Proof of Theorem 7.3, p. 45 — for c = −|y − x| the optimal martingale plans form a convex set
ProvedMartOT.HN.optimizers_convexconvexityhobson-neubergermartingale-optimal-transportp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be probability measures on in convex order and . If and are optimal martingale transport plans for , then so is
This is the "linear structure of the optimization problem" that lets Lemma 5.6 conclude uniqueness in Theorem 7.3. Costs are finite here because and have finite first moments.
Formalization Note The mixture is with non-negative extended-real weights.
Preamble
import Mathlib import Definitions.Def_MartOT_Var_Setting
Formal statement
namespace MartOT.HN
open MeasureTheory
theorem optimizers_convex (μ ν : Measure ℝ) [IsProbabilityMeasure μ]
[IsProbabilityMeasure ν] (hμν : MartOT.Var.ConvexLE μ ν) (π π' : Measure (ℝ × ℝ))
(hπ : MartOT.Var.IsOptimal (fun x y => -|y - x|) μ ν π) (hπ' : MartOT.Var.IsOptimal (fun x y => -|y - x|) μ ν π')
(t : ℝ) (ht : t ∈ Set.Icc (0 : ℝ) 1) :
MartOT.Var.IsOptimal (fun x y => -|y - x|) μ ν (ENNReal.ofReal t • π + ENNReal.ofReal (1 - t) • π') := by sorry
end MartOT.HN
Source
arXiv:1208.1509v2, proof of Theorem 7.3, p. 45 (last paragraph: "the set of solutions is convex")
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.