pith. sign in

arxiv: 2406.00648 · v3 · submitted 2024-06-02 · 🧮 math.OC

Level proximal subdifferential, variational convexity, and pointwise quadratic approximation

Honglin Luo , Xianfu Wang , Ziyuan Wang , Xinmin Yang This is my paper

Pith reviewed 2026-05-24 00:10 UTC · model grok-4.3

classification 🧮 math.OC
keywords level proximal subdifferentialvariational convexityproximal mappingproximal gradient methodpointwise quadratic approximationvariational analysisnonconvex optimizationfirm nonexpansiveness
0
0 comments X

The pith

Variational convexity holds exactly when the proximal mapping is locally firmly nonexpansive.

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

The paper shows that a function satisfies variational convexity precisely when its proximal mapping is locally firmly nonexpansive or its level proximal subdifferential is locally relatively monotone. This equivalence is then applied to prove that the proximal gradient method converges to local minimizers rather than stopping at critical points. The same subdifferential is used to study existence, single-valuedness, integration, and pointwise quadratic approximation properties. A reader cares because these characterizations turn a local geometric property into a practical test for when standard algorithms succeed on nonconvex problems.

Core claim

Level proximal subdifferential, introduced by Rockafellar for nonconvex functions, is studied systematically. Variational convexity of a function is characterized by local firm nonexpansiveness of its proximal mappings or by local relative monotonicity of the level proximal subdifferential. These characterizations imply that proximal gradient methods and related algorithms converge locally to minimizers when variational sufficiency holds. The paper also establishes results on existence, single-valuedness, integration of the subdifferential, and quantifies pointwise quadratic approximation or Lipschitz smoothness.

What carries the argument

The level proximal subdifferential, a set-valued mapping that captures proximal information at a given level, used to translate variational convexity into monotonicity or nonexpansiveness conditions on proximal mappings.

If this is right

  • Proximal gradient methods converge to local minimizers (not merely critical points) for variationally convex functions.
  • Variational sufficiency becomes a verifiable condition that upgrades algorithm behavior from stationarity to local optimality.
  • Pointwise quadratic approximation of a function can be quantified directly through the level proximal subdifferential.
  • Existence and single-valuedness of the level proximal subdifferential follow from the same monotonicity conditions used for convexity.
  • Integration theorems recover the original function from its level proximal subdifferential under the local monotonicity assumption.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same local monotonicity test could be checked numerically to certify that a given nonconvex problem admits reliable local convergence of first-order methods.
  • Connections may exist to other subdifferential notions that already appear in convergence analyses of proximal algorithms.
  • The framework might extend to composite problems or to second-order conditions that strengthen the quadratic approximation results.
  • Local relative monotonicity could serve as a practical surrogate for checking variational convexity in high-dimensional settings where direct verification is hard.

Load-bearing premise

The level proximal subdifferential must be well-defined and obey the basic calculus rules that let monotonicity and nonexpansiveness properties transfer between the subdifferential and the proximal mapping.

What would settle it

A concrete function whose proximal mapping is locally firmly nonexpansive yet fails to be variationally convex, or whose level proximal subdifferential fails local relative monotonicity while the function is variationally convex.

Figures

Figures reproduced from arXiv: 2406.00648 by Honglin Luo, Xianfu Wang, Xinmin Yang, Ziyuan Wang.

Figure 1
Figure 1. Figure 1: Correspondence between convex and variationally [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 1
Figure 1. Figure 1: (ii) One distinguished feature of Theorem 4.6(viii)⇒(i) is that only one λ is required instead of all 0 < λ < λ¯; see [20, Theorem 3.2]. This is exactly what we need in Corollary 5.9 and Theorem 5.12 for variational sufficiency. (iii) Recently, in [19] among many other results Kanhn, Khoa, Mordukhovich and Phat have extended the equivalence of Theorem 4.6(i)⇔(vii) to infinite-dimensional spaces. Next we tu… view at source ↗
Figure 2
Figure 2. Figure 2: Proximal operators and Moreau envelope in Example [PITH_FULL_IMAGE:figures/full_fig_p029_2.png] view at source ↗
read the original abstract

Level proximal subdifferential was introduced by Rockafellar recently for studying proximal mappings of possibly nonconvex functions. In this paper a systematic study of level proximal subdifferential is given. We characterize variational convexity of a function by local firm nonexpansiveness of proximal mappings or local relative monotonicity of level proximal subdifferential, and use them to study local convergence of proximal gradient method and others for variationally convex functions. Variational sufficiency guarantees that proximal gradient method converges to local minimizers rather than just critical points. We also investigate the existence, single-valuedness and integration of level proximal subdifferential, and quantify pointwise quadratic approximation (or Lipschitz smoothness) of a function. As a powerful tool, level proximal subdifferential provides deep insights into variational analysis and optimization.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper provides a systematic study of the level proximal subdifferential (recently introduced by Rockafellar) for possibly nonconvex functions. It characterizes variational convexity via local firm nonexpansiveness of proximal mappings or local relative monotonicity of the level proximal subdifferential, applies these to local convergence analysis of the proximal gradient method (and others) for variationally convex functions, shows that variational sufficiency ensures convergence to local minimizers, and further investigates existence/single-valuedness/integration of the subdifferential along with pointwise quadratic approximation/Lipschitz smoothness.

Significance. If the characterizations and convergence results hold in the stated generality, the work supplies new variational-analytic tools linking subdifferential monotonicity properties to algorithm behavior in nonconvex settings, with potential to clarify when proximal methods reach local minimizers. The quantification of pointwise quadratic approximation adds a concrete link to smoothness notions.

major comments (2)
  1. [Systematic study section] Systematic study section (characterizations of variational convexity): The claimed equivalences between variational convexity and local relative monotonicity of the level proximal subdifferential (or local firm nonexpansiveness of proximal mappings) invoke calculus rules and monotonicity properties of the level subdifferential; it is unclear whether these identities hold without extra regularity (e.g., prox-regularity or a fixed level parameter) that is not implied by the variational-convexity hypothesis alone. This is load-bearing for all subsequent claims.
  2. [Local convergence section] Local convergence section (proximal gradient method): The assertion that variational sufficiency guarantees convergence to local minimizers (rather than merely critical points) rests on the preceding equivalences; the precise step-size conditions, neighborhood size, and any implicit lower-semicontinuity assumptions needed for the argument must be stated explicitly, as gaps here would undermine the main algorithmic contribution.
minor comments (2)
  1. The abstract refers to convergence results for 'proximal gradient method and others'; the relevant section should explicitly enumerate the additional methods analyzed.
  2. Notation for the level proximal subdifferential and the level parameter should be introduced with a dedicated preliminary subsection to improve readability before the characterizations.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful and constructive review. We address each major comment below.

read point-by-point responses

  1. Referee: [Systematic study section] Systematic study section (characterizations of variational convexity): The claimed equivalences between variational convexity and local relative monotonicity of the level proximal subdifferential (or local firm nonexpansiveness of proximal mappings) invoke calculus rules and monotonicity properties of the level subdifferential; it is unclear whether these identities hold without extra regularity (e.g., prox-regularity or a fixed level parameter) that is not implied by the variational-convexity hypothesis alone. This is load-bearing for all subsequent claims.

    Authors: The equivalences are derived directly from the definition of variational convexity together with the construction of the level proximal subdifferential; no prox-regularity or other regularity beyond the variational-convexity hypothesis is used. The level parameter is selected locally (sufficiently small relative to the point under consideration) and is not required to be fixed globally. The relevant calculus rules follow from the monotonicity properties that are equivalent to variational convexity by definition. We will add a clarifying remark after the main characterization theorem to make this explicit and to reference the precise propositions used. revision: partial

  2. Referee: [Local convergence section] Local convergence section (proximal gradient method): The assertion that variational sufficiency guarantees convergence to local minimizers (rather than merely critical points) rests on the preceding equivalences; the precise step-size conditions, neighborhood size, and any implicit lower-semicontinuity assumptions needed for the argument must be stated explicitly, as gaps here would undermine the main algorithmic contribution.

    Authors: We agree that the hypotheses should be stated more explicitly. The proof relies on the local firm nonexpansiveness established earlier, which keeps iterates inside a neighborhood on which variational convexity holds; the step-size must be positive and smaller than a constant determined by the local modulus of the proximal mapping, the initial point must lie in a sufficiently small ball around the local minimizer, and the function is assumed lower semicontinuous (as is standard for the proximal mapping to be well-defined). We will revise the statement of the convergence theorem and the surrounding discussion to list these conditions explicitly. revision: yes

Circularity Check

0 steps flagged

No circularity: characterizations derived from external Rockafellar definition

full rationale

The paper defines no quantities in terms of their own outputs. All central equivalences (variational convexity ↔ local firm nonexpansiveness of proximal mappings, or local relative monotonicity of level proximal subdifferential) are derived from the externally introduced level proximal subdifferential (Rockafellar) plus standard calculus rules. No parameters are fitted then relabeled as predictions, no self-citation chains carry the load-bearing steps, and no ansatz is smuggled via prior work by the same authors. The derivation chain is therefore self-contained against the stated external definition and does not reduce any claimed result to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper relies on standard variational analysis background rather than new free parameters or invented entities. The level proximal subdifferential itself originates from Rockafellar and is treated as given.

axioms (2)
  • domain assumption Functions are proper lower semicontinuous (standard in proximal analysis)
    Invoked implicitly for subdifferential definitions and proximal mappings to be well-defined.
  • domain assumption Rockafellar's definition and basic properties of level proximal subdifferential hold
    The entire study rests on this recent external definition.

pith-pipeline@v0.9.0 · 5659 in / 1324 out tokens · 17475 ms · 2026-05-24T00:10:38.723408+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

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

  1. Bregman level proximal subdifferentials and new characterizations of Bregman proximal operators

    math.OC 2025-06 unverdicted novelty 6.0

    Defines Bregman level proximal subdifferentials whose resolvents recover Bregman proximal operators under a range assumption, yielding new characterizations and properties without convexity.

Reference graph

Works this paper leans on

48 extracted references · 48 canonical work pages · cited by 1 Pith paper

  1. [1]

    Attouch and J

    H. Attouch and J. Bolte , On the convergence of the proximal algorithm for nonsmooth functions involving analytic features , Mathematical Programming, 116 (2009), pp. 5–16. 32

    work page 2009

  2. [2]

    Attouch, J

    H. Attouch, J. Bolte, and B. F. Svaiter , Convergence of descent methods for semi- algebraic and tame problems: proximal algorithms, forward –backward splitting, and regularized gauss–seidel methods, Mathematical Programming, 137 (2013), pp. 91–129

    work page 2013

  3. [3]

    Bachir, A

    M. Bachir, A. Daniilidis, and J.-P. Penot , Lower subdifferentiability and integration , Set-Valued Analysis, 10 (2002), pp. 89–108

    work page 2002

  4. [4]

    H. H. Bauschke, J. Bolte, and M. Teboulle , A descent lemma beyond Lipschitz gra- dient continuity: first-order methods revisited and applic ations, Mathematics of Operations Research, 42 (2017), pp. 330–348

    work page 2017

  5. [5]

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

    work page 2017

  6. [6]

    H. H. Bauschke, W. M. Moursi, and X. W ang , Generalized monotone operators and their averaged resolvents , Mathematical Programming, 189 (2021), pp. 55–74

    work page 2021

  7. [7]

    Beck , First-order Methods in Optimization , MOS-SIAM Series on Optimization, SIAM, Philadelphia, PA, 2017

    A. Beck , First-order Methods in Optimization , MOS-SIAM Series on Optimization, SIAM, Philadelphia, PA, 2017

    work page 2017

  8. [8]

    Benoist and A

    J. Benoist and A. Daniilidis , Integration of Fenchel subdifferentials of epi-pointed func - tions, SIAM Journal on Optimization, 12 (2002), pp. 575–582

    work page 2002

  9. [9]

    Benoist and J.-B

    J. Benoist and J.-B. Hiriart-Urruty , What is the subdifferential of the closed convex hull of a function? , SIAM Journal on Mathematical Analysis, 27 (1996), pp. 1661 –1679

    work page 1996

  10. [10]

    Bernard and L

    F. Bernard and L. Thibault , Prox-regular functions in Hilbert spaces , Journal of Mathe- matical Analysis and Applications, 303 (2005), pp. 1–14

    work page 2005

  11. [11]

    Bolte, S

    J. Bolte, S. Sabach, and M. Teboulle , Proximal alternating linearized minimization for nonconvex and nonsmooth problems , Mathematical Programming, 146 (2014), pp. 459–494

    work page 2014

  12. [12]

    J. M. Borwein, R. Girgensohn, and X. W ang , On the construction of H¨ older and prox- imal subderivatives , Canadian Mathematical Bulletin, 41 (1998), pp. 497–507

    work page 1998

  13. [13]

    J. Chen, X. W ang, and C. Planiden , A proximal average for prox-bounded functions , SIAM Journal on Optimization, 30 (2020), pp. 1366–1390

    work page 2020

  14. [14]

    F. H. Clarke, Y. S. Ledyaev, and P. R. Wolenski , Proximal analysis and minimization principles, Journal of Mathematical Analysis and Applications, 196 (1 995), pp. 722–735

    work page

  15. [15]

    F. H. Clarke and R. M. Redheffer , The proximal subgradient and constancy , Canadian Mathematical Bulletin, 36 (1993), pp. 30–32

    work page 1993

  16. [16]

    F abian and V

    M. F abian and V. Zizler, An elementary approach to some questions in higher order smo oth- ness in Banach spaces , Extracta Mathematicae, 14 (1999), pp. 295–328

    work page 1999

  17. [17]

    Hiriart-Urruty and C

    J.-B. Hiriart-Urruty and C. Lemar ´ echal, Fundamentals of Convex Analysis , Springer Science & Business Media, 2004

    work page 2004

  18. [18]

    Jourani, L

    A. Jourani, L. Thibault, and D. Zagrodny , Differential properties of the Moreau enve- lope, Journal of Functional Analysis, 266 (2014), pp. 1185–1237

    work page 2014

  19. [19]

    P. D. Khanh, V. V. H. Khoa, B. S. Mordukhovich, and V. T. Phat , Variational and strong variational convexity in infinite-dimensional v ariational analysis, SIAM Journal on Optimization, 34 (2024), pp. 2756–2787. 33

    work page 2024

  20. [20]

    P. D. Khanh, B. S. Mordukhovich, and V. T. Phat , Variational convexity of func- tions and variational sufficiency in optimization , SIAM Journal on Optimization, 33 (2023), pp. 1121–1158

    work page 2023

  21. [21]

    Kreyszig , Introductory Functional Analysis with Applications , Wiley Classics Library, John Wiley & Sons, Inc., New York, 1989

    E. Kreyszig , Introductory Functional Analysis with Applications , Wiley Classics Library, John Wiley & Sons, Inc., New York, 1989

    work page 1989

  22. [22]

    H. Luo, X. W ang, and X. Yang , Various notions of nonexpansiveness coincide for proximal mappings of functions , SIAM Journal on Optimization, 34 (2024), pp. 642–653

    work page 2024

  23. [23]

    B. S. Mordukhovich , Variational Analysis and Generalized Differentiation I: Bas ic Theory, Springer-Verlag, Berlin, 2006

    work page 2006

  24. [24]

    B. S. Mordukhovich and T. T. A. Nghia , Local monotonicity and full stability for para- metric variational systems , SIAM Journal on Optimization, 26 (2016), pp. 1032–1059

    work page 2016

  25. [25]

    Moreau , Proximit´ e et dualit´ e dans un espace hilbertien , Bulletin de la Soci´ et´ e Math´ ematique de France, 93 (1965), pp

    J.-J. Moreau , Proximit´ e et dualit´ e dans un espace hilbertien , Bulletin de la Soci´ et´ e Math´ ematique de France, 93 (1965), pp. 273–299

    work page 1965

  26. [26]

    Nesterov , Lectures on Convex Optimization , Springer Optimization and Its Applications, Springer, Cham, second ed., 2018

    Y. Nesterov , Lectures on Convex Optimization , Springer Optimization and Its Applications, Springer, Cham, second ed., 2018

    work page 2018

  27. [27]

    Parikh and S

    N. Parikh and S. Boyd , Proximal algorithms , Foundations and Trends in Optimization, 1 (2013), pp. 123–231

    work page 2013

  28. [28]

    Penot , Proximal mappings, Journal of Approximation Theory, 94 (1998), pp

    J.-P. Penot , Proximal mappings, Journal of Approximation Theory, 94 (1998), pp. 203–221

    work page 1998

  29. [29]

    Penot and M

    J.-P. Penot and M. L. Bougeard , Approximation and decomposition properties of some classes of locally D.C. functions , Mathematical Programming, 41 (1988), pp. 195–227

    work page 1988

  30. [30]

    Planiden and X

    C. Planiden and X. W ang , Strongly convex functions, Moreau envelopes, and the gener ic nature of convex functions with strong minimizers , SIAM Journal on Optimization, 26 (2016), pp. 1341–1364

    work page 2016

  31. [31]

    R. A. Poliquin , Integration of subdifferentials of nonconvex functions , Nonlinear Analysis: Theory, Methods & Applications, 17 (1991), pp. 385–398

    work page 1991

  32. [32]

    R. A. Poliquin and R. T. Rockafellar , Prox-regular functions in variational analysis , Transactions of the American Mathematical Society, 348 (19 96), pp. 1805–1838

    work page

  33. [33]

    R. T. Rockafellar , Monotone operators and the proximal point algorithm , SIAM Journal on Control and Optimization, 14 (1976), pp. 877–898

    work page 1976

  34. [34]

    R. T. Rockafellar , Variational convexity and the local monotonicity of subgra dient map- pings, Vietnam Journal of Mathematics, 47 (2019), pp. 547–561

    work page 2019

  35. [35]

    , Characterizing firm nonexpansiveness of prox mappings both l ocally and globally , Journal of Nonlinear and Convex Analysis, 22 (2021), pp. 887–899

    work page 2021

  36. [36]

    R. T. Rockafellar , Augmented lagrangians and hidden convexity in sufficient con ditions for local optimality , Mathematical Programming, 198 (2023), pp. 159–194

    work page 2023

  37. [37]

    , Convergence of augmented lagrangian methods in extensions b eyond nonlinear program- ming, Mathematical Programming, 199 (2023), pp. 375–420

    work page 2023

  38. [38]

    R. T. Rockafellar and R. J. B. Wets , Variational Analysis, Springer, Berlin, 1998. 34

    work page 1998

  39. [39]

    E. K. Ryu and W. Yin , Large-scale Convex Optimization: Algorithms & Analyses via Mono- tone Operators, Cambridge University Press, 2022

    work page 2022

  40. [40]

    Soloviov, Duality for nonconvex optimization and its applications , Analysis Mathematica, 19 (1993), pp

    V. Soloviov, Duality for nonconvex optimization and its applications , Analysis Mathematica, 19 (1993), pp. 297–315

    work page 1993

  41. [41]

    Str ¨omberg, Duality between Fr´ echet differentiability and strong convexity, Positivity, 15 (2011), pp

    T. Str ¨omberg, Duality between Fr´ echet differentiability and strong convexity, Positivity, 15 (2011), pp. 527–536

    work page 2011

  42. [42]

    Thibault and D

    L. Thibault and D. Zagrodny , Integration of subdifferentials of lower semicontinuous functions on Banach spaces , Journal of Mathematical Analysis and Applications, 189 (1 995), pp. 33–58

    work page

  43. [43]

    W ang, C

    S. W ang, C. Ding, Y. Zhang, and X. Zhao , Strong variational sufficiency for nonlinear semidefinite programming and its implications , SIAM Journal on Optimization, 33 (2023), pp. 2988–3011

    work page 2023

  44. [44]

    W ang, On Chebyshev functions and Klee functions , Journal of Mathematical Analysis and Applications, 368 (2010), pp

    X. W ang, On Chebyshev functions and Klee functions , Journal of Mathematical Analysis and Applications, 368 (2010), pp. 293–310

    work page 2010

  45. [45]

    W ang and Z

    X. W ang and Z. W ang , Every proximal mapping is a resolvent of level proximal subd iffer- ential, Optimization Letters, 18 (2024), pp. 1237–1252

    work page 2024

  46. [46]

    W ang, A

    Z. W ang, A. Themelis, H. Ou, and X. W ang , A mirror inertial forward-reflected-backward splitting: convergence analysis beyond convexity and Lipsc hitz smoothness , Journal of Opti- mization Theory and Applications, 203 (2024), pp. 1127–115 9

    work page 2024

  47. [47]

    Wu and J

    Z. Wu and J. J. Ye , Some results on integration of subdifferentials , Nonlinear Analysis: Theory, Methods & Applications, 39 (2000), pp. 955–976

    work page 2000

  48. [48]

    Zalinescu , Convex Analysis in General Vector Spaces , World Scientific Publishing Co., Inc., River Edge, NJ, 2002

    C. Zalinescu , Convex Analysis in General Vector Spaces , World Scientific Publishing Co., Inc., River Edge, NJ, 2002. 35

    work page 2002