(8), p. 15 — the objective of problem (7) is linear:
ProvedConstrNestedLogit.Card.eq_8Let be a nest of a nested logit instance in which a customer who chooses the nest always buys (no within-nest no-purchase weight), so that , and let the preference weights of the nest be positive. Let , with . Then for every assortment and every ,
This is the identity that makes problem (7) linear in the assortment, and it turns (7) under cardinality constraints into the unit-weight knapsack problem (9).
Formalization Note The published V includes a within-nest no-purchase weight vnp i, set to here, which is the paper's main model. Positivity of the is not restated on the page but is part of the model (, p. 10); it guarantees for . The case holds through . The identity is stated for every real ; the paper uses it for .
import Mathlib import Definitions.Def_NestedLogitVariants_LP_Model open NestedLogitVariants.LP
namespace ConstrNestedLogit.Card
/-- Equation (8) (p. 15): with no within-nest no-purchase weight (`v_i0 = 0`, so
`V_i(S) = ∑_{j ∈ S} v_ij`) and positive preference weights, the objective of problem (7) is linear
in the assortment: `V_i(S)(R_i(S) − u) = ∑_{j ∈ S} v_ij (r_ij − u)`, including `S = ∅`. -/
theorem eq_8 {ι : Type*} {n : ℕ} (I : Instance ι n) (i : ι)
(hvnp : I.vnp i = 0) (hv : ∀ j, 0 < I.v i j) (S : Finset (Fin n)) (u : ℝ) :
V I i S * (R I i S - u) = ∑ j ∈ S, I.v i j * (I.r i j - u) := by sorry
end ConstrNestedLogit.Card
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.