Lemma A.2, p. 27 — AM = B iff im B ⊆ im A and M = A†B + (I − A†A)T for some T
ProvedQuadMatIneq.Reduced.lemma_A_2linear-matrix-equationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1pseudo-inverse
Let , and . Then
This is the classical description of all solutions of a linear matrix equation through the Moore–Penrose pseudo-inverse; it converts the factorization of Lemma A.1 into an explicit formula for the solution matrix.
Formalization Note The printed dimensions (, , ) are inconsistent with the term . The statement here uses the dimensions forced by the formula and by its use in the proof of Theorem 3.3: has columns, has as many rows as , and . The image inclusion is stated for the linear maps and .
Preamble
import Mathlib import Definitions.Def_QuadMatIneq_Reduced_QMI open Matrix open scoped MatrixOrder
Formal statement
namespace QuadMatIneq.Reduced
theorem lemma_A_2 {α β γ : Type*} [Fintype α] [Fintype β] [Fintype γ] [DecidableEq β]
(A : Matrix α β ℝ) (B : Matrix α γ ℝ) (M : Matrix β γ ℝ) :
A * M = B ↔
LinearMap.range B.mulVecLin ≤ LinearMap.range A.mulVecLin ∧
∃ T : Matrix β γ ℝ, M = pinv A * B + (1 - pinv A * A) * T := by sorry
end QuadMatIneq.Reduced
Source
van Waarde–Camlibel–Eising–Trentelman, arXiv:2203.12959v3, Lemma A.2, p. 27 (dimensions corrected, see Formalization Note)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.