<aside> 💡 use when you have a small training/validation set

</aside>

*k-*fold method overview

1. split data into $k$ non overlapping subsets i.e “folds”
2. choose a fold to be the validation set, rest of folds used to train
3. Evaluate performance on this split
4. Repeat until all possible validation folds are chosen
5. Return average performance on $k$-fold cross validation

Applications

• Less noisy estimate of hyperparameter effects on model