Two-sided sub-Gaussian tail bound
Provedsubgaussian_abs_ge_leRole. Atomic two-sided sub-Gaussian tail bound — the base case of the concentration toolkit (single variable), reused in Hoeffding/Azuma/McDiarmid and the empirical-process concentration of exact matrix completion.
Claim. If a real random variable on a probability space is sub-Gaussian with parameter (HasSubgaussianMGF X c), then for every ,
Proof. The one-sided Chernoff bound HasSubgaussianMGF.measure_ge_le gives . The same applied to (HasSubgaussianMGF.neg, same parameter) gives the left tail. Then with measureReal_mono and the union bound measureReal_union_le.
Source. Standard two-sided sub-Gaussian concentration. R. Vershynin, High-Dimensional Probability, CUP 2018, Prop. 2.5.2 (sub-Gaussian tail); S. Boucheron, G. Lugosi, P. Massart, Concentration Inequalities, OUP 2013, §2.3. Derived from Mathlib's one-sided HasSubgaussianMGF.measure_ge_le by the union bound.
import Mathlib.Probability.Moments.SubGaussian import Mathlib.MeasureTheory.Measure.Real open MeasureTheory ProbabilityTheory Real open scoped NNReal ENNReal
theorem subgaussian_abs_ge_le
{Ω : Type*} {mΩ : MeasurableSpace Ω} {μ : Measure Ω} [IsProbabilityMeasure μ]
{X : Ω → ℝ} {c : ℝ≥0} (h : HasSubgaussianMGF X c μ)
{ε : ℝ} (hε : 0 ≤ ε) :
μ.real {ω | ε ≤ |X ω|} ≤ 2 * Real.exp (-ε ^ 2 / (2 * c)) := by sorry