A nonnegative scalar σ associated with a linear map that quantifies how much the map stretches or compresses vectors in specific orthogonal directions; for a real matrix A, σ is a square root of a nonnegative eigenvalue of A^T A.

Principle

Singular values measure norm amplification on orthogonal axes and are ordered to capture the anisotropic scaling properties of a linear operator independently of sign or phase.

Demonstration

For A = diag(2,-3) acting on R^2, the singular values are 2 and 3, reflecting absolute stretch along the coordinate axes regardless of orientation reversal in one axis.

Misapplication

Treating singular values as if they carried signed or phase information like some other spectral quantities leads to incorrect conclusions about directional sign or rotation; they quantify magnitude only.

Consequence

Control of singular values yields bounds on operator norms, informs numerical conditioning via the ratio of largest to smallest nonzero singular value, and quantifies sensitivity of linear systems to perturbation.

Reversal

Unlike signed or complex-valued invariants that record orientation or phase, singular values are always nonnegative and summarize magnitude of stretching without direction sign.

Boundary

Defined for rectangular matrices and bounded linear maps between inner-product spaces; their existence and ordering rely on an inner-product structure and representations of A^T A (or the appropriate adjoint composition).

Semantic Tension

Often compared with direction-sensitive spectrum information; singular values capture magnitude of action on norms, while direction- and phase-sensitive quantities capture orientation or rotation effects.

Synthesis

A singular value is a nonnegative measure of how a linear map amplifies the norm of vectors in particular orthogonal directions, summarizing anisotropic scaling into ordered magnitudes.