Package 'detourr'

Initiate a detour visualisation

Description

This function initialises a detour object which, along with the tour_path and show_functions is used to build a detour visualisation.

Usage

detour(.data, mapping)

Arguments

.data	a data frame, tibble, or crosstalk::SharedData object
mapping	a mapping of data columns to aesthetic values using the tour_aes function. The only required aesthetic is projection, which determines which columns are used to generate the tour path and supports tidy selection.

Value

A detour object containing information about the tour path and its parameters

Examples

detour(tourr::flea, tour_aes(projection = -species, colour = species)) |>
  tour_path(grand_tour(3), fps = 60) |>
  show_scatter(alpha = 0.7, axes = FALSE)

---

Shiny bindings for detourr

Description

Output and render functions for using detourr with shiny. The output function used must match both the display method and tour dim used, or it will lead to strange behavour.

Usage

displayScatter3dOutput(output_id, width = "100%", height = "400px")

displayScatter2dOutput(output_id, width = "100%", height = "400px")

shinyRenderDisplayScatter2d(expr, env = parent.frame(), quoted = FALSE)

shinyRenderDisplayScatter3d(expr, env = parent.frame(), quoted = FALSE)

Arguments

output_id	output variable to read from
width, height	Must be a valid CSS unit (like "100%", "400px", "auto") or a number, which will be coerced to a string and have "px" appended.
expr	an expression that generates a detourr widget
env	The environment in which to evaluate expr.
quoted	Is expr a quoted expression (with quote())? This is useful if you want to save an expression in a variable.

Value

An output or render function that enables the use of the widget within shiny applications

---

Test for detour-ness

Description

Test for detour-ness

Usage

is_detour(x)

Arguments

x

an object

Value

TRUE or FALSE

---

Embeddings of images in the MNIST test set

Description

Two datasets are available; mnist_embeddings_8d contains 8-dimensional embedding vectors and mnist_embeddings_32d contains 32-dimensional embedding vectors.

The neural network that produced these embeddings was created using TensorFlow (Abadi et al. (2016)) with a variation of the code found in this example: https://www.tensorflow.org/addons/tutorials/losses_triplet

Usage

mnist_embeddings_32d

mnist_embeddings_8d

Format

An object of class tbl_df (inherits from tbl, data.frame) with 10000 rows and 34 columns.

An object of class tbl_df (inherits from tbl, data.frame) with 10000 rows and 10 columns.

Details

A data frame with 10,000 rows and p variables:

• id: sequential ID or row number of the image

• label: the digit 0, 1, ..., 9

• X1–Xp: elements 1–p of the embedding vector

References

LeCun, Y (1998). The MNIST database of handwritten digits. http://yann.lecun.com/exdb/mnist/.

Abadi, M, P Barham, J Chen, Z Chen, A Davis, J Dean, M Devin, S Ghemawat, G Irving, M Isard, et al. (2016). TensorFlow: A System for Large-Scale Machine Learning. In: 12th USENIX symposium on operating systems design and implementation (OSDI 16), pp.265–283.

---

2D and 3D Sage Tour Display

Description

An implementation of the Sage Tour described in Laa et al., 2021. It uses a radial transformation on the projected data so that the relative volume is preserved when the data is projected. I.e. a uniform distribution in the original space will remain uniformly distributed in the projected space. Includes both 2D and 3D variations.

Usage

show_sage(
  x,
  ...,
  palette = viridis,
  center = TRUE,
  axes = TRUE,
  edges = NULL,
  paused = TRUE,
  scale_factor = NULL,
  gamma = 1,
  R = NULL
)

Arguments

x	a detour object
...	used to support aesthetic parameters for the plot, including size: point size, defaults to 1 alpha: point opacity, defaults to 1 background_colour: defaults to "white"
palette	Colour palette to use with the colour aesthetic. Can be: A character vector of R colours. This should match the number of levels of the colour aesthetic, or the number of bins to use for continuous colours. A function which takes the number of colours to use as input and returns a character vector of colour names and / or hex values as output.
center	If TRUE, center the projected data to (0, 0, 0).
axes	Can be one of: TRUE draw axes and use column names for axis labels FALSE do not draw axes or labels NULL draw axes with no labels An unnamed vector of labels with the same length as cols A named vector in the form c("h" = "head"), where head is renamed to h
edges	A two column numeric matrix giving indices of ends of lines.
paused	whether the widget should be initialised in the 'paused' state
scale_factor	used as a multiplier for the point coordinates so they are displayed on a sensible range. Defaults to the reciprocal of maximum distance from a point to the origin.
gamma	the gamma parameter for scaling the effective dimensionality for the sage tour radial transformation. defaults to 1
R	scale for the radial transformation. Defaults to scale_factor times the maximum distance from the origin to each row of data. If the default scale_factor is used this will result in R=1. Because the R and scale_factor parameters interact with one another, it is recommended to leave scale_factor at its default value, and modify R if needed.

Details

This display method produces an interactive scatterplot animation which supports both 2D and 3D tours. Linked selection and filtering is also supported using crosstalk. The set of interactive controls available are:

• A timeline with a play / pause button and indicators at the position of each basis used. The basis indicators can be hovered with the mouse to show the index of the basis, or clicked to jump to that basis. The timeline also allows for clicking and dragging of the scrubber to move to any individual frame of the animation.

• Orbit controls. For the 2D variant, this allows the projection to be rotated by clicking and dragging from left to right. For the 3D variant, full orbit controls are available by clicking and dragging. For both orbit and pan controls, the scroll wheel can be used to zoom.

• Pan controls, which work similarly to orbit controls but move the camera laterally / vertically rather than rotating

• Resetting of the orbit and pan controls

• Selection and highlighting. Multiple selection is possible by using the shift key

• Colouring / brushing of highlighted points

Value

An object of class htmlwidget

References

Laa, U., Cook, D., & Lee, S. (2021). Burning sage: Reversing the curse of dimensionality in the visualization of high-dimensional data. Journal of Computational and Graphical Statistics, 1-10.

See Also

show_scatter

Examples

detour(tourr::flea, tour_aes(projection = -species, colour = species)) |>
  tour_path(grand_tour(3), fps = 60) |>
  show_sage(gamma = 2)

---

2D and 3D Scatter Plot Display for Tours

Description

Display method for a high performance 2D or 3D scatter plot. Performance is achieved through the use of Three.js / WebGL, and the 2D or 3D variant is selected automatically based on the tour generator provided.

Usage

show_scatter(
  x,
  ...,
  palette = viridis,
  center = TRUE,
  axes = TRUE,
  edges = NULL,
  paused = TRUE,
  scale_factor = NULL
)

Arguments

x	a detour object
...	used to support aesthetic parameters for the plot, including size: point size, defaults to 1 alpha: point opacity, defaults to 1 background_colour: defaults to "white"
palette	Colour palette to use with the colour aesthetic. Can be: A character vector of R colours. This should match the number of levels of the colour aesthetic, or the number of bins to use for continuous colours. A function which takes the number of colours to use as input and returns a character vector of colour names and / or hex values as output.
center	If TRUE, center the projected data to (0, 0, 0).
axes	Can be one of: TRUE draw axes and use column names for axis labels FALSE do not draw axes or labels NULL draw axes with no labels An unnamed vector of labels with the same length as cols A named vector in the form c("h" = "head"), where head is renamed to h
edges	A two column numeric matrix giving indices of ends of lines.
paused	whether the widget should be initialised in the 'paused' state
scale_factor	used as a multiplier for the point coordinates so they are displayed on a sensible range. Defaults to the reciprocal of maximum distance from a point to the origin.

Details

This display method produces an interactive scatterplot animation which supports both 2D and 3D tours. Linked selection and filtering is also supported using crosstalk. The set of interactive controls available are:

• A timeline with a play / pause button and indicators at the position of each basis used. The basis indicators can be hovered with the mouse to show the index of the basis, or clicked to jump to that basis. The timeline also allows for clicking and dragging of the scrubber to move to any individual frame of the animation.

• Orbit controls. For the 2D variant, this allows the projection to be rotated by clicking and dragging from left to right. For the 3D variant, full orbit controls are available by clicking and dragging. For both orbit and pan controls, the scroll wheel can be used to zoom.

• Pan controls, which work similarly to orbit controls but move the camera laterally / vertically rather than rotating

• Resetting of the orbit and pan controls

• Selection and highlighting. Multiple selection is possible by using the shift key

• Colouring / brushing of highlighted points

Value

An object of class htmlwidget

Examples

detour(tourr::flea, tour_aes(projection = -species, colour = species)) |>
  tour_path(grand_tour(3), fps = 60) |>
  show_scatter(alpha = 0.7, axes = FALSE)

---

Internal method for 2D and 3D Scatter Plot Display

Description

Internal method for 2D and 3D Scatter Plot Display

Usage

show_scatter_internal(
  x,
  ...,
  palette = viridisLite::viridis,
  center = TRUE,
  axes = TRUE,
  edges = NULL,
  paused = TRUE,
  scale_factor = NULL
)

Arguments

x	a detour object
...	used to support aesthetic parameters for the plot, including size: point size, defaults to 1 alpha: point opacity, defaults to 1 background_colour: defaults to "white"
palette	Colour palette to use with the colour aesthetic. Can be: A character vector of R colours. This should match the number of levels of the colour aesthetic, or the number of bins to use for continuous colours. A function which takes the number of colours to use as input and returns a character vector of colour names and / or hex values as output.
center	If TRUE, center the projected data to (0, 0, 0).
axes	Can be one of: TRUE draw axes and use column names for axis labels FALSE do not draw axes or labels NULL draw axes with no labels An unnamed vector of labels with the same length as cols A named vector in the form c("h" = "head"), where head is renamed to h
edges	A two column numeric matrix giving indices of ends of lines.
paused	whether the widget should be initialised in the 'paused' state
scale_factor	used as a multiplier for the point coordinates so they are displayed on a sensible range. Defaults to the reciprocal of maximum distance from a point to the origin.

Details

This display method produces an interactive scatterplot animation which supports both 2D and 3D tours. Linked selection and filtering is also supported using crosstalk. The set of interactive controls available are:

• A timeline with a play / pause button and indicators at the position of each basis used. The basis indicators can be hovered with the mouse to show the index of the basis, or clicked to jump to that basis. The timeline also allows for clicking and dragging of the scrubber to move to any individual frame of the animation.

• Orbit controls. For the 2D variant, this allows the projection to be rotated by clicking and dragging from left to right. For the 3D variant, full orbit controls are available by clicking and dragging. For both orbit and pan controls, the scroll wheel can be used to zoom.

• Pan controls, which work similarly to orbit controls but move the camera laterally / vertically rather than rotating

• Resetting of the orbit and pan controls

• Selection and highlighting. Multiple selection is possible by using the shift key

• Colouring / brushing of highlighted points

Value

An object of class htmlwidget

---

2D and 3D Slice Display for Tours

Description

An implementation of the Slice Tour described in Laa et al., 2020. Points close to the projection plane are highlighted, and those far away are faded out.

Usage

show_slice(
  x,
  ...,
  palette = viridis,
  center = TRUE,
  axes = TRUE,
  edges = NULL,
  paused = TRUE,
  scale_factor = NULL,
  slice_relative_volume = 0.1,
  anchor = NULL
)

Arguments

x	a detour object
...	used to support aesthetic parameters for the plot, including size: point size, defaults to 1 alpha: point opacity, defaults to 1 background_colour: defaults to "white"
palette	Colour palette to use with the colour aesthetic. Can be: A character vector of R colours. This should match the number of levels of the colour aesthetic, or the number of bins to use for continuous colours. A function which takes the number of colours to use as input and returns a character vector of colour names and / or hex values as output.
center	If TRUE, center the projected data to (0, 0, 0).
axes	Can be one of: TRUE draw axes and use column names for axis labels FALSE do not draw axes or labels NULL draw axes with no labels An unnamed vector of labels with the same length as cols A named vector in the form c("h" = "head"), where head is renamed to h
edges	A two column numeric matrix giving indices of ends of lines.
paused	whether the widget should be initialised in the 'paused' state
scale_factor	used as a multiplier for the point coordinates so they are displayed on a sensible range. Defaults to the reciprocal of maximum distance from a point to the origin.
slice_relative_volume	number default 0.1. Controls the relative volume of the slice and thus the number of points which are highlighted. This is an approximate value and is only accurate for values << 1
anchor	anchor vector of length p to offset the projection plane when calculating the distance from each point to the projection plane.

Details

This display method produces an interactive scatterplot animation which supports both 2D and 3D tours. Linked selection and filtering is also supported using crosstalk. The set of interactive controls available are:

• A timeline with a play / pause button and indicators at the position of each basis used. The basis indicators can be hovered with the mouse to show the index of the basis, or clicked to jump to that basis. The timeline also allows for clicking and dragging of the scrubber to move to any individual frame of the animation.

• Orbit controls. For the 2D variant, this allows the projection to be rotated by clicking and dragging from left to right. For the 3D variant, full orbit controls are available by clicking and dragging. For both orbit and pan controls, the scroll wheel can be used to zoom.

• Pan controls, which work similarly to orbit controls but move the camera laterally / vertically rather than rotating

• Resetting of the orbit and pan controls

• Selection and highlighting. Multiple selection is possible by using the shift key

• Colouring / brushing of highlighted points

Value

An object of class htmlwidget

References

Laa, U., Cook, D., & Valencia, G. (2020). A slice tour for finding hollowness in high-dimensional data. Journal of Computational and Graphical Statistics, 29(3), 681-687.

See Also

show_scatter

Examples

x <- geozoo::torus(p = 4, n = 10000)$points |>
  tibble::as_tibble(.name_repair = "unique")

detour(x, tour_aes(projection = everything())) |>
  tour_path(grand_tour(2)) |>
  show_slice(slice_relative_volume = 0.1)

---

Aesthetic mapping for tours

Description

Aesthetic mapping for tours describing how variables in the data are mapped to visual properties of the tour animation.

Usage

tour_aes(...)

Arguments

...	list of name-value pairs in the form 'aesthetic = variable'. Variables are evaluated using tidyselect syntax.

Value

a list of quosures

Examples

detour(tourr::flea, tour_aes(projection = -species, colour = species)) |>
  tour_path(grand_tour(3), fps = 60) |>
  show_scatter(alpha = 0.7, axes = FALSE)

---

Generate a tour path for a detour object

Description

This function takes a detour object as an input, and generates a sequence of projection matrices for the tour. The return value is another detour object with the tour path and other metadata attached. This can then be passed to a ⁠show_*⁠#' function to generate the detour visualisation.

Usage

tour_path(
  x,
  tour_path = grand_tour(2),
  start = NULL,
  aps = 1,
  fps = 30,
  max_bases = 10
)

Arguments

x	a detour object
tour_path	tour path generator, defaults to 2d grand tour
start	projection to start at, if not specified, uses default associated with tour path
aps	target angular velocity (in radians per second)
fps	target frames per second
max_bases	the maximum number of bases to generate

Value

A detour object containing information about the tour path and its parameters

---