Lemma 1 — cluster points of difference quotients come from generalized gradients
ProvedMifflinSemismooth.Composition.lemma1generalized-gradientnonsmooth-analysisp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be Lipschitz on an open set containing . For a direction , positive numbers , and shifts , every accumulation point of the difference quotients has the form
The lemma connects directional limiting values to the generalized gradient at the base point.
Formalization Note The accumulation point is a MapClusterPt; no existence of a full quotient limit is assumed. The paper uses for positive convergence to zero.
Preamble
import Mathlib import Definitions.Def_MifflinSemismooth_Extremal_Basic open Filter Topology
Formal statement
namespace MifflinSemismooth.Composition
theorem lemma1 {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 : EuclideanSpace ℝ (Fin n)) (t : ℕ → ℝ)
(htpos : ∀ k, 0 < t k) (htlim : Tendsto t atTop (𝓝 0))
(h : ℕ → EuclideanSpace ℝ (Fin n))
(hhlim : Tendsto h atTop (𝓝 0)) (Fstar : ℝ)
(hcluster : MapClusterPt Fstar atTop
(fun k => (F (x + h k + t k • d) - F (x + h k)) / t k)) :
∃ g ∈ MifflinSemismooth.Extremal.genGrad F x, Fstar = inner ℝ g d := by sorry
end MifflinSemismooth.Composition
Source
Mifflin, Semismooth and semiconvex functions in constrained optimization, IIASA Research Report RR-76-21 (December 1976), p. 4, Lemma 1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.