Learning.IsAlgEnvSeq.condExp_noiseSum_increment
No docstring.
Learning.IsAlgEnvSeq.condExp_noiseSum_increment.{u_1, u_2, u_3} {Ξ© : Type u_1} {π : Type u_2} {π¨ : Type u_3} {mΞ© : MeasurableSpace Ξ©} {mπ : MeasurableSpace π} {mπ¨ : MeasurableSpace π¨} [NormedAddCommGroup π¨] [NormedSpace β π¨] {P : MeasureTheory.Measure Ξ©} [MeasureTheory.IsFiniteMeasure P] {A : β β Ξ© β π} {Y : β β Ξ© β π¨} {alg : Algorithm π π¨} {env : Environment π π¨} [MeasurableSingletonClass π] [SecondCountableTopology π¨] [BorelSpace π¨] [CompleteSpace π¨] (h : IsAlgEnvSeq A Y alg env P) (k : π) (i : β) (hint : MeasureTheory.Integrable (Y i) P) : P[Set.indicator {Ο | A i Ο = k} fun Ο => Y i Ο - Environment.means env A Y (A i Ο) i Ο | β(filtrationAction h) i] =α΅[P] 0Learning.IsAlgEnvSeq.condExp_noiseSum_increment.{u_1, u_2, u_3} {Ξ© : Type u_1} {π : Type u_2} {π¨ : Type u_3} {mΞ© : MeasurableSpace Ξ©} {mπ : MeasurableSpace π} {mπ¨ : MeasurableSpace π¨} [NormedAddCommGroup π¨] [NormedSpace β π¨] {P : MeasureTheory.Measure Ξ©} [MeasureTheory.IsFiniteMeasure P] {A : β β Ξ© β π} {Y : β β Ξ© β π¨} {alg : Algorithm π π¨} {env : Environment π π¨} [MeasurableSingletonClass π] [SecondCountableTopology π¨] [BorelSpace π¨] [CompleteSpace π¨] (h : IsAlgEnvSeq A Y alg env P) (k : π) (i : β) (hint : MeasureTheory.Integrable (Y i) P) : P[Set.indicator {Ο | A i Ο = k} fun Ο => Y i Ο - Environment.means env A Y (A i Ο) i Ο | β(filtrationAction h) i] =α΅[P] 0
Code
lemma IsAlgEnvSeq.condExp_noiseSum_increment [CompleteSpace π¨]
(h : IsAlgEnvSeq A Y alg env P) (k : π) (i : β) (hint : Integrable (Y i) P) :
P[{Ο | A i Ο = k}.indicator (fun Ο β¦ Y i Ο - env.means A Y (A i Ο) i Ο) | h.filtrationAction i]
=α΅[P] 0Proof
by
let c : Ξ© β β := actionIndicator A k i
let g : Ξ© β π¨ := fun Ο β¦ Y i Ο - env.means A Y (A i Ο) i Ο
have h_smul : c β’ g = {Ο | A i Ο = k}.indicator (fun Ο β¦ Y i Ο - env.means A Y (A i Ο) i Ο) := by
ext Ο
by_cases hΟ : A i Ο = k <;> simp [c, g, actionIndicator, hΟ]
have hAG : Measurable[h.filtrationAction i] (A i) := h.adapted_action_filtrationAction i
have hcG : StronglyMeasurable[h.filtrationAction i] c :=
(h.adapted_actionIndicator_filtrationAction k i).stronglyMeasurable
have hgint : Integrable g P := hint.sub (h.integrable_means_action hint)
have hcint : Integrable (c β’ g) P := by
rw [h_smul]
exact integrable_noiseSum_increment h hint k
have hcondg : P[g | h.filtrationAction i] =α΅[P] 0 := by
refine (condExp_sub hint (h.integrable_means_action hint) _).trans ?_
have h1 := h.condExp_feedback i hint
grw [h1]
rw [condExp_of_stronglyMeasurable]
Β· simp
Β· exact h.adapted_means_filtrationAction.stronglyAdapted i
Β· exact h.integrable_means_action hint
have hpull := condExp_smul_of_aestronglyMeasurable_left hcG.aestronglyMeasurable hcint hgint
filter_upwards [hpull, hcondg] with Ο hp hcg
rw [β h_smul, hp]
simp only [Pi.smul_apply', hcg, Pi.ofNat_apply, smul_eq_zero]
rcases eq_or_ne (A i Ο) k with hak | hak
Β· simp
Β· simp [c, actionIndicator, hak]Actions: Source Β· Open Issue
New in v4.34.0-rc2-14-gf86702d (2026-08-25), and its meaning has not changed since.
Self-contained, with its dependencies inlined and proofs replaced by sorry: download the raw file Β· open it in the Lean web editor.
Dependency graph
Audit surface: 8 project declarations, 94 external constants
β Proved: no sorry anywhere in its closure
This is the tool's own reading of one build's recorded axioms, and it is not robust against an author who wants it to pass. Checking meant to be relied on should go through Comparator, which replays the proof through the kernel from an export against an explicit list of permitted axioms.