Appendix B.1, p. 33 — (Σ̃ + λI)(β* − b) = (γ̃ − Σ̃b) − λb and ‖β* − b‖₂ ≤ (‖γ̃ − Σ̃b‖₂ + λ‖b‖₂)/(σ_min(Σ̃) + λ)
ProvedSynDigits.Ridge.population_coefficient_errorerror-boundp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1perturbation-boundridge-regression
Let be symmetric positive semidefinite, , and with , where is the smallest singular value. Let . Then
and
Combined with the structural split, this yields the second part of Theorem 4.1, the bound on the structural error .
Formalization Note The hypothesis makes invertible, so is the genuine solution of rather than Lean's junk value .
Differences from the source:
- generalization: Σ̃ = (1/n)𝔼[ỸᵀỸ], γ̃ = (1/n)𝔼[ỸᵀỸ_v] for a random twin matrix → any positive semidefinite Sig, any vector gam; reason: every such expectation is symmetric PSD, so the statement specializes to the paper's; the argument uses only (Σ̃+λI)β* = γ̃ and Σ̃ ⪰ 0
- encoding: (Σ̃+λI)⁻¹ exists (implicit) → 0 < sMin Sig + lam; reason: for PSD Σ̃ this excludes exactly the singular Σ̃ + λI, where the paper's β* and the bound's denominator are undefined; Lean's inverse returns 0 there
- encoding: λ ≥ 0 (Ridge penalty, implicit) → 0 ≤ lam; reason: Ridge penalties are nonnegative; the printed λ‖b‖₂ equals ‖λb‖₂ only for λ ≥ 0
- encoding: vectors in ℝⁿ, n×m matrices, ‖·‖₂, spectral norm, σ_min → Fin n → ℝ, Matrix (Fin n) (Fin m) ℝ, vnorm, GaussianMatrix.specNorm, GaussianMatrix.sMin; reason: Lean representation of the same objects; same instances
Preamble
import Mathlib import Definitions.Def_GaussianMatrix_basic import Definitions.Def_SynDigits_Ridge_Model open Matrix GaussianMatrix
Formal statement
namespace SynDigits.Ridge
theorem population_coefficient_error {m : ℕ} (Sig : Matrix (Fin m) (Fin m) ℝ)
(gam b : Fin m → ℝ) (lam : ℝ)
(hSig : Sig.PosSemidef) (hlam : 0 ≤ lam) (hpos : 0 < sMin Sig + lam) :
(Sig + lam • (1 : Matrix (Fin m) (Fin m) ℝ)) *ᵥ (betaStar Sig gam lam - b)
= (gam - Sig *ᵥ b) - lam • b ∧
vnorm (betaStar Sig gam lam - b)
≤ 1 / (sMin Sig + lam) * (vnorm (gam - Sig *ᵥ b) + lam * vnorm b) := by sorry
end SynDigits.Ridge
Source
Fan, Huang, Peng, Wang, Wu, SYN-DIGITS: A Synthetic Control Framework for Calibrated Digital Twin Simulation, arXiv:2604.07513v1, p. 33, Appendix B.1, proof of Theorem 4.1, "To bound ||b − β*||₂"
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.