tsne: t-SNE Embedding

In maotai: Tools for Matrix Algebra, Optimization and Inference

Description

This function is a simple wrapper of Rtsne function for t-Stochastic Neighbor Embedding for finding low-dimensional structure of the data embedded in the high-dimensional space.

Usage

tsne(data, ndim = 2, ...)

Arguments

data	an (n\times p) matrix whose rows are observations.
ndim	an integer-valued target dimension.
...	extra parameters to be used in Rtsne function.

Value

a named list containing

embed

an (n\times ndim) matrix whose rows are embedded observations.

stress

discrepancy between embedded and origianl data as a measure of error.

Examples

## use simple example of iris dataset 
data(iris) 
mydat = as.matrix(iris[,1:4])
mylab = as.factor(iris[,5])

## run t-SNE and MDS for comparison
iris.cmds = cmds(mydat, ndim=2)
iris.tsne = tsne(mydat, ndim=2)

## extract coordinates and class information
cx = iris.cmds$embed # embedded coordinates of CMDS
tx = iris.tsne$embed #                         t-SNE

## visualize
#  main title
mc = paste("CMDS with STRESS=",round(iris.cmds$stress,4),sep="")
mt = paste("tSNE with STRESS=",round(iris.tsne$stress,4),sep="")

#  draw a figure
opar <- par(no.readonly=TRUE)
par(mfrow=c(1,2))
plot(cx, col=mylab, pch=19, main=mc)
plot(tx, col=mylab, pch=19, main=mt)
par(opar)

maotai documentation built on Feb. 3, 2022, 5:09 p.m.