Regularity and Coercive Decay of the Tilted Objective
ProvedFunctionalInequalities.Analysis.tiltedObjective_data_readableThis is a technical attainment lemma for a Legendre transform.
Let be a finite-dimensional real inner-product normed space, let be a ProperPotential, and assume that is lower semicontinuous. If lies in the interior of , define . Then:
- is upper semicontinuous;
- for every ;
- for at least one ; and
- there exist with such that, for every ,
The conclusion supplies the hypotheses for a separate upper-semicontinuous maximum theorem; it does not itself assert existence of a maximizer.
Original Lean declaration header
Within namespace FunctionalInequalities.Analysis. Shown in its original module context. Proof omitted.
theorem tiltedObjective_data
(φ : E → EReal) (hφ : ProperPotential φ)
(hlsc : LowerSemicontinuous φ) (y : E)
(hy : y ∈ interior {q : E | legendre φ q ≠ ⊤}) :
let G : E → EReal := fun z => (inner ℝ z y : EReal) - φ z
UpperSemicontinuous G ∧
(∀ z, G z ≠ ⊤) ∧
(∃ z, G z ≠ ⊥) ∧
∃ (a b : ℝ), 0 < a ∧ ∀ z,
G z ≤ ((b - a * ‖z‖ : ℝ) : EReal)
Module navigation
Module: Analysis/Convex/LegendreVariation/Attainment.
Definition modules: Analysis/Convex/Legendre.
Supporting theorems: exists_coercive_tilt_of_mem_interior_legendre_domain.
Previous: exists_coercive_tilt_of_mem_interior_legendre_domain · Next: exists_legendre_maximizer_with_gradient_identity.
Source excerpt, lines 280–289 · Frozen source 408915f.
import Definitions.Def_KLSModular_FunctionalInequalities_Analysis_Convex_LegendreVariation_Attainment
variable {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℝ E]
[FiniteDimensional ℝ E]
theorem FunctionalInequalities.Analysis.tiltedObjective_data_readable
(φ : E → EReal) (hφ : ProperPotential φ)
(hlsc : LowerSemicontinuous φ) (y : E)
(hy : y ∈ interior {q : E | legendre φ q ≠ ⊤}) :
let G : E → EReal := fun z => (inner ℝ z y : EReal) - φ z
UpperSemicontinuous G ∧
(∀ z, G z ≠ ⊤) ∧
(∃ z, G z ≠ ⊥) ∧
∃ (a b : ℝ), 0 < a ∧ ∀ z,
G z ≤ ((b - a * ‖z‖ : ℝ) : EReal) := by sorry