two_point_hoeffding_mgf_bound
Provedanalysisconcentrationprobability
Two-point Hoeffding MGF inequality (Hoeffding 1963; Boucheron-Lugosi-Massart, Concentration Inequalities, OUP 2013, Lemma 2.2). For a real number and reals , the two-point convex combination of exponentials is bounded by the exponential of the mean plus a Gaussian-type term: . Equivalently, a two-point random variable taking value with probability and with probability has a sub-Gaussian MGF with variance proxy .
Preamble
import Mathlib.Analysis.SpecialFunctions.Log.Deriv import Mathlib.Analysis.SpecialFunctions.Exp import Mathlib.Analysis.Calculus.MeanValue open scoped BigOperators
Formal statement
theorem two_point_hoeffding_mgf_bound (p a b : ℝ) (hp0 : 0 ≤ p) (hp1 : p ≤ 1) :
p * Real.exp a + (1 - p) * Real.exp b ≤
Real.exp (p * a + (1 - p) * b + (a - b) ^ 2 / 8) := by sorrySource
Hoeffding, W. (1963), Probability inequalities for sums of bounded random variables, JASA 58(301):13-30; Boucheron, Lugosi, Massart, Concentration Inequalities, OUP 2013, Lemma 2.2.
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.