pith. sign in

arxiv: 1907.06140 · v1 · pith:IACPZY7Enew · submitted 2019-07-13 · 🧮 math.OC

Bilevel Optimization and Variational Analysis

Boris S. Mordukhovich This is my paper

Pith reviewed 2026-05-24 21:48 UTC · model grok-4.3

classification 🧮 math.OC
keywords bilevel optimizationvariational analysisgeneralized differentiationnecessary optimality conditionsLipschitzian dataoptimistic models
0
0 comments X

The pith

Variational analysis and generalized differentiation derive necessary optimality conditions for bilevel optimization with Lipschitzian data.

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

The paper establishes a self-contained approach using variational analysis to derive necessary optimality conditions in bilevel optimization problems where the data is Lipschitz continuous. This matters because bilevel problems model hierarchical decision-making in which one level optimizes subject to the solution of another optimization problem. The method focuses primarily on optimistic models but the tools extend to pessimistic versions as well. It also identifies and discusses some open problems remaining in the area.

Core claim

This chapter presents a self-contained approach of variational analysis and generalized differentiation to deriving necessary optimality in problems of bilevel optimization with Lipschitzian data. We mainly concentrate on optimistic models, although the developed machinery also applies to pessimistic versions. Some open problems are posed and discussed.

What carries the argument

The self-contained approach of variational analysis and generalized differentiation to derive necessary optimality conditions.

If this is right

  • Necessary optimality conditions follow for optimistic bilevel optimization problems under the Lipschitz assumption.
  • The same variational machinery yields conditions for pessimistic bilevel optimization problems.
  • Open problems in bilevel optimization can be formulated and analyzed within this generalized differentiation framework.

Where Pith is reading between the lines

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

  • The conditions could be specialized to produce explicit stationarity systems for particular classes of bilevel problems that arise in applications.
  • The approach may link to existing variational tools for single-level problems with equilibrium constraints.
  • Numerical schemes that exploit the derived conditions could be tested on benchmark bilevel instances.

Load-bearing premise

The functions and mappings in the bilevel problems satisfy a Lipschitz continuity condition.

What would settle it

A concrete bilevel optimization problem with Lipschitzian data for which no set of necessary optimality conditions follows from the generalized differentiation rules developed in the approach would falsify the claim.

read the original abstract

This chapter presents a self-contained approach of variational analysis and generalized differentiation to deriving necessary optimality in problems of bilevel optimization with Lipschitzian data. We mainly concentrate on optimistic models, although the developed machinery also applies to pessimistic versions. Some open problems are posed and discussed.

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 manuscript presents a self-contained variational-analysis framework, relying on generalized differentiation, for deriving necessary optimality conditions in optimistic bilevel optimization problems whose data functions are assumed Lipschitz continuous; the same machinery is indicated to extend to pessimistic models, and several open problems are posed.

Significance. If the central derivations are valid, the work would supply a unified generalized-differentiation route to first-order conditions for nonsmooth bilevel problems, potentially streamlining existing results that rely on value-function or KKT reformulations and offering a template for handling Lipschitzian data without explicit smoothing.

major comments (2)
  1. [Abstract] Abstract and introductory sections: the claim that Lipschitz continuity of the given data is sufficient for direct application of the coderivative/subdifferential calculus to the optimistic bilevel problem is not automatically justified. Standard sum and chain rules for the value function or argmin mapping require additional metric regularity or calmness of the inner solution map at the reference point; these properties do not follow from Lipschitz continuity alone and must be stated as explicit qualification conditions or derived from the problem data.
  2. [Introduction / main development] The manuscript does not appear to supply a concrete verification (or counter-example) showing that the required qualification conditions hold under the stated Lipschitz hypothesis for a representative class of bilevel problems; without such verification the applicability range of the derived necessary conditions remains unclear.
minor comments (2)
  1. [Abstract] The abstract mentions that the machinery also applies to pessimistic models; a brief indication of the necessary modifications would help readers assess the scope.
  2. [Concluding section] Open problems are listed; numbering them and cross-referencing them to the preceding derivations would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and the constructive comments, which help clarify the scope of the results. We respond to each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract and introductory sections: the claim that Lipschitz continuity of the given data is sufficient for direct application of the coderivative/subdifferential calculus to the optimistic bilevel problem is not automatically justified. Standard sum and chain rules for the value function or argmin mapping require additional metric regularity or calmness of the inner solution map at the reference point; these properties do not follow from Lipschitz continuity alone and must be stated as explicit qualification conditions or derived from the problem data.

    Authors: We agree that Lipschitz continuity of the data alone does not automatically guarantee the metric regularity or calmness of the argmin mapping needed to invoke the full coderivative and subdifferential calculus rules without qualification. The manuscript develops the necessary conditions under the assumption that such qualification conditions hold at the reference point (as is standard in variational analysis), but we acknowledge that this should be stated explicitly rather than left implicit. In the revised manuscript we will add a remark or short subsection listing the precise qualification conditions (e.g., calmness of the solution map) required for the chain and sum rules applied to the value function and argmin mapping. revision: yes

  2. Referee: [Introduction / main development] The manuscript does not appear to supply a concrete verification (or counter-example) showing that the required qualification conditions hold under the stated Lipschitz hypothesis for a representative class of bilevel problems; without such verification the applicability range of the derived necessary conditions remains unclear.

    Authors: We accept the point that an explicit verification for a representative class would strengthen the exposition. While the paper concentrates on the general variational framework, we will insert a brief illustrative example (e.g., a bilevel problem with a linear or strongly convex inner objective) in which calmness of the argmin mapping can be verified directly from the Lipschitz data and the problem structure. This will indicate the range of problems to which the derived conditions apply without additional assumptions. revision: yes

Circularity Check

0 steps flagged

No circularity: self-contained variational derivation on external Lipschitz assumption

full rationale

The paper explicitly frames its contribution as a self-contained application of established generalized differentiation rules from variational analysis to bilevel problems under a Lipschitz continuity hypothesis on the data. No equation or step is exhibited that defines a quantity in terms of itself, renames a fitted input as a prediction, or reduces the central necessary optimality conditions to a self-citation chain whose validity depends on the present result. The Lipschitz assumption is stated as an external input permitting application of prior calculus; any additional qualification conditions required for the rules to hold are not shown to be smuggled in by definition or by the author's prior uniqueness theorems. This satisfies the default expectation of non-circularity for a methods chapter that applies existing tools to a new problem class.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the Lipschitzian data assumption is mentioned but its precise role cannot be audited.

pith-pipeline@v0.9.0 · 5547 in / 1026 out tokens · 29720 ms · 2026-05-24T21:48:11.482527+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages

  1. [1]

    T. Q. Bao, P. Gupta and B. S. Mordukhovich, Necessary cond itions for multiobjective optimization with equilibrium constraints, J. Optim. Theory Appl. 135 (2007), 179–203

    work page 2007

  2. [2]

    Benita, S

    F. Benita, S. Dempe and P. Mehlitz, Bilevel optimal contr ol problems with pure state constraints and finite-dimensional lower level, SIAM J. Optim. 26 (2016), 564–588

    work page 2016

  3. [3]

    F. H. Clarke, Optimization and Nonsmooth Analysis , Wiley-Interscience, New York, 1983

    work page 1983

  4. [4]

    Colson, P

    B. Colson, P. Marcotte and G. Savard, An overview of bilev el optimization, Ann. Oper. Res. 153 (2007), 235–256

    work page 2007

  5. [5]

    Dempe, Foundations of Bilevel Programming, Kluwer, Dordrecht, The Netherlands, 2003

    S. Dempe, Foundations of Bilevel Programming, Kluwer, Dordrecht, The Netherlands, 2003

    work page 2003

  6. [6]

    Dempe and J

    S. Dempe and J. Dutta, Is bilevel programming a special ca se of mathematical programming with complementarity constraints?, Math. Program. 131 (2012), 37–48

    work page 2012

  7. [7]

    Dempe, J

    S. Dempe, J. Dutta and B. S. Mordukhovich, New necessary o ptimality conditions in opti- mistic bilevel programming, Optimization 56 (2007), 577–604

    work page 2007

  8. [8]

    Dempe, V

    S. Dempe, V. Kalashnikov, G. A. P´ erez-Vald´ es and N. Kalashnikova, Bilevel Programming Problems, Springer, New York, 2015

    work page 2015

  9. [9]

    Dempe, B

    S. Dempe, B. S. Mordukhovich and A. B. Zemkoho, Sensitivi ty analysis for two-level value functions with applications to bilevel programming, SIAM J. Optimization 22 (2012), 1309– 1343

    work page 2012

  10. [10]

    Dempe, B

    S. Dempe, B. S. Mordukhovich and A. B. Zemkoho, Necessar y optimality conditions in pessimistic bilevel programming, Optimization 63 (2014), 505–533. 20

    work page 2014

  11. [11]

    Dempe, B

    S. Dempe, B. S. Mordukhovich and A. B. Zemkoho, Two-leve l value function approach to optimistic and pessimistic bilevel programs, Optimization 68 (2019), 433–455

    work page 2019

  12. [12]

    Dempe and A

    S. Dempe and A. B. Zemkoho, The bilevel programming prob lem: reformulations, constraint qualifcations and optimality conditions, Math. Program. 138 (2013), 447–473

    work page 2013

  13. [13]

    Dempe and A

    S. Dempe and A. B. Zemkoho, KKT reformulation and necess ary conditions for optimality in nonsmooth bilevel optimization, SIAM J. Optim. 24 (2014), 1639–1669

    work page 2014

  14. [14]

    N. Dinh, B. S. Mordukhovich and T. T. A. Nghia, Subdifferen tials of value functions and optimality conditions for some classes of DC and bilevel infi nite and semi-infinite programs, Math. Program. 123 (2010), 101–138

    work page 2010

  15. [15]

    Henrion and T

    R. Henrion and T. Surowiec, On calmness conditions in co nvex bilevel programming, Applic. Anal. 90 (2011), 951–970

    work page 2011

  16. [16]

    A. Y. Kruger and B. S. Mordukhovich, Extremal points and the Euler equation in nons- mooth optimization, Dokl. Akad. Nauk BSSR 24 (1980), 684–687

    work page 1980

  17. [17]

    M. B. Lignola and J. Morgan, Inner regularizations and v iscosity solutions for pessimistic bilevel optimization problems, J. Optim. Theory Appl. 173 (2017), 183-202

    work page 2017

  18. [18]

    B. S. Mordukhovich, Maximum principle in problems of ti me optimal control with nons- mooth constraints, J. Appl. Math. Mech. 40 (1976), 960–969

    work page 1976

  19. [19]

    B. S. Mordukhovich, Sensitivity analysis in nonsmooth optimization, in Theoretical Aspects of Industrial Design , edited by D. A. Field and V. Komkov, SIAM Proc. Appl. Math. 58, pp. 32–46, Philadelphia, Pennsylvania, 1992

    work page 1992

  20. [20]

    B. S. Mordukhovich, Complete characterization of open ness, metric regularity, and Lips- chitzian properties of multifunctions, Trans. Amer. Math. Soc. 340 (1993), 1–35

    work page 1993

  21. [21]

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

    work page 2006

  22. [22]

    B. S. Mordukhovich, Variational Analysis and Applications , Springer, Cham, Switzerland, 2018

    work page 2018

  23. [23]

    B. S. Mordukhovich, N. M. Nam and H. M. Phan, Variational analysis of marginal functions with applications to bilevel programming, J. Optim. Theory Appl. 152 (2011), 557–586

    work page 2011

  24. [24]

    B. S. Mordukhovich, N. M. Nam and N. D. Yen, Fr´ echet subd ifferential calculus and opti- mality conditions in nondifferentiable programming, Optimization 55 (2006), 685–396

    work page 2006

  25. [25]

    B. S. Mordukhovich and T. T. A. Nghia, Subdifferentials of nonconvex supremum functions and their applications to semi-infinite and infinite program s with Lipschitzian data, SIAM J. Optim. 23 (2013), 406–431

    work page 2013

  26. [26]

    B. S. Mordukhovich and T. T. A. Nghia, Nonsmooth cone-co nstrained optimization with applications to semi-infinite programming, Math. Oper. Res. 39 (2014), 301–337

    work page 2014

  27. [27]

    J. V. Outrata, On the numerical solution of a class of Sta ckelberg problems, ZOR–Methods Models Oper. Res. 34 (1990), 255–277

    work page 1990

  28. [28]

    P´ erez-Aros, Subdifferential formulae for the supremum of an arbitrary family of functions, SIAM J

    P. P´ erez-Aros, Subdifferential formulae for the supremum of an arbitrary family of functions, SIAM J. Optim. 29 (2019), 1714–1743. 21

    work page 2019

  29. [29]

    S. M. Robinson, Generalized equations and their soluti ons, I: basic theory, Math. Program. Study 10 (1979), 128–141

    work page 1979

  30. [30]

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

    work page 1998

  31. [31]

    J. J. Ye and D. L. Zhu, Optimality conditions for bilevel programming problems, Optimiza- tion 33 (1995), 9–27

    work page 1995

  32. [32]

    J. J. Ye and D. L. Zhu, New necessary optimality conditio ns for bilevel programs by com- bining MPEC and the value function approach, SIAM J. Optim. 20 (2010), 1885–1905

    work page 2010

  33. [33]

    Zemkoho, Solving ill-posed bilevel programs, Set-Valued Var

    A. Zemkoho, Solving ill-posed bilevel programs, Set-Valued Var. Anal. 24 (2016), 423–448

    work page 2016

  34. [34]

    A. J. Zaslavski, Necessary optimality conditions for b ilevel minimization problems, Nonlin- ear Anal. 75 (2012), 1655–1678

    work page 2012

  35. [35]

    Zhang, Multistage bilevel programming problems, Optimization 52 (2003), 605–616

    R. Zhang, Multistage bilevel programming problems, Optimization 52 (2003), 605–616. 22

    work page 2003