Gradient Boosting
In plain terms
An ensemble that builds trees sequentially to correct prior errors.
Going deeper
Gradient boosting adds models one at a time, each focusing on the residual errors of the ensemble so far, yielding strong predictors. It learns from mistakes iteratively.
Under the hood
Powerful for tabular data but sensitive to tuning and prone to overfitting without regularization.
Examples
XGBoost, LightGBM, tabular ML, Kaggle competitions, ranking.
Think of it like...
Like a team where each new member specializes in fixing the previous one's mistakes.