L1 regularization in discriminant analysis

I am just getting acquainted with the discrim package and with discriminant analysis in general (discriminant analysis and its varieties are a little new to me).

I suppose my question is fairly straight forward: are there any computational engines (and corresponding tuning parameters) supported by the discrim package that will shrink model coefficients down to zero?

I see here that using the mda engine will allow for ridge/weight decay regularization. Is there anything available that would be more akin to L1 regularization? I am well-acquainted with lasso regression using multinom_reg() and logistic_reg() + glmnet, but I was wondering if something similar was available for discriminant analysis in the tidymodels framework.


We haven't added any L1 methods to the package but can. sparseLDAcan be added.

Nice. Do you see this being on the short list of things to do for the package? I can add this as a feature request on GitHub.

Btw thanks for all the work the tidymodels team has put into this whole ecosystem, really enjoying it.

A feature request on GitHub would be great. We often rotate between packages so it might be a month or so before I get to it.

1 Like

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.