mlr3pipelines

mlr3pipelines copied to clipboard

Suggestion for ID suggestion improvement

4

I think a relatively common user error (at least for not so experienced users) is that they forget the prefixes when setting graph parameters: (e.g. `nrounds` instead of `classif.xgboost.nrounds` below....

sebffischer

Add examples for `PipeOpFixFactors` and `PipeOpTargetInvert`

2

advieser

Many of our pipeops fit this mold: impute, encode, YeoJohnson, BoxCox (?), binning. Should PipeOpTaskPreproc(Simple), possibly both.

mb706

PipeOpImputeOOR behaviour for missing only during predict

3

Basically https://github.com/mlr-org/mlr3pipelines/issues/677 again. Currently we naively turn `NA` into the `.__MISSING__` level for factor columns. If things are only missing during prediction, this introduces a new factor level. In the...

mb706

Error: `colapply` only raises errors during parallel optimization

5

## description When using multisession with future, graph learner containing `colapply` raises an error, which never exists before and exists in sequential mode. ## error message ``` > # not...

Vinnish-A

Allow multiple input/output types in `as.data.table.DictionaryPipeOp`

2

It would be good if it were possible to specify multiple different allowed `task_type`s to be displayed in `input.type.train`, `input.type.predict`, `output.type.train`, and `output.type.predict`, when calling the S3 method `as.data.table.DictionaryPipeOp` for...

advieser

refactor: use more `ps$set_values()`

1

@mb706 check if you want/like this, this standardises to use more `ps$set_values()` like in the other mlr3 packages

m-muecke

feat: hotstart

1

Nothing to merge yet but a poc we should discuss. `mlr3tuning` is ready to use hotstarting efficiently and @sebffischer wants to use hotstarting for torch. This currently only works for...

be-marc

lona-k

`GraphLearner` should implement custom `$configure()` and `Graph` should also implement `$configure()`

2

`mlr3`'s `$configure()` allows to set both hyperparameters and fields. It would be nice if `GraphLearner` provided its own `$configure()` method that also allows to set fields of `PipeOp`s, something like...

sebffischer