tapkee (version 1.2)

Tapkee: Tapkee wrapper

Description

R wrapper for the 'tapkee' dimension reduction library

Usage

Tapkee(data, method="pca", td=2, verbose=FALSE, add="", prefix="Dim", rm=TRUE)

Arguments

data

R numerical matrix or data frame (will be converted into matrix)

method

'tapkee' method, run "system('tapkee -h')" for the list, default is "pca"

td

Number of dimensions to output, default is 2

verbose

If TRUE, 'tapkee' is verbose, defalut is FALSE

add

'tapkee' additional arguments as character string: see "system('tapkee -h')"

prefix

Variable name prefix in the resulted data frame, default is "Dim"

rm

Remove temp files (but temp folder will be removed anyway in the end of R session), default is TRUE

Value

Data frame with number of columns equal to number of dimensions given and "prefix" column names prefixes.

Details

Interface (wrapper) for the 'tapkee', flexible and efficient C++ template library for dimension reduction. 'tapkee' is extremely fast comparing with other DR tools.

For methods used in 'tapkee', run 'vignette(tapkee_methods)'.

Users should install 'tapkee' independently from author Web site (https://github.com/lisitsyn/tapkee) or associated GitHub (https://github.com/lisitsyn/tapkee). Run 'package?tapkee' or help("tapkee-package") for details related with your operation system. If 'tapkee' is not installed, Tapkee() will fail gracefully and output the input data with warning.

Please note that "[warning] The neighborhood graph is not connected" message in most cases means that 'tapkee' run was unsuccessful. As a result, Tapkee() might return the matrix of NaN's. One of possible workarounds is to specify the higher number of neigbors ('-k' option, default is 10). See below for the example.

Note that the wrapper catches only one (main) type of 'tapkee' utility outputs. For other possible output types (see 'tapkee -h' for explanation), run 'tapkee' without wrapper.

References

Sergey Lisitsyn and Christian Widmer and Fernando J. Iglesias Garcia. Tapkee: An Efficient Dimension Reduction Library. Journal of Machine Learning Research, 14: 2355-2359, 2013.

See Also

tapkee-package

Examples

# NOT RUN {
## 'tapkee' vs. R base functions
system.time(Tapkee(iris[, -5], method="mds"))
system.time(cmdscale(dist(iris[, -5])))

## How to use 'add' option
plot(Tapkee(iris[, -5], "isomap", add="-k 47"), col=iris[, 5])

## 'tapkee' methods as of March 2019:
TM <- c(
"lle", # 1) locally_linear_embedding (lle),
"npe", # 2) neighborhood_preserving_embedding (npe),
"ltsa", # 3) local_tangent_space_alignment (ltsa),
"lltsa", # 4) linear_local_tangent_space_alignment (lltsa), 
"hlle", # 5) hessian_locally_linear_embedding (hlle),
"la", # 6) laplacian_eigenmaps (la),
"lpp", # 7) locality_preserving_projections (lpp),
"dm", # 8) diffusion_map (dm),
"isomap", # 9) isomap (isomap),
"l-isomap", # 10) landmark_isomap (l-isomap),
"mds", # 11) multidimensional_scaling (mds),
"l-mds", # 12) landmark_multidimensional_scaling (l-mds),
"spe", # 13) stochastic_proximity_embedding (spe),
"kpca", # 14) kernel_pca (kpca),
"pca", # 15) pca (pca),
"ra", # 16) random_projection (ra),
"fa", # 17) factor_analysis (fa),
"t-sne", # 18) t-stochastic_neighborhood_embedding (t-sne),
"ms") # 19) manifold_sculpting (ms)

## Iris example
oldpar <- par(mfrow=c(4, 5), mar=c(1, 1, 3, 1), xaxt="n", yaxt="n")
for (n in c(1:18)) {
plot(Tapkee(iris[, -5], method=TM[n], add="-k 50"),
 col=iris[, 5], pch=20, main=TM[n], xlab="", ylab="")
}
plot(iris[, 1:2], col=iris[, 5], pch=20, main="iris[, 1:2]", xlab="", ylab="")
par(oldpar)

## Generate typical 3D data
SR <- Gen.dr.data("swissroll")
SC <- Gen.dr.data("scurve")
HX <- Gen.dr.data("helix")
SS <- Gen.dr.data("ssphere")

## This will separate colors better
COL <- rainbow(1100)[1:1000]

## 3D plot (if no 'scatterplot3d' package, plots XY axes)
T3D <- function(dat, title, col=COL) {
 if (requireNamespace("scatterplot3d", quietly = TRUE)) {
 scatterplot3d::scatterplot3d(dat, color=col, main=title, pch=20, xlab="", ylab="", zlab="",
  axis=FALSE, tick.marks=FALSE, label.tick.marks=FALSE, mar=c(1, 1, 3, 1))
} else {
 plot(dat[, 1:2], col=col, main=paste(title, "(2D projection)"), pch=20)
 warning("Please install 'scatterplot3d' package to see the 3D plot")
}}

## Swiss Roll
oldpar <- par(mfrow=c(4, 5))
T3D(SR, title="Swiss Roll")
for (n in 1:18) plot(Tapkee(SR, method=TM[n]), col=COL, pch=20, main=TM[n],
 xlab="", ylab="", xaxt="n", yaxt="n")
par(oldpar)

## S-Curve
oldpar <- par(mfrow=c(4, 5))
T3D(SC, title="S-Curve")
for (n in 1:18) plot(Tapkee(SC, method=TM[n]), col=COL, pch=20, main=TM[n],
 xlab="", ylab="", xaxt="n", yaxt="n")
par(oldpar)

## Helix
oldpar <- par(mfrow=c(4, 5))
T3D(HX, title="Helix")
for (n in 1:18) plot(Tapkee(HX, method=TM[n]), col=COL, pch=20,
 main=TM[n], xlab="", ylab="", xaxt="n", yaxt="n")
par(oldpar)

## Severed Sphere
oldpar <- par(mfrow=c(4, 5))
T3D(SS, title="Severed Sphere", col=rainbow(nrow(SS)))
for (n in 1:18) plot(Tapkee(SS, method=TM[n]), col=rainbow(nrow(SS)), pch=20,
 main=TM[n], xlab="", ylab="", xaxt="n", yaxt="n")
par(oldpar)

# }