Display (2.4) — cosh(A) ≼ e^{A²/2} for every self-adjoint matrix A
ProvedMatrixTail.Gaussian.display2_4matrix-analysismatrix-inequalityp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
For every self-adjoint complex matrix ,
where and are defined through the eigen-decomposition of and is the semidefinite order.
This is the matrix version of the scalar inequality , transferred to matrices by the spectral calculus. It is what bounds the mgf of a Rademacher-modulated matrix in Lemma 4.3.
Formalization Note Both matrix functions are Mathlib's continuous functional calculus cfc; is the scalar multiple .
Preamble
import Mathlib import Definitions.Def_MatrixTail_Gaussian_Model open MeasureTheory ProbabilityTheory open scoped MatrixOrder ComplexOrder ENNReal
Formal statement
namespace MatrixTail.Gaussian
/-- **Display (2.4)**, Tropp, arXiv:1004.4389v7, §2.4, p. 8: for each self-adjoint matrix `A`,
`cosh(A) ≼ e^{A²/2}`. (Cited in the proof of Lemma 4.3, p. 15: "the second relation is (2.4)".)
Formalization Note. Complex Hermitian `d × d` matrices; `cosh(A)` and `e^{A²/2}` are Mathlib's `cfc` of
`Real.cosh` and `Real.exp`; `≼` is the Loewner order `≤` under `MatrixOrder`; `A²/2` is written
`(1/2 : ℝ) • A ^ 2`. -/
theorem display2_4 {d : ℕ} (A : Matrix (Fin d) (Fin d) ℂ) (hA : A.IsHermitian) :
cfc Real.cosh A ≤ MatrixTail.Master.mexp ((1 / 2 : ℝ) • A ^ 2) := by sorry
end MatrixTail.Gaussian
Source
Tropp, User-Friendly Tail Bounds for Sums of Random Matrices, arXiv:1004.4389v7, p. 8, §2.4, display (2.4)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.