Pith. sign in

REVIEW 3 major objections 4 minor 39 references

On second-order weak sharp minima of general nonconvex set-constrained optimization problems

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper proves second-order necessary and sufficient conditions for quadratic growth toward the solution set in nonconvex set-constrained optimization, replacing the classical support function by a lower generalized support function…

desk verdict Real extension of second-order weak-sharp-minima theory, but Theorem 4.4 overclaims to infeasible points; the fix is simple and the paper deserves conditional acceptance. read the letter →

arxiv 2507.12682 v1 pith:MJD4OZRY submitted 2025-07-16 math.OC

Xiaoxiao Ma , Wei Ouyang , Jane Ye , Binbin Zhang This is my paper
classification math.OC MSC 90C2690C4649J5249J53
keywords second-orderweaksharpminimaquadraticgrowthlowergeneralizedsupportfunctionoutertangentsetasymptoticconemetricsubregularityconstraintqualificationnonconvexset-constrainedoptimizationdirectionalmultiplier
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

The paper asks when a feasible point of min f(x) subject to g(x) in K is a second-order weak sharp minimizer, meaning f grows at least quadratically with the distance to the full solution set S rather than only to the point itself. It proves necessary conditions: under directional metric subregularity, every critical direction d that is an epsilon-proximal normal to S must satisfy a curvature inequality with some M-multiplier, where the classical support function is replaced by a lower generalized support function that is smaller and agrees with the classical one on convex sets. It also proves point-based sufficient conditions using second-order tangent sets defined relative to the solution set. Collectively these extend the classical second-order theory to constraint sets K that are nonconvex and to second-order tangent sets that are nonconvex or empty, and remove the uniform second-order regularity and uniform critical-cone approximation assumptions used in earlier sufficient results.

What carries the argument

The load-bearing object is the lower generalized support function, defined for a closed set S and a subset A by liminf as $\lambda$' approaches $\lambda$ of the infimum of <$\lambda$', u> over u in the preimage of the normal cone of S at $\lambda$' intersected with A. Because it is built from normal-cone preimages, it respects the geometry of nonconvex sets, stays below the classical support function, and reduces to the classical support function when A is convex. It is paired with the outer second-order tangent set $T^{2}$_A(x;d) and the asymptotic second-order tangent cone T''_A(x;d), the latter always nonempty as a cone. The proofs push the quadratic growth inequality through Taylor expansions along arcs x + t d + (1/2) $t^{2}$ w, using epsilon-proximal normals to control distance to S, and use directional metric subregularity to convert implicit feasible-set conditions into explicit multiplier conditions.

What would settle it

Take any feasible arc of the form x + t d + (1/2) $t^{2}$ w lying in Phi with w in $T^{2}$_Phi(x;d) and d in C(x) intersected with the epsilon-proximal normal cone; if for some such arc the scaled increase f(x + t d + (1/2)$t^{2}$ w) - f(x) is smaller than kappa $t^{2}$ (1-2epsilon)^2 ||d||^2 for arbitrarily small t, then the quadratic growth condition (2) fails and the point cannot be a second-order weak sharp minimizer, directly testing the core necessity inequality.

Watch

Extended reading notes

Core claim

The central claim is that the classical second-order characterization of weak sharp minima remains valid in a generalized form even when K and the second-order tangent set $T^{2}$_K are nonconvex and even when $T^{2}$_K is empty. The classical support function is replaced by the lower generalized support function, which is no larger than the classical support function and coincides with it for convex sets. When directional metric subregularity holds at x in direction d, there is a directional Mordukhovich multiplier $\lambda$ such that the lower generalized support function of the asymptotic second-order tangent cone is nonpositive and, if $T^{2}$_K is nonempty, the curvature inequality with the lower generalized support function of $T^{2}$_K holds with the constant 2 kappa (1-2epsilon)^2. On the sufficient side, the paper shows that directional sign conditions on the asymptotic second-order tangent cone plus a strict curvature inequality against the second-order tangent set with respect to the solution set S guarantee the quadratic growth f(x) >= f(xbar) + kappa dist(x,S)^2 locally, with no uniform regularity assumptions.

Load-bearing premise

The sufficient-condition half assumes that the constraint mapping is locally injective on the solution set: whenever g(x_k) approaches g(xbar) with x_k in S, one must have x_k approaching xbar, and if two nearby solutions share the same constraint value the linearized description of the tangent cone with respect to S and the distance estimate can fail.

Editorial extensions

If this is right

  • Necessary conditions hold for nonconvex K and for nonconvex or empty second-order tangent sets, so disjunctive programs, mathematical programs with equilibrium constraints, and cone-complementarity problems now have second-order tests without convexifying the constraint geometry.
  • The epsilon in [0,1/2) interpolates between proximal normals, where the curvature constant is 2 kappa, and looser epsilon-proximal normals, where the constant shrinks to 2 kappa (1-2epsilon)^2, so the theorem states exactly how much curvature certainty is lost when the direction class is widened.
  • The sufficient conditions are point-based and require only a directional multiplier lambda satisfying two local inequalities, not uniform second-order regularity of K or uniform approximation of critical cones.
  • When S is a singleton, the conditions reduce to existing estimates for isolated local minimizers, confirming that the results are genuine extensions of the earlier framework.
  • Under directional nondegeneracy the multiplier is unique and the lower generalized support function can be replaced by the classical support function, giving a clean scalar inequality with the classical support function of T^2_K.

Reading between the lines

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

  • Editorial inference: because the lower generalized support function is defined through normal-cone preimages, the same inequality should specialize to the classical support function not only for convex T^2_K but for any T^2_K that is normal-cone representable, which suggests a testable extension to prox-regular sets.
  • Editorial inference: the assumption that g(x_k) approaching g(xbar) forces x_k to approach xbar is a properness or calmness condition on g over S; replacing it by metric subregularity of the solution map would yield a parameterized version of the sufficient condition that the paper does not state.
  • Editorial inference: the factor (1-2epsilon)^2 quantifies a tradeoff between the size of the admissible normal directions and the strength of the curvature certificate, implying that an algorithm producing only epsilon-proximal normal directions can still certify quadratic growth but only with a tolerance-dependent constant.
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

3 major / 4 minor

Summary. The paper studies local second-order weak sharp minima for the nonconvex set-constrained problem min f(x) subject to g(x) in K, where K is closed but not necessarily convex and the second-order tangent set need not be convex or nonempty. It derives neighborhood necessary conditions (Theorems 3.2, 3.3, 3.8, 3.9) using directional MSCQ, directional multipliers, and the lower generalized support function, and point-based sufficient conditions (Theorem 4.4, Corollary 4.5) using second-order tangent sets and asymptotic second-order tangent cones with respect to the solution set S. The paper claims to remove assumptions such as convexity of K, convexity/nonemptiness of the second-order tangent set, and uniform second-order regularity.

Significance. If the results are correct, they extend the classical second-order weak-sharp-minima theory of Bonnans and Shapiro to a substantially broader nonconvex setting and sharpen the necessary conditions by using directional multipliers and the lower generalized support function. The sufficient conditions are point-based and do not rely on uniform second-order regularity or uniform approximation of critical cones, which is a genuine improvement over earlier results. The paper builds on the authors' own prior work [26] and on Gfrerer-Ye-Zhou [18] for key tools, but the weak-sharp-minima inequalities are new derivations rather than restatements. No numerical experiments are claimed or needed for this theoretical contribution.

major comments (3)
  1. [Theorem 4.4, Eq. (37)] The statement asserts the quadratic growth inequality f(x) - f(xbar) >= kappa [dist(x,S)]^2 for all x in B_delta(xbar), but the proof only treats sequences {x_k} contained in Phi intersect B_{1/k}(xbar). As written the statement is false at infeasible points. For example, take f identically 0, g(x)=x, K=[0,infinity), so Phi=S=[0,infinity) and xbar=0; the hypotheses of Theorem 4.4 are vacuous because T_{Phi,S}(0) intersect N_S(0) = {0}, yet for any infeasible x=-delta/2 the inequality fails. The intended and provable statement is obtained by replacing B_delta(xbar) with Phi intersect B_delta(xbar), and the existing proof does establish this feasible version. This correction is necessary for the theorem to be true.
  2. [Corollary 4.5] The same overclaim appears in the conclusion of Corollary 4.5: the inequality f(x) - f(xbar) >= kappa [dist(x,S)]^2 is stated for all x in B_delta(xbar), while the proof inherits only the feasible version from Theorem 4.4. In addition, condition (ii) contains a typo: the second argument of the second-order tangent set T^2_{K,g_epsilon(S)} is written as g(xbar)d, but it should be grad g(xbar)d, as in condition (i) and in the preceding inclusion (35).
  3. [Section 3, observation before Lemma 3.7] The proof of Theorem 3.8 applies Lemma 3.7 at each x in S near xbar, and Lemma 3.7 requires MSCQ at x in direction d. The paper asserts without proof that if MSCQ holds at xbar then there exists delta' > 0 such that MSCQ holds at all x in S intersect B_{delta'}(xbar). This stability of directional metric subregularity is not immediate from Definition 2.4 and is load-bearing for the necessary conditions in Theorem 3.8. Please provide a proof of this observation or state the stronger assumption that MSCQ holds at every relevant x in direction d; if the observation is not true in general, the proof of Theorem 3.8 needs to be reworked.
minor comments (4)
  1. [Proposition 4.1, proof of (30)] In the proof of the distance estimate, the line 'then d' in T''_{Phi,S}(xbar)' should read 'then d' in T_{Phi,S}(xbar)', since the construction x'_k = uk + tk d' + o(tk) defines a first-order tangent vector with respect to S, not an asymptotic second-order tangent vector.
  2. [Definition 3.6] The definition of the lower generalized support function contains the phrase 'nonenmpty', which appears to be a typo for 'nonempty'.
  3. [Example 3.4] The example writes S = [0,1/2] x {0}, but the feasible set is the closed annulus 1/4 <= (x1-1)^2 + x2^2 <= 1; the verification that all x in S are indeed contained in Phi would be clearer if stated explicitly.
  4. [Section 5] The concluding remark that computing second-order tangent sets with respect to level sets remains challenging is helpful, but the paper could also note explicitly whether the sufficient conditions in Theorem 4.4 can be checked through the inclusions (35)-(36) without computing these level-set tangent sets directly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the second-order weak-sharp-minima inequalities are derived from the defining quadratic-growth inequality and Taylor expansions, not assumed by construction.

full rationale

The paper's derivation chain is not circular. In Theorem 3.2, part (ii) is proved by taking a feasible arc x_k = x + t_k d + (1/2)t_k^2 w_k, applying the defining inequality f(x_k) - f(xbar) >= kappa dist(x_k, S)^2 together with Lemma 3.1's lower bound on dist(x_k, S), and comparing the Taylor expansion with the support-function term; the result is a derived inequality, not a restatement. The explicit multiplier form in Theorem 3.8 uses the same weak-sharp inequality through Theorem 3.2 and imports from [26, Prop. 4.6] and [18] only the calculus of directional multipliers and lower generalized support functions, whose assumptions do not include the weak-sharp-minimizer conclusion being proved. On the sufficient side, Theorem 4.4 assumes the negation of quadratic growth, constructs a violating feasible sequence, and uses second-order Taylor expansions to contradict conditions (i)-(ii); Corollary 4.5 inherits this proof and only uses inclusions (35)-(36) to move to K-tangent sets. The self-citations to [26] and [18] are to prior published theorems with different targets (local optimality, directional multipliers), and under the review rules those count as independent mathematical support. A separate correctness issue exists: Theorem 4.4 and Corollary 4.5 state the inequality for all x in B_delta(xbar), although the proof and Definition (2) support only feasible points; this is a validity defect, not a circularity.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

No fitted parameters appear in this purely analytic paper. The main assumptions are constraint qualifications (MSCQ, DirRCQ, directional nondegeneracy) and two structural conditions on the constraint mapping and the solution set boundary. The solution-set-relative tangent sets are new definitions but are not free parameters or invented physical entities.

assumptions (6)
  • standard math Standard definitions and properties of tangent cones, normal cones, support functions, and limiting normal cones from variational analysis.
    Background results are taken from [2,24,31] and used throughout without proof.
  • domain assumption MSCQ (metric subregularity constraint qualification) holds at xbar in direction d for g(x) in K.
    Used in Theorem 3.8 to convert implicit necessary conditions into explicit multiplier conditions; weaker than Robinson CQ but not universally satisfied.
  • domain assumption DirRCQ (directional Robinson constraint qualification) holds at every x in B_delta(xbar) in direction d.
    Used in Theorem 3.9 to invoke conic linear programming duality with no dual gap, citing [2, Theorem 2.187].
  • domain assumption Directional nondegeneracy condition (26) holds at xbar in direction d.
    Used in Corollary 3.11 to ensure the multiplier set is a singleton and the lower generalized support function coincides with the classical support function.
  • domain assumption Condition (29) in Proposition 4.1: g(x_k) -> g(xbar) implies x_k -> xbar for sequences in S.
    Local injectivity or properness of g over the solution set; needed for equality in the S-relative tangent cone linearization and for distance estimate (30). This is not flagged in the abstract.
  • domain assumption Assumption in Theorem 4.4: for every d in T_Phi,S(xbar) intersect N_S(xbar) \ {0}, one has grad f(x) d = 0 for all x in bd(S) close to xbar.
    Used in the contradiction proof of Theorem 4.4 to remove first-order terms in the Taylor expansion along projection directions onto S.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On second-order weak sharp minima of general nonconvex set-constrained optimization problems." pith.science (2026). https://pith.science/paper/MJD4OZRY

@misc{pith2026250712682,
  author       = {Pith},
  title        = {Pith review of: On second-order weak sharp minima of general nonconvex set-constrained optimization problems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MJD4OZRY}},
  note         = {Machine review of arXiv:2507.12682}
}
read the original abstract

This paper explores local second-order weak sharp minima for a broad class of nonconvex optimization problems. We propose novel second-order optimality conditions formulated through the use of classical and lower generalized support functions. These results are based on asymptotic second-order tangent cones and outer second-order tangent sets. Specifically, our findings eliminate the necessity of assuming convexity in the constraint set and/or the outer second-order tangent set, or the nonemptiness of the outer second-order tangent set. Furthermore, unlike traditional approaches, our sufficient conditions do not rely on strong assumptions such as the uniform second-order regularity of the constraint set and the property of uniform approximation of the critical cones.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [26]

    Ouyang, J.J

    W. Ouyang, J.J. Ye and B. Zhang , New second-order optimality conditions for directional optimality of a general set-constrained optimization problem, SIAM J. Op- tim., 35 (2025), pp. 1274-1299

    work page 2025

  2. [18]

    Gfrerer, J

    H. Gfrerer, J. J. Ye, and J. Zhou, Second-order optimality conditions for noncon- vex set-constrained optimization problems, Math. Oper. Res., 47 (2022), pp. 2344-2365

    work page 2022

  3. [1]

    Bonnans and A.D

    J.F. Bonnans and A.D. Ioffe , Quadratic growth and stability in convex program- ming problems with multiple solutions, J. Convex Anal., 2 (1995), pp. 41-57

    work page 1995

  4. [2]

    J. F. Bonnans and A. Shapiro , Perturbation Analysis of Optimization Problems, Springer, New York, 2000

    work page 2000

  5. [3]

    Burke and S

    J.V. Burke and S. Deng , Weak sharp minima revisited, I: basic theory, Control Cybernet., 31 (2002), pp. 439-469

    work page 2002

  6. [4]

    Burke and S

    J.V. Burke and S. Deng , Weak sharp minima revisited, II: application to linear regularity and error bounds, Math. Program., 104 (2005), pp. 235-261

    work page 2005

  7. [5]

    Burke and S

    J.V. Burke and S. Deng , Weak sharp minima revisited, III: error bounds for differentiable convex inclusions, Math. Program., 116 (2009), pp. 37-56

    work page 2009

  8. [6]

    Burke and M.C

    J.V. Burke and M.C. Ferris , Weak sharp minima in mathematical programming, SIAM J. Control Optim., 31 (1993), pp. 1340-1359

    work page 1993

Show all 39 references
  1. [7]

    Burke and M.C

    J.V. Burke and M.C. Ferris , A Gauss–Newton method for convex composite optimization, Math. Program., 71 (1995), pp. 179-194

    1995

  2. [8]

    Burke and J.J

    J.V. Burke and J.J. Mor ´e, On the identification of active constraints, SIAM J. Numer. Anal., 25 (1988), pp. 1197-1211

    1988

  3. [9]

    Chao, D.F

    D. Chao, D.F. Sun and J. J. Ye, First order optimality conditions for mathematical programs with semidefinite cone complementarity constraints, Math. Program., 147 (2014), pp. 539-579

    2014

  4. [10]

    Deng, Some remarks on finite termination of descent methods, Pac

    S. Deng, Some remarks on finite termination of descent methods, Pac. J. Optim., 1 (2005), pp. 31-37

    2005

  5. [11]

    Durea and R

    M. Durea and R. Strugariu , Necessary optimality conditions for weak sharp minima in set-valued optimization, Nonlinear Anal., 73 (2010), pp, 2148-2157

    2010

  6. [12]

    Ferris, Weak sharp minima and penalty functions in mathematical program- ming, Ph.D

    M.C. Ferris, Weak sharp minima and penalty functions in mathematical program- ming, Ph.D. Dissertation, University of Cambridge, Cambridge, UK, 1988

    1988

  7. [13]

    Ferris, Finite termination of the proximal point algorithm, Math

    M.C. Ferris, Finite termination of the proximal point algorithm, Math. Program., 50 (1991), pp. 359-366. 20

    1991

  8. [14]

    Gfrerer, On directional metric regularity, subregularity and optimality conditions for nonsmooth mathematical programs, Set-Valued Var

    H. Gfrerer, On directional metric regularity, subregularity and optimality conditions for nonsmooth mathematical programs, Set-Valued Var. Anal., 21 (2013), pp. 151-176

    2013

  9. [15]

    Gfrerer , Optimality conditions for disjunctive programs based on generalized differentiation with application to mathematical programs with equilibrium constraints, SIAM J

    H. Gfrerer , Optimality conditions for disjunctive programs based on generalized differentiation with application to mathematical programs with equilibrium constraints, SIAM J. Optim., 24 (2014), pp. 898-931

    2014

  10. [16]

    Gfrerer and D

    H. Gfrerer and D. Klatte, Lipschitz and H¨ older stability of optimization problems and generalized equations, Math. Program., 158 (2016), pp. 35-75

    2016

  11. [17]

    Gfrerer and J

    H. Gfrerer and J. J. Ye , New sharp necessary optimality conditions for mathe- matical programs with equilibrium constraints, Set-Valued Var. Anal., 28 (2020), pp. 395-426

    2020

  12. [19]

    Ginchev and B

    I. Ginchev and B. S. Mordukhovich, On directionally dependent subdifferentials, Mathematics Research Reports, 76 (2010)

    2010

  13. [20]

    Jime´nez and V

    B. Jime´nez and V. Novo, Optimality conditions in differentiable vector optimiza- tion via second-order tangent sets, Appl. Math. Optim., 49 (2004), pp. 123-144

    2004

  14. [21]

    Klatte, On quantitative stability for non-isolated minima, Control Cybernet., 23 (1994), pp

    D. Klatte, On quantitative stability for non-isolated minima, Control Cybernet., 23 (1994), pp. 183-200

    1994

  15. [22]

    C. Li, B.S. Mordukhovich, J. W ang and J.C. Yao , Weak sharp minima on Riemannian manifolds, SIAM J. Optim., 21 (2011), pp. 1523-1560

    2011

  16. [23]

    Luo, X.X

    H.L. Luo, X.X. Huang and J.W. Peng , Generalized weak sharp minima in cone- constrained convex optimization with applications, Comput. Optim. Appl., 53 (2012), pp. 807-821

    2012

  17. [24]

    B. S. Mordukhovich, Variational Analysis and Generalized Differentiation,I: Basic Theory, II: Applications, Springer Berlin, Heidelberg, 2006

    2006

  18. [25]

    Ng and X.Y

    K.F. Ng and X.Y. Zheng , Global weak sharp minima on Banach spaces, SIAM J. Control Optim., 41 (2003), pp. 1868-1885

    2003

  19. [27]

    Penot , Second-order conditions for optimization problems with constraints, SIAM J

    J.-P. Penot , Second-order conditions for optimization problems with constraints, SIAM J. Control Optim., 37 (1998), pp. 303-318

    1998

  20. [28]

    E. D. Rahmo , Characterizations of weak sharp minima for lower- C1 functions, J. Math. Anal. Appl., 397 (2013), pp. 619-627

    2013

  21. [29]

    Robinson, Some continuity properties of polyhedral multifunctions, Math Pro- gram Study, 14 (1981), pp

    S.M. Robinson, Some continuity properties of polyhedral multifunctions, Math Pro- gram Study, 14 (1981), pp. 206-214

    1981

  22. [30]

    R. T. Rockafellar, Convex Analysis, Princeton University Press, Princeton, 1970. 21

    1970

  23. [31]

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

    1998

  24. [32]

    Studniarski and D.E

    M. Studniarski and D.E. W ard, Weak sharp minima: characterizations and suf- ficient conditions, SIAM J. Control Optim., 38 (1999), pp. 219-236

    1999

  25. [33]

    Ye and J.C

    J.J. Ye and J.C. Zhou, First order optimality conditions for mathematical programs with second-order cone complementarity constraints, SIAM J. Optim., 26 (2016), pp. 2820-2846

    2016

  26. [34]

    Ye and J.C

    J.J. Ye and J.C. Zhou , Verifiable sufficient conditions for the error bound property of second-order cone complementarity problems, Math. Program., 171 (2018), pp. 361- 395

    2018

  27. [35]

    Zheng and K.F

    X.Y. Zheng and K.F. Ng , Strong KKT conditions and weak sharp minima in convex-composite optimization, Math. Program., 126 (2009),pp. 259-279

    2009

  28. [36]

    Zheng and X.Q

    X.Y. Zheng and X.Q. Yang , Weak sharp minima for semi-infinite optimization problems with applications, SIAM J. Optim., 18 (2007), pp. 573-588

    2007

  29. [37]

    Zheng and X.Q

    X.Y. Zheng and X.Q. Yang , Global weak sharp minima for convex (semi-)infinite optimization problems, J. Math. Anal. Appl., 348 (2008), pp. 1021-1028

    2008

  30. [38]

    Zhou, B.S

    J. Zhou, B.S. Mordukhovich and N. Xiu , Complete characterizations of local weak sharp minima with applications to semi-infinite optimization and complementar- ity, Nonlinear Anal., 75 (2012), pp. 1700-1718

    2012

  31. [39]

    Zhou and C

    J. Zhou and C. W ang, New characterizations of weak sharp minima , Optim. Lett., 6 (2012), pp. 1773-1785. 22

    2012

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.