Theorem 1 — some assortment in is optimal for (RO)
ProvedRankRobustAssort.exists_optimal_mem_SHatassortment-optimizationchoice-modelsp2o-gran-per-chapterp2o-plan-paperp2o-v1ranking-based-choicerobust-optimization
Consider a firm with products (0 the no-purchase option) and revenues , past assortments with normalized purchase frequencies , a radius and an or error norm, and suppose the set of ranking-based choice models consistent with the data is nonempty. Then there exists an assortment
that is, an assortment revenue-ordered relative to the past assortments that is optimal for the robust problem (RO) .
The theorem restricts the search for an optimal robust assortment from all assortments to the collection , whose size depends on the structure of the past assortments; the algorithms and reformulations of the paper are built on it.
Differences from the source:
- encoding: standing assumption stated once on p.8 → explicit hypothesis; reason: the paper assumes it throughout; same instances
- encoding: → infimum of ; reason: with the set is nonempty, bounded below by 0 and its infimum is attained
- encoding: products , rankings as bijections, past assortments → products , rankings as permutations, past assortments indexed by ; reason: relabeling only
Preamble
import Mathlib import Definitions.Def_RankRobustAssort_IsOptimalRO import Definitions.Def_RankRobustAssort_SHat
Formal statement
namespace RankRobustAssort
/-- **Theorem 1.** There exists an assortment `S ∈ Ŝ` that is optimal for `(RO)`:
for every problem instance (revenues `0 = r₀ < r₁ < ⋯ < r_n`, any past assortments
`S₁,…,S_M ⊆ 𝒩₀`, normalized sales data `v`, radius `η ≥ 0`, `ℓ₁` or `ℓ∞` norm) whose consistent
set `𝒰` is nonempty, some `S ∈ Ŝ` satisfies `min_{λ∈𝒰} ℛ^λ(S') ≤ min_{λ∈𝒰} ℛ^λ(S)` for every
`S' ∈ 𝒮`.
Sturt, arXiv:2112.05010v4, Theorem 1, p.12.
Formalization Note: `𝒰 ≠ ∅` is the paper's standing assumption (p.8), taken as the hypothesis
`hU`; without it `worstCase` would be the junk value `sInf ∅ = 0`. Optimality (`IsOptimalRO`) is
against all assortments containing `0`. -/
theorem exists_optimal_mem_SHat {n M : ℕ} (P : Problem n M) (hU : (consistentSet P).Nonempty) :
∃ S ∈ SHat P, IsOptimalRO P S := by sorry
end RankRobustAssort
Source
Sturt, The value of robust assortment optimization under ranking-based choice models, arXiv:2112.05010v4, Theorem 1, p.12
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.