floating point decimal #1086
Replies: 2 comments 2 replies
-
I guess I'd like to understand the larger context here. Am I understanding correctly that the issue is not that the parameters are in fact discrete and that you want to round, but that the acquisition function optimization provides suggestions that are essentially identical to training observations? I that is the case then a few comments / questions:
|
Beta Was this translation helpful? Give feedback.
-
Thanks for the quick respond @Balandat |
Beta Was this translation helpful? Give feedback.
-
Hi
Sometimes, the acquisition function optimisation recommend sample that are very close (differ only in high decimal values), for example these two new recommended solutions
in most cases such accuracy is not practical. Is it possible to pass a rounding argument to "optimize_acqf" to get more different points?
Many thanks
Beta Was this translation helpful? Give feedback.
All reactions