File size: 13,182 Bytes
d76ed8a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 | # Copyright The Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import List, Optional, Union
import torch
from torch import Tensor
from typing_extensions import Literal
from torchmetrics.functional.classification.roc import (
binary_roc,
multiclass_roc,
multilabel_roc,
)
from torchmetrics.utilities.enums import ClassificationTask
def _binary_eer_compute(fpr: Tensor, tpr: Tensor) -> Tensor:
"""Compute Equal Error Rate (EER) for binary classification task."""
diff = fpr - (1 - tpr)
idx = torch.argmin(torch.abs(diff))
return (fpr[idx] + (1 - tpr[idx])) / 2
def _eer_compute(
fpr: Union[Tensor, List[Tensor]],
tpr: Union[Tensor, List[Tensor]],
) -> Tensor:
"""Compute Equal Error Rate (EER)."""
if isinstance(fpr, Tensor) and isinstance(tpr, Tensor) and fpr.ndim == 1:
return _binary_eer_compute(fpr, tpr)
return torch.stack([_binary_eer_compute(f, t) for f, t in zip(fpr, tpr)])
def binary_eer(
preds: Tensor,
target: Tensor,
thresholds: Optional[Union[int, List[float], Tensor]] = None,
ignore_index: Optional[int] = None,
validate_args: bool = True,
) -> Tensor:
r"""Compute Equal Error Rate (EER) for binary classification task.
.. math::
\text{EER} = \frac{\text{FAR} + (1 - \text{FRR})}{2}, \text{where} \min_t abs(FAR_t-FRR_t)
The Equal Error Rate (EER) is the point where the False Positive Rate (FPR) and True Positive Rate (TPR) are
equal, or in practise minimized. A lower EER value signifies higher system accuracy.
Args:
preds: Tensor with predictions
target: Tensor with true labels
thresholds:
Can be one of:
- If set to `None`, will use a non-binned approach where thresholds are dynamically calculated from
all the data. Most accurate but also most memory consuming approach.
- If set to an `int` (larger than 1), will use that number of thresholds linearly spaced from
0 to 1 as bins for the calculation.
- If set to an `list` of floats, will use the indicated thresholds in the list as bins for the calculation
- If set to an 1d `tensor` of floats, will use the indicated thresholds in the tensor as
bins for the calculation.
ignore_index:
Specifies a target value that is ignored and does not contribute to the metric calculation
validate_args: bool indicating if input arguments and tensors should be validated for correctness.
Set to ``False`` for faster computations
Returns:
A single scalar with the eer score
Example:
>>> from torchmetrics.functional.classification import binary_eer
>>> preds = torch.tensor([0, 0.5, 0.7, 0.8])
>>> target = torch.tensor([0, 1, 1, 0])
>>> binary_eer(preds, target, thresholds=None)
tensor(0.5000)
>>> binary_eer(preds, target, thresholds=5)
tensor(0.7500)
"""
fpr, tpr, _ = binary_roc(preds, target, thresholds, ignore_index, validate_args)
return _eer_compute(fpr, tpr)
def multiclass_eer(
preds: Tensor,
target: Tensor,
num_classes: int,
thresholds: Optional[Union[int, List[float], Tensor]] = None,
average: Optional[Literal["micro", "macro"]] = None,
ignore_index: Optional[int] = None,
validate_args: bool = True,
) -> Tensor:
r"""Compute Equal Error Rate (EER) for multiclass classification task.
.. math::
\text{EER} = \frac{\text{FAR} + (1 - \text{FRR})}{2}, \text{where} \min_t abs(FAR_t-FRR_t)
The Equal Error Rate (EER) is the point where the False Positive Rate (FPR) and True Positive Rate (TPR) are
equal, or in practise minimized. A lower EER value signifies higher system accuracy.
Args:
preds: Tensor with predictions
target: Tensor with true labels
num_classes: Integer specifying the number of classes
thresholds:
Can be one of:
- If set to `None`, will use a non-binned approach where thresholds are dynamically calculated from
all the data. Most accurate but also most memory consuming approach.
- If set to an `int` (larger than 1), will use that number of thresholds linearly spaced from
0 to 1 as bins for the calculation.
- If set to an `list` of floats, will use the indicated thresholds in the list as bins for the calculation
- If set to an 1d `tensor` of floats, will use the indicated thresholds in the tensor as
bins for the calculation.
average:
If aggregation of should be applied. The aggregation is applied to underlying ROC curves.
By default, eer is not aggregated and a score for each class is returned. If `average` is set to ``"micro"``
, the metric will aggregate the curves by one hot encoding the targets and flattening the predictions,
considering all classes jointly as a binary problem. If `average` is set to ``"macro"``, the metric will
aggregate the curves by first interpolating the curves from each class at a combined set of thresholds and
then average over the classwise interpolated curves. See `averaging curve objects`_ for more info on the
different averaging methods.
ignore_index:
Specifies a target value that is ignored and does not contribute to the metric calculation
validate_args: bool indicating if input arguments and tensors should be validated for correctness.
Set to ``False`` for faster computations.
Returns:
If `average=None|"none"` then a 1d tensor of shape (n_classes, ) will be returned with eer score per class.
If `average="macro"|"micro"` then a single scalar is returned.
Example:
>>> from torchmetrics.functional.classification import multiclass_eer
>>> preds = torch.tensor([[0.75, 0.05, 0.05, 0.05, 0.05],
... [0.05, 0.75, 0.05, 0.05, 0.05],
... [0.05, 0.05, 0.75, 0.05, 0.05],
... [0.05, 0.05, 0.05, 0.75, 0.05]])
>>> target = torch.tensor([0, 1, 3, 2])
>>> multiclass_eer(preds, target, num_classes=5, average="macro", thresholds=None)
tensor(0.4667)
>>> multiclass_eer(preds, target, num_classes=5, average=None, thresholds=None)
tensor([0.0000, 0.0000, 0.6667, 0.6667, 1.0000])
>>> multiclass_eer(preds, target, num_classes=5, average="macro", thresholds=5)
tensor(0.4667)
>>> multiclass_eer(preds, target, num_classes=5, average=None, thresholds=5)
tensor([0.0000, 0.0000, 0.6667, 0.6667, 1.0000])
"""
fpr, tpr, _ = multiclass_roc(preds, target, num_classes, thresholds, average, ignore_index, validate_args)
return _eer_compute(fpr, tpr)
def multilabel_eer(
preds: Tensor,
target: Tensor,
num_labels: int,
thresholds: Optional[Union[int, List[float], Tensor]] = None,
ignore_index: Optional[int] = None,
validate_args: bool = True,
) -> Tensor:
r"""Compute Equal Error Rate (EER) for multilabel classification task.
.. math::
\text{EER} = \frac{\text{FAR} + (1 - \text{FRR})}{2}, \text{where} \min_t abs(FAR_t-FRR_t)
The Equal Error Rate (EER) is the point where the False Positive Rate (FPR) and True Positive Rate (TPR) are
equal, or in practise minimized. A lower EER value signifies higher system accuracy.
Args:
preds: Tensor with predictions
target: Tensor with true labels
num_labels: Integer specifying the number of labels
thresholds:
Can be one of:
- If set to `None`, will use a non-binned approach where thresholds are dynamically calculated from
all the data. Most accurate but also most memory consuming approach.
- If set to an `int` (larger than 1), will use that number of thresholds linearly spaced from
0 to 1 as bins for the calculation.
- If set to an `list` of floats, will use the indicated thresholds in the list as bins for the calculation
- If set to an 1d `tensor` of floats, will use the indicated thresholds in the tensor as
bins for the calculation.
ignore_index:
Specifies a target value that is ignored and does not contribute to the metric calculation
validate_args: bool indicating if input arguments and tensors should be validated for correctness.
Set to ``False`` for faster computations.
Returns:
A 1d tensor of shape (n_classes, ) will be returned with eer score per label.
Example:
>>> from torchmetrics.functional.classification import multilabel_eer
>>> preds = torch.tensor([[0.75, 0.05, 0.35],
... [0.45, 0.75, 0.05],
... [0.05, 0.55, 0.75],
... [0.05, 0.65, 0.05]])
>>> target = torch.tensor([[1, 0, 1],
... [0, 0, 0],
... [0, 1, 1],
... [1, 1, 1]])
>>> multilabel_eer(preds, target, num_labels=3, thresholds=None)
tensor([0.5000, 0.5000, 0.1667])
>>> multilabel_eer(preds, target, num_labels=3, thresholds=5)
tensor([0.5000, 0.7500, 0.1667])
"""
fpr, tpr, _ = multilabel_roc(preds, target, num_labels, thresholds, ignore_index, validate_args)
return _eer_compute(fpr, tpr)
def eer(
preds: Tensor,
target: Tensor,
task: Literal["binary", "multiclass", "multilabel"],
thresholds: Optional[Union[int, List[float], Tensor]] = None,
num_classes: Optional[int] = None,
num_labels: Optional[int] = None,
average: Optional[Literal["micro", "macro"]] = None,
ignore_index: Optional[int] = None,
validate_args: bool = True,
) -> Union[Tensor, List[Tensor]]:
"""Compute Equal Error Rate (EER) metric.
This function is a simple wrapper to get the task specific versions of this metric, which is done by setting the
``task`` argument to either ``'binary'``, ``'multiclass'`` or ``'multilabel'``. See the documentation of
:func:`~torchmetrics.functional.classification.binary_eer`,
:func:`~torchmetrics.functional.classification.multiclass_eer` and
:func:`~torchmetrics.functional.classification.multilabel_eer` for the specific details of
each argument influence and examples.
Args:
preds: Predictions from model (logits or probabilities)
target: Ground truth labels
task: Type of task, either 'binary', 'multiclass' or 'multilabel'
thresholds: Thresholds used for computing the ROC curve
num_classes: Number of classes (for multiclass task)
num_labels: Number of labels (for multilabel task)
average: Method to average EER over multiple classes/labels
ignore_index: Specify a target value that is ignored
validate_args: Bool indicating whether to validate input arguments
Legacy Example:
>>> from torchmetrics.functional.classification import eer
>>> preds = torch.tensor([0.13, 0.26, 0.08, 0.19, 0.34])
>>> target = torch.tensor([0, 0, 1, 1, 1])
>>> eer(preds, target, task='binary')
tensor(0.5833)
>>> preds = torch.tensor([[0.90, 0.05, 0.05],
... [0.05, 0.90, 0.05],
... [0.05, 0.05, 0.90],
... [0.85, 0.05, 0.10],
... [0.10, 0.10, 0.80]])
>>> target = torch.tensor([0, 1, 1, 2, 2])
>>> eer(preds, target, task='multiclass', num_classes=3, )
tensor([0.0000, 0.4167, 0.4167])
"""
task = ClassificationTask.from_str(task)
if task == ClassificationTask.BINARY:
return binary_eer(preds, target, thresholds, ignore_index, validate_args)
if task == ClassificationTask.MULTICLASS:
if not isinstance(num_classes, int):
raise ValueError(f"`num_classes` is expected to be `int` but `{type(num_classes)} was passed.`")
return multiclass_eer(preds, target, num_classes, thresholds, average, ignore_index, validate_args)
if task == ClassificationTask.MULTILABEL:
if not isinstance(num_labels, int):
raise ValueError(f"`num_labels` is expected to be `int` but `{type(num_labels)} was passed.`")
return multilabel_eer(preds, target, num_labels, thresholds, ignore_index, validate_args)
raise ValueError(f"Task {task} not supported.")
|