Early Stopping

Early stopping directly combats overfitting by monitoring the model's performance on an additional set of examples called the validation set (validation set). When the objective function on the validation set stops decreasing for a certain number of iterations and begins to worsen, training is interrupted because the model may be starting to overfit the problem.