Machinelearning: ColumnInfo as an API parameter
The use of ColumnInfo in the public API can be rather jarring. I'd like to clarify when we should use it in the catalog API, and if the parameter name could be made to be more consistent with the API, such as ColumnOptions or Options.
rogancarr
All 3 comments
Whatever we decide to call it, we also have MeanVarColumn, LogMeanVarColumn, BinningColumn, SupervisedBinningColumn and MinMaxColumn in NormalizingEstimator so they should be renamed to same thing. (Easy to miss if we sweep just ColumnInfo).
We also have SimpleColumnInfo which we need to address as well.
Ivanidzo4ka
on 15 Feb 2019
I think the suggestion ColumnOptions could work. It relates to the naming convention that we use for trainers, and it also indicates that this is column-specific compared to the trainers.
I would like however to hear @TomFinley 's opinion on this.
artidoro
on 21 Feb 2019
I could definitely get behind ColumnOptions. It seems a far superior choice, so, let me say I agree wholeheartedly.
TomFinley
on 22 Feb 2019
Related issues
darren-zdc
路
3Comments
daholste
路
3Comments
daholste
路
4Comments
sethreidnz
路
3Comments
samueleresca
路
3Comments
Most helpful comment
Whatever we decide to call it, we also have
MeanVarColumn,LogMeanVarColumn,BinningColumn,SupervisedBinningColumnandMinMaxColumninNormalizingEstimatorso they should be renamed to same thing. (Easy to miss if we sweep just ColumnInfo).We also have
SimpleColumnInfowhich we need to address as well.