Ensemble methods that train models in sequence, each concentrating on the mistakes of its predecessors.