theorem
BooleanFourier.majority_is_stablest_theorem01
(ρ : ℝ)
(hρ₀ : 0 < ρ)
(hρ₁ : ρ < 1)
(ε : ℝ)
(hε : 0 < ε)
:
∃ δ > 0,
∀ (n : ℕ) (f : (Fin n → Bool) → ℝ),
HasBoundedRange01 f →
boolExpectation f = 1 / 2 →
(∀ (i : Fin n), influenceReal f i ≤ δ) → noiseStability ρ f ≤ halfspaceNoiseStability01 ρ (1 / 2) + ε
theorem
BooleanFourier.majority_is_stablest_general
(ρ : ℝ)
(hρ₀ : 0 ≤ ρ)
(hρ₁ : ρ < 1)
(ε : ℝ)
(hε : 0 < ε)
:
∃ δ > 0,
∀ (n : ℕ) (f : (Fin n → Bool) → ℝ),
HasBoundedRange f → maxInfluence f ≤ δ → noiseStability ρ f ≤ halfspaceNoiseStability ρ (boolExpectation f) + ε