(4.6) — h(z) ≥ h(0) + h′(0; z) and h′(0; z) ≥ −h′(0; −z)
ProvedDualityStability.Stability.eq_4_6convex-analysisdirectional-derivativep2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a real vector space and a convex function (convex epigraph) with finite. Let be its directional derivative (4.3). Then for every ,
These two elementary facts about directional derivatives of a convex function turn an upper bound on near into a lower bound on near , which is the stability part of Theorem 1.
Formalization Note Both inequalities are in EReal. The paper states (4.6) for "a convex function" and applies it to ; the statement here is the general one for any convex finite at . The right-hand side is a negation, so no undefined form arises; is a sum of a finite number and an extended real.
Preamble
import Mathlib import Definitions.Def_DualityStability_Stability_ConvexFunction import Definitions.Def_DualityStability_Stability_StablySet
Formal statement
namespace DualityStability.Stability
/-- Rockafellar (1967), (4.6), p. 176: two elementary facts about the directional derivatives of a
convex function `h : F → [−∞, +∞]` with `h(0)` finite:
`h(z) ≥ h(0) + h′(0; z)` and `h′(0; z) ≥ −h′(0; −z)` for all `z`. -/
theorem eq_4_6 {F : Type*} [AddCommGroup F] [Module ℝ F] (h : F → EReal) (hconv : ConvexFn h)
(h0_bot : h 0 ≠ ⊥) (h0_top : h 0 ≠ ⊤) :
∀ z : F, h 0 + dirDeriv0 h z ≤ h z ∧ -dirDeriv0 h (-z) ≤ dirDeriv0 h z := by sorry
end DualityStability.Stability
Source
Rockafellar, Duality and Stability in Extremum Problems Involving Convex Functions, Pacific J. Math. 21 (1967), p. 176, (4.6), proof of Theorem 1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.