Lemma 4.6.3 — Rademacher matrix mgf and cgf bounds
ProvedTroppMatrixConcentration.ch4_rademacher_mgf_cgfLet be a fixed Hermitian complex matrix, , and let be a measurable Rademacher random variable on a probability space, taking and with probabilities one half each. For every real ,
Both relations are in positive-semidefinite order. The actual two-point distribution is required, not merely its mean and variance. No sign restriction is placed on .
import Definitions.Def_TroppMatrixConcentration_ch4_scalar_laws import Definitions.Def_TroppMatrixConcentration_dilation open MeasureTheory ProbabilityTheory open scoped Matrix.Norms.L2Operator
namespace TroppMatrixConcentration
theorem ch4_rademacher_mgf_cgf {Ω : Type*} [MeasurableSpace Ω]
(μ : Measure Ω) [IsProbabilityMeasure μ] {d : ℕ} [NeZero d]
(A : Matrix (Fin d) (Fin d) ℂ) (hA : A.IsHermitian)
(g : Ω → ℝ) (hMeas : Measurable g) (hLaw : rademacherLaw μ g) (θ : ℝ) :
loewnerLE (∫ ω, matrixExp ((θ * g ω) • A) ∂μ) (matrixExp ((θ ^ 2 / 2) • A ^ 2)) ∧
loewnerLE (matrixLog (∫ ω, matrixExp ((θ * g ω) • A) ∂μ)) ((θ ^ 2 / 2) • A ^ 2) := by sorry
end TroppMatrixConcentration
Read-back
What the Lean code literally says, in plain math · gpt-6
For every type equipped with a measurable space, every probability measure on that space (so ), every natural number , every complex matrix indexed by satisfying , every measurable function whose pushforward measure is , and every real number , put and . Then both and hold. Here is the unit point mass at , the integral is the matrix-valued Bochner integral, scalar multiplication uses real scalars acting on complex entries, is the matrix product, and is the matrix exponential. The notation means exactly that is Hermitian and is a nonnegative real number for every . The function is the real continuous functional calculus applied to the total real logarithm: on a Hermitian matrix with a unitary diagonalization , it gives , where for and ; the real functional calculus returns zero on non-Hermitian matrices (and in general when its scalar function is not continuous on the real spectrum). The law of is an equality of measures, so its values outside are permitted on a measurable null set; it gives each of the events and measure . The zero dimension is excluded, whereas , , singular Hermitian , and all negative, zero, and positive are included; if or , then and , and both displayed inequalities are equalities. No separate integrability or positive-definiteness hypothesis is imposed; under the stated law the integrand takes two matrix values almost everywhere and is positive definite. An empty cannot carry the required probability measure.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.