Proof of Theorem 4, p. 13 — for 0 ≤ ξ ≤ v′₋(a), v(t) − ξt has a maximizer in [a, b], so v^*(ξ) is finite
ProvedNonlinSSD.DualFunctional.exists_maximizer_mem_Iccconvex-conjugatep2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-dominance
Let , and let be a real number with
Then the function attains its maximum over at some point :
and consequently the concave conjugate is finite, .
This is the pointwise fact behind the finite case of Theorem 4: on the domain , the inner maximization defining can be restricted to the compact interval .
Formalization Note The paper states this for the random variable almost surely; here it is stated for a deterministic , which is the content of "the function has a maximizer in " at each . is EReal-valued, and the conclusion gives its value as a real number.
Preamble
import Mathlib import Definitions.Def_NonlinSSD_DualFunctional_Basic open MeasureTheory
Formal statement
namespace NonlinSSD.DualFunctional
theorem exists_maximizer_mem_Icc (a b : ℝ) (hab : a ≤ b) (v : ℝ → ℝ) (hv : v ∈ NonlinSSD.Optimality.U1 a b)
(ξ : ℝ) (hξ0 : 0 ≤ ξ) (hξc : ξ ≤ leftDeriv v a) :
∃ t₀ ∈ Set.Icc a b, (∀ t : ℝ, v t - ξ * t ≤ v t₀ - ξ * t₀) ∧
concaveConj v ξ = ((ξ * t₀ - v t₀ : ℝ) : EReal) := by sorry
end NonlinSSD.DualFunctional
Source
Dentcheva, Ruszczyński, Optimality and duality theory for stochastic optimization problems with nonlinear dominance constraints, author manuscript (rev. April 2003; Math. Program. 2004, DOI 10.1007/s10107-003-0453-z), p. 13, proof of Theorem 4
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.