§2, p. 3 — if the cone of active gradients is ℝⁿ, KKT holds for every objective
ProvedStrictCQ.Scaled.activeCone_univ_kktconstraint-qualificationkktp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let , be a constraint system on and a feasible point. Suppose the cone of active gradients is the whole space:
Then for every function , the point satisfies the KKT conditions for .
This is the first half of the proof that (2.3) is a strict constraint qualification for Scaled-AKKT: when the cone is full, KKT holds independently of the objective function.
Formalization Note No differentiability of is assumed: the statement uses only the vector , which Mathlib's gradient returns for any function.
Preamble
import Mathlib import Definitions.Def_StrictCQ_Scaled_Setting import Definitions.Def_StrictCQ_Scaled_Conditions open Filter Topology open scoped InnerProductSpace
Formal statement
namespace StrictCQ.Scaled
theorem activeCone_univ_kkt {n m p : ℕ} (C : Constraints n m p)
(xs : EuclideanSpace ℝ (Fin n)) (hxs : xs ∈ C.feasible)
(hcone : C.activeCone xs = Set.univ) (f : EuclideanSpace ℝ (Fin n) → ℝ) :
C.IsKKT f xs := by sorry
end StrictCQ.Scaled
Source
Andreani, Martínez, Ramos & Silva, Strict constraint qualifications and sequential optimality conditions for constrained optimization, Optimization Online 5197 (version of November 12, 2015), p. 3, §2 (paragraph after (2.3))
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.