Finite almost-sure partition transfers independent event probabilities
ProvedStabilityDichotomies.Bijunctive.finite_almost_sure_partition_independent_event_probabilityPartition the first factor of an independent product probability space into finitely many measurable events, up to a null set. If an event in the product is described on each partition cell by a measurable event in the second factor, its probability equals the sum of the cell probability times the corresponding second-factor event probability. This transfers conditional priority orderings to independent exponential-shift event costs.
Preamble
import Mathlib open scoped Classical ENNReal BigOperators open MeasureTheory ProbabilityTheory Set Function
Formal statement
theorem StabilityDichotomies.Bijunctive.finite_almost_sure_partition_independent_event_probability {Ω Δ ι : Type}
[MeasurableSpace Ω] [MeasurableSpace Δ] [Fintype ι]
(μ : Measure Ω) (ν : Measure Δ) [IsProbabilityMeasure μ] [IsProbabilityMeasure ν]
(P : ι → Set Ω) (hP : ∀ i, MeasurableSet (P i)) (hdis : Pairwise (Disjoint on P))
(hcover : ∀ᵐ ω ∂μ, ∃ i, ω ∈ P i)
(E : ι → Set Δ) (hE : ∀ i, MeasurableSet (E i)) (T : Set (Ω × Δ))
(hsection : ∀ i ω δ, ω ∈ P i → ((ω,δ) ∈ T ↔ δ ∈ E i)) :
((μ.prod ν) T).toReal = ∑ i : ι, μ.real (P i) * ν.real (E i) := by
sorrySource
Wang and Yoshida, Stability Dichotomies for Boolean CSPs, STOC 2027, https://github.com/sigact-stoc/stoc2027/releases/download/papers-1/535.pdf, proof of Theorem 4.9, conditional ordering and independent shifts.