Remark 2.2 — MFCQ iff the active inequality and all equality gradients are positive-linearly independent
ProvedMPECRelax.ScholtesConv.remark_2_2constraint-qualificationsp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1theorems-of-the-alternative
Let be a feasible point of a nonlinear program (2) with finitely many inequality constraints and equality constraints , and let be its active set. Then
This dual characterisation of MFCQ is the form used throughout the convergence analysis of relaxation methods: MPEC-MFCQ is applied as positive-linear independence of a family of gradients.
Formalization Note The statement concerns only the gradient values at ; no differentiability is assumed (for non-differentiable data, Mathlib's gradient is , and the equivalence is still a statement about those vectors). Positive-linear independence is the negation of positive-linear dependence (Definition 2.1) with and all equality indices.
Preamble
import Mathlib import Definitions.Def_MPECRelax_ScholtesConv_NLP
Formal statement
namespace MPECRelax.ScholtesConv
theorem remark_2_2 {n : ℕ} {ι κ : Type} [Fintype ι] [Fintype κ] (P : NLP n ι κ) (x : E n)
(hx : P.Feasible x) :
P.IsMFCQ x ↔ ¬ P.PosLinDep x (P.activeSet x) Set.univ := by sorry
end MPECRelax.ScholtesConv
Source
Hoheisel, Kanzow, Schwartz, Theoretical and numerical comparison of relaxation methods for mathematical programs with complementarity constraints, Preprint 299, Univ. Würzburg, Sept. 2010, p. 4, Remark 2.2
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.