Rademacher contraction from squared to linear coordinate sums
ProvedTalagrandCore.contraction_mainconcentration-inequalitiesempirical-processesprobabilitytalagrand
For coefficients bounded by one in absolute value, replacing selected signed quadratic terms by twice the corresponding signed linear terms can only increase the expected finite supremum. In particular,
This is the induction form of the Rademacher contraction principle needed for the variance-process bound.
Formalization Note Rademacher signs are encoded by Boolean coordinates with parameter .
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 contraction_main (u : ι → κ → ℝ) (hu : ∀ a x, |u a x| ≤ 1) (S : Finset κ) :
Ex (1/2 : ℝ) (fun s : κ → Bool =>
Finset.univ.sup' Finset.univ_nonempty (fun a : ι =>
(∑ x ∈ S, sg (s x) * u a x ^ 2) + ∑ x ∈ Sᶜ, sg (s x) * (2 * u a x))) ≤
Ex (1/2 : ℝ) (fun s : κ → Bool =>
Finset.univ.sup' Finset.univ_nonempty
(fun a : ι => ∑ x : κ, sg (s x) * (2 * u a x))) := 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.