def recall(
outputs: torch.Tensor,
targets: torch.Tensor,
argmax_dim: int = -1,
eps: float = 1e-7,
num_classes: Optional[int] = None,
) -> Union[float, torch.Tensor]:
"""
Multiclass recall score.
Args:
outputs: estimated targets as predicted by a model
with shape [bs; ..., (num_classes or 1)]
targets: ground truth (correct) target values
with shape [bs; ..., 1]
argmax_dim: int, that specifies dimension for argmax transformation
in case of scores/probabilities in ``outputs``
eps: float. Epsilon to avoid zero division.
num_classes: int, that specifies number of classes if it known
Returns:
Tensor: recall for every class
Examples:
.. code-block:: python
import torch
from catalyst import metrics
metrics.recall(
outputs=torch.tensor([
[1, 0, 0],
[0, 1, 0],
[0, 0, 1],
]),
targets=torch.tensor([0, 1, 2]),
)
# tensor([1., 1., 1.])
.. code-block:: python
import torch
from catalyst import metrics
metrics.recall(
outputs=torch.tensor([[0, 0, 1, 1, 0, 1, 0, 1]]),
targets=torch.tensor([[0, 1, 0, 1, 0, 0, 1, 1]]),
)
# tensor([0.5000, 0.5000]
"""
_, recall_score, _, _ = precision_recall_fbeta_support(
outputs=outputs,
targets=targets,
argmax_dim=argmax_dim,
eps=eps,
num_classes=num_classes,
)
return recall_score
def fbeta_score(
outputs: torch.Tensor,
targets: torch.Tensor,
beta: float = 1.0,
eps: float = 1e-7,
argmax_dim: int = -1,
num_classes: Optional[int] = None,
) -> Union[float, torch.Tensor]:
"""Counts fbeta score for given ``outputs`` and ``targets``.
Args:
outputs: A list of predicted elements
targets: A list of elements that are to be predicted
beta: beta param for f_score
eps: epsilon to avoid zero division
argmax_dim: int, that specifies dimension for argmax transformation
in case of scores/probabilities in ``outputs``
num_classes: int, that specifies number of classes if it known
Raises:
ValueError: If ``beta`` is a negative number.
Returns:
float: F_beta score.
Example:
.. code-block:: python
import torch
from catalyst import metrics
metrics.fbeta_score(
outputs=torch.tensor([
[1, 0, 0],
[0, 1, 0],
[0, 0, 1],
]),
targets=torch.tensor([0, 1, 2]),
beta=1,
)
# tensor([1., 1., 1.]), # per class fbeta
"""
if beta < 0:
raise ValueError("beta parameter should be non-negative")
_p, _r, fbeta, _ = precision_recall_fbeta_support(
outputs=outputs,
targets=targets,
beta=beta,
eps=eps,
argmax_dim=argmax_dim,
num_classes=num_classes,
)
return fbeta
def precision(
outputs: torch.Tensor,
targets: torch.Tensor,
argmax_dim: int = -1,
eps: float = 1e-7,
num_classes: Optional[int] = None,
) -> Union[float, torch.Tensor]:
"""
Multiclass precision score.
Args:
outputs: estimated targets as predicted by a model
with shape [bs; ..., (num_classes or 1)]
targets: ground truth (correct) target values
with shape [bs; ..., 1]
argmax_dim: int, that specifies dimension for argmax transformation
in case of scores/probabilities in ``outputs``
eps: float. Epsilon to avoid zero division.
num_classes: int, that specifies number of classes if it known
Returns:
Tensor: precision for every class
Examples:
>>> precision(
>>> outputs=torch.tensor([
>>> [1, 0, 0],
>>> [0, 1, 0],
>>> [0, 0, 1],
>>> ]),
>>> targets=torch.tensor([0, 1, 2]),
>>> beta=1,
>>> )
tensor([1., 1., 1.])
>>> precision(
>>> outputs=torch.tensor([[0, 0, 1, 1, 0, 1, 0, 1]]),
>>> targets=torch.tensor([[0, 1, 0, 1, 0, 0, 1, 1]]),
>>> )
tensor([0.5000, 0.5000]
"""
precision_score, _, _, _, = precision_recall_fbeta_support(
outputs=outputs,
targets=targets,
argmax_dim=argmax_dim,
eps=eps,
num_classes=num_classes,
)
return precision_score