Regularization

Any technique that discourages a model from fitting noise in the training data too closely (penalizing large weights, dropping units at random, stopping training early) — the general defense against overfitting.