matrix_trace_duality_inequality
Provedconvex-analysislinear-algebramatrix-completion
Trace duality (von Neumann inequality), Candès–Recht 2009, Lemma 3.2, first part (arXiv:0805.4471, p.15). For any two real matrices , the Frobenius (trace) inner product is bounded by the product of the spectral (operator) norm of and the nuclear norm of :
Here (matrixInner), is the operator norm (spectralNorm, the largest singular value), and is the nuclear norm (nuclearNorm, the sum of singular values). This is the statement that the nuclear and spectral norms are dual to one another. It follows from the von Neumann trace inequality together with ; equivalently, writing in its SVD, and for unit vectors .
Preamble
import Definitions.Def_matrix_completion_tangent open MatrixCompletion
Formal statement
theorem matrix_trace_duality_inequality {n₁ n₂ : ℕ} (A B : Matrix (Fin n₁) (Fin n₂) ℝ) : matrixInner A B ≤ spectralNorm A * nuclearNorm B := by sorrySource
Candès–Recht 2009, arXiv:0805.4471, Lemma 3.2 (p.15)