Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 1 — robust regression with feature-wise uncertainty set (2) is equivalent to ℓ¹-regularized regression (3)

Proved
RobustRegLasso.FeatureWise.theorem_1_robust_eq_l1_regularized

by mikedeng1 · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

lassop2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1regressionrobust-optimization

Let n≥1n\ge 1n≥1 and mmm be integers, let A∈Rn×mA\in\mathbb R^{n\times m}A∈Rn×m have columns a1,…,ama_1,\dots,a_ma1​,…,am​, let b∈Rnb\in\mathbb R^nb∈Rn, and let c1,…,cm≥0c_1,\dots,c_m\ge 0c1​,…,cm​≥0. Let U={(δ1,…,δm):∥δi∥2≤ci, i=1,…,m}\mathcal U=\{(\delta_1,\dots,\delta_m):\|\delta_i\|_2\le c_i,\ i=1,\dots,m\}U={(δ1​,…,δm​):∥δi​∥2​≤ci​, i=1,…,m} be the feature-wise uncoupled uncertainty set (2). Then the robust regression problem (1),

min⁡x∈Rm{max⁡ΔA∈U∥b−(A+ΔA)x∥2},\min_{x\in\mathbb R^m}\Bigl\{\max_{\Delta A\in\mathcal U}\|b-(A+\Delta A)x\|_2\Bigr\},x∈Rmmin​{ΔA∈Umax​∥b−(A+ΔA)x∥2​},

is equivalent to the ℓ¹-regularized regression problem (3),

min⁡x∈Rm{∥b−Ax∥2+∑i=1mci∣xi∣}.\min_{x\in\mathbb R^m}\Bigl\{\|b-Ax\|_2+\sum_{i=1}^m c_i|x_i|\Bigr\}.x∈Rmmin​{∥b−Ax∥2​+i=1∑m​ci​∣xi​∣}.

Precisely:

  1. a vector x∈Rmx\in\mathbb R^mx∈Rm minimizes the robust objective R(x)=sup⁡ΔA∈U∥b−(A+ΔA)x∥2R(x)=\sup_{\Delta A\in\mathcal U}\|b-(A+\Delta A)x\|_2R(x)=supΔA∈U​∥b−(A+ΔA)x∥2​ if and only if it minimizes L(x)=∥b−Ax∥2+∑ici∣xi∣L(x)=\|b-Ax\|_2+\sum_i c_i|x_i|L(x)=∥b−Ax∥2​+∑i​ci​∣xi​∣;
  2. the two problems have the same optimal value, inf⁡xR(x)=inf⁡xL(x)\inf_x R(x)=\inf_x L(x)infx​R(x)=infx​L(x).

The theorem identifies robustness to feature-wise bounded perturbations of the data matrix with ℓ¹ regularization: with equal budgets ci=cc_i=cci​=c it is the Lasso with the unsquared loss.

Formalization Note "Equivalent" is read as in the proof ("Minimizing over xxx on both sides proves the theorem"): same minimizers and same optimal value. No minimizer is claimed to exist, since the paper does not show one. The robust objective RRR is the supremum over U\mathcal UU in EReal, and the optimal values are infima in EReal, so neither carries a junk value. The paper leaves n≥1n\ge1n≥1 implicit; it is a hypothesis (0 < n). The budgets ci≥0c_i\ge0ci​≥0 are the paper's standing assumption on (2).

Preamble
import Mathlib
import Definitions.Def_RobustRegLasso_FeatureWise_Basic
Formal statement
namespace RobustRegLasso.FeatureWise

theorem theorem_1_robust_eq_l1_regularized {n m : ℕ} (hn : 0 < n)
    (a : Fin m → EuclideanSpace ℝ (Fin n)) (b : EuclideanSpace ℝ (Fin n)) (c : Fin m → ℝ)
    (hc : ∀ i, 0 ≤ c i) :
    (∀ x : Fin m → ℝ, IsMinOn (robustObjective a b c) Set.univ x ↔
        IsMinOn (l1RegularizedObjective a b c) Set.univ x) ∧
      ⨅ x : Fin m → ℝ, robustObjective a b c x =
        ⨅ x : Fin m → ℝ, ((l1RegularizedObjective a b c x : ℝ) : EReal) := by sorry

end RobustRegLasso.FeatureWise
Source
Xu, Caramanis, Mannor, Robust Regression and Lasso, arXiv:0811.1790v1, p. 4, Theorem 1
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · 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