Apply MAD-scaled SVD residual filter to new samples

Projects test samples into the frozen PC space estimated by fit_robust_pca_residual, then returns the residual (test sample minus reconstruction from top-n_components PCs) as a low-rank-removed representation.

All training features must be present in x_test.

Usage

apply_robust_pca_residual(fit, x_test, return_reconstructed = FALSE)

Arguments

fit

A robust_pca_residual_fit object from fit_robust_pca_residual.

x_test

Numeric matrix (samples \(\times\) features). All values must be non-negative.

return_reconstructed

Logical. If TRUE, returns a list with components scores, reconstructed, and residual. If FALSE (default), returns only the residual matrix.

Value

The residual matrix (samples \(\times\) features) on the MAD-scaled log scale, or a list when return_reconstructed = TRUE.