embed_plot: Embedding Plot

In jlmelville/sneer: Stochastic Neighbor Embedding Experiments in R

Embedding Plot

Description

Plots the embedded coordinates, with each point colored by a specified color.

Usage

embed_plot(coords, x = NULL, colors = NULL,
  color_scheme = grDevices::rainbow, num_colors = 15, limits = NULL,
  top = NULL, cex = 1, title = NULL, text = NULL, equal_axes = FALSE)

Arguments

coords	Matrix of embedded coordinates, with as many rows as observations, and 2 columns; or the return value of running sneer.
x	Either a data frame or a column that can be used to derive a suitable vector of colors. Ignored if colors is provided.
colors	Vector containing colors for each coordinate.
color_scheme	Either a color ramp function, or the name of a ColorBrewer scheme. See 'Details'.
num_colors	Number of unique colors to map to from x, if x is a numeric vector. Otherwise ignored.
limits	The range that the colors should map over when mapping from a numeric vector. If not specified, then the range of x. This is useful if there is some external absolute scale that should be used. Ignored if x is not a numeric vector.
top	If not NULL, only the specified number of points will be displayed, corresponding to those with the highest values in vec, after sorting by decreasing order.
cex	Size of the points. Ignored if text is provided.
title	Title for the plot.
text	Vector of label text to display instead of a point. If the labels are long or the data set is large, this is unlikely to be very legible, but is occasionally useful.
equal_axes	If TRUE, the X and Y axes are set to have the same extents.

Details

The x argument can be used to provide a suitable vector of colors from either a data frame or vector.

If a data frame is provided, then a vector of colors will be looked for. If it's present, it will be used as the colors argument directly. Otherwise, a factor column will be looked for, and each level will be mapped to a different color. Otherwise, one color will be used for each point. If more than one column of a type is found in the data frame, the last one encountered is used.

If a vector is provided, a similar procedure to the data frame is used when mapping from its content to a vector of colors. Additionally, a numeric vector can be provided, which will be linearly mapped to a color scheme.

The color_scheme parameter can be one of either a color ramp function, accepting an integer n as an argument and returning n colors, or the name of a ColorBrewer color scheme. Probably should be one of the "Qualitative" set.

For some applicable color ramp functions, see the Palettes help page in the grDevices package (e.g. by running the ?rainbow command).

Note

Use of this function with ColorBrewer qualitative palette names requires that the RColorBrewer package be installed.

More information on ColorBrewer is available at its website, http://www.colorbrewer2.org.

Examples

## Not run: 
# Embed with PCA
pca_iris <- sneer(iris, method = "pca", ret = c("dy"))
# Visualize the resulting embedding, colored by iris species, using the
# rainbow color scheme
embed_plot(pca_iris$coords, x = iris$Species, color_scheme = rainbow)

# same as above, save typing '$coords' each time
embed_plot(pca_iris, iris$Species, color_scheme = rainbow)

# topo.colors scheme
embed_plot(pca_iris$coords, x = iris$Species, color_scheme = topo.colors)

# Pass in data frame and it will use any factor column it finds
embed_plot(pca_iris$coords, x = iris)

#' library("RColorBrewer")
# Use the "Dark2" ColorBrewer scheme
embed_plot(pca_iris$coords, x = iris, color_scheme = "Dark2")

# Can plot the category names instead of points, but looks bad if they're
# long (or the dataset is large)
embed_plot(pca_iris$coords, x = iris$Species, cex = 0.5, text = iris$Species)

tsne_iris <- sneer(iris, method = "tsne", ret = c("dx", "dy", "deg"))
# how well is the 32 nearest neighborhood preserved for each point?
nbr_pres_32 <- nbr_pres(tsne_iris$dx, tsne_iris$dy, 32)
# visualize preservation, use absolute scale of 0-1 for colors.
embed_plot(tsne_iris$coords, x = nbr_pres_32, limits = c(0, 1))

# visualize 10 points with the hightest degree centrality
embed_plot(tsne_iris$coords, x = tsne_iris$deg, top = 10)

## End(Not run)

jlmelville/sneer documentation built on Sept. 8, 2024, 9:58 p.m.