bluecast.conformal_prediction.nonconformity_measures¶
Module Contents¶
Functions¶
|
|
|
Calculate Hinge loss per row. |
|
Calculate margin nonconformity score per row. |
|
Calculate Brier score per row. |
- bluecast.conformal_prediction.nonconformity_measures.convert_to_numpy(y_true: pandas.Series, y_hat: numpy.ndarray | pandas.Series | pandas.DataFrame) Tuple[numpy.ndarray, numpy.ndarray]¶
- bluecast.conformal_prediction.nonconformity_measures.hinge_loss(y_true: pandas.Series, y_hat: numpy.ndarray | pandas.Series | pandas.DataFrame) numpy.ndarray¶
Calculate Hinge loss per row.
To compute the nonconformity score, take the probability score of the true class and subtract it from 1. :param y_true: True labels :param y_hat: Predicted probabilities :return: Hinge loss per row
- bluecast.conformal_prediction.nonconformity_measures.margin_nonconformity_measure(y_true: pandas.Series, y_hat: numpy.ndarray | pandas.Series | pandas.DataFrame) numpy.ndarray¶
Calculate margin nonconformity score per row.
The margin nonconformity measure is defined as the difference between the predicted probability of the most likely incorrect class label and the predicted probability of the true label. :param y_true: True labels :param y_hat: Predicted probabilities :return: Margin nonconformity score loss per row
- bluecast.conformal_prediction.nonconformity_measures.brier_score(y_true: pandas.Series, y_hat: numpy.ndarray | pandas.Series | pandas.DataFrame) numpy.ndarray¶
Calculate Brier score per row.
It calculates the squared difference between the predicted probabilities and the actual binary results. The score’s values can range from 0 (perfect accuracy) to 1 (complete inaccuracy). :param y_true: True labels :param y_hat: Predicted probabilities :return: Brier score loss per row