Regularization

Statistical learning techniques that constrain how complex a fitted model may become, including sparsity limits, magnitude penalties and pruning, in order to improve results on data the model was not fitted to.