(1.8), p. 40 — ½Σ_{x,y}(φ(x)+φ(y))²Q(x,y) = E(φ²) + ⟨φ, Pφ⟩_π
ProvedDiaconisStroock.OddPaths.eq_1_8dirichlet-formidentitymarkov-chainp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a transition matrix on a finite set and a stationary distribution of , and write . For every function ,
where is the expectation under , , and .
This identity converts the sum over edges produced by the path argument into the quadratic form , which equals on an eigenfunction with eigenvalue .
Formalization Note Only stochasticity of and stationarity of are assumed; reversibility and irreducibility, standing assumptions of the paper, are not needed for this identity and are omitted, which makes the statement more general.
Preamble
import Mathlib import Definitions.Def_mm_spectral open MarkovMixing
Formal statement
namespace DiaconisStroock.OddPaths
theorem eq_1_8 {V : Type*} [Fintype V] [DecidableEq V] (P : Matrix V V ℝ) (hP : IsStochastic P)
(π : V → ℝ) (hπ : IsStationary P π) :
∀ φ : V → ℝ, 2⁻¹ * ∑ x, ∑ y, (φ x + φ y) ^ 2 * edgeMeasure P π x y =
innerPi π φ φ + innerPi π φ (P.mulVec φ) := by sorry
end DiaconisStroock.OddPaths
Source
Diaconis and Stroock, Geometric bounds for eigenvalues of Markov chains, Ann. Appl. Probab. 1 (1991), p. 40, proof of Proposition 2, (1.8), https://doi.org/10.1214/aoap/1177005980
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.