Proof of Theorem 3.1, p. 7 — a type contribution is bounded by the MMNL optimum
ProvedCustAssort.Value.thm31_type_boundassortment-optimizationcustomizationmultinomial-logitp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Fix positive product revenues, nonnegative preference weights, nonnegative type arrival weights, and a cardinality budget . For any customer type and any carried set with , its personalized expected contribution satisfies
The type-optimal subassortment is feasible as a common MMNL offer, whose other type contributions are nonnegative. This is the typewise comparison used for the upper bound in Theorem 3.1.
Formalization Note The source applies this comparison to a CAP-optimal ; the statement records the same comparison for every feasible . Normalization of is not needed for this intermediate inequality.
Preamble
import Mathlib import Definitions.Def_CustAssort_Value_Setting
Formal statement
namespace CustAssort.Value
/-- The typewise upper bound in the proof of Theorem 3.1, page 7. -/
theorem thm31_type_bound {n m : ℕ} (K : ℕ) (r : Fin n → ℝ)
(v : Fin n → Fin m → ℝ) (θ : Fin m → ℝ)
(hr : ∀ i, 0 < r i) (hv : ∀ i j, 0 ≤ v i j) (hθ : ∀ j, 0 ≤ θ j) :
∀ (q : Fin m) (S : Finset (Fin n)), S.card ≤ K →
θ q * CustAssort.AugGreedy.fj v r q S ≤ zMMNL θ v r K := by sorry
end CustAssort.Value
Source
El Housni & Topaloglu, Joint Assortment Optimization and Customization under a Mixture of Multinomial Logit Models: Value of Personalized Assortments, SSRN 3830082 (version of December 7, 2021), p. 7, proof of Theorem 3.1, displayed type-q comparison
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.