Unconditional full-path Markov–Tonelli formula for arbitrary past and future Lévy rewards
ProvedAvramDividend.Classical.chat04_unconditional_history_future_path_markov_tonellifull-pathindependent-incrementskilled-semigroupmarkov-tonellitwo-sided-exit
For any spectrally negative Lévy process, time s, F_s-measurable past-state statistic H and measurable nonnegative payoff depending on H and the entire shifted future path, the expected payoff factors as an iterated integral over the law of H and the original whole Lévy path. This theorem has no finite-cylinder independence assumptions, unlike the older conditional chat04_arbitrary_history_full_future_path_markov_tonelli. The exact formula applies to genuinely history-killed rewards and is a direct semigroup input for the remaining positive-capital killed q-scale renewal identity a885f0b3-3960-424d-97e4-abbcffe268ba, then stopped exit 09663d0d and parent 3e1ce2c8.
Preamble
import Mathlib.MeasureTheory.Integral.Lebesgue.Map import Mathlib.MeasureTheory.Measure.Prod import Mathlib.Probability.Independence.Process.Basic import Theorems.Thm_AvramDividend_Classical_levy_deterministicTime_history_future_product_law open MeasureTheory ProbabilityTheory open scoped NNReal ENNReal open AvramDividend.Classical
Formal statement
theorem AvramDividend.Classical.chat04_unconditional_history_future_path_markov_tonelli
{Ω : Type*} [mΩ : MeasurableSpace Ω] {P : Measure Ω}
{𝓕 : Filtration ℝ≥0 mΩ}
{E : Type*} [MeasurableSpace E]
(X : SpectrallyNegativeLevy P 𝓕)
(s : ℝ≥0)
(H : Ω → E) (hH : Measurable[𝓕 s] H)
(f : E × (ℝ≥0 → ℝ) → ℝ≥0∞) (hf : Measurable f) :
(∫⁻ ω, f (H ω,
(fun t : ℝ≥0 => X.X (s + t) ω - X.X s ω)) ∂P) =
∫⁻ h, ∫⁻ y, f (h, y)
∂(Measure.map (fun ω => (fun t : ℝ≥0 => X.X t ω)) P)
∂(Measure.map H P) := by sorrySource
Composition of Proved unconditional entire-path stationary joint product law cd9853a4-d75b-4779-b22b-4c4db8f41256, pinned Mathlib lintegral_map and lintegral_prod. Root-connected through the killed q-scale semigroup and two-sided stopped exit identities.