Proposition 5.2 — the absolute-deviation evaluation function is convex
ProvedKallMayerSLP.RiskMeasures.proposition_5_2convex-analysisp2o-gran-per-chapterp2o-plan-bookp2o-v1risk-measuresstochastic-programming
Let be an -dimensional random vector on a probability space whose expected value exists. Then the evaluation function
is convex on .
measures the absolute deviation between and ; its convexity makes the models (5.16) and (5.17) convex programs.
Formalization Note "The expected value exists" is integrability of every component of and of .
Differences from the source:
- encoding: A(x) := 𝔼[|ηᵀx − ξ|] → rhoA P (fun ω => η ω ⬝ᵥ x − ξ ω), with rhoA P ϑ = ∫ |ϑ| ∂P; reason: the same function
- encoding: x ∈ ℝⁿ, ηᵀx → x : Fin n → ℝ, η ω ⬝ᵥ x; reason: Lean convention; same instances
Preamble
import Mathlib import Definitions.Def_KallMayerSLP_RiskMeasures_rhoA open MeasureTheory open scoped ENNReal
Formal statement
namespace KallMayerSLP.RiskMeasures
/-- **Proposition 5.2** (convexity of the absolute-deviation evaluation function).
Kall & Mayer, *Stochastic Linear Programming* (2005), Chapter 2, §5.2, (5.14)–(5.15) and Prop. 5.2, pp.169–170 (PDF 179–180), with the standing assumption
of §5 (p.166, PDF 176) that the expected value of `(ηᵀ, ξ)` exists. Then
`A(x) := 𝔼[|ηᵀx − ξ|]` is a convex function of `x ∈ ℝⁿ`.
**Formalization Note.** `A(x) = ρ_A(ηᵀx − ξ)` with `ρ_A(ϑ) = 𝔼|ϑ|`. Vectors are `Fin n → ℝ`,
`ηᵀx` is `η ω ⬝ᵥ x`; "the expected value exists" is integrability of each component of `η` and of
`ξ`. -/
theorem proposition_5_2 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
{n : ℕ} (η : Ω → Fin n → ℝ) (ξ : Ω → ℝ)
(hη : ∀ i, Integrable (fun ω => η ω i) P) (hξ : Integrable ξ P) :
ConvexOn ℝ Set.univ (fun x : Fin n → ℝ => rhoA P (fun ω => η ω ⬝ᵥ x - ξ ω)) := by sorry
end KallMayerSLP.RiskMeasures
Source
Kall & Mayer, Stochastic Linear Programming (2005), Chapter 2, §5.2, (5.14)–(5.15), p.169 (PDF 179), and Proposition 5.2, p.170 (PDF 180)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.