Lemma A3 — everywhere finite and differentiable monotone conjugate
ProvedBregmanPPA.IneqMult.lemmaA3convex-conjugatemonotone-conjugatep2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be closed, proper, and essentially strictly convex, finite on the strictly positive orthant. If the image of its subdifferential contains that orthant, then its monotone conjugate has full domain and is differentiable everywhere:
This result is the appendix bridge from general convex analysis to Lemma 3's Bregman-function case.
Formalization Note Essential strict convexity is strict convexity on every convex subset of . Full domain states that every extended-real value is finite before differentiability is asserted of its real form.
Preamble
import Mathlib import Definitions.Def_BregmanPPA_IneqMult_Program open Filter Topology InertialFB.IFB
Formal statement
namespace BregmanPPA.IneqMult
/-- Lemma A3, pp. 223–224: full domain and differentiability of the monotone conjugate. -/
theorem lemmaA3 {m : ℕ} (F : E m → EReal)
(hproper : IsProperFn F) (hconvex : IsConvexFn F)
(hclosed : LowerSemicontinuous F)
(hstrict : EssentiallyStrictlyConvex F)
(hfinite : ∀ p : E m, p ∈ posOrthant m → F p ≠ ⊤)
(him : ∀ z : E m, z ∈ posOrthant m →
∃ p : E m, IsSubgradient F p z) :
(∀ z : E m, monoConjE F z ≠ ⊤ ∧ monoConjE F z ≠ ⊥) ∧
Differentiable ℝ (fun z : E m => (monoConjE F z).toReal) := by sorry
end BregmanPPA.IneqMult
Source
Eckstein, Nonlinear proximal point algorithms using Bregman functions, with applications to convex programming, Math. Oper. Res. 18(1) (1993), pp. 223–224, Lemma A3, https://doi.org/10.1287/moor.18.1.202
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.