Proof of Theorem 5, p. 40 — for a convex cone K, φ ≼_K ψ almost surely implies E[φ] ≼_K E[ψ]
ProvedDRConvexOpt.Lifting.cone_integral_leconic-optimizationexpectationjensen-inequalityp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a probability measure on a measurable space , let be a proper cone, and let be -integrable. If for -almost every , that is, almost surely, then
This conic form of monotonicity of the expectation is the step of the Lifting Theorem's proof that turns the almost-sure constraint into the bound .
Formalization Note The paper says "since is a convex cone"; the statement assumes the full proper-cone hypothesis of the paper, of which only closedness and convexity are relevant.
Preamble
import Mathlib import Definitions.Def_DRConvexOpt_Lifting_Setting
Formal statement
namespace DRConvexOpt.Lifting
open MeasureTheory
/-- Proof of Theorem 5, p. 40: since K is a convex cone, φ ≼_K ψ almost surely implies
E[φ] ≼_K E[ψ] (for a probability measure and integrable φ, ψ). -/
theorem cone_integral_le {Ω : Type*} [MeasurableSpace Ω] (μ : Measure Ω)
[IsProbabilityMeasure μ] {nM : ℕ} (K : Set (Fin nM → ℝ)) (hK : DRConvexOpt.Reform.IsProperCone K)
(φ ψ : Ω → (Fin nM → ℝ)) (hφ : Integrable φ μ) (hψ : Integrable ψ μ)
(h : ∀ᵐ ω ∂μ, ψ ω - φ ω ∈ K) :
∫ ω, ψ ω ∂μ - ∫ ω, φ ω ∂μ ∈ K := by sorry
end DRConvexOpt.Lifting
Source
Wiesemann, Kuhn & Sim, Distributionally Robust Convex Optimization, Optimization Online preprint 3757 (version of September 22, 2013), p. 40, proof of Theorem 5, "Since K is a convex cone, g(z̃) ≼_K ũ P-a.s. implies that E_P[g(z̃)] ≼_K E_P[ũ]"
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.