Prepare the data for whitening — rs_prepare

Prepares the data for subsequent usage in ICA. Incorrect use can cause kernel crashes. Wrapper around the Rust functions with type checks are provided in the package.

Usage

rs_prepare_whitening(x, fast_svd, seed, rank, oversampling, n_power_iter)

Arguments

x: Numeric matrix to whiten. The columns are centred and the whitening happens over the columns.
fast_svd: Boolean. Shall a randomised SVD be used. This is way faster on larger data sets.
seed: Integer. Only relevant with fast_svd is set to TRUE.
rank: Integer. How many ranks to use for the fast SVD approximation. If you supply NULL, it will default to 10L. Only relevant with fast_svd is set to TRUE.
oversampling: Integer. Oversampling parameter to make the approximation more precise. If you supply NULL, it will default to 10L. Only relevant with fast_svd is set to TRUE.
n_power_iter: Integer. Number of power iterations for the randomised SVD. If you supply NULL, it will default to 2L. Only relevant with fast_svd is set to TRUE.

Value

A list containing:

x - The column-centred input, transposed.
k - The whitening matrix K. With fast_svd = TRUE it can carry more than rank rows; the caller trims it.