Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem A.1, pp. 742–743 — the chain K(x, y) = π(y) has α = (1 − 2π_*)/log(1/π_* − 1)

Proved
LogSobolevMC.TwoPoint.theorem_A_1

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

complete-graphlog-sobolev-inequalitymarkov-chainsp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let π\piπ be a probability measure on a finite set X\mathcal XX with π(x)>0\pi(x)>0π(x)>0 for every xxx, and set π∗=min⁡Xπ\pi_*=\min_{\mathcal X}\piπ∗​=minX​π. Consider the Markov chain K(x,y)=π(y)K(x,y)=\pi(y)K(x,y)=π(y), which has invariant measure π\piπ. If π∗<1/2\pi_*<1/2π∗​<1/2, its log-Sobolev constant α\alphaα is

α=1−2π∗log⁡(1/π∗−1).\alpha=\frac{1-2\pi_*}{\log(1/\pi_*-1)}.α=log(1/π∗​−1)1−2π∗​​.

In particular, for the uniform measure π≡1/∣X∣\pi\equiv1/|\mathcal X|π≡1/∣X∣ on a set with ∣X∣≥3|\mathcal X|\ge3∣X∣≥3 points,

α=1−2/∣X∣log⁡(∣X∣−1).\alpha=\frac{1-2/|\mathcal X|}{\log(|\mathcal X|-1)}.α=log(∣X∣−1)1−2/∣X∣​.

This is the exact log-Sobolev constant of the simplest nontrivial chain, the one that forgets its state in one step. Since its spectral gap is 111, it shows that α\alphaα can be of smaller order than λ/2\lambda/2λ/2 (by a factor log⁡(1/π∗)\log(1/\pi_*)log(1/π∗​)) and gives the comparison bound of Corollary A.4 for every finite chain.

Formalization Note The page assumes only that π\piπ is positive. The hypothesis π∗<1/2\pi_*<1/2π∗​<1/2 (and ∣X∣≥3|\mathcal X|\ge3∣X∣≥3 in the uniform case) is added and disclosed: π∗≤1/2\pi_*\le1/2π∗​≤1/2 always holds when ∣X∣≥2|\mathcal X|\ge2∣X∣≥2, and at π∗=1/2\pi_*=1/2π∗​=1/2 (the uniform measure on two points) the printed formula is 0/00/00/0, which Lean would evaluate to 000, while the true value is 1/21/21/2 (Theorem A.2 at θ=1/2\theta=1/2θ=1/2); on a one-point set π∗=1\pi_*=1π∗​=1 and the formula is meaningless. The infimum (A.2) is Lean's sInf of the set of ratios.

Preamble
import Mathlib
import Definitions.Def_mm_basic
import Definitions.Def_mm_lower
import Definitions.Def_LogSobolevMC_TwoPoint_Setting
Formal statement
namespace LogSobolevMC.TwoPoint

open MarkovMixing

/-- Theorem A.1 (pp. 742–743): for a positive probability `π` on a finite set with
`π_* = min π < 1/2`, the chain `K(x, y) = π(y)` has invariant measure `π` and log-Sobolev constant
`(1 − 2π_*)/log(1/π_* − 1)`; in particular, for the uniform `π ≡ 1/|𝒳|` with `|𝒳| ≥ 3`,
`α = (1 − 2/|𝒳|)/log(|𝒳| − 1)`. -/
theorem theorem_A_1 {V : Type*} [Fintype V] [DecidableEq V] [Nonempty V]
    (π : V → ℝ) (hπ : IsDist π) (hπpos : ∀ x, 0 < π x) (hhalf : piMin π < 1 / 2) :
    IsStationary (indepChain π) π ∧
    LogSobolevMC.ChiSquare.logSobolev (indepChain π) π = (1 - 2 * piMin π) / Real.log (1 / piMin π - 1) ∧
      (3 ≤ Fintype.card V →
        LogSobolevMC.ChiSquare.logSobolev (indepChain (uniformDist V)) (uniformDist V) =
          (1 - 2 / (Fintype.card V : ℝ)) / Real.log ((Fintype.card V : ℝ) - 1)) := by sorry

end LogSobolevMC.TwoPoint
Source
Diaconis and Saloff-Coste, Logarithmic Sobolev inequalities for finite Markov chains, Ann. Appl. Probab. 6 (1996), pp. 742–743, Theorem A.1
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me