by Tianjun Liao,
Daniel Molina, Marco A. Montes de Oca, and Thomas Stützle
|Submitted to Evolutionary Computation Journal|
The benchmark functions and some of the algorithms proposed for the special session on real parameter optimization of the 2005 IEEE Congress on Evolutionary Computation (CEC'05) have played and still play an important role in the assessment of the
state of the art in continuous optimization. In this note we first show that, if bound constraints are not enforced for the final reported solutions, state-of-the-art algorithms
produce on a majority of the CEC'05 benchmark functions infeasible best candidate solutions, even though the optima of the CEC'05 functions are within the specified
bounds. This observation has important implications on algorithm comparisons. In fact, this note also draws the attention to the fact that authors may have drawn wrong
conclusions from experiments using the CEC'05 problems.
Additional tables are given that show the results of "kbf" version. This version search outside the bounds without restrictions but the final solution reported is the best feasible solution that has been identified during the search process.
please see the supp-kbf.pdf
Additional tables are given that show the median results and more detailed information such as the (the best, 0.25 quartile, median, 0.75 quartile and worst error values for each function.
please see the supp-median-others.pdf
See the allsolutions.tar.gz. For a quick look, here are the final solutions of IPOP-CMA-ES and MA-LSch-CMA (both acb and ncb versions) obtained in f18 of dimension 30. See the quicklysee-example-f18.tar.gz