Lemma 2.5 (second part), p. 707 — ∂_t Ent_π(H_t f)|_{t=0} = −ℰ(f, log f)
ProvedLogSobolevMC.Entropy.lemma_2_5_entLet be a Markov kernel on a finite set with invariant probability , for all , let be the associated semigroup and the Dirichlet form. For a function with for all , write . Then
This is the entropy counterpart of the formula ; applied to the adjoint chain it gives the time derivative of the entropy along the semigroup, which drives the proof of Theorem 3.6.
Formalization Note The derivative is a two-sided derivative at (the series defining makes sense for all real ). The paper states the lemma for nonnegative ; the Lean statement asks , because is at a zero of on the page but Lean's Real.log 0 = 0. The normalisation under which the paper calls an entropy is not assumed: invariance of makes the formula hold without it.
import Mathlib import Definitions.Def_mm_basic import Definitions.Def_mm_continuous import Definitions.Def_LogSobolevMC_Entropy_Setting
namespace LogSobolevMC.Entropy
theorem lemma_2_5_ent {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)
(f : V → ℝ) (hf : ∀ x, 0 < f x) :
HasDerivAt (fun t : ℝ => entF π (LogSobolevMC.ChiSquare.heatOp K t f))
(-LogSobolevMC.ChiSquare.dirichlet K π f (fun x => Real.log (f x))) 0 := by sorry
end LogSobolevMC.Entropy
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.