AutoSamplingTheory.TechnicalLemmas.StochasticProcesses.StoppingGraphNull
6 named declarations scanned from AutoSamplingTheory/TechnicalLemmas/StochasticProcesses/StoppingGraphNull.lean.
Declarations
def AutoSamplingTheory.TechnicalLemmas.StochasticProcesses.StoppingGraphNull.stoppingGraph Compiled Not mapped
- The graph of a possibly-infinite nonnegative stopping time in `Omega × ℝ≥0`.
def stoppingGraph (tau : Omega → WithTop ℝ≥0) : Set (Omega × ℝ≥0) :=
{z | (z.2 : WithTop ℝ≥0) = tau z.1}
/-- A Chewi stopping time has a measurable graph in product space. -/
AutoSamplingTheory/TechnicalLemmas/StochasticProcesses/StoppingGraphNull.lean:33published source at 77184245109a
theorem AutoSamplingTheory.TechnicalLemmas.StochasticProcesses.StoppingGraphNull.measurableSet_stoppingGraph Compiled Not mapped
- A Chewi stopping time has a measurable graph in product space.
theorem measurableSet_stoppingGraph
(tau : Omega → WithTop ℝ≥0)
(htau : IsChewiStoppingTime filtration tau) :
MeasurableSet (stoppingGraph tau) := by
have ht : Measurable (fun z : Omega × ℝ≥0 => (z.2 : WithTop ℝ≥0)) :=
(measurable_snd : Measurable (fun z : Omega × ℝ≥0 => z.2)).withTop_coe
have htau' : MeasureTheory.IsStoppingTime filtration tau := htau
have hτ : Measurable (fun z : Omega × ℝ≥0 => tau z.1) :=
htau'.measurable'.comp measurable_fst
have heq :
stoppingGraph tau =
{z : Omega × ℝ≥0 | (z.2 : WithTop ℝ≥0) ≤ tau z.1} ∩
{z : Omega × ℝ≥0 | tau z.1 ≤ (z.2 : WithTop ℝ≥0)} := by
ext z
constructor
· intro hz
change (z.2 : WithTop ℝ≥0) = tau z.1 at hz
exact ⟨hz.le, hz.ge⟩
· rintro ⟨hz₁, hz₂⟩
change (z.2 : WithTop ℝ≥0) = tau z.1
exact le_antisymm hz₁ hz₂
rw [heq]
exact (measurableSet_le ht hτ).inter (measurableSet_le hτ ht)
/-- Every fixed-sample-path section of the stopping graph has zero stopped
Lebesgue-time measure. The proof avoids choosing an `untop`: if a finite
section point exists, injectivity of the `WithTop` coercion makes the whole
section that singleton; otherwise the section is empty. -/
AutoSamplingTheory/TechnicalLemmas/StochasticProcesses/StoppingGraphNull.lean:37published source at 77184245109a
theorem AutoSamplingTheory.TechnicalLemmas.StochasticProcesses.StoppingGraphNull.timeSection_stoppingGraph_zero Compiled Not mapped
- Every fixed-sample-path section of the stopping graph has zero stopped Lebesgue-time measure. The proof avoids choosing an `untop`: if a finite section point exists, injectivity of the `WithTop` coercion makes the whole section that singleton; otherwise the section is empty.
theorem timeSection_stoppingGraph_zero
(tau : Omega → WithTop ℝ≥0) (T : ℝ≥0) (omega : Omega) :
(TimeMeasure.upTo T) (Prod.mk omega ⁻¹' stoppingGraph tau) = 0 := by
by_cases hex : ∃ s : ℝ≥0, (s : WithTop ℝ≥0) = tau omega
· obtain ⟨s, hs⟩ := hex
have hset : Prod.mk omega ⁻¹' stoppingGraph tau = ({s} : Set ℝ≥0) := by
ext r
change ((r : WithTop ℝ≥0) = tau omega) ↔ r = s
constructor
· intro hr
apply WithTop.coe_eq_coe.mp
exact hr.trans hs.symm
· intro hrs
subst r
exact hs
rw [hset, TimeMeasure.upTo_singleton]
· have hset : Prod.mk omega ⁻¹' stoppingGraph tau = (∅ : Set ℝ≥0) := by
ext r
change ((r : WithTop ℝ≥0) = tau omega) ↔ False
constructor
· intro hr
exact hex ⟨r, hr⟩
· intro hfalse
exact hfalse.elim
rw [hset, measure_empty]
/-- The stopping-time graph is null under Chewi's product process-time
measure. -/
AutoSamplingTheory/TechnicalLemmas/StochasticProcesses/StoppingGraphNull.lean:65published source at 77184245109a
theorem AutoSamplingTheory.TechnicalLemmas.StochasticProcesses.StoppingGraphNull.processTimeMeasure_stoppingGraph_zero Compiled Not mapped
- The stopping-time graph is null under Chewi's product process-time measure.
theorem processTimeMeasure_stoppingGraph_zero
(tau : Omega → WithTop ℝ≥0)
(htau : IsChewiStoppingTime filtration tau) (T : ℝ≥0) :
processTimeMeasure mu T (stoppingGraph tau) = 0 := by
rw [processTimeMeasure,
Measure.prod_apply (measurableSet_stoppingGraph tau htau)]
simp_rw [timeSection_stoppingGraph_zero tau T]
simp
/-- Almost every product-space point avoids the stopping-time graph. -/
AutoSamplingTheory/TechnicalLemmas/StochasticProcesses/StoppingGraphNull.lean:93published source at 77184245109a
theorem AutoSamplingTheory.TechnicalLemmas.StochasticProcesses.StoppingGraphNull.ae_notMem_stoppingGraph Compiled Not mapped
- Almost every product-space point avoids the stopping-time graph.
theorem ae_notMem_stoppingGraph
(tau : Omega → WithTop ℝ≥0)
(htau : IsChewiStoppingTime filtration tau) (T : ℝ≥0) :
∀ᵐ z ∂processTimeMeasure mu T, z ∉ stoppingGraph tau :=
measure_eq_zero_iff_ae_notMem.mp
(processTimeMeasure_stoppingGraph_zero tau htau T)
/-- Pointwise form of `ae_notMem_stoppingGraph`: almost every product-space
point has time coordinate different from the stopping time. -/
AutoSamplingTheory/TechnicalLemmas/StochasticProcesses/StoppingGraphNull.lean:103published source at 77184245109a
theorem AutoSamplingTheory.TechnicalLemmas.StochasticProcesses.StoppingGraphNull.ae_time_ne_stoppingTime Compiled Not mapped
- Pointwise form of `ae_notMem_stoppingGraph`: almost every product-space point has time coordinate different from the stopping time.
theorem ae_time_ne_stoppingTime
(tau : Omega → WithTop ℝ≥0)
(htau : IsChewiStoppingTime filtration tau) (T : ℝ≥0) :
∀ᵐ z ∂processTimeMeasure mu T,
(z.2 : WithTop ℝ≥0) ≠ tau z.1 := by
filter_upwards [ae_notMem_stoppingGraph (mu := mu) tau htau T] with z hz
exact hz
end StoppingGraphNull
end StochasticProcesses
end TechnicalLemmas
end AutoSamplingTheory
AutoSamplingTheory/TechnicalLemmas/StochasticProcesses/StoppingGraphNull.lean:112published source at 77184245109a