Lemma 6.6.2 — Matrix Bernstein mgf and cgf bounds
ProvedTroppMatrixConcentration.bernstein_mgf_cgfLet be a measurable Hermitian random matrix on a probability space, with , finite second moment, , and almost surely, where . For , put . Then, in the semidefinite order,
Only a one-sided eigenvalue bound is assumed. Finite second moment makes the expectation of well-defined under the regularity convention of Section 2.2.1.
import Definitions.Def_TroppMatrixConcentration_probability import Definitions.Def_TroppMatrixConcentration_dilation import Mathlib.Analysis.Convex.Function open MeasureTheory ProbabilityTheory open scoped Matrix.Norms.L2Operator
namespace TroppMatrixConcentration
theorem bernstein_mgf_cgf {Ω : Type*} [MeasurableSpace Ω]
(μ : Measure Ω) [IsProbabilityMeasure μ] {d : ℕ} [NeZero d]
(X : Ω → Matrix (Fin d) (Fin d) ℂ) (L θ : ℝ)
(hL : 0 < L) (hθ : 0 < θ) (hθL : θ < 3 / L)
(hMeas : Measurable X) (hHerm : ∀ᵐ ω ∂μ, (X ω).IsHermitian)
(hL2 : MemLp X 2 μ) (hMean : (∫ ω, X ω ∂μ) = 0)
(hBound : ∀ᵐ ω ∂μ, lambdaMax (X ω) ≤ L) :
loewnerLE (∫ ω, matrixExp (θ • X ω) ∂μ)
(matrixExp (((θ ^ 2 / 2) / (1 - θ * L / 3)) • (∫ ω, X ω ^ 2 ∂μ))) ∧
loewnerLE (matrixLog (∫ ω, matrixExp (θ • X ω) ∂μ))
(((θ ^ 2 / 2) / (1 - θ * L / 3)) • (∫ ω, X ω ^ 2 ∂μ)) := by sorry
end TroppMatrixConcentration
Read-back
What the Lean code literally says, in plain math · gpt-6
For every measurable space , every probability measure on , every positive integer , every measurable map (with rows and columns indexed by ), and real numbers satisfying and , suppose that is Hermitian for -almost every , that belongs to for the matrix operator norm (equivalently here, , together with the almost-everywhere strong measurability included in membership), that , and that almost everywhere. Here is defined as the real supremum of the real spectrum ; on Hermitian matrices of this positive dimension it is the largest eigenvalue, whereas the total real supremum is for an empty spectrum. Write , where the square is matrix multiplication, , and . The conclusion is the conjunction and . The notation means that is Hermitian and for every ; is the matrix exponential defined by the norm-convergent exponential series; and is the real continuous functional calculus of the total real logarithm, which for a Hermitian matrix applies to each nonzero eigenvalue and sends a zero eigenvalue to , and is for a non-Hermitian matrix. All matrix integrals are Bochner integrals, with the total-integral convention assigning to a nonintegrable function. The bound on is an almost-everywhere upper bound on its largest eigenvalue, with no lower eigenvalue bound or operator-norm bound imposed; exceptional null sets are allowed. The dimension , the values and , and the endpoint are excluded, so the denominator defining is strictly positive; zero and zero second moment are not excluded.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.