Proof of Theorem 6.1, (15), p. 37 — d(t) = λh(y⁺ − t) + (1 − λ)h(y⁻ − t) − h(y′ − t) is strictly decreasing when h′ is strictly convex
ProvedMartOT.Opt.d_strictAnticonvexitymartingale-optimal-transportp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1real-analysis
Let be differentiable with strictly convex derivative . Let and with . Then the function
is strictly decreasing on .
For the cost this is the function (15), ; its monotonicity is the analytic heart of the optimality of the left-curtain coupling.
Preamble
import Mathlib import Definitions.Def_MartOT_Var_Setting
Formal statement
namespace MartOT.Opt
open MeasureTheory
theorem d_strictAnti (h : ℝ → ℝ) (hd : Differentiable ℝ h)
(hconv : StrictConvexOn ℝ Set.univ (deriv h)) (ym yp y' l : ℝ)
(hym : ym < y') (hyp : y' < yp) (hl0 : 0 < l) (hl1 : l < 1)
(hcomb : l * yp + (1 - l) * ym = y') :
StrictAnti (fun t : ℝ => l * h (yp - t) + (1 - l) * h (ym - t) - h (y' - t)) := by sorry
end MartOT.Opt
Source
arXiv:1208.1509v2, proof of Theorem 6.1, display (15) and the lines after it, p. 37
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.