Proposition 1(b), p. 3 — F⁰(x;d) = max [⟨g,d⟩ : g ∈ ∂F(x)]
ProvedMifflinSemismooth.Extremal.prop1_bgeneralized-gradientnonsmooth-analysisp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1support-function
Let be open and let be Lipschitz on with constant . Let . Then for every direction the generalized directional derivative is the largest value of over the generalized gradient, and this largest value is attained:
So is the support function of . In the proof of Theorem 2 this identity is applied to the extremal function .
Formalization Note. "max" is IsGreatest of the image of under , asserting both the bound and its attainment. The Lipschitz hypothesis also ensures that the real limsup defining is not a default value.
Preamble
import Mathlib import Definitions.Def_ClarkeGradients_Shared_genDirDeriv import Definitions.Def_MifflinSemismooth_Extremal_Basic open Filter Topology
Formal statement
namespace MifflinSemismooth.Extremal
/-- Mifflin (1976), Proposition 1(b), p. 3: for `F` Lipschitz on the open set `B` and `x ∈ B`,
`F⁰(x; d) = max [⟨g, d⟩ : g ∈ ∂F(x)]` for every `d`. -/
theorem prop1_b {n : ℕ} (F : EuclideanSpace ℝ (Fin n) → ℝ) (B : Set (EuclideanSpace ℝ (Fin n)))
(hB : IsOpen B) (K : NNReal) (hFB : LipschitzOnWith K F B) (x : EuclideanSpace ℝ (Fin n))
(hx : x ∈ B) :
∀ d, IsGreatest ((fun g => inner ℝ g d) '' genGrad F x)
(ClarkeGradients.Shared.genDirDeriv F x d) := by sorry
end MifflinSemismooth.Extremal
Source
Mifflin, Semismooth and semiconvex functions in constrained optimization, IIASA Research Report RR-76-21 (December 1976), p. 3, Proposition 1(b)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.