Proof of Theorem 2, p. 23 — squared Euclidean distance from to the ideal allocation
ProvedSeqLoss.Equidistance.sq_dist_idealcost-sharingeuclidean-distancep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let have total , let , and let be the -th ideal allocation. For every ,
The first term does not depend on , so equidistance from reduces to the last two terms being independent of .
Formalization Note The paper writes the identity for ; it holds for every real vector and every real , and is stated so. Agents are 0-based, so is the sum over indices .
Preamble
import Mathlib import Definitions.Def_SeqLoss_FixedFraction_Setting import Definitions.Def_SeqLoss_Equidistance_IdealAllocations
Formal statement
namespace SeqLoss.Equidistance
theorem sq_dist_ideal {n : ℕ} (ℓ x : Fin n → ℝ) (k : ℕ) (hk1 : 1 ≤ k) (hkn : k ≤ n) :
euclDist x (ideal ℓ k) ^ 2 =
(∑ i, x i ^ 2)
- 2 * SeqLoss.FixedFraction.total ℓ / (k : ℝ) * (∑ i ∈ Finset.univ.filter (fun i : Fin n => i.val < k), x i)
+ SeqLoss.FixedFraction.total ℓ ^ 2 / (k : ℝ) := by sorry
end SeqLoss.Equidistance
Source
Gudmundsson, Hougaard & Ko, Sharing Sequentially Triggered Losses, accepted manuscript (2023), p. 23, Appendix A, proof of Theorem 2, first display
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.