R/visualizeParetoOptimalFront.R

In smoof: Single and Multi-Objective Optimization Test Functions

#' @title
#' Pareto-optimal front visualization.
#'
#' @description
#' Quickly visualize the Pareto-optimal front of a bi-criteria objective
#' function by calling the EMOA \code{\link[mco]{nsga2}} and extracting the
#' approximated Pareto-optimal front.
#'
#' @param fn [\code{smoof_multi_objective_function}]\cr
#'   Multi-objective smoof function.
#' @param ... [any]\cr
#'   Arguments passed to \code{\link[mco]{nsga2}}.
#' @examples
#' # Here we visualize the Pareto-optimal front of the bi-objective ZDT3 function
#' fn = makeZDT3Function(dimensions = 3L)
#' vis = visualizeParetoOptimalFront(fn)
#'
#' # Alternatively we can pass some more algorithm parameters to the NSGA2 algorithm
#' vis = visualizeParetoOptimalFront(fn, popsize = 1000L)
#' @return [\code{\link[ggplot2]{ggplot}}]
#' @export
visualizeParetoOptimalFront = function(fn, ...) {
  n.objectives = getNumberOfObjectives(fn)
  if (!isMultiobjective(fn)) {
    stopf("Visualization of approximated Pareto-optimal front only possible fo multi-objective
      functions with two objectives at the moment.")
  }

  if (!requireNamespace("mco", quietly = TRUE))
    stopf("Package \"mco\" needed for this function to work.")

  par.set = ParamHelpers::getParamSet(fn)

  # get approximated Pareto front
  res = mco::nsga2(fn,
    idim = getNumberOfParameters(fn),
    odim = n.objectives,
    lower.bounds = getLower(par.set),
    upper.bounds = getUpper(par.set),
    ...
  )
  eff.points = res$value[res$pareto.optimal, ]

  # transform to ggplot-friendly format
  eff.points = as.data.frame(eff.points)
  colnames(eff.points) = c("f1", "f2")

  pl = ggplot(eff.points, mapping = aes_string(x = "f1", y = "f2"))
  pl = pl + geom_line(colour = "darkgray")
  pl = pl + xlab(expression(f[1])) + ylab(expression(f[2]))
  pl = pl + ggtitle(sprintf("Objective space with shape of Pareto-optimal\n
    front for the bi-criteria %s", getName(fn)))
  return(pl)
}