Proposition 3, p. 16 ā for r ā„ 0 the distributionally robust predictor Ä_r is continuous on X Ć š«
ProvedDROOptimal.Predictor.proposition_3continuitydistributionally-robust-optimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1relative-entropy
Let be compact, , continuous in for each , and let
be the distributionally robust predictor (10). If , then is (jointly) continuous on ; that is, belongs to the class of data-driven predictors.
Continuity is half of feasibility in (5) and is needed for to be a competitor at all.
Formalization Note Continuity is with respect to the product of the subspace topologies on and on the simplex .
Preamble
import Mathlib import Definitions.Def_DROOptimal_Predictor_Setting
Formal statement
namespace DROOptimal.Predictor
/-- Proposition 3 (Continuity of Ä_r), p. 16: under the standing assumptions of §2 (p. 5), if r ā„ 0
then the distributionally robust predictor Ä_r is continuous on X Ć š«. -/
theorem proposition_3 {n d : ā} (X : Set (EuclideanSpace ā (Fin n))) (hX : IsCompact X)
(γ : ā„X ā Fin d ā ā) (hγ : ā i, Continuous (fun x : ā„X => γ x i))
(r : ā) (hr : 0 ⤠r) :
IsPredictor (drPredictor γ r) := by sorry
end DROOptimal.Predictor
Source
Van Parys, Mohajerin Esfahani & Kuhn, From Data to Decisions: Distributionally Robust Optimization is Optimal, arXiv:1704.04118v3, p. 16, Proposition 3 (and Remark 2, p. 15, for r = 0)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.