Cramér–Chernoff optimisation:
ProvedCramerChernoff.gaussian_exponent_optimal_lower_boundCramér–Chernoff optimisation of the Gaussian-shaped MGF bound. For and all , the Chernoff exponent satisfies , with equality at the optimiser . Equivalently , by completing the square .
Role in the entropy method. This is the final Cramér–Chernoff closure step (Boucheron–Lugosi–Massart, Concentration Inequalities, §2.3 and the end of the proof of Theorem 5.3). Herbst's argument produces the Gaussian-shaped log-MGF bound ; the exponential Markov inequality (ProbabilityTheory.measure_ge_le_exp_mul_mgf in Mathlib) gives for every ; minimising the exponent with this lemma yields the sub-Gaussian tail . Together with the Herbst integration brick this closes the Gaussian half of the entropy method; the Bennett/Bousquet half replaces by the modified-LSI exponent (the remaining irreducible step).
import Mathlib.Analysis.SpecialFunctions.Exp import Mathlib.Analysis.SpecialFunctions.Log.Basic import Mathlib.Tactic open Real
theorem CramerChernoff.gaussian_exponent_optimal_lower_bound (C t lam : ℝ) (hC : 0 < C) :
- t ^ 2 / (4 * C) ≤ C * lam ^ 2 - lam * t := by sorry