Theorem 3.9 (i), p. 21 — a.s. every cluster point of estimated optimal solutions minimizes Ef
ProvedDupacovaWets.Consistency.clusterPt_mem_argminThe standing data are those of §3: is a Polish space with its Borel -field, a probability measure on it, a probability space with an increasing sequence of sub--fields , random probability measures on , and an integrand with domain ; , and the measures satisfy Assumptions 3.4 and 3.5. Write and , with the convention of p. 10.
Under Assumptions 3.4 and 3.5 there is with such that for every : if for , then every cluster point of satisfies
i.e. it is an optimal solution of the true problem.
This is the solution part of the consistency theorem, valid without assuming that has a unique minimizer.
Formalization Note Lean writes as EuclideanSpace ℝ (Fin n), and the sample index is shifted: Lean's Pν k ζ and 𝔽 k are the paper's and , while is kept as the separate argument P. A cluster point is a point such that every neighbourhood of contains for infinitely many (MapClusterPt). On the page (i) and (ii) share one set ; as separate items each has its own, and intersecting the two recovers a common one. The printed "" is read as . The paper also says that is the support of ; this hypothesis is not used in the proofs of §3 and is omitted, which makes the statement stronger.
import Mathlib import Definitions.Def_DupacovaWets_Consistency_expect import Definitions.Def_DupacovaWets_Consistency_ExtendedFunctions import Definitions.Def_DupacovaWets_Consistency_Assumptions open MeasureTheory Filter Topology
namespace DupacovaWets.Consistency
/-- Dupačová–Wets (WP-86-41), p. 21, Theorem 3.9 (i): there is `Z₀ ∈ F` with `μ(Z \ Z₀) = 0`
such that for every `ζ ∈ Z₀`, every cluster point of every sequence `x^ν ∈ argmin E^ν f(·, ζ)`
belongs to `argmin Ef`. (`Pν k` is the paper's `P^{k+1}`.) -/
theorem clusterPt_mem_argmin {n : ℕ} {Ξ : Type*} [TopologicalSpace Ξ] [PolishSpace Ξ] [MeasurableSpace Ξ] [BorelSpace Ξ]
(P : Measure Ξ) [IsProbabilityMeasure P]
{Z : Type*} [mZ : MeasurableSpace Z] (μ : Measure Z) [IsProbabilityMeasure μ]
(𝔽 : Filtration ℕ mZ) (Pν : ℕ → Z → Measure Ξ)
(f : EuclideanSpace ℝ (Fin n) → Ξ → EReal) (S : Set (EuclideanSpace ℝ (Fin n)))
(h34 : Assumption3_4 f S) (h35 : Assumption3_5 f S P μ 𝔽 Pν) :
∃ Z₀ : Set Z, MeasurableSet Z₀ ∧ μ Z₀ᶜ = 0 ∧ ∀ ζ ∈ Z₀,
∀ xs : ℕ → EuclideanSpace ℝ (Fin n), (∀ k, xs k ∈ argminSet (expectFn (Pν k ζ) f)) →
∀ xhat : EuclideanSpace ℝ (Fin n), MapClusterPt xhat atTop xs →
xhat ∈ argminSet (expectFn P f) := by sorry
end DupacovaWets.Consistency
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.