Extending strong extraction from flat to weighted sources #
Seeded test probabilities are linear in the source weights. The exact decomposition into uniform fixed-size supports therefore transfers the flat-source guarantee without any error loss. Conversely every nonempty flat support is a normalized capped weighting on the ambient finite type.
theorem
Algebraic.Cutwidth.Extractor.Internal.probabilityWeight_flatWeight
{α : Type u_1}
[Fintype α]
(P : Finset α)
(nonempty : P.Nonempty)
:
theorem
Algebraic.Cutwidth.Extractor.Internal.cappedWeight_flatWeight
{α : Type u_1}
(P : Finset α)
{K : ℕ}
(size : K ≤ P.card)
:
CappedWeight (flatWeight P) K
theorem
Algebraic.Cutwidth.Extractor.Internal.weightedSeededTestProb_mixture
{α : Type u_1}
{ι : Type u_2}
{Seed : Type u_3}
{Ω : Type u_4}
[Fintype α]
[Fintype ι]
[Fintype Seed]
(w : ι → ℝ)
(p : ι → α → ℝ)
(E : α → Seed → Ω)
(T : Finset (Seed × Ω))
:
weightedSeededTestProb (fun (x : α) => ∑ i : ι, w i * p i x) E T = ∑ i : ι, w i * weightedSeededTestProb (p i) E T
theorem
Algebraic.Cutwidth.Extractor.Internal.flatStrong_of_weighted
{α : Type u_1}
{Seed : Type u_2}
{Ω : Type u_3}
[Fintype α]
[Fintype Seed]
[Fintype Ω]
{E : α → Seed → Ω}
{K : ℕ}
{ε : ℝ}
(extract : WeightedStrongSeededExtractor E K ε)
: