Type:	Package
Title:	Graphical Visualizations for Multi-Objective Optimization
Version:	0.1.1
Description:	Visualization of multi-dimensional data arising in multi-objective optimization, including plots of the empirical attainment function (EAF), M. López-Ibáñez, L. Paquete, and T. Stützle (2010) <doi:10.1007/978-3-642-02538-9_9>, and symmetric Vorob'ev expectation and deviation, M. Binois, D. Ginsbourger, O. Roustant (2015) <doi:10.1016/j.ejor.2014.07.032>, among others.
Depends:	R (≥ 4.0)
Imports:	Rdpack, collapse (≥ 2.0.8), grDevices, graphics, matrixStats, moocore (≥ 0.1.5)
Suggests:	extrafont, viridisLite, spelling, testthat (≥ 3.0.0), withr
License:	LGPL-2 | LGPL-2.1 | LGPL-3 [expanded from: LGPL (≥ 2)]
BugReports:	https://github.com/multi-objective/mooplot/issues/
URL:	https://multi-objective.github.io/mooplot/r/, https://github.com/multi-objective/mooplot/
LazyLoad:	true
Encoding:	UTF-8
RoxygenNote:	7.3.2
RdMacros:	Rdpack
Config/testthat/edition:	3
Language:	en-GB
NeedsCompilation:	no
Packaged:	2025-05-11 15:12:52 UTC; manu
Author:	Manuel López-Ibáñez [aut, cre], Carlos Fonseca [ctb], Luís Paquete [ctb], Mickaël Binois [ctb]
Maintainer:	Manuel López-Ibáñez <manuel.lopez-ibanez@manchester.ac.uk>
Repository:	CRAN
Date/Publication:	2025-05-12 10:20:02 UTC

mooplot: Graphical Visualizations for Multi-Objective Optimization

Description

Visualization of multi-dimensional data arising in multi-objective optimization, including plots of the empirical attainment function (EAF), M. López-Ibáñez, L. Paquete, and T. Stützle (2010) doi: 10.1007/978-3-642-02538-9_9, and symmetric Vorob'ev expectation and deviation, M. Binois, D. Ginsbourger, O. Roustant (2015) doi: 10.1016/j.ejor.2014.07.032, among others.

Author(s)

Maintainer: Manuel López-Ibáñez manuel.lopez-ibanez@manchester.ac.uk (ORCID)

Other contributors:

• Carlos Fonseca [contributor]

• Luís Paquete [contributor]

• Mickaël Binois [contributor]

See Also

Useful links:

• https://multi-objective.github.io/mooplot/r/

• https://github.com/multi-objective/mooplot/

• Report bugs at https://github.com/multi-objective/mooplot/issues/

Interactively choose according to empirical attainment function differences

Description

Creates the same plot as eafdiffplot() but waits for the user to click in one of the sides. Then it returns the rectangles the give the differences in favour of the chosen side. These rectangles may be used for interactive decision-making as shown in Diaz and López-Ibáñez (2021). The function moocore::choose_eafdiff() may be used in a non-interactive context.

Usage

choose_eafdiffplot(
  data_left,
  data_right,
  intervals = 5,
  maximise = c(FALSE, FALSE),
  title_left = deparse(substitute(data_left)),
  title_right = deparse(substitute(data_right)),
  ...
)

Arguments

Value

matrix() where the first 4 columns give the coordinates of two corners of each rectangle and the last column. In both cases, the last column gives the positive differences in favor of the chosen side.

References

Juan Esteban Diaz, Manuel López-Ibáñez (2021). “Incorporating Decision-Maker's Preferences into the Automatic Configuration of Bi-Objective Optimisation Algorithms.” European Journal of Operational Research, 289(3), 1209–1222. doi: 10.1016/j.ejor.2020.07.059.

See Also

moocore::read_datasets(), eafdiffplot(), moocore::whv_rect()

Examples

library(moocore)
extdata_dir <- system.file(package="moocore", "extdata")
A1 <- read_datasets(file.path(extdata_dir, "wrots_l100w10_dat"))
A2 <- read_datasets(file.path(extdata_dir, "wrots_l10w100_dat"))
if (interactive()) {
  rectangles <- choose_eafdiffplot(A1, A2, intervals = 5)
} else { # Choose A1
  rectangles <- eafdiff(A1, A2, intervals = 5, rectangles = TRUE)
  rectangles <- choose_eafdiff(rectangles, left = TRUE)
}
reference <- c(max(A1[, 1], A2[, 1]), max(A1[, 2], A2[, 2]))
x <- split.data.frame(A1[,1:2], A1[,3])
hv_A1 <- sapply(split.data.frame(A1[, 1:2], A1[, 3]),
                 hypervolume, reference=reference)
hv_A2 <- sapply(split.data.frame(A2[, 1:2], A2[, 3]),
                 hypervolume, reference=reference)
boxplot(list(A1=hv_A1, A2=hv_A2), main = "Hypervolume")

whv_A1 <- sapply(split.data.frame(A1[, 1:2], A1[, 3]),
                 whv_rect, rectangles=rectangles, reference=reference)
whv_A2 <- sapply(split.data.frame(A2[, 1:2], A2[, 3]),
                 whv_rect, rectangles=rectangles, reference=reference)
boxplot(list(A1=whv_A1, A2=whv_A2), main = "Weighted hypervolume")

Plot empirical attainment function differences

Description

Plot the differences between the empirical attainment functions (EAFs) of two data sets as a two-panel plot, where the left side shows the values of the left EAF minus the right EAF and the right side shows the differences in the other direction.

Usage

eafdiffplot(
  data_left,
  data_right,
  col = c("#FFFFFF", "#808080", "#000000"),
  intervals = 5L,
  percentiles = 50,
  full.eaf = FALSE,
  type = "area",
  legend.pos = if (full.eaf) "bottomleft" else "topright",
  maximise = c(FALSE, FALSE),
  title_left,
  title_right,
  xlim = NULL,
  ylim = NULL,
  cex = par("cex"),
  cex.lab = par("cex.lab"),
  cex.axis = par("cex.axis"),
  grand.lines = TRUE,
  sci.notation = FALSE,
  left.panel.last = NULL,
  right.panel.last = NULL,
  ...
)

Arguments

Details

This function calculates the differences between the EAFs of two data sets, and plots on the left the differences in favour of the left data set, and on the right the differences in favour of the right data set. By default, it also plots the grand best and worst attainment surfaces, that is, the 0%- and 100%-attainment surfaces over all data. These two surfaces delimit the area where differences may exist. In addition, it also plots the 50%-attainment surface of each data set.

With type = "point", only the points where there is a change in the value of the EAF difference are plotted. This means that for areas where the EAF differences stays constant, the region will appear in white even if the value of the differences in that region is large. This explains "white holes" surrounded by black points.

With type = "area", the area where the EAF differences has a certain value is plotted. The idea for the algorithm to compute the areas was provided by Carlos M. Fonseca. The implementation uses R polygons, which some PDF viewers may have trouble rendering correctly (See https://cran.r-project.org/doc/FAQ/R-FAQ.html#Why-are-there-unwanted-borders). Plots (should) look correct when printed.

Large differences that appear when using type = "point" may seem to disappear when using type = "area". The explanation is the points size is independent of the axes range, therefore, the plotted points may seem to cover a much larger area than the actual number of points. On the other hand, the areas size is plotted with respect to the objective space, without any extra borders. If the range of an area becomes smaller than one-pixel, it won't be visible. As a consequence, zooming in or out certain regions of the plots does not change the apparent size of the points, whereas it affects considerably the apparent size of the areas.

Value

Returns a representation of the EAF differences (invisibly).

References

Viviane Grunert da Fonseca, Carlos M. Fonseca, Andreia O. Hall (2001). “Inferential Performance Assessment of Stochastic Optimisers and the Attainment Function.” In Eckart Zitzler, Kalyanmoy Deb, Lothar Thiele, Carlos A. Coello Coello, David Corne (eds.), Evolutionary Multi-criterion Optimization, EMO 2001, volume 1993 of Lecture Notes in Computer Science, 213–225. Springer, Berlin~/ Heidelberg. doi: 10.1007/3-540-44719-9_15.

Manuel López-Ibáñez, Luís Paquete, Thomas Stützle (2010). “Exploratory Analysis of Stochastic Local Search Algorithms in Biobjective Optimization.” In Thomas Bartz-Beielstein, Marco Chiarandini, Luís Paquete, Mike Preuss (eds.), Experimental Methods for the Analysis of Optimization Algorithms, 209–222. Springer, Berlin~/ Heidelberg. doi: 10.1007/978-3-642-02538-9_9.

See Also

read_datasets() eafplot() pdf_crop()

Examples

## NOTE: The plots in the website look squashed because of how pkgdown
## generates them. They should look fine when you generate them yourself.
extdata_dir <- system.file(package="moocore", "extdata")
A1 <- read_datasets(file.path(extdata_dir, "ALG_1_dat.xz"))
A2 <- read_datasets(file.path(extdata_dir, "ALG_2_dat.xz"))
# These take time
eafdiffplot(A1, A2, full.eaf = TRUE)
if (requireNamespace("viridisLite", quietly=TRUE)) {
  viridis_r <- function(n) viridisLite::viridis(n, direction=-1)
  eafdiffplot(A1, A2, type = "area", col = viridis_r)
} else {
  eafdiffplot(A1, A2, type = "area")
}
A1 <- read_datasets(file.path(extdata_dir, "wrots_l100w10_dat"))
A2 <- read_datasets(file.path(extdata_dir, "wrots_l10w100_dat"))
eafdiffplot(A1, A2, type = "point", sci.notation = TRUE, cex.axis=0.6)

# A more complex example
DIFF <- eafdiffplot(A1, A2, col = c("white", "blue", "red"), intervals = 5,
                    type = "point",
                    title_left=expression("W-RoTS," ~ lambda==100 * "," ~ omega==10),
                    title_right=expression("W-RoTS," ~ lambda==10 * "," ~ omega==100),
                    right.panel.last={
                      abline(a = 0, b = 1, col = "red", lty = "dashed")})

## Save the values to a file.
# DIFF$right[,3] <- -DIFF$right[,3]
# write.table(rbind(DIFF$left,DIFF$right),
#             file = "wrots_l100w10_dat-wrots_l10w100_dat-diff.txt",
#             quote = FALSE, row.names = FALSE, col.names = FALSE)

Plot the Empirical Attainment Function for two objectives

Description

Computes and plots the Empirical Attainment Function (EAF), either as attainment surfaces for certain percentiles or as points.

Usage

eafplot(x, ...)

## Default S3 method:
eafplot(
  x,
  sets,
  groups = NULL,
  percentiles = c(0, 50, 100),
  attsurfs = NULL,
  maximise = c(FALSE, FALSE),
  type = "point",
  xlab = NULL,
  ylab = NULL,
  xlim = NULL,
  ylim = NULL,
  log = "",
  col = NULL,
  lty = c("dashed", "solid", "solid", "solid", "dashed"),
  lwd = 1.75,
  pch = NA,
  cex.pch = par("cex"),
  las = par("las"),
  legend.pos = paste0(ifelse(maximise[1L], "bottom", "top"), ifelse(rep_len(maximise,
    2L)[2L], "left", "right")),
  legend.txt = NULL,
  extra.points = NULL,
  extra.legend = NULL,
  extra.pch = 4:25,
  extra.lwd = 0.5,
  extra.lty = NA,
  extra.col = "black",
  xaxis.side = "below",
  yaxis.side = "left",
  axes = TRUE,
  sci.notation = FALSE,
  ...
)

## S3 method for class 'list'
eafplot(x, ...)

Arguments

Details

This function can be used to plot random sets of points like those obtained by different runs of biobjective stochastic optimisation algorithms (López-Ibáñez et al. 2010). An EAF curve represents the boundary separating points that are known to be attainable (that is, dominated in Pareto sense) in at least a fraction (quantile) of the runs from those that are not (Grunert da Fonseca et al. 2001). The median EAF represents the curve where the fraction of attainable points is 50%. In single objective optimisation the function can be used to plot the profile of solution quality over time of a collection of runs of a stochastic optimizer (López-Ibáñez et al. 2025).

Value

The attainment surfaces computed (invisibly).

Methods (by class)

• eafplot(default): Main function

• eafplot(list): List interface for lists of data.frames or matrices

References

Viviane Grunert da Fonseca, Carlos M. Fonseca, Andreia O. Hall (2001). “Inferential Performance Assessment of Stochastic Optimisers and the Attainment Function.” In Eckart Zitzler, Kalyanmoy Deb, Lothar Thiele, Carlos A. Coello Coello, David Corne (eds.), Evolutionary Multi-criterion Optimization, EMO 2001, volume 1993 of Lecture Notes in Computer Science, 213–225. Springer, Berlin~/ Heidelberg. doi: 10.1007/3-540-44719-9_15.

Manuel López-Ibáñez, Luís Paquete, Thomas Stützle (2010). “Exploratory Analysis of Stochastic Local Search Algorithms in Biobjective Optimization.” In Thomas Bartz-Beielstein, Marco Chiarandini, Luís Paquete, Mike Preuss (eds.), Experimental Methods for the Analysis of Optimization Algorithms, 209–222. Springer, Berlin~/ Heidelberg. doi: 10.1007/978-3-642-02538-9_9.

Manuel López-Ibáñez, Diederick Vermetten, Johann Dreo, Carola Doerr (2025). “Using the Empirical Attainment Function for Analyzing Single-objective Black-box Optimization Algorithms.” IEEE Transactions on Evolutionary Computation. doi: 10.1109/TEVC.2024.3462758.

See Also

moocore::read_datasets() eafdiffplot() pdf_crop()

Examples

extdata_path <- system.file(package = "moocore", "extdata")
A1 <- read_datasets(file.path(extdata_path, "ALG_1_dat.xz"))
A2 <- read_datasets(file.path(extdata_path, "ALG_2_dat.xz"))
eafplot(A1, percentiles = 50, sci.notation = TRUE, cex.axis=0.6)
# The attainment surfaces are returned invisibly.
attsurfs <- eafplot(list(A1 = A1, A2 = A2), percentiles = 50)
str(attsurfs)

## Save as a PDF file.
# dev.copy2pdf(file = "eaf.pdf", onefile = TRUE, width = 5, height = 4)


## Using extra.points

data(HybridGA, package="moocore")
data(SPEA2relativeVanzyl, package="moocore")
eafplot(SPEA2relativeVanzyl, percentiles = c(25, 50, 75),
        xlab = expression(C[E]), ylab = "Total switches", xlim = c(320, 400),
        extra.points = HybridGA$vanzyl, extra.legend = "Hybrid GA")

data(SPEA2relativeRichmond, package="moocore")
eafplot (SPEA2relativeRichmond, percentiles = c(25, 50, 75),
         xlab = expression(C[E]), ylab = "Total switches",
         xlim = c(90, 140), ylim = c(0, 25),
         extra.points = HybridGA$richmond, extra.lty = "dashed",
         extra.legend = "Hybrid GA")

eafplot (SPEA2relativeRichmond, percentiles = c(25, 50, 75),
         xlab = expression(C[E]), ylab = "Total switches",
         xlim = c(90, 140), ylim = c(0, 25), type = "area",
         extra.points = HybridGA$richmond, extra.lty = "dashed",
         extra.legend = "Hybrid GA", legend.pos = "bottomright")

data(SPEA2minstoptimeRichmond, package="moocore")
SPEA2minstoptimeRichmond[,2] <- SPEA2minstoptimeRichmond[,2] / 60
eafplot (SPEA2minstoptimeRichmond, xlab = expression(C[E]),
         ylab = "Minimum idle time (minutes)", maximise = c(FALSE, TRUE),
         las = 1, log = "y", main = "SPEA2 (Richmond)",
         legend.pos = "bottomright")

data(tpls50x20_1_MWT, package="moocore")
eafplot(tpls50x20_1_MWT[, c(2,3)], sets = tpls50x20_1_MWT[,4L],
        groups = tpls50x20_1_MWT[["algorithm"]])

Remove whitespace margins from a PDF file (and maybe embed fonts)

Description

Remove whitespace margins using https://ctan.org/pkg/pdfcrop and optionally embed fonts using grDevices::embedFonts(). You may install pdfcrop using TinyTeX (https://cran.r-project.org/package=tinytex) with tinytex::tlmgr_install('pdfcrop').

Usage

pdf_crop(
  filename,
  mustWork = FALSE,
  pdfcrop = Sys.which("pdfcrop"),
  embed_fonts = FALSE
)

Arguments

Details

You may also wish to consider extrafont::embed_fonts() (https://cran.r-project.org/package=extrafont).

library(extrafont)
# If you need to specify the path to Ghostscript (probably not needed in Linux)
Sys.setenv(R_GSCMD = "C:/Program Files/gs/gs9.56.1/bin/gswin64c.exe")
embed_fonts("original.pdf", outfile = "new.pdf")

As an alternative, saving the PDF with grDevices::cairo_pdf() should already embed the fonts.

Value

No return value, called for side effects

See Also

grDevices::embedFonts() extrafont::embed_fonts() grDevices::cairo_pdf()

Examples

extdata_path <- system.file(package = "moocore", "extdata")
A1 <- read_datasets(file.path(extdata_path, "wrots_l100w10_dat"))
A2 <- read_datasets(file.path(extdata_path, "wrots_l10w100_dat"))
filename <- tempfile("eafplot", fileext=".pdf")
pdf(file = filename, onefile = TRUE, width = 5, height = 4)
eafplot(list(A1 = A1, A2 = A2), percentiles = 50, sci.notation = TRUE)
dev.off()
try(pdf_crop(filename)) # This may fail if pdfcrop is not installed.

Objects exported from other packages

Description

These objects are imported from other packages. Follow the links below to see their documentation.

moocore

read_datasets

Plot the symmetric deviation function.

Description

The symmetric deviation function is the probability for a given target in the objective space to belong to the symmetric difference between the Vorob'ev expectation and a realization of the (random) attained set.

Usage

symdevplot(
  x,
  sets,
  ve,
  threshold,
  nlevels = 11,
  ve.col = "blue",
  xlim = NULL,
  ylim = NULL,
  legend.pos = "topright",
  main = "Symmetric deviation function",
  col.fun = function(n) gray(seq(0, 0.9, length.out = n)^2)
)

Arguments

Value

No return value, called for side effects

Author(s)

Mickael Binois

References

M Binois, D Ginsbourger, O Roustant (2015). “Quantifying uncertainty on Pareto fronts with Gaussian process conditional simulations.” European Journal of Operational Research, 243(2), 386–394. doi: 10.1016/j.ejor.2014.07.032.

C. Chevalier (2013), Fast uncertainty reduction strategies relying on Gaussian process models, University of Bern, PhD thesis.

Ilya Molchanov (2005). Theory of Random Sets. Springer.

See Also

moocore::vorob_t() moocore::vorob_dev() eafplot()

Examples

data(CPFs, package = "moocore")
res <- moocore::vorob_t(CPFs, reference = c(2, 200))
print(res$threshold)

## Display Vorob'ev expectation and attainment function
# First style
eafplot(CPFs[,1:2], sets = CPFs[,3], percentiles = c(0, 25, 50, 75, 100, res$threshold),
        main = substitute(paste("Empirical attainment function, ",beta,"* = ", a, "%"),
                          list(a = formatC(res$threshold, digits = 2, format = "f"))))

# Second style
eafplot(CPFs[,1:2], sets = CPFs[,3], percentiles = c(0, 20, 40, 60, 80, 100),
        col = gray(seq(0.8, 0.1, length.out = 6)^0.5), type = "area",
        legend.pos = "bottomleft", extra.points = res$ve, extra.col = "cyan",
        extra.legend = "VE", extra.lty = "solid", extra.pch = NA, extra.lwd = 2,
        main = substitute(paste("Empirical attainment function, ",beta,"* = ", a, "%"),
                          list(a = formatC(res$threshold, digits = 2, format = "f"))))
# Vorob'ev deviation
VD <- moocore::vorob_dev(CPFs, reference = c(2, 200), ve = res$ve)
# Display the symmetric deviation function.
symdevplot(CPFs, ve = res$ve, threshold = res$threshold, nlevels = 11)
# Levels are adjusted automatically if too large.
symdevplot(CPFs, ve = res$ve, threshold = res$threshold, nlevels = 200, legend.pos = "none")

# Use a different palette.
symdevplot(CPFs, ve = res$ve, threshold = res$threshold, nlevels = 11, col.fun = heat.colors)