Corollary 4.3, p. 7 — (V^σ)ᴴV^σ = W^σ
ProvedTensorBTD.Cogradient.corollary_4_3gramianhadamard-productkhatri-rao-productp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be complex matrices with rows and a common set of columns, , and let be any set of modes. With the Khatri–Rao string of (3.2) and
the Hadamard product of the Gramians of the factor matrices outside ,
In the proof of Theorem 4.4 it is applied with , to compute .
Formalization Note. Modes are Fin P ⊕ Fin Q; the rows of are indexed by the tuple of indices of the modes outside . The column set is an arbitrary finite type (in the mission it is the columns of the CPD).
Preamble
import Mathlib import Definitions.Def_TensorBTD_Cogradient_Setting
Formal statement
namespace TensorBTD.Cogradient
open Matrix
/-- Corollary 4.3, p. 7: `(V^σ)ᴴ V^σ = W^σ` for every set `σ` of modes and all factor matrices
with a common finite column set `K`. -/
theorem corollary_4_3 {P Q : ℕ} {I : Fin P ⊕ Fin Q → ℕ} {K : Type*} [Fintype K]
(σ : Finset (Fin P ⊕ Fin Q)) (A : (n : Fin P ⊕ Fin Q) → Matrix (Fin (I n)) K ℂ) :
(V σ A)ᴴ * V σ A = W σ A := by sorry
end TensorBTD.Cogradient
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, Corollary 4.3; (3.2), p. 4
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.