Theorem 1, p. 15 — optimal nondominated offers and monotone largest index
ProvedChoiceRM.Policy.theorem_1Consider a finite set of fare products with a proper choice model, nonnegative fares and a constant arrival probability . List every nondominated offer set exactly once as , in nondecreasing purchase-probability order. Let be the optimal value with periods and units remaining, and its marginal capacity value. At every , a greatest index maximizing exists. Every maximizing index gives an optimal offer set, and
For fixed , the greatest optimal index is nondecreasing with remaining capacity . For fixed , it is nonincreasing with the number of periods remaining. This gives the paper's ordered optimal policy, including the specified rule for ties.
Formalization Note Fin n indexes products, including the empty offer set among subsets. The ordered family is exhaustive and has no repeats. Time counts backward as periods remaining. The assumptions state that arrival is a probability; avoid undefined decisions and marginal values at zero. The fare-order convention is omitted because no result in Section 2 uses it.
import Mathlib import Definitions.Def_RevenueManagement_singleResource import Definitions.Def_ChoiceRM_Policy_Dominance import Definitions.Def_ChoiceRM_Policy_Value
namespace ChoiceRM.Policy
open RevenueManagement
/-- Theorem 1 (p. 15): choose a nondominated set attaining the stage maximum;
the greatest optimal index rises with capacity and falls with time remaining. -/
theorem theorem_1 {n m : ℕ} (lam : ℝ) (hlam0 : 0 ≤ lam) (hlam1 : lam ≤ 1)
(P : Finset (Fin n) → Fin n → ℝ) (hP : IsChoiceModel P)
(r : Fin n → ℝ) (hr : ∀ j, 0 ≤ r j)
(Sq : Fin m → Finset (Fin n)) (hSq : IsOrderedNondominated P r Sq) :
(∀ t x, 1 ≤ t → 1 ≤ x →
∃ k : Fin m,
IsLargestMaximizer P r Sq (deltaV lam P r (t - 1) x) k ∧
(∀ l : Fin m,
(∀ i : Fin m,
expRevenue P r (Sq i) - purchaseProb P (Sq i) * deltaV lam P r (t - 1) x ≤
expRevenue P r (Sq l) - purchaseProb P (Sq l) * deltaV lam P r (t - 1) x) →
IsOptimalOffer lam P r t x (Sq l)) ∧
V lam P r t x =
lam * (expRevenue P r (Sq k) - purchaseProb P (Sq k) *
deltaV lam P r (t - 1) x) + V lam P r (t - 1) x) ∧
(∀ t x x' (hx : 1 ≤ x) (hxx' : x ≤ x') (ht : 1 ≤ t)
(k k' : Fin m),
IsLargestMaximizer P r Sq (deltaV lam P r (t - 1) x) k →
IsLargestMaximizer P r Sq (deltaV lam P r (t - 1) x') k' →
k ≤ k') ∧
(∀ x t t' (hx : 1 ≤ x) (ht : 1 ≤ t) (htt' : t ≤ t')
(k k' : Fin m),
IsLargestMaximizer P r Sq (deltaV lam P r (t - 1) x) k →
IsLargestMaximizer P r Sq (deltaV lam P r (t' - 1) x) k' →
k' ≤ k) := by sorry
end ChoiceRM.Policy
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.