scikit-learn/scikit-learn

Huber Loss for HistGradientBoostingRegressor

Open

#31.542 aperta il 13 giu 2025

Vedi su GitHub
 (24 commenti) (0 reazioni) (0 assegnatari)Python (27.020 fork)batch import
HardNew Featurehelp wanted

Metriche repository

Star
 (66.084 star)
Metriche merge PR
 (Merge medio 10g) (90 PR mergiate in 30 g)

Descrizione

Describe the workflow you want to enable

Huber loss is available as an option for GradientBoostingRegressor and works great when training on data with frequent outliers (thank you!). HistGradientBoostingRegressor however does not support Huber loss, which may be required when scaling to larger datasets.

Describe your proposed solution

Add HuberLoss as an option for the HistGradientBoostingRegressor class.

Describe alternatives you've considered, if relevant

Possibly allow custom loss functions for the HistGradientBoostingRegressor

Additional context

No response

Guida contributor