import numpy as np
from captum.log import log_usage
from captum._utils.typing import TensorOrTupleOfTensorsGeneric
from sklearn.metrics import average_precision_score
from typing import Tuple
from .base import _base_white_box_metric
def _auprc(
attributions: Tuple[np.ndarray],
true_attributions: Tuple[np.ndarray],
attributions_subset: Tuple[np.ndarray],
) -> Tuple[float]:
return tuple(
average_precision_score(true_attr, attr)
for true_attr, attr in zip(true_attributions, attributions)
)
[docs]@log_usage()
def auprc(
attributions: TensorOrTupleOfTensorsGeneric,
true_attributions: TensorOrTupleOfTensorsGeneric,
normalize: bool = True,
) -> Tuple[float]:
"""
Area under precision-recall.
This is the standard area under the precision-recall curve.
Higher is better.
Args:
attributions (tensor or tuple of tensors):
The attributions with respect to each input feature.
Attributions will always be
the same size as the provided inputs, with each value
providing the attribution of the corresponding input index.
If a single tensor is provided as inputs, a single float
is returned. If a tuple is provided for inputs, a tuple of
float is returned.
true_attributions (tensor or tuple of tensors):
True attributions to be used as a benchmark. Should be of
the same format as the attributions.
normalize (bool): Whether to normalize the attributions before
computing the metric or not. Default: True
Returns:
(float or tuple or floats): The auprc metric.
Examples:
>>> import torch as th
>>> from tint.metrics.white_box import auprc
<BLANKLINE>
>>> attr = th.rand(8, 7, 5)
>>> true_attr = th.randint(2, (8, 7, 5))
<BLANKLINE>
>>> auprc_ = auprc(attr, true_attr)
"""
return _base_white_box_metric(
metric=_auprc,
attributions=attributions,
true_attributions=true_attributions,
normalize=normalize,
hard_labels=True,
)