troppca.linsp {Rtropical}

Tropical Principal Component Analysis by Tropical Linear Space

Description

Approximate the principal component as a tropical linear space for a given data matrix and returns the results as an object of class troppca.

Usage

troppca.linsp(x, pcs = 2, iteration = list(), ncores = 2)

Arguments

x

a data matrix, of size n x e, with each row an observation vector. e is the dimension of the tropical space

pcs

a numeric value indicating the order of principal component. (default: 2)

iteration

a list with arguments controlling the iteration of the algorithm.

exhaust

a logical variable indicating if to iterate over all possible combinations of the linear space based on the given data matrix x. If FALSE, please input a number of iteration for niter. If TRUE, please enter 0 for niter and this function will iterate over all possible combinations of linear space. This could be time consuming when x is large. (default: FALSE)

niter

a numeric variable indicating the number of iterations. (default: 100)

ncores

a numeric value indicating the number of threads utilized for multi-cored CPUs. (default: 2)

Value

A list of S3 class "troppca", including:

pc

The principal component as a tropical linear space

obj

The tropical PCA objective, the sum of tropical distance from each point to the projection.

projection

The projections of all data points.

type

The geometry of principal component.

Examples

library(Rfast)
n <- 100
e <- 10
sig2 <- 1
x <- rbind(rmvnorm(n, mu = c(5, -5, rep(0, e - 2)), sigma = diag(sig2, e)))
troppca_fit <- troppca.linsp(x)