stopOnEvals

[mariusbommert]

The first mu function evaluations do not seem to be counted for the stopping criterion stopOnEvals, like the following example shows:

library(ecr)
library(smoof)

dtlz2 = makeDTLZ2Function(dimensions = 2, n.objectives = 2)
tmp = nsga2(fitness.fun = dtlz2, n.objectives = 2, n.dim = 2,
            minimize = TRUE, lower = getLowerBoxConstraints(dtlz2),
            upper = getUpperBoxConstraints(dtlz2), mu = 20, lambda = 10,
            terminators = list(stopOnEvals(10)), log.pop = TRUE)
tmp$last.population

In this example at least 20 function evaluations are used for the last population, but the budget of function evaluations is limited to 10 evaluations by stopOnEvals. The same behavior can be observed for stopOnEvals > mu. It would be nice to have at least a hint in the description that the first mu function evaluations are not counted towards the limit which is given by stopOnEvals.

[mariusbommert]

Probably I did a mistake and there is only a problem for stopOnEvals < mu which is not so relvant.