boosting

• Iteratively add a weak model to a strong model to improve performance of the strong model.
• The weak model is added to the ensemble with a weight proportional to its accuracy
• The training data examples that are misclassified by the ensemble are weighted more in the loss