I propose that we remove the --scale,-s option from Seidr algorithms where it has been shown that scaling usually improves predictions (All ensemble regression + pearson correlation) and instead introduce a --no-scale option for advanced users.
Since this is an API-breaking change, I'll put it up here for discussion.
As per the internal discussion, the option --scale will be made default where applicable. Setting --scale will have no effect. A new option --no-scale will be added to turn scaling off if desired
I propose that we remove the
--scale,-s
option from Seidr algorithms where it has been shown that scaling usually improves predictions (All ensemble regression + pearson correlation) and instead introduce a--no-scale
option for advanced users.Since this is an API-breaking change, I'll put it up here for discussion.
IMO,
seidr
should provide "best-effort" defaults