- Compatibility with upcoming 'paradox' release.
- Change the vignette to not use the holdout task.
- Removed new functionality for survival tasks added in
0.4.0
. A dependency,mlr3proba
was removed from CRAN for now. The functionality will be added back whenmlr3proba
is re-introduced to CRAN. Users who wish to usemcboost
forsurvival
are adviced to use version0.4.1
usetogether with the GitHub version ofmlr3proba
. - Improved stability of unit tests and example checks on CRAN.
- Fixed unit error in unit tests that led to non-passing unit tests with new mlr3proba version.
- [Experimental] mcboost now has experimental support for survival tasks.
See
MCBoostSurv
and the corresponding vignette "MCBoostSurv - Basics" for more information. - We have published an article about mcboost in the Journal of Open Source Software: "https://joss.theoj.org/papers/10.21105/joss.03453". See citation("mcboost") for the citation info.
- Auditors can now also update weights if correlations are negative by switching the sign of the update direction as intended in the paper.
- Minor adaptions to improve stability of unit tests
- Minor adpations to improve stability of unit tests
- Fixed a bug for additive weight updates, were updates went in the wrong direction.
- Added new parameter
eval_fulldata
that allows to compute auditor effect across the full sample (as opposed to the bucket).
- First CRAN-ready version of the package.
- Added a
NEWS.md
file to track changes to the package.