| Title: | Facilitate Exploration of touRR optimisatioN |
| Version: | 0.1.0 |
| Description: | Diagnostic plots for optimisation, with a focus on projection pursuit. These show paths the optimiser takes in the high-dimensional space in multiple ways: by reducing the dimension using principal component analysis, and also using the tour to show the path on the high-dimensional space. Several botanical colour palettes are included, reflecting the name of the package. A paper describing the methodology can be found at https://journal.r-project.org/archive/2021/RJ-2021-105/index.html. |
| License: | MIT + file LICENSE |
| Encoding: | UTF-8 |
| LazyData: | true |
| URL: | https://github.com/huizezhang-sherry/ferrn/ |
| BugReports: | https://github.com/huizezhang-sherry/ferrn/issues |
| Imports: | rlang (≥ 0.1.2), dplyr, magrittr, scales, gganimate, ggplot2, tibble, purrr, tourr, stringr, ggrepel, ggforce, tidyr, cli, progress, glue, GpGp, |
| RoxygenNote: | 7.3.1 |
| Depends: | R (≥ 2.10) |
| Suggests: | roxygen2, covr, pkgdown, testthat, forcats, patchwork, future.apply, |
| Language: | en-GB |
| NeedsCompilation: | no |
| Packaged: | 2024-06-23 18:20:02 UTC; hzha400 |
| Author: | H. Sherry Zhang 
[aut, cre],
Dianne Cook [aut],
Ursula Laa [aut],
Nicolas Langrené
[aut],
Patricia Menéndez
[aut] |
| Maintainer: | H. Sherry Zhang <huizezhangsh@gmail.com> |
| Repository: | CRAN |
| Date/Publication: | 2024-06-23 22:30:07 UTC |
ferrn: Facilitate Exploration of touRR optimisatioN
Description
Diagnostic plots for optimisation, with a focus on projection pursuit. These show paths the optimiser takes in the high-dimensional space in multiple ways: by reducing the dimension using principal component analysis, and also using the tour to show the path on the high-dimensional space. Several botanical colour palettes are included, reflecting the name of the package. A paper describing the methodology can be found at https://journal.r-project.org/archive/2021/RJ-2021-105/index.html.
Author(s)
Maintainer: H. Sherry Zhang huizezhangsh@gmail.com (ORCID)
Authors:
Dianne Cook dicook@monash.edu (ORCID)
Ursula Laa ursula.laa@boku.ac.at (ORCID)
Nicolas Langrené nicolaslangrene@uic.edu.cn (ORCID)
Patricia Menéndez patricia.menendez@unimelb.edu.au (ORCID)
See Also
Useful links:
Report bugs at https://github.com/huizezhang-sherry/ferrn/issues
Pipe operator
Description
See magrittr::%>% for details.
Usage
lhs %>% rhs
Value
this is a magrittr pipe!
A ggproto for drawing anchor points
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_anchor(dt, anchor_size = 3, anchor_alpha = 0.5, anchor_color = NULL, ...)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
anchor_size |
numeric; the size of the anchor points |
anchor_alpha |
numeric; the alpha of the anchor points |
anchor_color |
the variable to be coloured by |
... |
other aesthetics inherent from |
Value
a wrapper for drawing anchor points in explore_space_pca()
See Also
Other draw functions:
add_anno(),
add_dir_search(),
add_end(),
add_interp(),
add_interp_last(),
add_interrupt(),
add_search(),
add_space(),
add_start(),
add_theo()
A ggproto for annotating the symmetry of the starting points
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_anno(dt, anno_color = "black", anno_lty = "dashed", anno_alpha = 0.1, ...)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
anno_color |
character; the colour of the annotation line |
anno_lty |
character; the linetype of the annotation line |
anno_alpha |
numeric; the alpha of the annotation line |
... |
other aesthetics inherent from |
Value
a wrapper for annotating the symmetry of start points in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_dir_search(),
add_end(),
add_interp(),
add_interp_last(),
add_interrupt(),
add_search(),
add_space(),
add_start(),
add_theo()
A ggproto for drawing directional search points
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_dir_search(dt, dir_size = 0.5, dir_alpha = 0.5, dir_color = NULL, ...)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
dir_size |
numeric; the size of the directional search points in pseudo derivative search |
dir_alpha |
numeric; the alpha of the directional search points in pseudo derivative search |
dir_color |
the variable to be coloured by |
... |
other aesthetics inherent from |
Value
a wrapper for drawing directional search points (used in pseudo derivative search) with buffer in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_end(),
add_interp(),
add_interp_last(),
add_interrupt(),
add_search(),
add_space(),
add_start(),
add_theo()
A ggproto for drawing start points
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_end(dt, end_size = 5, end_alpha = 1, end_color = NULL, ...)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
end_size |
numeric; the size of the end point |
end_alpha |
numeric; the alpha of the end point |
end_color |
the variable to be coloured by |
... |
other aesthetics inherent from |
Value
a wrapper for drawing end points in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_dir_search(),
add_interp(),
add_interp_last(),
add_interrupt(),
add_search(),
add_space(),
add_start(),
add_theo()
A ggproto for drawing interpolation path
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_interp(
dt,
interp_size = 1.5,
interp_alpha = NULL,
interp_color = NULL,
interp_group = NULL,
...
)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
interp_size |
numeric; the size of the interpolation path |
interp_alpha |
numeric; the alpha of the interpolation path |
interp_color |
the variable to be coloured by |
interp_group |
the variable to label different interpolation path |
... |
other aesthetics inherent from |
Value
a wrapper for drawing the interpolation points in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_dir_search(),
add_end(),
add_interp_last(),
add_interrupt(),
add_search(),
add_space(),
add_start(),
add_theo()
A ggproto for drawing finish points
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_interp_last(
dt,
interp_last_size = 3,
interp_last_alpha = 1,
interp_last_color = NULL,
...
)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
interp_last_size |
numeric; the size of the last interpolation points in each iteration |
interp_last_alpha |
numeric; the alpha of the last interpolation points in each iteration |
interp_last_color |
the variable to be coloured by |
... |
other aesthetics inherent from |
Value
a wrapper for drawing the last interpolation points of each iteration in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_dir_search(),
add_end(),
add_interp(),
add_interrupt(),
add_search(),
add_space(),
add_start(),
add_theo()
A ggproto for annotating the interrupted path
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_interrupt(
dt,
interrupt_size = 0.5,
interrupt_alpha = NULL,
interrupt_color = NULL,
interrupt_group = NULL,
interrupt_linetype = "dashed",
...
)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
interrupt_size |
numeric; the size of the interruption path |
interrupt_alpha |
numeric; the alpha of the interruption path |
interrupt_color |
the variable to be coloured by |
interrupt_group |
the variable to label different interruption |
interrupt_linetype |
character; the linetype to annotate the interruption |
... |
other aesthetics inherent from |
Value
a wrapper for annotating the interruption in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_dir_search(),
add_end(),
add_interp(),
add_interp_last(),
add_search(),
add_space(),
add_start(),
add_theo()
A ggproto for drawing search points
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_search(dt, search_size = 0.5, search_alpha = 0.5, search_color = NULL, ...)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
search_size |
numeric; the size of the search points |
search_alpha |
numeric; the alpha of the anchor points |
search_color |
the variable to be coloured by |
... |
other aesthetics inherent from |
Value
a wrapper for drawing search points in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_dir_search(),
add_end(),
add_interp(),
add_interp_last(),
add_interrupt(),
add_space(),
add_start(),
add_theo()
A ggproto for drawing circle
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_space(
dt,
space_alpha = 0.5,
space_fill = "grey92",
space_color = "white",
cent_size = 1,
cent_alpha = 1,
cent_color = "black",
...
)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
space_alpha |
numeric; the alpha of the basis space |
space_fill |
character; the colour of the space filling |
space_color |
character; the colour of the space brim |
cent_size |
numeric; the size of the centre point |
cent_alpha |
numeric; an alpha of the centre point |
cent_color |
character; the colour of the centre point |
... |
other aesthetics inherent from |
Value
a wrapper for drawing the space in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_dir_search(),
add_end(),
add_interp(),
add_interp_last(),
add_interrupt(),
add_search(),
add_start(),
add_theo()
Examples
library(ggplot2)
space <- tibble::tibble(x0 = 0, y0 = 0, r = 5)
ggplot() +
add_space(space) +
theme_void() +
theme(aspect.ratio = 1)
A ggproto for drawing start points
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_start(dt, start_size = 5, start_alpha = 1, start_color = NULL, ...)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
start_size |
numeric; the size of start point |
start_alpha |
numeric; the alpha of start point |
start_color |
the variable to be coloured by |
... |
other aesthetics inherent from |
Value
a wrapper for drawing start points in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_dir_search(),
add_end(),
add_interp(),
add_interp_last(),
add_interrupt(),
add_search(),
add_space(),
add_theo()
Examples
library(ggplot2)
# construct the space and start df for plotting
space <- tibble::tibble(x0 = 0, y0 = 0, r = 5)
holes_1d_geo %>%
compute_pca() %>%
purrr::pluck("aug") %>%
clean_method() %>%
get_start()
A ggproto for drawing the theoretical basis, if applicable
Description
This is a wrapper function used by explore_space_pca() and
should be be called directly by the user
Usage
add_theo(
dt,
theo_label = "*",
theo_size = 25,
theo_alpha = 0.8,
theo_color = "#000000",
...
)
Arguments
dt |
A data object from the running the optimisation algorithm in guided tour |
theo_label |
character; a symbol to label the theoretical point |
theo_size |
numeric; the size of the theoretical point |
theo_alpha |
numeric; the alpha of the theoretical point |
theo_color |
character; the colour of the theoretical point in hex |
... |
other aesthetics inherent from |
Value
a wrapper for drawing theoretical points in explore_space_pca()
See Also
Other draw functions:
add_anchor(),
add_anno(),
add_dir_search(),
add_end(),
add_interp(),
add_interp_last(),
add_interrupt(),
add_search(),
add_space(),
add_start()
Bind random bases in the projection bases space
Description
Given the orthonormality constraint, the projection bases live in a high dimensional hollow sphere. Generating random points on the sphere is useful to perceive the data object in the high dimensional space.
Usage
bind_random(dt, n = 500, seed = 1)
Arguments
dt |
a data object collected by the projection pursuit guided tour optimisation in the |
n |
numeric; the number of random bases to generate in each dimension by geozoo |
seed |
numeric; a seed for generating reproducible random bases from geozoo |
Value
a tibble object containing both the searched and random bases
See Also
Other bind:
bind_random_matrix(),
bind_theoretical()
Examples
bind_random(holes_1d_better) %>% tail(5)
Bind random bases in the projection bases space as a matrix
Description
Bind random bases in the projection bases space as a matrix
Usage
bind_random_matrix(basis, n = 500, d = 1, front = FALSE, seed = 1)
Arguments
basis |
a matrix returned by |
n |
numeric; the number of random bases to generate in each dimension by geozoo |
d |
numeric; dimension of the basis, d = 1, 2, ... |
front |
logical; if the random bases should be bound before or after the original bases |
seed |
numeric; a seed for generating reproducible random bases from geozoo |
Value
matrix
a matrix containing both the searched and random bases
See Also
Other bind:
bind_random(),
bind_theoretical()
Examples
data <- get_basis_matrix(holes_1d_geo)
bind_random_matrix(data) %>% tail(5)
Bind the theoretical best record
Description
The theoretical best basis is usually known for a simulated problem. Augment this information into the data object allows for evaluating the performance of optimisation against the theory.
Usage
bind_theoretical(dt, matrix, index, raw_data)
Arguments
dt |
a data object collected by the projection pursuit guided tour optimisation in the |
matrix |
a matrix of the theoretical basis |
index |
the index function used to calculate the index value |
raw_data |
a tibble of the original data used to calculate the index value |
Value
a tibble object containing both the searched and theoretical best bases
See Also
Other bind:
bind_random(),
bind_random_matrix()
Examples
best <- matrix(c(0, 1, 0, 0, 0), nrow = 5)
tail(holes_1d_better %>% bind_theoretical(best, tourr::holes(), raw_data = boa5), 1)
A customised colour palette based on Australian botanies
Description
Available colours in the palettes
Usage
botanical_palettes
botanical_pal(palette = "fern", reverse = FALSE)
Arguments
palette |
Colour palette from the botanical_palette |
reverse |
logical, if the colour should be reversed |
Format
An object of class list of length 5.
Value
a function for interpolating colour in the botanical palette
Clean method names
Description
Clean method names
Usage
clean_method(dt)
Arguments
dt |
a data object |
Value
a tibble with method cleaned
Examples
head(clean_method(holes_1d_better), 5)
Plot the PCA projection of the projection bases space
Description
Plot the PCA projection of the projection bases space
Usage
explore_space_start(dt, group = NULL, pca = TRUE, ...)
explore_space_end(dt, group = NULL, pca = TRUE, ...)
explore_space_pca(
dt,
details = FALSE,
pca = TRUE,
group = NULL,
color = NULL,
facet = NULL,
...,
animate = FALSE
)
Arguments
dt |
a data object collected by the projection pursuit guided tour optimisation in |
group |
the variable to label different runs of the optimiser(s) |
pca |
logical; if PCA coordinates need to be computed for the data |
... |
other arguments passed to |
details |
logical; if components other than start, end and interpolation need to be shown |
color |
the variable to be coloured by |
facet |
the variable to be faceted by |
animate |
logical; if the interpolation path needs to be animated |
Value
a ggplot2 object
See Also
Other main plot functions:
explore_space_tour(),
explore_trace_interp(),
explore_trace_search()
Examples
dplyr::bind_rows(holes_1d_geo, holes_1d_better) %>%
bind_theoretical(matrix(c(0, 1, 0, 0, 0), nrow = 5),
index = tourr::holes(), raw_data = boa5
) %>%
explore_space_pca(group = method, details = TRUE) +
scale_color_discrete_botanical()
## Not run:
best <- matrix(c(0, 1, 0, 0, 0), nrow = 5)
dt <- bind_theoretical(holes_1d_jellyfish, best, tourr::holes(), raw_data = boa5)
explore_space_start(dt)
explore_space_end(dt, group = loop, theo_size = 10, theo_color = "#FF0000")
explore_space_pca(
dt, facet = loop, interp_size = 0.5, theo_size = 10,
start_size = 1, end_size = 3
)
## End(Not run)
Plot the grand tour animation of the bases space in high dimension
Description
Plot the grand tour animation of the bases space in high dimension
Usage
explore_space_tour(..., axes = "bottomleft")
prep_space_tour(
dt,
group = NULL,
flip = FALSE,
n_random = 2000,
color = NULL,
rand_size = 1,
rand_color = "#D3D3D3",
point_size = 1.5,
end_size = 5,
theo_size = 3,
theo_shape = 17,
theo_color = "black",
palette = botanical_palettes$fern,
...
)
Arguments
... |
other argument passed to |
axes |
see [tourr::animate_xy()] |
dt |
a data object collected by the projection pursuit guided tour optimisation in |
group |
the variable to label different runs of the optimiser(s) |
flip |
logical; if the sign flipping need to be performed |
n_random |
numeric; the number of random basis to generate |
color |
the variable to be coloured by |
rand_size |
numeric; the size of random points |
rand_color |
character; the color hex code for random points |
point_size |
numeric; the size of points searched by the optimiser(s) |
end_size |
numeric; the size of end points |
theo_size |
numeric; the size of theoretical point(s) |
theo_shape |
numeric; the shape symbol in the basic plot |
theo_color |
character; the color of theoretical point(s) |
palette |
the colour palette to be used |
Value
explore_space_tour()an animation of the search path in the high-dimensional sphere
prep_space_tour()a list containing various components needed for producing the animation
See Also
Other main plot functions:
explore_space_start(),
explore_trace_interp(),
explore_trace_search()
Examples
explore_space_tour(dplyr::bind_rows(holes_1d_better, holes_1d_geo),
group = method, palette = botanical_palettes$fern[c(1, 6)]
)
Plot the trace the search progression
Description
Trace the index value of search/ interpolation points in guided tour optimisation
Usage
explore_trace_interp(
dt,
iter = NULL,
color = NULL,
group = NULL,
cutoff = 50,
target_size = 3,
interp_size = 1,
accuracy_x = 5,
accuracy_y = 0.01
)
Arguments
dt |
a data object collected by the projection pursuit guided tour optimisation in |
iter |
the variable to be plotted on the x-axis |
color |
the variable to be coloured by |
group |
the variable to label different runs of the optimiser(s) |
cutoff |
numeric; if the number of interpolating points is smaller than |
target_size |
numeric; the size of target points in the interpolation |
interp_size |
numeric; the size of interpolation points |
accuracy_x |
numeric; If the difference of two neighbour x-labels is smaller than |
accuracy_y |
numeric; the precision of y-axis label |
Value
a ggplot object for diagnosing how the index value progresses during the interpolation
See Also
Other main plot functions:
explore_space_start(),
explore_space_tour(),
explore_trace_search()
Examples
# Compare the trace of interpolated points in two algorithms
holes_1d_better %>%
explore_trace_interp(interp_size = 2) +
scale_color_continuous_botanical(palette = "fern")
Plot the count in each iteration
Description
Plot the count in each iteration
Usage
explore_trace_search(
dt,
iter = NULL,
color = NULL,
cutoff = 15,
extend_lower = 0.95,
...
)
Arguments
dt |
a data object collected by the projection pursuit guided tour optimisation in |
iter |
the variable to be plotted on the x-axis |
color |
the variable to be coloured by |
cutoff |
numeric; if the number of searches in one iteration is smaller than |
extend_lower |
a numeric for extending the y-axis to display text labels |
... |
arguments passed into geom_label_repel() for displaying text labels |
Value
a ggplot object for diagnosing how many points the optimiser(s) have searched
See Also
Other main plot functions:
explore_space_start(),
explore_space_tour(),
explore_trace_interp()
Examples
# Summary plots for search points in two algorithms
library(patchwork)
library(dplyr)
library(ggplot2)
p1 <- holes_1d_better %>% explore_trace_search() +
scale_color_continuous_botanical(palette = "fern")
p2 <- holes_2d_better_max_tries %>% explore_trace_search() +
scale_color_continuous_botanical(palette = "daisy")
p1 / p2
Helper functions for 'explore_space_pca()'
Description
Helper functions for 'explore_space_pca()'
Usage
flip_sign(dt, group = NULL, ...)
compute_pca(dt, group = NULL, random = TRUE, flip = TRUE, ...)
Arguments
dt |
a data object collected by the projection pursuit guided tour optimisation in |
group |
the variable to label different runs of the optimiser(s) |
... |
other arguments received from |
random |
logical; if random bases from the basis space need to be added to the data |
flip |
logical; if the sign flipping need to be performed |
Value
flip_sign(): a list containing a matrix of all the bases, a logical
value indicating whether a flip of sign is performed, and a data frame of
the original dataset.
compute_pca(): a list containing the PCA summary and a data frame
with PC coordinates augmented.
Examples
dt <- dplyr::bind_rows(holes_1d_geo, holes_1d_better)
flip_sign(dt, group = method) %>% str(max = 1)
compute_pca(dt, group = method)
Better label formatting to avoid overlapping
Description
Better label formatting to avoid overlapping
Usage
format_label(labels, accuracy)
Arguments
labels |
a numerical vector of labels |
accuracy |
the accuracy of the label |
Value
a vector of adjusted labels
Examples
format_label(c(0.87, 0.87, 0.9, 0.93, 0.95), 0.01)
format_label(c(0.87, 0.87, 0.9, 0.93, 0.95, 0.96, 0.96), 0.01)
Functions to get components from the data collecting object
Description
Functions to get components from the data collecting object
Usage
get_best(dt, group = NULL)
get_start(dt, group = NULL)
get_interp(dt, group = NULL)
get_interp_last(dt, group = NULL)
get_anchor(dt, group = NULL)
get_search(dt)
get_dir_search(dt, ratio = 5, ...)
get_space_param(dt, ...)
get_theo(dt)
get_interrupt(dt, group = NULL, precision = 0.001)
get_search_count(dt, iter = NULL, group = NULL)
get_basis_matrix(dt)
Arguments
dt |
a data object collected by the projection pursuit guided tour optimisation in the |
group |
the variable to label different runs of the optimiser(s) |
ratio |
numeric; a buffer value to deviate directional search points from the anchor points |
... |
other arguments passed to |
precision |
numeric; if the index value of the last interpolating point and the anchor point differ by |
iter |
the variable to be counted by |
Details
get_best: extract the best basis found by the optimiser(s)
get_start: extract the start point of the optimisation
get_interp: extract the interpolation points
get_interp_last: extract the last point in each interpolation
get_anchor: extract the anchor points on the geodesic path
get_search: extract search points in the optimisation (for
search_geodesic)
get_dir_search: extract directional search points (for
search_geodesic)
get_space_param: estimate the radius of the background circle
based on the randomly generated points. The space of projected bases is a
circle when reduced to 2D. A radius is estimated using the largest distance
from the bases in the data object to the centre point.
get_theo: extract the theoretical basis, if exist
get_interrupt: extract the end point of the interpolation and the
target point in the iteration when an interruption happens. The optimiser
can find better basis on the interpolation path, an interruption is
implemented to stop further interpolation from the highest point to the
target point. This discrepancy is highlighted in the PCA plot.
get_search_count: summarise the number of search points in each iteration
get_basis_matrix: extract all the bases as a matrix
Value
a tibble object containing the best basis found by the optimiser(s)
Examples
get_search(holes_1d_geo)
get_anchor(holes_1d_geo)
get_start(holes_1d_better)
get_interrupt(holes_1d_better)
get_interp(holes_1d_better) %>% head()
get_basis_matrix(holes_1d_better) %>% head()
get_best(dplyr::bind_rows(holes_1d_better, holes_1d_geo), group = method)
get_search_count(holes_1d_better)
get_search_count(dplyr::bind_rows(holes_1d_better, holes_1d_geo), group = method)
get_interp_last(holes_1d_better)
get_interp_last(dplyr::bind_rows(holes_1d_better, holes_1d_geo), group = method)
res <- holes_1d_geo %>% compute_pca() %>% purrr::pluck("aug")
get_dir_search(res)
best <- matrix(c(0, 1, 0, 0, 0), nrow = 5)
holes_1d_better %>%
bind_theoretical(best, tourr::holes(), raw_data = boa5) %>%
get_theo()
Data objects collected during the projection pursuit optimisation
Description
Simulated data to demonstrate the usage of four diagnostic plots in the package, users can create their own guided tour data objects and diagnose with the visualisation designed in this package.
Usage
holes_1d_geo
holes_1d_better
holes_1d_jellyfish
holes_2d_better
holes_2d_better_max_tries
Format
An object of class tbl_df (inherits from tbl, data.frame) with 416 rows and 8 columns.
An object of class tbl_df (inherits from tbl, data.frame) with 79 rows and 8 columns.
An object of class tbl_df (inherits from tbl, data.frame) with 2500 rows and 8 columns.
An object of class tbl_df (inherits from tbl, data.frame) with 98 rows and 8 columns.
An object of class tbl_df (inherits from tbl, data.frame) with 1499 rows and 8 columns.
Details
The prefix holes_* indicates the use of holes index in the guided tour.
The suffix *_better/geo/jellyfish indicates the optimiser used:
search_better, search_geodesic, search_jellyfish.
Examples
holes_1d_better %>%
explore_trace_interp(interp_size = 2) +
scale_color_continuous_botanical(palette = "fern")
Plot the projection from the optimisation data collected from projection pursuit
Description
Plot the projection from the optimisation data collected from projection pursuit
Usage
plot_projection(dt, data, cols = NULL)
compute_projection(dt, data, cols = NULL)
Arguments
dt |
a data object collected by the projection pursuit guided tour optimisation in |
data |
the original data |
cols |
additional columns to include in the plot |
Value
a ggplot object
Examples
holes_1d_jellyfish |> get_best() |> plot_projection(data = boa5)
Function to calculate smoothness and squintability
Description
Function to calculate smoothness and squintability
Usage
sample_bases(
idx,
data = sine1000,
n_basis = 300,
parallel = FALSE,
best = matrix(c(0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1), nrow = 6),
min_proj_dist = NA,
step_size = NA,
seed = 123
)
## S3 method for class 'basis_df'
print(x, width = NULL, ...)
## S3 method for class 'basis_df'
tbl_sum(x)
calc_smoothness(
basis_df,
start_params = c(0.001, 0.5, 2, 2),
other_gp_params = NULL,
verbose = FALSE
)
## S3 method for class 'smoothness_res'
print(x, width = NULL, ...)
## S3 method for class 'smoothness_res'
tbl_sum(x)
calc_squintability(
basis_df,
method = c("ks", "nls"),
scale = TRUE,
bin_width = 0.005,
other_params = NULL
)
## S3 method for class 'squintability_res'
print(x, width = NULL, ...)
## S3 method for class 'squintability_res'
tbl_sum(x)
fit_ks(basis_df, idx, other_params = NULL)
fit_nls(basis_df, other_params = NULL)
Arguments
idx |
character, the name of projection pursuit index function, e.g. "holes" |
data |
a matrix or data frame, the high dimensional data to be projected |
n_basis |
numeric, the number of random bases to generate |
parallel |
logic, whether to use parallel computing for calculating the index. Recommend for the stringy index. |
best |
a matrix, the theoretical/ empirical best projection matrix to calculate the projection distance from the simulated random bases. |
min_proj_dist |
only for squintability, the threshold for projection distance for the random basis to be considered in sampling |
step_size |
numeric, step size for interpolating from each random basis to the best basis, recommend 0.005 |
seed |
numeric, seed for sampling random bases |
x |
objects with specialised printing methods |
width |
only used when |
... |
further arguments passed to or from other methods. |
basis_df |
the basis data frame returned from |
start_params |
list, the starting parameters for the Gaussian process for smoothness |
other_gp_params |
list, additional parameters to be passed to [GpGp::fit_model()] for calculating smoothness |
verbose |
logical, whether to print optimisation progression when fitting the Gaussian process |
method |
either "ks" (kernel smoothing) or "nls" (non-linear least square) for calculating squintability. |
scale |
logic, whether to scale the index value to 0-1 in squintability |
bin_width |
numeric, the bin width to average the index value before fitting the kernel, recommend to set as the same as 'step' parameter |
other_params |
list additional parameters for fitting kernel smoothing or non-linear least square, see [stats::ksmooth()] and [stats::nls()] for details |
Examples
## Not run:
library(GpGp)
library(fields)
library(tourr)
basis_smoothness <- sample_bases(idx = "holes")
calc_smoothness(basis_smoothness)
basis_squint <- sample_bases(idx = "holes", n_basis = 100, step_size = 0.01, min_proj_dist = 1.5)
calc_squintability(basis_squint, method = "ks", bin_width = 0.01)
## End(Not run)
continuous scale colour function
Description
continuous scale colour function
Discrete scale colour function
continuous scale fill function
discrete scale fill function
Usage
scale_color_continuous_botanical(palette = "fern", reverse = FALSE, ...)
scale_color_discrete_botanical(palette = "fern", reverse = FALSE, ...)
scale_fill_continuous_botanical(palette = "fern", reverse = FALSE, ...)
scale_fill_discrete_botanical(palette = "fern", reverse = FALSE, ...)
Arguments
palette |
colour palette from the botanical_palette |
reverse |
logical; if the colour should be reversed |
... |
other arguments passed into scale_color_gradientn |
Value
a wrapper for continuous scales in the botanical palette
a wrapper for discrete scales in the botanical palette
a wrapper for continuous fill in the botanical palette
a wrapper for discrete fill in the botanical palette
Simulated sine, pipe, and gaussian mixture
Description
Simulated sine and pipe data for calculating optimisation features. Each dataset has 1000 observations and the last two columns contain the intended structure with the rest being noise.
Usage
sine1000
sine1000_8d
pipe1000
pipe1000_8d
pipe1000_10d
pipe1000_12d
boa
boa5
boa6
Format
An object of class matrix (inherits from array) with 1000 rows and 6 columns.
An object of class matrix (inherits from array) with 1000 rows and 8 columns.
An object of class matrix (inherits from array) with 1000 rows and 6 columns.
An object of class matrix (inherits from array) with 1000 rows and 8 columns.
An object of class matrix (inherits from array) with 1000 rows and 10 columns.
An object of class matrix (inherits from array) with 1000 rows and 12 columns.
An object of class tbl_df (inherits from tbl, data.frame) with 1000 rows and 10 columns.
An object of class tbl_df (inherits from tbl, data.frame) with 1000 rows and 5 columns.
An object of class tbl_df (inherits from tbl, data.frame) with 1000 rows and 6 columns.
Examples
library(ggplot2)
library(tidyr)
library(dplyr)
boa %>%
pivot_longer(cols = x1:x10, names_to = "var", values_to = "value") %>%
mutate(var = forcats::fct_relevel(as.factor(var), paste0("x", 1:10))) %>%
ggplot(aes(x = value)) +
geom_density() +
facet_wrap(vars(var))
sine1000 |> ggplot(aes(x = V5, y = V6)) + geom_point() + theme(aspect.ratio = 1)
pipe1000_8d |> ggplot(aes(x = V5, y = V6)) + geom_point() + theme(aspect.ratio = 1)
pipe1000_8d |> ggplot(aes(x = V7, y = V8)) + geom_point() + theme(aspect.ratio = 1)
A specific theme for trace plots
Description
A specific theme for trace plots
Usage
theme_fern()
Value
a ggplot2 theme for explore_trace_interp()