Multi-parametric approach

Totally dumb question, that could lead into something more interesting:
With multi-dimensional approaches like stuff with descriptors, it is often difficult to know which set of descriptors to play with first. I would agree doing so usually means I am wrongly willing to think with only a few dimensions; but it is often useful as a starting point for exploration.

In the help patches, would proposing one possible order of parameter tweak make sense ? I mean proposing one or several possible experimental paths for playing with parameters ? A regressive approach only to learn how parameters interact one another.

1 Like

This is an interesting interface proposal. We could do a rough sample and see which descriptor are maximally sparse for instance… or combinations. I think @weefuzzy and @groma have many ideas on this for the 2nd toolset.