SVD-based Moore-Penrose pseudo-inverse — svd_pseudo

Computes $A^{+}$ via SVD, optionally using the split-merge variant for large matrices. Singular values below .Machine$double.eps * d[1] * max(dim(A)) are treated as zero.

Usage

svd_pseudo_inverse(A, use_split_merge = TRUE, verbose = FALSE)

Arguments

A: Numeric square matrix.
use_split_merge: Logical. If TRUE and nrow(A) >= 100, use split_merge_svd_row(); otherwise use base svd().
verbose: Logical. If TRUE, print method and condition number.

Value

A list with elements inverse (the pseudo-inverse matrix), condition_number, effective_rank, and method (character: "standard SVD" or "split-merge SVD").

Details

This is an internal helper for invert_matrix(); end users should call invert_matrix() instead.