remove save.population.at stuff, work on autoplot function

R/autoplotECRResult.R

@@ -26,8 +26,12 @@ autoplot.ecr_single_objective_result = function(object, xlim = NULL, ylim = NULL
                                , log.fitness = FALSE, complete.trace = FALSE, ...) {
   assertFlag(show.process, na.ok = FALSE)
   assertFlag(complete.trace, na.ok = FALSE)
-  obj.fun = object$task$fitness.fun
-  n.params = getNumberOfParameters(obj.fun)
+  assertFlag(log.fitness, na.ok = FALSE)
+
+  if (is.null(object$opt.path)) {
+    stopf("Cannot plot optimization trace, since obviously no logging took place.")
+  }
+
 
   op = as.data.frame(object$opt.path)
   # we start with the second dob, since otherwise there is not enough info to plot
@@ -39,24 +43,40 @@ autoplot.ecr_single_objective_result = function(object, xlim = NULL, ylim = NULL
     pl.trace = plotTrace(op[which(op$dob <= dob), ], xlim, ylim, log.fitness, ...)
     pl.trace = pl.trace + ggtitle(sprintf("Optimization trace for function '%s'", getName(obj.fun)))
     if (show.process) {
-      if (n.params > 2L || isMultiobjective(obj.fun)) {
-        stopf("Visualization not possible for multi-objective functions or functions with greater than 2 parameters.")
+      if (object$final.opt.state$control$representation == "custom") {
+        stopf("Process cannot be visualized if custom representation was used.")
+      }
+      obj.fun = object$task$fitness.fun
+      task = object$task
+      par.set = smoof::getParamSet(obj.fun)
+      n.params = getNumberOfParameters(obj.fun)
+
+      if (n.params > 2L) {
+        stopf("Visualization not possible for functions with more than 2 parameters.")
       }
-      if (!length(object$control$save.population.at)) {
-        stopf("Cannot visualize population since no population was stored! Take a glance a the 'save.population.at' control parameter.")
+
+      if (hasDiscrete(par.set)) {
+        stopf("Visualization for mixed/discrete decision spaces not supported at the moment.")
+      }
+
+      if (isMultiobjective(obj.fun)) {
+        stopf("Visualization not possible for multi-objective functions at the moment.")
       }
+
+      # call smoof plot function
       pl.fun = autoplot(obj.fun)
-      population = object$population.storage[[paste0("gen.", dob)]]
+
+      # get interesting stuff out of opt.path in ggplot2 friendly format
+      df.points = getOptPathX(op, dob = dob)
+      y.name = task$objective.names
+      df.points[[y.name]] = getOptPathY(op, dob = dob)
+      x.names = getParamIds(par.set, with.nr = TRUE, repeated = TRUE)
       if (n.params == 2L) {
-        df.points = as.data.frame(do.call(rbind, population$individuals))
-        colnames(df.points) = paste("x", 1:n.params, sep = "")
-        df.points$y = as.numeric(population$fitness)
-        pl.fun = pl.fun + geom_point(data = df.points, aes_string(x = "x1", y = "x2"), colour = "tomato")
+        pl.fun = pl.fun + geom_point(data = df.points, aes_string(x = x.names[1L], y = x.names[2L]), colour = "tomato")
       } else {
-        fitness = as.numeric(population$fitness)
-        df.points = data.frame(x = do.call(c, population$individuals), y = fitness)
-        pl.fun = pl.fun + geom_point(data = df.points, aes_string(x = "x", y = "y"), colour = "tomato")
-        pl.fun = pl.fun + geom_hline(yintercept = min(fitness), linetype = "dashed", colour = "gray")
+        pl.fun = pl.fun + geom_point(data = df.points, aes_string(x = x.names, y = y.name), colour = "tomato")
+        opt.dir.fun = if (task$minimkze) min else max
+        pl.fun = pl.fun + geom_hline(yintercept = opt.dir.fun(df.points[[y.name]]), linetype = "dashed", colour = "gray")
       }
 
       #FIXME: this seems to fail!

R/optState.R

@@ -26,13 +26,6 @@ setupOptState = function(task, population, control) {
     opt.state$best.value = best$fitness
   }
 
-  # optional logging of population
-  opt.state$population.storage = namedList(paste0("gen.", control$save.population.at))
-
-  if (0 %in% control$save.population.at) {
-    opt.state$population.storage[[paste0("gen.", as.character(0))]] = population
-  }
-
   # construct opt path
   y.names = paste0("y", seq(task$n.objectives))
 
@@ -65,18 +58,11 @@ updateOptState = function(opt.state, population, control) {
   # update population
   opt.state$population = population
 
-  #FIXME/TODO: handle opt.path update!
-
   # save best-so-far solution in single-objective case
   if (task$n.objectives == 1L) {
     updateOptStateBestIndividual(opt.state)
   }
 
-  # update populaton storage
-  if (opt.state$iter %in% control$save.population.at) {
-    opt.state$population.storage[[paste0("gen.", as.character(opt.state$iter))]] = population
-  }
-
   invisible()
 }
 

R/setupECRControl.R

@@ -23,8 +23,6 @@
 #'   Number of fittest individuals of the current generation that shall be copied to the
 #'   next generation without changing. Default is 0. Keep in mind, that the algorithm
 #'   does not care about this option if the \code{survival.strategy} is set to 'plus'.
-#' @param save.population.at [\code{integer}]\cr
-#'   Which populations should be saved? Default is none.
 #' @param monitor [\code{function}]\cr
 #'   Monitoring function. Default is \code{consoleMonitor}.
 #' @param stopping.conditions [\code{list}]\cr
@@ -59,7 +57,6 @@ setupECRControl = function(
   representation,
   survival.strategy = "plus",
   n.elite = 0L,
-  save.population.at = integer(0),
   monitor = makeConsoleMonitor(),
   stopping.conditions = list(),
   logger = makeNullMonitor(),
@@ -75,10 +72,6 @@ setupECRControl = function(
   assertList(custom.constants, unique = TRUE, any.missing = FALSE, all.missing = FALSE)
   assertFlag(vectorized.evaluation, na.ok = FALSE)
 
-  if (length(save.population.at) > 0) {
-    assertIntegerish(save.population.at, lower = 0L, any.missing = FALSE)
-  }
-
   # If the survival strategy is (mu + lambda), than the number of generated offspring in each iteration
   # must greater or equal to the population size
   if (survival.strategy == "comma" && n.offspring < n.population) {
@@ -125,7 +118,6 @@ setupECRControl = function(
     representation = representation,
     survival.strategy = survival.strategy,
     n.elite = n.elite,
-    save.population.at = save.population.at,
     stopping.conditions = stopping.conditions,
     monitor = monitor,
     logger = logger,

R/setupResult.R

@@ -60,7 +60,7 @@ setupResult.ecr_single_objective_opt_state = function(opt.state, stop.object, co
     control = control,
     best.param = opt.state$best.param,
     best.value = opt.state$best.value,
-    opt.path = coalesce(opt.state$opt.path, NA),
+    opt.path = opt.state$opt.path,
     last.population = opt.state$population,
     population.storage = opt.state$population.storage,
     message = stop.object$message,
@@ -94,7 +94,7 @@ setupResult.ecr_multi_objective_opt_state = function(opt.state, stop.object, con
     final.opt.state = opt.state,
     task = opt.state$task,
     control = control,
-    opt.path = coalesce(opt.state$opt.path, NA),
+    opt.path = opt.state$opt.path,
     pareto.idx = pareto.idx,
     pareto.front = t(fitness[, pareto.idx, drop = FALSE]),
     pareto.set = population[pareto.idx],

examples/ex_doTheEvolution.R

@@ -58,10 +58,11 @@ ctrl = setupECRControl(
   n.population = 100L,
   n.offspring = 10L,
   survival.strategy = "plus",
-  save.population.at = 0:100L,
   representation = "float",
+  logger = makeOptPathLoggingMonitor(),
   stopping.conditions = setupStoppingConditions(max.iter = 100L)
 )
+
 ctrl = setupEvolutionaryOperators(
  ctrl,
  parent.selector = makeRouletteWheelSelector(),

inst/examples/onesAndZeros_example.R

@@ -10,13 +10,14 @@ load_all(".", reset = TRUE)
 # objective function
 obj.fn = makeSingleObjectiveFunction(
   name = "Number of Ones",
-  fn = function(x, y) {
-    (length(x) - sum(x)) + sum(y)
+  fn = function(x1, x2) {
+    (length(x1) - sum(x1)) + sum(x2)
   },
   par.set = makeParamSet(
-    makeIntegerVectorParam("x", len = 20L, lower = 0, upper = 1),
-    makeIntegerVectorParam("y", len = 15L, lower = 0, upper = 1)
-  )
+    makeIntegerVectorParam("x1", len = 20L, lower = 0, upper = 1),
+    makeIntegerVectorParam("x2", len = 15L, lower = 0, upper = 1)
+  ),
+  has.simple.signature = FALSE
 )
 
 makeOptimumAppearsStoppingCondition = function(opt.fitness = 0) {
@@ -36,7 +37,7 @@ control = setupECRControl(
   n.offspring = 100L,
   n.mating.pool = 100L,
   representation = "binary",
-  save.population.at = c(0L, 10L, 20L, 30L, 40L, 50L, 60L),
+  logger = makeOptPathLoggingMonitor(step = 10L),
   monitor = makeConsoleMonitor(1L),
   stopping.conditions = list(
     makeOptimumAppearsStoppingCondition()
@@ -55,4 +56,4 @@ control = setupEvolutionaryOperators(
 # names(as.data.frame(res$opt.path))
 res = doTheEvolution(obj.fn, control = control)
 print(res)
-#autoplot(res, complete.trace = TRUE)
+autoplot(res, complete.trace = TRUE)

tests/testthat/helper_zzz.R

@@ -1,4 +1,4 @@
-set.seed(1)
+set.seed(5453)
 
 makeOneMinFunction = function(dimensions) {
   assertInteger(dimensions, len = 1L, lower = 2L, upper = 100L)

tests/testthat/test_autoplot.R

@@ -33,7 +33,6 @@
 #       representation = "float",
 #       survival.strategy = "plus",
 #       # FIXME: Throws an error if not complete population is saved
-#       save.population.at = 0:15,
 #       monitor = makeNullMonitor(),
 #       stopping.conditions = list(makeMaximumIterationsStoppingCondition(max.iter = 15L))
 #     )

tests/testthat/test_ecr.R

@@ -191,19 +191,11 @@ test_that("ecr can handle initial populations", {
   control = setupECRControl(
     n.population = 3L,
     n.offspring = 1L,
-    save.population.at = 0L,
     representation = "float",
     monitor = makeNullMonitor(),
     stopping.conditions = setupStoppingConditions(max.iter = 1L)
   )
 
   # stop if initial population is to large
   expect_error(doTheEvolution(fn, control, c(initial.population, c(2, 2.5))), "exceeds", ignore.case = TRUE)
-
-  # check if initial population is indeed used
-  res = doTheEvolution(fn, control, initial.population)
-  first.population = res$population.storage[[1L]]
-  for (i in seq(length(first.population))) {
-    expect_true(all(first.population$individuals[[i]] == initial.population[[i]]), info = sprintf("%i-th element not equal to intial population.", i))
-  }
 })