Michel Lang

In https://github.com/mlr-org/mlr3misc/commit/7f56ff2e0992508e25a04e3af0ed20fbd17eacfc I introduced default names for the elements returned by a dictionaries (which is sometimes very convenient). This breaks some tests in pipelines, see https://github.com/mlr-org/mlr3misc/pull/68.

_Originally posted by @spietras in https://github.com/mlr-org/mlr3/issues/792#issuecomment-1145899186_

Change default of robustify pipeline?

1

The encoding is set to one-hot encoding per default. This results in colinear features, which many stats models do not like at all. Wouldn't treatment encoding be the better default?...

Otherwise everything gets deterministic (unless executed in parallel).

Implement base_learner method for GraphLearners

2

To simplify accessing fields in simple (linear) pipelines. https://github.com/mlr-org/mlr3/issues/653

Store timings for each pipeline step

2

@bommert requested that we record the execution time for each step in the GraphLearner. I agree that this would be helpful and suggest introducing a flag to turn this on...

Manually bagging ranger predictions

1

See https://stackoverflow.com/questions/65972567/how-can-i-save-all-prediction-from-ranger-learner-with-predit-all-true-into-a and https://github.com/mlr-org/mlr3learners/pull/172. Should this be a pipe operator in mlr3pipelines? Maybe a nice use case to demonstrate how to write custom pipe ops?

Feature Request: Send line and move cursor down

7

For most REPLs, I find it quite convenient to execute code line by line. I've written a small helper function to make send the current line and then move the...

Plan for debugging (and traceback)

3

I know there is the `"transparent"` plan, but I'm still not very helpful tracebacks: ```r library(future) plan("transparent") f = function(x) if (x > 5) stop("whoops") else x g = function(x)...