This page contains results of IPOP-CMA-ES variants on CEC 2017 benchmark for single objective
real-parameter numerical optimization. Results in 30 dimensions were used in the article: Handling bound constraints in CMA-ES: An experimental study. Swarm and Evolutionary Computation, vol. 52, no. 100627, 2020, DOI: 10.1016/j.swevo.2019.100627.
Results of IPOP-CMA-ES that used its default bound constraint handling (transformation)
Results of the verion 2.7.1 of IPOP-CMA-ES which was allowed for up to 9 restarts. The initial step size (sigma) was set to 0.3(u-l), where u and l are upper and lower bounds of the search space.
Objective function error values (Fi(x)-Fi(x*)) (CEC tables with Best, Worst, Median, Mean and Std) for 10, 30, 50 and 100 dimensions.
The best objective function error value (Fi(x)-Fi(x*)) for each function and for each run. Runs are in columns, objective function numbers are in rows. The results for 10, 30, 50 and 100 dimensions.
Results of IPOP-CMA-ES that used Darwinian reflection for bound constraint handling
Objective function error values (Fi(x)-Fi(x*)) (CEC tables with Best, Worst, Median, Mean and Std) for 10, 30, 50 and 100 dimensions.
The best objective function error value (Fi(x)-Fi(x*)) for each function and for each run. Runs are in columns, objective function numbers are in rows. The results for 10, 30, 50 and 100 dimensions.
CEC 2017 functions renumbering
Originally CEC 2017 consisted of 30 functions. Then, function number 2 was excluded from the competition. At the time of writing this note (March 2021), an updated problem definitions renumbered functions formerly used in 2017 (old function 3 is now called 2, etc.).