Theorem 2 — convergence to uniformity from every starting point
ProvedSmithHitAndRun.SymMixing.theorem_2_strongly_mixingLet be finite nonzero content on and a Markov kernel with a jointly measurable symmetric density relative to . Suppose for every pair . Write and for the -step transition probability. Then for every starting point and every measurable ,
This is the paper's meaning of strongly mixing: setwise convergence to the uniform law from each point. It supplies the qualitative convergence target behind the Random Directions Algorithm. Formalization Note The abstract measurable space represents the whole region and its content; symmetry of encodes the consequence Smith asserts for symmetric mixing algorithms satisfying Assumption (a). The statement asks for every , and does not assume stationarity, uniqueness, or a convergence bound.
import Mathlib.Probability.Kernel.Invariance import Definitions.Def_MarkovIterKernel import Definitions.Def_SmithHitAndRun_SymMixing_uniformLaw import Definitions.Def_SmithHitAndRun_SymMixing_IsSymmetricMixingKernel open MeasureTheory ProbabilityTheory Filter
namespace SmithHitAndRun.SymMixing
/-- Smith (1984), Theorem 2, p. 1301: setwise convergence from every start. -/
theorem theorem_2_strongly_mixing {α : Type*} [MeasurableSpace α]
(V : Measure α) [IsFiniteMeasure V] (hV : V ≠ 0)
(P : Kernel α α) [IsMarkovKernel P] (f : α → α → ENNReal)
(h : IsSymmetricMixingKernel V P f) (hpos : ∀ x y, 0 < f y x) :
∀ x : α, ∀ A : Set α, MeasurableSet A →
Tendsto (fun m : ℕ => MarkovChainCLT.iterKernel P m x A)
atTop (nhds (uniformLaw V A)) := by sorry
end SmithHitAndRun.SymMixing
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.