Theorem 2.6, p. 12 (necessity) — a martingale transport plan in Π_M(µ, ν) forces µ ⪯C ν (Jensen)
ProvedMartOT.Abs.theorem_2_6_necessityconvex-orderjensenmartingale-optimal-transport
Let , i.e. finite measures on with finite first moment. If there exists a martingale transport plan , then : for every convex ,
This is the necessity half of Theorem 2.6 of the paper, which it derives from Jensen's inequality: .
Formalization Note is IsMartingalePlan (marginals and for all bounded Borel , the paper's (4)); is ConvexLE (Definition 2.1), whose integrals are computed as in EReal. The measures need not be probability measures, as in the paper's class .
Preamble
import Mathlib import Definitions.Def_MartOT_Var_Setting
Formal statement
namespace MartOT.Abs
open MeasureTheory
theorem theorem_2_6_necessity (μ ν : Measure ℝ) (hμ : MartOT.Var.InM μ) (hν : MartOT.Var.InM ν)
(π : Measure (ℝ × ℝ)) (hπ : MartOT.Var.IsMartingalePlan μ ν π) :
MartOT.Var.ConvexLE μ ν := by sorry
end MartOT.Abs
Source
arXiv:1208.1509v2 (Beiglböck–Juillet, On a problem of optimal transport under marginal martingale constraints), Theorem 2.6, p. 12 — necessity, proved on p. 12 by Jensen's inequality