-
Notifications
You must be signed in to change notification settings - Fork 818
/
sample_distortion_metric.py
153 lines (125 loc) · 6.61 KB
/
sample_distortion_metric.py
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
from functools import partial
import numpy as np
import scipy.spatial.distance as scdist
from aif360.datasets import StructuredDataset
from aif360.metrics import DatasetMetric, utils
class SampleDistortionMetric(DatasetMetric):
"""Class for computing metrics based on two StructuredDatasets."""
def __init__(self, dataset, distorted_dataset, unprivileged_groups=None,
privileged_groups=None):
"""
Args:
dataset (StructuredDataset): A StructuredDataset.
distorted_dataset (StructuredDataset): A StructuredDataset.
privileged_groups (list(dict)): Privileged groups. Format is a list
of `dicts` where the keys are `protected_attribute_names` and
the values are values in `protected_attributes`. Each `dict`
element describes a single group. See examples for more details.
unprivileged_groups (list(dict)): Unprivileged groups in the same
format as `privileged_groups`.
Raises:
TypeError: `dataset` and `distorted_dataset` must be
:obj:`~aif360.datasets.StructuredDataset` types.
"""
# sets self.dataset, self.unprivileged_groups, self.privileged_groups
super(SampleDistortionMetric, self).__init__(dataset,
unprivileged_groups=unprivileged_groups,
privileged_groups=privileged_groups)
if isinstance(distorted_dataset, StructuredDataset):
self.distorted_dataset = distorted_dataset
else:
raise TypeError("'distorted_dataset' should be a StructuredDataset")
with dataset.temporarily_ignore('features', 'labels', 'scores'):
if dataset != distorted_dataset:
raise ValueError("The two datasets may differ in features and "
"labels/scores only.")
def total(self, dist, privileged):
distance, weights = dist(privileged=privileged, returned=True)
return np.sum(distance * weights)
def average(self, dist, privileged):
distance, weights = dist(privileged=privileged, returned=True)
return np.average(distance, weights=weights)
def maximum(self, dist, privileged):
return np.max(dist(privileged=privileged))
def euclidean_distance(self, privileged=None, returned=False):
"""Compute the average Euclidean distance between the samples from the
two datasets.
"""
condition = self._to_condition(privileged)
distance, mask = utils.compute_distance(self.dataset.features,
self.distorted_dataset.features, self.dataset.protected_attributes,
self.dataset.protected_attribute_names, dist_fun=scdist.euclidean,
condition=condition)
if returned:
return distance, self.dataset.instance_weights[mask]
return distance
def manhattan_distance(self, privileged=None, returned=False):
"""Compute the average Manhattan distance between the samples from the
two datasets.
"""
condition = self._to_condition(privileged)
distance, mask = utils.compute_distance(self.dataset.features,
self.distorted_dataset.features, self.dataset.protected_attributes,
self.dataset.protected_attribute_names, dist_fun=scdist.cityblock,
condition=condition)
if returned:
return distance, self.dataset.instance_weights[mask]
return distance
def mahalanobis_distance(self, privileged=None, returned=False):
"""Compute the average Mahalanobis distance between the samples from the
two datasets.
"""
condition = self._to_condition(privileged)
X_orig = self.dataset.features
X_distort = self.distorted_dataset.features
dist_fun = partial(scdist.mahalanobis,
VI=np.linalg.inv(np.cov(np.vstack([X_orig, X_distort]).T)).T)
distance, mask = utils.compute_distance(X_orig, X_distort,
self.dataset.protected_attributes,
self.dataset.protected_attribute_names, dist_fun=dist_fun,
condition=condition)
if returned:
return distance, self.dataset.instance_weights[mask]
return distance
def total_euclidean_distance(self, privileged=None):
return self.total(self.euclidean_distance, privileged=privileged)
def total_manhattan_distance(self, privileged=None):
return self.total(self.manhattan_distance, privileged=privileged)
def total_mahalanobis_distance(self, privileged=None):
return self.total(self.mahalanobis_distance, privileged=privileged)
def average_euclidean_distance(self, privileged=None):
return self.average(self.euclidean_distance, privileged=privileged)
def average_manhattan_distance(self, privileged=None):
return self.average(self.manhattan_distance, privileged=privileged)
def average_mahalanobis_distance(self, privileged=None):
return self.average(self.mahalanobis_distance, privileged=privileged)
def maximum_euclidean_distance(self, privileged=None):
return self.maximum(self.euclidean_distance, privileged=privileged)
def maximum_manhattan_distance(self, privileged=None):
return self.maximum(self.manhattan_distance, privileged=privileged)
def maximum_mahalanobis_distance(self, privileged=None):
return self.maximum(self.mahalanobis_distance, privileged=privileged)
def mean_euclidean_distance_difference(self, privileged=None):
"""Difference of the averages."""
return self.difference(
self.average(self.euclidean_distance, privileged=privileged))
def mean_manhattan_distance_difference(self, privileged=None):
"""Difference of the averages."""
return self.difference(
self.average(self.manhattan_distance, privileged=privileged))
def mean_mahalanobis_distance_difference(self, privileged=None):
"""Difference of the averages."""
return self.difference(
self.average(self.mahalanobis_distance, privileged=privileged))
def mean_euclidean_distance_ratio(self, privileged=None):
"""Ratio of the averages."""
return self.ratio(
self.average(self.euclidean_distance, privileged=privileged))
def mean_manhattan_distance_ratio(self, privileged=None):
"""Ratio of the averages."""
return self.ratio(
self.average(self.manhattan_distance, privileged=privileged))
def mean_mahalanobis_distance_ratio(self, privileged=None):
"""Ratio of the averages."""
return self.ratio(
self.average(self.mahalanobis_distance, privileged=privileged))