x_k\}$ converges to $x^*$. Moreover, when the desingularization function is $\chi(t)=ct^\kappa$, the reference values $R_k$ converge Q-linearly to $\psi(x^*)$ and the iterates converge R-linearly to $x^*$ for $\kappa\in[1/2,1)$, while for $\kappa\in(0,1/2)$ the stated polynomial bounds $R_k-\psi(x^*)\le\eta_1 k^{-1/(1-2\kappa)}$ and $\|x_k-x^*\|\le\eta_2 k^{-\kappa/(1-2\kappa)}$ hold. This is the first whole-sequence and rate result for a nonmonotone proximal gradient method that does not assume a global Lipschitz gradient or an a priori bound on the iterates.

Load-bearing premise

The argument assumes that some subsequence of the iterates already converges to a finite limit point $x^*$ at which the Kurdyka--Lojasiewicz property holds; if the iterates drift to infinity without an accumulation point, the theorem says nothing and the paper does not prove boundedness.

Editorial extensions

If this is right

Nonmonotone proximal gradient methods with the mean-type rule now carry the same whole-sequence convergence guarantee as monotone methods once the KL property holds at one accumulation point.
The rate dichotomy is explicit: $\kappa\in[1/2,1)$ gives Q-linear convergence of $R_k$ and R-linear convergence of $x_k$, while $\kappa\in(0,1/2)$ gives the polynomial estimates $k^{-1/(1-2\kappa)}$ and $k^{-\kappa/(1-2\kappa)}$.
No global Lipschitz constant of $f'$ has to be known or estimated; only local Lipschitz continuity is used, so the theory covers objectives like exponentials and higher-degree polynomials.
The result applies to the generalized projected gradient method for minimization over a nonempty closed set $S$ that is not necessarily convex, with the projection possibly set-valued.
If a generated sequence has any accumulation point satisfying KL, the whole sequence converges to that point; in particular, two distinct accumulation points cannot both carry the KL property.

Reading between the lines

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

A natural extension outside the paper's scope is to add coercivity or bounded-level-set conditions that guarantee the existence of a convergent subsequence, since the theorem itself only applies once such a subsequence exists.
The fixed memory length $m$ grows as $p_{\min}$ shrinks, so tuning $p_{\min}$ may trade off larger nonmonotone steps against longer KL windows; this is a quantitative prediction that could be tested numerically on sparse-regression problems.
The same reference-value recursion may extend to other nonmonotone first-order algorithms, such as inertial or Bregman proximal variants, whenever their reference values decrease monotonically; this is an extrapolation the paper does not make.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The paper does what it says: it gives the first KL-based whole-sequence convergence and rate results for the mean-type nonmonotone proximal gradient method when f' is only locally Lipschitz and the iterates are not assumed bounded in advance. That is a genuine gap, and the proof is a careful adaptation of the monotone machinery from Jia–Kanzow–Mehlitz and the recent framework of Qian et al. The main ingredients are Lemma 4.1, which shows accepted stepsizes stay bounded away from zero near an accumulation point using only local Lipschitzness, and Lemma 4.3, which bounds the limiting subgradient by the step norm. The induction argument in Theorem 4.5 is intricate but readable, and the rate results in Theorem 4.6 match what one would expect from the monotone theory. I followed the proof through the key steps and did not find an internal contradiction. Credit is due for stating the assumptions explicitly and for not hiding the conditional nature of the result.

The soft spots are real but not disqualifying. First, the whole-sequence convergence is conditional on the existence of a convergent subsequence with limit x*; the paper does not prove boundedness. That is clearly stated in Theorem 4.5, so it is not a hidden gap, but it does mean the headline “no a priori boundedness” should be read as “boundedness replaced by a subsequential convergence assumption.” Second, Theorem 3.4(c) from De Marchi, that R_k converges monotonically to psi(x*), is load-bearing and imported without proof. The authors note it follows from [14], and I see no evidence of a hidden global Lipschitz requirement there, but the dependency is substantial. Third, Lemma 4.4's stated hypotheses are slightly weaker than what Case 2 uses; the applications in Theorem 4.5 do supply the needed condition, so this is a minor presentation issue, not a mathematical flaw.

This paper is for researchers working on nonconvex, nonsmooth optimization and KL-based convergence theory. It deserves a serious referee: the claims are precise, the proof is detailed, and the result closes a known gap. I would send it out rather than desk reject, and I would expect acceptance after minor revisions to clarify the conditional nature of the main theorem and the reliance on [14].

Referee Report

0 major / 5 minor

Summary. The paper studies the composite minimization problem min ψ=f+φ with f continuously differentiable and φ proper lower semicontinuous extended-valued, using a nonmonotone proximal gradient method with the Zhang–Hager mean-type reference rule (Algorithm 3.2). Under Assumption 3.1 (ψ bounded below on dom(φ), φ bounded below by an affine function, and f' locally Lipschitz), the method has a finite stepsize inner loop and the iterate gaps satisfy ‖x_{k+1}−x_k‖→0. The main results, Theorems 4.5 and 4.6, state that if a subsequence {x_k}_K converges to some x* and ψ satisfies the Kurdyka–Łojasiewicz property at x*, then the entire sequence converges to x*; moreover, for desingularization χ(t)=ct^κ, the reference values R_k converge Q-linearly and the iterates x_k R-linearly when κ∈[1/2,1), and both converge polynomially at the stated rates when κ∈(0,1/2). The proof develops a stepsize lower-bound lemma (Lemma 4.1) using only local Lipschitz continuity, a limiting-subgradient bound (Lemma 4.3), and a nonmonotone adaptation of the summable-distance argument from [30].

Significance. If the results hold, they represent a meaningful advance in the convergence theory of nonmonotone proximal gradient methods: they show that the mean-type nonmonotone rule inherits essentially the same global and rate-of-convergence properties as its monotone counterpart without a global Lipschitz assumption and without an a priori bound on the iterates. The proofs are detailed and transparent, with no fitted constants or tuned parameters; all algorithmic parameters are interval-constrained inputs. The paper is honest about the conditional nature of its main theorems: convergence is not asserted unconditionally, but explicitly conditioned on the existence of a convergent subsequence and on the KL property at its limit. The rate results match the known monotone theory, which further supports the plausibility of the claims. The only external load-bearing ingredient is Theorem 3.4(c) imported from [14], whose exact hypotheses should be made explicit.

minor comments (5)

[Section 4, Theorem 3.4] Theorem 3.4(c), the monotone convergence of {R_k} to ψ(x*), is imported from [14] and is used in Lemmas 4.2, 4.4, and Theorem 4.6. Please state the precise result in [14] that implies this, and explicitly confirm that it holds under exactly Assumption 3.1 without additional hypotheses such as an a priori bound on the iterates or a global Lipschitz condition.
[Section 4, proof of Lemma 4.1] In the proof of Lemma 4.1, the phrase 'the boundedness condition from Assumption 3.1 for φ' is imprecise: Assumption 3.1(b) states that φ is bounded below by an affine function, not that φ is bounded. This wording should be changed to avoid confusion.
[Section 4, proof of Lemma 4.4] In the proof of Lemma 4.4, after applying the inequality √(x+y) ≤ √x + √y, the expression √(R_k − R_{k+n}) appears; this is a typo and should read √(R_k − R_{k+m}), consistent with the definition of m.
[Section 4, proof of Theorem 4.6] The step from s_{k+1} ≤ (1−p_min)s_k + p_min a (s_k−s_{k+1})^{1/(2(1−κ))} to s_k ≤ (1/p_min + a)(s_k−s_{k+1})^{min{1,1/(2(1−κ))}} is presented as 'This implies'; expanding this rearrangement would improve readability, since the exponent manipulation is central to the subsequent application of Lemma 2.2.
[Abstract and Introduction] The abstract and introduction state that the results are obtained 'without an a priori knowledge of the boundedness of the iterates'; this is accurate, but readers should be reminded that the theorems are conditional on the existence of a convergent subsequence. The theorem statements are clear, but an explicit sentence in the abstract would prevent an overstrong reading.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the whole-sequence KL convergence proof is self-contained and imports only weaker, external subsequential results.

full rationale

The paper's central claim is that, conditional on a given convergent subsequence {x_k}_K with limit x* and the KL property of psi at x*, the entire sequence converges and has the stated rates. This is not assumed in the input; it is proved. The proof does not fit any parameter to a target conclusion: gamma_k, alpha_k, beta_k, and p_k are interval-constrained algorithm inputs, and the desingularization chi(t)=c t^kappa is a hypothesis of the rate statement, not a fitted quantity. The main external ingredient is Theorem 3.4, especially part (c) that R_k converges monotonically to psi(x*), imported from De Marchi [14]. That result is weaker than the theorem being proved (subsequential stationarity and reference-value convergence versus whole-sequence convergence), it is external rather than self-citational, and its use is explicit. The authors' own prior works [19] and [21] supply the monotone template and basic lemmas, but the nonmonotone KL argument in Lemmas 4.1-4.4 and Theorems 4.5-4.6 is carried out in this paper and does not reduce to those citations. The subsequence assumption is stated transparently in the theorem and in the surrounding text, and the paper does not claim to prove boundedness without it. I therefore find no step where an output is equivalent by construction to an input, no fitted quantity renamed as a prediction, and no load-bearing self-citation chain.

Assumptions & free parameters 1 free parameters · 9 assumptions · 0 invented entities

The central claim rests on standard variational analysis machinery, the KL hypothesis, and a prior subsequential convergence theorem from [14]. No constants are fitted to data; the theorem is parameter-free in the sense that the algorithm parameters only need to lie in the stated intervals. The only imported nonstandard item is Theorem 3.4 from [14], which is cited rather than reproved.

free parameters (1)

Algorithm hyperparameters (gamma_min, gamma_max, alpha_min, alpha_max, beta_min, beta_max, p_min) = not fitted; user-chosen intervals
These are design parameters of Algorithm 3.2, not fitted to data. The convergence proof only needs them to lie in the stated open intervals; no numerical fitting is done.

assumptions (9)

domain assumption Assumption 3.1(a): psi is bounded from below on dom(phi).
Guarantees that R_k and psi(x_k) have finite lower limits, which is used throughout Lemma 3.3, Theorem 3.4, and the rate analysis.
domain assumption Assumption 3.1(b): phi is bounded from below by an affine function.
Ensures the proximal subproblems (1.2) have solutions and controls phi from below in Lemma 4.1 when trial points move.
domain assumption Assumption 3.1(c): f' is locally Lipschitz continuous.
Replaces global Lipschitz continuity; used in Lemma 4.1 to keep accepted stepsizes bounded away from zero near an accumulation point and in Lemma 4.3 to bound subgradients.
domain assumption The Kurdyka-Lojasiewicz property holds at an accumulation point x* with desingularization function chi.
This is the engine for whole-sequence convergence and rates; the paper assumes it rather than proving it for broad function classes.
domain assumption There exists a convergent subsequence {x_k}_K with limit x*.
Theorem 4.5 starts with a subsequence converging to x*; without it the argument says nothing, and the paper does not prove boundedness.
domain assumption The algorithm generates an infinite sequence, meaning the termination criterion in line 5 never triggers.
The paper states this explicitly; otherwise the method returns only an epsilon-stationary point and the convergence theory does not apply.
standard math Theorem 3.4 from De Marchi [14]: every accumulation point is M-stationary, psi(x_k) converges to psi(x*), and R_k converges monotonically to psi(x*).
Imported without proof; Lemma 4.2 and the proof of Theorem 4.5 rely on R_k converging to psi(x*) to make the constant alpha small.
standard math Standard variational analysis facts, including the sum rule (2.1) and robustness of the limiting subdifferential.
Used to identify M-stationarity and to derive the subgradient bound (4.10) used in Lemma 4.3.
standard math Lemma 2.2 from [1] on rates implied by s_k^alpha <= beta (s_k - s_{k+1}).
Converts the inequalities in the proof of Theorem 4.6 into the stated Q-linear and polynomial rates.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Convergence of Nonmonotone Proximal Gradient Methods under the Kurdyka-Lojasiewicz Property without a Global Lipschitz Assumption." pith.science (2026). https://pith.science/paper/TCNVTCIU

@misc{pith2026241112376,
  author       = {Pith},
  title        = {Pith review of: Convergence of Nonmonotone Proximal Gradient Methods under the Kurdyka-Lojasiewicz Property without a Global Lipschitz Assumption},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TCNVTCIU}},
  note         = {Machine review of arXiv:2411.12376}
}

read the original abstract

We consider the composite minimization problem with the objective function being the sum of a continuously differentiable and a merely lower semicontinuous and extended-valued function. The proximal gradient method is probably the most popular solver for this class of problems. Its convergence theory typically requires that either the gradient of the smooth part of the objective function is globally Lipschitz continuous or the (implicit or explicit) a priori assumption that the iterates generated by this method are bounded. Some recent results show that, without these assumptions, the proximal gradient method, combined with a monotone stepsize strategy, is still globally convergent with a suitable rate-of-convergence under the Kurdyka-Lojasiewicz property. For a nonmonotone stepsize strategy, there exist some attempts to verify similar convergence results, but, so far, they need stronger assumptions. This paper is the first which shows that nonmonotone proximal gradient methods for composite optimization problems share essentially the same nice global and rate-of-convergence properties as its monotone counterparts, still without assuming a global Lipschitz assumption and without an a priori knowledge of the boundedness of the iterates.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

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

Monotone and nonmonotone linearized block coordinate descent methods for nonsmooth composite optimization problems
math.OC 2025-06 conditional novelty 6.0 of 10

Two linearized block coordinate descent algorithms for nonsmooth composite optimization converge in expectation to a stationary point at O(1/ε²) rate.
Convergence of linesearch-based generalized conditional gradient methods without smoothness assumptions
math.OC 2025-05 accept novelty 6.0 of 10

Two linesearch-based generalized conditional gradient methods, including a parameter-free variant from Ito et al., are proven to converge to stationary points without any Lipschitz or Hölder gradient assumption.
Convergence analysis of nonmonotone proximal gradient methods under local Lipschitz continuity and Kurdyka--{\L}ojasiewicz property
math.OC 2024-11 conditional novelty 6.0 of 10

Nonmonotone proximal gradient methods with average or max line search converge, with rates, under local gradient Lipschitz continuity and the Kurdyka-Lojasiewicz property, without boundedness of iterates.

Reference graph

Works this paper leans on

34 extracted references · 15 canonical work pages · cited by 3 Pith papers

[14]

Proximal gradient methods beyond monoto ny

A. De Marchi. “Proximal gradient methods beyond monoto ny”. In: Journal of Non- smooth Analysis and Optimization 4 (2023)

work page 2023
[30]

Convergence of ZH-typ e nonmonotone descent method for Kurdyka-Łojasiewicz optimization problems

Y. Qian, T. Tao, S. Pan, and H. Qi. “Convergence of ZH-typ e nonmonotone descent method for Kurdyka-Łojasiewicz optimization problems”. I n: arXiv:2406.05740 (2024)

arXiv 2024
[1]

Ac celerating the DC al- gorithm for smooth functions

F. J. Aragón Artacho, R. M. T. Fleming, and P. T. Vuong. “Ac celerating the DC al- gorithm for smooth functions”. In: Mathematical Programming 169.1 (2018), pp. 95–

work page 2018
[2]

On the convergence of the proxim al algorithm for nons- mooth functions involving analytic features

H. Attouch and J. Bolte. “On the convergence of the proxim al algorithm for nons- mooth functions involving analytic features”. In: Mathematical Programming 116.1 (2009), pp. 5–16. doi: 10.1007/s10107-007-0133-5

doi:10.1007/s10107-007-0133-5 2009
[3]

Proxi mal alternating mini- mization and projection methods for nonconvex problems: An approach based on the Kurdyka-Łojasiewicz inequality

H. Attouch, J. Bolte, P. Redont, and A. Soubeyran. “Proxi mal alternating mini- mization and projection methods for nonconvex problems: An approach based on the Kurdyka-Łojasiewicz inequality”. In: Mathematics of Operations Research 35.2 (2010), pp. 438–457. doi: 10.1287/moor.1100.0449

work page arXiv 2010
[4]

Convergence of d escent methods for semi-algebraic and tame problems, proximal algorithms, fo rward-backward split- ting, and regularized Gauss–Seidel methods

H. Attouch, J. Bolte, and B. F. Svaiter. “Convergence of d escent methods for semi-algebraic and tame problems, proximal algorithms, fo rward-backward split- ting, and regularized Gauss–Seidel methods”. In: Mathematical Programming 137 (2013), pp. 91 –129. doi: 10.1007/s10107-011-0484-9 . 18

doi:10.1007/s10107-011-0484-9 2013
[5]

A. Beck. First-Order Methods in Optimization. SIAM, 2017. doi: 10.1137/1.9781611974997

doi:10.1137/1.9781611974997 2017
[6]

Nonmonotone spectral projected gradient methods on convex sets

E. G. Birgin, J. M. Martínez, and M. Raydan. “Nonmonotone spectral projected gradient methods on convex sets”. In: SIAM Journal on Optimization 10.4 (2000), pp. 1196–1211

work page 2000

Show all 34 references

[7]

The Łojasiewicz i nequality for nonsmooth sub- analytic functions with applications to subgradient dynam ical systems

J. Bolte, A. Daniilidis, and A. Lewis. “The Łojasiewicz i nequality for nonsmooth sub- analytic functions with applications to subgradient dynam ical systems”. In: SIAM Journal on Optimization 17.4 (2007), pp. 1205–1223. doi: 10.1137/050644641

2007 doi
[8]

Clarke subgradients of stratiﬁ- able functions

J. Bolte, A. Daniilidis, A. Lewis, and M. Shiota. “Clarke subgradients of stratiﬁ- able functions”. In: SIAM Journal on Optimization 18.2 (2007), pp. 556–572. doi: 10.1137/060670080

2007 doi
[9]

Proximal alternat ing linearized minimiza- tion for nonconvex and nonsmooth problems

J. Bolte, S. Sabach, and M. Teboulle. “Proximal alternat ing linearized minimiza- tion for nonconvex and nonsmooth problems”. In: Mathematical Programming 146 (2014), pp. 459 –494. doi: 10.1007/s10107-013-0701-9

2014 doi
[10]

An inertial Tseng’s type pro ximal algorithm for non- smooth and nonconvex optimization problems

R. I. Boţ and E. R. Csetnek. “An inertial Tseng’s type pro ximal algorithm for non- smooth and nonconvex optimization problems”. In: Journal of Optimization Theory and Applications 171.2 (2016), pp. 600–616. doi: 10.1007/s10957-015-0730-z

2016 doi
[11]

An inertial fo rward–backward algorithm for the minimization of the sum of two nonconvex functions

R. I. Boţ, E. R. Csetnek, and S. C. László. “An inertial fo rward–backward algorithm for the minimization of the sum of two nonconvex functions”. In: EURO Journal on Computational Optimization 4.1 (2016), pp. 3–25. doi: 10.1007/s13675-015-0045-8

2016 doi
[12]

From sparse solutions of systems of equations to sparse modeling of signals and images

A. M. Bruckstein, D. L. Donoho, and M. Elad. “From sparse solutions of systems of equations to sparse modeling of signals and images”. In: SIAM Review 51.1 (2009), pp. 34–81. doi: 10.1137/060657704

2009 doi
[13]

Exact reconstruction of sparse signals via nonconvex minimization

R. Chartrand. “Exact reconstruction of sparse signals via nonconvex minimization”. In: IEEE Signal Processing Letters 14.10 (2007), pp. 707–710. doi: 10.1109/LSP.2007.898300

2007
[15]

Constra ined composite op- timization and augmented Lagrangian methods

A. De Marchi, X. Jia, C. Kanzow, and P. Mehlitz. “Constra ined composite op- timization and augmented Lagrangian methods”. In: Mathematical Programming 201.1 (2023), pp. 863–896

2023
[16]

A concave optimization-based approach for sparse portfolio selecti on

D. Di Lorenzo, G. Liuzzi, F. Rinaldi, F. Schoen, and M. Sc iandrone. “A concave optimization-based approach for sparse portfolio selecti on”. In: Optimization Meth- ods and Software 27.6 (2012), pp. 983–1000. doi: 10.1080/10556788.2011.577773

2012
[17]

Dumitrescu and P

B. Dumitrescu and P. Irofti. Dictionary Learning Algorithms and Applications . Springer, 2018

2018
[18]

A nonmonoton e line search technique for Newton’s method

L. Grippo, F. Lampariello, and S. Lucidi. “A nonmonoton e line search technique for Newton’s method”. In: SIAM Journal on Numerical Analysis 23.4 (1986), pp. 707– 716

1986
[19]

Convergence Analysi s of the Proximal Gradient Method in the Presence of the Kurdyka–Łojasiewicz Property Without Global Lips- chitz Assumptions

X. Jia, C. Kanzow, and P. Mehlitz. “Convergence Analysi s of the Proximal Gradient Method in the Presence of the Kurdyka–Łojasiewicz Property Without Global Lips- chitz Assumptions”. In: SIAM Journal on Optimization 33.4 (2023), pp. 3038–3056. doi: 10.1137/23M1548293. url: htt...

2023 doi
[20]

An augm ented Lagrangian method for optimization problems with structured geometri c constraints

X. Jia, C. Kanzow, P. Mehlitz, and G. Wachsmuth. “An augm ented Lagrangian method for optimization problems with structured geometri c constraints”. In: Math- ematical Programming (2022). doi: 10.1007/s10107-022-01870-z . 19

2022 doi
[21]

Convergence properties of mo notone and nonmono- tone proximal gradient methods revisited

C. Kanzow and P. Mehlitz. “Convergence properties of mo notone and nonmono- tone proximal gradient methods revisited”. In: Journal of Optimization Theory and Applications 195.2 (2022), pp. 624–646. doi: 10.1007/s10957-022-02101-3

2022 doi
[22]

On gradients of functions deﬁnable in o-mi nimal structures

K. Kurdyka. “On gradients of functions deﬁnable in o-mi nimal structures”. In: An- nales de l’institut Fourier 48.3 (1998), pp. 769–783. doi: 10.5802/aif.1638

1998 doi
[23]

Joint power and admissi on control: non-convex ℓq approximation and an eﬀective polynomial time deﬂation app roach

Y.-F. Liu, Y.-H. Dai, and S. Ma. “Joint power and admissi on control: non-convex ℓq approximation and an eﬀective polynomial time deﬂation app roach”. In: IEEE Transactions on Signal Processing 63.14 (2015), pp. 3641–3656. doi: 10.1109/TSP.2015.2428224

2015
[24]

On ℓq optimization and matrix completion

G. Marjanovic and V. Solo. “On ℓq optimization and matrix completion”. In: IEEE Transactions on Signal Processing 60.11 (2012), pp. 5714–5724. doi: 10.1109/TSP.2012.2212015

2012
[25]

Markovsky

I. Markovsky. Low Rank Approximation . Springer, 2018

2018
[26]

B. S. Mordukhovich. Variational Analysis and Applications . Springer, 2018. doi: 10.1007/978-3-319-92775-6

2018 doi
[27]

Local convergence of the heavy-ball method an d iPiano for non-convex optimization

P. Ochs. “Local convergence of the heavy-ball method an d iPiano for non-convex optimization”. In: Journal of Optimization Theory and Applications 177.1 (2018), pp. 153–180. doi: 10.1007/s10957-018-1272-y

2018 doi
[28]

iPiano: Inertial proximal algorithm for nonconvex optimization

P. Ochs, Y. Chen, T. Brox, and T. Pock. “iPiano: Inertial proximal algorithm for nonconvex optimization”. In: SIAM Journal on Imaging Sciences 7.2 (2014), pp. 1388–1419. doi: 10.1137/130942954

2014 doi
[29]

Convergence of a class of nonmonoton e descent methods for KL optimization problems

Y. Qian and S. Pan. “Convergence of a class of nonmonoton e descent methods for KL optimization problems”. In: arXiv preprint arXiv:2201.00639 (2022)

2022 arXiv
[31]

R. T. Rockafellar and R. J.-B. Wets. Variational Analysis . Springer, 2009. doi: 10.1007/978-3-642-02431-3

2009 doi
[32]

Variational ana lysis perspective on linear convergence of some ﬁrst order methods for nonsmooth convex optimization problems

J. J. Ye, X. Yuan, S. Zeng, and J. Zhang. “Variational ana lysis perspective on linear convergence of some ﬁrst order methods for nonsmooth convex optimization problems”. In: Set-Valued and Variational Analysis (2021), pp. 1–35

2021
[33]

A nonmonotone line search tech nique and its ap- plication to unconstrained optimization

H. Zhang and W. W. Hager. “A nonmonotone line search tech nique and its ap- plication to unconstrained optimization”. In: SIAM Journal on Optimization 14.4 (2004), pp. 1043–1056. 20

2004
[118]

doi: 10.1007/s10107-017-1180-1

doi

Pith tools

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

[1] [14]

Proximal gradient methods beyond monoto ny

A. De Marchi. “Proximal gradient methods beyond monoto ny”. In: Journal of Non- smooth Analysis and Optimization 4 (2023)

work page 2023

[2] [30]

Convergence of ZH-typ e nonmonotone descent method for Kurdyka-Łojasiewicz optimization problems

Y. Qian, T. Tao, S. Pan, and H. Qi. “Convergence of ZH-typ e nonmonotone descent method for Kurdyka-Łojasiewicz optimization problems”. I n: arXiv:2406.05740 (2024)

arXiv 2024

[3] [1]

Ac celerating the DC al- gorithm for smooth functions

F. J. Aragón Artacho, R. M. T. Fleming, and P. T. Vuong. “Ac celerating the DC al- gorithm for smooth functions”. In: Mathematical Programming 169.1 (2018), pp. 95–

work page 2018

[4] [2]

On the convergence of the proxim al algorithm for nons- mooth functions involving analytic features

H. Attouch and J. Bolte. “On the convergence of the proxim al algorithm for nons- mooth functions involving analytic features”. In: Mathematical Programming 116.1 (2009), pp. 5–16. doi: 10.1007/s10107-007-0133-5

doi:10.1007/s10107-007-0133-5 2009

[5] [3]

Proxi mal alternating mini- mization and projection methods for nonconvex problems: An approach based on the Kurdyka-Łojasiewicz inequality

H. Attouch, J. Bolte, P. Redont, and A. Soubeyran. “Proxi mal alternating mini- mization and projection methods for nonconvex problems: An approach based on the Kurdyka-Łojasiewicz inequality”. In: Mathematics of Operations Research 35.2 (2010), pp. 438–457. doi: 10.1287/moor.1100.0449

work page arXiv 2010

[6] [4]

Convergence of d escent methods for semi-algebraic and tame problems, proximal algorithms, fo rward-backward split- ting, and regularized Gauss–Seidel methods

H. Attouch, J. Bolte, and B. F. Svaiter. “Convergence of d escent methods for semi-algebraic and tame problems, proximal algorithms, fo rward-backward split- ting, and regularized Gauss–Seidel methods”. In: Mathematical Programming 137 (2013), pp. 91 –129. doi: 10.1007/s10107-011-0484-9 . 18

doi:10.1007/s10107-011-0484-9 2013

[7] [5]

A. Beck. First-Order Methods in Optimization. SIAM, 2017. doi: 10.1137/1.9781611974997

doi:10.1137/1.9781611974997 2017

[8] [6]

Nonmonotone spectral projected gradient methods on convex sets

E. G. Birgin, J. M. Martínez, and M. Raydan. “Nonmonotone spectral projected gradient methods on convex sets”. In: SIAM Journal on Optimization 10.4 (2000), pp. 1196–1211

work page 2000