§2.2 — comparison inequality
ProvedHighMagSampling.SK.chiSq_ge_four_tv_sqchi-square-divergencep2o-gran-per-chapterp2o-plan-paperp2o-v1total-variation
Let be probability measures on a measurable space . The total variation distance is over measurable , and the -divergence is when and otherwise. Then
Together with Lemma 1 this converts the contraction of the lazy down-up walk into the total-variation guarantee of Theorem 4.
Formalization Note. TV and are the published PolySampIso.tvDist and PolySampIso.chiSq; is -valued, and the inequality is stated in .
Differences from the source:
- reading: densities on a general → the Radon–Nikodym derivative , and when ; reason: §2.1 writes the formulas with " a counting measure if is discrete"; this is their measure-theoretic meaning.
- encoding: → (
PolySampIso.tvDist); reason: equal for probability measures (pass to the complement). - encoding: in →
ENNReal.ofReal (4 * TV²) ≤ χ²in ; reason: may be ; same instances.
Preamble
import Mathlib import Definitions.Def_PolySampIso_Divergences open MeasureTheory open scoped ENNReal
Formal statement
namespace HighMagSampling.SK
/-- §2.2, p. 11, the comparison inequality: `χ²(μ ‖ π) ≥ 4 TV(μ, π)²` for probability measures on
a general measurable space. -/
theorem chiSq_ge_four_tv_sq {Ω : Type*} [MeasurableSpace Ω] (μ π : Measure Ω)
[IsProbabilityMeasure μ] [IsProbabilityMeasure π] :
ENNReal.ofReal (4 * PolySampIso.tvDist μ π ^ 2) ≤ PolySampIso.chiSq μ π := by sorry
end HighMagSampling.SK
Source
Kumar, Sarkar, Tian, Zhu, High-Magnetization Sampling at Low Temperatures, arXiv:2609.08873v1, p. 11, §2.2, comparison inequality; p. 9, §2.1 (TV, χ²)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.