Several objectives
Generated by python -m hone.benchmarks --multi with hone 0.11.0 (the multi-objective branch): 7 problems, 10 seeds, budgets of 10, 20, 50, 100 and 1,000 evaluations per dimension.
10 evaluations per dimension
| Problem |
nsga2 |
bayesian-parego |
bayesian-models |
random |
weighted-sums |
| zdt1-8d |
0.332 |
0.162 |
0.405 |
0.597 |
0.581 |
| rotated-zdt1-8d |
0.332 |
0.159 |
0.419 |
0.622 |
0.562 |
| zdt2-8d |
0.437 |
0.204 |
0.433 |
0.543 |
0.543 |
| zdt3-8d |
0.289 |
0.256 |
0.413 |
0.549 |
0.538 |
| zdt4-6d |
0.877 |
0.877 |
0.877 |
0.877 |
0.877 |
| dtlz2-7d |
0.494 |
0.376 |
0.405 |
0.598 |
0.604 |
| welded-beam-2 |
0.325 |
0.0872 |
0.0876 |
0.283 |
0.359 |
20 evaluations per dimension
| Problem |
nsga2 |
bayesian-parego |
bayesian-models |
random |
weighted-sums |
| zdt1-8d |
0.171 |
0.0837 |
0.402 |
0.559 |
0.173 |
| rotated-zdt1-8d |
0.171 |
0.0956 |
0.416 |
0.577 |
0.222 |
| zdt2-8d |
0.248 |
0.0983 |
0.319 |
0.543 |
0.224 |
| zdt3-8d |
0.132 |
0.172 |
0.336 |
0.448 |
0.213 |
| zdt4-6d |
0.877 |
0.877 |
0.877 |
0.877 |
0.877 |
| dtlz2-7d |
0.388 |
0.229 |
0.126 |
0.531 |
0.525 |
| welded-beam-2 |
0.221 |
0.051 |
0.0455 |
0.231 |
0.0664 |
50 evaluations per dimension
| Problem |
nsga2 |
bayesian-parego |
bayesian-models |
random |
weighted-sums |
| zdt1-8d |
0.0673 |
n/a |
n/a |
0.447 |
0.0663 |
| rotated-zdt1-8d |
0.0719 |
n/a |
n/a |
0.475 |
0.0644 |
| zdt2-8d |
0.23 |
n/a |
n/a |
0.517 |
0.162 |
| zdt3-8d |
0.0537 |
n/a |
n/a |
0.398 |
0.122 |
| zdt4-6d |
0.877 |
n/a |
n/a |
0.877 |
0.877 |
| dtlz2-7d |
0.238 |
n/a |
n/a |
0.438 |
0.522 |
| welded-beam-2 |
0.094 |
0.0224 |
0.0245 |
0.134 |
0.0208 |
100 evaluations per dimension
| Problem |
nsga2 |
bayesian-parego |
bayesian-models |
random |
weighted-sums |
| zdt1-8d |
0.0381 |
n/a |
n/a |
0.385 |
0.0613 |
| rotated-zdt1-8d |
0.0219 |
n/a |
n/a |
0.439 |
0.062 |
| zdt2-8d |
0.0372 |
n/a |
n/a |
0.462 |
0.162 |
| zdt3-8d |
0.0219 |
n/a |
n/a |
0.371 |
0.0579 |
| zdt4-6d |
0.877 |
n/a |
n/a |
0.877 |
0.877 |
| dtlz2-7d |
0.141 |
n/a |
n/a |
0.367 |
0.521 |
| welded-beam-2 |
0.0561 |
n/a |
n/a |
0.112 |
0.0144 |
1000 evaluations per dimension
| Problem |
nsga2 |
bayesian-parego |
bayesian-models |
random |
weighted-sums |
| zdt1-8d |
0.00127 |
n/a |
n/a |
0.241 |
0.0128 |
| rotated-zdt1-8d |
0.00133 |
n/a |
n/a |
0.251 |
0.0129 |
| zdt2-8d |
0.00128 |
n/a |
n/a |
0.353 |
0.034 |
| zdt3-8d |
0.00117 |
n/a |
n/a |
0.195 |
0.0114 |
| zdt4-6d |
0.484 |
n/a |
n/a |
0.877 |
0.877 |
| dtlz2-7d |
0.0208 |
n/a |
n/a |
0.207 |
0.16 |
| welded-beam-2 |
0.00423 |
n/a |
n/a |
0.0543 |
0.0142 |
Median hypervolume gap over the runs: the true front's hypervolume less the run's, with objectives scaled from 0 at the true front's ideal to 1 at its nadir and a reference point of 1.1. 0 is the true front; lower is better.