Priors for tuning

[PhilippPro]

Can you include the possibility of priors instead of only ranges to your package? Would be awesome, although I think that it is not easy to implement this in mlrMBO...

I know that you can put in functions, but this is not super flexible.

[jakob-r]

Well it's exactly that. Just that the function would be more complex than 2^x.
After all your param space would be [0,1] and the trafo would be the quantile of your prior distribution.
So pretty easy but not so easy to store in the database at the moment.

That is the paper you have in mind, right?

[PhilippPro]

Yes, for example, probably there are also others than Jans Paper.