Hi,
Are there any news regarding this feature:
?
This will be very useful for me - to let the bayesian optimization some “prior” knowledge of good runs, instead of letting it shoot in the dark before it gets to a reasonable hyperparameters region.