def _generate_inner_folds(self):
self.inner_folds = FoldInfo.generate_folds(self.cross_validaton_info.inner_cv,
self._validation_X,
self._validation_y,
self._validation_kwargs)
self.cross_validaton_info.inner_folds[self.outer_fold_id] = {f.fold_id: f for f in self.inner_folds}
def test_no_cv_strategy_eval_final_performance_false(self):
test_size = 0.15
fold_list = FoldInfo.generate_folds(None,
self.X,
self.y,
self.kwargs,
eval_final_performance=False,
test_size=test_size)
# check that we have only one outer fold, that is split in training and test according to test size
self.assertEqual(len(fold_list), 1)
self.assertEqual(len(fold_list[0].train_indices), self.num_subjects)
self.assertEqual(len(fold_list[0].test_indices), 0)
def base_assertions(self, cv, nr_of_folds, eval_final_performance=True):
fold_list = FoldInfo.generate_folds(
cv,
self.X,
self.y,
self.kwargs,
eval_final_performance=eval_final_performance)
self.assertTrue(len(fold_list) == nr_of_folds)
for generated_fold in fold_list:
self.assertEqual(len(generated_fold.train_indices),
(nr_of_folds - 1) *
(self.num_subjects / nr_of_folds))
self.assertEqual(len(generated_fold.test_indices),
(self.num_subjects / nr_of_folds))
# we always start with 1 for the investigator
self.assertEqual(fold_list[0].fold_nr, 1)
return fold_list