[ENH]: Implement a Repetead Group K Fold CV

[fraimondo]

Which feature do you want to include?

We need a repeated (non-deterministic) group k-fold

How do you imagine this integrated in julearn?

Something like this, from @kaurao using ChatGPT

import numpy as np
from sklearn.model_selection import GroupKFold

class RepeatedGroupKFold:
    def __init__(self, n_splits=5, n_repeats=5, random_state=None):
        self.n_splits = n_splits
        self.n_repeats = n_repeats
        self.random_state = np.random.RandomState(random_state)

    def split(self, X, y=None, groups=None):
        if groups is None:
            raise ValueError("Groups must be provided for GroupKFold.")

        unique_groups = np.unique(groups)

        for repeat in range(self.n_repeats):
            # Shuffle groups before each repeat
            shuffled_groups = self.random_state.permutation(unique_groups)

            folds = np.array_split(shuffled_groups, self.n_splits)

            for fold_groups in folds:
                test_idx = np.isin(groups, fold_groups)
                train_idx = ~test_idx

                yield np.where(train_idx)[0], np.where(test_idx)[0]

    def get_n_splits(self, X=None, y=None, groups=None):
        return self.n_splits * self.n_repeats

Do you have a sample code that implements this outside of julearn?

Anything else to say?

No response