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
For minimizing my black-box function y I can always maximize -y.
However, since the analytic acquisition functions (noisy/constrained)EI, PI and UCB support minization via maximize=False while the others don't, I was wondering if you plan to implement the maximize argument for each acquisition.
Would a negating OutcomeTransform or AcquisitionObjective be an alternative?
The text was updated successfully, but these errors were encountered:
We could support that, but it seems kind of a lot of machinery for flipping a sign.
Would a negating OutcomeTransform or AcquisitionObjective be an alternative?
Sure. we could give AcquisitionObjective a negate kwarg to carry through. I think that would make more sense than doing this in the outcome transforms.
We could support that, but it seems kind of a lot of machinery for flipping a sign.
Definitively. My thinking was that if there is way to support minimization independent of the acquisition, then not having the maximize kwarg in the analytic acquistions woud make it feel more consistent. I know where to look now, thanks!
For minimizing my black-box function
y
I can always maximize-y
.However, since the analytic acquisition functions (noisy/constrained)EI, PI and UCB support minization via
maximize=False
while the others don't, I was wondering if you plan to implement themaximize
argument for each acquisition.Would a negating
OutcomeTransform
orAcquisitionObjective
be an alternative?The text was updated successfully, but these errors were encountered: