Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 4.1, p. 7 — vec(C^(q)E)ᵀ = vec(C^(q))ᵀF^(q) and ∂/∂vec(C^(q))ᵀ = ∂/∂vec(A^(P+q))ᵀ·F^(q)ᵀ

Proved
TensorBTD.Gramian.proposition_4_1

by mikedeng1 · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

block-term-decompositionchain-rulejacobiankronecker-productp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1tensor-decomposition

Let E∈RR×R′E\in\mathbb R^{R\times R'}E∈RR×R′ be the block diagonal matrix diag(11×L1,…,11×LR)\mathrm{diag}(1_{1\times L_1},\dots,1_{1\times L_R})diag(11×L1​​,…,11×LR​​) and F(q)=E⊗IIP+qF^{(q)} = E\otimes\mathbb I_{I_{P+q}}F(q)=E⊗IIP+q​​. Then:

  1. for every C∈CIP+q×RC\in\mathbb C^{I_{P+q}\times R}C∈CIP+q​×R,
vec(C⋅E)T=vec(C)T⋅F(q);\mathrm{vec}(C\cdot E)^{\mathrm T} = \mathrm{vec}(C)^{\mathrm T}\cdot F^{(q)};vec(C⋅E)T=vec(C)T⋅F(q);
  1. (chain rule (4.3b)) for every tensor T\mathcal TT and every vector of unknowns z=(vecA(1),…,vecA(P),vecC(1),…,vecC(Q))z = (\mathrm{vec}A^{(1)},\dots,\mathrm{vec}A^{(P)},\mathrm{vec}C^{(1)},\dots,\mathrm{vec}C^{(Q)})z=(vecA(1),…,vecA(P),vecC(1),…,vecC(Q)) of the (rank-Lr∘L_r\circLr​∘ rank-1) BTD, the columns of the BTD Jacobian J=∂ vec(FBTD)/∂zTJ = \partial\,\mathrm{vec}(\mathcal F_{\mathrm{BTD}})/\partial z^{\mathrm T}J=∂vec(FBTD​)/∂zT that belong to vec(C(q))\mathrm{vec}(C^{(q)})vec(C(q)) are
∂ vec(FBTD)∂ vec(C(q))T=∂ vec(FBTD)∂ vec(A(P+q))T⋅F(q)T,\frac{\partial\,\mathrm{vec}(\mathcal F_{\mathrm{BTD}})}{\partial\,\mathrm{vec}(C^{(q)})^{\mathrm T}} = \frac{\partial\,\mathrm{vec}(\mathcal F_{\mathrm{BTD}})}{\partial\,\mathrm{vec}(A^{(P+q)})^{\mathrm T}}\cdot F^{(q)\mathrm T},∂vec(C(q))T∂vec(FBTD​)​=∂vec(A(P+q))T∂vec(FBTD​)​⋅F(q)T,

where the right-hand Jacobian is that of the unstructured CPD residual (all NNN factor matrices free) evaluated at the factor matrices of zzz, i.e. with A(P+q)=C(q)EA^{(P+q)} = C^{(q)}EA(P+q)=C(q)E.

The proposition lets one ignore the structure A(P+q)=C(q)EA^{(P+q)} = C^{(q)}EA(P+q)=C(q)E when differentiating, and then correct by the fixed matrix F(q)F^{(q)}F(q).

Formalization Note. Part (4.3a) of the proposition (the matrix form ∂/∂C(q)=∂/∂A(P+q)⋅ET\partial/\partial C^{(q)} = \partial/\partial A^{(P+q)}\cdot E^{\mathrm T}∂/∂C(q)=∂/∂A(P+q)⋅ET of the cogradient) is not formalized here; this mission formalizes the vec identity and the Jacobian form (4.3b), which is what Theorem 4.5 uses. Row vectors times matrices are Mathlib's vecMul; vec is column-major, indexed by (column, row). The left-hand side of part 2 is the submatrix of JJJ at the columns (q,(r,i))(q,(r,i))(q,(r,i)), the right-hand side the submatrix of the unstructured CPD Jacobian at the columns (P+q,((r,l),i))(P+q,((r,l),i))(P+q,((r,l),i)), times F(q)TF^{(q)\mathrm T}F(q)T. Since A(P+q)A^{(P+q)}A(P+q) is not an unknown of the BTD, ∂/∂A(P+q)\partial/\partial A^{(P+q)}∂/∂A(P+q) is read as the derivative of the unstructured CPD residual, as in the proof of Theorem 4.5.

Preamble
import Mathlib
import Definitions.Def_TensorBTD_Gramian_Setting
Formal statement
namespace TensorBTD.Gramian

open Matrix

theorem proposition_4_1 {P Q R : ℕ} {I : Fin P ⊕ Fin Q → ℕ} {L : Fin R → ℕ} (q : Fin Q) :
    (∀ C : Matrix (Fin (I (.inr q))) (Fin R) ℂ,
      Matrix.vec (C * E L) = Matrix.vecMul (Matrix.vec C) (F I L q)) ∧
    (∀ (T : Tensor I) (z : Unk I L → ℂ),
      (jac (residual T) z).submatrix id
          (fun y : Fin R × Fin (I (.inr q)) => (⟨.inr q, y⟩ : Unk I L)) =
        (jac (cpdResidual T) (fullVec z)).submatrix id
            (fun y : Col L × Fin (I (.inr q)) => (⟨.inr q, y⟩ : GIdx I L)) *
          (F I L q)ᵀ) := by sorry

end TensorBTD.Gramian
Source
Sorber, Van Barel & De Lathauwer, Optimization-Based Algorithms for Tensor Decompositions, SIAM J. Optim. 23(2) (2013), doi:10.1137/120868323, authors' manuscript (Lirias), p. 7, Proposition 4.1 (vec identity and (4.3b))
Human review
  • Endorsed by Shuze Chen · Oct 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 9, 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