Isotropic Standardization of a Compact Posterior
ProvedFunctionalInequalities.KLS.Isotropic.exists_inverseSqrt_equiv_isotropic_standardized_modularThis is a canonical standardization statement related to the posterior in Equation (67) of Bizeul–Klartag–Lehec (arXiv:2610.05474v1), Equation (67), p. 19.
Let be a finite-dimensional real inner-product space with Borel measurable structure. Let be an isotropic probability measure with compact support, let , and let be continuous linear. Then there exists a continuous linear equivalence such that
and the pushforward by of the posterior centered at its mean is isotropic:
The posterior is compactPosterior μ hc c Q, and the covariance operator is Covariance.operator. The conclusion is existential and applies only under the compact-support and finite-dimensional hypotheses shown.
Superseded upload. This snapshot was replaced by FunctionalInequalities.KLS.Isotropic.exists_inverseSqrt_equiv_isotropic_standardized_readable: open the readable replacement. The replacement preserves the frozen source and mathematical statement, with a compact declaration context. This historical node remains available for existing imports.
import Definitions.Def_KLSModular_FunctionalInequalities_KLS_BKL_Posterior_Standardization
noncomputable section
open MeasureTheory
open FunctionalInequalities.Probability.Posterior FunctionalInequalities.Analysis
open scoped RealInnerProductSpace ContDiff
open Set
section
namespace FunctionalInequalities
end FunctionalInequalities
open FunctionalInequalities
section
namespace FunctionalInequalities.KLS
end FunctionalInequalities.KLS
open FunctionalInequalities FunctionalInequalities.KLS
variable {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℝ E]
[FiniteDimensional ℝ E] [MeasurableSpace E] [BorelSpace E]
section CanonicalStandardisation
theorem FunctionalInequalities.KLS.Isotropic.exists_inverseSqrt_equiv_isotropic_standardized_modular (μ : ProbabilityMeasure E)
(hc : IsCompact (μ : Measure E).support) (hiso : KLS.Isotropic μ) (c : E) (Q : E →L[ℝ] E) :
∃ W : E ≃L[ℝ] E,
(W : E →L[ℝ] E) = inverseSqrt (Covariance.operator (compactPosterior μ hc c Q : Measure E)) ∧
KLS.Isotropic (Probability.mapContinuousLinearEquiv
(Probability.translateLaw (compactPosterior μ hc c Q) (-posteriorMean μ c Q)) W) := by sorry