§2.2, p. 4 — M = diag(b)(I − AA†)diag(b) is positive semidefinite
ProvedPhaseCut.Tight.Mmat_posSemidefp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1phase-retrievalpositive-semidefiniteprojector
Let be injective, with pseudoinverse , and let be arbitrary. Then the Hermitian matrix
is positive semidefinite.
The matrix is the orthogonal projector onto the orthogonal complement of the range of . Positive semidefiniteness of makes the quadratic problem (2), over unit-modulus , and its relaxation PhaseCut bounded below by ; it is used in the proof of Proposition 4.2.
Formalization Note The paper states this for every . Here is encoded as , which is the pseudoinverse only when is injective, so injectivity is added as a hypothesis; it is the standing assumption of the whole mission.
Preamble
import Mathlib import Definitions.Def_PhaseCut_Tight_Defs
Formal statement
namespace PhaseCut.Tight
open Matrix
open scoped ComplexOrder
/-- §2.2, p. 4: for injective `A` (so that `pinv A` is `A†`) and any real `b`, the matrix
`M = diag(b)(I − AA†)diag(b)` is positive semidefinite. -/
theorem Mmat_posSemidef {n p : ℕ} (A : Matrix (Fin n) (Fin p) ℂ) (b : Fin n → ℝ)
(hA : Function.Injective A.mulVec) :
(Mmat A b).PosSemidef := by sorry
end PhaseCut.Tight
Source
Waldspurger, d'Aspremont & Mallat, arXiv:1206.0102v3, §2.2, pp. 3–4 (sentence after (2); cited p. 4)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.