Learn R Programming

seriation (version 1.5.6)

register_umap: Register Seriation Based on 1D UMAP

Description

Use uniform manifold approximation and projection (UMAP) to embed the data on the number line and create a order for seriate().

Usage

register_umap()

Arguments

Value

Nothing.

Details

Registers the method "umap" for seriate(). This method applies 1D UMAP to a data matrix or a distance matrix and extracts the order from the 1D embedding.

Control parameter n_epochs can be increased to find a better embedding.

The returned seriation permutation vector has an attribute named embedding containing the umap embedding.

Note: Package umap needs to be installed.

References

McInnes, L and Healy, J, UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction, ArXiv e-prints 1802.03426, 2018.

See Also

umap::umap() in umap.

Other seriation: register_DendSer(), register_GA(), register_optics(), register_smacof(), register_tsne(), registry_for_seriaiton_methods, seriate(), seriate_best()

Examples

Run this code

if (FALSE) {
register_umap()

## distances
get_seriation_method("dist", "umap")

data(SupremeCourt)
d <- as.dist(SupremeCourt)

o <- seriate(d, method = "umap", verbose = TRUE)
pimage(d, o)

# look at the returned embedding and plot it
attr(o[[1]], "configuration")
plot_config(o)

## matrix
get_seriation_method("matrix", "umap")

data("Zoo")
Zoo[,"legs"] <- (Zoo[,"legs"] > 0)
x <- as.matrix(Zoo[,-17])
label <- rownames(Zoo)
class <- Zoo$class

o <- seriate(x, method = "umap", verbose = TRUE)
pimage(x, o)

plot_config(o[[1]], col = class)
}

Run the code above in your browser using DataLab