§3, PART 1, p. 9 —
ProvedScenarioExact.PartOne.subset_StildeLet and () be convex, let , and suppose that for every number of constraints and every the program (minimise over ) has a unique solution . Fix . Write for the set of multi-extractions whose support constraints are exactly , and for the set where are not violated by the solution generated by . Then
This is the first half of the fact, stated on p. 8 and proved at the end of PART 1 (p. 9), that and coincide up to a probability-zero set. It holds for every multi-extraction; no probability enters. The paper's step "the first constraints are still the support constraints for the problem with constraints" uses Proposition 2.2 (at most support constraints), which is why convexity is assumed.
Formalization Note. The paper leaves implicit that its Assumption 1 holds for every number of constraints; we state existence and uniqueness for every and every sample, and take a solution family θs k ω assumed to solve (so θs k ω is ). Support constraints are the published IsSupportConstraint (removal admits a feasible point of strictly smaller cost), which under uniqueness is the paper's "its removal changes the solution". Closedness, measurability and full support are not needed and not assumed. Indexes are 0-based.
import Mathlib import Definitions.Def_ScenarioApproach_Generalization_violation import Definitions.Def_ScenarioApproach_Generalization_scenarioProgram import Definitions.Def_ScenarioApproach_Generalization_supportConstraint import Definitions.Def_ScenarioExact_PartOne_Basic open MeasureTheory ScenarioApproach.Generalization open scoped ENNReal
namespace ScenarioExact.PartOne
/-- §3, PART 1, p. 9 (claim on p. 8): `S_𝓘̄ ⊆ S̃_𝓘̄`. Deterministic: for every multi-extraction
whose support constraints are exactly the first `d`, the solution generated by the first `d`
constraints violates none of the others. -/
theorem subset_Stilde {d m : ℕ} {Δ : Type*}
(c : EuclideanSpace ℝ (Fin d)) (Θ : Set (EuclideanSpace ℝ (Fin d)))
(Θδ : Δ → Set (EuclideanSpace ℝ (Fin d)))
(hΘ_convex : Convex ℝ Θ) (hΘδ_convex : ∀ δ, Convex ℝ (Θδ δ))
(hexu : ∀ (k : ℕ) (ω : Fin k → Δ), ∃! θ, IsSolution c Θ Θδ ω θ)
(θs : (k : ℕ) → (Fin k → Δ) → EuclideanSpace ℝ (Fin d))
(hθs : ∀ (k : ℕ) (ω : Fin k → Δ), IsSolution c Θ Θδ ω (θs k ω))
(hm : d ≤ m) :
S c Θ Θδ θs (Ibar d m) ⊆ Stilde Θδ θs hm := by sorry
end ScenarioExact.PartOne
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.