Gradient Boosting

Residuals

Prediction errors of the current model, calculated as the difference between observed values and predictions, on which the next weak learner is trained in Gradient Boosting.

← Geri