Closed hfrick closed 8 months ago
Because of the conversion of x
from a (sparse) matix to a data frame in parsnip::.convert_xy_to_form_fit()
, we can't use sparse matrices for coxnet models so I'm gonna change the encoding to reflect that.
https://github.com/tidymodels/parsnip/blob/b305a8e229ce145868593a8ae0fd9998fcf6a48e/R/convert_data.R#L230-L232
This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.
Currently, we use a parsnip encoding of
allow_sparse = TRUE
for glmnet but I think with us funneling everything through parsnip with a supposed formula interface for coxnet models (via our owncensored::coxnet_train()
) that's not actually possible, i.e. parsnip's conversion functions don't let a sparse matrix through.So we might have to change that encoding or come up with something clever.