Version 0.9.2 on CRAN

See NEWS.md for details.
Summary:

• Patch fixing slow gbt.save() functionality.
• Obtain XGBoost and LightGBM hyperparameters from gbt.complexity().
• Include attribute "offset" in gbt.train() and predict().
• Throw error when gb-loss-approximation deviates from true loss. Suggest lower learning_rate.
• Solves $\arg\min_\eta \sum_i l(y_i, g^{-1}(offset_i+\eta))$ numerically instead of simple average to obtain initial prediction
• Deprecate warnings for zero-inflation
• Decrease package size drastically by stripping debugging information

Version 0.9.1 on CRAN

Version 0.9.1. on CRAN is a patch, adressing the problem on Solaris and UBSAN:

(Solaris) Adressing comment from Prof Brian Ripley at the CRAN Team:

• An ERROR on Solaris. Caused by taking the logarithm of type int.

Fixed by casting the problematic int to double in the logarithm.
Package passed rhub::check_on_solaris() after this.

(UBSAN) Comments from CRAN maintainer Uwe Ligges:

• Thanks, we still see with clang-UBSAN gcc-UBSAN
• agtboost.Rcheck/agtboost-Ex.Rout:../inst/include/tree.hpp:127:32:
• runtime error: member call on null pointer of type 'struct node'

This is fixed by not calling methods from NULL pointers.
Package passed rhub::check_with_sanitizers() after this.

First release on CRAN

First version

Article: An information criterion for gradient tree boosting

Version used for article "An information criterion for gradient tree boosting"
Known things to change for future releases:

• Transform of predictions to conditional expectations (e.g. logit-raw to probabilities)
• Greedy function complexity algorithm (not used in article!)