Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§6.5, p. 47 — the covariance-selection X-update is Q diag((λi+λi2+4ρ)/(2ρ)) QTQ\,\mathrm{diag}\bigl((\lambda_i+\sqrt{\lambda_i^2+4\rho})/(2\rho)\bigr)\,Q^TQdiag((λi​+λi2​+4ρ​)/(2ρ))QT

Proved
BoydADMM.L1.covsel_X_update

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

admmcovariance-selectioneigenvalue-decompositionp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-bookp2o-v1

Let S,Z,U∈Rn×nS,Z,U\in\mathbb R^{n\times n}S,Z,U∈Rn×n with SSS and Z−UZ-UZ−U symmetric, and let ρ>0\rho>0ρ>0. Let Q∈Rn×nQ\in\mathbb R^{n\times n}Q∈Rn×n with QTQ=QQT=IQ^TQ=QQ^T=IQTQ=QQT=I and μ1,…,μn∈R\mu_1,\dots,\mu_n\in\mathbb Rμ1​,…,μn​∈R satisfy

ρ(Z−U)−S=Q diag(μ1,…,μn) QT.\rho(Z-U)-S=Q\,\mathrm{diag}(\mu_1,\dots,\mu_n)\,Q^T .ρ(Z−U)−S=Qdiag(μ1​,…,μn​)QT.

Put

X=Q diag(μ1+μ12+4ρ2ρ,…,μn+μn2+4ρ2ρ) QT.X=Q\,\mathrm{diag}\Bigl(\frac{\mu_1+\sqrt{\mu_1^2+4\rho}}{2\rho},\dots,\frac{\mu_n+\sqrt{\mu_n^2+4\rho}}{2\rho}\Bigr)\,Q^T .X=Qdiag(2ρμ1​+μ12​+4ρ​​,…,2ρμn​+μn2​+4ρ​​)QT.

Then XXX is symmetric positive definite, and it is the unique minimizer over the symmetric positive definite matrices X′∈S++nX'\in\mathbb S^n_{++}X′∈S++n​ of

Tr⁡(SX′)−log⁡det⁡X′+ρ2∥X′−Z+U∥F2.\operatorname{Tr}(SX')-\log\det X'+\frac\rho2\|X'-Z+U\|_F^2 .Tr(SX′)−logdetX′+2ρ​∥X′−Z+U∥F2​.

With Z=ZkZ=Z^kZ=Zk and U=UkU=U^kU=Uk this is the X-update of ADMM for sparse inverse covariance selection, minimize Tr⁡(SX)−log⁡det⁡X+λ∥X∥1\operatorname{Tr}(SX)-\log\det X+\lambda\|X\|_1Tr(SX)−logdetX+λ∥X∥1​: the only update of the chapter that is neither a linear solve nor a thresholding, it costs one symmetric eigenvalue decomposition.

Formalization Note The book's eigenvalues λi\lambda_iλi​ are called μi\mu_iμi​. The statement holds for every orthogonal QQQ and diagonal factor with ρ(Z−U)−S=Q diag(μ) QT\rho(Z-U)-S=Q\,\mathrm{diag}(\mu)\,Q^Tρ(Z−U)−S=Qdiag(μ)QT, not just one chosen by Mathlib. The symmetry of SSS (an empirical covariance (1/N)∑iaiaiT(1/N)\sum_ia_ia_i^T(1/N)∑i​ai​aiT​) and of Z−UZ-UZ−U (the ADMM iterates are symmetric) is an explicit hypothesis; SSS is not required to be positive semidefinite. The minimization is over Matrix.PosDef, the domain of log⁡det⁡\log\detlogdet (p. 46); singular matrices are not competitors. ∥⋅∥F2\|\cdot\|_F^2∥⋅∥F2​ is the sum of squared entries.

Preamble
import Mathlib
import Definitions.Def_BoydADMM_L1_Basic
import Definitions.Def_BoydADMM_L1_CovSel

open Matrix
Formal statement
namespace BoydADMM.L1

/-- §6.5, p. 47 (the goal): let `S` and `Z − U` be symmetric, `ρ > 0`, and
`ρ(Z − U) − S = Q diag(μ) Qᵀ` with `QᵀQ = QQᵀ = I`. Then `X = Q X̃ Qᵀ`,
`X̃_ii = (μ_i + √(μ_i² + 4ρ))/(2ρ)`, is positive definite and is the unique minimizer of
`Tr(SX) − log det X + (ρ/2)‖X − Z + U‖_F²` over the positive definite matrices. -/
theorem covsel_X_update {n : ℕ} (S Z U Q : Matrix (Fin n) (Fin n) ℝ) (ρ : ℝ) (μ : Fin n → ℝ)
    (hρ : 0 < ρ) (hS : S.IsSymm) (hZU : (Z - U).IsSymm)
    (hQ₁ : Qᵀ * Q = 1) (hQ₂ : Q * Qᵀ = 1)
    (hdecomp : ρ • (Z - U) - S = Q * Matrix.diagonal μ * Qᵀ) :
    IsUniqueMinimizerOn (covselXObjective S Z U ρ) {X | X.PosDef} (covselX ρ Q μ) := by sorry

end BoydADMM.L1
Source
Boyd, Parikh, Chu, Peleato, Eckstein, Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers, Found. Trends Mach. Learn. 3(1) (2011), p. 47, §6.5 (X-update in closed form); objective on p. 46
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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