by LukeBernese · Aug 13, 2026 · Mathlib c5ea003 (Lean v4.30.0)
machine-learningmatrix-completionoptimization
Let R(X)=∑i=1d(∥Xi∥−α)+4 be the row regularizer with threshold α>0. Its gradient is ∇R(X)=ΓX (the companion milestone), so the directional derivative of R at a point Y along V is ⟨∇R(Y),V⟩. This statement is the second-order half of the same calculus: as the base point slides along the direction V, that directional derivative is itself differentiable at s=0, with
dsd⟨∇R(X+sV),V⟩s=0=⟨V,∇2R(X)[V]⟩.
Equivalently, s↦R(X+sV) is twice differentiable at 0 and its second derivative is the Hessian quadratic form vec(V)⊤∇2R(X)vec(V). Row by row, for the radial function x↦h(∥x∥) with h(t)=(t−α)+4, the Hessian is
h′′(t)uu⊤+th′(t)(I−uu⊤),t=∥Xi∥,u=Xi/t,
with h′(t)=4(t−α)+3 and h′′(t)=12(t−α)+2, which is exactly the quantity the mission's model layer records.
Formalization Note The statement is phrased as a directional derivative of the directional gradient, mirroring the gradient milestone, so that no norm has to be fixed on matrix space and no second-order Frechet machinery is needed. The hypothesis α>0 is what makes the formula correct at rows with Xi=0: there R vanishes identically in a neighbourhood, and the formula's 0/0 terms are 0 under Lean's division convention, so both sides are zero. This is the missing analytic input to the first- and second-order optimality conditions (Chen-Li Lemma 4.3): the smooth part of the objective is a quartic polynomial along any line, so the regularizer is the only term whose second derivative needs an argument.
Preamble
import Definitions.Def_MCNoSpuriousMinModel
import Mathlib.Analysis.Calculus.Deriv.Basic
open Matrix MatrixCompletion.NoSpuriousMin
Formal statement
theorem MatrixCompletion.NoSpuriousMin.regularizer_hessian
{d r : ℕ} (α : ℝ) (hα : 0 < α) (X V : Matrix (Fin d) (Fin r) ℝ) :
HasDerivAt (fun s : ℝ => innerM (regGrad α (X + s • V)) V) (regHessQF α X V) 0 := by sorry
Source
Ge, Lee, Ma 2016, Matrix Completion has No Spurious Local Minimum, https://arxiv.org/abs/1605.07272 (v4), p. 11, Proposition 5.2 (the Hessian half; the printed exponent 4 in the gradient is corrected to 3 per the derivative of (t-alpha)_+^4 and the rank-1 form on its p. 9). Explicit Hessian form: Ge, Jin, Zheng 2017, No Spurious Local Minima in Nonconvex Low Rank Problems, https://arxiv.org/abs/1704.00708, Lemma 18. Used by: Chen, Li 2019, Model-free Nonconvex Matrix Completion, JMLR 20(142), https://arxiv.org/abs/1711.01742 (v3), pp. 16-19, Lemma 4.3 (second-order optimality condition).