AFTD Auto-Formalizing Theoretical Domains

CommunicationComplexity.TVDistance.exists_event_abs_signedMeasureDiff_eq_half_totalVariation

theoremverifiedimported Lean

Attainment of half the total variation by a single event. Let $\mu$ and $\nu$ be probability measures on a measurable space $\Omega$, and let $\mu - \nu$ denote their signed measure difference. Then there exists a measurable set $S \subseteq \Omega$ at which \[ \abs{(\mu - \nu)(S)} = \tfrac{1}{2}\,\abs{\mu - \nu}(\Omega), \] where $\abs{\mu - \nu}$ is the total variation measure of $\mu - \nu$; that is, some event realizes exactly half the total variation of $\mu - \nu$ over the whole space.

Statement

set_option relaxedAutoImplicit false in
set_option autoImplicit false in
open MeasureTheory in
lemma CommunicationComplexity.TVDistance.exists_event_abs_signedMeasureDiff_eq_half_totalVariation
    {Ω : Type*} [MeasurableSpace Ω]
    (μ ν : Measure Ω) [IsProbabilityMeasure μ] [IsProbabilityMeasure ν] :
    ∃ S : {S : Set Ω // MeasurableSet S},
      |signedMeasureDiff μ ν (S : Set Ω)| =
        (1 / 2 : ℝ) * (signedMeasureDiff μ ν).totalVariation.real Set.univ
Source
TCSlib, `CommunicationComplexity.TVDistance.exists_event_abs_signedMeasureDiff_eq_half_totalVariation`
Lean proof
By Lucy Horowitz, Timothe Kasriel, Mihir Singhal, from github.com/Shilun-Allan-Li/tcslib at c8a02591b1e5 (Copyright (c) 2026 Lucy Horowitz, Timothe Kasriel, and Mihir Singhal. All rights reserved, Apache-2.0). Compiled here against leanprover/lean4:v4.33.1, Mathlib v4.33.1; written for leanprover/lean4:v4.25.0.
Verified
07 Oct 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 3
Used by 1

Lean source view module on GitHub

set_option relaxedAutoImplicit false inset_option autoImplicit false inopen MeasureTheory in/-- Some measurable event attains half the total variation mass of `μ − ν`: the complement of a Hahn-decomposition negative set, on which `N` vanishes and `P` has full mass. **Proof sketch.** Take a measurable set `S` from the Hahn decomposition of `μ − ν` such that the positive part `P` vanishes on `S` and the negative part `N` vanishes on `Sᶜ`; the event is `Sᶜ`. (1) Since `P(S) = 0`, `P(Sᶜ) = P(Ω)`. (2) `N(Sᶜ) = 0`. (3) Hence `(μ − ν)(Sᶜ) = P(Sᶜ) − N(Sᶜ) = P(Ω) ≥ 0`, which is half the total variation mass. -/lemma CommunicationComplexity.TVDistance.exists_event_abs_signedMeasureDiff_eq_half_totalVariation    {Ω : Type*} [MeasurableSpace Ω]    (μ ν : Measure Ω) [IsProbabilityMeasure μ] [IsProbabilityMeasure ν] :    ∃ S : {S : Set Ω // MeasurableSet S},      |signedMeasureDiff μ ν (S : Set Ω)| =        (1 / 2 : ℝ) * (signedMeasureDiff μ ν).totalVariation.real Set.univ := by  let s := signedMeasureDiff μ ν  obtain ⟨S, hS, -, -, hPzero, hNzero⟩ :=    s.toJordanDecomposition.exists_compl_positive_negative  refine ⟨⟨Sᶜ, hS.compl⟩, ?_⟩  let P := s.toJordanDecomposition.posPart  let N := s.toJordanDecomposition.negPart  -- Step 1: `P(S) = 0`, so `P(Sᶜ) = P(Ω)`  have hPzero_real : P.real S = 0 := by    rw [measureReal_def, show P S = 0 by simpa [P] using hPzero, ENNReal.toReal_zero]  have hPcompl : P.real Sᶜ = P.real Set.univ := by    rw [measureReal_compl hS, hPzero_real]    ring  -- Step 2: `N(Sᶜ) = 0`  have hNcompl : N.real Sᶜ = 0 := by    rw [measureReal_def, hNzero, ENNReal.toReal_zero]  have hnonneg : 0 ≤ P.real Set.univ := measureReal_nonneg  -- Step 3: `(μ − ν)(Sᶜ) = P(Ω)`, the half total variation  rw [signedMeasure_apply_eq_posPart_sub_negPart s hS.compl]  rw [half_totalVariation_real_univ_eq_posPart_real_univ μ ν]  simp [s, P, N, hPcompl, hNcompl, abs_of_nonneg hnonneg]
See it in the graphDiscuss this resultChallenge it