torcheval.metrics.functional.weighted_calibration

torcheval.metrics.functional.weighted_calibration(input: Tensor, target: Tensor, weight: float | int | Tensor = 1.0, *, num_tasks: int = 1) → Tensor

Compute weighted calibration metric. When weight is not provided, it calculates the unweighted calibration. Its class version is torcheval.metrics.WeightedCalibration.

weighted_calibration = sum(input * weight) / sum(target * weight)

Parameters:

• input (Tensor) – Tensor of input values.

• target (Tensor) – Tensor of binary targets

• weight (optional) – Float or Int or Tensor of input weights. It is default to 1.0. If weight is a Tensor, its size should match the input tensor size.

• num_tasks (int) – Number of tasks that need weighted_calibration calculation. Default value is 1.

Returns:

The return value of weighted calibration for each task (num_tasks,).

Return type:

Tensor

Raises:

ValueError – If value of weight is neither a float nor a int nor a torch.Tensor that matches the input tensor size.

Examples:

>>> import torch
>>> from torcheval.metrics.functional import weighted_calibration
>>> weighted_calibration(torch.tensor([0.8, 0.4, 0.3, 0.8, 0.7, 0.6]),torch.tensor([1, 1, 0, 0, 1, 0]))
tensor([1.2])

>>> weighted_calibration(torch.tensor([0.8, 0.4, 0.3, 0.8, 0.7, 0.6]),torch.tensor([1, 1, 0, 0, 1, 0]), torch.tensor([0.5, 1., 2., 0.4, 1.3, 0.9]))
tensor([1.1321])

>>> weighted_calibration(
        torch.tensor([[0.8, 0.4], [0.8, 0.7]]),
        torch.tensor([[1, 1], [0, 1]]),
        num_tasks=2,
    ),
>>> tensor([0.6000, 1.5000])