noise_operator
The noise operator T_rho f on the Boolean cube is the conditional expectation E[f(X) | Y = y] where each bit of X is flipped independently with probability (1 - rho) / 2.
Statement
open scoped BigOperators in noncomputable def noise_operator (n : ℕ) (ρ : ℝ) (f : (Fin n → Bool) → ℝ) (y : Fin n → Bool) : ℝ
- Defined in
- scout/2610-10618-the-hellinger-conjecture-for-antipodal-boolean-functions.md:162
- Verified
- 01 Nov 2026
- Axioms
- none listed
- Used by 2
- hellinger_ZThe Hellinger affinity Z_rho(f) of a Boolean function f is the expected value over uniform Y of sqrt(1 - (T_rho f(Y))^2).
- antipodal_hellinger_signed_dictator_equalityFor every real number ρ with 0 < ρ < 1, every natural number n with 0 < n, and every signed dictator g : (Fin n → Bool) → ℝ satisfying…
Read back from the Lean
The declaration defines noise_operator, which takes a natural number n, a real number ρ, a function f : (Fin n → Bool) → ℝ, and a point y : Fin n → Bool, and returns the real number given by the sum over all x : Fin n → Bool of
(((1 + ρ) / 2) ^ (n - hammingDist x y) * ((1 - ρ) / 2) ^ (hammingDist x y)) * f x,
where hammingDist x y is the Hamming distance between x and y, n - hammingDist x y is natural number subtraction, and the powers are natural powers of real numbers.
Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.
Lean source view module on GitHub
open scoped BigOperators in /-- The Bonami-Beckner noise operator T_rho f(y) = E[f(X) | Y = y] for the binary symmetric channel. -/ noncomputable def noise_operator (n : ℕ) (ρ : ℝ) (f : (Fin n → Bool) → ℝ) (y : Fin n → Bool) : ℝ := ∑ x : Fin n → Bool, (((1 + ρ) / 2) ^ (n - hammingDist x y) * ((1 - ρ) / 2) ^ (hammingDist x y)) * f x