rademacher_expectation_power_mean
Provedinequalityjensenpower-meanprobabilityrademacher
Power-mean (Jensen) inequality for the uniform Rademacher sign average. The map is the expectation under the uniform probability measure on the discrete cube of sign assignments (each of the assignments has weight , and these weights sum to ). For any nonnegative integrand and exponents , the -th moment is dominated by the -th moment raised to the power :
Equivalently for on a probability space (nesting of norms / power-mean monotonicity), a consequence of Jensen's inequality applied to the convex map . This is the analytic step that lets one pass from an even-integer moment down to a real moment in the noncommutative Khintchine assembly.
Preamble
import Definitions.Def_matrix_completion_rademacher open MatrixCompletion open scoped BigOperators
Formal statement
theorem rademacher_expectation_power_mean
{n1 n2 : Nat} (r s : ℝ) (hr : 0 < r) (hrs : r ≤ s)
(F : Finset (Fin n1 × Fin n2) → ℝ) (hF : ∀ eps, 0 ≤ F eps) :
rademacherExpectation (fun eps => (F eps) ^ r)
≤ Real.rpow (rademacherExpectation (fun eps => (F eps) ^ s)) (r / s) := by sorrySource
Power-mean / nesting of L^p norms on a probability space (Jensen's inequality). Used in CR2009 (arXiv:0805.4471) section 6.1 to pass from the even-integer Buchholz moment to a general real moment q >= 2.