Train many weak classifiers incrementally by increasing the error cost associated to misclassified elements at previous steps. The final predictor is a weighted sum of weak predictors.

Train many weak classifiers on different subsets of a training set, then average their predictions.