Precisionrecallcurve

precisionrecallcurve

Metric PrecisionRecallCurve from torchmetrics (torchmetrics.PrecisionRecallCurve)

When to invoke this skill

The user has predictions + ground truth and asks to evaluate with PrecisionRecallCurve, or mentions torchmetrics.PrecisionRecallCurve directly, or wants the standard torchmetrics implementation.

Reference signature

from torchmetrics import PrecisionRecallCurve

# PrecisionRecallCurve(task: Literal['binary', 'multiclass', 'multilabel'], thresholds: Union[int, list[float], torch.Tensor, NoneType] = None, num_classes: Optional[int] = None, num_labels: Optional[int] = None, ignore_index: Optional[int] = None, validate_args: bool = True, **kwargs: Any) -> torchmetrics.metric.Metric

Library docstring

Compute the precision-recall curve.

The curve consist of multiple pairs of precision and recall values evaluated at different thresholds, such that the
tradeoff between the two values can been seen.

This function is a simple wrapper to get the task specific versions of this metric, which is done by setting the
``task`` argument to either ``'binary'``, ``'multiclass'`` or ``'multilabel'``. See the documentation of
:class:`~torchmetrics.classification.BinaryPrecisionRecallCurve`,
:class:`~torchmetrics.classification.MulticlassPrecisionRecallCurve` and
:class:`~torchmetrics.classification.MultilabelPrecisionRecallCurve` for the specific details of each argument
influence and examples.

Legacy Example:
    >>> pred = torch.tensor([0, 0.1, 0.8, 0.4])
    >>> target = torch.tensor([0, 1, 1, 0])
    >>> pr_curve = PrecisionRecallCurve(task="binary")
    >>> precision, recall, thresholds = pr_curve(pred, target)
    >>> precision
    tensor([0.5000, 0.6667, 0.5000, 1.0000, 1.0000])
    >>> recall
    tensor([1.0000, 1.0000, 0.5000, 0.5000, 0.0000])
    >>> thresholds
    tensor([0.0000, 0.1000, 0.4000, 0.8000])

    >>> pred = torch.tensor([[0.75, 0.05, 0.05, 0.05, 0.05],
    ...                      [0.05, 0.75, 0.05, 0.05, 0.05],
    ...                      [0.05, 0.05, 0.75, 0.05, 0.05],
    ...                      [0.05, 0.05, 0.05, 0.75, 0.05]])
    >>> target = torch.tensor([0, 1, 3, 2])
    >>> pr_curve = PrecisionRecallCurve(task="multiclass", num_classes=5)
    >>> precision, recall, thresholds = pr_curve(pred, target)
    >>> precision
    [tensor([0.2500, 1.0000, 1.0000]), tensor([0.2500, 1.0000, 1.0000]), tensor([0.2500, 0.0000, 1.0000]),
     tensor([0.2500, 0.0000, 1.0000]), tensor([0., 1.])]
    >>> recall
    [tensor([1., 1., 0.]), tensor([1., 1., 0.]), tensor([1., 0., 0.]), tensor([1., 0., 0.]), tensor([nan, 0.])]
    >>> thresholds
    [tensor([0.0500, 0.7500]), tensor([0.0500, 0.7500]), tensor([0.0500, 0.7500]), tensor([0.0500, 0.7500]),
     tensor(0.0500)]

Quick recipe

import torchmetrics as _m
score = _m.PrecisionRecallCurve(y_true, y_pred)

Don'ts

• Don't reimplement when the library version handles edge cases (NaN, ties, empty inputs) better than a hand-rolled formula.
• Always check the library version's argument order — sklearn is (y_true, y_pred) while torchmetrics is (preds, target).