bernoulli_moment_from_two_sided_mgf
Provedconcentration-inequalitiesmatrix-completionprobability
Moment-from-MGF on the Bernoulli powerset measure (Cramér–Chernoff). Let be any real statistic of the observation set under the Bernoulli powerset measure with inclusion probability . If at some parameter both and , then for every real the -th absolute moment obeys
This is the discrete Cramér–Chernoff device that converts a two-sided moment generating function bound into a moment bound, with NO integration: pointwise (from the elementary maximum ), then take expectation using nonnegativity of the Bernoulli weights and the two MGF bounds.
Preamble
import Definitions.Def_matrix_completion_neumann import Mathlib.Analysis.SpecialFunctions.Pow.Real import Mathlib.Analysis.SpecialFunctions.Exp open MatrixCompletion open scoped BigOperators Classical
Formal statement
theorem bernoulli_moment_from_two_sided_mgf {n₁ n₂ : ℕ} (p : ℝ) (hp0 : 0 ≤ p) (hp1 : p ≤ 1)
(Z : Finset (Fin n₁ × Fin n₂) → ℝ) (lam M q : ℝ)
(hlam : 0 < lam) (hq : 0 < q)
(hMGFpos : bernoulliExpectation p (fun Omega => Real.exp (lam * Z Omega)) ≤ M)
(hMGFneg : bernoulliExpectation p (fun Omega => Real.exp (-(lam * Z Omega))) ≤ M) :
bernoulliExpectation p (fun Omega => |Z Omega| ^ q) ≤
2 * (q / (lam * Real.exp 1)) ^ q * M := by sorrySource
Cramér–Chernoff moment method; Boucheron–Lugosi–Massart, Concentration Inequalities, OUP 2013, Ch. 2.