 K-fold cross-validation

K-fold cross-validation partitions observations into $K$ folds, trains $K$ times while holding out one fold at a time, and combines the held-out losses.