Proof of Proposition 3.15, p. 17, first display — dcov((µ₁ × δ(y₁) + µ₂ × δ(y₂))/2) = −d(y₁, y₂) D(µ₁ − µ₂)/8
ProvedDistCov.Converse.dcov_twoPointLawdistance-covariancenegative-typep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let and be separable metric spaces, let be Borel probability measures on with finite first moments, and let be points of . Put
Then
In the paper this identity is used twice: with it gives a law with negative distance covariance, and with a law with zero distance covariance. The sign of is not assumed here.
Formalization Note. Separability of both spaces is the standing assumption of Errata (i). The marginals of are and and have finite first moments, so all integrals in are genuine. The hypothesis is kept as on the page, although the identity also holds (with both sides ) when .
Preamble
import Mathlib import Definitions.Def_DistCov_Converse_Setting open MeasureTheory
Formal statement
namespace DistCov.Converse
theorem dcov_twoPointLaw
{X : Type*} [MetricSpace X] [MeasurableSpace X] [BorelSpace X] [SecondCountableTopology X]
{Y : Type*} [MetricSpace Y] [MeasurableSpace Y] [BorelSpace Y] [SecondCountableTopology Y]
(μ₁ μ₂ : Measure X) [IsProbabilityMeasure μ₁] [IsProbabilityMeasure μ₂]
(hμ₁ : DistCov.Indep.FiniteFirstMoment μ₁) (hμ₂ : DistCov.Indep.FiniteFirstMoment μ₂)
(y₁ y₂ : Y) (hy : y₁ ≠ y₂) :
DistCov.Indep.dcov (twoPointLaw μ₁ μ₂ y₁ y₂) = -(dist y₁ y₂) * DistCov.Indep.Ddiff μ₁ μ₂ / 8 := by sorry
end DistCov.Converse
Source
Lyons, Distance covariance in metric spaces, arXiv:1106.5758 (version of 19 Dec. 2020), p. 17, proof of Proposition 3.15, first display
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.