torcheval.metrics.WindowedBinaryNormalizedEntropy

class torcheval.metrics.WindowedBinaryNormalizedEntropy(*, from_logits: bool = False, num_tasks: int = 1, max_num_updates: int = 100, enable_lifetime: bool = True, device: Optional[device] = None)

The windowed version of BinaryNormalizedEntropy that provides both windowed and liftime values. Windowed value is calculated from the input and target of the last window_size number of update() calls. Lifetime value is calculated from all past input and target of update() calls.

Compute the normalized binary cross entropy between predicted input and ground-truth binary target.

Parameters:

• from_logits (bool) – A boolean indicator whether the predicted value y_pred is a floating-point logit value (i.e., value in [-inf, inf] when from_logits=True) or a probablity value (i.e., value in [0., 1.] when from_logits=False) Default value is False.
• num_tasks (int) – Number of tasks that need BinaryNormalizedEntropy calculation. Default value is 1. BinaryNormalizedEntropy for each task will be calculated independently.
• max_num_updates (int) – The max window size that can accommodate the number of updates.
• enable_lifetime (bool) – A boolean indicator whether to calculate lifetime values.

Examples:

>>> import torch
>>> from torcheval.metrics import WindowedBinaryNormalizedEntropy

>>> metric = WindowedBinaryNormalizedEntropy(max_num_updates=2)
>>> metric.update(torch.tensor([0.2, 0.3]), torch.tensor([1.0, 0.0]))
>>> metric.update(torch.tensor([0.5, 0.6]), torch.tensor([1.0, 1.0]))
>>> metric.update(torch.tensor([0.6, 0.2]), torch.tensor([0.0, 1.0]))
>>> metric.num_examples, metric.windowed_num_examples
(tensor([6.], dtype=torch.float64), tensor([[2., 2.]], dtype=torch.float64))
>>> metric.compute()
(tensor([1.4914], dtype=torch.float64), tensor([1.6581], dtype=torch.float64))

>>> metric = WindowedBinaryNormalizedEntropy(max_num_updates=2, enable_lifetime=False)
>>> metric.update(torch.tensor([0.2, 0.3]), torch.tensor([1.0, 0.0]))
>>> metric.update(torch.tensor([0.5, 0.6]), torch.tensor([1.0, 1.0]))
>>> metric.update(torch.tensor([0.6, 0.2]), torch.tensor([0.0, 1.0]))
>>> metric.windowed_num_examples
tensor([[2., 2.]], dtype=torch.float64)
>>> metric.compute()
tensor([1.6581], dtype=torch.float64)

>>> metric = WindowedBinaryNormalizedEntropy(max_num_updates=2, num_tasks=2)
>>> metric.update(torch.tensor([[0.2, 0.3], [0.5, 0.1]]), torch.tensor([[1.0, 0.0], [0.0, 1.0]]))
>>> metric.update(torch.tensor([[0.8, 0.3], [0.6, 0.1]]), torch.tensor([[1.0, 1.0], [1.0, 0.0]]))
>>> metric.update(torch.tensor([[0.5, 0.1], [0.3, 0.9]]), torch.tensor([[0.0, 1.0], [0.0, 0.0]]))
>>> metric.num_examples, metric.windowed_num_examples
(tensor([6., 6.], dtype=torch.float64),
tensor([[2., 2.],
        [2., 2.]], dtype=torch.float64))
>>> metric.compute()
(tensor([1.6729, 1.6421], dtype=torch.float64),
tensor([1.9663, 1.4562], dtype=torch.float64))

__init__(*, from_logits: bool = False, num_tasks: int = 1, max_num_updates: int = 100, enable_lifetime: bool = True, device: Optional[device] = None) → None

Initialize a metric object and its internal states.

Use self._add_state() to initialize state variables of your metric class. The state variables should be either torch.Tensor, a list of torch.Tensor, or a dictionary with torch.Tensor as values

Methods

__init__(*[, from_logits, num_tasks, ...])	Initialize a metric object and its internal states.
compute()	Return the normalized binary cross entropy.
load_state_dict(state_dict[, strict])	Loads metric state variables from state_dict.
merge_state(metrics)	Merge the metric state with its counterparts from other metric instances.
reset()	Reset the metric state variables to their default value.
state_dict()	Save metric state variables in state_dict.
to(device, *args, **kwargs)	Move tensors in metric state variables to device.
update(input, target, *[, weight])	Update the metric state with the total entropy, total number of examples and total number of positive targets.

Attributes

device

The last input device of Metric.to().