Lemma 12 (first claim), p. 45 — for p ≥ p_k the robust risk of a two-point Z is its larger value z₁
ProvedUniformDRO.LowerBound.lemma_12_acressie-readdistributionally-robust-optimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let , , and . Let , , and let take the value with probability and with probability . If , then
So once the larger value has mass at least , the Cressie–Read ball of radius contains the point mass at , and the worst case is the largest value of .
Formalization Note. is the robust risk of the two-point law in the likelihood-ratio form (3) of the mission's definition file.
Preamble
import Mathlib import Definitions.Def_UniformDRO_LowerBound_Setting
Formal statement
namespace UniformDRO.LowerBound
theorem lemma_12_a (k ρ : ℝ) (hk : 1 < k) (hρ : 0 < ρ) (z₀ z₁ p : ℝ) (hz : z₀ ≤ z₁)
(hp : p ∈ Set.Icc (0 : ℝ) 1) (hpk : pk k ρ ≤ p) :
UniformDRO.Concentration.robustRisk k ρ (twoPointLaw z₀ z₁ p) id = z₁ := by sorry
end UniformDRO.LowerBound
Source
Duchi & Namkoong, arXiv:1810.08750v6, Lemma 12 (first claim), p. 45; proof App. D.1.1, pp. 46–47
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.