API Reference

pysloth.ccpd_function(x_train_cal: numpy.ndarray, y_train_cal: numpy.ndarray, x_test: numpy.ndarray, y_grid: numpy.ndarray, k: int, n_delta: int) → Tuple[numpy.ndarray, numpy.ndarray]

Cross Conformal Predictive Distributions

Parameters

• x_train_cal (np.ndarray) – Train and calibration features

• y_train_cal (np.ndarray) – Train and calibration target

• x_test (np.ndarray) – Test features

• y_grid (np.ndarray) – Target grid

• k (int) – Number of fold

• n_delta (int) – Number of discrete values in an interval

Returns

Q folds and p folds

Return type

Tuple[np.ndarray, np.ndarray]

pysloth.crps_function(y_range: numpy.ndarray, q: numpy.ndarray, y: float) → float

Continuously Ranked Probabilistic System

Parameters

• y_range (np.ndarray) – Range over which Q was computed

• q (np.ndarray) – Distribution function

• y (float) – Target

Returns

Sum

Return type

float

pysloth.scpd_function(x_train: numpy.ndarray, x_cal: numpy.ndarray, y_train: numpy.ndarray, y_cal: numpy.ndarray, x_test: numpy.ndarray, y_test: numpy.ndarray, y_grid: numpy.ndarray, k: int, n_delta: int, shuffle_ind: bool = True) → Tuple[numpy.ndarray, numpy.ndarray]

Split Conformal Predictive Distributions

Parameters

• x_train (np.ndarray) – Training features

• x_cal (np.ndarray) – Calibration features

• y_train (np.ndarray) – Training target

• y_cal (np.ndarray) – Calibration target

• x_test (np.ndarray) – Test features

• y_test (np.ndarray) – Test target

• y_grid (np.ndarray) – Target grid

• k (int) – Number of fold

• n_delta (int) – Number of discrete values in an interval

• shuffle_ind (bool) – Whether to shuffle indicators (default=True)

Returns

Output of split-conformal transducer (Q) and CRPS

Return type

Tuple[np.ndarray, np.ndarray]