The Chernoff bound
ProvedFamousTheorems.chernoff_boundThe Chernoff bound. Let be a real random variable on a finite measure space, , and suppose is integrable. Then for every real ,
Optimising over gives exponentially small tail bounds. This is the starting point of large deviations theory (Cramér's theorem) and of the concentration inequalities of Hoeffding and Bernstein type used throughout probability, statistics and the analysis of randomized algorithms.
Formalization note. Mathlib's ProbabilityTheory.measure_ge_le_exp_mul_mgf. mgf X μ t is the moment generating function and μ.real is the measure as a real number. The inequality is Markov's inequality applied to .
Preamble
import Mathlib
Formal statement
namespace FamousTheorems
open MeasureTheory ProbabilityTheory
theorem chernoff_bound {Ω : Type*} {m : MeasurableSpace Ω} {X : Ω → ℝ} {μ : Measure Ω} {t : ℝ} [IsFiniteMeasure μ]
(ε : ℝ) (ht : 0 ≤ t) (h_int : Integrable (fun ω => Real.exp (t * X ω)) μ) :
μ.real {ω | ε ≤ X ω} ≤ Real.exp (-t * ε) * mgf X μ t := by sorry
end FamousTheoremsSource
Listed in Mathlib's curated theorem manifests (docs/1000.yaml); formalized in Mathlib as `ProbabilityTheory.measure_ge_le_exp_mul_mgf`. Proof here reduces to that Mathlib result.