Lemma 2(b) — Interval assignment for arbitrary prediction values
ProvedFBMSP.lemma2b_general_menu_intervalsFix a nonempty, nondecreasing list of nonnegative prices and an integrable mean-zero MHR error law. There exists a nondecreasing assignment of every real prediction value to a menu index that maximizes the pointwise expected revenue. Each nonempty assignment fiber is an interval: if two ordered endpoints receive the same index, all prediction values between them receive that index too.
This fixed-menu statement retains arbitrary prediction values and therefore supports both atomic and continuous prediction laws. It asserts existence of a suitable interval assignment; ties need not make every optimal assignment interval.
import Definitions.Def_FBMSP_Model
namespace FBMSP
theorem lemma2b_general_menu_intervals {k : ℕ}
(hk : 0 < k) (η : MeasureTheory.Measure ℝ)
(hη : ValidNoise η) (hMHR : HasMHR η)
(p : Fin k → ℝ) (hp : Monotone p) (hpNonneg : ∀ j, 0 ≤ p j) :
∃ a : ℝ → Fin k, Monotone a ∧
(∀ (x : ℝ) (j : Fin k), postedRevenue η x (p j) ≤
postedRevenue η x (p (a x))) ∧
(∀ x₁ x₂ x₃ : ℝ, x₁ ≤ x₂ → x₂ ≤ x₃ →
a x₁ = a x₃ → a x₂ = a x₁) := by sorry
end FBMSPRead-back
What the Lean code literally says, in plain math · gpt-6-astra
For every natural , let be a probability measure on with integrable identity function and , and write and . Assume that is absolutely continuous with respect to Lebesgue measure and that is convex and is concave on , meaning for and . For any nondecreasing nonnegative real price vector , there exists a function such that whenever , for every real and every , and for all real , equality implies . Thus the selected slot maximizes posted revenue within the supplied vector at every real argument, and the set assigned to any fixed slot is an interval in the order-convex sense, possibly empty. The arguments range over all of , including negative values; there is no prediction measure or almost-everywhere restriction. Duplicate prices, zero prices, ties, and zero sale probabilities are permitted, and the theorem does not require uniqueness of the selection or prescribe the inclusion of interval endpoints. The case is excluded.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.