Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 7.5.1 — Intrinsic dimension trace inequality

Proved
TroppMatrixConcentration.ch7_intrinsic_dimension

by tc · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

matrix-concentrationrandom-matrices

For every real function φ\varphiφ convex on [0,∞)[0,\infty)[0,∞) with φ(0)=0\varphi(0)=0φ(0)=0, and every positive semidefinite complex d×dd\times dd×d matrix AAA, d≥1d\ge1d≥1,

tr⁡φ(A)≤r(A)φ(∥A∥).\operatorname{tr}\varphi(A)\le r(A)\varphi(\|A\|).trφ(A)≤r(A)φ(∥A∥).

The function of the matrix is defined by spectral functional calculus. Zero matrices are included using the totalized convention r(0)=0r(0)=0r(0)=0, so both sides are zero. No nonnegativity assumption on φ\varphiφ is imposed.

Preamble
import Definitions.Def_TroppMatrixConcentration_ch7_intrinsic

open MeasureTheory ProbabilityTheory
open scoped Matrix.Norms.L2Operator ComplexOrder

Formal statement
namespace TroppMatrixConcentration

theorem ch7_intrinsic_dimension {d : ℕ} [NeZero d]
    (φ : ℝ → ℝ) (hConvex : ConvexOn ℝ (Set.Ici 0) φ) (hZero : φ 0 = 0)
    (A : Matrix (Fin d) (Fin d) ℂ) (hPSD : A.PosSemidef) :
    traceFunction φ A ≤ intrinsicDimension A * φ (spectralNorm A) := by sorry

end TroppMatrixConcentration
Source
Joel A. Tropp, An Introduction to Matrix Concentration Inequalities, arXiv:1501.01571v1 (7 January 2015); https://arxiv.org/abs/1501.01571v1; Lemma 7.5.1, printed pp. 112–113.
Read-back

What the Lean code literally says, in plain math · gpt-6

For every natural number d≠0d\ne 0d=0, every function φ:R→R\varphi:\mathbb R\to\mathbb Rφ:R→R that is convex on [0,∞)[0,\infty)[0,∞) and satisfies φ(0)=0\varphi(0)=0φ(0)=0, and every complex d×dd\times dd×d matrix AAA, with rows and columns indexed by {0,…,d−1}\{0,\ldots,d-1\}{0,…,d−1}, assume that AAA is positive semidefinite, meaning that A=A∗A=A^*A=A∗ and z∗Azz^*Azz∗Az is real and nonnegative for every z∈Cdz\in\mathbb C^dz∈Cd. Then Re⁡tr⁡(φ(A))≤(Re⁡tr⁡(A)/∥A∥2→2) φ(∥A∥2→2)\operatorname{Re}\operatorname{tr}(\varphi(A))\le (\operatorname{Re}\operatorname{tr}(A)/\|A\|_{2\to2})\,\varphi(\|A\|_{2\to2})Retr(φ(A))≤(Retr(A)/∥A∥2→2​)φ(∥A∥2→2​), where ∥A∥2→2\|A\|_{2\to2}∥A∥2→2​ is the operator norm induced by the Euclidean norm and φ(A)\varphi(A)φ(A) is the real continuous functional calculus of AAA: if A=Udiag⁡(λ1,…,λd)U∗A=U\operatorname{diag}(\lambda_1,\ldots,\lambda_d)U^*A=Udiag(λ1​,…,λd​)U∗ is a unitary diagonalization, then φ(A)=Udiag⁡(φ(λ1),…,φ(λd))U∗\varphi(A)=U\operatorname{diag}(\varphi(\lambda_1),\ldots,\varphi(\lambda_d))U^*φ(A)=Udiag(φ(λ1​),…,φ(λd​))U∗. Equivalently, the asserted inequality is ∑i=1dφ(λi)≤((∑i=1dλi)/max⁡iλi) φ(max⁡iλi)\sum_{i=1}^d\varphi(\lambda_i)\le ((\sum_{i=1}^d\lambda_i)/\max_i\lambda_i)\,\varphi(\max_i\lambda_i)∑i=1d​φ(λi​)≤((∑i=1d​λi​)/maxi​λi​)φ(maxi​λi​), with eigenvalues counted with multiplicity. Division is total, with division by zero defined to be zero; in particular, A=0A=0A=0 is allowed and the inequality then reads 0≤00\le 00≤0. The case d=0d=0d=0 is excluded, but singular nonzero matrices are allowed; no nonnegativity of φ\varphiφ or restriction on its values at negative arguments is assumed.

Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by tc · Oct 7, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me