Hammingdistance

hammingdistance

Metric HammingDistance from torchmetrics (torchmetrics.HammingDistance)

When to invoke this skill

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

Reference signature

from torchmetrics import HammingDistance

# HammingDistance(task: Literal['binary', 'multiclass', 'multilabel'], threshold: float = 0.5, num_classes: Optional[int] = None, num_labels: Optional[int] = None, average: Optional[Literal['micro', 'macro', 'weighted', 'none']] = 'micro', multidim_average: Optional[Literal['global', 'samplewise']] = 'global', top_k: Optional[int] = 1, ignore_index: Optional[int] = None, validate_args: bool = True, **kwargs: Any) -> torchmetrics.metric.Metric

Library docstring

Compute the average `Hamming distance`_ (also known as Hamming loss).

.. math::
    \text{Hamming distance} = \frac{1}{N \cdot L} \sum_i^N \sum_l^L 1(y_{il} \neq \hat{y}_{il})

Where :math:`y` is a tensor of target values, :math:`\hat{y}` is a tensor of predictions,
and :math:`\bullet_{il}` refers to the :math:`l`-th label of the :math:`i`-th sample of that
tensor.

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.BinaryHammingDistance`,
:class:`~torchmetrics.classification.MulticlassHammingDistance` and
:class:`~torchmetrics.classification.MultilabelHammingDistance` for the specific details of each argument influence
and examples.

Legacy Example:
    >>> from torch import tensor
    >>> target = tensor([[0, 1], [1, 1]])
    >>> preds = tensor([[0, 1], [0, 1]])
    >>> hamming_distance = HammingDistance(task="multilabel", num_labels=2)
    >>> hamming_distance(preds, target)
    tensor(0.2500)

Quick recipe

import torchmetrics as _m
score = _m.HammingDistance(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).