Data (results of the experiments):
  1. Supplementary materials for the paper include: starting points used in the experiments; results of tuned and untuned NL-SHADE-LBC, NL-SHADE-RSP-MID, and S-LSHADE-DP on CEC2022 benchmark; R and Python code that calculates the proposed quality measure and ranking; results of EA4Eig with various combinations of disabled components; the source code of listed algorithms.
  2. NL-SHADE-LBC results on CEC2022: default, tuned for CEC, tuned for the percentage measure. Explanation is in the paper.
  3. NL-SHADE-RSP-MID results on CEC2022: default, tuned for CEC, tuned for the percentage measure. Explanation is in the paper.
  4. S-LSHADE-DP results on CEC2022: default, tuned for CEC, tuned for the percentage measure. Explanation is in the paper.
  5. Corrected EA4eig results on CEC2022. In the original version, the decision path that used jSO and Eigen crossover did not respect the bounds of the search space. A more detailed explanation is in the paper.
  6. Results of EA4Eig with various combinations of disabled components on CEC2022. A more detailed explanation is in the paper.
  7. Supplementary materials for the paper include: scripts that run several versions of CMA-ES using COCO platform; results of the experiments generated by COCO. Examined versions of CMA-ES differ in performance.
  8. Results of IPOP-CMA-ES variants with two bound constraint handling methods on CEC 2017 benchmark. A more detailed explanation is in the paper.