mb706
mb706
Another solution for https://github.com/mlr-org/mlr3pipelines/issues/564 . May influence efficiency. Should only be used when imputation-methods are feature-independent, however.
somehow automatically recognize when a column is close to collinear with another column and drop it; could be useful for linear models
e.g. check whether training failed for anything that is connected.
* empty multiplicities without nesting should now work, but need testing * nested empty multiplicities probably need to keep track of "hidden" nesting depth, e.g. with an attribute.
input: table with id column (not row id), time column t, feature columns f_i with entries f_i,id,t, and target value t_id,t. The PipeOp generates for each id,t (for which it...
mlr3 doesn't give us an error any more here: https://github.com/mlr-org/mlr3pipelines/blob/6bfe8a8a30767f72395720ef531f8a505e65a903/tests/testthat/test_pipeop_featureunion.R#L105 Need to investigate if this is a problem.
We should try to reduce CRAN tests to those that interact with non-mlr3-ecosystem packages (so their maintainers have to give us prior notice about breaking mlr3pipelines). mlr3-internal things are checked...
We want to be able to execute Graphs inside recipes workflows (or whatever they are called) for interoperability with tidymodels.
... once we have competent checking of backend equality.
PipeOpFeatureUnion currently uses `Task$data()` to get the data to cbind. We can save some space and performance if we just cbind the `DataBackend` directly. Difficulty: The `DataBackend`s do not contain...