You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As of 2023, we have now the sbox-cost suite which implements hard box constraints with the functions of the bbob suite where also some of the functions were changed. In discussions during GECCO 2023, the criticism of the original suites/setup was that without checking the publication, we don't know (from the suite name) in which setting the algorithm was run, soft or hard box constraints. A simple half-fix to this criticism is to add the bbob and bbob-biobj and ... suites such that they return NaN outside the box. We would still not know what the algorithms do on soft-constrained suites. Adding such suites should be a pretty simple copy-paste job on the experimentation side. The post-processing then needs to define/respect the compatibilities between suites.
The text was updated successfully, but these errors were encountered:
As of 2023, we have now the
sbox-cost
suite which implements hard box constraints with the functions of thebbob
suite where also some of the functions were changed. In discussions during GECCO 2023, the criticism of the original suites/setup was that without checking the publication, we don't know (from the suite name) in which setting the algorithm was run, soft or hard box constraints. A simple half-fix to this criticism is to add thebbob
andbbob-biobj
and ... suites such that they returnNaN
outside the box. We would still not know what the algorithms do on soft-constrained suites. Adding such suites should be a pretty simple copy-paste job on the experimentation side. The post-processing then needs to define/respect the compatibilities between suites.The text was updated successfully, but these errors were encountered: