AFTD Auto-Formalizing Theoretical Domains

has_log_concave_density

def

A probability measure μ on ℝ has a log-concave density if it is given by μ = volume.withDensity f for some measurable nonnegative function f : ℝ → ℝ with ∫ f = 1 such that f((1 - t)x + ty) ≥ (fx)^(1 - t) * (fy)^t for all x, y ∈ ℝ and t ∈ [0, 1].

Statement

def has_log_concave_density (μ : MeasureTheory.Measure ℝ) : Prop
Defined in
Saumard & Wellner 2014, "Log-concavity and strong log-concavity: a review", Section 2
Verified
15 Nov 2026
Axioms
none listed

Read back from the Lean

The property has_log_concave_density μ asserts that there exists a function $f : \mathbb{R} \to \mathbb{R}$ such that: 1. $f$ is Borel-measurable; 2. $f(x) \ge 0$ for all $x \in \mathbb{R}$; 3. The integral of $f$ over $\mathbb{R}$ with respect to the Lebesgue measure (volume) is $1$; 4. For all $x, y \in \mathbb{R}$ and all $t \in \mathbb{R}$ satisfying $0 \le t \le 1$, $$(f(x))^{1 - t} \cdot (f(y))^t \le f((1 - t)x + ty)$$ (where powers are real powers); and 5. $\mu$ equals the measure on $\mathbb{R}$ obtained by integrating against Lebesgue measure with density function $x \mapsto \mathrm{ofReal}(f(x))$.

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

/-- A probability measure on the real line has a log-concave density if it is absolutely continuous with respect to Lebesgue measure with a nonnegative log-concave density integrating to 1. -/
def has_log_concave_density (μ : MeasureTheory.Measure ℝ) : Prop := ∃ f : ℝ → ℝ, Measurable f ∧ (∀ x, 0 ≤ f x) ∧ (∫ x, f x) = 1 ∧
    (∀ (x y : ℝ) (t : ℝ), 0 ≤ t → t ≤ 1 →
      (f x) ^ (1 - t) * (f y) ^ t ≤ f ((1 - t) * x + t * y)) ∧
    μ = MeasureTheory.volume.withDensity (fun x ↦ ENNReal.ofReal (f x))
See it in the graphDiscuss this resultChallenge it