R/KPC.R

In Luisiglm/KBoost: Inference of gene regulatory networks from gene expression data

#' Function to calculate the principal components of a kernel.
#'
#' @param K an NxN numeric matrix with the Kernel matrix.
#' @param thr a positive scalar which is a threshold to discard eigen-vectors based on eigen-values.
#' @export
#' @return the kernel principal components
#' @examples
#' x = rnorm(100,0,1)
#' k = RBF_K(x,1)
#' k_ = kernel_normal(k)
#' kpca = KPC(k,1e-8)
KPC <- function(K,thr){
    # Perform the eigen decomposition.
    Ei <- eigen(K)
    # Ei is a structure with eigen values and eigen vectors.
    Pass <- (Re(Ei$values)>thr)
    # Project the eigen-vectors on the Kernel matrix.
    if (sum(Pass) > 0){
        KPC <- K%*%Re(Ei$vectors[,Pass])
        # Set their euclidean norm to one.
        for (i in seq_len(dim(KPC)[2])){
            KPC[,i] <- KPC[,i] / sum(KPC[,i]^2)^(1/2)
        }
        return(KPC)
    } else {
        stop("No Eigen-values lower than thr. You can try using a lower thr")
    }
}