Feature request: parameter optimisation rounding

KNIME Analytics Platform
1

Dear KNIMErs,

Here's another feature request: the parameter optimisation loop start should be rounding the generated variables to the specified significant digits.

In the current, unrounded state the variable values are hard to use for conditional filtering, row/column naming, etc. I'm aware of an acceptable workaround for the time being, but it involves three nodes (var to row, round, row to var). Really shouldn't be necessary to do this in the first place, though.

Thanks
E

2

Dear E.,

what is the reason you want to do this rounding? Renaming columns to the variables seems to be a really rare use case.

Best regards, Iris 

3

Hi Iris,

The main rationale is for variable transformations where 0 = undefined, like by-the-book Box Cox power transformations. I have to catch if lambda == 0, but due to lack of precise rounding this will never be the case. :-(

Thanks
E