/
_fairness_metrics.py
291 lines (225 loc) · 8.89 KB
/
_fairness_metrics.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
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
286
287
288
289
290
291
# Copyright (c) Microsoft Corporation and Fairlearn contributors.
# Licensed under the MIT License.
"""Metrics for measuring fairness."""
from ._base_metrics import false_positive_rate, selection_rate, true_positive_rate
from ._metric_frame import MetricFrame
def demographic_parity_difference(
y_true, y_pred, *, sensitive_features, method="between_groups", sample_weight=None
) -> float:
"""Calculate the demographic parity difference.
The demographic parity difference is defined as the difference
between the largest and the smallest group-level selection rate,
:math:`E[h(X) | A=a]`, across all values :math:`a` of the sensitive feature(s).
The demographic parity difference of 0 means that all groups have the same selection rate.
Read more in the :ref:`User Guide <disparity_metrics>`.
Parameters
----------
y_true : array-like
Ground truth (correct) labels.
y_pred : array-like
Predicted labels :math:`h(X)` returned by the classifier.
sensitive_features : array-like
The sensitive features over which demographic parity should be assessed
method : str
How to compute the differences. See :func:`fairlearn.metrics.MetricFrame.difference`
for details.
sample_weight : array-like
The sample weights
Returns
-------
float
The demographic parity difference
"""
sel_rate = MetricFrame(
metrics=selection_rate,
y_true=y_true,
y_pred=y_pred,
sensitive_features=sensitive_features,
sample_params={"sample_weight": sample_weight},
)
result = sel_rate.difference(method=method)
return result
def demographic_parity_ratio(
y_true, y_pred, *, sensitive_features, method="between_groups", sample_weight=None
) -> float:
"""Calculate the demographic parity ratio.
The demographic parity ratio is defined as the ratio
between the smallest and the largest group-level selection rate,
:math:`E[h(X) | A=a]`, across all values :math:`a` of the sensitive feature(s).
The demographic parity ratio of 1 means that all groups have the same selection rate.
Read more in the :ref:`User Guide <disparity_metrics>`.
Parameters
----------
y_true : array-like
Ground truth (correct) labels.
y_pred : array-like
Predicted labels :math:`h(X)` returned by the classifier.
sensitive_features : array-like
The sensitive features over which demographic parity should be assessed
method : str
How to compute the differences. See :func:`fairlearn.metrics.MetricFrame.ratio`
for details.
sample_weight : array-like
The sample weights
Returns
-------
float
The demographic parity ratio
"""
sel_rate = MetricFrame(
metrics=selection_rate,
y_true=y_true,
y_pred=y_pred,
sensitive_features=sensitive_features,
sample_params={"sample_weight": sample_weight},
)
result = sel_rate.ratio(method=method)
return result
def equalized_odds_difference(
y_true, y_pred, *, sensitive_features, method="between_groups", sample_weight=None
) -> float:
"""Calculate the equalized odds difference.
The greater of two metrics: `true_positive_rate_difference` and
`false_positive_rate_difference`. The former is the difference between the
largest and smallest of :math:`P[h(X)=1 | A=a, Y=1]`, across all values :math:`a`
of the sensitive feature(s). The latter is defined similarly, but for
:math:`P[h(X)=1 | A=a, Y=0]`.
The equalized odds difference of 0 means that all groups have the same
true positive, true negative, false positive, and false negative rates.
Read more in the :ref:`User Guide <disparity_metrics>`.
Parameters
----------
y_true : array-like
Ground truth (correct) labels.
y_pred : array-like
Predicted labels :math:`h(X)` returned by the classifier.
sensitive_features : array-like
The sensitive features over which demographic parity should be assessed
method : str
How to compute the differences. See :func:`fairlearn.metrics.MetricFrame.difference`
for details.
sample_weight : array-like
The sample weights
Returns
-------
float
The equalized odds difference
"""
eo = _get_eo_frame(y_true, y_pred, sensitive_features, sample_weight)
return max(eo.difference(method=method))
def equalized_odds_ratio(
y_true, y_pred, *, sensitive_features, method="between_groups", sample_weight=None
) -> float:
"""Calculate the equalized odds ratio.
The smaller of two metrics: `true_positive_rate_ratio` and
`false_positive_rate_ratio`. The former is the ratio between the
smallest and largest of :math:`P[h(X)=1 | A=a, Y=1]`, across all values :math:`a`
of the sensitive feature(s). The latter is defined similarly, but for
:math:`P[h(X)=1 | A=a, Y=0]`.
The equalized odds ratio of 1 means that all groups have the same
true positive, true negative, false positive, and false negative rates.
Read more in the :ref:`User Guide <disparity_metrics>`.
Parameters
----------
y_true : array-like
Ground truth (correct) labels.
y_pred : array-like
Predicted labels :math:`h(X)` returned by the classifier.
sensitive_features : array-like
The sensitive features over which demographic parity should be assessed
method : str
How to compute the differences. See :func:`fairlearn.metrics.MetricFrame.ratio`
for details.
sample_weight : array-like
The sample weights
Returns
-------
float
The equalized odds ratio
"""
eo = _get_eo_frame(y_true, y_pred, sensitive_features, sample_weight)
return min(eo.ratio(method=method))
def _get_eo_frame(y_true, y_pred, sensitive_features, sample_weight) -> MetricFrame:
fns = {"tpr": true_positive_rate, "fpr": false_positive_rate}
sw_dict = {"sample_weight": sample_weight}
sp = {"tpr": sw_dict, "fpr": sw_dict}
eo = MetricFrame(
metrics=fns,
y_true=y_true,
y_pred=y_pred,
sensitive_features=sensitive_features,
sample_params=sp,
)
return eo
def equal_opportunity_difference(
y_true, y_pred, *, sensitive_features, method="between_groups", sample_weight=None
) -> float:
"""Calculate the equal opportunity difference.
The equal opportunity difference is defined as the difference
between the largest and the smallest group-level true positive rates,
:math:`E[h(X) | A=a]`, across all values :math:`a` of the sensitive feature(s).
The equal opportunity difference of 0 means that all groups have the same true positive rate.
Read more in the :ref:`User Guide <disparity_metrics>`.
Parameters
----------
y_true : array-like
Ground truth (correct) labels.
y_pred : array-like
Predicted labels :math:`h(X)` returned by the classifier.
sensitive_features : array-like
The sensitive features over which equal opportunity should be assessed
method : str
How to compute the differences. See :func:`fairlearn.metrics.MetricFrame.difference`
for details.
sample_weight : array-like
The sample weights
Returns
-------
float
The equal opportunity difference
"""
tpr = MetricFrame(
metrics=true_positive_rate,
y_true=y_true,
y_pred=y_pred,
sensitive_features=sensitive_features,
sample_params={"sample_weight": sample_weight},
)
result = tpr.difference(method=method)
return result
def equal_opportunity_ratio(
y_true, y_pred, *, sensitive_features, method="between_groups", sample_weight=None
) -> float:
"""Calculate the equal opportunity ratio.
The equal opportunity ratio is defined as the ratio
between the smallest and the largest group-level true positive rate,
:math:`E[h(X) | A=a]`, across all values :math:`a` of the sensitive feature(s).
The equal opportunity ratio of 1 means that all groups have the same true positive rate.
Read more in the :ref:`User Guide <disparity_metrics>`.
Parameters
----------
y_true : array-like
Ground truth (correct) labels.
y_pred : array-like
Predicted labels :math:`h(X)` returned by the classifier.
sensitive_features : array-like
The sensitive features over which equal opportunity should be assessed
method : str
How to compute the differences. See :func:`fairlearn.metrics.MetricFrame.ratio`
for details.
sample_weight : array-like
The sample weights
Returns
-------
float
The equal opportunity ratio
"""
tpr = MetricFrame(
metrics=true_positive_rate,
y_true=y_true,
y_pred=y_pred,
sensitive_features=sensitive_features,
sample_params={"sample_weight": sample_weight},
)
result = tpr.ratio(method=method)
return result