search

slds-lmu / paper_2023_survival_benchmark

Benchmark for Burk et al. (2024)
https://projects.lukasburk.de/survival_benchmark/
GNU General Public License v3.0
4 stars 0 forks source link

Early stopping for XGBoost when using mlr3mbo #12

Closed jemus42 closed 1 week ago

jemus42 commented 1 month ago

This is currently broken in the latest GitHub versions, which are required since CRAN versions of mlr3 and mlr3mbo are not compatible if I remember correctly, as I'm getting:

Learner (surv.xgboost.cox): Configure field 'validate' to enable early stopping."
[436] "This happened PipeOp surv.xgboost.cox's $train())}"

Last I tracked this down in a regression context I ended up with https://github.com/mlr-org/mlr3tuning/issues/452 and https://github.com/mlr-org/mlr3mbo/issues/163 respectively, which in short means that I'll need Marc's help and a bit more stability testing before I can start things.

jemus42 commented 1 week ago

fixed as of recently