AFTD Auto-Formalizing Theoretical Domains

Bonami.second_moment_decomp

theoremverifiedimported Lean

Second-moment decomposition over the last coordinate. Let $f : \{0,1\}^{n+1} \to \bbr$ be a Boolean function on $n+1$ variables. Define its average and half-difference over the last coordinate as the functions on $n$ variables \[ g(x) = \tfrac12\bigl(f(x,0) + f(x,1)\bigr), \qquad h(x) = \tfrac12\bigl(f(x,0) - f(x,1)\bigr). \] Then, with the expectation on the left taken under the uniform measure on $\{0,1\}^{n+1}$ and those on the right under the uniform measure on $\{0,1\}^n$, the second moment of $f$ decomposes as \[ \E\bigl[f^2\bigr] = \E\bigl[g^2\bigr] + \E\bigl[h^2\bigr]. \]

Statement

set_option autoImplicit true in
set_option relaxedAutoImplicit true in
open BooleanAnalysis in
lemma Bonami.second_moment_decomp {n : ℕ} (f : BooleanFunc (n + 1)) :
    expect (fun x => f x ^ 2) =
      expect (fun x => avgLast f x ^ 2) + expect (fun x => diffLast f x ^ 2)
Source
TCSlib, `Bonami.second_moment_decomp`
Lean proof
By Owen McGinty (from the file's git history), from github.com/Shilun-Allan-Li/tcslib at c8a02591b1e5 (Apache-2.0). Compiled here against leanprover/lean4:v4.33.1, Mathlib v4.33.1; written for leanprover/lean4:v4.25.0.
Verified
07 Oct 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 9
  • BooleanAnalysis.uniformWeightThe uniform probability measure on $\{0,1\}^n$ assigns weight $2^{-n}$ to each point.
  • BooleanAnalysis.expectThe expectation of $f$ under the uniform measure: \[ \E[f] \;=\; 2^{-n}\sum_{x\in\{0,1\}^n} f(x). \]
  • Bonami.diffLastThe half-difference $\mathrm{diffLast}\,f$ of $f$ over its last coordinate is the function on $n$ variables given by…
  • BooleanAnalysis.BoolCubeThe Boolean hypercube of dimension $n$ is $\{0,1\}^n$, represented as the function type $\mathrm{Fin}\,n \to \mathrm{Bool}$.
  • Bonami.avgLastThe average $\mathrm{avgLast}\,f$ of $f$ over its last coordinate is the function on $n$ variables given by…
  • Bonami.restrictLastGiven a Boolean function $f$ on $n+1$ variables and a bit $b$, the restriction $\mathrm{restrictLast}\,f\,b$ is the function on $n$…
  • Bonami.sum_boolCube_succSplitting a sum over the hypercube on its last coordinate. Let $n$ be a natural number and let $\varphi \colon \{0,1\}^{n+1} \to \bbr$ be…
  • Bonami.restrictLast_true_eqRecovering the true-restriction from average and half-difference. Let $f : \{0,1\}^{n+1} \to \bbr$ be a Boolean function on $n+1$…
  • BooleanAnalysis.BooleanFuncA Boolean function of arity $n$ is a function $f : \{0,1\}^n \to \bbr$.
Used by 1
  • Bonami.bonami_expectBonami inequality in expectation form. Let $f : \{0,1\}^n \to \bbr$ be a Boolean function of degree at most $k$, meaning that every…

Lean source view module on GitHub

set_option autoImplicit true inset_option relaxedAutoImplicit true inopen BooleanAnalysis in/-- Decomposes a second moment into the average and half-difference components. **Source:** [OD14, Cor. 9.6 (proof), Eq. (9.2)]. -/lemma Bonami.second_moment_decomp {n : ℕ} (f : BooleanFunc (n + 1)) :    expect (fun x => f x ^ 2) =      expect (fun x => avgLast f x ^ 2) + expect (fun x => diffLast f x ^ 2) := by  unfold expect  ring_nf  have h_expand :      ∑ x : BoolCube (n + 1), f x ^ 2 =        ∑ x : BoolCube n, (avgLast f x + diffLast f x) ^ 2 +        ∑ x : BoolCube n, (avgLast f x - diffLast f x) ^ 2 := by    convert sum_boolCube_succ (fun x => f x ^ 2) using 1    congr! 2    · exact congr_arg (· ^ 2) (by        unfold avgLast diffLast restrictLast        ring)    · simpa only [restrictLast] using        congrArg (fun y : ℝ => y ^ 2) (restrictLast_true_eq f _).symm  simp_all +decide [add_sq, sub_sq, Finset.sum_add_distrib, Finset.mul_sum _ _ _]  ring_nf  norm_num [← Finset.mul_sum _ _ _, ← Finset.sum_mul, uniformWeight]  ring
See it in the graphDiscuss this resultChallenge it