Coverage Error

coverage-error

Metric coverage_error from scikit-learn (sklearn.metrics.coverage_error)

When to invoke this skill

The user has predictions + ground truth and asks to evaluate with coverage_error, or mentions sklearn.metrics.coverage_error directly, or wants the standard scikit-learn implementation.

Reference signature

from sklearn.metrics import coverage_error

# coverage_error(y_true, y_score, *, sample_weight=None)

Library docstring

Coverage error measure.

Compute how far we need to go through the ranked scores to cover all
true labels. The best value is equal to the average number
of labels in ``y_true`` per sample.

Ties in ``y_scores`` are broken by giving maximal rank that would have
been assigned to all tied values.

Note: Our implementation's score is 1 greater than the one given in
Tsoumakas et al., 2010. This extends it to handle the degenerate case
in which an instance has 0 true labels.

Read more in the :ref:`User Guide <coverage_error>`.

Parameters
----------
y_true : array-like of shape (n_samples, n_labels)
    True binary labels in binary indicator format.

y_score : array-like of shape (n_samples, n_labels)
    Target scores, can either be probability estimates of the positive
    class, confidence values, or non-thresholded measure of decisions
    (as returned by "decision_function" on some classifiers).
    For :term:`decision_function` scores, values greater than or equal to
    zero should indicate the positive class.

sample_weight : array-like of shape (n_samples,), default=None
    Sample weights.

Returns
-------
coverage_error : float
    The coverage error.

References
----------
.. [1] Tsoumakas, G., Katakis, I., & Vlahavas, I. (2010).
       Mining multi-label data. In Data mining and knowledge discovery
       handbook (pp. 667-685). Springer US.

Examples
--------
>>> from sklearn.metrics import coverage_error
>>> y_true = [[1, 0, 0], [0, 1, 1]]
>>> y_score = [[1, 0, 0], [0, 1, 1]]
>>> coverage_error(y_true, y_score)
1.5

Quick recipe

import sklearn.metrics as _m
score = _m.coverage_error(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).