Lemma 3, PDF p. 11 — on ℐ there is an optimal affine solution ŷ(b) = P̂b + q̂ with all q̂_i equal
ProvedAffinePolicies.TwoGap.lemma_3Let be even and consider the instance of (6). Then there exist a first-stage decision , a matrix and a vector such that together with the affine policy
is an optimal affine solution of (feasible, and with worst-case cost no larger than that of any feasible affine solution), and
The lemma lets the proof of Theorem 2 work with a single scalar for the constant part of the policy.
Formalization Note The lemma asserts that the minimum defining is attained, by a solution with constant intercept. Optimality is stated directly: the solution is feasible, and its worst-case cost is bounded by every bound achieved by any feasible affine solution. The first-stage vector is part of the solution (it plays no role here since and ). The hypothesis excludes the empty instance.
import Mathlib import Definitions.Def_AffinePolicies_TwoGap_Setting
namespace AffinePolicies.TwoGap
theorem lemma_3 (m : ℕ) (hm_even : Even m) (hm0 : 0 < m) :
∃ (x : Fin m → ℝ) (P : Matrix (Fin m) (Fin m) ℝ) (q : Fin m → ℝ),
AffinePolicies.Simplex.IsOptimalAff (A6 m) (B6 m) (c6 m) (d6 m) (U6 m) x P q ∧ ∀ i j, q i = q j := by sorry
end AffinePolicies.TwoGap
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.