I have a use case where our end users are interested in customizing the feature reduction options in the Visual/Auto ML UI. Dataiku already offers end users the capability to create custom python models in the algorithm section, customize the metrics for model evaluation, and also to customize the cross-validation strategy. Some users want a method for feature reduction focused on the cut-off of correlation values and not for specific top # of features (i.e. with the out of the box correlation with target method). This would be helpful to allow end users to adjust where possible.