Proof of Theorem 3, p. 24 — the liability game is convex
ProvedSeqLoss.LiabilityGame.liabilityGame_convexconvex-gamecooperative-gameliabilityp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a loss problem with agents and let be its liability game ( if the initiator , and otherwise). Then is a convex game: and for all coalitions ,
Convexity is what makes the Shapley value of lie in its core and, for big boss games, coincide with the nucleolus (Muto et al. 1988, Theorem 4.5).
Formalization Note. The paper shows convexity in the form of increasing marginal contributions, for ; the statement uses the platform's published IsConvexGame, which is the equivalent supermodular form together with . The hypothesis for all is the paper's .
Preamble
import Mathlib import Definitions.Def_Supermodularity_Monotonicity_SupermodularOn import Definitions.Def_Supermodularity_Cooperative_IsConvexGame import Definitions.Def_SeqLoss_FixedFraction_Setting import Definitions.Def_SeqLoss_LiabilityGame_Game
Formal statement
namespace SeqLoss.LiabilityGame
theorem liabilityGame_convex {n : ℕ} [NeZero n] (ℓ : Fin n → ℝ) (hℓ : SeqLoss.FixedFraction.IsLoss ℓ) :
Supermodularity.Cooperative.IsConvexGame (liabilityGame ℓ) := by sorry
end SeqLoss.LiabilityGame
Source
Gudmundsson, Hougaard & Ko, Sharing Sequentially Triggered Losses, accepted manuscript (2023), p. 24, Appendix A, proof of Theorem 3, first paragraph
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.