Lemma 3.1, p. 715 — 2α ≤ λ for any chain
ProvedLogSobolevMC.Metropolis.lemma_3_1log-sobolev-inequalitymarkov-chainsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1spectral-gap
Let be a Markov kernel on a finite set with an invariant probability charging every point. Its log-Sobolev constant and spectral gap satisfy
Together with a lower bound on obtained otherwise, this caps the log-Sobolev constant by the spectral gap; in Example 3.3 it turns the bound and the known gap into the equality .
Formalization Note The page's section assumes irreducibility; the statement does not need it and is stated without it (a stronger statement). On a one-point space both sides are in Lean.
Preamble
import Mathlib import Definitions.Def_mm_basic import Definitions.Def_LogSobolevMC_Metropolis_Setting
Formal statement
namespace LogSobolevMC.Metropolis
/-- Lemma 3.1, p. 715: for any finite chain `K` with positive invariant probability `π`,
the log-Sobolev constant and the spectral gap satisfy `2α ≤ λ`. -/
theorem lemma_3_1 {V : Type*} [Fintype V] [DecidableEq V]
(K : Matrix V V ℝ) (hK : MarkovMixing.IsStochastic K)
(π : V → ℝ) (hπ : MarkovMixing.IsStationary K π) (hπpos : ∀ x, 0 < π x) :
2 * LogSobolevMC.ChiSquare.logSobolev K π ≤ LogSobolevMC.ChiSquare.gap K π := by sorry
end LogSobolevMC.Metropolis
Source
Diaconis and Saloff-Coste, Logarithmic Sobolev inequalities for finite Markov chains, Ann. Appl. Probab. 6 (1996), p. 715, Lemma 3.1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.