My idea is to eventually add a more barebones mlrMBO version of the "calculate_sample_size" that is easier to collaborate on and to implement new ideas like the multiple criteria. I think this would hugely open up design space since we would not rely on mlpwr which is in some places specialized for power analysis stuff.
We could also have a look if there are other general backbones for surrogate modeling / Bayesian optimization. mlrMBO at least seems to have a nice documentation: https://github.com/mlr-org/mlrMBO
My idea is to eventually add a more barebones mlrMBO version of the "calculate_sample_size" that is easier to collaborate on and to implement new ideas like the multiple criteria. I think this would hugely open up design space since we would not rely on mlpwr which is in some places specialized for power analysis stuff.
We could also have a look if there are other general backbones for surrogate modeling / Bayesian optimization. mlrMBO at least seems to have a nice documentation: https://github.com/mlr-org/mlrMBO