autotune n_masks for rise

[cwmeijer]

Output of rise can be added repeatedly so you don't have to know the final n_masks in advance. We can keep adding masks until the output doesn't change much anymore. Make sure other hyperparameters (p_keep and n_features) have already been optimized (see dianna-ai/dianna#24).

[elboyran]

@cwmeijer , this is already implemented, right?

[cwmeijer]

It isn't. We've only got autotuning for p_keep (#24) at the moment.

[loostrum]

@laurasootes and I have been experimenting with this as part of the chemistry project we're working on (in the dianna-deeprank branch). There, a 3D volume needs to be masked so way more masks are needed, but we ran into memory issues when pre-generating the masks. Instead we now create masks in batches and add to the output repetitively. This is done with a moving average, so it's possible to compare to the existing result and determine whether or not it changes significantly. For the next sprint we could try adding this to DIANNA, but we can discuss tomorrow in the planning meeting.