Theorem 3.1 — limits of stationary points of Scholtes' relaxation are C-stationary under MPEC-MFCQ
ProvedMPECRelax.ScholtesConv.theorem_3_1Consider the MPEC (1) with continuously differentiable data . Let , let be a stationary point of Scholtes' relaxed program
and let such that MPEC-MFCQ holds at . Then
that is, is feasible for (1) and there are multipliers satisfying the weak-stationarity conditions of Definition 2.3(a) together with for all .
Scholtes' theorem made this conclusion under MPEC-LICQ; the paper weakens the assumption to MPEC-MFCQ, under which the KKT multipliers of the relaxed programs need not converge.
Formalization Note The paper states ""; we state , nonincreasing and . The paper states " is a stationary point of "; we state that is a KKT point of as an NLP, feasibility included (p. 5). The standing assumption of p. 1 is the hypothesis hP. MPEC-MFCQ is Definition 2.4 (MFCQ for TNLP). Feasibility of is not assumed; it is part of the conclusion.
import Mathlib import Definitions.Def_MPECRelax_ScholtesConv_NLP import Definitions.Def_MPECRelax_ScholtesConv_MPEC import Definitions.Def_MPECRelax_ScholtesConv_Scholtes open Filter Topology
namespace MPECRelax.ScholtesConv
theorem theorem_3_1 {n m p l : ℕ} (P : MPEC n m p l) (hP : P.IsC1)
(t : ℕ → ℝ) (ht_pos : ∀ k, 0 < t k) (ht_anti : Antitone t)
(ht_lim : Tendsto t atTop (𝓝 0))
(x : ℕ → E n) (hx : ∀ k, (P.RS (t k)).IsKKTPoint (x k))
(xs : E n) (hxs : Tendsto x atTop (𝓝 xs)) (hCQ : P.MPEC_MFCQ xs) :
P.IsCStationary xs := by sorry
end MPECRelax.ScholtesConv
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.