Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Corollary 7.3.2 — Intrinsic Bernstein expectation bound

Proved
TroppMatrixConcentration.ch7_intrinsic_bernstein_expectation

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

matrix-concentrationrandom-matrices

There exists a universal real constant C>0C>0C>0 such that the following holds for every probability space and every finite independent family of measurable complex m×nm\times nm×n random matrices SkS_kSk​, m,n≥1m,n\ge1m,n≥1, with mean zero and ∥Sk∥≤L\|S_k\|\le L∥Sk​∥≤L almost surely for L≥0L\ge0L≥0. Set Z=∑kSkZ=\sum_kS_kZ=∑k​Sk​. For semidefinite upper bounds V1⪰EZZ∗V_1\succeq\mathbb EZZ^*V1​⪰EZZ∗ and V2⪰EZ∗ZV_2\succeq\mathbb EZ^*ZV2​⪰EZ∗Z, at least one nonzero, set r=r(diag⁡(V1,V2))r=r(\operatorname{diag}(V_1,V_2))r=r(diag(V1​,V2​)) and v=max⁡{∥V1∥,∥V2∥}v=\max\{\|V_1\|,\|V_2\|\}v=max{∥V1​∥,∥V2​∥}. Then

E∥Z∥≤C(vlog⁡(1+r)+Llog⁡(1+r)).\mathbb E\|Z\|\le C\left(\sqrt{v\log(1+r)}+L\log(1+r)\right).E∥Z∥≤C(vlog(1+r)​+Llog(1+r)).

The constant is quantified before all probability spaces, dimensions, summands, bounds, and proxies, so it cannot depend on any of them. No explicit numerical constant is specified by the corollary.

Preamble
import Definitions.Def_TroppMatrixConcentration_ch7_intrinsic

open MeasureTheory ProbabilityTheory
open scoped Matrix.Norms.L2Operator ComplexOrder

Formal statement
namespace TroppMatrixConcentration

theorem ch7_intrinsic_bernstein_expectation :
    ∃ C : ℝ, 0 < C ∧
    ∀ {Ω : Type*} [MeasurableSpace Ω] (μ : Measure Ω) [IsProbabilityMeasure μ]
      {m n N : ℕ} [NeZero m] [NeZero n]
      (S : Fin N → Ω → Matrix (Fin m) (Fin n) ℂ) (L : ℝ), 0 ≤ L →
    ∀ (V₁ : Matrix (Fin m) (Fin m) ℂ) (V₂ : Matrix (Fin n) (Fin n) ℂ),
      (V₁ ≠ 0 ∨ V₂ ≠ 0) →
      (∀ k, Measurable (S k)) → iIndepFun S μ →
      (∀ k, (∫ ω, S k ω ∂μ) = 0) →
      (∀ k, ∀ᵐ ω ∂μ, spectralNorm (S k ω) ≤ L) →
      loewnerLE (∫ ω, (∑ k, S k ω) * (∑ k, S k ω).conjTranspose ∂μ) V₁ →
      loewnerLE (∫ ω, (∑ k, S k ω).conjTranspose * (∑ k, S k ω) ∂μ) V₂ →
      let r := intrinsicDimension (Matrix.fromBlocks V₁ 0 0 V₂)
      let v := max (spectralNorm V₁) (spectralNorm V₂)
      (∫ ω, spectralNorm (∑ k, S k ω) ∂μ) ≤
        C * (Real.sqrt (v * Real.log (1 + r)) + L * Real.log (1 + r)) := 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; Corollary 7.3.2, equation (7.3.3), printed p. 109; Section 7.7.4, printed pp. 117–118.
Read-back

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

There exists a real constant C>0C>0C>0 such that the following holds for every measurable space Ω\OmegaΩ, every probability measure μ\muμ on it, every pair of positive natural numbers m,nm,nm,n, every natural number NNN (including N=0N=0N=0), every family of functions Sk:Ω→Cm×nS_k:\Omega\to\mathbb C^{m\times n}Sk​:Ω→Cm×n indexed by k∈{0,…,N−1}k\in\{0,\ldots,N-1\}k∈{0,…,N−1}, every real number L≥0L\ge0L≥0, and every pair of matrices V1∈Cm×mV_1\in\mathbb C^{m\times m}V1​∈Cm×m and V2∈Cn×nV_2\in\mathbb C^{n\times n}V2​∈Cn×n. Suppose that at least one of V1,V2V_1,V_2V1​,V2​ is nonzero; each SkS_kSk​ is measurable for the Borel measurable structure on matrices; the family (Sk)k(S_k)_k(Sk​)k​ is jointly independent under μ\muμ; ∫ΩSk(ω) dμ(ω)=0\int_\Omega S_k(\omega)\,d\mu(\omega)=0∫Ω​Sk​(ω)dμ(ω)=0 for every kkk; and, for every kkk, ∥Sk(ω)∥2→2≤L\|S_k(\omega)\|_{2\to2}\le L∥Sk​(ω)∥2→2​≤L for μ\muμ-almost every ω\omegaω. Put Z(ω)=∑k=0N−1Sk(ω)Z(\omega)=\sum_{k=0}^{N-1}S_k(\omega)Z(ω)=∑k=0N−1​Sk​(ω), and suppose additionally that both V1−∫ΩZ(ω)Z(ω)∗ dμ(ω)V_1-\int_\Omega Z(\omega)Z(\omega)^*\,d\mu(\omega)V1​−∫Ω​Z(ω)Z(ω)∗dμ(ω) and V2−∫ΩZ(ω)∗Z(ω) dμ(ω)V_2-\int_\Omega Z(\omega)^*Z(\omega)\,d\mu(\omega)V2​−∫Ω​Z(ω)∗Z(ω)dμ(ω) are positive semidefinite, where ∗^*∗ denotes conjugate transpose. Define the block diagonal matrix D=(V100V2)D=\begin{pmatrix}V_1&0\\0&V_2\end{pmatrix}D=(V1​0​0V2​​), the real numbers r=Re⁡(tr⁡D)/∥D∥2→2r=\operatorname{Re}(\operatorname{tr}D)/\|D\|_{2\to2}r=Re(trD)/∥D∥2→2​ and v=max⁡{∥V1∥2→2,∥V2∥2→2}v=\max\{\|V_1\|_{2\to2},\|V_2\|_{2\to2}\}v=max{∥V1​∥2→2​,∥V2​∥2→2​}, and interpret every matrix norm here as the operator norm induced by Euclidean norms. Then ∫Ω∥Z(ω)∥2→2 dμ(ω)≤C(vlog⁡(1+r)+Llog⁡(1+r))\int_\Omega\|Z(\omega)\|_{2\to2}\,d\mu(\omega)\le C\bigl(\sqrt{v\log(1+r)}+L\log(1+r)\bigr)∫Ω​∥Z(ω)∥2→2​dμ(ω)≤C(vlog(1+r)​+Llog(1+r)). The same CCC works simultaneously for all these choices. The case L=0L=0L=0 is included; if N=0N=0N=0, the sum defining ZZZ is the zero matrix and the hypotheses quantified over kkk are vacuous, while the two positive-semidefinite inequalities and the requirement that at least one of V1,V2V_1,V_2V1​,V2​ be nonzero still apply. The integrals are Bochner integrals; no separate integrability hypotheses are stated, although the measurability and almost-sure boundedness assumptions on this finite family over a probability space give integrability of the quantities used. The denominator defining rrr is nonzero under the stated nonzero-matrix hypothesis.

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