Statement (informal)
The governed claim has no exact TeX statement, so this section shows its informal summary. The Lean wrapper below checks a narrower declaration.
This page traces one bounded independent finite-sum Hoeffding claim from the informal theorem statement to the checked Lean declaration, then shows which diagnostics test the assumptions. It does not claim that the whole concentration page or the finite-class uniform-convergence theorem is fully formalized.
The governed claim has no exact TeX statement, so this section shows its informal summary. The Lean wrapper below checks a narrower declaration.
theorem hoeffdingBoundedFiniteSumTail
{Ω ι : Type*} [MeasurableSpace Ω] {μ : Measure Ω} [IsProbabilityMeasure μ]
{X : ι → Ω → ℝ} (hIndep : iIndepFun X μ)
{a b : ι → ℝ} {s : Finset ι}
(hMeas : ∀ i ∈ s, AEMeasurable (X i) μ)
(hBound : ∀ i ∈ s, ∀ᵐ ω ∂μ, X i ω ∈ Set.Icc (a i) (b i))
{ε : ℝ} (hε : 0 ≤ ε) :
μ.real {ω | ε ≤ ∑ i ∈ s, (X i ω - μ[X i])}
≤ Real.exp
(-ε ^ 2 / (2 * (↑(∑ i ∈ s, ((‖b i - a i‖₊ / 2) ^ (2 : ℕ))) : ℝ)))The proof composes existing mathlib facts rather than reproving concentration theory from scratch. That is intentional: the checked artifact is the exact TheoremPath-facing declaration.
Hoeffding's Lemma
claim:subgaussian-random-variables::hoeffding-lemma
TheoremPath.Probability.Concentration.hoeffdingLemmaBoundedCenteredSubgaussianMGF
Finite Union Bound for Probability Measures
claim:kolmogorov-probability-axioms::probability-measure-finite-union-bound
TheoremPath.Probability.KolmogorovAxioms.probabilityMeasureFiniteUnionBound
Uniform Convergence for Finite Classes
claim:uniform-convergence::finite-class-uniform-convergence
TheoremPath.LearningTheory.UniformConvergence.finiteClassEpsilonRepresentativeHighProbabilityFromOneSidedTails