Randscore

randscore

Metric RandScore from torchmetrics (torchmetrics.clustering.RandScore)

When to invoke this skill

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

Reference signature

from torchmetrics.clustering import RandScore

# RandScore(**kwargs: Any) -> None

Library docstring

Compute `Rand Score`_ (alternatively known as Rand Index).

.. math::
    RS(U, V) = \text{number of agreeing pairs} / \text{number of pairs}

The number of agreeing pairs is every :math:`(i, j)` pair of samples where :math:`i \in U` and :math:`j \in V`
(the predicted and true clusterings, respectively) that are in the same cluster for both clusterings. The metric is
symmetric, therefore swapping :math:`U` and :math:`V` yields the same rand score.

This clustering metric is an extrinsic measure, because it requires ground truth clustering labels, which may not
be available in practice since clustering in generally is used for unsupervised learning.

As input to ``forward`` and ``update`` the metric accepts the following input:

- ``preds`` (:class:`~torch.Tensor`): single integer tensor with shape ``(N,)`` with predicted cluster labels
- ``target`` (:class:`~torch.Tensor`): single integer tensor with shape ``(N,)`` with ground truth cluster labels

As output of ``forward`` and ``compute`` the metric returns the following output:

- ``rand_score`` (:class:`~torch.Tensor`): A tensor with the Rand Score

Args:
    kwargs: Additional keyword arguments, see :ref:`Metric kwargs` for more info.

Example::
    >>> import torch
    >>> from torchmetrics.clustering import RandScore
    >>> preds = torch.tensor([2, 1, 0, 1, 0])
    >>> target = torch.tensor([0, 2, 1, 1, 0])
    >>> metric = RandScore()
    >>> metric(preds, target)
    tensor(0.6000)

Quick recipe

import torchmetrics.clustering as _m
score = _m.RandScore(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).