[LRNRQ] Update oblique random survival forests with aorsf

[bcjaeger]

Algorithm

Random Forest

Package

aorsf

Supported types

• classif
• clust
• dens
• regr
• surv

I have checked that this is not already implemented in

• mlr3
• mlr3learners
• mlr3extralearners
• Other core packages (e.g. mlr3proba, mlr3keras)

Why do I think this is a useful learner?

Oblique random survival forests are implemented in ml3extralearners using the obliqueRSF package, but I have a faster version of this algorithm available in the aorsf R package. (over 500 times faster in my preliminary benchmarks).

Further Optional Comments

The aorsf package was developed to be more user friendly and runs faster than obliqueRSF. It may even give more accurate predictions. I will mark obliqueRSF as deprecated at some point in the future and will direct any interested users to aorsf instead. I am also happy to help by submitting a pull request.

[sebffischer]

Hey @bcjaeger, thanks for giving the heads up!

Will this package be on CRAN and if so, when can I expect this?

[bcjaeger]

Hi @sebffischer, no problem at all. Thanks for the quick response!

I am planning to submit aorsf to CRAN after making possibly major revisions based on ROpenSci's review (submission here: ropensci/software-review#532). In a best case scenario, the review will be done in 2 weeks and there will be no recommended changes and I'll have it on CRAN in 3-4 weeks. In a more realistic scenario, I'll make some changes after the review and it will be on CRAN in 2-3 months. Should I wait until the package is on CRAN before submitting my pull request?

[sebffischer]

I think waiting for the ROpenSci review would make sense, in case they catch anything important, CRAN I don't care so much as we have other non-CRAN packages. But if you think the package is stable enough and want to add it immediately this is also ok for me!

And don't hesitate to contact me if you encounter any problems!

[bcjaeger]

Closing this now that the pull request is merged. Thank you for all your help!