Documentation

TauCeti.Probability.Martingale.LevyDownwardEventuallyConst

Lévy's downward theorem on an eventually constant filtration #

This file is the Exchangeability-roadmap worked example (TauCetiRoadmap/Exchangeability/README.md, "Worked examples"): the Lévy downward theorem specializes correctly to an eventually constant decreasing filtration. It is a consistency test of the flagship Layer-4 result MeasureTheory.tendsto_ae_condExp_iInf, not a de Finetti example.

Fix an antitone filtration 𝔽 : ℕ → MeasurableSpace Ω that is constant from some index N on, 𝔽 n = 𝔽 N for n ≥ N. Two independent computations must agree:

The capstone tendsto_ae_condExp_of_eventually_const runs the general theorem and rewrites its limit through the intersection collapse, recovering exactly the elementary limit: the two routes agree, which is what "specializes correctly" means.

No material from cameronfreer/exchangeability is used; this is a Tau-Ceti-local check of the flagship theorem stated in TauCeti.Probability.Martingale.Convergence.

theorem TauCeti.Probability.iInf_measurableSpace_eq_of_antitone_of_eventually_const {Ω : Type u_2} {𝔽 : ℕ → MeasurableSpace Ω} (h_anti : Antitone 𝔽) {N : ℕ} (h_const : ∀ (n : ℕ), N ≤ n → 𝔽 n = 𝔽 N) :
⨅ (n : ℕ), 𝔽 n = 𝔽 N

The intersection of an antitone filtration that is constant from index N on collapses to the stabilized σ-algebra 𝔽 N. This is the order fact behind the Lévy-downward specialization.

theorem TauCeti.Probability.condExp_eq_of_eventually_const {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} {𝔽 : ℕ → MeasurableSpace Ω} {N : ℕ} (h_const : ∀ (n : ℕ), N ≤ n → 𝔽 n = 𝔽 N) (f : Ω → ℝ) {n : ℕ} (hn : N ≤ n) :
μ[f | 𝔽 n] = μ[f | 𝔽 N]

Past the stabilization index the conditional expectations are literally equal: for N ≤ n the σ-algebras 𝔽 n and 𝔽 N coincide, so μ[f | 𝔽 n] and μ[f | 𝔽 N] are the same term.

theorem TauCeti.Probability.tendsto_condExp_of_eventually_const {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} {𝔽 : ℕ → MeasurableSpace Ω} {N : ℕ} (h_const : ∀ (n : ℕ), N ≤ n → 𝔽 n = 𝔽 N) (f : Ω → ℝ) (ω : Ω) :
Filter.Tendsto (fun (n : ℕ) => μ[f | 𝔽 n] ω) Filter.atTop (nhds (μ[f | 𝔽 N] ω))

Elementary convergence. Along a filtration constant from index N on, the conditional expectations μ[f | 𝔽 n] reach μ[f | 𝔽 N] exactly and stay there, so they converge to it at every point. No integrability of f or finiteness of μ is needed: the sequence is eventually constant.

theorem TauCeti.Probability.condExp_iInf_eq_of_eventually_const {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} {𝔽 : ℕ → MeasurableSpace Ω} (h_anti : Antitone 𝔽) {N : ℕ} (h_const : ∀ (n : ℕ), N ≤ n → 𝔽 n = 𝔽 N) (f : Ω → ℝ) :
μ[f | ⨅ (n : ℕ), 𝔽 n] = μ[f | 𝔽 N]

The Lévy downward limit collapses. For an antitone filtration constant from index N on, the reverse-martingale limit μ[f | ⨅ n, 𝔽 n] produced by tendsto_ae_condExp_iInf is the conditional expectation onto the stabilized σ-algebra 𝔽 N.

theorem TauCeti.Probability.tendsto_ae_condExp_of_eventually_const {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} {𝔽 : ℕ → MeasurableSpace Ω} [MeasureTheory.IsFiniteMeasure μ] (h_anti : Antitone 𝔽) (h_le0 : 𝔽 0 ≤ inferInstance) {N : ℕ} (h_const : ∀ (n : ℕ), N ≤ n → 𝔽 n = 𝔽 N) (f : Ω → ℝ) :
∀ᵐ (ω : Ω) ∂μ, Filter.Tendsto (fun (n : ℕ) => μ[f | 𝔽 n] ω) Filter.atTop (nhds (μ[f | 𝔽 N] ω))

Lévy's downward theorem specializes correctly to an eventually constant filtration.

Run the flagship MeasureTheory.tendsto_ae_condExp_iInf on an antitone filtration that is constant from index N on, then collapse its limit through condExp_iInf_eq_of_eventually_const. The a.e. limit it delivers is μ[f | 𝔽 N] — exactly the value the sequence already reaches by the elementary tendsto_condExp_of_eventually_const. The general and elementary computations agree.