ProbabilityTheory.HasBernsteinMGF
structureimported Lean
$\texttt{ProbabilityTheory.HasBernsteinMGF}\ X\ \mu\ c\ t_{\max}$ is a predicate on a random variable $X : \Omega \to \mathbb{R}$, a measure $\mu$, and real parameters $c, t_{\max} \ge 0$. It asserts that for every $t$ with $|t| \le t_{\max}$, the exponential moment $e^{tX}$ is $\mu$-integrable, and the moment generating function satisfies $\mathbf{E}_\mu[e^{tX}] \le e^{c t^2}$. The pair $(c, t_{\max})$ parametrizes a ``two-parameter sub-exponential'' family: $c$ controls the quadratic growth near zero and $t_{\max}$ bounds the radius of validity.
Statement
set_option relaxedAutoImplicit false in set_option autoImplicit false in open MeasureTheory ProbabilityTheory Real in variable {Ω : Type*} {mΩ : MeasurableSpace Ω} in structure ProbabilityTheory.HasBernsteinMGF (X : Ω → ℝ) (μ : Measure Ω) (c tmax : ℝ) : Prop
- Defined in
- TCSlib, `ProbabilityTheory.HasBernsteinMGF`
- Lean proof
- By Ganesh Sankar, from github.com/Shilun-Allan-Li/tcslib at
c8a02591b1e5(Copyright (c) 2026 Ganesh Sankar. All rights reserved, Apache-2.0). Compiled here against leanprover/lean4:v4.33.1, Mathlib v4.33.1; written for leanprover/lean4:v4.25.0.ProbabilityTheory.HasBernsteinMGF← ProbabilityTheory.HasBernsteinMGF (TCSlib/LearningTheory/JohnsonLindenstrauss/Bernstein.lean:68): verbatim
- Verified
- 07 Oct 2026
- Axioms
- none listed
- Used by 6
- ProbabilityTheory.HasBernsteinMGF.measure_le_leLower-tail Bernstein bound. Let $X : \Omega \to \bbr$ be a random variable on a probability space carrying a finite measure $\mu$, and…
- hasBernsteinMGF_centered_chi_squaredBernstein MGF bound for a centered chi-squared summand. Fix an integer $k \ge 1$, and let $Y$ be a real-valued random variable on a…
- ProbabilityTheory.HasBernsteinMGF.measure_ge_leUpper-tail Bernstein bound. Let $X : \Omega \to \bbr$ be a random variable on a finite measure space $(\Omega, \mu)$, and suppose $X$…
- ProbabilityTheory.HasBernsteinMGF.sum_of_iIndepFunSub-exponential MGF bound for independent sums. Let $(X_i)_{i \in \iota}$ be a family of measurable real random variables on a probability…
- ProbabilityTheory.HasBernsteinMGF.measure_abs_gt_leTwo-sided Bernstein tail bound. Let $X : \Omega \to \bbr$ be a random variable on a probability space $(\Omega, \mu)$, and let…
- chi_squared_tailChi-squared tail bound for independent $N(0,1/k)$ variables. Let $\mu$ be a probability measure on $\Omega$, let $k \ge 1$, and let…
Lean source view module on GitHub
set_option relaxedAutoImplicit false in set_option autoImplicit false in open MeasureTheory ProbabilityTheory Real in variable {Ω : Type*} {mΩ : MeasurableSpace Ω} in /-- A real random variable `X` on the measure space `(Ω, μ)` has a *Bernstein-type moment generating function* with parameters `(c, tmax)` if, for every real `t` with `|t| ≤ tmax`, the function `exp(t·X)` is `μ`-integrable and its mean — the moment generating function of `X` at `t` — is at most `exp(c·t²)`. This is the sub-exponential MGF condition of [Ver18, Prop 2.7.1(e)] (`E exp(λX) ≤ exp(K²λ²)` for `|λ| ≤ 1/K`), with `c = K²` and `tmax = 1/K`. Deviation from the source: the two parameters `c` and `tmax` are explicit and independent of each other (the source ties both to a single sub-exponential norm `K`), and the integrability of `exp(t·X)` on the range is recorded as a field rather than derived. `HasBernsteinMGF X μ c tmax` says the MGF of `X` under `μ` is bounded by the Gaussian envelope `exp(c t²)` on `|t| ≤ tmax`, with `exp(t · X)` integrable on the same range. The pair `(c, tmax)` parametrizes a "two-parameter sub-exponential" family: `c` controls the local quadratic behavior near zero; `tmax` bounds the radius of the MGF's domain (which may be finite for genuinely sub-exponential — but not sub-Gaussian — RVs). For sub-Gaussian `X` (with parameter `σ²`) one can take `tmax = ∞` formally; this predicate is more useful when `tmax < ∞`. -/ structure ProbabilityTheory.HasBernsteinMGF (X : Ω → ℝ) (μ : Measure Ω) (c tmax : ℝ) : Prop where integrable : ∀ t : ℝ, |t| ≤ tmax → Integrable (fun ω => Real.exp (t * X ω)) μ mgf_le : ∀ t : ℝ, |t| ≤ tmax → mgf X μ t ≤ Real.exp (c * t ^ 2)