Computationally expensive feature selection algorithms like RFE

[MostafaBouzari]

Dear Community, Feature selection is an important part of the ML pipeline. Often these algorithms are time consuming and not efficient to run on a dataframe/dataset with extraordinary columns (e.g.: 500 or 5000). in those cases, algorithms like Recursive Feature Elimination (RFE) are Impossible to run and train on different module.

We can use some of these different on cpu time-consuming and computationally expensive algorithms into cuML.

that would help us a lot.

tnx.

[dantegd]

Thanks for the request @MostafaBouzari! It is very useful to know that RFE and related algos are of interest, we will be considering them for development in a future release.