AFTD Auto-Formalizing Theoretical Domains

mutual_f_information_subadditive_full_support_zero

theoremverified

Sanity: the zero generator f = 0 is subadditive on full-support distributions (every mutual f-information is 0).

Statement

theorem mutual_f_information_subadditive_full_support_zero :
    mutual_f_information_subadditive_full_support (fun _ => 0)
Source
sanity lemma for mutual_f_information_subadditive_full_support
Verified
23 Oct 2026
Axioms
Classical.choiceQuot.soundpropext
Built on 2
  • mutual_f_information_subadditive_full_supportSubadditivity of mutual f-information on full-support distributions: for all finite X, Y, Z, every p_X > 0 and channels p_{Y|X} > 0,…
  • mutual_f_informationFor a joint probability mass function p_{XY} on finite sets, the mutual f-information is Ĩ_f(X; Y) = D_f(p_{XY} ‖ p_X p_Y) = Σ_{x,y}…

The proof, explained

To show that the zero generator f = 0 is subadditive on full-support distributions, we introduce the arbitrary finite spaces and distributions. Unfolding the definition of mutual f-information reveals that every summand contains a factor of f(·) = 0, causing all three mutual information terms to vanish. The subadditivity inequality reduces to 0 ≤ 0 + 0, which holds trivially.

  1. Let X, Y, and Z be finite types equipped with a prior distribution pX and transition kernels PY and PZ having full support and summing to 1.

    We introduce the arbitrary finite types X, Y, and Z, the prior distribution pX on X, the transition kernels PY : X → Y → ℝ and PZ : X → Z → ℝ, and the six hypotheses asserting that pX, PY, and PZ are strictly positive everywhere and sum to 1.

    line 3

      intro X Y Z _ _ _ pX PY PZ _ _ _ _ _ _
  2. The subadditivity inequality holds for the zero generator function. automation

    Unfolding the definition of mutual f-information, each summand contains a factor of f evaluated at a likelihood ratio. Since f is identically zero (fun _ => 0), every summand is zero, so all three mutual f-information quantities evaluate to 0. The inequality simplifies to 0 ≤ 0 + 0, which holds trivially.

    line 4

    uses mutual_f_information

      simp [mutual_f_information]
mutual_f_information
The mutual f-information of a joint distribution p on X × Y with generator f : ℝ → ℝ, defined as ∑ x, ∑ y, (marginal_X x) * (marginal_Y y) * f(p x y / ((marginal_X x) * (marginal_Y y))).

Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.

Lean source view module on GitHub

theorem mutual_f_information_subadditive_full_support_zero :    mutual_f_information_subadditive_full_support (fun _ => 0) := by  intro X Y Z _ _ _ pX PY PZ _ _ _ _ _ _  simp [mutual_f_information]
See it in the graphDiscuss this resultChallenge it