subtopic

Gradient Boosting & HistGradientBoosting (sklearn)

Fit GradientBoosting and HistGradientBoosting with early stopping; tune learning_rate and max_depth. Concepts: Additive trees fitted to residuals/gradients; learning rate; early stopping Project A: HistGradientBoosting with class weights; compare cost with logistic regression. Project B: HistGradientBoostingRegressor: big drop in MAE vs linear; quantile=0.9 gives a P90 forecast for peak planning.