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.
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
Rtsnefunction.
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
# \donttest{
## 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)
# }