Tensorized entropy inequality on a finite Bernoulli cube
ProvedTalagrandCore.entropy_tensor_psiLet be the product Bernoulli measure on a finite Boolean cube, let be a real-valued function, and for each coordinate let be independent of that coordinate. With , the entropy of satisfies
This is the finite-product entropy tensorization interface used by both tails of the Talagrand argument.
Formalization Note Expectations are exact weighted sums on the Boolean cube, and coordinate independence is expressed by FiberConst.
Preamble
import Definitions.Def_talagrand_finite_bool_core open MeasureTheory open scoped Classical BigOperators
Formal statement
namespace TalagrandCore
variable {κ ι : Type} [DecidableEq κ] [Fintype κ]
variable [Fintype ι] [Nonempty ι]
theorem entropy_tensor_psi (p : NNReal) (hp : p ≤ 1)
(Z : (κ → Bool) → ℝ) (lam : ℝ)
(c : κ → (κ → Bool) → ℝ) (hc : ∀ x, FiberConst x (c x)) :
lam * Ex (p : ℝ) (fun ω => Z ω * Real.exp (lam * Z ω)) -
Ex (p : ℝ) (fun ω => Real.exp (lam * Z ω)) *
Real.log (Ex (p : ℝ) (fun ω => Real.exp (lam * Z ω))) ≤
∑ x : κ, Ex (p : ℝ)
(fun ω => Real.exp (lam * Z ω) * psi (lam * (Z ω - c x ω))) := by sorry
end TalagrandCoreSource
Michel Ledoux, On Talagrand’s deviation inequalities for product measures, ESAIM Probability and Statistics 1 (1996), Theorems 2.4–2.5, pp. 63–87. Formal Lean proof extracted from Prove2Me accepted submission bf106d23-ff42-48f5-a837-a8528101c849 by tianyipeng.