Holdout Method

Split dataset into training set, validation set and test set based on a simple fraction (e.g. 70-20-10).

The estimates from this method are highly dependent on the partioning of the data.

We can thus use random sampling to repeat the holdout method $k$ times and create an average over a lot of partitions, this is called K-Fold Cross Validation.

• Leave-One-Out K-Fold Cross Validation