Proof of Theorem 5.1, first case — giving all items to the top bidder is a 2√m-approximation
ProvedComplementFreeCA.ValueQuery.case_one_all_to_oneLet be normalized, monotone, complement-free valuations on bundles of and let be an allocation (pairwise disjoint bundles). Call bidder large if and small if . Suppose the large bidders carry at least as much welfare as the small ones:
Then for every bidder maximizing ,
This is the first case of the proof of Theorem 5.1: there are at most large bidders, so one of them is worth , and giving all items to the top bidder achieves it.
Formalization Note The paper writes the conclusion for "the bidder that maximizes "; the conclusion here is for a bidder maximizing , which is the allocation the algorithm compares, and follows by monotonicity. The allocation is arbitrary, not only optimal. At the statement holds trivially (all values are ).
import Mathlib import Definitions.Def_ComplementFreeCA_ValueQuery_Basic open Finset
namespace ComplementFreeCA.ValueQuery
/-- Proof of Theorem 5.1, first case (p. 11): if the bidders whose bundle has at least `√m`
items carry at least as much welfare as the others, then `|OPT| ≤ 2√m · v_{i₀}(M)` for a bidder
`i₀` maximizing `v_i(M)`. -/
theorem case_one_all_to_one {n m : ℕ} (v : Fin n → Finset (Fin m) → ℝ)
(hv : ∀ i, IsCFValuation (v i)) (O : Fin n → Finset (Fin m)) (hO : IsAllocation O)
(hcase : ∑ i ∈ univ.filter (fun i => ((O i).card : ℝ) < Real.sqrt m), v i (O i) ≤
∑ i ∈ univ.filter (fun i => Real.sqrt m ≤ ((O i).card : ℝ)), v i (O i))
(i₀ : Fin n) (hi₀ : ∀ i, v i univ ≤ v i₀ univ) :
welfare v O ≤ 2 * Real.sqrt m * v i₀ univ := by sorry
end ComplementFreeCA.ValueQuery
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.