toPlot: toPlot

In tudob/wfg: WFG toolkit implementation package

Description

A visualization used alongside the WFG package. It shows for each point in the discretized search-space which objective-values will be realized by it (as color).
At the moment the first two search-space dimensions are selected (this should be parameterized in the future).
It is possible to show an RGB- or Red-Blue-plot (chosen because of common red-green blindness).
Red-Blue: Red shows a large value in the first objective, Blue shows a large value in the second objective.
Green dots show the non-dominated individuals of an optimization algorithm (which can internally run multiple times), it is surprisingly instructive.
toPlot22 is a wrapper taking a wfg spec instead of an arbitrary function.

Usage

toPlot(func, ranked = TRUE, inDim = 5, opt.algo = NA, ...)

toPlot22(spec, ranked = TRUE, opt.algo = NA, ...)

Arguments

func	The function to be shown
ranked	Decides whether the colors are chosen based on ranked objective-values. This is recommended to distinguish colors but it does not allow interpretation of their absolute values.
inDim	Number of search-space dimensions; needed for the optimization algorithm
opt.algo	An optimization algorithm may be specified. Its non-dominated individuals are shown.
...	Parameters to be passed through to the plot function
spec	In plot22 - the spec for the test function to plot

Value

nothing

Examples

toPlot(wfgWrap(2, c(tDecept, aperture = 0.25, sLinear) ) )

toPlot22( c(tFlat, from=0.4, to=0.9, sLinear) )

toPlot22( c(tMulti, hill.size=0, num.minima=100, sLinear) )

toPlot22( c(tNonsep, degree=10, sLinear) )

# finally an example of showing an optimization algorithm on the objective-landscape:
# (the points returned by the algo will be filtered so only the non-dominated individuals are shown)
# install.packages("mco")
library(mco)
nsga2wrap = function(func, numEvals=1000) {
  # divide numEvals into popsize*generations:
  popsize = floor(sqrt(numEvals)/4)*4
  generations = floor(numEvals/popsize)
  inDim = 2
  return ( nsga2(func, inDim, 2, lower.bounds=rep(0, 5), upper.bounds=rep(1, 5),
                 popsize=popsize, generations=generations) )
}
toPlot22( c(tFlat, from=0.4, to=0.9, sLinear), opt=nsga2wrap)
toPlot22( c(tDecept, aperture = 0.25, sLinear), opt=nsga2wrap)

tudob/wfg documentation built on June 1, 2019, 2:54 a.m.