ProbabilityTheory.HasBernsteinMGF.sum_of_iIndepFun
Sub-exponential MGF bound for independent sums. Let $(X_i)_{i \in \iota}$ be a family of measurable real random variables on a probability space $(\Omega, \mu)$ that are mutually independent, and let $s$ be a finite set of indices. Fix a radius $t_{\max}$, and suppose that for each $i \in s$ there is a constant $c_i$ such that $e^{tX_i}$ is $\mu$-integrable and $\mathbf{E}_\mu[e^{tX_i}] \le e^{c_i t^2}$ for every $t$ with $|t| \le t_{\max}$. Then the sum $\sum_{i \in s} X_i$ satisfies the same two-parameter sub-exponential bound with radius $t_{\max}$ and constant $\sum_{i \in s} c_i$: for every $t$ with $|t| \le t_{\max}$, the moment $e^{t \sum_{i \in s} X_i}$ is $\mu$-integrable and $\mathbf{E}_\mu\!\left[e^{t \sum_{i \in s} X_i}\right] \le e^{(\sum_{i \in s} c_i)\, t^2}$.
Statement
set_option relaxedAutoImplicit false in set_option autoImplicit false in open MeasureTheory ProbabilityTheory Real in variable {Ω : Type*} {mΩ : MeasurableSpace Ω} in variable {X : Ω → ℝ} {μ : Measure Ω} {c tmax : ℝ} in lemma ProbabilityTheory.HasBernsteinMGF.sum_of_iIndepFun {ι : Type*} {X : ι → Ω → ℝ} {μ : Measure Ω} [IsProbabilityMeasure μ] (h_indep : iIndepFun X μ) (h_meas : ∀ i, Measurable (X i)) {c : ι → ℝ} {tmax : ℝ} {s : Finset ι} (h_bern : ∀ i ∈ s, HasBernsteinMGF (X i) μ (c i) tmax) : HasBernsteinMGF (fun ω => ∑ i ∈ s, X i ω) μ (∑ i ∈ s, c i) tmax
- Source
- TCSlib, `ProbabilityTheory.HasBernsteinMGF.sum_of_iIndepFun`
- 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.sum_of_iIndepFun← ProbabilityTheory.HasBernsteinMGF.sum_of_iIndepFun (TCSlib/LearningTheory/JohnsonLindenstrauss/Bernstein.lean:88): verbatim
- Verified
- 07 Oct 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 1
- ProbabilityTheory.HasBernsteinMGF$\texttt{ProbabilityTheory.HasBernsteinMGF}\ X\ \mu\ c\ t_{\max}$ is a predicate on a random variable $X : \Omega \to \mathbb{R}$, a…
- Used by 1
- 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
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
set_option relaxedAutoImplicit false inset_option autoImplicit false inopen MeasureTheory ProbabilityTheory Real invariable {Ω : Type*} {mΩ : MeasurableSpace Ω} invariable {X : Ω → ℝ} {μ : Measure Ω} {c tmax : ℝ} in/-- **Closure under independent sums.** If `X i` (`i` ranging over a finite set `s`) is a family of mutually independent measurable real random variables on a probability space, and each `X i` has a Bernstein-type MGF with parameters `(c i, tmax)` for a common radius `tmax`, then the sum `∑ i ∈ s, X i` has a Bernstein-type MGF with parameters `(∑ i ∈ s, c i, tmax)`. This is the MGF-of-a-sum step in [Ver18, proof of Thm 2.8.1]. Only independence is assumed, not identical distribution. **Proof sketch.** Step 1: integrability of `exp(t·∑ X i)` for `|t| ≤ tmax` follows from the per-summand integrability via independence (Mathlib's `iIndepFun.integrable_exp_mul_sum`). Step 2: by independence the moment generating function of the sum at `t` is the product of the individual ones; bound each factor by `exp(c i · t²)` and collect the product of exponentials into `exp((∑ c i)·t²)`. -/lemma ProbabilityTheory.HasBernsteinMGF.sum_of_iIndepFun {ι : Type*} {X : ι → Ω → ℝ} {μ : Measure Ω} [IsProbabilityMeasure μ] (h_indep : iIndepFun X μ) (h_meas : ∀ i, Measurable (X i)) {c : ι → ℝ} {tmax : ℝ} {s : Finset ι} (h_bern : ∀ i ∈ s, HasBernsteinMGF (X i) μ (c i) tmax) : HasBernsteinMGF (fun ω => ∑ i ∈ s, X i ω) μ (∑ i ∈ s, c i) tmax := by refine ⟨?_, ?_⟩ · -- Step 1: integrability of `exp(t · ∑ X_i)` on `|t| ≤ tmax` from per-element -- integrability. intro t ht have h1 : Integrable (fun ω => Real.exp (t * (∑ i ∈ s, X i) ω)) μ := by refine h_indep.integrable_exp_mul_sum h_meas (fun i hi => ?_) exact (h_bern i hi).integrable t ht convert h1 using 1 funext ω rw [Finset.sum_apply] · -- Step 2: MGF bound: -- `mgf (∑ X_i) μ t = ∏ mgf (X_i) μ t ≤ ∏ exp(c_i t²) = exp(∑ c_i · t²)`. intro t ht have h_pi_eq : (fun ω => ∑ i ∈ s, X i ω) = ∑ i ∈ s, X i := by funext ω; rw [Finset.sum_apply] rw [h_pi_eq, h_indep.mgf_sum h_meas] calc ∏ i ∈ s, mgf (X i) μ t ≤ ∏ i ∈ s, Real.exp (c i * t ^ 2) := by apply Finset.prod_le_prod · exact fun i _ => mgf_nonneg · exact fun i hi => (h_bern i hi).mgf_le t ht _ = Real.exp ((∑ i ∈ s, c i) * t ^ 2) := by rw [← Real.exp_sum, Finset.sum_mul]