Proof of Theorem 2, p. 33 — Δ″: (S′)^Γ(A, y) − (S′)^Γ(L, y) = ∫ (α − w(z − y))(𝟙_{A∖L} − 𝟙_{L∖A}) h dz
ProvedMultivarScores.LevelSet.delta_pp_eqconsistent-scoring-functionlevel-setsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1scoring-rules
Let be measurable with , and integrable, let , and let be Borel sets. Fix a realization such that is integrable. Then the difference of level set scores is
This is the quantity of the proof of Theorem 2, with and . Taking the expectation over turns it into display (19).
Formalization Note The paper's partition uses while its target is ; the statement is given for an arbitrary Borel , which covers the target . The integrability of is Assumption 1(c) for this ; without it the Bochner integrals would be . and are kept from Theorem 2's hypotheses.
Preamble
import Mathlib import Definitions.Def_MultivarScores_LevelSet_Setting
Formal statement
namespace MultivarScores.LevelSet
open MeasureTheory
theorem delta_pp_eq {d : ℕ} (w h : MultivarScores.L2Rule.E d → ℝ)
(hw_meas : Measurable w) (hw_nonneg : ∀ z, 0 ≤ w z)
(hh_meas : Measurable h) (hh_nonneg : ∀ z, 0 ≤ h z) (hh_int : Integrable h)
(α : ℝ) (hα : 0 < α) (A L : Set (MultivarScores.L2Rule.E d)) (hA : MeasurableSet A) (hL : MeasurableSet L)
(y : MultivarScores.L2Rule.E d) (hwy_int : Integrable (fun z => w (z - y) * h z)) :
scoreΓ w h α A y - scoreΓ w h α L y =
∫ z, (α - w (z - y)) * ((A \ L).indicator 1 z - (L \ A).indicator 1 z) * h z := by sorry
end MultivarScores.LevelSet
Source
Meng, Taylor, Ben Taieb, Li, Scores for Multivariate Distributions and Level Sets, arXiv:2002.09578v6, p. 33, proof of Theorem 2, display of Δ″ (B₁, B₂, Σ±±)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.