spectral_norm_le_singular_value_zero
Provedlinear-algebramatrix-analysisoperator-norm
The spectral (operator) norm of a real matrix equals its largest singular value, and in particular is at most it: , where is the zeroth (largest) entry of the singular-value sequence of the associated Euclidean linear map . Singular values are antitone, so is the maximum. This is the geometric core of the operator-norm/Schatten-norm comparison: writing , one has , since is a positive self-adjoint operator whose top eigenvalue is .
Preamble
import Definitions.Def_matrix_completion_tangent open MatrixCompletion
Formal statement
theorem spectral_norm_le_singular_value_zero :
∀ {n₁ n₂ : ℕ} (Y : Matrix (Fin n₁) (Fin n₂) ℝ),
spectralNorm Y ≤ (Matrix.toEuclideanLin Y).singularValues 0 := by sorrySource
Candes & Recht, Exact matrix completion via convex optimization, CACM 55.6 (2012), Section 6.1.