Spearmancorrcoef

spearmancorrcoef

Metric SpearmanCorrCoef from torchmetrics (torchmetrics.SpearmanCorrCoef)

When to invoke this skill

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

Reference signature

from torchmetrics import SpearmanCorrCoef

# SpearmanCorrCoef(num_outputs: int = 1, **kwargs: Any) -> None

Library docstring

Compute `spearmans rank correlation coefficient`_.

.. math:
    r_s = = \frac{cov(rg_x, rg_y)}{\sigma_{rg_x} * \sigma_{rg_y}}

where :math:`rg_x` and :math:`rg_y` are the rank associated to the variables :math:`x` and :math:`y`.
Spearmans correlations coefficient corresponds to the standard pearsons correlation coefficient calculated
on the rank variables.

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

- ``preds`` (:class:`~torch.Tensor`): Predictions from model in float tensor with shape ``(N,d)``
- ``target`` (:class:`~torch.Tensor`): Ground truth values in float tensor with shape ``(N,d)``

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

- ``spearman`` (:class:`~torch.Tensor`): A tensor with the spearman correlation(s)

Args:
    num_outputs: Number of outputs in multioutput setting
    kwargs: Additional keyword arguments, see :ref:`Metric kwargs` for more info.

Example (single output regression):
    >>> from torch import tensor
    >>> from torchmetrics.regression import SpearmanCorrCoef
    >>> target = tensor([3, -0.5, 2, 7])
    >>> preds = tensor([2.5, 0.0, 2, 8])
    >>> spearman = SpearmanCorrCoef()
    >>> spearman(preds, target)
    tensor(1.0000)

Example (multi output regression):
    >>> from torchmetrics.regression import SpearmanCorrCoef
    >>> target = tensor([[3, -0.5], [2, 7]])
    >>> preds = tensor([[2.5, 0.0], [2, 8]])
    >>> spearman = SpearmanCorrCoef(num_outputs=2)
    >>> spearman(preds, target)
    tensor([1.0000, 1.0000])

Quick recipe

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