mlr3filters icon indicating copy to clipboard operation
mlr3filters copied to clipboard

Published 20 hours ago • verified publisher icon mlr-org

Missing filter / featsel methods

Open pat-s opened this issue 6 years ago • 8 comments

Filters

Pkg

No pkg

  • [x] AUC

  • [ ] generic permutation

  • [x] univariate.model.score

stats

  • [x] anova

  • [x] kruskal

  • [x] linear.correlation

  • [x] rank.correlation

  • [x] variance

FSelector

Do we want to have these filters in again? Slow and Java problems..

FSelectorRcpp

  • [x] information.gain
  • [x] gain.ratio
  • [x] symmetrical.uncertainty

Learner integrated filters

  • [x] ranger.impurity

  • [x] ranger.permutation

  • [x] cforest.importance

Do we want to add the ramdomForest and randomForestSRC ones?

mRMRe

~~- [ ] mrmr~~ -> slow and no support for classif tasks https://github.com/mlr-org/mlr/issues/2604

praznik

  • [x] CMIM

  • [x] DISR

  • [x] JMI

  • [x] JMIM

  • [x] MIM

  • [x] MRMR

  • [x] NJMIM

care

  • [x] carscore

spFSR

Need to check.

Ensemble filters

  • [ ] Min

  • [ ] Mean

  • [ ] Median

  • [ ] Max

  • [ ] Borda

  • [ ] Borda-staircase

  • [ ] Borda-power

pat-s avatar Nov 15 '18 22:11 pat-s

For the filters: I'd start with stats / no pkg, then try to connect the modern filter packages (FSelectorRcpp and praznik).

We don't need 3 forest filters, we can solve this more generically by extending mlr3 learners with methods to extract feature scores.

mllg avatar Nov 16 '18 07:11 mllg

we can solve this more generically by extending mlr3 learners with methods to extract feature scores.

This is really important to be able to use all kinds of embedded feature selection directly by the learner.

ja-thomas avatar Nov 16 '18 09:11 ja-thomas

I don't really see a reason to use the Java FSelector package when there is FSelectoRcpp.

ja-thomas avatar Nov 16 '18 09:11 ja-thomas

I don't really see a reason to use the Java FSelector package when there is FSelectoRcpp.

The later does not have all filters of the former. See https://mlr.mlr-org.com/articles/tutorial/filter_methods.html.

pat-s avatar Nov 16 '18 09:11 pat-s

Well, with this argument we have to include all possible filters :smile:

I would suggest we start without it, and if people complain/open issues we can still add them later. Or are there any really important filters not yet in FSelectorRcpp?

ja-thomas avatar Nov 16 '18 09:11 ja-thomas

there never is (should be) pressure to include everything, include what is most important

berndbischl avatar Nov 16 '18 10:11 berndbischl

My comment was more meant to be a comparison, not a statement that we should do it :)

pat-s avatar Nov 16 '18 15:11 pat-s

NB: All learners which have some sort of "importance" are now supported via FilterVariableImportance.

mllg avatar Apr 25 '19 14:04 mllg