Lemma 1, p. 23 — if γ is continuous on compact X × Ξ, then c(x,ℙ) is continuous on X × 𝒫
ProvedDROOptimal.Continuous.lemma_1data-driven-optimizationdistributionally-robust-optimizationlarge-deviationsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1relative-entropy
Throughout, and are compact, is jointly continuous (the standing assumptions of §2, p. 5, and §5, p. 23), and is the set of Borel probability distributions on with the topology of weak convergence.
Lemma 1 (Continuity of model-based predictors). The model-based predictor
is continuous on , where carries the Euclidean topology and the weak topology.
Thus belongs to the class of data-driven predictors, and model-based prescriptors exist. The lemma is used in the continuity proof for (Proposition 6).
Preamble
import Mathlib import Definitions.Def_DROOptimal_Continuous_Setting
Formal statement
namespace DROOptimal.Continuous
/-- Lemma 1 (p. 23): if `γ` is continuous on the compact set `X × Ξ`, then `c(x, ℙ)` is continuous
on `X × 𝒫` (weak topology on 𝒫). -/
theorem lemma_1 {d n : ℕ} {Ξ : Set (EuclideanSpace ℝ (Fin d))} {X : Set (EuclideanSpace ℝ (Fin n))}
(hX : IsCompact X) (hΞ : IsCompact Ξ) (γ : ↥X → ↥Ξ → ℝ)
(hγ : Continuous (fun p : ↥X × ↥Ξ => γ p.1 p.2)) :
IsPredictor (cost γ) := by sorry
end DROOptimal.Continuous
Source
Van Parys, Mohajerin Esfahani & Kuhn, From Data to Decisions: Distributionally Robust Optimization is Optimal, arXiv:1704.04118v3, p. 23, Lemma 1; proof p. 29
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.