Example 3.1, p. 715 — the two-point chain has π ≡ 1/2, λ = 2 and α = λ/2 = 1
ProvedLogSobolevMC.Metropolis.example_3_1log-sobolev-inequalitymarkov-chainsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1spectral-gap
Consider the two-point space with kernel (and ). Its stationary measure is , its spectral gap is , and its log-Sobolev constant is
The two-point chain is the building block of the hypercube walk (Example 3.2); it shows that the inequality of Lemma 3.1 is sharp.
Formalization Note is encoded as Fin 2; and are the variational infima of the mission's Setting.
Preamble
import Mathlib import Definitions.Def_LogSobolevMC_Metropolis_Setting import Definitions.Def_LogSobolevMC_Metropolis_Chains
Formal statement
namespace LogSobolevMC.Metropolis
/-- Example 3.1, p. 715: the two-point chain `K(−1, 1) = K(1, −1) = 1` on `{−1, 1}` has
stationary measure `π ≡ 1/2`, spectral gap `λ = 2` and log-Sobolev constant `α = λ/2 = 1`. -/
theorem example_3_1 :
MarkovMixing.IsStationary swap2 (fun _ => (1 / 2 : ℝ)) ∧
LogSobolevMC.ChiSquare.gap swap2 (fun _ => (1 / 2 : ℝ)) = 2 ∧
LogSobolevMC.ChiSquare.logSobolev swap2 (fun _ => (1 / 2 : ℝ)) = LogSobolevMC.ChiSquare.gap swap2 (fun _ => (1 / 2 : ℝ)) / 2 ∧
LogSobolevMC.ChiSquare.logSobolev swap2 (fun _ => (1 / 2 : ℝ)) = 1 := by sorry
end LogSobolevMC.Metropolis
Source
Diaconis and Saloff-Coste, Logarithmic Sobolev inequalities for finite Markov chains, Ann. Appl. Probab. 6 (1996), p. 715, Example 3.1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.