pflm.fpca.utils.get_eigen_analysis_results

get_eigen_analysis_results(reg_cov: ndarray, is_upper_triangular: bool = False) → tuple[ndarray, ndarray]

Compute eigenvalues and eigenvectors of a covariance matrix.

Parameters:

reg_covnp.ndarray of shape (nt, nt)

Regularized covariance matrix. Dtype determines the LAPACK routine (float64 -> f64 backend; otherwise f32).

is_upper_triangularbool, default=False

Whether reg_cov contains only the upper triangular part (packed in a full matrix). If True, the routine will treat the lower part as unspecified.

Returns:

eig_lambdanp.ndarray of shape (k,)

Sorted eigenvalues (descending) filtered to positive and finite values.

eig_vectornp.ndarray of shape (nt, k)

Corresponding eigenvectors (columns) aligned with eig_lambda.

Warns:

UserWarning

If the LAPACK routine fails (info != 0) or eigenvalues contain NaN or negative values.

Notes

• Very small eigenvalues (<= 10 * eps) are discarded.

• On failure, the function returns (None, None).