p. 41 — E_ℚ[ℓ(Y) | (X, Y) ∈ S] ≤ sup_{y∈𝒴} ℓ(y) whenever ℚ(S) > 0
ProvedCondPortOT.SingUninf.condExp_le_supconditional-expectationdistributionally-robust-optimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a probability measure on , let be measurable with , and let be any function. Then
both sides taken in , where the conditional expectation is .
This is the upper-bound half of Proposition 3.1 (with the fiber ): a conditional law of is supported on , so it cannot average above its supremum. It is stated for a general measurable so that it applies to every fiber .
Formalization Note The integrals are lower Lebesgue integrals of the positive and negative parts; no measurability or integrability of is assumed.
Preamble
import Mathlib import Definitions.Def_CondPortOT_SingUninf_Setting open MeasureTheory
Formal statement
namespace CondPortOT.SingUninf
theorem condExp_le_sup {X Y : Type*} [MeasurableSpace X] [MeasurableSpace Y]
(Q : Measure (X × Y)) [IsProbabilityMeasure Q] (S : Set (X × Y)) (hS : MeasurableSet S)
(hQS : 0 < Q S) (L : Y → ℝ) :
condExp Q S L ≤ ⨆ y, (L y : EReal) := by sorry
end CondPortOT.SingUninf
Source
Nguyen et al., Robustifying conditional portfolio decisions via optimal transport, arXiv:2103.16451v3, p. 41, proof of Proposition 3.1, last display
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.