quantlab.model.walk_forward_training.cv_mean_metrics#
- quantlab.model.walk_forward_training.cv_mean_metrics(results)[source]#
Average every
train_*/val_*/test_*metric over folds.Each mean is keyed
cv_mean_{key}. Only finite numeric values count (a recorded null is skipped); a metric with no finite value in any fold averages to NaN.cv_n_foldsis added.- Parameters:
- resultslist[dict]
Each fold’s metrics.
- Returns:
- dict
The means and
cv_n_folds; empty when no fold carries a metric, in which case no summary run is opened.
Examples
>>> cv_mean_metrics([{"test_ic": 0.1}, {"test_ic": 0.3}, {"test_ic": None}]) {'cv_mean_test_ic': 0.2, 'cv_n_folds': 3}