search

crflynn / skgrf

scikit-learn compatible Python bindings for grf (generalized random forests) C++ random forest library
https://skgrf.readthedocs.io/en/stable/
GNU General Public License v3.0
31 stars 7 forks source link

Forest Classifier #42

Closed crflynn closed 3 years ago

crflynn commented 3 years ago

Probability forest implemented but not yet released in grf

crflynn commented 3 years ago

@erikcs Do you have an ETA on a new release?

erikcs commented 3 years ago

Hopefully by the end of this month (July at the latest). The final release addition is https://github.com/grf-labs/grf/pull/988, and after that we will prepare a CRAN release (v2.0).

erikcs commented 3 years ago

Hi @crflynn, the 2.0 release is out now, and as you see there are a few breaking changes. From the wrapper side the orthog.boosting argument is removed from causal_forest, and explicit support for sparse training data is removed (the data class is now concrete).

We haven't documented all the C++ changes since we strictly speaking only follow Semantic Versioning for the R package, however considering we don't want to cause trouble for repos like this which maintains wrappers - the most prominent internal changes are just removing sparse data support and using some Eigen arrays instead of std::vectors in internals, I don't think wrappers should be much affected by this release.

crflynn commented 3 years ago

Thanks for the heads up. It looks like most of the big changes are on the R side so it should be relatively painless here.