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.CommunicationComplexity.TVDistance.exists_event_abs_signedMeasureDiff_eq_half_totalVariation← CommunicationComplexity.TVDistance.exists_event_abs_signedMeasureDiff_eq_half_totalVariation (TCSlib/CommunicationComplexity/NewmanTheorem/TVDistance.lean:186): verbatim
- Verified
- 07 Oct 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 3
- CommunicationComplexity.TVDistance.half_totalVariation_real_univ_eq_posPart_real_univHalf the total variation equals the positive Jordan part. Let $\mu$ and $\nu$ be probability measures on a measurable space $\Omega$, and…
- CommunicationComplexity.signedMeasureDiffGiven two probability measures $\mu$ and $\nu$ on a measurable space $\Omega$, the signed measure difference $\mu - \nu$ is the…
- CommunicationComplexity.TVDistance.signedMeasure_apply_eq_posPart_sub_negPartSigned measure via Jordan decomposition. Let $s$ be a signed measure on a measurable space $\Omega$, and let $s = s^+ - s^-$ be its Jordan…
- Used by 1
- CommunicationComplexity.TVDistance.exists_event_abs_measureReal_sub_eq_half_totalVariationAttainment of the total-variation distance. Let $\mu$ and $\nu$ be probability measures on a measurable space $\Omega$, and let…
Lean source view module on GitHub
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
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]