numpy

NumPy implementation of Metrics.

Functions

numpy_categorical_classwise_ece

Compute the classwise ECE for a categorical distribution via its class probabilities.

numpy_categorical_expected_calibration_error

Compute the confidence ECE for a categorical distribution via its class probabilities.

numpy_classwise_ece

Compute the classwise expected calibration error for NumPy arrays.

numpy_expected_calibration_error

Compute the confidence expected calibration error for NumPy arrays.

numpy_log_probabilities

Compute the log probabilities of a NumPy array.

numpy_one_hot_argmax

One-hot encode the first maximum of each probability vector of a NumPy array.

numpy_precision_recall_curve

Compute precision-recall curve along the last axis.

numpy_probabilities

Return a NumPy array of probabilities unchanged.

numpy_roc_curve

Compute ROC curve along the last axis.