Lévy's upward theorem
ProvedFamousTheorems.levy_upward_theoremLévy's upward theorem. Let be a filtration on a finite measure space , and let . For every real function ,
Lévy's upward theorem (1937) is one of the basic martingale convergence theorems: conditional expectations along an increasing flow of information converge to the conditional expectation given all of it. It implies the Lévy zero–one law and is used throughout probability, ergodic theory and statistics.
Formalization note. Mathlib's MeasureTheory.tendsto_ae_condExp. μ[g | m] is the conditional expectation, and ⨆ n, ℱ n is the σ-algebra generated by the filtration. Mathlib defines the conditional expectation of a non-integrable function to be , so the statement holds trivially for such and needs no integrability hypothesis.
Preamble
import Mathlib
Formal statement
namespace FamousTheorems
open MeasureTheory
theorem levy_upward_theorem {Ω : Type*} {m0 : MeasurableSpace Ω} {μ : Measure Ω} {ℱ : Filtration ℕ m0} [IsFiniteMeasure μ]
(g : Ω → ℝ) :
∀ᵐ x ∂μ, Filter.Tendsto (fun n => μ[g | ℱ n] x) Filter.atTop (nhds (μ[g | ⨆ n, ℱ n] x)) := by sorry
end FamousTheoremsSource
Listed in Mathlib's curated theorem manifests (docs/1000.yaml); formalized in Mathlib as `MeasureTheory.tendsto_ae_condExp`. Proof here reduces to that Mathlib result.