Gavish–Donoho Optimal Adaptive PCA Denoising

Filters high-dimensional correlated features by hard-thresholding singular values at the Gavish–Donoho (2014) optimal threshold.

Usage

lmdt_pca(X)

Arguments

X

A numeric matrix of covariates (n x p).

Value

A matrix of retained principal components.

Examples

set.seed(42)
X <- matrix(rnorm(100 * 20), nrow = 100, ncol = 20)
X_clean <- lmdt_pca(X)
ncol(X_clean)
#> [1] 2