Figure 2.

Visualization of the five-times-repeated three-fold cross-validation. For each hold-out test set, the remaining data is split into three different train and validation sets. Each datapoint is used as test data at some point.