In pabloOnate/iraceplot: Plots for visualizing the data produced by the irace package

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(iraceplot)

General information

The provided execution of irace statistics are:

• Iterations: r {length(iraceResults$allElites)}
• Configurations: r {nrow(iraceResults$allConfigurations)}
• Instances: r {nrow(iraceResults$experiments)}
• Experiments: r {nrow(iraceResults$experimentLog)}
• Elite configurations: r {length(iraceResults$allElites[[length(iraceResults$allElites)]])}

Elite configurations

The final best configurations found by irace are:

knitr::kable(iraceResults$allConfigurations[iraceResults$allElites[[length(iraceResults$allElites)]],])

The parallel_coord function creates a parallel coordinate plot of a set of selected configurations.

parallel_coord(iraceResults)

Parameter values sampling

The frequency of the parameter values sampled by irace may indicate high performing areas:

sampling_frequency(iraceResults)

Testing performance

The test performance allows to evaluate if the configurations found by irace.

The final elite configuration performance on the test set:

if (iraceResults$scenario$testInstances == "" || is.null(iraceResults$scenario$testInstances)) {
  cat("No test performed.\n")
} else {
  boxplot_test(iraceResults, type = "best")
}

Iteration elite configuration performance on the test set:

if (iraceResults$scenario$testIterationElites) {
   boxplot_test(iraceResults, type = "all")
} else{
  cat("No iteration test performed.\n")
}

Training performance

Final elite configurations performance on the training set.

boxplot_training(iraceResults)

pabloOnate/iraceplot documentation built on Aug. 29, 2021, 2:32 a.m.