`scipy.stats._multivariate:_eigvalsh_to_eps`

source: /scipy/stats/_multivariate.py :72

Signature

def _eigvalsh_to_eps ( spectrum , cond = None , rcond = None )

Summary

Determine which eigenvalues are "small" given the spectrum.

Extended Summary

This is for compatibility across various linear algebra functions that should agree about whether or not a Hermitian matrix is numerically singular and what is its numerical matrix rank. This is designed to be compatible with scipy.linalg.pinvh.

Parameters

spectrum : 1d ndarray: Array of eigenvalues of a Hermitian matrix.
cond, rcond : float, optional: Cutoff for small eigenvalues. Singular values smaller than rcond * largest_eigenvalue are considered zero. If None or -1, suitable machine precision is used.

Returns

eps : float: Magnitude cutoff for numerical negligibility.

Aliases

scipy.stats._multivariate._eigvalsh_to_eps