§4.1, p. 14 — for r ≥ 0 the supremum in (10) defining ĉ_r(x,ℙ′) is attained
ProvedDROOptimal.Predictor.eq_10_attaineddistributionally-robust-optimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1relative-entropy
Let be compact, , continuous in for each , , and
the distributionally robust predictor (10). If , then for every and there is with
The proof of Theorem 4 starts from such a maximizer (display (14)).
Formalization Note The supremum is a real sSup; the statement asserts that it equals a value of at a feasible model, i.e. that it is a maximum. The standing assumptions of §2 (compact , continuous ) are carried although only the fixed decision matters here.
Preamble
import Mathlib import Definitions.Def_DROOptimal_Predictor_Setting
Formal statement
namespace DROOptimal.Predictor
/-- §4.1, p. 14, after Definition 6: for r ≥ 0 the supremum in (10) is attained, i.e. for every
x ∈ X and ℙ′ ∈ 𝒫 there is ℙ₀ ∈ 𝒫 with I(ℙ′, ℙ₀) ≤ r and ĉ_r(x, ℙ′) = c(x, ℙ₀). -/
theorem eq_10_attained {n d : ℕ} (X : Set (EuclideanSpace ℝ (Fin n))) (hX : IsCompact X)
(γ : ↥X → Fin d → ℝ) (hγ : ∀ i, Continuous (fun x : ↥X => γ x i))
(r : ℝ) (hr : 0 ≤ r) (x : ↥X) (ℙ' : Δ d) :
∃ ℙ₀ : Δ d, relEntropy ℙ' ℙ₀ ≤ (r : EReal) ∧ drPredictor γ r x ℙ' = cost γ x ℙ₀ := by sorry
end DROOptimal.Predictor
Source
Van Parys, Mohajerin Esfahani & Kuhn, From Data to Decisions: Distributionally Robust Optimization is Optimal, arXiv:1704.04118v3, p. 14, §4.1, after Definition 6 (unnumbered claim on (10))
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.