metrics¶
Metrics with backend dispatch for NumPy, PyTorch, and JAX.
Submodules¶
Functions¶
Compute top-1 classification accuracy. |
|
Compute area under a curve using the trapezoid rule. |
|
Compute the mean width of per-class probability intervals. |
|
Compute average precision (area under the precision-recall curve). |
|
Compute the mean Brier loss of the predictions, the average of |
|
Compute the Brier loss |
|
Compute the classwise expected calibration error (classwise-ECE) [KPerelloNKangsepp+19]. |
|
Empirical convex-hull coverage for distribution-valued targets. |
|
Compute the empirical coverage of a predicted set against true labels. |
|
Compute the average size of a predicted set. |
|
Compute the confidence expected calibration error (ECE) [GPSW17]. |
|
Compute the false negative rate |
|
Compute the false positive rate |
|
Compute the mean log loss of the predictions, the average of |
|
Compute the log loss |
|
Compute precision-recall curve. |
|
Compute area under the ROC curve. |
|
Compute receiver operating characteristic (ROC) curve. |
|
Compute the mean spherical loss of the predictions, the average of |
|
Compute the spherical loss |
|
Compute the mean zero-one loss of the predictions, the average of |
|
Compute the zero-one loss |