Lemma 3.1, p. 6 — y ↦ max(0, y) is slantly differentiable on ℝⁿ with slanting function G_m for every δ
ProvedPDASNewton.Local.lemma_3_1max-functionp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1semismooth-newtonslant-differentiability
Fix and let as in (3.2), where for , for and . Then the componentwise maximum from to is slantly differentiable on with slanting function : for every ,
The value of where a coordinate vanishes is arbitrary; the semismooth Newton method of the paper uses .
Formalization Note is Fin n → ℝ with the sup norm; the limit is stated as a little-o relation.
Preamble
import Mathlib import Definitions.Def_PDASNewton_Local_Setting
Formal statement
namespace PDASNewton.Local
/-- Lemma 3.1, p. 6: the map `y ↦ max(0, y)` from `ℝⁿ` to `ℝⁿ` (componentwise) is slantly
differentiable on `ℝⁿ`, and `G_m` of (3.2) is a slanting function for every `δ ∈ ℝⁿ`. -/
theorem lemma_3_1 {n : ℕ} (δ : Fin n → ℝ) :
IsSlantingFunction (fun y : Fin n → ℝ => fun i => max 0 (y i)) (Gm δ) Set.univ := by sorry
end PDASNewton.Local
Source
Hintermüller, Ito, Kunisch, The primal-dual active set strategy as a semismooth Newton method, HAL hal-01660511v1, p. 6, (3.2) and Lemma 3.1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.